diff options
Diffstat (limited to 'src/gui/rhi')
44 files changed, 32140 insertions, 12353 deletions
diff --git a/src/gui/rhi/MiniEngine_LICENSE.txt b/src/gui/rhi/MiniEngine_LICENSE.txt new file mode 100644 index 0000000000..b8b569d774 --- /dev/null +++ b/src/gui/rhi/MiniEngine_LICENSE.txt @@ -0,0 +1,22 @@ +The MIT License (MIT) + +Copyright (c) 2015 Microsoft + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. + diff --git a/src/gui/rhi/cs_mipmap_p.h b/src/gui/rhi/cs_mipmap_p.h new file mode 100644 index 0000000000..317cbe7b2e --- /dev/null +++ b/src/gui/rhi/cs_mipmap_p.h @@ -0,0 +1,939 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef CS_MIPMAP_P_H +#define CS_MIPMAP_P_H + +// +// W A R N I N G +// ------------- +// +// This file is not part of the Qt API. It exists for the convenience +// of other Qt classes. This header file may change from version to +// version without notice, or even be removed. +// +// We mean it. +// + +#include <QtCore/private/qglobal_p.h> + +#ifdef Q_OS_WIN + +#include <qt_windows.h> + +#if 0 +// +// Generated by Microsoft (R) HLSL Shader Compiler 10.1 +// +// +// Buffer Definitions: +// +// cbuffer CB0 +// { +// +// uint SrcMipLevel; // Offset: 0 Size: 4 +// uint NumMipLevels; // Offset: 4 Size: 4 +// float2 TexelSize; // Offset: 8 Size: 8 +// +// } +// +// +// Resource Bindings: +// +// Name Type Format Dim HLSL Bind Count +// ------------------------------ ---------- ------- ----------- -------------- ------ +// BilinearClamp sampler NA NA s0 1 +// SrcMip texture float4 2d t0 1 +// OutMip1 UAV float4 2d u0 1 +// OutMip2 UAV float4 2d u1 1 +// OutMip3 UAV float4 2d u2 1 +// OutMip4 UAV float4 2d u3 1 +// CB0 cbuffer NA NA cb0 1 +// +// +// +// Input signature: +// +// Name Index Mask Register SysValue Format Used +// -------------------- ----- ------ -------- -------- ------- ------ +// no Input +// +// Output signature: +// +// Name Index Mask Register SysValue Format Used +// -------------------- ----- ------ -------- -------- ------- ------ +// no Output +cs_5_0 +dcl_globalFlags refactoringAllowed +dcl_constantbuffer CB0[1], immediateIndexed +dcl_sampler s0, mode_default +dcl_resource_texture2d (float,float,float,float) t0 +dcl_uav_typed_texture2d (float,float,float,float) u0 +dcl_uav_typed_texture2d (float,float,float,float) u1 +dcl_uav_typed_texture2d (float,float,float,float) u2 +dcl_uav_typed_texture2d (float,float,float,float) u3 +dcl_input vThreadIDInGroupFlattened +dcl_input vThreadID.xy +dcl_temps 6 +dcl_tgsm_structured g0, 4, 64 +dcl_tgsm_structured g1, 4, 64 +dcl_tgsm_structured g2, 4, 64 +dcl_tgsm_structured g3, 4, 64 +dcl_thread_group 8, 8, 1 +utof r0.xy, vThreadID.xyxx +add r0.xy, r0.xyxx, l(0.250000, 0.250000, 0.000000, 0.000000) +mul r0.zw, r0.xxxy, cb0[0].zzzw +utof r1.x, cb0[0].x +sample_l_indexable(texture2d)(float,float,float,float) r2.xyzw, r0.zwzz, t0.xyzw, s0, r1.x +mul r3.xyz, cb0[0].zwzz, l(0.500000, 0.500000, 0.500000, 0.000000) +mov r3.w, l(0) +mad r3.xyzw, cb0[0].zwzw, r0.xyxy, r3.zwxy +sample_l_indexable(texture2d)(float,float,float,float) r4.xyzw, r3.xyxx, t0.xyzw, s0, r1.x +add r2.xyzw, r2.xyzw, r4.xyzw +mov r3.x, l(0) +mul r3.y, cb0[0].w, l(0.500000) +mad r0.xy, cb0[0].zwzz, r0.xyxx, r3.xyxx +sample_l_indexable(texture2d)(float,float,float,float) r0.xyzw, r0.xyxx, t0.xyzw, s0, r1.x +add r0.xyzw, r0.xyzw, r2.xyzw +sample_l_indexable(texture2d)(float,float,float,float) r1.xyzw, r3.zwzz, t0.xyzw, s0, r1.x +add r0.xyzw, r0.xyzw, r1.xyzw +mul r1.xyzw, r0.xyzw, l(0.250000, 0.250000, 0.250000, 0.250000) +store_uav_typed u0.xyzw, vThreadID.xyyy, r1.xyzw +ieq r2.x, cb0[0].y, l(1) +if_nz r2.x + ret +endif +store_structured g0.x, vThreadIDInGroupFlattened.x, l(0), r1.x +store_structured g1.x, vThreadIDInGroupFlattened.x, l(0), r1.y +store_structured g2.x, vThreadIDInGroupFlattened.x, l(0), r1.z +store_structured g3.x, vThreadIDInGroupFlattened.x, l(0), r1.w +sync_g_t +and r2.x, vThreadIDInGroupFlattened.x, l(9) +if_z r2.x + iadd r2.xyz, vThreadIDInGroupFlattened.xxxx, l(1, 8, 9, 0) + ld_structured r3.x, r2.x, l(0), g0.xxxx + ld_structured r3.y, r2.x, l(0), g1.xxxx + ld_structured r3.z, r2.x, l(0), g2.xxxx + ld_structured r3.w, r2.x, l(0), g3.xxxx + ld_structured r4.x, r2.y, l(0), g0.xxxx + ld_structured r4.y, r2.y, l(0), g1.xxxx + ld_structured r4.z, r2.y, l(0), g2.xxxx + ld_structured r4.w, r2.y, l(0), g3.xxxx + ld_structured r5.x, r2.z, l(0), g0.xxxx + ld_structured r5.y, r2.z, l(0), g1.xxxx + ld_structured r5.z, r2.z, l(0), g2.xxxx + ld_structured r5.w, r2.z, l(0), g3.xxxx + mad r0.xyzw, r0.xyzw, l(0.250000, 0.250000, 0.250000, 0.250000), r3.xyzw + add r0.xyzw, r4.xyzw, r0.xyzw + add r0.xyzw, r5.xyzw, r0.xyzw + mul r1.xyzw, r0.xyzw, l(0.250000, 0.250000, 0.250000, 0.250000) + ushr r0.xyzw, vThreadID.xyyy, l(1, 1, 1, 1) + store_uav_typed u1.xyzw, r0.xyzw, r1.xyzw + store_structured g0.x, vThreadIDInGroupFlattened.x, l(0), r1.x + store_structured g1.x, vThreadIDInGroupFlattened.x, l(0), r1.y + store_structured g2.x, vThreadIDInGroupFlattened.x, l(0), r1.z + store_structured g3.x, vThreadIDInGroupFlattened.x, l(0), r1.w +endif +ieq r0.x, cb0[0].y, l(2) +if_nz r0.x + ret +endif +sync_g_t +and r0.x, vThreadIDInGroupFlattened.x, l(27) +if_z r0.x + iadd r0.xyz, vThreadIDInGroupFlattened.xxxx, l(2, 16, 18, 0) + ld_structured r2.x, r0.x, l(0), g0.xxxx + ld_structured r2.y, r0.x, l(0), g1.xxxx + ld_structured r2.z, r0.x, l(0), g2.xxxx + ld_structured r2.w, r0.x, l(0), g3.xxxx + ld_structured r3.x, r0.y, l(0), g0.xxxx + ld_structured r3.y, r0.y, l(0), g1.xxxx + ld_structured r3.z, r0.y, l(0), g2.xxxx + ld_structured r3.w, r0.y, l(0), g3.xxxx + ld_structured r4.x, r0.z, l(0), g0.xxxx + ld_structured r4.y, r0.z, l(0), g1.xxxx + ld_structured r4.z, r0.z, l(0), g2.xxxx + ld_structured r4.w, r0.z, l(0), g3.xxxx + add r0.xyzw, r1.xyzw, r2.xyzw + add r0.xyzw, r3.xyzw, r0.xyzw + add r0.xyzw, r4.xyzw, r0.xyzw + mul r1.xyzw, r0.xyzw, l(0.250000, 0.250000, 0.250000, 0.250000) + ushr r0.xyzw, vThreadID.xyyy, l(2, 2, 2, 2) + store_uav_typed u2.xyzw, r0.xyzw, r1.xyzw + store_structured g0.x, vThreadIDInGroupFlattened.x, l(0), r1.x + store_structured g1.x, vThreadIDInGroupFlattened.x, l(0), r1.y + store_structured g2.x, vThreadIDInGroupFlattened.x, l(0), r1.z + store_structured g3.x, vThreadIDInGroupFlattened.x, l(0), r1.w +endif +ieq r0.x, cb0[0].y, l(3) +if_nz r0.x + ret +endif +sync_g_t +if_z vThreadIDInGroupFlattened.x + ld_structured r0.x, l(4), l(0), g0.xxxx + ld_structured r0.y, l(4), l(0), g1.xxxx + ld_structured r0.z, l(4), l(0), g2.xxxx + ld_structured r0.w, l(4), l(0), g3.xxxx + ld_structured r2.x, l(32), l(0), g0.xxxx + ld_structured r2.y, l(32), l(0), g1.xxxx + ld_structured r2.z, l(32), l(0), g2.xxxx + ld_structured r2.w, l(32), l(0), g3.xxxx + ld_structured r3.x, l(36), l(0), g0.xxxx + ld_structured r3.y, l(36), l(0), g1.xxxx + ld_structured r3.z, l(36), l(0), g2.xxxx + ld_structured r3.w, l(36), l(0), g3.xxxx + add r0.xyzw, r0.xyzw, r1.xyzw + add r0.xyzw, r2.xyzw, r0.xyzw + add r0.xyzw, r3.xyzw, r0.xyzw + mul r0.xyzw, r0.xyzw, l(0.250000, 0.250000, 0.250000, 0.250000) + ushr r1.xyzw, vThreadID.xyyy, l(3, 3, 3, 3) + store_uav_typed u3.xyzw, r1.xyzw, r0.xyzw +endif +ret +// Approximately 111 instruction slots used +#endif + +inline constexpr BYTE g_csMipmap[] = +{ + 68, 88, 66, 67, 133, 122, + 5, 181, 163, 163, 140, 185, + 158, 179, 4, 65, 180, 238, + 158, 10, 1, 0, 0, 0, + 60, 17, 0, 0, 5, 0, + 0, 0, 52, 0, 0, 0, + 200, 2, 0, 0, 216, 2, + 0, 0, 232, 2, 0, 0, + 160, 16, 0, 0, 82, 68, + 69, 70, 140, 2, 0, 0, + 1, 0, 0, 0, 88, 1, + 0, 0, 7, 0, 0, 0, + 60, 0, 0, 0, 0, 5, + 83, 67, 0, 1, 0, 0, + 100, 2, 0, 0, 82, 68, + 49, 49, 60, 0, 0, 0, + 24, 0, 0, 0, 32, 0, + 0, 0, 40, 0, 0, 0, + 36, 0, 0, 0, 12, 0, + 0, 0, 0, 0, 0, 0, + 28, 1, 0, 0, 3, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 1, 0, 0, 0, 1, 0, + 0, 0, 42, 1, 0, 0, + 2, 0, 0, 0, 5, 0, + 0, 0, 4, 0, 0, 0, + 255, 255, 255, 255, 0, 0, + 0, 0, 1, 0, 0, 0, + 13, 0, 0, 0, 49, 1, + 0, 0, 4, 0, 0, 0, + 5, 0, 0, 0, 4, 0, + 0, 0, 255, 255, 255, 255, + 0, 0, 0, 0, 1, 0, + 0, 0, 13, 0, 0, 0, + 57, 1, 0, 0, 4, 0, + 0, 0, 5, 0, 0, 0, + 4, 0, 0, 0, 255, 255, + 255, 255, 1, 0, 0, 0, + 1, 0, 0, 0, 13, 0, + 0, 0, 65, 1, 0, 0, + 4, 0, 0, 0, 5, 0, + 0, 0, 4, 0, 0, 0, + 255, 255, 255, 255, 2, 0, + 0, 0, 1, 0, 0, 0, + 13, 0, 0, 0, 73, 1, + 0, 0, 4, 0, 0, 0, + 5, 0, 0, 0, 4, 0, + 0, 0, 255, 255, 255, 255, + 3, 0, 0, 0, 1, 0, + 0, 0, 13, 0, 0, 0, + 81, 1, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 1, 0, 0, 0, 1, 0, + 0, 0, 66, 105, 108, 105, + 110, 101, 97, 114, 67, 108, + 97, 109, 112, 0, 83, 114, + 99, 77, 105, 112, 0, 79, + 117, 116, 77, 105, 112, 49, + 0, 79, 117, 116, 77, 105, + 112, 50, 0, 79, 117, 116, + 77, 105, 112, 51, 0, 79, + 117, 116, 77, 105, 112, 52, + 0, 67, 66, 48, 0, 171, + 171, 171, 81, 1, 0, 0, + 3, 0, 0, 0, 112, 1, + 0, 0, 16, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 232, 1, 0, 0, + 0, 0, 0, 0, 4, 0, + 0, 0, 2, 0, 0, 0, + 252, 1, 0, 0, 0, 0, + 0, 0, 255, 255, 255, 255, + 0, 0, 0, 0, 255, 255, + 255, 255, 0, 0, 0, 0, + 32, 2, 0, 0, 4, 0, + 0, 0, 4, 0, 0, 0, + 2, 0, 0, 0, 252, 1, + 0, 0, 0, 0, 0, 0, + 255, 255, 255, 255, 0, 0, + 0, 0, 255, 255, 255, 255, + 0, 0, 0, 0, 45, 2, + 0, 0, 8, 0, 0, 0, + 8, 0, 0, 0, 2, 0, + 0, 0, 64, 2, 0, 0, + 0, 0, 0, 0, 255, 255, + 255, 255, 0, 0, 0, 0, + 255, 255, 255, 255, 0, 0, + 0, 0, 83, 114, 99, 77, + 105, 112, 76, 101, 118, 101, + 108, 0, 100, 119, 111, 114, + 100, 0, 171, 171, 0, 0, + 19, 0, 1, 0, 1, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 244, 1, 0, 0, 78, 117, + 109, 77, 105, 112, 76, 101, + 118, 101, 108, 115, 0, 84, + 101, 120, 101, 108, 83, 105, + 122, 101, 0, 102, 108, 111, + 97, 116, 50, 0, 171, 171, + 1, 0, 3, 0, 1, 0, + 2, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 55, 2, 0, 0, + 77, 105, 99, 114, 111, 115, + 111, 102, 116, 32, 40, 82, + 41, 32, 72, 76, 83, 76, + 32, 83, 104, 97, 100, 101, + 114, 32, 67, 111, 109, 112, + 105, 108, 101, 114, 32, 49, + 48, 46, 49, 0, 73, 83, + 71, 78, 8, 0, 0, 0, + 0, 0, 0, 0, 8, 0, + 0, 0, 79, 83, 71, 78, + 8, 0, 0, 0, 0, 0, + 0, 0, 8, 0, 0, 0, + 83, 72, 69, 88, 176, 13, + 0, 0, 80, 0, 5, 0, + 108, 3, 0, 0, 106, 8, + 0, 1, 89, 0, 0, 4, + 70, 142, 32, 0, 0, 0, + 0, 0, 1, 0, 0, 0, + 90, 0, 0, 3, 0, 96, + 16, 0, 0, 0, 0, 0, + 88, 24, 0, 4, 0, 112, + 16, 0, 0, 0, 0, 0, + 85, 85, 0, 0, 156, 24, + 0, 4, 0, 224, 17, 0, + 0, 0, 0, 0, 85, 85, + 0, 0, 156, 24, 0, 4, + 0, 224, 17, 0, 1, 0, + 0, 0, 85, 85, 0, 0, + 156, 24, 0, 4, 0, 224, + 17, 0, 2, 0, 0, 0, + 85, 85, 0, 0, 156, 24, + 0, 4, 0, 224, 17, 0, + 3, 0, 0, 0, 85, 85, + 0, 0, 95, 0, 0, 2, + 0, 64, 2, 0, 95, 0, + 0, 2, 50, 0, 2, 0, + 104, 0, 0, 2, 6, 0, + 0, 0, 160, 0, 0, 5, + 0, 240, 17, 0, 0, 0, + 0, 0, 4, 0, 0, 0, + 64, 0, 0, 0, 160, 0, + 0, 5, 0, 240, 17, 0, + 1, 0, 0, 0, 4, 0, + 0, 0, 64, 0, 0, 0, + 160, 0, 0, 5, 0, 240, + 17, 0, 2, 0, 0, 0, + 4, 0, 0, 0, 64, 0, + 0, 0, 160, 0, 0, 5, + 0, 240, 17, 0, 3, 0, + 0, 0, 4, 0, 0, 0, + 64, 0, 0, 0, 155, 0, + 0, 4, 8, 0, 0, 0, + 8, 0, 0, 0, 1, 0, + 0, 0, 86, 0, 0, 4, + 50, 0, 16, 0, 0, 0, + 0, 0, 70, 0, 2, 0, + 0, 0, 0, 10, 50, 0, + 16, 0, 0, 0, 0, 0, + 70, 0, 16, 0, 0, 0, + 0, 0, 2, 64, 0, 0, + 0, 0, 128, 62, 0, 0, + 128, 62, 0, 0, 0, 0, + 0, 0, 0, 0, 56, 0, + 0, 8, 194, 0, 16, 0, + 0, 0, 0, 0, 6, 4, + 16, 0, 0, 0, 0, 0, + 166, 142, 32, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 86, 0, 0, 6, 18, 0, + 16, 0, 1, 0, 0, 0, + 10, 128, 32, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 72, 0, 0, 141, 194, 0, + 0, 128, 67, 85, 21, 0, + 242, 0, 16, 0, 2, 0, + 0, 0, 230, 10, 16, 0, + 0, 0, 0, 0, 70, 126, + 16, 0, 0, 0, 0, 0, + 0, 96, 16, 0, 0, 0, + 0, 0, 10, 0, 16, 0, + 1, 0, 0, 0, 56, 0, + 0, 11, 114, 0, 16, 0, + 3, 0, 0, 0, 230, 138, + 32, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 2, 64, + 0, 0, 0, 0, 0, 63, + 0, 0, 0, 63, 0, 0, + 0, 63, 0, 0, 0, 0, + 54, 0, 0, 5, 130, 0, + 16, 0, 3, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 50, 0, 0, 10, + 242, 0, 16, 0, 3, 0, + 0, 0, 230, 142, 32, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 70, 4, 16, 0, + 0, 0, 0, 0, 230, 4, + 16, 0, 3, 0, 0, 0, + 72, 0, 0, 141, 194, 0, + 0, 128, 67, 85, 21, 0, + 242, 0, 16, 0, 4, 0, + 0, 0, 70, 0, 16, 0, + 3, 0, 0, 0, 70, 126, + 16, 0, 0, 0, 0, 0, + 0, 96, 16, 0, 0, 0, + 0, 0, 10, 0, 16, 0, + 1, 0, 0, 0, 0, 0, + 0, 7, 242, 0, 16, 0, + 2, 0, 0, 0, 70, 14, + 16, 0, 2, 0, 0, 0, + 70, 14, 16, 0, 4, 0, + 0, 0, 54, 0, 0, 5, + 18, 0, 16, 0, 3, 0, + 0, 0, 1, 64, 0, 0, + 0, 0, 0, 0, 56, 0, + 0, 8, 34, 0, 16, 0, + 3, 0, 0, 0, 58, 128, + 32, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 1, 64, + 0, 0, 0, 0, 0, 63, + 50, 0, 0, 10, 50, 0, + 16, 0, 0, 0, 0, 0, + 230, 138, 32, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 70, 0, 16, 0, 0, 0, + 0, 0, 70, 0, 16, 0, + 3, 0, 0, 0, 72, 0, + 0, 141, 194, 0, 0, 128, + 67, 85, 21, 0, 242, 0, + 16, 0, 0, 0, 0, 0, + 70, 0, 16, 0, 0, 0, + 0, 0, 70, 126, 16, 0, + 0, 0, 0, 0, 0, 96, + 16, 0, 0, 0, 0, 0, + 10, 0, 16, 0, 1, 0, + 0, 0, 0, 0, 0, 7, + 242, 0, 16, 0, 0, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 2, 0, 0, 0, + 72, 0, 0, 141, 194, 0, + 0, 128, 67, 85, 21, 0, + 242, 0, 16, 0, 1, 0, + 0, 0, 230, 10, 16, 0, + 3, 0, 0, 0, 70, 126, + 16, 0, 0, 0, 0, 0, + 0, 96, 16, 0, 0, 0, + 0, 0, 10, 0, 16, 0, + 1, 0, 0, 0, 0, 0, + 0, 7, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 70, 14, 16, 0, 1, 0, + 0, 0, 56, 0, 0, 10, + 242, 0, 16, 0, 1, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 2, 64, + 0, 0, 0, 0, 128, 62, + 0, 0, 128, 62, 0, 0, + 128, 62, 0, 0, 128, 62, + 164, 0, 0, 6, 242, 224, + 17, 0, 0, 0, 0, 0, + 70, 5, 2, 0, 70, 14, + 16, 0, 1, 0, 0, 0, + 32, 0, 0, 8, 18, 0, + 16, 0, 2, 0, 0, 0, + 26, 128, 32, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 1, 0, + 0, 0, 31, 0, 4, 3, + 10, 0, 16, 0, 2, 0, + 0, 0, 62, 0, 0, 1, + 21, 0, 0, 1, 168, 0, + 0, 8, 18, 240, 17, 0, + 0, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 0, 0, 0, 0, 10, 0, + 16, 0, 1, 0, 0, 0, + 168, 0, 0, 8, 18, 240, + 17, 0, 1, 0, 0, 0, + 10, 64, 2, 0, 1, 64, + 0, 0, 0, 0, 0, 0, + 26, 0, 16, 0, 1, 0, + 0, 0, 168, 0, 0, 8, + 18, 240, 17, 0, 2, 0, + 0, 0, 10, 64, 2, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 42, 0, 16, 0, + 1, 0, 0, 0, 168, 0, + 0, 8, 18, 240, 17, 0, + 3, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 0, 0, 0, 0, 58, 0, + 16, 0, 1, 0, 0, 0, + 190, 24, 0, 1, 1, 0, + 0, 6, 18, 0, 16, 0, + 2, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 9, 0, 0, 0, 31, 0, + 0, 3, 10, 0, 16, 0, + 2, 0, 0, 0, 30, 0, + 0, 9, 114, 0, 16, 0, + 2, 0, 0, 0, 6, 64, + 2, 0, 2, 64, 0, 0, + 1, 0, 0, 0, 8, 0, + 0, 0, 9, 0, 0, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 3, 0, 0, 0, 10, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 3, 0, 0, 0, 10, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 3, 0, 0, 0, 10, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 3, 0, 0, 0, 10, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 4, 0, 0, 0, 26, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 4, 0, 0, 0, 26, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 4, 0, 0, 0, 26, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 4, 0, 0, 0, 26, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 5, 0, 0, 0, 42, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 5, 0, 0, 0, 42, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 5, 0, 0, 0, 42, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 5, 0, 0, 0, 42, 0, + 16, 0, 2, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 50, 0, + 0, 12, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 2, 64, 0, 0, 0, 0, + 128, 62, 0, 0, 128, 62, + 0, 0, 128, 62, 0, 0, + 128, 62, 70, 14, 16, 0, + 3, 0, 0, 0, 0, 0, + 0, 7, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 4, 0, 0, 0, + 70, 14, 16, 0, 0, 0, + 0, 0, 0, 0, 0, 7, + 242, 0, 16, 0, 0, 0, + 0, 0, 70, 14, 16, 0, + 5, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 56, 0, 0, 10, 242, 0, + 16, 0, 1, 0, 0, 0, + 70, 14, 16, 0, 0, 0, + 0, 0, 2, 64, 0, 0, + 0, 0, 128, 62, 0, 0, + 128, 62, 0, 0, 128, 62, + 0, 0, 128, 62, 85, 0, + 0, 9, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 5, + 2, 0, 2, 64, 0, 0, + 1, 0, 0, 0, 1, 0, + 0, 0, 1, 0, 0, 0, + 1, 0, 0, 0, 164, 0, + 0, 7, 242, 224, 17, 0, + 1, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 70, 14, 16, 0, 1, 0, + 0, 0, 168, 0, 0, 8, + 18, 240, 17, 0, 0, 0, + 0, 0, 10, 64, 2, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 10, 0, 16, 0, + 1, 0, 0, 0, 168, 0, + 0, 8, 18, 240, 17, 0, + 1, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 0, 0, 0, 0, 26, 0, + 16, 0, 1, 0, 0, 0, + 168, 0, 0, 8, 18, 240, + 17, 0, 2, 0, 0, 0, + 10, 64, 2, 0, 1, 64, + 0, 0, 0, 0, 0, 0, + 42, 0, 16, 0, 1, 0, + 0, 0, 168, 0, 0, 8, + 18, 240, 17, 0, 3, 0, + 0, 0, 10, 64, 2, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 58, 0, 16, 0, + 1, 0, 0, 0, 21, 0, + 0, 1, 32, 0, 0, 8, + 18, 0, 16, 0, 0, 0, + 0, 0, 26, 128, 32, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 1, 64, 0, 0, + 2, 0, 0, 0, 31, 0, + 4, 3, 10, 0, 16, 0, + 0, 0, 0, 0, 62, 0, + 0, 1, 21, 0, 0, 1, + 190, 24, 0, 1, 1, 0, + 0, 6, 18, 0, 16, 0, + 0, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 27, 0, 0, 0, 31, 0, + 0, 3, 10, 0, 16, 0, + 0, 0, 0, 0, 30, 0, + 0, 9, 114, 0, 16, 0, + 0, 0, 0, 0, 6, 64, + 2, 0, 2, 64, 0, 0, + 2, 0, 0, 0, 16, 0, + 0, 0, 18, 0, 0, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 2, 0, 0, 0, 10, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 2, 0, 0, 0, 10, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 2, 0, 0, 0, 10, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 2, 0, 0, 0, 10, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 3, 0, 0, 0, 26, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 3, 0, 0, 0, 26, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 3, 0, 0, 0, 26, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 3, 0, 0, 0, 26, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 4, 0, 0, 0, 42, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 4, 0, 0, 0, 42, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 4, 0, 0, 0, 42, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 4, 0, 0, 0, 42, 0, + 16, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 0, 0, + 0, 7, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 1, 0, 0, 0, + 70, 14, 16, 0, 2, 0, + 0, 0, 0, 0, 0, 7, + 242, 0, 16, 0, 0, 0, + 0, 0, 70, 14, 16, 0, + 3, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 0, 0, 0, 7, 242, 0, + 16, 0, 0, 0, 0, 0, + 70, 14, 16, 0, 4, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 56, 0, + 0, 10, 242, 0, 16, 0, + 1, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 2, 64, 0, 0, 0, 0, + 128, 62, 0, 0, 128, 62, + 0, 0, 128, 62, 0, 0, + 128, 62, 85, 0, 0, 9, + 242, 0, 16, 0, 0, 0, + 0, 0, 70, 5, 2, 0, + 2, 64, 0, 0, 2, 0, + 0, 0, 2, 0, 0, 0, + 2, 0, 0, 0, 2, 0, + 0, 0, 164, 0, 0, 7, + 242, 224, 17, 0, 2, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 1, 0, 0, 0, + 168, 0, 0, 8, 18, 240, + 17, 0, 0, 0, 0, 0, + 10, 64, 2, 0, 1, 64, + 0, 0, 0, 0, 0, 0, + 10, 0, 16, 0, 1, 0, + 0, 0, 168, 0, 0, 8, + 18, 240, 17, 0, 1, 0, + 0, 0, 10, 64, 2, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 26, 0, 16, 0, + 1, 0, 0, 0, 168, 0, + 0, 8, 18, 240, 17, 0, + 2, 0, 0, 0, 10, 64, + 2, 0, 1, 64, 0, 0, + 0, 0, 0, 0, 42, 0, + 16, 0, 1, 0, 0, 0, + 168, 0, 0, 8, 18, 240, + 17, 0, 3, 0, 0, 0, + 10, 64, 2, 0, 1, 64, + 0, 0, 0, 0, 0, 0, + 58, 0, 16, 0, 1, 0, + 0, 0, 21, 0, 0, 1, + 32, 0, 0, 8, 18, 0, + 16, 0, 0, 0, 0, 0, + 26, 128, 32, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 1, 64, 0, 0, 3, 0, + 0, 0, 31, 0, 4, 3, + 10, 0, 16, 0, 0, 0, + 0, 0, 62, 0, 0, 1, + 21, 0, 0, 1, 190, 24, + 0, 1, 31, 0, 0, 2, + 10, 64, 2, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 0, 0, 0, 0, 1, 64, + 0, 0, 4, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 0, 0, 0, 0, 1, 64, + 0, 0, 4, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 0, 0, 0, 0, 1, 64, + 0, 0, 4, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 0, 0, 0, 0, 1, 64, + 0, 0, 4, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 2, 0, 0, 0, 1, 64, + 0, 0, 32, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 2, 0, 0, 0, 1, 64, + 0, 0, 32, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 2, 0, 0, 0, 1, 64, + 0, 0, 32, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 2, 0, 0, 0, 1, 64, + 0, 0, 32, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 167, 0, + 0, 9, 18, 0, 16, 0, + 3, 0, 0, 0, 1, 64, + 0, 0, 36, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 0, 0, 0, 0, 167, 0, + 0, 9, 34, 0, 16, 0, + 3, 0, 0, 0, 1, 64, + 0, 0, 36, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 1, 0, 0, 0, 167, 0, + 0, 9, 66, 0, 16, 0, + 3, 0, 0, 0, 1, 64, + 0, 0, 36, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 2, 0, 0, 0, 167, 0, + 0, 9, 130, 0, 16, 0, + 3, 0, 0, 0, 1, 64, + 0, 0, 36, 0, 0, 0, + 1, 64, 0, 0, 0, 0, + 0, 0, 6, 240, 17, 0, + 3, 0, 0, 0, 0, 0, + 0, 7, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 70, 14, 16, 0, 1, 0, + 0, 0, 0, 0, 0, 7, + 242, 0, 16, 0, 0, 0, + 0, 0, 70, 14, 16, 0, + 2, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 0, 0, 0, 7, 242, 0, + 16, 0, 0, 0, 0, 0, + 70, 14, 16, 0, 3, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 56, 0, + 0, 10, 242, 0, 16, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 2, 64, 0, 0, 0, 0, + 128, 62, 0, 0, 128, 62, + 0, 0, 128, 62, 0, 0, + 128, 62, 85, 0, 0, 9, + 242, 0, 16, 0, 1, 0, + 0, 0, 70, 5, 2, 0, + 2, 64, 0, 0, 3, 0, + 0, 0, 3, 0, 0, 0, + 3, 0, 0, 0, 3, 0, + 0, 0, 164, 0, 0, 7, + 242, 224, 17, 0, 3, 0, + 0, 0, 70, 14, 16, 0, + 1, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 21, 0, 0, 1, 62, 0, + 0, 1, 83, 84, 65, 84, + 148, 0, 0, 0, 111, 0, + 0, 0, 6, 0, 0, 0, + 0, 0, 0, 0, 2, 0, + 0, 0, 22, 0, 0, 0, + 5, 0, 0, 0, 5, 0, + 0, 0, 4, 0, 0, 0, + 6, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 4, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 2, 0, 0, 0, + 0, 0, 0, 0, 2, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 3, 0, 0, 0, + 0, 0, 0, 0, 4, 0, + 0, 0 +}; + +#endif // Q_OS_WIN + +#endif // CS_MIPMAP_P_H diff --git a/src/gui/rhi/cs_tdr_p.h b/src/gui/rhi/cs_tdr_p.h deleted file mode 100644 index 534e793ad2..0000000000 --- a/src/gui/rhi/cs_tdr_p.h +++ /dev/null @@ -1,228 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: http://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef CS_TDR_P_H -#define CS_TDR_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists for the convenience -// of other Qt classes. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include <QtCore/qglobal.h> - -#ifdef Q_OS_WIN - -#include <qt_windows.h> - -#if 0 -// -// Generated by Microsoft (R) HLSL Shader Compiler 10.1 -// -// -// Buffer Definitions: -// -// cbuffer ConstantBuffer -// { -// -// uint zero; // Offset: 0 Size: 4 -// -// } -// -// -// Resource Bindings: -// -// Name Type Format Dim HLSL Bind Count -// ------------------------------ ---------- ------- ----------- -------------- ------ -// uav UAV uint buf u0 1 -// ConstantBuffer cbuffer NA NA cb0 1 -// -// -// -// Input signature: -// -// Name Index Mask Register SysValue Format Used -// -------------------- ----- ------ -------- -------- ------- ------ -// no Input -// -// Output signature: -// -// Name Index Mask Register SysValue Format Used -// -------------------- ----- ------ -------- -------- ------- ------ -// no Output -cs_5_0 -dcl_globalFlags refactoringAllowed -dcl_constantbuffer CB0[1], immediateIndexed -dcl_uav_typed_buffer (uint,uint,uint,uint) u0 -dcl_input vThreadID.x -dcl_thread_group 256, 1, 1 -loop - breakc_nz cb0[0].x - store_uav_typed u0.xyzw, vThreadID.xxxx, cb0[0].xxxx -endloop -ret -// Approximately 5 instruction slots used -#endif - -const BYTE g_killDeviceByTimingOut[] = -{ - 68, 88, 66, 67, 217, 62, - 220, 38, 136, 51, 86, 245, - 161, 96, 18, 35, 141, 17, - 26, 13, 1, 0, 0, 0, - 164, 2, 0, 0, 5, 0, - 0, 0, 52, 0, 0, 0, - 100, 1, 0, 0, 116, 1, - 0, 0, 132, 1, 0, 0, - 8, 2, 0, 0, 82, 68, - 69, 70, 40, 1, 0, 0, - 1, 0, 0, 0, 144, 0, - 0, 0, 2, 0, 0, 0, - 60, 0, 0, 0, 0, 5, - 83, 67, 0, 1, 0, 0, - 0, 1, 0, 0, 82, 68, - 49, 49, 60, 0, 0, 0, - 24, 0, 0, 0, 32, 0, - 0, 0, 40, 0, 0, 0, - 36, 0, 0, 0, 12, 0, - 0, 0, 0, 0, 0, 0, - 124, 0, 0, 0, 4, 0, - 0, 0, 4, 0, 0, 0, - 1, 0, 0, 0, 255, 255, - 255, 255, 0, 0, 0, 0, - 1, 0, 0, 0, 0, 0, - 0, 0, 128, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 1, 0, 0, 0, - 0, 0, 0, 0, 117, 97, - 118, 0, 67, 111, 110, 115, - 116, 97, 110, 116, 66, 117, - 102, 102, 101, 114, 0, 171, - 128, 0, 0, 0, 1, 0, - 0, 0, 168, 0, 0, 0, - 16, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 208, 0, 0, 0, 0, 0, - 0, 0, 4, 0, 0, 0, - 2, 0, 0, 0, 220, 0, - 0, 0, 0, 0, 0, 0, - 255, 255, 255, 255, 0, 0, - 0, 0, 255, 255, 255, 255, - 0, 0, 0, 0, 122, 101, - 114, 111, 0, 100, 119, 111, - 114, 100, 0, 171, 0, 0, - 19, 0, 1, 0, 1, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 213, 0, 0, 0, 77, 105, - 99, 114, 111, 115, 111, 102, - 116, 32, 40, 82, 41, 32, - 72, 76, 83, 76, 32, 83, - 104, 97, 100, 101, 114, 32, - 67, 111, 109, 112, 105, 108, - 101, 114, 32, 49, 48, 46, - 49, 0, 73, 83, 71, 78, - 8, 0, 0, 0, 0, 0, - 0, 0, 8, 0, 0, 0, - 79, 83, 71, 78, 8, 0, - 0, 0, 0, 0, 0, 0, - 8, 0, 0, 0, 83, 72, - 69, 88, 124, 0, 0, 0, - 80, 0, 5, 0, 31, 0, - 0, 0, 106, 8, 0, 1, - 89, 0, 0, 4, 70, 142, - 32, 0, 0, 0, 0, 0, - 1, 0, 0, 0, 156, 8, - 0, 4, 0, 224, 17, 0, - 0, 0, 0, 0, 68, 68, - 0, 0, 95, 0, 0, 2, - 18, 0, 2, 0, 155, 0, - 0, 4, 0, 1, 0, 0, - 1, 0, 0, 0, 1, 0, - 0, 0, 48, 0, 0, 1, - 3, 0, 4, 4, 10, 128, - 32, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 164, 0, - 0, 7, 242, 224, 17, 0, - 0, 0, 0, 0, 6, 0, - 2, 0, 6, 128, 32, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 22, 0, 0, 1, - 62, 0, 0, 1, 83, 84, - 65, 84, 148, 0, 0, 0, - 5, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 1, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 2, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, - 1, 0, 0, 0 -}; - -#endif // Q_OS_WIN - -#endif // CS_TDR_P_H diff --git a/src/gui/rhi/mipmap.hlsl b/src/gui/rhi/mipmap.hlsl new file mode 100644 index 0000000000..ac293e07f9 --- /dev/null +++ b/src/gui/rhi/mipmap.hlsl @@ -0,0 +1,117 @@ +// Copyright (c) Microsoft. All rights reserved. +// This code is licensed under the MIT License (MIT). +// THIS CODE IS PROVIDED *AS IS* WITHOUT WARRANTY OF +// ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING ANY +// IMPLIED WARRANTIES OF FITNESS FOR A PARTICULAR +// PURPOSE, MERCHANTABILITY, OR NON-INFRINGEMENT. + +RWTexture2D<float4> OutMip1 : register(u0); +RWTexture2D<float4> OutMip2 : register(u1); +RWTexture2D<float4> OutMip3 : register(u2); +RWTexture2D<float4> OutMip4 : register(u3); +Texture2D<float4> SrcMip : register(t0); +SamplerState BilinearClamp : register(s0); + +cbuffer CB0 : register(b0) +{ + uint SrcMipLevel; // Texture level of source mip + uint NumMipLevels; // Number of OutMips to write: [1, 4] + float2 TexelSize; // 1.0 / OutMip1.Dimensions +} + +// The reason for separating channels is to reduce bank conflicts in the +// local data memory controller. A large stride will cause more threads +// to collide on the same memory bank. +groupshared float gs_R[64]; +groupshared float gs_G[64]; +groupshared float gs_B[64]; +groupshared float gs_A[64]; + +void StoreColor( uint Index, float4 Color ) +{ + gs_R[Index] = Color.r; + gs_G[Index] = Color.g; + gs_B[Index] = Color.b; + gs_A[Index] = Color.a; +} + +float4 LoadColor( uint Index ) +{ + return float4( gs_R[Index], gs_G[Index], gs_B[Index], gs_A[Index]); +} + +[numthreads( 8, 8, 1 )] +void csMain( uint GI : SV_GroupIndex, uint3 DTid : SV_DispatchThreadID ) +{ + // Use 4 bilinear samples to guarantee we don't undersample when downsizing by more than 2x + // in both directions. + float2 UV1 = TexelSize * (DTid.xy + float2(0.25, 0.25)); + float2 O = TexelSize * 0.5; + float4 Src1 = SrcMip.SampleLevel(BilinearClamp, UV1, SrcMipLevel); + Src1 += SrcMip.SampleLevel(BilinearClamp, UV1 + float2(O.x, 0.0), SrcMipLevel); + Src1 += SrcMip.SampleLevel(BilinearClamp, UV1 + float2(0.0, O.y), SrcMipLevel); + Src1 += SrcMip.SampleLevel(BilinearClamp, UV1 + float2(O.x, O.y), SrcMipLevel); + Src1 *= 0.25; + + OutMip1[DTid.xy] = Src1; + + // A scalar (constant) branch can exit all threads coherently. + if (NumMipLevels == 1) + return; + + // Without lane swizzle operations, the only way to share data with other + // threads is through LDS. + StoreColor(GI, Src1); + + // This guarantees all LDS writes are complete and that all threads have + // executed all instructions so far (and therefore have issued their LDS + // write instructions.) + GroupMemoryBarrierWithGroupSync(); + + // With low three bits for X and high three bits for Y, this bit mask + // (binary: 001001) checks that X and Y are even. + if ((GI & 0x9) == 0) + { + float4 Src2 = LoadColor(GI + 0x01); + float4 Src3 = LoadColor(GI + 0x08); + float4 Src4 = LoadColor(GI + 0x09); + Src1 = 0.25 * (Src1 + Src2 + Src3 + Src4); + + OutMip2[DTid.xy / 2] = Src1; + StoreColor(GI, Src1); + } + + if (NumMipLevels == 2) + return; + + GroupMemoryBarrierWithGroupSync(); + + // This bit mask (binary: 011011) checks that X and Y are multiples of four. + if ((GI & 0x1B) == 0) + { + float4 Src2 = LoadColor(GI + 0x02); + float4 Src3 = LoadColor(GI + 0x10); + float4 Src4 = LoadColor(GI + 0x12); + Src1 = 0.25 * (Src1 + Src2 + Src3 + Src4); + + OutMip3[DTid.xy / 4] = Src1; + StoreColor(GI, Src1); + } + + if (NumMipLevels == 3) + return; + + GroupMemoryBarrierWithGroupSync(); + + // This bit mask would be 111111 (X & Y multiples of 8), but only one + // thread fits that criteria. + if (GI == 0) + { + float4 Src2 = LoadColor(GI + 0x04); + float4 Src3 = LoadColor(GI + 0x20); + float4 Src4 = LoadColor(GI + 0x24); + Src1 = 0.25 * (Src1 + Src2 + Src3 + Src4); + + OutMip4[DTid.xy / 8] = Src1; + } +} diff --git a/src/gui/rhi/qrhi.cpp b/src/gui/rhi/qrhi.cpp index 4ae75b5df5..a39709c726 100644 --- a/src/gui/rhi/qrhi.cpp +++ b/src/gui/rhi/qrhi.cpp @@ -1,68 +1,37 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhi_p_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhi_p.h" #include <qmath.h> #include <QLoggingCategory> -#include "qrhinull_p_p.h" +#include "qrhinull_p.h" #ifndef QT_NO_OPENGL -#include "qrhigles2_p_p.h" +#include "qrhigles2_p.h" #endif #if QT_CONFIG(vulkan) -#include "qrhivulkan_p_p.h" +#include "qrhivulkan_p.h" #endif #ifdef Q_OS_WIN -#include "qrhid3d11_p_p.h" +#include "qrhid3d11_p.h" +#include "qrhid3d12_p.h" #endif -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) -#include "qrhimetal_p_p.h" +#if QT_CONFIG(metal) +#include "qrhimetal_p.h" #endif +#include <memory> + QT_BEGIN_NAMESPACE Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") /*! \class QRhi - \internal - \inmodule QtGui + \ingroup painting-3D + \inmodule QtGuiPrivate + \inheaderfile rhi/qrhi.h + \since 6.6 \brief Accelerated 2D/3D graphics API abstraction. @@ -73,50 +42,55 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \l{https://developer.apple.com/metal/}{Metal}, and \l{https://www.khronos.org/vulkan/}{Vulkan}. - Some of the main design goals are: - - \list - - \li Simple, minimal, understandable, extensible. Follow the proven path of the - Qt Quick scenegraph. - - \li Aim to be a product - and in the bigger picture, part of a product (Qt) - - that is usable out of the box both by internal (such as, Qt Quick) and, - eventually, external users. - - \li Not a complete 1:1 wrapper for any of the underlying APIs. The feature set - is tuned towards the needs of Qt's 2D and 3D offering (QPainter, Qt Quick, Qt - 3D Studio). Iterate and evolve in a sustainable manner. - - \li Intrinsically cross-platform, without reinventing: abstracting - cross-platform aspects of certain APIs (such as, OpenGL context creation and - windowing system interfaces, Vulkan instance and surface management) is not in - scope here. These are delegated to the existing QtGui facilities (QWindow, - QOpenGLContext, QVulkanInstance) and its backing QPA architecture. - - \endlist + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qrhi.h>}. Each QRhi instance is backed by a backend for a specific graphics API. The selection of the backend is a run time choice and is up to the application or library that creates the QRhi instance. Some backends are available on multiple platforms (OpenGL, Vulkan, Null), while APIs specific to a given platform are only available when running on the platform in question (Metal - on macOS/iOS/tvOS, Direct3D on Windows). + on macOS/iOS, Direct3D on Windows). The available backends currently are: \list - \li OpenGL 2.1 or OpenGL ES 2.0 or newer. Some extensions are utilized when - present, for example to enable multisample framebuffers. + \li OpenGL 2.1 / OpenGL ES 2.0 or newer. Some extensions and newer core + specification features are utilized when present, for example to enable + multisample framebuffers or compute shaders. Operating in core profile + contexts is supported as well. If necessary, applications can query the + \l{QRhi::Feature}{feature flags} at runtime to check for features that are + not supported in the OpenGL context backing the QRhi. The OpenGL backend + builds on QOpenGLContext, QOpenGLFunctions, and the related cross-platform + infrastructure of the Qt GUI module. - \li Direct3D 11.1 + \li Direct3D 11.1 or newer, with Shader Model 5.0 or newer. When the D3D + runtime has no support for 11.1 features or Shader Model 5.0, + initialization using an accelerated graphics device will fail, but using + the + \l{https://learn.microsoft.com/en-us/windows/win32/direct3darticles/directx-warp}{software + adapter} is still an option. - \li Metal + \li Direct3D 12 on Windows 10 version 1703 and newer, with Shader Model 5.0 + or newer. Qt requires ID3D12Device2 to be present, hence the requirement + for at least version 1703 of Windows 10. The D3D12 device is by default + created with specifying a minimum feature level of + \c{D3D_FEATURE_LEVEL_11_0}. - \li Vulkan 1.0, optionally with some extensions that are part of Vulkan 1.1 + \li Metal 1.2 or newer. - \li Null - A "dummy" backend that issues no graphics calls at all. + \li Vulkan 1.0 or newer, optionally utilizing some Vulkan 1.1 level + features. + + \li Null, a "dummy" backend that issues no graphics calls at all. \endlist @@ -126,15 +100,60 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") are then generated from that, together with reflection information (inputs, outputs, shader resources). This is then packed into easily and efficiently serializable QShader instances. The compilers and tools to generate such - shaders are not part of QRhi, but the core classes for using such shaders, - QShader and QShaderDescription, are. + shaders are not part of QRhi and the Qt GUI module, but the core classes + for using such shaders, QShader and QShaderDescription, are. The APIs and + tools for performing compilation and translation are part of the Qt Shader + Tools module. + + See the \l{RHI Window Example} for an introductory example of creating a + portable, cross-platform application that performs accelerated 3D rendering + onto a QWindow using QRhi. + + \section1 An Impression of the API + + To provide a quick look at the API with a short yet complete example that + does not involve window-related setup, the following is a complete, + runnable cross-platform application that renders 20 frames off-screen, and + then saves the generated images to files after reading back the texture + contents from the GPU. For an example that renders on-screen, which then + involves setting up a QWindow and a swapchain, refer to the + \l{RHI Window Example}. - \section2 Design Fundamentals + For brevity, the initialization of the QRhi is done based on the platform: + the sample code here chooses Direct 3D 12 on Windows, Metal on macOS and + iOS, and Vulkan otherwise. OpenGL and Direct 3D 11 are never used by this + application, but support for those could be introduced with a few + additional lines. + + \snippet rhioffscreen/main.cpp 0 + + The result of the application is 20 \c PNG images (frame0.png - + frame19.png). These contain a rotating triangle with varying opacity over a + green background. + + The vertex and fragment shaders are expected to be processed and packaged + into \c{.qsb} files. The Vulkan-compatible GLSL source code is the + following: + + \e color.vert + \snippet rhioffscreen/color.vert 0 + + \e color.frag + \snippet rhioffscreen/color.frag 0 + + To manually compile and transpile these shaders to a number of targets + (SPIR-V, HLSL, MSL, GLSL) and generate the \c{.qsb} files the application + loads at run time, run \c{qsb --qt6 color.vert -o color.vert.qsb} and + \c{qsb --qt6 color.frag -o color.frag.qsb}. Alternatively, the Qt Shader + Tools module offers build system integration for CMake, the + \c qt_add_shaders() CMake function, that can achieve the same at build time. + + \section1 Design Fundamentals A QRhi cannot be instantiated directly. Instead, use the create() function. Delete the QRhi instance normally to release the graphics device. - \section3 Resources + \section2 Resources Instances of classes deriving from QRhiResource, such as, QRhiBuffer, QRhiTexture, etc., encapsulate zero, one, or more native graphics @@ -142,10 +161,10 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") functions of the QRhi, such as, newBuffer(), newTexture(), newTextureRenderTarget(), newSwapChain(). - \badcode - vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertexData)); - if (!vbuf->create()) { error } - ... + \code + QRhiBuffer *vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertexData)); + if (!vbuf->create()) { error(); } + // ... delete vbuf; \endcode @@ -154,9 +173,10 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \li The returned value from functions like newBuffer() is always owned by the caller. - \li Just creating a QRhiResource subclass never allocates or initializes any - native resources. That is only done when calling the \c create() function of a - subclass, for example, QRhiBuffer::create() or QRhiTexture::create(). + \li Just creating an instance of a QRhiResource subclass never allocates or + initializes any native resources. That is only done when calling the + \c create() function of a subclass, for example, QRhiBuffer::create() or + QRhiTexture::create(). \li The exceptions are QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor(), @@ -180,15 +200,15 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \li Note that this does not mean that a QRhiResource can freely be destroy()'ed or deleted within a frame (that is, in a - \l{QRhiCommandBuffer::beginFrame()}{beginFrame()} - - \l{QRhiCommandBuffer::endFrame()}{endFrame()} section). As a general rule, - all referenced QRhiResource objects must stay unchanged until the frame is - submitted by calling \l{QRhiCommandBuffer::endFrame()}{endFrame()}. To ease - this, QRhiResource::deleteLater() is provided as a convenience. + \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} + section). As a general rule, all referenced QRhiResource objects must stay + unchanged until the frame is submitted by calling + \l{QRhi::endFrame()}{endFrame()}. To ease this, + QRhiResource::deleteLater() is provided as a convenience. \endlist - \section3 Command buffers and deferred command execution + \section2 Command buffers and deferred command execution Regardless of the design and capabilities of the underlying graphics API, all QRhi backends implement some level of command buffers. No @@ -215,7 +235,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") As a general rule, all referenced QRhiResource objects must stay valid and unmodified until the frame is submitted by calling - \l{QRhiCommandBuffer::endFrame()}{endFrame()}. On the other hand, calling + \l{QRhi::endFrame()}{endFrame()}. On the other hand, calling \l{QRhiResource::destroy()}{destroy()} or deleting the QRhiResource are always safe once the frame is submitted, regardless of the status of the underlying native resources (which may still be in use by the GPU - but @@ -223,10 +243,20 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") Unlike APIs like OpenGL, upload and copy type of commands cannot be mixed with draw commands. The typical renderer will involve a sequence similar to - the following: \c{(re)create resources} - \c{begin frame} - \c{record - uploads and copies} - \c{start renderpass} - \c{record draw calls} - \c{end - renderpass} - \c{end frame}. Recording copy type of operations happens via - QRhiResourceUpdateBatch. Such operations are committed typically on + the following: + + \list + \li (re)create resources + \li begin frame + \li record/issue uploads and copies + \li start recording a render pass + \li record draw calls + \li end render pass + \li end frame + \endlist + + Recording copy type of operations happens via QRhiResourceUpdateBatch. Such + operations are committed typically on \l{QRhiCommandBuffer::beginPass()}{beginPass()}. When working with legacy rendering engines designed for OpenGL, the @@ -245,7 +275,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") remain the primary way of operating since this is what fits Qt's various UI technologies best. - \section3 Threading + \section2 Threading A QRhi instance and the associated resources can be created and used on any thread but all usage must be limited to that one single thread. When @@ -274,7 +304,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") (gui) thread, but becomes important when a separate, dedicated render thread is used. - \section3 Resource synchronization + \section2 Resource synchronization QRhi does not expose APIs for resource barriers or image layout transitions. Such synchronization is done implicitly by the backends, where @@ -294,7 +324,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") different access (one for load, one for store) is supported even within the same pass. - \section3 Resource reuse + \section2 Resource reuse From the user's point of view a QRhiResource is reusable immediately after calling QRhiResource::destroy(). With the exception of swapchains, calling @@ -314,26 +344,39 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") though there is a good chance that under the hood the QRhiBuffer is now backed by a whole new native buffer. - \badcode - ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 256); + \code + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 256); ubuf->create(); - srb = rhi->newShaderResourceBindings() + QRhiShaderResourceBindings *srb = rhi->newShaderResourceBindings() srb->setBindings({ QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage | QRhiShaderResourceBinding::FragmentStage, ubuf) }); srb->create(); - ... + // ... // now in a later frame we need to grow the buffer to a larger size ubuf->setSize(512); ubuf->create(); // same as ubuf->destroy(); ubuf->create(); - // that's it, srb needs no changes whatsoever + // srb needs no changes whatsoever, any references in it to ubuf + // stay valid. When it comes to internal details, such as that + // ubuf may now be backed by a completely different native buffer + // resource, that is is recognized and handled automatically by the + // next setShaderResources(). \endcode - \section3 Pooled objects + QRhiTextureRenderTarget offers the same contract: calling + QRhiCommandBuffer::beginPass() is safe even when one of the render target's + associated textures or renderbuffers has been rebuilt (by calling \c + create() on it) since the creation of the render target object. This allows + the application to resize a texture by setting a new pixel size on the + QRhiTexture and calling create(), thus creating a whole new native texture + resource underneath, without having to update the QRhiTextureRenderTarget + as that will be done implicitly in beginPass(). + + \section2 Pooled objects In addition to resources, there are pooled objects as well, such as, QRhiResourceUpdateBatch. An instance is retrieved via a \c next function, @@ -343,24 +386,25 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") QRhiCommandBuffer::beginPass() or QRhiCommandBuffer::endPass(). These functions take care of returning the batch to the pool. Alternatively, a batch can be "canceled" and returned to the pool without processing by - calling QRhiResourceUpdateBatch::destroy(). + calling QRhiResourceUpdateBatch::release(). A typical pattern is thus: - \badcode + \code QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); - ... + // ... resUpdates->updateDynamicBuffer(ubuf, 0, 64, mvp.constData()); if (!image.isNull()) { resUpdates->uploadTexture(texture, image); image = QImage(); } - ... + // ... QRhiCommandBuffer *cb = m_sc->currentFrameCommandBuffer(); + // note the last argument cb->beginPass(swapchain->currentFrameRenderTarget(), clearCol, clearDs, resUpdates); \endcode - \section3 Swapchain specifics + \section2 Swapchain specifics QRhiSwapChain features some special semantics due to the peculiar nature of swapchains. @@ -391,30 +435,174 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \endlist - \section3 Ownership + \section2 Ownership The general rule is no ownership transfer. Creating a QRhi with an already existing graphics device does not mean the QRhi takes ownership of the device object. Similarly, ownership is not given away when a device or texture object is "exported" via QRhi::nativeHandles() or - QRhiTexture::nativeHandles(). Most importantly, passing pointers in structs + QRhiTexture::nativeTexture(). Most importantly, passing pointers in structs and via setters does not transfer ownership. - \section2 Troubleshooting + \section1 Troubleshooting and Profiling - Errors are printed to the output via qWarning(). Additional debug messages - can be enabled via the following logging categories. Messages from these - categories are not printed by default unless explicitly enabled via - QRhi::EnableProfiling or the facilities of QLoggingCategory (such as, the - \c QT_LOGGING_RULES environment variable). + \section2 Error reporting + + Functions such as \l QRhi::create() and the resource classes' \c create() + member functions (e.g., \l QRhiBuffer::create()) indicate failure with the + return value (\nullptr or + \c false, respectively). When working with QShader, \l QShader::fromSerialized() + returns an invalid QShader (for which \l{QShader::isValid()}{isValid()} returns + \c false) when the data passed to the function cannot be successfully deserialized. + Some functions, beginFrame() in particular, may also sometimes report "soft failures", + such as \l FrameOpSwapChainOutOfDate, which do not indicate an unrecoverable error, + but rather should be seen as a "try again later" response. + + Warnings and errors may get printed at any time to the debug output via + qWarning(). It is therefore always advisable to inspect the output of the + application. + + Additional debug messages can be enabled via the following logging + categories. Messages from these categories are not printed by default + unless explicitly enabled via QLoggingCategory or the \c QT_LOGGING_RULES + environment variable. For better interoperation with Qt Quick, the + environment variable \c{QSG_INFO} also enables these debug prints. \list \li \c{qt.rhi.general} \endlist - It is strongly advised to inspect the output with the logging categories - (\c{qt.rhi.*}) enabled whenever a QRhi-based application is not behaving as - expected. + Additionally, applications can query the \l{QRhi::backendName()}{QRhi + backend name} and + \l{QRhi::driverInfo()}{graphics device information} from a successfully + initialized QRhi. This can then be printed to the user or stored in the + application logs even in production builds, if desired. + + \section2 Investigating rendering problems + + When the rendering results are not as expected, or the application is + experiencing problems, always consider checking with the the native 3D + APIs' debug and validation facilities. QRhi itself features limited error + checking since replicating the already existing, vast amount of + functionality in the underlying layers is not reasonable. + + \list + + \li For Vulkan, controlling the + \l{https://github.com/KhronosGroup/Vulkan-ValidationLayers}{Vulkan + Validation Layers} is not in the scope of the QRhi, but rather can be + achieved by configuring the \l QVulkanInstance with the appropriate layers. + For example, call \c{instance.setLayers({ "VK_LAYER_KHRONOS_validation" });} + before invoking \l{QVulkanInstance::create()}{create()} on the QVulkanInstance. + (note that this assumes that the validation layers are actually installed + and available, e.g. from the Vulkan SDK) By default, QVulkanInstance conveniently + redirects the Vulkan debug messages to qDebug, meaning the validation messages get + printed just like other Qt warnings. + + \li With Direct 3D 11 and 12, a graphics device with the debug layer + enabled can be requested by toggling the \c enableDebugLayer flag in the + appropriate \l{QRhiD3D11InitParams}{init params struct}. The messages appear on the + debug output, which is visible in Qt Creator's messages panel or via a tool + such as \l{https://learn.microsoft.com/en-us/sysinternals/downloads/debugview}{DebugView}. + + \li For Metal, controlling Metal Validation is outside of QRhi's scope. + Rather, to enable validation, run the application with the environment + variable \c{METAL_DEVICE_WRAPPER_TYPE=1} set, or run the application within + XCode. There may also be further settings and environment variable in modern + XCode and macOS versions. See for instance + \l{https://developer.apple.com/documentation/metal/diagnosing_metal_programming_issues_early}{this + page}. + + \endlist + + \section2 Frame captures and performance profiling + + A Qt application rendering with QRhi to a window while relying on a 3D API + under the hood, is, from the windowing and graphics pipeline perspective at + least, no different from any other (non-Qt) applications using the same 3D + API. This means that tools and practices for debugging and profiling + applications involving 3D graphics, such as games, all apply to such a Qt + application as well. + + A few examples of tools that can provide insights into the rendering + internals of Qt applications that use QRhi, which includes Qt Quick and Qt + Quick 3D based projects as well: + + \list + + \li \l{https://renderdoc.org/}{RenderDoc} allows taking frame captures and + introspecting the recorded commands and pipeline state on Windows and Linux + for applications using OpenGL, Vulkan, D3D11, or D3D12. When trying to + figure out why some parts of the 3D scene do not show up as expected, + RenderDoc is often a fast and efficient way to check the pipeline stages + and the related state and discover the missing or incorrect value. It is + also a tool that is actively used when developing Qt itself. + + \li For NVIDIA-based systems, + \l{https://developer.nvidia.com/nsight-graphics}{Nsight Graphics} provides + a graphics debugger tool on Windows and Linux. In addition to investigating the commands + in the frame and the pipeline, the vendor-specific tools allow looking at timings and + hardware performance information, which is not something simple frame captures can provide. + + \li For AMD-based systems, the \l{https://gpuopen.com/rgp/}{Radeon GPU + Profiler} can be used to gain deeper insights into the application's + rendering and its performance. + + \li As QRhi supports Direct 3D 12, using + \l{https://devblogs.microsoft.com/pix/download/}{PIX}, a performance tuning + and debugging tool for DirectX 12 games on Windows is an option as well. + + \li On macOS, + \l{https://developer.apple.com/documentation/metal/debugging_tools/viewing_your_gpu_workload_with_the_metal_debugger}{the + XCode Metal debugger} can be used to take and introspect frame + captures, to investigate performance details, and debug shaders. In macOS 13 it is also possible + to enable an overlay that displays frame rate and other information for any Metal-based window by + setting the environment variable \c{MTL_HUD_ENABLED=1}. + + \endlist + + On mobile and embedded platforms, there may be vendor and platform-specific + tools, provided by the GPU or SoC vendor, available to perform performance + profiling of application using OpenGL ES or Vulkan. + + When capturing frames, remember that objects and groups of commands can be + named via debug markers, as long as \l{QRhi::EnableDebugMarkers}{debug + markers were enabled} for the QRhi, and the graphics API in use supports + this. To annotate the command stream, call + \l{QRhiCommandBuffer::debugMarkBegin()}{debugMarkBegin()}, + \l{QRhiCommandBuffer::debugMarkEnd()}{debugMarkEnd()} and/or + \l{QRhiCommandBuffer::debugMarkMsg()}{debugMarkMsg()}. + This can be particularly useful in larger frames with multiple render passes. + Resources are named by calling \l{QRhiResource::setName()}{setName()} before create(). + + To perform basic timing measurements on the CPU and GPU side within the + application, \l QElapsedTimer and + \l QRhiCommandBuffer::lastCompletedGpuTime() can be used. The latter is + only available with select graphics APIs at the moment and requires opting + in via the \l QRhi::EnableTimestamps flag. + + \section2 Resource leak checking + + When destroying a QRhi object without properly destroying all buffers, + textures, and other resources created from it, warnings about this are + printed to the debug output whenever the application is a debug build, or + when the \c QT_RHI_LEAK_CHECK environment variable is set to a non-zero + value. This is a simple way to discover design issues around resource + handling within the application rendering logic. Note however that some + platforms and underlying graphics APIs may perform their own allocation and + resource leak detection as well, over which Qt will have no direct control. + For example, when using Vulkan, the memory allocator may raise failing + assertions in debug builds when resources that own graphics memory + allocations are not destroyed before the QRhi. In addition, the Vulkan + validation layer, when enabled, will issue warnings about native graphics + resources that were not released. Similarly, with Direct 3D warnings may + get printed about unreleased COM objects when the application does not + destroy the QRhi and its resources in the correct order. + + \sa {RHI Window Example}, QRhiCommandBuffer, QRhiResourceUpdateBatch, + QRhiShaderResourceBindings, QShader, QRhiBuffer, QRhiTexture, + QRhiRenderBuffer, QRhiSampler, QRhiTextureRenderTarget, + QRhiGraphicsPipeline, QRhiComputePipeline, QRhiSwapChain */ /*! @@ -425,6 +613,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value Vulkan \value OpenGLES2 \value D3D11 + \value D3D12 \value Metal */ @@ -432,17 +621,18 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \enum QRhi::Flag Describes what special features to enable. - \value EnableProfiling Enables gathering timing (CPU, GPU) and resource - (QRhiBuffer, QRhiTexture, etc.) information and additional metadata. See - QRhiProfiler. Avoid enabling in production builds as it may involve a - performance penalty. Also enables debug messages from the \c{qt.rhi.*} - logging categories. - \value EnableDebugMarkers Enables debug marker groups. Without this frame debugging features like making debug groups and custom resource name visible in external GPU debugging tools will not be available and functions - like QRhiCommandBuffer::debugMarkBegin() will become a no-op. Avoid - enabling in production builds as it may involve a performance penalty. + like QRhiCommandBuffer::debugMarkBegin() will become no-ops. Avoid enabling + in production builds as it may involve a small performance impact. Has no + effect when the QRhi::DebugMarkers feature is not reported as supported. + + \value EnableTimestamps Enables GPU timestamp collection. When not set, + QRhiCommandBuffer::lastCompletedGpuTime() always returns 0. Enable this + only when needed since there may be a small amount of extra work involved + (e.g. timestamp queries), depending on the underlying graphics API. Has no + effect when the QRhi::Timestamps feature is not reported as supported. \value PreferSoftwareRenderer Indicates that backends should prefer choosing an adapter or physical device that renders in software on the CPU. @@ -458,16 +648,35 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value EnablePipelineCacheDataSave Enables retrieving the pipeline cache contents, where applicable. When not set, pipelineCacheData() will return - an empty blob always. Opting in is relevant in particular with backends - where additional, potentially time consuming work is needed to maintain the - data structures with the serialized, binary versions of shader programs. An - example is OpenGL, where the "pipeline cache" is simulated by retrieving - and loading shader program binaries. With backends where retrieving and - restoring the pipeline cache contents is not supported, the flag has no - effect. With some backends (such as, OpenGL) there are additional, - disk-based caching mechanisms for shader binaries. Writing to those may get - disabled whenever this flag is set since storing program binaries (OpenGL) - to multiple caches is not sensible. + an empty blob always. With backends where retrieving and restoring the + pipeline cache contents is not supported, the flag has no effect and the + serialized cache data is always empty. The flag provides an opt-in + mechanism because the cost of maintaining the related data structures is + not insignificant with some backends. With Vulkan this feature maps + directly to VkPipelineCache, vkGetPipelineCacheData and + VkPipelineCacheCreateInfo::pInitialData. With Direct3D 11 there is no real + pipline cache, but the results of HLSL->DXBC compilations are stored and + can be serialized/deserialized via this mechanism. This allows skipping the + time consuming D3DCompile() in future runs of the applications for shaders + that come with HLSL source instead of offline pre-compiled bytecode. This + can provide a huge boost in startup and load times, if there is a lot of + HLSL source compilation happening. With OpenGL the "pipeline cache" is + simulated by retrieving and loading shader program binaries (if supported + by the driver). With OpenGL there are additional, disk-based caching + mechanisms for shader/program binaries provided by Qt. Writing to those may + get disabled whenever this flag is set since storing program binaries to + multiple caches is not sensible. + + \value SuppressSmokeTestWarnings Indicates that, with backends where this + is relevant, certain, non-fatal QRhi::create() failures should not + produce qWarning() calls. For example, with D3D11, passing this flag + makes a number of warning messages (that appear due to QRhi::create() + failing) to become categorized debug prints instead under the commonly used + \c{qt.rhi.general} logging category. This can be used by engines, such as + Qt Quick, that feature fallback logic, i.e. they retry calling create() + with a different set of flags (such as, \l PreferSoftwareRenderer), in order + to hide the unconditional warnings from the output that would be printed + when the first create() attempt had failed. */ /*! @@ -493,21 +702,34 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") Flag values to indicate what features are supported by the backend currently in use. \value MultisampleTexture Indicates that textures with a sample count larger - than 1 are supported. + than 1 are supported. In practice this feature will be unsupported with + OpenGL ES versions older than 3.1, and OpenGL older than 3.0. \value MultisampleRenderBuffer Indicates that renderbuffers with a sample - count larger than 1 are supported. + count larger than 1 are supported. In practice this feature will be + unsupported with OpenGL ES 2.0, and may also be unsupported with OpenGL 2.x + unless the relevant extensions are present. \value DebugMarkers Indicates that debug marker groups (and so QRhiCommandBuffer::debugMarkBegin()) are supported. \value Timestamps Indicates that command buffer timestamps are supported. - Relevant for QRhiProfiler::gpuFrameTimes(). - - \value Instancing Indicates that instanced drawing is supported. - - \value CustomInstanceStepRate Indicates that instance step rates other than - 1 are supported. + Relevant for QRhiCommandBuffer::lastCompletedGpuTime(). This can be + expected to be supported on Metal, Vulkan, Direct 3D 11 and 12, and OpenGL + contexts of version 3.3 or newer. However, with some of these APIs support + for timestamp queries is technically optional, and therefore it cannot be + guaranteed that this feature is always supported with every implementation + of them. + + \value Instancing Indicates that instanced drawing is supported. In + practice this feature will be unsupported with OpenGL ES 2.0 and OpenGL + 3.2 or older. + + \value CustomInstanceStepRate Indicates that instance step rates other + than 1 are supported. In practice this feature will always be unsupported + with OpenGL. In addition, running with Vulkan 1.0 without + VK_EXT_vertex_attribute_divisor will also lead to reporting false for this + feature. \value PrimitiveRestart Indicates that restarting the assembly of primitives when encountering an index value of 0xFFFF @@ -535,7 +757,8 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") index buffer offsets (\c{indexOffset + firstIndex * indexComponentSize}) that are not 4 byte aligned are supported. When not supported, attempting to issue a \l{QRhiCommandBuffer::drawIndexed()}{drawIndexed()} with a - non-aligned effective offset may lead to unspecified behavior. + non-aligned effective offset may lead to unspecified behavior. Relevant in + particular for Metal, where this will be reported as unsupported. \value NPOTTextureRepeat Indicates that the \l{QRhiSampler::Repeat}{Repeat} wrap mode and mipmap filtering modes are @@ -546,8 +769,12 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value RedOrAlpha8IsRed Indicates that the \l{QRhiTexture::RED_OR_ALPHA8}{RED_OR_ALPHA8} format maps to a one component 8-bit \c red format. This is the case for all backends except - OpenGL, where \c{GL_ALPHA}, a one component 8-bit \c alpha format, is used - instead. This is relevant for shader code that samples from the texture. + OpenGL when using either OpenGL ES or a non-core profile context. There + \c{GL_ALPHA}, a one component 8-bit \c alpha format, is used + instead. Using the special texture format allows having a single code + path for creating textures, leaving it up to the backend to decide the + actual format, while the feature flag can be used to pick the + appropriate shader variant for sampling the texture. \value ElementIndexUint Indicates that 32-bit unsigned integer elements are supported in the index buffer. In practice this is true everywhere except @@ -556,12 +783,13 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") index buffer. \value Compute Indicates that compute shaders, image load/store, and - storage buffers are supported. + storage buffers are supported. OpenGL older than 4.3 and OpenGL ES older + than 3.1 have no compute support. \value WideLines Indicates that lines with a width other than 1 are supported. When reported as not supported, the line width set on the graphics pipeline state is ignored. This can always be false with some - backends (D3D11, Metal). With Vulkan, the value depends on the + backends (D3D11, D3D12, Metal). With Vulkan, the value depends on the implementation. With OpenGL, wide lines are not supported in core profile contexts. @@ -574,50 +802,56 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") require the point size to be set in the shader explicitly whenever drawing points, even when the size is 1, as they do not automatically default to 1. - \value BaseVertex Indicates that \l{QRhiCommandBuffer::drawIndexed()}{drawIndexed()} - supports the \c vertexOffset argument. When reported as not supported, the - vertexOffset value in an indexed draw is ignored. + \value BaseVertex Indicates that + \l{QRhiCommandBuffer::drawIndexed()}{drawIndexed()} supports the \c + vertexOffset argument. When reported as not supported, the vertexOffset + value in an indexed draw is ignored. In practice this feature will be + unsupported with OpenGL and OpenGL ES versions lower than 3.2, and with + Metal on older iOS devices, including the iOS Simulator. \value BaseInstance Indicates that instanced draw commands support the \c firstInstance argument. When reported as not supported, the firstInstance - value is ignored and the instance ID starts from 0. + value is ignored and the instance ID starts from 0. In practice this feature + will be unsupported with OpenGL, and with Metal on older iOS devices, + including the iOS Simulator. \value TriangleFanTopology Indicates that QRhiGraphicsPipeline::setTopology() - supports QRhiGraphicsPipeline::TriangleFan. + supports QRhiGraphicsPipeline::TriangleFan. In practice this feature will be + unsupported with Metal and Direct 3D 11/12. \value ReadBackNonUniformBuffer Indicates that \l{QRhiResourceUpdateBatch::readBackBuffer()}{reading buffer contents} is supported for QRhiBuffer instances with a usage different than - UniformBuffer. While this is supported in the majority of cases, it will be - unsupported with OpenGL ES older than 3.0. + UniformBuffer. In practice this feature will be unsupported with OpenGL ES + 2.0. \value ReadBackNonBaseMipLevel Indicates that specifying a mip level other than 0 is supported when reading back texture contents. When not supported, specifying a non-zero level in QRhiReadbackDescription leads to returning an all-zero image. In practice this feature will be unsupported with OpenGL - ES 2.0, while it will likely be supported everywhere else. + ES 2.0. - \value TexelFetch Indicates that texelFetch() is available in shaders. In - practice this will be reported as unsupported with OpenGL ES 2.0 and OpenGL - 2.x contexts, because GLSL 100 es and versions before 130 do not support - this function. + \value TexelFetch Indicates that texelFetch() and textureLod() are available + in shaders. In practice this will be reported as unsupported with OpenGL ES + 2.0 and OpenGL 2.x contexts, because GLSL 100 es and versions before 130 do + not support these functions. \value RenderToNonBaseMipLevel Indicates that specifying a mip level other than 0 is supported when creating a QRhiTextureRenderTarget with a QRhiTexture as its color attachment. When not supported, create() will fail whenever the target mip level is not zero. In practice this feature will be - unsupported with OpenGL ES 2.0, while it will likely be supported everywhere - else. + unsupported with OpenGL ES 2.0. \value IntAttributes Indicates that specifying input attributes with signed and unsigned integer types for a shader pipeline is supported. When not supported, build() will succeed but just show a warning message and the values of the target attributes will be broken. In practice this feature - will be unsupported with OpenGL ES 2.0 and OpenGL 2.x, while it will likely - be supported everywhere else. + will be unsupported with OpenGL ES 2.0 and OpenGL 2.x. \value ScreenSpaceDerivatives Indicates that functions such as dFdx(), - dFdy(), and fwidth() are supported in shaders. + dFdy(), and fwidth() are supported in shaders. In practice this feature will + be unsupported with OpenGL ES 2.0 without the GL_OES_standard_derivatives + extension. \value ReadBackAnyTextureFormat Indicates that reading back texture contents can be expected to work for any QRhiTexture::Format. Backends @@ -660,6 +894,148 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") texture is supported. This can be unsupported with Vulkan 1.0 due to relying on VK_IMAGE_CREATE_2D_ARRAY_COMPATIBLE_BIT which is a Vulkan 1.1 feature. + + \value TextureArrays Indicates that texture arrays are supported and + QRhi::newTextureArray() is functional. Note that even when texture arrays + are not supported, arrays of textures are still available as those are two + independent features. + + \value Tessellation Indicates that the tessellation control and evaluation + stages are supported. When reported as supported, the topology of a + QRhiGraphicsPipeline can be set to + \l{QRhiGraphicsPipeline::Patches}{Patches}, the number of control points + can be set via + \l{QRhiGraphicsPipeline::setPatchControlPointCount()}{setPatchControlPointCount()}, + and shaders for tessellation control and evaluation can be specified in the + QRhiShaderStage list. Tessellation shaders have portability issues between + APIs (for example, translating GLSL/SPIR-V to HLSL is problematic due to + the way hull shaders are structured, whereas Metal uses a somewhat + different tessellation pipeline than others), and therefore unexpected + issues may still arise, even though basic functionality is implemented + across all the underlying APIs. For Direct 3D in particular, handwritten + HLSL hull and domain shaders must be injected into each QShader for the + tessellation control and evaluation stages, respectively, since qsb cannot + generate these from SPIR-V. Note that isoline tessellation should be + avoided as it will not be supported by all backends. The maximum patch + control point count portable between backends is 32. + + \value GeometryShader Indicates that the geometry shader stage is + supported. When supported, a geometry shader can be specified in the + QRhiShaderStage list. Geometry Shaders are considered an experimental + feature in QRhi and can only be expected to be supported with Vulkan, + Direct 3D, OpenGL (3.2+) and OpenGL ES (3.2+), assuming the implementation + reports it as supported at run time. Geometry shaders have portability + issues between APIs, and therefore no guarantees can be given for a + universal solution. They will never be supported with Metal. Whereas with + Direct 3D a handwritten HLSL geometry shader must be injected into each + QShader for the geometry stage since qsb cannot generate this from SPIR-V. + + \value TextureArrayRange Indicates that for + \l{QRhi::newTextureArray()}{texture arrays} it is possible to specify a + range that is exposed to the shaders. Normally all array layers are exposed + and it is up to the shader to select the layer (via the third coordinate + passed to texture() when sampling the \c sampler2DArray). When supported, + calling QRhiTexture::setArrayRangeStart() and + QRhiTexture::setArrayRangeLength() before + \l{QRhiTexture::create()}{building} or + \l{QRhiTexture::createFrom()}{importing} the native texture has an effect, + and leads to selecting only the specified range from the array. This will + be necessary in special cases, such as when working with accelerated video + decoding and Direct 3D 11, because a texture array with both + \c{D3D11_BIND_DECODER} and \c{D3D11_BIND_SHADER_RESOURCE} on it is only + usable as a shader resource if a single array layer is selected. Note that + all this is applicable only when the texture is used as a + QRhiShaderResourceBinding::SampledTexture or + QRhiShaderResourceBinding::Texture shader resource, and is not compatible + with image load/store. This feature is only available with some backends as + it does not map well to all graphics APIs, and it is only meant to provide + support for special cases anyhow. In practice the feature can be expected to + be supported with Direct3D 11/12 and Vulkan. + + \value NonFillPolygonMode Indicates that setting a PolygonMode other than + the default Fill is supported for QRhiGraphicsPipeline. A common use case + for changing the mode to Line is to get wireframe rendering. This however + is not available as a core OpenGL ES feature, and is optional with Vulkan + as well as some mobile GPUs may not offer the feature. + + \value OneDimensionalTextures Indicates that 1D textures are supported. + In practice this feature will be unsupported on OpenGL ES. + + \value OneDimensionalTextureMipmaps Indicates that generating 1D texture + mipmaps are supported. In practice this feature will be unsupported on + backends that do not report support for + \l{OneDimensionalTextures}, Metal, and Direct 3D 12. + + \value HalfAttributes Indicates that specifying input attributes with half + precision (16bit) floating point types for a shader pipeline is supported. + When not supported, build() will succeed but just show a warning message + and the values of the target attributes will be broken. In practice this + feature will be unsupported in some OpenGL ES 2.0 and OpenGL 2.x + implementations. Note that while Direct3D 11/12 does support half precision + input attributes, it does not support the half3 type. The D3D backends pass + half3 attributes as half4. To ensure cross platform compatibility, half3 + inputs should be padded to 8 bytes. + + \value RenderToOneDimensionalTexture Indicates that 1D texture render + targets are supported. In practice this feature will be unsupported on + backends that do not report support for + \l{OneDimensionalTextures}, and Metal. + + \value ThreeDimensionalTextureMipmaps Indicates that generating 3D texture + mipmaps are supported. In practice this feature will be unsupported with + Direct 3D 12. + + \value MultiView Indicates that multiview, see e.g. + \l{https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_multiview.html}{VK_KHR_multiview} + is supported. With OpenGL ES 2.0, Direct 3D 11, and OpenGL (ES) + implementations without \c{GL_OVR_multiview2} this feature will not be + supported. With Vulkan 1.1 and newer, and Direct 3D 12 multiview is + typically supported. When reported as supported, creating a + QRhiTextureRenderTarget with a QRhiColorAttachment that references a texture + array and has \l{QRhiColorAttachment::setMultiViewCount()}{multiViewCount} + set enables recording a render pass that uses multiview rendering. In addition, + any QRhiGraphicsPipeline used in that render pass must have + \l{QRhiGraphicsPipeline::setMultiViewCount()}{the same view count set}. Note that + multiview is only available in combination with 2D texture arrays. It cannot + be used to optimize the rendering into individual textures (e.g. two, for + the left and right eyes). Rather, the target of a multiview render pass is + always a texture array, automatically rendering to the layer (array element) + corresponding to each view. Therefore this feature implies \l TextureArrays + as well. Multiview rendering is not supported in combination with + tessellation or geometry shaders. See QRhiColorAttachment::setMultiViewCount() + for further details on multiview rendering. This enum value has been introduced in Qt 6.7. + + \value TextureViewFormat Indicates that setting a + \l{QRhiTexture::setWriteViewFormat()}{view format} on a QRhiTexture is + effective. When reported as supported, setting the read (sampling) or write + (render target / image load-store) view mode changes the texture's viewing + format. When unsupported, setting a view format has no effect. Note that Qt + has no knowledge or control over format compatibility or resource view rules + in the underlying 3D API and its implementation. Passing in unsuitable, + incompatible formats may lead to errors and unspecified behavior. This is + provided mainly to allow "casting" rendering into a texture created with an + sRGB format to non-sRGB to avoid the unwanted linear->sRGB conversion on + shader writes. Other types of casting may or may not be functional, + depending on the underlying API. Currently implemented for Vulkan and Direct + 3D 12. With D3D12 the feature is available only if + \c CastingFullyTypedFormatSupported is supported, see + \l{https://microsoft.github.io/DirectX-Specs/d3d/RelaxedCasting.html} (and + note that QRhi always uses fully typed formats for textures.) This enum + value has been introduced in Qt 6.8. + + \value ResolveDepthStencil Indicates that resolving a multisample depth or + depth-stencil texture is supported. Otherwise, + \l{QRhiTextureRenderTargetDescription::setDepthResolveTexture()}{setting a + depth resolve texture} is not functional and must be avoided. Direct 3D 11 + and 12 have no support for resolving depth/depth-stencil formats, and + therefore this feature will never be supported with those. Vulkan 1.0 has no + API to request resolving a depth-stencil attachment. Therefore, with Vulkan + this feature will only be supported with Vulkan 1.2 and up, and on 1.1 + implementations with the appropriate extensions present. This feature is + provided for the rare case when resolving into a non-multisample depth + texture becomes necessary, for example when rendering into an + OpenXR-provided depth texture (XR_KHR_composition_layer_depth). This enum + value has been introduced in Qt 6.8. */ /*! @@ -747,26 +1123,52 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value MaxThreadGroupZ The maximum size of a work/thread group in the Z dimension. Effectively the maximum value of \c local_size_z in the compute shader. Typically 64 or 256. + + \value TextureArraySizeMax Maximum texture array size. Typically in range + 256 - 2048. Attempting to \l{QRhi::newTextureArray()}{create a texture + array} with more elements will likely fail. + + \value MaxUniformBufferRange The number of bytes that can be exposed from a + uniform buffer to the shaders at once. On OpenGL ES 2.0 and 3.0 + implementations this may be as low as 3584 bytes (224 four component, 32 + bits per component vectors). Elsewhere the value is typically 16384 (1024 + vec4s) or 65536 (4096 vec4s). + + \value MaxVertexInputs The number of input attributes to the vertex shader. + The location in a QRhiVertexInputAttribute must be in range \c{[0, + MaxVertexInputs-1]}. The value may be as low as 8 with OpenGL ES 2.0. + Elsewhere, typical values are 16, 31, or 32. + + \value MaxVertexOutputs The maximum number of outputs (4 component vector + \c out variables) from the vertex shader. The value may be as low as 8 with + OpenGL ES 2.0, and 15 with OpenGL ES 3.0 and some Metal devices. Elsewhere, + a typical value is 32. */ /*! \class QRhiInitParams - \internal \inmodule QtGui + \since 6.6 \brief Base class for backend-specific initialization parameters. Contains fields that are relevant to all backends. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! \class QRhiDepthStencilClearValue - \internal \inmodule QtGui + \since 6.6 \brief Specifies clear values for a depth or stencil buffer. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiDepthStencilClearValue::QRhiDepthStencilClearValue() + \fn QRhiDepthStencilClearValue::QRhiDepthStencilClearValue() = default Constructs a depth/stencil clear value with depth clear value 1.0f and stencil clear value 0. @@ -783,37 +1185,45 @@ QRhiDepthStencilClearValue::QRhiDepthStencilClearValue(float d, quint32 s) } /*! + \fn float QRhiDepthStencilClearValue::depthClearValue() const + \return the depth clear value. In most cases this is 1.0f. + */ + +/*! + \fn void QRhiDepthStencilClearValue::setDepthClearValue(float d) + Sets the depth clear value to \a d. + */ + +/*! + \fn quint32 QRhiDepthStencilClearValue::stencilClearValue() const + \return the stencil clear value. In most cases this is 0. + */ + +/*! + \fn void QRhiDepthStencilClearValue::setStencilClearValue(quint32 s) + Sets the stencil clear value to \a s. + */ + +/*! + \fn bool QRhiDepthStencilClearValue::operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + \return \c true if the values in the two QRhiDepthStencilClearValue objects \a a and \a b are equal. - - \relates QRhiDepthStencilClearValue */ -bool operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept -{ - return a.depthClearValue() == b.depthClearValue() - && a.stencilClearValue() == b.stencilClearValue(); -} /*! + \fn bool QRhiDepthStencilClearValue::operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + \return \c false if the values in the two QRhiDepthStencilClearValue objects \a a and \a b are equal; otherwise returns \c true. - \relates QRhiDepthStencilClearValue */ -bool operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiDepthStencilClearValue::qHash(const QRhiDepthStencilClearValue &v, size_t seed = 0) noexcept - \relates QRhiDepthStencilClearValue + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiDepthStencilClearValue &v, size_t seed) noexcept -{ - return seed * (uint(qFloor(qreal(v.depthClearValue()) * 100)) + v.stencilClearValue()); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiDepthStencilClearValue &v) @@ -828,8 +1238,8 @@ QDebug operator<<(QDebug dbg, const QRhiDepthStencilClearValue &v) /*! \class QRhiViewport - \internal \inmodule QtGui + \since 6.6 \brief Specifies a viewport rectangle. Used with QRhiCommandBuffer::setViewport(). @@ -839,20 +1249,23 @@ QDebug operator<<(QDebug dbg, const QRhiDepthStencilClearValue &v) Typical usage is like the following: - \badcode + \code const QSize outputSizeInPixels = swapchain->currentPixelSize(); const QRhiViewport viewport(0, 0, outputSizeInPixels.width(), outputSizeInPixels.height()); - cb->beginPass(swapchain->currentFrameRenderTarget(), { 0, 0, 0, 1 }, { 1, 0 }); + cb->beginPass(swapchain->currentFrameRenderTarget(), Qt::black, { 1.0f, 0 }); cb->setGraphicsPipeline(ps); cb->setViewport(viewport); - ... + // ... \endcode + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setViewport(), QRhi::clipSpaceCorrMatrix(), QRhiScissor */ /*! - \fn QRhiViewport::QRhiViewport() + \fn QRhiViewport::QRhiViewport() = default Constructs a viewport description with an empty rectangle and a depth range of 0.0f - 1.0f. @@ -878,40 +1291,59 @@ QRhiViewport::QRhiViewport(float x, float y, float w, float h, float minDepth, f } /*! + \fn std::array<float, 4> QRhiViewport::viewport() const + \return the viewport x, y, width, and height. + */ + +/*! + \fn void QRhiViewport::setViewport(float x, float y, float w, float h) + Sets the viewport's position and size to \a x, \a y, \a w, and \a h. + + \note Viewports are specified in a coordinate system that has its origin in + the bottom-left. + */ + +/*! + \fn float QRhiViewport::minDepth() const + \return the minDepth value of the depth range of the viewport. + */ + +/*! + \fn void QRhiViewport::setMinDepth(float minDepth) + Sets the \a minDepth of the depth range of the viewport. + By default this is set to 0.0f. + */ + +/*! + \fn float QRhiViewport::maxDepth() const + \return the maxDepth value of the depth range of the viewport. + */ + +/*! + \fn void QRhiViewport::setMaxDepth(float maxDepth) + Sets the \a maxDepth of the depth range of the viewport. + By default this is set to 1.0f. + */ + +/*! + \fn bool QRhiViewport::operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept + \return \c true if the values in the two QRhiViewport objects \a a and \a b are equal. - - \relates QRhiViewport */ -bool operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept -{ - return a.viewport() == b.viewport() - && a.minDepth() == b.minDepth() - && a.maxDepth() == b.maxDepth(); -} /*! + \fn bool QRhiViewport::operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept + \return \c false if the values in the two QRhiViewport objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiViewport */ -bool operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiViewport::qHash(const QRhiViewport &v, size_t seed = 0) noexcept - \relates QRhiViewport + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiViewport &v, size_t seed) noexcept -{ - const std::array<float, 4> r = v.viewport(); - return seed + uint(r[0]) + uint(r[1]) + uint(r[2]) + uint(r[3]) - + uint(qFloor(qreal(v.minDepth()) * 100)) + uint(qFloor(qreal(v.maxDepth()) * 100)); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiViewport &v) @@ -931,8 +1363,8 @@ QDebug operator<<(QDebug dbg, const QRhiViewport &v) /*! \class QRhiScissor - \internal \inmodule QtGui + \since 6.6 \brief Specifies a scissor rectangle. Used with QRhiCommandBuffer::setScissor(). Setting a scissor rectangle is @@ -946,11 +1378,14 @@ QDebug operator<<(QDebug dbg, const QRhiViewport &v) appropriate. Therefore, any rendering logic targeting OpenGL can feed scissor rectangles into QRhiScissor as-is, without any adaptation. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setScissor(), QRhiViewport */ /*! - \fn QRhiScissor::QRhiScissor() + \fn QRhiScissor::QRhiScissor() = default Constructs an empty scissor. */ @@ -971,37 +1406,37 @@ QRhiScissor::QRhiScissor(int x, int y, int w, int h) } /*! + \fn std::array<int, 4> QRhiScissor::scissor() const + \return the scissor position and size. + */ + +/*! + \fn void QRhiScissor::setScissor(int x, int y, int w, int h) + Sets the scissor position and size to \a x, \a y, \a w, \a h. + + \note The position is always expected to be specified in a coordinate + system that has its origin in the bottom-left corner, like OpenGL. + */ + +/*! + \fn bool QRhiScissor::operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept + \return \c true if the values in the two QRhiScissor objects \a a and \a b are equal. - - \relates QRhiScissor */ -bool operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept -{ - return a.scissor() == b.scissor(); -} /*! + \fn bool QRhiScissor::operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept + \return \c false if the values in the two QRhiScissor objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiScissor */ -bool operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiScissor::qHash(const QRhiScissor &v, size_t seed = 0) noexcept - \relates QRhiScissor + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiScissor &v, size_t seed) noexcept -{ - const std::array<int, 4> r = v.scissor(); - return seed + uint(r[0]) + uint(r[1]) + uint(r[2]) + uint(r[3]); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiScissor &s) @@ -1019,8 +1454,8 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) /*! \class QRhiVertexInputBinding - \internal \inmodule QtGui + \since 6.6 \brief Describes a vertex input binding. Specifies the stride (in bytes, must be a multiple of 4), the @@ -1035,10 +1470,10 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) Now let's assume also that 3 component vertex positions \c{(x, y, z)} and 2 component texture coordinates \c{(u, v)} are provided in a non-interleaved - format in a buffer (or separate buffers even). Definining two bindings + format in a buffer (or separate buffers even). Defining two bindings could then be done like this: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 3 * sizeof(float) }, @@ -1055,7 +1490,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) vertices, assuming we have a single buffer with first the positions and then the texture coordinates: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBindings[] = { { cubeBuf, 0 }, { cubeBuf, 36 * 3 * sizeof(float) } @@ -1071,6 +1506,9 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) \note the stride must always be a multiple of 4. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setVertexInput() */ @@ -1083,7 +1521,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) */ /*! - \fn QRhiVertexInputBinding::QRhiVertexInputBinding() + \fn QRhiVertexInputBinding::QRhiVertexInputBinding() = default Constructs a default vertex input binding description. */ @@ -1095,7 +1533,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) \note \a stepRate other than 1 is only supported when QRhi::CustomInstanceStepRate is reported to be supported. */ -QRhiVertexInputBinding::QRhiVertexInputBinding(quint32 stride, Classification cls, int stepRate) +QRhiVertexInputBinding::QRhiVertexInputBinding(quint32 stride, Classification cls, quint32 stepRate) : m_stride(stride), m_classification(cls), m_instanceStepRate(stepRate) @@ -1103,38 +1541,54 @@ QRhiVertexInputBinding::QRhiVertexInputBinding(quint32 stride, Classification cl } /*! + \fn quint32 QRhiVertexInputBinding::stride() const + \return the stride in bytes. + */ + +/*! + \fn void QRhiVertexInputBinding::setStride(quint32 s) + Sets the stride to \a s. + */ + +/*! + \fn QRhiVertexInputBinding::Classification QRhiVertexInputBinding::classification() const + \return the input data classification. + */ + +/*! + \fn void QRhiVertexInputBinding::setClassification(Classification c) + Sets the input data classification \a c. By default this is set to PerVertex. + */ + +/*! + \fn quint32 QRhiVertexInputBinding::instanceStepRate() const + \return the instance step rate. + */ + +/*! + \fn void QRhiVertexInputBinding::setInstanceStepRate(quint32 rate) + Sets the instance step \a rate. By default this is set to 1. + */ + +/*! + \fn bool QRhiVertexInputBinding::operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + \return \c true if the values in the two QRhiVertexInputBinding objects \a a and \a b are equal. - - \relates QRhiVertexInputBinding */ -bool operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept -{ - return a.stride() == b.stride() - && a.classification() == b.classification() - && a.instanceStepRate() == b.instanceStepRate(); -} /*! + \fn bool QRhiVertexInputBinding::operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + \return \c false if the values in the two QRhiVertexInputBinding objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiVertexInputBinding */ -bool operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiVertexInputBinding::qHash(const QRhiVertexInputBinding &v, size_t seed = 0) noexcept - \relates QRhiVertexInputBinding + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiVertexInputBinding &v, size_t seed) noexcept -{ - return seed + v.stride() + v.classification(); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) @@ -1150,14 +1604,15 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) /*! \class QRhiVertexInputAttribute - \internal \inmodule QtGui + \since 6.6 \brief Describes a single vertex input element. The members specify the binding number, location, format, and offset for a single vertex input element. - \note For HLSL it is assumed that the vertex shader uses + \note For HLSL it is assumed that the vertex shader translated from SPIR-V + uses \c{TEXCOORD<location>} as the semantic for each input. Hence no separate semantic name and index. @@ -1173,7 +1628,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) non-interleaved format in a buffer (or separate buffers even). Once two bindings are defined, the attributes could be specified as: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 3 * sizeof(float) }, @@ -1190,7 +1645,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) vertices, assuming we have a single buffer with first the positions and then the texture coordinates: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBindings[] = { { cubeBuf, 0 }, { cubeBuf, 36 * 3 * sizeof(float) } @@ -1202,7 +1657,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) binding, with multiple attributes referring to that same buffer binding point: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 5 * sizeof(float) } @@ -1215,11 +1670,14 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) and then: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBinding(interleavedCubeBuf, 0); cb->setVertexInput(0, 1, &vbufBinding); \endcode + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setVertexInput() */ @@ -1242,10 +1700,30 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) \value SInt3 Three component signed integer vector \value SInt2 Two component signed integer vector \value SInt Signed integer + \value Half4 Four component half precision (16 bit) float vector + \value Half3 Three component half precision (16 bit) float vector + \value Half2 Two component half precision (16 bit) float vector + \value Half Half precision (16 bit) float + \value UShort4 Four component unsigned short (16 bit) integer vector + \value UShort3 Three component unsigned short (16 bit) integer vector + \value UShort2 Two component unsigned short (16 bit) integer vector + \value UShort Unsigned short (16 bit) integer + \value SShort4 Four component signed short (16 bit) integer vector + \value SShort3 Three component signed short (16 bit) integer vector + \value SShort2 Two component signed short (16 bit) integer vector + \value SShort Signed short (16 bit) integer + + \note Support for half precision floating point attributes is indicated at + run time by the QRhi::Feature::HalfAttributes feature flag. + + \note Direct3D 11/12 supports 16 bit input attributes, but does not support + the Half3, UShort3 or SShort3 types. The D3D backends pass through Half3 as + Half4, UShort3 as UShort4, and SShort3 as SShort4. To ensure cross platform + compatibility, 16 bit inputs should be padded to 8 bytes. */ /*! - \fn QRhiVertexInputAttribute::QRhiVertexInputAttribute() + \fn QRhiVertexInputAttribute::QRhiVertexInputAttribute() = default Constructs a default vertex input attribute description. */ @@ -1270,39 +1748,85 @@ QRhiVertexInputAttribute::QRhiVertexInputAttribute(int binding, int location, Fo } /*! + \fn int QRhiVertexInputAttribute::binding() const + \return the binding point index. + */ + +/*! + \fn void QRhiVertexInputAttribute::setBinding(int b) + Sets the binding point index to \a b. + By default this is set to 0. + */ + +/*! + \fn int QRhiVertexInputAttribute::location() const + \return the location of the vertex input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setLocation(int loc) + Sets the location of the vertex input element to \a loc. + By default this is set to 0. + */ + +/*! + \fn QRhiVertexInputAttribute::Format QRhiVertexInputAttribute::format() const + \return the format of the vertex input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setFormat(Format f) + Sets the format of the vertex input element to \a f. + By default this is set to Float4. + */ + +/*! + \fn quint32 QRhiVertexInputAttribute::offset() const + \return the byte offset for the input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setOffset(quint32 ofs) + Sets the byte offset for the input element to \a ofs. + */ + +/*! + \fn int QRhiVertexInputAttribute::matrixSlice() const + + \return the matrix slice if the input element corresponds to a row or + column of a matrix, or -1 if not relevant. + */ + +/*! + \fn void QRhiVertexInputAttribute::setMatrixSlice(int slice) + + Sets the matrix \a slice. By default this is set to -1, and should be set + to a >= 0 value only when this attribute corresponds to a row or column of + a matrix (for example, a 4x4 matrix becomes 4 vec4s, consuming 4 + consecutive vertex input locations), in which case it is the index of the + row or column. \c{location - matrixSlice} must always be equal to the \c + location for the first row or column of the unrolled matrix. + */ + +/*! + \fn bool QRhiVertexInputAttribute::operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + \return \c true if the values in the two QRhiVertexInputAttribute objects \a a and \a b are equal. - - \relates QRhiVertexInputAttribute */ -bool operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept -{ - return a.binding() == b.binding() - && a.location() == b.location() - && a.format() == b.format() - && a.offset() == b.offset(); -} /*! + \fn bool QRhiVertexInputAttribute::operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + \return \c false if the values in the two QRhiVertexInputAttribute objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiVertexInputAttribute */ -bool operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiVertexInputAttribute::qHash(const QRhiVertexInputAttribute &v, size_t seed = 0) noexcept - \relates QRhiVertexInputAttribute + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiVertexInputAttribute &v, size_t seed) noexcept -{ - return seed + uint(v.binding()) + uint(v.location()) + uint(v.format()) + v.offset(); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiVertexInputAttribute &a) @@ -1317,53 +1841,235 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputAttribute &a) } #endif +QRhiVertexInputAttribute::Format QRhiImplementation::shaderDescVariableFormatToVertexInputFormat(QShaderDescription::VariableType type) const +{ + switch (type) { + case QShaderDescription::Vec4: + return QRhiVertexInputAttribute::Float4; + case QShaderDescription::Vec3: + return QRhiVertexInputAttribute::Float3; + case QShaderDescription::Vec2: + return QRhiVertexInputAttribute::Float2; + case QShaderDescription::Float: + return QRhiVertexInputAttribute::Float; + + case QShaderDescription::Int4: + return QRhiVertexInputAttribute::SInt4; + case QShaderDescription::Int3: + return QRhiVertexInputAttribute::SInt3; + case QShaderDescription::Int2: + return QRhiVertexInputAttribute::SInt2; + case QShaderDescription::Int: + return QRhiVertexInputAttribute::SInt; + + case QShaderDescription::Uint4: + return QRhiVertexInputAttribute::UInt4; + case QShaderDescription::Uint3: + return QRhiVertexInputAttribute::UInt3; + case QShaderDescription::Uint2: + return QRhiVertexInputAttribute::UInt2; + case QShaderDescription::Uint: + return QRhiVertexInputAttribute::UInt; + + case QShaderDescription::Half4: + return QRhiVertexInputAttribute::Half4; + case QShaderDescription::Half3: + return QRhiVertexInputAttribute::Half3; + case QShaderDescription::Half2: + return QRhiVertexInputAttribute::Half2; + case QShaderDescription::Half: + return QRhiVertexInputAttribute::Half; + + default: + Q_UNREACHABLE_RETURN(QRhiVertexInputAttribute::Float); + } +} + +quint32 QRhiImplementation::byteSizePerVertexForVertexInputFormat(QRhiVertexInputAttribute::Format format) const +{ + switch (format) { + case QRhiVertexInputAttribute::Float4: + return 4 * sizeof(float); + case QRhiVertexInputAttribute::Float3: + return 4 * sizeof(float); // vec3 still takes 16 bytes + case QRhiVertexInputAttribute::Float2: + return 2 * sizeof(float); + case QRhiVertexInputAttribute::Float: + return sizeof(float); + + case QRhiVertexInputAttribute::UNormByte4: + return 4 * sizeof(quint8); + case QRhiVertexInputAttribute::UNormByte2: + return 2 * sizeof(quint8); + case QRhiVertexInputAttribute::UNormByte: + return sizeof(quint8); + + case QRhiVertexInputAttribute::UInt4: + return 4 * sizeof(quint32); + case QRhiVertexInputAttribute::UInt3: + return 4 * sizeof(quint32); // ivec3 still takes 16 bytes + case QRhiVertexInputAttribute::UInt2: + return 2 * sizeof(quint32); + case QRhiVertexInputAttribute::UInt: + return sizeof(quint32); + + case QRhiVertexInputAttribute::SInt4: + return 4 * sizeof(qint32); + case QRhiVertexInputAttribute::SInt3: + return 4 * sizeof(qint32); // uvec3 still takes 16 bytes + case QRhiVertexInputAttribute::SInt2: + return 2 * sizeof(qint32); + case QRhiVertexInputAttribute::SInt: + return sizeof(qint32); + + case QRhiVertexInputAttribute::Half4: + return 4 * sizeof(qfloat16); + case QRhiVertexInputAttribute::Half3: + return 4 * sizeof(qfloat16); // half3 still takes 8 bytes + case QRhiVertexInputAttribute::Half2: + return 2 * sizeof(qfloat16); + case QRhiVertexInputAttribute::Half: + return sizeof(qfloat16); + + case QRhiVertexInputAttribute::UShort4: + return 4 * sizeof(quint16); + case QRhiVertexInputAttribute::UShort3: + return 4 * sizeof(quint16); // ivec3 still takes 8 bytes + case QRhiVertexInputAttribute::UShort2: + return 2 * sizeof(quint16); + case QRhiVertexInputAttribute::UShort: + return sizeof(quint16); + + case QRhiVertexInputAttribute::SShort4: + return 4 * sizeof(qint16); + case QRhiVertexInputAttribute::SShort3: + return 4 * sizeof(qint16); // uvec3 still takes 8 bytes + case QRhiVertexInputAttribute::SShort2: + return 2 * sizeof(qint16); + case QRhiVertexInputAttribute::SShort: + return sizeof(qint16); + + default: + Q_UNREACHABLE_RETURN(1); + } +} + /*! \class QRhiVertexInputLayout - \internal \inmodule QtGui + \since 6.6 \brief Describes the layout of vertex inputs consumed by a vertex shader. The vertex input layout is defined by the collections of QRhiVertexInputBinding and QRhiVertexInputAttribute. + + As an example, let's assume that we have a single buffer with 3 component + vertex positions and 2 component UV coordinates interleaved (\c x, \c y, \c + z, \c u, \c v), that the position and UV are expected at input locations 0 + and 1 by the vertex shader, and that the vertex buffer will be bound at + binding point 0 using + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()} later on: + + \code + QRhiVertexInputLayout inputLayout; + inputLayout.setBindings({ + { 5 * sizeof(float) } + }); + inputLayout.setAttributes({ + { 0, 0, QRhiVertexInputAttribute::Float3, 0 }, + { 0, 1, QRhiVertexInputAttribute::Float2, 3 * sizeof(float) } + }); + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiVertexInputLayout::QRhiVertexInputLayout() + \fn QRhiVertexInputLayout::QRhiVertexInputLayout() = default Constructs an empty vertex input layout description. */ /*! + \fn void QRhiVertexInputLayout::setBindings(std::initializer_list<QRhiVertexInputBinding> list) + Sets the bindings from the specified \a list. + */ + +/*! + \fn template<typename InputIterator> void QRhiVertexInputLayout::setBindings(InputIterator first, InputIterator last) + Sets the bindings using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::cbeginBindings() const + \return a const iterator pointing to the first item in the binding list. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::cendBindings() const + \return a const iterator pointing just after the last item in the binding list. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::bindingAt(qsizetype index) const + \return the binding at the given \a index. + */ + +/*! + \fn qsizetype QRhiVertexInputLayout::bindingCount() const + \return the number of bindings. + */ + +/*! + \fn void QRhiVertexInputLayout::setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) + Sets the attributes from the specified \a list. + */ + +/*! + \fn template<typename InputIterator> void QRhiVertexInputLayout::setAttributes(InputIterator first, InputIterator last) + Sets the attributes using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::cbeginAttributes() const + \return a const iterator pointing to the first item in the attribute list. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::cendAttributes() const + \return a const iterator pointing just after the last item in the attribute list. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::attributeAt(qsizetype index) const + \return the attribute at the given \a index. + */ + +/*! + \fn qsizetype QRhiVertexInputLayout::attributeCount() const + \return the number of attributes. + */ + +/*! + \fn bool QRhiVertexInputLayout::operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + \return \c true if the values in the two QRhiVertexInputLayout objects \a a and \a b are equal. - - \relates QRhiVertexInputLayout */ -bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept -{ - return a.m_bindings == b.m_bindings && a.m_attributes == b.m_attributes; -} /*! + \fn bool QRhiVertexInputLayout::operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + \return \c false if the values in the two QRhiVertexInputLayout objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiVertexInputLayout */ -bool operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiVertexInputLayout::qHash(const QRhiVertexInputLayout &v, size_t seed = 0) noexcept - \relates QRhiVertexInputLayout + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiVertexInputLayout &v, size_t seed) noexcept -{ - return qHash(v.m_bindings, seed) + qHash(v.m_attributes, seed); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiVertexInputLayout &v) @@ -1378,9 +2084,39 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputLayout &v) /*! \class QRhiShaderStage - \internal \inmodule QtGui + \since 6.6 \brief Specifies the type and the shader code for a shader stage in the pipeline. + + When setting up a QRhiGraphicsPipeline, a collection of shader stages are + specified. The QRhiShaderStage contains a QShader and some associated + metadata, such as the graphics pipeline stage, and the + \l{QShader::Variant}{shader variant} to select. There is no need to specify + the shader language or version because the QRhi backend in use at runtime + will take care of choosing the appropriate shader version from the + collection within the QShader. + + The typical usage is in combination with + QRhiGraphicsPipeline::setShaderStages(), shown here with a simple approach + to load the QShader from \c{.qsb} files generated offline or at build time: + + \code + QShader getShader(const QString &name) + { + QFile f(name); + return f.open(QIODevice::ReadOnly) ? QShader::fromSerialized(f.readAll()) : QShader(); + } + + QShader vs = getShader("material.vert.qsb"); + QShader fs = getShader("material.frag.qsb"); + pipeline->setShaderStages({ + { QRhiShaderStage::Vertex, vs }, + { QRhiShaderStage::Fragment, fs } + }); + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1388,18 +2124,63 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputLayout &v) Specifies the type of the shader stage. \value Vertex Vertex stage - \value Fragment Fragment (pixel) stage - \value Compute Compute stage (this may not always be supported at run time) + + \value TessellationControl Tessellation control (hull shader) stage. Must + be used only when the QRhi::Tessellation feature is supported. + + \value TessellationEvaluation Tessellation evaluation (domain shader) + stage. Must be used only when the QRhi::Tessellation feature is supported. + + \value Fragment Fragment (pixel shader) stage + + \value Compute Compute stage. Must be used only when the QRhi::Compute + feature is supported. + + \value Geometry Geometry stage. Must be used only when the + QRhi::GeometryShader feature is supported. */ /*! - \fn QRhiShaderStage::QRhiShaderStage() + \fn QRhiShaderStage::QRhiShaderStage() = default Constructs a shader stage description for the vertex stage with an empty QShader. */ /*! + \fn QRhiShaderStage::Type QRhiShaderStage::type() const + \return the type of the stage. + */ + +/*! + \fn void QRhiShaderStage::setType(Type t) + + Sets the type of the stage to \a t. Setters should rarely be needed in + pratice. Most applications will likely use the QRhiShaderStage constructor + in most cases. + */ + +/*! + \fn QShader QRhiShaderStage::shader() const + \return the QShader to be used for this stage in the graphics pipeline. + */ + +/*! + \fn void QRhiShaderStage::setShader(const QShader &s) + Sets the shader collection \a s. + */ + +/*! + \fn QShader::Variant QRhiShaderStage::shaderVariant() const + \return the requested shader variant. + */ + +/*! + \fn void QRhiShaderStage::setShaderVariant(QShader::Variant v) + Sets the requested shader variant \a v. + */ + +/*! Constructs a shader stage description with the \a type of the stage and the \a shader. @@ -1416,38 +2197,24 @@ QRhiShaderStage::QRhiShaderStage(Type type, const QShader &shader, QShader::Vari } /*! + \fn bool QRhiShaderStage::operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + \return \c true if the values in the two QRhiShaderStage objects \a a and \a b are equal. - - \relates QRhiShaderStage */ -bool operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept -{ - return a.type() == b.type() - && a.shader() == b.shader() - && a.shaderVariant() == b.shaderVariant(); -} /*! + \fn bool QRhiShaderStage::operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + \return \c false if the values in the two QRhiShaderStage objects \a a and \a b are equal; otherwise returns \c true. - - \relates QRhiShaderStage */ -bool operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept -{ - return !(a == b); -} /*! - \return the hash value for \a v, using \a seed to seed the calculation. + \fn size_t QRhiShaderStage::qHash(const QRhiShaderStage &v, size_t seed = 0) noexcept - \relates QRhiShaderStage + \return the hash value for \a v, using \a seed to seed the calculation. */ -size_t qHash(const QRhiShaderStage &v, size_t seed) noexcept -{ - return v.type() + qHash(v.shader(), seed) + v.shaderVariant(); -} #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) @@ -1463,12 +2230,14 @@ QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) /*! \class QRhiColorAttachment - \internal \inmodule QtGui + \since 6.6 \brief Describes the a single color attachment of a render target. A color attachment is either a QRhiTexture or a QRhiRenderBuffer. The - former, when texture() is set, is used in most cases. + former, i.e. when texture() is set, is used in most cases. + QRhiColorAttachment is commonly used in combination with + QRhiTextureRenderTargetDescription. \note texture() and renderBuffer() cannot be both set (be non-null at the same time). @@ -1483,7 +2252,7 @@ QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) When targeting a non-multisample texture, the layer() and level() indicate the targeted layer (face index \c{0-5} for cubemaps) and mip level. For 3D textures layer() specifies the slice (one 2D image within the 3D texture) - to render to. + to render to. For texture arrays layer() is the array index. When texture() or renderBuffer() is multisample, resolveTexture() can be set optionally. When set, samples are resolved automatically into that @@ -1495,10 +2264,15 @@ QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) \note when resolving is enabled, the multisample data may not be written out at all. This means that the multisample texture() must not be used afterwards with shaders for sampling when resolveTexture() is set. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiTextureRenderTargetDescription */ /*! - \fn QRhiColorAttachment::QRhiColorAttachment() + \fn QRhiColorAttachment::QRhiColorAttachment() = default Constructs an empty color attachment description. */ @@ -1522,9 +2296,194 @@ QRhiColorAttachment::QRhiColorAttachment(QRhiRenderBuffer *renderBuffer) } /*! + \fn QRhiTexture *QRhiColorAttachment::texture() const + + \return the texture this attachment description references, or \nullptr if + there is none. + */ + +/*! + \fn void QRhiColorAttachment::setTexture(QRhiTexture *tex) + + Sets the texture \a tex. + + \note texture() and renderBuffer() cannot be both set (be non-null at the + same time). + */ + +/*! + \fn QRhiRenderBuffer *QRhiColorAttachment::renderBuffer() const + + \return the renderbuffer this attachment description references, or + \nullptr if there is none. + + In practice associating a QRhiRenderBuffer with a QRhiColorAttachment makes + the most sense when setting up multisample rendering via a multisample + \l{QRhiRenderBuffer::Type}{color} renderbuffer that is then resolved into a + non-multisample texture at the end of the render pass. + */ + +/*! + \fn void QRhiColorAttachment::setRenderBuffer(QRhiRenderBuffer *rb) + + Sets the renderbuffer \a rb. + + \note texture() and renderBuffer() cannot be both set (be non-null at the + same time). + */ + +/*! + \fn int QRhiColorAttachment::layer() const + \return the layer index (cubemap face or array layer). 0 by default. + */ + +/*! + \fn void QRhiColorAttachment::setLayer(int layer) + Sets the \a layer index. + */ + +/*! + \fn int QRhiColorAttachment::level() const + \return the mip level. 0 by default. + */ + +/*! + \fn void QRhiColorAttachment::setLevel(int level) + Sets the mip \a level. + */ + +/*! + \fn QRhiTexture *QRhiColorAttachment::resolveTexture() const + + \return the resolve texture this attachment description references, or + \nullptr if there is none. + + Setting a non-null resolve texture is applicable when the attachment + references a multisample texture or renderbuffer. The QRhiTexture in the + resolveTexture() is then a non-multisample 2D texture (or texture array) + with the same size (but a sample count of 1). The multisample content is + automatically resolved into this texture at the end of each render pass. + */ + +/*! + \fn void QRhiColorAttachment::setResolveTexture(QRhiTexture *tex) + + Sets the resolve texture \a tex. + + \a tex is expected to be a 2D texture or a 2D texture array. In either + case, resolving targets a single mip level of a single layer (array + element) of \a tex. The mip level and array layer are specified by + resolveLevel() and resolveLayer(). + + An exception is \l{setMultiViewCount()}{multiview}: when the color + attachment is associated with a texture array and multiview is enabled, the + resolve texture must also be a texture array with sufficient elements for + all views. In this case all elements that correspond to views are resolved + automatically; the behavior is similar to the following pseudo-code: + \badcode + for (i = 0; i < multiViewCount(); ++i) + resolve texture's layer() + i into resolveTexture's resolveLayer() + i + \endcode + + Setting a non-multisample texture to resolve a multisample texture or + renderbuffer automatically at the end of the render pass is often + preferable to working with multisample textures (and not setting a resolve + texture), because it avoids the need for writing dedicated fragment shaders + that work exclusively with multisample textures (\c sampler2DMS, \c + texelFetch, etc.), and rather allows using the same shader as one would if + the attachment's texture was not multisampled to begin with. This comes at + the expense of an additional resource (the non-multisample \a tex). + */ + +/*! + \fn int QRhiColorAttachment::resolveLayer() const + \return the currently set resolve texture layer. Defaults to 0. + */ + +/*! + \fn void QRhiColorAttachment::setResolveLayer(int layer) + Sets the resolve texture \a layer to use. + */ + +/*! + \fn int QRhiColorAttachment::resolveLevel() const + \return the currently set resolve texture mip level. Defaults to 0. + */ + +/*! + \fn void QRhiColorAttachment::setResolveLevel(int level) + Sets the resolve texture mip \a level to use. + */ + +/*! + \fn int QRhiColorAttachment::multiViewCount() const + + \return the currently set number of views. Defaults to 0 which indicates + the render target with this color attachment is not going to be used with + multiview rendering. + + \since 6.7 + */ + +/*! + \fn void QRhiColorAttachment::setMultiViewCount(int count) + + Sets the view \a count. Setting a value larger than 1 indicates that the + render target with this color attachment is going to be used with multiview + rendering. The default value is 0. Values smaller than 2 indicate no + multiview rendering. + + When \a count is set to \c 2 or greater, the color attachment must be + associated with a 2D texture array. layer() and multiViewCount() together + define the range of texture array elements that are targeted during + multiview rendering. + + For example, if \c layer is \c 0 and \c multiViewCount is \c 2, the texture + array must have 2 (or more) elements, and the multiview rendering will + target elements 0 and 1. The \c{gl_ViewIndex} variable in the shaders has a + value of \c 0 or \c 1 then, where view \c 0 corresponds to the texture array + element \c 0, and view \c 1 to the array element \c 1. + + \note Setting a \a count larger than 1, using a texture array as texture(), + and calling \l{QRhiCommandBuffer::beginPass()}{beginPass()} on a + QRhiTextureRenderTarget with this color attachment implies multiview + rendering for the entire render pass. multiViewCount() should not be set + unless multiview rendering is wanted. Multiview cannot be used with texture + types other than 2D texture arrays. (although 3D textures may work, + depending on the graphics API and backend; applications are nonetheless + advised not to rely on that and only use 2D texture arrays as the render + targets of multiview rendering) + + See + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt}{GL_OVR_multiview} + for more details regarding multiview rendering. Do note that Qt requires + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview2.txt}{GL_OVR_multiview2} + as well, when running on OpenGL (ES). + + Multiview rendering is available only when the + \l{QRhi::MultiView}{MultiView} feature is reported as supported from + \l{QRhi::isFeatureSupported()}{isFeatureSupported()}. + + \note For portability, be aware of limitations that exist for multiview + rendering with some of the graphics APIs. It is recommended that multiview + render passes do not rely on any of the features that + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt}{GL_OVR_multiview} + declares as unsupported. The one exception is shader stage outputs other + than \c{gl_Position} depending on \c{gl_ViewIndex}: that can be relied on + (even with OpenGL) because QRhi never reports multiview as supported without + \c{GL_OVR_multiview2} also being present. + + \note Multiview rendering is not supported in combination with tessellation + or geometry shaders, even though some implementations of some graphics APIs + may allow this. + + \since 6.7 + */ + +/*! \class QRhiTextureRenderTargetDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes the color and depth or depth/stencil attachments of a render target. A texture render target has zero or more textures as color attachments, @@ -1533,10 +2492,97 @@ QRhiColorAttachment::QRhiColorAttachment(QRhiRenderBuffer *renderBuffer) \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be non-null at the same time). + + Let's look at some example usages in combination with + QRhiTextureRenderTarget. + + Due to the constructors, the targeting a texture (and no depth/stencil + buffer) is simple: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(256, 256), 1, QRhiTexture::RenderTarget); + texture->create(); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ texture })); + \endcode + + The following creates a texture render target that is set up to target mip + level #2 of a texture: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1, QRhiTexture::RenderTarget | QRhiTexture::MipMapped); + texture->create(); + QRhiColorAttachment colorAtt(texture); + colorAtt.setLevel(2); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ colorAtt }); + \endcode + + Another example, this time to render into a depth texture: + + \code + QRhiTexture *shadowMap = rhi->newTexture(QRhiTexture::D32F, QSize(1024, 1024), 1, QRhiTexture::RenderTarget); + shadowMap->create(); + QRhiTextureRenderTargetDescription rtDesc; + rtDesc.setDepthTexture(shadowMap); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget(rtDesc); + \endcode + + A very common case, having a texture as the color attachment and a + renderbuffer as depth/stencil to enable depth testing: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1. QRhiTexture::RenderTarget); + texture->create(); + QRhiRenderBuffer *depthStencil = rhi->newRenderBuffer(QRhiRenderBuffer::DepthStencil, QSize(512, 512)); + depthStencil->create(); + QRhiTextureRenderTargetDescription rtDesc({ texture }, depthStencil); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget(rtDesc); + \endcode + + Finally, to enable multisample rendering in a portable manner (so also + supporting OpenGL ES 3.0), using a QRhiRenderBuffer as the (multisample) + color buffer and then resolving into a regular (non-multisample) 2D + texture. To enable depth testing, a depth-stencil buffer, which also must + use the same sample count, is used as well: + + \code + QRhiRenderBuffer *colorBuffer = rhi->newRenderBuffer(QRhiRenderBuffer::Color, QSize(512, 512), 4); // 4x MSAA + colorBuffer->create(); + QRhiRenderBuffer *depthStencil = rhi->newRenderBuffer(QRhiRenderBuffer::DepthStencil, QSize(512, 512), 4); + depthStencil->create(); + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1, QRhiTexture::RenderTarget); + texture->create(); + QRhiColorAttachment colorAtt(colorBuffer); + colorAtt.setResolveTexture(texture); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ colorAtt, depthStencil }); + \endcode + + \note when multisample resolving is enabled, the multisample data may not be + written out at all. This means that the multisample texture in a color + attachment must not be used afterwards with shaders for sampling (or other + purposes) whenever a resolve texture is set, since the multisample color + buffer is merely an intermediate storage then that gets no data written back + on some GPU architectures at all. See + \l{QRhiTextureRenderTarget::Flag}{PreserveColorContents} for more details. + + \note When using setDepthTexture(), not setDepthStencilBuffer(), and the + depth (stencil) data is not of interest afterwards, set the + DoNotStoreDepthStencilContents flag on the QRhiTextureRenderTarget. This + allows indicating to the underlying 3D API that the depth/stencil data can + be discarded, leading potentially to better performance with tiled GPU + architectures. When the depth-stencil buffer is a QRhiRenderBuffer (and also + for the multisample color texture, see previous note) this is implicit, but + with a depth (stencil) QRhiTexture the intention needs to be declared + explicitly. By default QRhi assumes that the data is of interest (e.g., the + depth texture is sampled in a shader afterwards). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiColorAttachment, QRhiTextureRenderTarget */ /*! - \fn QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription() + \fn QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription() = default Constructs an empty texture render target description. */ @@ -1578,9 +2624,125 @@ QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription(const QRh } /*! + \fn void QRhiTextureRenderTargetDescription::setColorAttachments(std::initializer_list<QRhiColorAttachment> list) + Sets the \a list of color attachments. + */ + +/*! + \fn template<typename InputIterator> void QRhiTextureRenderTargetDescription::setColorAttachments(InputIterator first, InputIterator last) + Sets the list of color attachments via the iterators \a first and \a last. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::cbeginColorAttachments() const + \return a const iterator pointing to the first item in the attachment list. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::cendColorAttachments() const + \return a const iterator pointing just after the last item in the attachment list. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::colorAttachmentAt(qsizetype index) const + \return the color attachment at the specified \a index. + */ + +/*! + \fn qsizetype QRhiTextureRenderTargetDescription::colorAttachmentCount() const + \return the number of currently set color attachments. + */ + +/*! + \fn QRhiRenderBuffer *QRhiTextureRenderTargetDescription::depthStencilBuffer() const + \return the renderbuffer used as depth-stencil buffer, or \nullptr if none was set. + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) + + Sets the \a renderBuffer for depth-stencil. Not mandatory, e.g. when no + depth test/write or stencil-related features are used within any graphics + pipelines in any of the render passes for this render target, it can be + left set to \nullptr. + + \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be + non-null at the same time). + + Using a QRhiRenderBuffer over a 2D QRhiTexture as the depth or + depth/stencil buffer is very common, and is the recommended approach for + applications. Using a QRhiTexture, and so setDepthTexture() becomes + relevant if the depth data is meant to be accessed (e.g. sampled in a + shader) afterwards, or when + \l{QRhiColorAttachment::setMultiViewCount()}{multiview rendering} is + involved (because then the depth texture must be a texture array). + + \sa setDepthTexture() + */ + +/*! + \fn QRhiTexture *QRhiTextureRenderTargetDescription::depthTexture() const + \return the currently referenced depth texture, or \nullptr if none was set. + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthTexture(QRhiTexture *texture) + + Sets the \a texture for depth-stencil. This is an alternative to + setDepthStencilBuffer(), where instead of a QRhiRenderBuffer a QRhiTexture + with a suitable type (e.g., QRhiTexture::D32F) is provided. + + \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be + non-null at the same time). + + \a texture can either be a 2D texture or a 2D texture array (when texture + arrays are supported). Specifying a texture array is relevant in particular + with + \l{QRhiColorAttachment::setMultiViewCount()}{multiview rendering}. + + \note If \a texture is a format with a stencil component, such as + \l QRhiTexture::D24S8, it will serve as the stencil buffer as well. + + \sa setDepthStencilBuffer() + */ + +/*! + \fn QRhiTexture *QRhiTextureRenderTargetDescription::depthResolveTexture() const + + \return the texture to which a multisample depth (or depth-stencil) texture + (or texture array) is resolved to. \nullptr if there is none, which is the + most common case. + + \since 6.8 + \sa QRhiColorAttachment::resolveTexture(), depthTexture() + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthResolveTexture(QRhiTexture *tex) + + Sets the depth (or depth-stencil) resolve texture \a tex. + + \a tex is expected to be a 2D texture or a 2D texture array with a format + matching the texture set via setDepthTexture(). + + \note Resolving depth (or depth-stencil) data is only functional when the + \l ResolveDepthStencil feature is reported as supported at run time. Support + for depth-stencil resolve is not universally available among the graphics + APIs. Designs assuming unconditional availability of depth-stencil resolve + are therefore non-portable, and should be avoided. + + \note As an additional limitation for OpenGL ES in particular, setting a + depth resolve texture may only be functional in combination with + setDepthTexture(), not with setDepthStencilBuffer(). + + \since 6.8 + \sa QRhiColorAttachment::setResolveTexture(), setDepthTexture() + */ + +/*! \class QRhiTextureSubresourceUploadDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes the source for one mip level in a layer in a texture upload operation. The source content is specified either as a QImage or as a raw blob. The @@ -1626,10 +2788,26 @@ QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription(const QRh QRhi::ImageDataStride is reported as \l{QRhi::isFeatureSupported()}{supported}. In practice this can be expected to be supported everywhere except for OpenGL ES 2.0. + + \note When a QImage is given, the stride returned from + QImage::bytesPerLine() is taken into account automatically. + + \warning When a QImage is given and the QImage does not own the underlying + pixel data, it is up to the caller to ensure that the associated data stays + valid until the end of the frame. (just submitting the resource update batch + is not sufficient, the data must stay valid until QRhi::endFrame() is called + in order to be portable across all backends) If this cannot be ensured, the + caller is strongly encouraged to call QImage::detach() on the image before + passing it to uploadTexture(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiTextureUploadDescription */ /*! - \fn QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription() + \fn QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription() = default Constructs an empty subresource description. @@ -1662,7 +2840,7 @@ QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription \a data can safely be destroyed or changed once this function returns. */ -QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription(const void *data, int size) +QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription(const void *data, quint32 size) : m_data(reinterpret_cast<const char *>(data), size) { } @@ -1677,10 +2855,109 @@ QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription } /*! + \fn QImage QRhiTextureSubresourceUploadDescription::image() const + \return the currently set QImage. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setImage(const QImage &image) + + Sets \a image. + Upon textures loading, the image data will be read as is, with no formats conversions. + + \note image() and data() cannot be both set at the same time. + */ + +/*! + \fn QByteArray QRhiTextureSubresourceUploadDescription::data() const + \return the currently set raw pixel data. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setData(const QByteArray &data) + + Sets \a data. + + \note image() and data() cannot be both set at the same time. + */ + +/*! + \fn quint32 QRhiTextureSubresourceUploadDescription::dataStride() const + \return the currently set data stride. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setDataStride(quint32 stride) + + Sets the data \a stride in bytes. By default this is 0 and not always + relevant. When providing raw data(), and the stride is not specified via + setDataStride(), the stride (row pitch, row length in bytes) of the + provided data must be equal to \c{width * pixelSize} where \c pixelSize is + the number of bytes used for one pixel, and there must be no additional + padding between rows. Otherwise, if there is additional space between the + lines, set a non-zero \a stride. All this is applicable only when raw image + data is provided, and is not necessary when working QImage since that has + its own \l{QImage::bytesPerLine()}{stride} value. + + \note Setting the stride via setDataStride() is only functional when + QRhi::ImageDataStride is reported as + \l{QRhi::isFeatureSupported()}{supported}. + + \note When a QImage is given, the stride returned from + QImage::bytesPerLine() is taken into account automatically and therefore + there is no need to set the data stride manually. + */ + +/*! + \fn QPoint QRhiTextureSubresourceUploadDescription::destinationTopLeft() const + \return the currently set destination top-left position. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setDestinationTopLeft(const QPoint &p) + Sets the destination top-left position \a p. + */ + +/*! + \fn QSize QRhiTextureSubresourceUploadDescription::sourceSize() const + + \return the source size in pixels. Defaults to a default-constructed QSize, + which indicates the entire subresource. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setSourceSize(const QSize &size) + + Sets the source \a size in pixels. + + \note Setting sourceSize() or sourceTopLeft() may trigger a QImage copy + internally, depending on the format and the backend. + */ + +/*! + \fn QPoint QRhiTextureSubresourceUploadDescription::sourceTopLeft() const + \return the currently set source top-left position. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setSourceTopLeft(const QPoint &p) + + Sets the source top-left position \a p. + + \note Setting sourceSize() or sourceTopLeft() may trigger a QImage copy + internally, depending on the format and the backend. + */ + +/*! \class QRhiTextureUploadEntry - \internal \inmodule QtGui - \brief Describes one layer (face for cubemaps) in a texture upload operation. + \since 6.6 + + \brief Describes one layer (face for cubemaps, slice for 3D textures, + element for texture arrays) in a texture upload operation. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1706,18 +2983,61 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, } /*! + \fn int QRhiTextureUploadEntry::layer() const + \return the currently set layer index (cubemap face, array layer). Defaults to 0. + */ + +/*! + \fn void QRhiTextureUploadEntry::setLayer(int layer) + Sets the \a layer. + */ + +/*! + \fn int QRhiTextureUploadEntry::level() const + \return the currently set mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureUploadEntry::setLevel(int level) + Sets the mip \a level. + */ + +/*! + \fn QRhiTextureSubresourceUploadDescription QRhiTextureUploadEntry::description() const + \return the currently set subresource description. + */ + +/*! + \fn void QRhiTextureUploadEntry::setDescription(const QRhiTextureSubresourceUploadDescription &desc) + Sets the subresource description \a desc. + */ + +/*! \class QRhiTextureUploadDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a texture upload operation. Used with QRhiResourceUpdateBatch::uploadTexture(). That function has two variants: one taking a QImage and one taking a QRhiTextureUploadDescription. The former is a convenience version, internally creating a QRhiTextureUploadDescription with a single image - targeting level 0 for layer 0. However, when cubemaps, pre-generated mip - images, or compressed textures are involved, applications will have to work - directly with this class instead. + targeting level 0 for layer 0. + + An example of the the common, simple case of wanting to upload the contents + of a QImage to a QRhiTexture with a matching pixel size: + + \code + QImage image(256, 256, QImage::Format_RGBA8888); + image.fill(Qt::green); // or could use a QPainter targeting image + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(256, 256)); + texture->create(); + QRhiResourceUpdateBatch *u = rhi->nextResourceUpdateBatch(); + u->uploadTexture(texture, image); + \endcode + + When cubemaps, pre-generated mip images, compressed textures, or partial + uploads are involved, applications will have to use this class instead. QRhiTextureUploadDescription also enables specifying batched uploads, which are useful for example when generating an atlas or glyph cache texture: @@ -1733,32 +3053,35 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, For example, specifying the faces of a cubemap could look like the following: - \badcode + \code QImage faces[6]; - ... - QList<QRhiTextureUploadEntry> entries; + // ... + QVarLengthArray<QRhiTextureUploadEntry, 6> entries; for (int i = 0; i < 6; ++i) entries.append(QRhiTextureUploadEntry(i, 0, faces[i])); - QRhiTextureUploadDescription desc(entries); + QRhiTextureUploadDescription desc; + desc.setEntries(entries.cbegin(), entries.cend()); resourceUpdates->uploadTexture(texture, desc); \endcode Another example that specifies mip images for a compressed texture: - \badcode - QRhiTextureUploadDescription desc; + \code + QList<QRhiTextureUploadEntry> entries; const int mipCount = rhi->mipLevelsForSize(compressedTexture->pixelSize()); for (int level = 0; level < mipCount; ++level) { const QByteArray compressedDataForLevel = .. - desc.append(QRhiTextureUploadEntry(0, level, compressedDataForLevel)); + entries.append(QRhiTextureUploadEntry(0, level, compressedDataForLevel)); } + QRhiTextureUploadDescription desc; + desc.setEntries(entries.cbegin(), entries.cend()); resourceUpdates->uploadTexture(compressedTexture, desc); \endcode With partial uploads targeting the same subresource, it is recommended to batch them into a single upload request, whenever possible: - \badcode + \code QRhiTextureSubresourceUploadDescription subresDesc(image); subresDesc.setSourceSize(QSize(10, 10)); subResDesc.setDestinationTopLeft(QPoint(50, 40)); @@ -1772,6 +3095,11 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, QRhiTextureUploadDescription desc({ entry, entry2}); resourceUpdates->uploadTexture(texture, desc); \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiResourceUpdateBatch */ /*! @@ -1804,9 +3132,39 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list } /*! + \fn void QRhiTextureUploadDescription::setEntries(std::initializer_list<QRhiTextureUploadEntry> list) + Sets the \a list of entries. + */ + +/*! + \fn template<typename InputIterator> void QRhiTextureUploadDescription::setEntries(InputIterator first, InputIterator last) + Sets the list of entries using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::cbeginEntries() const + \return a const iterator pointing to the first item in the entry list. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::cendEntries() const + \return a const iterator pointing just after the last item in the entry list. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::entryAt(qsizetype index) const + \return the entry at \a index. + */ + +/*! + \fn qsizetype QRhiTextureUploadDescription::entryCount() const + \return the number of entries. + */ + +/*! \class QRhiTextureCopyDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a texture-to-texture copy operation. An empty pixelSize() indicates that the entire subresource is to be copied. @@ -1820,12 +3178,15 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list sourceTopLeft(), and destinationTopLeft() must fit the source and destination textures, respectively. The behavior is undefined otherwise. - With cubemap and 3D textures one face or slice can be copied at a time. The - face or slice is specified by the source and destination layer indices. - With mipmapped textures one mip level can be copied at a time. The source - and destination layer and mip level indices can differ, but the size and - position must be carefully controlled to avoid out of bounds copies, in - which case the behavior is undefined. + With cubemaps, 3D textures, and texture arrays one face or slice can be + copied at a time. The face or slice is specified by the source and + destination layer indices. With mipmapped textures one mip level can be + copied at a time. The source and destination layer and mip level indices can + differ, but the size and position must be carefully controlled to avoid out + of bounds copies, in which case the behavior is undefined. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1835,9 +3196,83 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list */ /*! + \fn QSize QRhiTextureCopyDescription::pixelSize() const + \return the size of the region to copy. + + \note An empty pixelSize() indicates that the entire subresource is to be + copied. A default constructed copy description therefore leads to copying + the entire subresource at level 0 of layer 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setPixelSize(const QSize &sz) + Sets the size of the region to copy to \a sz. + */ + +/*! + \fn int QRhiTextureCopyDescription::sourceLayer() const + \return the source array layer (cubemap face or array layer index). Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceLayer(int layer) + Sets the source array \a layer. + */ + +/*! + \fn int QRhiTextureCopyDescription::sourceLevel() const + \return the source mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceLevel(int level) + Sets the source mip \a level. + */ + +/*! + \fn QPoint QRhiTextureCopyDescription::sourceTopLeft() const + \return the source top-left position (in pixels). Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceTopLeft(const QPoint &p) + Sets the source top-left position to \a p. + */ + +/*! + \fn int QRhiTextureCopyDescription::destinationLayer() const + \return the destination array layer (cubemap face or array layer index). Default to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationLayer(int layer) + Sets the destination array \a layer. + */ + +/*! + \fn int QRhiTextureCopyDescription::destinationLevel() const + \return the destionation mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationLevel(int level) + Sets the destination mip \a level. + */ + +/*! + \fn QPoint QRhiTextureCopyDescription::destinationTopLeft() const + \return the destionation top-left position in pixels. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationTopLeft(const QPoint &p) + Sets the destination top-left position \a p. + */ + +/*! \class QRhiReadbackDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a readback (reading back texture contents from possibly GPU-only memory) operation. The source of the readback operation is either a QRhiTexture or the @@ -1855,10 +3290,13 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list \note Multisample textures cannot be read back. Readbacks are supported for multisample swapchain buffers however. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiReadbackDescription::QRhiReadbackDescription() + \fn QRhiReadbackDescription::QRhiReadbackDescription() = default Constructs an empty texture readback description. @@ -1882,32 +3320,140 @@ QRhiReadbackDescription::QRhiReadbackDescription(QRhiTexture *texture) } /*! + \fn QRhiTexture *QRhiReadbackDescription::texture() const + + \return the QRhiTexture that is read back. Can be left set to \nullptr + which indicates that the backbuffer of the current swapchain is to be used + instead. + */ + +/*! + \fn void QRhiReadbackDescription::setTexture(QRhiTexture *tex) + + Sets the texture \a tex as the source of the readback operation. + + Setting \nullptr is valid too, in which case the current swapchain's + current backbuffer is used. (but then the readback cannot be issued in a + non-swapchain-based frame) + + \note Multisample textures cannot be read back. Readbacks are supported for + multisample swapchain buffers however. + + \note Textures used in readbacks must be created with + QRhiTexture::UsedAsTransferSource. + + \note Swapchains used in readbacks must be created with + QRhiSwapChain::UsedAsTransferSource. + */ + +/*! + \fn int QRhiReadbackDescription::layer() const + + \return the currently set array layer (cubemap face, array index). Defaults to 0. + + Applicable only when the source of the readback is a QRhiTexture. + */ + +/*! + \fn void QRhiReadbackDescription::setLayer(int layer) + Sets the array \a layer to read back. + */ + +/*! + \fn int QRhiReadbackDescription::level() const + + \return the currently set mip level. Defaults to 0. + + Applicable only when the source of the readback is a QRhiTexture. + */ + +/*! + \fn void QRhiReadbackDescription::setLevel(int level) + Sets the mip \a level to read back. + */ + +/*! \class QRhiReadbackResult - \internal \inmodule QtGui - \brief Describes the results of a potentially asynchronous readback operation. + \since 6.6 + \brief Describes the results of a potentially asynchronous buffer or texture readback operation. When \l completed is set, the function is invoked when the \l data is available. \l format and \l pixelSize are set upon completion together with \l data. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiReadbackResult::completed + + Callback that is invoked upon completion, on the thread the QRhi operates + on. Can be left set to \nullptr, in which case no callback is invoked. + */ + +/*! + \variable QRhiReadbackResult::format + + Valid only for textures, the texture format. + */ + +/*! + \variable QRhiReadbackResult::pixelSize + + Valid only for textures, the size in pixels. + */ + +/*! + \variable QRhiReadbackResult::data + + The buffer or image data. + + \sa QRhiResourceUpdateBatch::readBackTexture(), QRhiResourceUpdateBatch::readBackBuffer() */ + /*! \class QRhiNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Base class for classes exposing backend-specific collections of native resource objects. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! \class QRhiResource - \internal \inmodule QtGui + \since 6.6 \brief Base class for classes encapsulating native resource objects. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiResource::Type QRhiResource::resourceType() const + \enum QRhiResource::Type + Specifies type of the resource. + + \value Buffer + \value Texture + \value Sampler + \value RenderBuffer + \value RenderPassDescriptor + \value SwapChainRenderTarget + \value TextureRenderTarget + \value ShaderResourceBindings + \value GraphicsPipeline + \value SwapChain + \value ComputePipeline + \value CommandBuffer + */ + +/*! + \fn virtual QRhiResource::Type QRhiResource::resourceType() const = 0 \return the type of the resource. */ @@ -1939,7 +3485,7 @@ QRhiResource::~QRhiResource() } /*! - \fn void QRhiResource::destroy() + \fn virtual void QRhiResource::destroy() = 0 Releases (or requests deferred releasing of) the underlying native graphics resources. Safe to call multiple times, subsequent invocations will be a @@ -1953,7 +3499,7 @@ QRhiResource::~QRhiResource() released until the frame is submitted by QRhi::endFrame(). The QRhiResource destructor also performs the same task, so calling this - function is not necessary before destroying a QRhiResource. + function is not necessary before deleting a QRhiResource. \sa deleteLater() */ @@ -1966,11 +3512,42 @@ QRhiResource::~QRhiResource() requirement of not altering QRhiResource objects that are referenced by the frame being recorded. + If the QRhi that created this object is already destroyed, the object is + deleted immediately. + + Using deleteLater() can be a useful convenience in many cases, and it + complements the low-level guarantee (that the underlying native graphics + objects are never destroyed until it is safe to do so and it is known for + sure that they are not used by the GPU in an still in-flight frame), by + offering a way to make sure the C++ object instances (of QRhiBuffer, + QRhiTexture, etc.) themselves also stay valid until the end of the current + frame. + + The following example shows a convenient way of creating a throwaway buffer + that is only used in one frame and gets automatically released in + endFrame(). (when it comes to the underlying native buffer(s), the usual + guarantee applies: the QRhi backend defers the releasing of those until it + is guaranteed that the frame in which the buffer is accessed by the GPU has + completed) + + \code + rhi->beginFrame(swapchain); + QRhiBuffer *buf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, 256); + buf->deleteLater(); // ! + u = rhi->nextResourceUpdateBatch(); + u->uploadStaticBuffer(buf, data); + // ... draw with buf + rhi->endFrame(); + \endcode + \sa destroy() */ void QRhiResource::deleteLater() { - m_rhi->addDeleteLater(this); + if (m_rhi) + m_rhi->addDeleteLater(this); + else + delete this; } /*! @@ -1984,11 +3561,10 @@ QByteArray QRhiResource::name() const /*! Sets a \a name for the object. - This has two uses: to get descriptive names for the native graphics + This allows getting descriptive names for the native graphics resources visible in graphics debugging tools, such as \l{https://renderdoc.org/}{RenderDoc} and - \l{https://developer.apple.com/xcode/}{XCode}, and in the output stream of - QRhiProfiler. + \l{https://developer.apple.com/xcode/}{XCode}. When it comes to naming native objects by relaying the name via the appropriate graphics API, note that the name is ignored when @@ -2005,7 +3581,6 @@ QByteArray QRhiResource::name() const void QRhiResource::setName(const QByteArray &name) { m_objectName = name; - m_objectName.replace(',', '_'); // cannot contain comma for QRhiProfiler } /*! @@ -2020,10 +3595,141 @@ quint64 QRhiResource::globalResourceId() const } /*! + \return the QRhi that created this resource. + + If the QRhi that created this object is already destroyed, the result is + \nullptr. + */ +QRhi *QRhiResource::rhi() const +{ + return m_rhi ? m_rhi->q : nullptr; +} + +/*! \class QRhiBuffer - \internal \inmodule QtGui + \since 6.6 \brief Vertex, index, or uniform (constant) buffer resource. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + A QRhiBuffer encapsulates zero, one, or more native buffer objects (such as + a \c VkBuffer or \c MTLBuffer). With some graphics APIs and backends + certain types of buffers may not use a native buffer object at all (e.g. + OpenGL if uniform buffer objects are not used), but this is transparent to + the user of the QRhiBuffer API. Similarly, the fact that some types of + buffers may use two or three native buffers underneath, in order to allow + efficient per-frame content update without stalling the GPU pipeline, is + mostly invisible to the applications and libraries. + + A QRhiBuffer instance is always created by calling + \l{QRhi::newBuffer()}{the QRhi's newBuffer() function}. This creates no + native graphics resources. To do that, call create() after setting the + appropriate options, such as the type, usage flags, size, although in most cases these + are already set based on the arguments passed to + \l{QRhi::newBuffer()}{newBuffer()}. + + \section2 Example usage + + To create a uniform buffer for a shader where the GLSL uniform block + contains a single \c mat4 member, and update the contents: + + \code + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 64); + if (!ubuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + QMatrix4x4 mvp; + // ... set up the modelview-projection matrix + batch->updateDynamicBuffer(ubuf, 0, 64, mvp.constData()); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + An example of creating a buffer with vertex data: + + \code + const float vertices[] = { -1.0f, -1.0f, 1.0f, -1.0f, 0.0f, 1.0f }; + QRhiBuffer *vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertices)); + if (!vbuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + batch->uploadStaticBuffer(vbuf, vertices); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + An index buffer: + + \code + static const quint16 indices[] = { 0, 1, 2 }; + QRhiBuffer *ibuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::IndexBuffer, sizeof(indices)); + if (!ibuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + batch->uploadStaticBuffer(ibuf, indices); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + \section2 Common patterns + + A call to create() destroys any existing native resources if create() was + successfully called before. If those native resources are still in use by + an in-flight frame (i.e., there's a chance they are still read by the GPU), + the destroying of those resources is deferred automatically. Thus a very + common and convenient pattern to safely increase the size of an already + initialized buffer is the following. In practice this drops and creates a + whole new set of native resources underneath, so it is not necessarily a + cheap operation, but is more convenient and still faster than the + alternatives, because by not destroying the \c buf object itself, all + references to it stay valid in other data structures (e.g., in any + QRhiShaderResourceBinding the QRhiBuffer is referenced from). + + \code + if (buf->size() < newSize) { + buf->setSize(newSize); + if (!buf->create()) { error(); } + } + // continue using buf, fill it with new data + \endcode + + When working with uniform buffers, it will sometimes be necessary to + combine data for multiple draw calls into a single buffer for efficiency + reasons. Be aware of the aligment requirements: with some graphics APIs + offsets for a uniform buffer must be aligned to 256 bytes. This applies + both to QRhiShaderResourceBinding and to the dynamic offsets passed to + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()}. Use the + \l{QRhi::ubufAlignment()}{ubufAlignment()} and + \l{QRhi::ubufAligned()}{ubufAligned()} functions to create portable code. + As an example, the following is an outline for issuing multiple (\c N) draw + calls with the same pipeline and geometry, but with a different data in the + uniform buffers exposed at binding point 0. This assumes the buffer is + exposed via + \l{QRhiShaderResourceBinding::uniformBufferWithDynamicOffset()}{uniformBufferWithDynamicOffset()} + which allows passing a QRhiCommandBuffer::DynamicOffset list to + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()}. + + \code + const int N = 2; + const int UB_SIZE = 64 + 4; // assuming a uniform block with { mat4 matrix; float opacity; } + const int ONE_UBUF_SIZE = rhi->ubufAligned(UB_SIZE); + const int TOTAL_UBUF_SIZE = N * ONE_UBUF_SIZE; + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, TOTAL_UBUF_SIZE); + if (!ubuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + for (int i = 0; i < N; ++i) { + batch->updateDynamicBuffer(ubuf, i * ONE_UBUF_SIZE, 64, matrix.constData()); + batch->updateDynamicBuffer(ubuf, i * ONE_UBUF_SIZE + 64, 4, &opacity); + } + // ... + // beginPass(), set pipeline, etc., and then: + for (int i = 0; i < N; ++i) { + QRhiCommandBuffer::DynamicOffset dynOfs[] = { { 0, i * ONE_UBUF_SIZE } }; + cb->setShaderResources(srb, 1, dynOfs); + cb->draw(36); + } + \endcode + + \sa QRhiResourceUpdateBatch, QRhi, QRhiCommandBuffer */ /*! @@ -2059,42 +3765,29 @@ quint64 QRhiResource::globalResourceId() const Flag values to specify how the buffer is going to be used. \value VertexBuffer Vertex buffer. This allows the QRhiBuffer to be used in - \l{setVertexInput()}{QRhiCommandBuffer::setVertexInput()}. + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()}. \value IndexBuffer Index buffer. This allows the QRhiBuffer to be used in - \l{setVertexInput()}{QRhiCommandBuffer::setVertexInput()}. + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()}. \value UniformBuffer Uniform buffer (also called constant buffer). This allows the QRhiBuffer to be used in combination with - \l{UniformBuffer}{QRhiShaderResourceBinding::UniformBuffer}. When + \l{QRhiShaderResourceBinding::UniformBuffer}{UniformBuffer}. When \l{QRhi::NonDynamicUniformBuffers}{NonDynamicUniformBuffers} is reported as not supported, this usage can only be combined with the type Dynamic. \value StorageBuffer Storage buffer. This allows the QRhiBuffer to be used - in combination with \l{BufferLoad}{QRhiShaderResourceBinding::BufferLoad}, - \l{BufferStore}{QRhiShaderResourceBinding::BufferStore}, or - \l{BufferLoadStore}{QRhiShaderResourceBinding::BufferLoadStore}. This usage + in combination with \l{QRhiShaderResourceBinding::BufferLoad}{BufferLoad}, + \l{QRhiShaderResourceBinding::BufferStore}{BufferStore}, or + \l{QRhiShaderResourceBinding::BufferLoadStore}{BufferLoadStore}. This usage can only be combined with the types Immutable or Static, and is only available when the \l{QRhi::Compute}{Compute feature} is reported as supported. */ /*! - \fn void QRhiBuffer::setSize(int sz) - - Sets the size of the buffer in bytes. The size is normally specified in - QRhi::newBuffer() so this function is only used when the size has to be - changed. As with other setters, the size only takes effect when calling - create(), and for already created buffers this involves releasing the previous - native resource and creating new ones under the hood. - - Backends may choose to allocate buffers bigger than \a sz in order to - fulfill alignment requirements. This is hidden from the applications and - size() will always report the size requested in \a sz. - */ - -/*! \class QRhiBuffer::NativeBuffer + \inmodule QtGui \brief Contains information about the underlying native resources of a buffer. */ @@ -2106,10 +3799,13 @@ quint64 QRhiResource::globalResourceId() const objects array are pointers to a GLuint. With Vulkan, the native handle is a VkBuffer, so the elements of the array are pointers to a VkBuffer. With Direct3D 11 and Metal the elements are pointers to a ID3D11Buffer or - MTLBuffer pointer, respectively. + MTLBuffer pointer, respectively. With Direct3D 12, the elements are + pointers to a ID3D12Resource. \note Pay attention to the fact that the elements are always pointers to the native buffer handle type, even if the native type itself is a pointer. + (so the elements are \c{VkBuffer *} on Vulkan, even though VkBuffer itself + is a pointer on 64-bit architectures). */ /*! @@ -2130,7 +3826,7 @@ quint64 QRhiResource::globalResourceId() const /*! \internal */ -QRhiBuffer::QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, int size_) +QRhiBuffer::QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, quint32 size_) : QRhiResource(rhi), m_type(type_), m_usage(usage_), m_size(size_) { @@ -2145,7 +3841,7 @@ QRhiResource::Type QRhiBuffer::resourceType() const } /*! - \fn bool QRhiBuffer::create() + \fn virtual bool QRhiBuffer::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2156,6 +3852,50 @@ QRhiResource::Type QRhiBuffer::resourceType() const */ /*! + \fn QRhiBuffer::Type QRhiBuffer::type() const + \return the buffer type. + */ + +/*! + \fn void QRhiBuffer::setType(Type t) + Sets the buffer's type to \a t. + */ + +/*! + \fn QRhiBuffer::UsageFlags QRhiBuffer::usage() const + \return the buffer's usage flags. + */ + +/*! + \fn void QRhiBuffer::setUsage(UsageFlags u) + Sets the buffer's usage flags to \a u. + */ + +/*! + \fn quint32 QRhiBuffer::size() const + + \return the buffer's size in bytes. + + This is always the value that was passed to setSize() or QRhi::newBuffer(). + Internally, the native buffers may be bigger if that is required by the + underlying graphics API. + */ + +/*! + \fn void QRhiBuffer::setSize(quint32 sz) + + Sets the size of the buffer in bytes. The size is normally specified in + QRhi::newBuffer() so this function is only used when the size has to be + changed. As with other setters, the size only takes effect when calling + create(), and for already created buffers this involves releasing the previous + native resource and creating new ones under the hood. + + Backends may choose to allocate buffers bigger than \a sz in order to + fulfill alignment requirements. This is hidden from the applications and + size() will always report the size requested in \a sz. + */ + +/*! \return the underlying native resources for this buffer. The returned value will be empty if exposing the underlying native resources is not supported by the backend. @@ -2214,7 +3954,7 @@ QRhiBuffer::NativeBuffer QRhiBuffer::nativeBuffer() depending on the backend. \warning When updating buffer data via this method, the update must be done - in every frame, otherwise backends that perform double or tripple buffering + in every frame, otherwise backends that perform double or triple buffering of resources may end up in unexpected behavior. \warning Partial updates are not possible with this approach since some @@ -2243,18 +3983,18 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() /*! \class QRhiRenderBuffer - \internal \inmodule QtGui + \since 6.6 \brief Renderbuffer resource. Renderbuffers cannot be sampled or read but have some benefits over textures in some cases: - A DepthStencil renderbuffer may be lazily allocated and be backed by + A \l DepthStencil renderbuffer may be lazily allocated and be backed by transient memory with some APIs. On some platforms this may mean the depth/stencil buffer uses no physical backing at all. - Color renderbuffers are useful since QRhi::MultisampleRenderBuffer may be + \l Color renderbuffers are useful since QRhi::MultisampleRenderBuffer may be supported even when QRhi::MultisampleTexture is not. How the renderbuffer is implemented by a backend is not exposed to the @@ -2268,6 +4008,9 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() QRhi provides automatic sizing behavior to match the color buffers, which means calling setPixelSize() and create() are not necessary for such renderbuffers. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2279,6 +4022,22 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() */ /*! + \struct QRhiRenderBuffer::NativeRenderBuffer + \inmodule QtGui + \brief Wraps a native renderbuffer object. + */ + +/*! + \variable QRhiRenderBuffer::NativeRenderBuffer::object + \brief 64-bit integer containing the native object handle. + + Used with QRhiRenderBuffer::createFrom(). + + With OpenGL the native handle is a GLuint value. \c object is expected to + be a valid OpenGL renderbuffer object ID. + */ + +/*! \enum QRhiRenderBuffer::Flag Flag values for flags() and setFlags() @@ -2315,7 +4074,7 @@ QRhiResource::Type QRhiRenderBuffer::resourceType() const } /*! - \fn bool QRhiRenderBuffer::create() + \fn virtual bool QRhiRenderBuffer::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2364,16 +4123,121 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) } /*! - \fn QRhiTexture::Format QRhiRenderBuffer::backingFormat() const + \fn QRhiRenderBuffer::Type QRhiRenderBuffer::type() const + \return the renderbuffer type. + */ + +/*! + \fn void QRhiRenderBuffer::setType(Type t) + Sets the type to \a t. + */ + +/*! + \fn QSize QRhiRenderBuffer::pixelSize() const + \return the pixel size. + */ + +/*! + \fn void QRhiRenderBuffer::setPixelSize(const QSize &sz) + Sets the size (in pixels) to \a sz. + */ + +/*! + \fn int QRhiRenderBuffer::sampleCount() const + \return the sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiRenderBuffer::setSampleCount(int s) + Sets the sample count to \a s. + */ + +/*! + \fn QRhiRenderBuffer::Flags QRhiRenderBuffer::flags() const + \return the flags. + */ + +/*! + \fn void QRhiRenderBuffer::setFlags(Flags f) + Sets the flags to \a f. + */ + +/*! + \fn virtual QRhiTexture::Format QRhiRenderBuffer::backingFormat() const = 0 \internal */ /*! \class QRhiTexture - \internal \inmodule QtGui + \since 6.6 \brief Texture resource. + + A QRhiTexture encapsulates a native texture object, such as a \c VkImage or + \c MTLTexture. + + A QRhiTexture instance is always created by calling + \l{QRhi::newTexture()}{the QRhi's newTexture() function}. This creates no + native graphics resources. To do that, call create() after setting the + appropriate options, such as the format and size, although in most cases + these are already set based on the arguments passed to + \l{QRhi::newTexture()}{newTexture()}. + + Setting the \l{QRhiTexture::Flags}{flags} correctly is essential, otherwise + various errors can occur depending on the underlying QRhi backend and + graphics API. For example, when a texture will be rendered into from a + render pass via QRhiTextureRenderTarget, the texture must be created with + the \l RenderTarget flag set. Similarly, when the texture is going to be + \l{QRhiResourceUpdateBatch::readBackTexture()}{read back}, the \l + UsedAsTransferSource flag must be set upfront. Mipmapped textures must have + the MipMapped flag set. And so on. It is not possible to change the flags + once create() has succeeded. To release the existing and create a new + native texture object with the changed settings, call the setters and call + create() again. This then might be a potentially expensive operation. + + \section2 Example usage + + To create a 2D texture with a size of 512x512 pixels and set its contents to all green: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512)); + if (!texture->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + QImage image(512, 512, QImage::Format_RGBA8888); + image.fill(Qt::green); + batch->uploadTexture(texture, image); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + \section2 Common patterns + + A call to create() destroys any existing native resources if create() was + successfully called before. If those native resources are still in use by + an in-flight frame (i.e., there's a chance they are still read by the GPU), + the destroying of those resources is deferred automatically. Thus a very + common and convenient pattern to safely change the size of an already + existing texture is the following. In practice this drops and creates a + whole new native texture resource underneath, so it is not necessarily a + cheap operation, but is more convenient and still faster than the + alternatives, because by not destroying the \c texture object itself, all + references to it stay valid in other data structures (e.g., in any + QShaderResourceBinding the QRhiTexture is referenced from). + + \code + // determine newSize, e.g. based on the swapchain's output size or other factors + if (texture->pixelSize() != newSize) { + texture->setPixelSize(newSize); + if (!texture->create()) { error(); } + } + // continue using texture, fill it with new data + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiResourceUpdateBatch, QRhi, QRhiTextureRenderTarget */ /*! @@ -2420,10 +4284,34 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) \value ThreeDimensional The texture is a 3D texture. Such textures should be created with the QRhi::newTexture() overload taking a depth in addition to width and height. A 3D texture can have mipmaps but cannot be - multisample. When rendering into a 3D texture, the layer specified in the - render target's color attachment refers to a slice in range [0..depth-1]. - The underlying graphics API may not support 3D textures at run time. - Support is indicated by the QRhi::ThreeDimensionalTextures feature. + multisample. When rendering into, or uploading data to a 3D texture, the \c + layer specified in the render target's color attachment or the upload + description refers to a single slice in range [0..depth-1]. The underlying + graphics API may not support 3D textures at run time. Support is indicated + by the QRhi::ThreeDimensionalTextures feature. + + \value TextureRectangleGL The texture should use the GL_TEXTURE_RECTANGLE + target with OpenGL. This flag is ignored with other graphics APIs. Just + like ExternalOES, this flag is useful when working with platform APIs where + native OpenGL texture objects received from the platform are wrapped in a + QRhiTexture, and the platform can only provide textures for a non-2D + texture target. + + \value TextureArray The texture is a texture array, i.e. a single texture + object that is a homogeneous array of 2D textures. Texture arrays are + created with QRhi::newTextureArray(). The underlying graphics API may not + support texture array objects at run time. Support is indicated by the + QRhi::TextureArrays feature. When rendering into, or uploading data to a + texture array, the \c layer specified in the render target's color + attachment or the upload description selects a single element in the array. + + \value OneDimensional The texture is a 1D texture. Such textures can be + created by passing a 0 height and depth to QRhi::newTexture(). Note that + there can be limitations on one dimensional textures depending on the + underlying graphics API. For example, rendering to them or using them with + mipmap-based filtering may be unsupported. This is indicated by the + QRhi::OneDimensionalTextures and QRhi::OneDimensionalTextureMipmaps + feature flags. */ /*! @@ -2453,6 +4341,19 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) \value RGBA32F Four components, 32-bit float per component. + \value R16F One component, 16-bit float. + + \value R32F One component, 32-bit float. + + \value RGB10A2 Four components, unsigned normalized 10 bit R, G, and B, + 2-bit alpha. This is a packed format so native endianness applies. Note + that there is no BGR10A2. This is because RGB10A2 maps to + DXGI_FORMAT_R10G10B10A2_UNORM with D3D, MTLPixelFormatRGB10A2Unorm with + Metal, VK_FORMAT_A2B10G10R10_UNORM_PACK32 with Vulkan, and + GL_RGB10_A2/GL_RGB/GL_UNSIGNED_INT_2_10_10_10_REV on OpenGL (ES). This is + the only universally supported RGB30 option. The corresponding QImage + formats are QImage::Format_BGR30 and QImage::Format_A2BGR30_Premultiplied. + \value D16 16-bit depth (normalized unsigned integer) \value D24 24-bit depth (normalized unsigned integer) @@ -2490,7 +4391,8 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) */ /*! - \class QRhiTexture::NativeTexture + \struct QRhiTexture::NativeTexture + \inmodule QtGui \brief Contains information about the underlying native resources of a texture. */ @@ -2499,9 +4401,10 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) \brief 64-bit integer containing the native object handle. With OpenGL, the native handle is a GLuint value, so \c object can then be - cast to a GLuint. With Vulkan, the native handle is a VkImage, so \c - object can be cast to a VkImage. With Direct3D 11 and Metal \c - object contains a ID3D11Texture2D or MTLTexture pointer, respectively. + cast to a GLuint. With Vulkan, the native handle is a VkImage, so \c object + can be cast to a VkImage. With Direct3D 11 and Metal \c object contains a + ID3D11Texture2D or MTLTexture pointer, respectively. With Direct3D 12 + \c object contains a ID3D12Resource pointer. */ /*! @@ -2515,9 +4418,10 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) \internal */ QRhiTexture::QRhiTexture(QRhiImplementation *rhi, Format format_, const QSize &pixelSize_, int depth_, - int sampleCount_, Flags flags_) + int arraySize_, int sampleCount_, Flags flags_) : QRhiResource(rhi), - m_format(format_), m_pixelSize(pixelSize_), m_depth(depth_), m_sampleCount(sampleCount_), m_flags(flags_) + m_format(format_), m_pixelSize(pixelSize_), m_depth(depth_), + m_arraySize(arraySize_), m_sampleCount(sampleCount_), m_flags(flags_) { } @@ -2530,7 +4434,7 @@ QRhiResource::Type QRhiTexture::resourceType() const } /*! - \fn bool QRhiTexture::create() + \fn virtual bool QRhiTexture::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2553,13 +4457,16 @@ QRhiTexture::NativeTexture QRhiTexture::nativeTexture() } /*! - Similar to create() except that no new native textures are created. Instead, - the native texture resources specified by \a src is used. + Similar to create(), except that no new native textures are created. + Instead, the native texture resources specified by \a src is used. This allows importing an existing native texture object (which must belong to the same device or sharing context, depending on the graphics API) from an external graphics engine. + \return true if the specified existing native texture object has been + successfully wrapped as a non-owning QRhiTexture. + \note format(), pixelSize(), sampleCount(), and flags() must still be set correctly. Passing incorrect sizes and other values to QRhi::newTexture() and then following it with a createFrom() expecting that the native texture @@ -2572,6 +4479,10 @@ QRhiTexture::NativeTexture QRhiTexture::nativeTexture() The opposite of this operation, exposing a QRhiTexture-created native texture object to a foreign engine, is possible via nativeTexture(). + \note When importing a 3D texture, or a texture array object, or, with + OpenGL ES, an external texture, it is then especially important to set the + corresponding flags (ThreeDimensional, TextureArray, ExternalOES) via + setFlags() before calling this function. */ bool QRhiTexture::createFrom(QRhiTexture::NativeTexture src) { @@ -2582,7 +4493,7 @@ bool QRhiTexture::createFrom(QRhiTexture::NativeTexture src) /*! With some graphics APIs, such as Vulkan, integrating custom rendering code that uses the graphics API directly needs special care when it comes to - image layouts. This function allows communicating the expected layout the + image layouts. This function allows communicating the expected \a layout the image backing the QRhiTexture is in after the native rendering commands. For example, consider rendering into a QRhiTexture's VkImage directly with @@ -2598,6 +4509,9 @@ bool QRhiTexture::createFrom(QRhiTexture::NativeTexture src) This function has no effect with QRhi backends where the underlying graphics API does not expose a concept of image layouts. + + \note With Vulkan \a layout is a \c VkImageLayout. With Direct 3D 12 \a + layout is a value composed of the bits from \c D3D12_RESOURCE_STATES. */ void QRhiTexture::setNativeLayout(int layout) { @@ -2605,10 +4519,196 @@ void QRhiTexture::setNativeLayout(int layout) } /*! + \fn QRhiTexture::Format QRhiTexture::format() const + \return the texture format. + */ + +/*! + \fn void QRhiTexture::setFormat(QRhiTexture::Format fmt) + + Sets the requested texture format to \a fmt. + + \note The value set is only taken into account upon the next call to + create(), i.e. when the underlying graphics resource are (re)created. + Setting a new value is futile otherwise and must be avoided since it can + lead to inconsistent state. + */ + +/*! + \fn QSize QRhiTexture::pixelSize() const + \return the size in pixels. + */ + +/*! + \fn void QRhiTexture::setPixelSize(const QSize &sz) + + Sets the texture size, specified in pixels, to \a sz. + + \note The value set is only taken into account upon the next call to + create(), i.e. when the underlying graphics resource are (re)created. + Setting a new value is futile otherwise and must be avoided since it can + lead to inconsistent state. The same applies to all other setters as well. + */ + +/*! + \fn int QRhiTexture::depth() const + \return the depth for 3D textures. + */ + +/*! + \fn void QRhiTexture::setDepth(int depth) + Sets the \a depth for a 3D texture. + */ + +/*! + \fn int QRhiTexture::arraySize() const + \return the texture array size. + */ + +/*! + \fn void QRhiTexture::setArraySize(int arraySize) + Sets the texture \a arraySize. + */ + +/*! + \fn int QRhiTexture::arrayRangeStart() const + + \return the first array layer when setArrayRange() was called. + + \sa setArrayRange() + */ + +/*! + \fn int QRhiTexture::arrayRangeLength() const + + \return the exposed array range size when setArrayRange() was called. + + \sa setArrayRange() +*/ + +/*! + \fn void QRhiTexture::setArrayRange(int startIndex, int count) + + Normally all array layers are exposed and it is up to the shader to select + the layer via the third coordinate passed to the \c{texture()} GLSL + function when sampling the \c sampler2DArray. When QRhi::TextureArrayRange + is reported as supported, calling setArrayRange() before create() or + createFrom() requests selecting only the specified range, \a count elements + starting from \a startIndex. The shader logic can then be written with this + in mind. + + \sa QRhi::TextureArrayRange + */ + +/*! + \fn Flags QRhiTexture::flags() const + \return the texture flags. + */ + +/*! + \fn void QRhiTexture::setFlags(Flags f) + Sets the texture flags to \a f. + */ + +/*! + \fn int QRhiTexture::sampleCount() const + \return the sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiTexture::setSampleCount(int s) + Sets the sample count to \a s. + */ + +/*! + \struct QRhiTexture::ViewFormat + \inmodule QtGui + \since 6.8 + \brief Specifies the view format for reading or writing from or to the texture. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiTexture::ViewFormat::format + */ + +/*! + \variable QRhiTexture::ViewFormat::srgb + */ + +/*! + \fn QRhiTexture::ViewFormat QRhiTexture::readViewFormat() const + \since 6.8 + \return the view format used when sampling the texture. When not called, the view + format is assumed to be the same as format(). + */ + +/*! + \fn void QRhiTexture::setReadViewFormat(const ViewFormat &fmt) + \since 6.8 + + Sets the shader resource view format (or the format of the view used for + sampling the texture) to \a fmt. By default the same format (and sRGB-ness) + is used as the texture itself, and in most cases this function does not need + to be called. + + This setting is only taken into account when the \l TextureViewFormat + feature is reported as supported. + + \note This functionality is provided to allow "casting" between + non-sRGB and sRGB in order to get the shader reads perform, or not perform, + the implicit sRGB conversions. Other types of casting may or may not be + functional. + */ + +/*! + \fn QRhiTexture::ViewFormat QRhiTexture::writeViewFormat() const + \since 6.8 + \return the view format used when writing to the texture and when using it + with image load/store. When not called, the view format is assumed to be the + same as format(). + */ + +/*! + \fn void QRhiTexture::setWriteViewFormat(const ViewFormat &fmt) + \since 6.8 + + Sets the render target view format to \a fmt. By default the same format + (and sRGB-ness) is used as the texture itself, and in most cases this + function does not need to be called. + + One common use case for providing a write view format is working with + externally provided textures that, outside of our control, use an sRGB + format with 3D APIs such as Vulkan or Direct 3D, but the rendering engine is + already prepared to handle linearization and conversion to sRGB at the end + of its shading pipeline. In this case what is wanted when rendering into + such a texture is a render target view (e.g. VkImageView) that has the same, + but non-sRGB format. (if e.g. from an OpenXR implementation one gets a + VK_FORMAT_R8G8B8A8_SRGB texture, it is likely that rendering into it should + be done using a VK_FORMAT_R8G8B8A8_UNORM view, if that is what the rendering + engine's pipeline requires; in this example one would call this function + with a ViewFormat that has a format of QRhiTexture::RGBA8 and \c srgb set to + \c false). + + This setting is only taken into account when the \l TextureViewFormat + feature is reported as supported. + + \note This functionality is provided to allow "casting" between + non-sRGB and sRGB in order to get the shader write not perform, or perform, + the implicit sRGB conversions. Other types of casting may or may not be + functional. + */ + +/*! \class QRhiSampler - \internal \inmodule QtGui + \since 6.6 \brief Sampler resource. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2665,14 +4765,91 @@ QRhiResource::Type QRhiSampler::resourceType() const } /*! + \fn QRhiSampler::Filter QRhiSampler::magFilter() const + \return the magnification filter mode. + */ + +/*! + \fn void QRhiSampler::setMagFilter(Filter f) + Sets the magnification filter mode to \a f. + */ + +/*! + \fn QRhiSampler::Filter QRhiSampler::minFilter() const + \return the minification filter mode. + */ + +/*! + \fn void QRhiSampler::setMinFilter(Filter f) + Sets the minification filter mode to \a f. + */ + +/*! + \fn QRhiSampler::Filter QRhiSampler::mipmapMode() const + \return the mipmap filter mode. + */ + +/*! + \fn void QRhiSampler::setMipmapMode(Filter f) + + Sets the mipmap filter mode to \a f. + + Leave this set to None when the texture has no mip levels, or when the mip + levels are not to be taken into account. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressU() const + \return the horizontal wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressU(AddressMode mode) + Sets the horizontal wrap \a mode. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressV() const + \return the vertical wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressV(AddressMode mode) + Sets the vertical wrap \a mode. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressW() const + \return the depth wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressW(AddressMode mode) + Sets the depth wrap \a mode. + */ + +/*! + \fn QRhiSampler::CompareOp QRhiSampler::textureCompareOp() const + \return the texture comparison function. + */ + +/*! + \fn void QRhiSampler::setTextureCompareOp(CompareOp op) + Sets the texture comparison function \a op. + */ + +/*! \class QRhiRenderPassDescriptor - \internal \inmodule QtGui + \since 6.6 \brief Render pass resource. A render pass, if such a concept exists in the underlying graphics API, is a collection of attachments (color, depth, stencil) and describes how those attachments are used. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2692,7 +4869,7 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const } /*! - \fn bool QRhiRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const + \fn virtual bool QRhiRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const = 0 \return true if the \a other QRhiRenderPassDescriptor is compatible with this one, meaning \c this and \a other can be used interchangebly in @@ -2713,11 +4890,21 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const \l{QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor()}{created} from the same QRhiTextureRenderTarget are always compatible. - \sa newCompatibleRenderPassDescriptor() + Similarly to QRhiShaderResourceBindings, compatibility can also be tested + without having two existing objects available. Extracting the opaque blob by + calling serializedFormat() allows testing for compatibility by comparing the + returned vector to another QRhiRenderPassDescriptor's + serializedFormat(). This has benefits in certain situations, because it + allows testing the compatibility of a QRhiRenderPassDescriptor with a + QRhiGraphicsPipeline even when the QRhiRenderPassDescriptor the pipeline was + originally built was is no longer available (but the data returned from its + serializedFormat() still is). + + \sa newCompatibleRenderPassDescriptor(), serializedFormat() */ /*! - \fn QRhiRenderPassDescriptor *QRhiRenderPassDescriptor::newCompatibleRenderPassDescriptor() const + \fn virtual QRhiRenderPassDescriptor *QRhiRenderPassDescriptor::newCompatibleRenderPassDescriptor() const = 0 \return a new QRhiRenderPassDescriptor that is \l{isCompatible()}{compatible} with this one. @@ -2738,6 +4925,24 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const */ /*! + \fn virtual QVector<quint32> QRhiRenderPassDescriptor::serializedFormat() const = 0 + + \return a vector of integers containing an opaque blob describing the data + relevant for \l{isCompatible()}{compatibility}. + + Given two QRhiRenderPassDescriptor objects \c rp1 and \c rp2, if the data + returned from this function is identical, then \c{rp1->isCompatible(rp2)}, + and vice versa hold true as well. + + \note The returned data is meant to be used for storing in memory and + comparisons during the lifetime of the QRhi the object belongs to. It is not + meant for storing on disk, reusing between processes, or using with multiple + QRhi instances with potentially different backends. + + \sa isCompatible() + */ + +/*! \return a pointer to a backend-specific QRhiNativeHandles subclass, such as QRhiVulkanRenderPassNativeHandles. The returned value is \nullptr when exposing the underlying native resources is not supported by the backend. @@ -2751,9 +4956,21 @@ const QRhiNativeHandles *QRhiRenderPassDescriptor::nativeHandles() /*! \class QRhiRenderTarget - \internal \inmodule QtGui + \since 6.6 \brief Represents an onscreen (swapchain) or offscreen (texture) render target. + + Applications do not create an instance of this class directly. Rather, it + is the subclass QRhiTextureRenderTarget that is instantiable by clients of + the API via \l{QRhi::newTextureRenderTarget()}{newTextureRenderTarget()}. + The other subclass is QRhiSwapChainRenderTarget, which is the type + QRhiSwapChain returns when calling + \l{QRhiSwapChain::currentFrameRenderTarget()}{currentFrameRenderTarget()}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiSwapChainRenderTarget, QRhiTextureRenderTarget */ /*! @@ -2765,21 +4982,25 @@ QRhiRenderTarget::QRhiRenderTarget(QRhiImplementation *rhi) } /*! - \return the resource type. - */ -QRhiResource::Type QRhiRenderTarget::resourceType() const -{ - return RenderTarget; -} - -/*! - \fn QSize QRhiRenderTarget::pixelSize() const + \fn virtual QSize QRhiRenderTarget::pixelSize() const = 0 \return the size in pixels. + + Valid only after create() has been called successfully. Until then the + result is a default-constructed QSize. + + With QRhiTextureRenderTarget the returned size is the size of the + associated attachments at the time of create(), in practice the size of the + first color attachment, or the depth/stencil buffer if there are no color + attachments. If the associated textures or renderbuffers are resized and + rebuilt afterwards, then pixelSize() performs an implicit call to create() + in order to rebuild the underlying data structures. This implicit check is + similar to what QRhiCommandBuffer::beginPass() does, and ensures that the + returned size is always up-to-date. */ /*! - \fn float QRhiRenderTarget::devicePixelRatio() const + \fn virtual float QRhiRenderTarget::devicePixelRatio() const = 0 \return the device pixel ratio. For QRhiTextureRenderTarget this is always 1. For targets retrieved from a QRhiSwapChain the value reflects the @@ -2788,29 +5009,94 @@ QRhiResource::Type QRhiRenderTarget::resourceType() const */ /*! - \class QRhiTextureRenderTarget + \fn virtual int QRhiRenderTarget::sampleCount() const = 0 + + \return the sample count or 1 if multisample antialiasing is not relevant for + this render target. + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiRenderTarget::renderPassDescriptor() const + + \return the associated QRhiRenderPassDescriptor. + */ + +/*! + \fn void QRhiRenderTarget::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + + Sets the QRhiRenderPassDescriptor \a desc for use with this render target. + */ + +/*! \internal + */ +QRhiSwapChainRenderTarget::QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain_) + : QRhiRenderTarget(rhi), + m_swapchain(swapchain_) +{ +} + +/*! + \class QRhiSwapChainRenderTarget + \inmodule QtGui + \since 6.6 + \brief Swapchain render target resource. + + When targeting the color buffers of a swapchain, active render target is a + QRhiSwapChainRenderTarget. This is what + QRhiSwapChain::currentFrameRenderTarget() returns. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiSwapChain + */ + +/*! + \return the resource type. + */ +QRhiResource::Type QRhiSwapChainRenderTarget::resourceType() const +{ + return SwapChainRenderTarget; +} + +/*! + \fn QRhiSwapChain *QRhiSwapChainRenderTarget::swapChain() const + + \return the swapchain object. + */ + +/*! + \class QRhiTextureRenderTarget \inmodule QtGui + \since 6.6 \brief Texture render target resource. A texture render target allows rendering into one or more textures, optionally with a depth texture or depth/stencil renderbuffer. + For multisample rendering the common approach is to use a renderbuffer as + the color attachment and set the non-multisample destination texture as the + \c{resolve texture}. + \note Textures used in combination with QRhiTextureRenderTarget must be created with the QRhiTexture::RenderTarget flag. The simplest example of creating a render target with a texture as its single color attachment: - \badcode - texture = rhi->newTexture(QRhiTexture::RGBA8, size, 1, QRhiTexture::RenderTarget); + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, size, 1, QRhiTexture::RenderTarget); texture->create(); - rt = rhi->newTextureRenderTarget({ texture }); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ texture }); rp = rt->newCompatibleRenderPassDescriptor(); rt->setRenderPassDescriptor(rt); rt->create(); // rt can now be used with beginPass() \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2825,7 +5111,19 @@ QRhiResource::Type QRhiRenderTarget::resourceType() const \value PreserveColorContents Indicates that the contents of the color attachments is to be loaded when starting a render pass, instead of clearing. This is potentially more expensive, especially on mobile (tiled) - GPUs, but allows preserving the existing contents between passes. + GPUs, but allows preserving the existing contents between passes. When doing + multisample rendering with a resolve texture set, setting this flag also + requests the multisample color data to be stored (written out) to the + multisample texture or render buffer. (for non-multisample rendering the + color data is always stored, but for MSAA storing the multisample data + decreases efficiency for certain GPU architectures, hence defaulting to not + writing it out) Note however that this is non-portable: in some cases there + is no intermediate multisample texture on the graphics API level, e.g. when + using OpenGL ES's \c{GL_EXT_multisampled_render_to_texture} as it is all + implicit, handled by the OpenGL ES implementation. In that case, + PreserveColorContents will likely have no effect. Therefore, avoid relying + on this flag when using multisample rendering and the color attachment is + using a multisample QRhiTexture (not QRhiRenderBuffer). \value PreserveDepthStencilContents Indicates that the contents of the depth texture is to be loaded when starting a render pass, instead @@ -2833,6 +5131,13 @@ QRhiResource::Type QRhiRenderTarget::resourceType() const (QRhiTextureRenderTargetDescription::depthTexture() is set) because depth/stencil renderbuffers may not have any physical backing and data may not be written out in the first place. + + \value DoNotStoreDepthStencilContents Indicates that the contents of the + depth texture does not need to be written out. Relevant only when a + QRhiTexture, not QRhiRenderBuffer, is used as the depth-stencil buffer, + because for QRhiRenderBuffer this is implicit. When a depthResolveTexture is + set, the flag is not relevant, because the behavior is then as if the flag + was set. This enum value is introduced in Qt 6.8. */ /*! @@ -2856,7 +5161,7 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const } /*! - \fn QRhiRenderPassDescriptor *QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor() + \fn virtual QRhiRenderPassDescriptor *QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor() = 0 \return a new QRhiRenderPassDescriptor that is compatible with this render target. @@ -2866,14 +5171,15 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const QRhiGraphicsPipeline::setRenderPassDescriptor(). A render pass descriptor describes the attachments (color, depth/stencil) and the load/store behavior that can be affected by flags(). A QRhiGraphicsPipeline can only - be used in combination with a render target that has the same + be used in combination with a render target that has a + \l{QRhiRenderPassDescriptor::isCompatible()}{compatible} QRhiRenderPassDescriptor set. Two QRhiTextureRenderTarget instances can share the same render pass descriptor as long as they have the same number and type of attachments. The associated QRhiTexture or QRhiRenderBuffer instances are not part of the render pass descriptor so those can differ in the two - QRhiTextureRenderTarget intances. + QRhiTextureRenderTarget instances. \note resources, such as QRhiTexture instances, referenced in description() must already have create() called on them. @@ -2882,7 +5188,7 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const */ /*! - \fn bool QRhiTextureRenderTarget::create() + \fn virtual bool QRhiTextureRenderTarget::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2906,9 +5212,29 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const */ /*! + \fn QRhiTextureRenderTargetDescription QRhiTextureRenderTarget::description() const + \return the render target description. + */ + +/*! + \fn void QRhiTextureRenderTarget::setDescription(const QRhiTextureRenderTargetDescription &desc) + Sets the render target description \a desc. + */ + +/*! + \fn QRhiTextureRenderTarget::Flags QRhiTextureRenderTarget::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiTextureRenderTarget::setFlags(Flags f) + Sets the flags to \a f. + */ + +/*! \class QRhiShaderResourceBindings - \internal \inmodule QtGui + \since 6.6 \brief Encapsulates resources for making buffer, texture, sampler resources visible to shaders. A QRhiShaderResourceBindings is a collection of QRhiShaderResourceBinding @@ -2929,19 +5255,19 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const QRhiShaderResourceBindings could be created and then passed to QRhiGraphicsPipeline::setShaderResourceBindings(): - \badcode - srb = rhi->newShaderResourceBindings(); + \code + QRhiShaderResourceBindings *srb = rhi->newShaderResourceBindings(); srb->setBindings({ QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage | QRhiShaderResourceBinding::FragmentStage, ubuf), QRhiShaderResourceBinding::sampledTexture(1, QRhiShaderResourceBinding::FragmentStage, texture, sampler) }); srb->create(); - ... - ps = rhi->newGraphicsPipeline(); - ... + // ... + QRhiGraphicsPipeline *ps = rhi->newGraphicsPipeline(); + // ... ps->setShaderResourceBindings(srb); ps->create(); - ... + // ... cb->setGraphicsPipeline(ps); cb->setShaderResources(); // binds srb \endcode @@ -2960,22 +5286,34 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const srb argument. As long as the layouts (so the number of bindings and the binding points) match between two QRhiShaderResourceBindings, they can both be used with the same pipeline, assuming the pipeline was created with one of - them in the first place. + them in the first place. See isLayoutCompatible() for more details. - \badcode - srb2 = rhi->newShaderResourceBindings(); - ... + \code + QRhiShaderResourceBindings *srb2 = rhi->newShaderResourceBindings(); + // ... cb->setGraphicsPipeline(ps); cb->setShaderResources(srb2); // binds srb2 \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \typedef QRhiShaderResourceBindingSet + \relates QRhi + \since 6.7 + + Synonym for QRhiShaderResourceBindings. +*/ + +/*! \internal */ QRhiShaderResourceBindings::QRhiShaderResourceBindings(QRhiImplementation *rhi) : QRhiResource(rhi) { + m_layoutDesc.reserve(BINDING_PREALLOC * QRhiShaderResourceBinding::LAYOUT_DESC_ENTRIES_PER_BINDING); } /*! @@ -2998,8 +5336,14 @@ QRhiResource::Type QRhiShaderResourceBindings::resourceType() const then safely be passed to QRhiCommandBuffer::setShaderResources(), and so be used with the pipeline in place of this QRhiShaderResourceBindings. - This function can be called before create() as well. The bindings must - already be set via setBindings() however. + \note This function must only be called after a successful create(), because + it relies on data generated during the baking of the underlying data + structures. This way the function can implement a comparison approach that + is more efficient than iterating through two binding lists and calling + QRhiShaderResourceBinding::isLayoutCompatible() on each pair. This becomes + relevant especially when this function is called at a high frequency. + + \sa serializedLayoutDescription() */ bool QRhiShaderResourceBindings::isLayoutCompatible(const QRhiShaderResourceBindings *other) const { @@ -3018,26 +5362,80 @@ bool QRhiShaderResourceBindings::isLayoutCompatible(const QRhiShaderResourceBind && m_layoutDesc == other->m_layoutDesc; } +/*! + \fn QVector<quint32> QRhiShaderResourceBindings::serializedLayoutDescription() const + + \return a vector of integers containing an opaque blob describing the layout + of the binding list, i.e. the data relevant for + \l{isLayoutCompatible()}{layout compatibility tests}. + + Given two objects \c srb1 and \c srb2, if the data returned from this + function is identical, then \c{srb1->isLayoutCompatible(srb2)}, and vice + versa hold true as well. + + \note The returned data is meant to be used for storing in memory and + comparisons during the lifetime of the QRhi the object belongs to. It is not + meant for storing on disk, reusing between processes, or using with multiple + QRhi instances with potentially different backends. + + \sa isLayoutCompatible() + */ + void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) { srb->m_layoutDescHash = 0; srb->m_layoutDesc.clear(); - for (const QRhiShaderResourceBinding &b : qAsConst(srb->m_bindings)) { - const QRhiShaderResourceBinding::Data *d = b.data(); - // must match QRhiShaderResourceBinding::isLayoutCompatible() - srb->m_layoutDescHash ^= uint(d->binding) ^ uint(d->stage) ^ uint(d->type); - srb->m_layoutDesc << uint(d->binding) << uint(d->stage) << uint(d->type); + auto layoutDescAppender = std::back_inserter(srb->m_layoutDesc); + for (const QRhiShaderResourceBinding &b : std::as_const(srb->m_bindings)) { + const QRhiShaderResourceBinding::Data *d = &b.d; + srb->m_layoutDescHash ^= uint(d->binding) ^ uint(d->stage) ^ uint(d->type) + ^ uint(d->arraySize()); + layoutDescAppender = d->serialize(layoutDescAppender); } } /*! + \fn void QRhiShaderResourceBindings::setBindings(std::initializer_list<QRhiShaderResourceBinding> list) + Sets the \a list of bindings. + */ + +/*! + \fn template<typename InputIterator> void QRhiShaderResourceBindings::setBindings(InputIterator first, InputIterator last) + Sets the list of bindings from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::cbeginBindings() const + \return a const iterator pointing to the first item in the binding list. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::cendBindings() const + \return a const iterator pointing just after the last item in the binding list. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::bindingAt(qsizetype index) const + \return the binding at the specified \a index. + */ + +/*! + \fn qsizetype QRhiShaderResourceBindings::bindingCount() const + \return the number of bindings. + */ + +/*! \class QRhiShaderResourceBinding - \internal \inmodule QtGui + \since 6.6 \brief Describes the shader resource for a single binding point. - A QRhiShaderResourceBinding cannot be constructed directly. Instead, use - the static functions uniformBuffer(), sampledTexture() to get an instance. + A QRhiShaderResourceBinding cannot be constructed directly. Instead, use the + static functions such as uniformBuffer() or sampledTexture() to get an + instance. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -3046,7 +5444,15 @@ void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) \value UniformBuffer Uniform buffer - \value SampledTexture Combined image sampler + \value SampledTexture Combined image sampler (a texture and sampler pair). + Even when the shading language associated with the underlying 3D API has no + support for this concept (e.g. D3D and HLSL), this is still supported + because the shader translation layer takes care of the appropriate + translation and remapping of binding points or shader registers. + + \value Texture Texture (separate) + + \value Sampler Sampler (separate) \value ImageLoad Image load (with GLSL this maps to doing imageLoad() on a single level - and either one or all layers - of a texture exposed to the @@ -3072,8 +5478,11 @@ void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) Flag values to indicate which stages the shader resource is visible in \value VertexStage Vertex stage - \value FragmentStage Fragment (pixel) stage + \value TessellationControlStage Tessellation control (hull shader) stage + \value TessellationEvaluationStage Tessellation evaluation (domain shader) stage + \value FragmentStage Fragment (pixel shader) stage \value ComputeStage Compute stage + \value GeometryStage Geometry stage */ /*! @@ -3083,14 +5492,18 @@ void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) For example, \c a and \c b below are not equal, but are compatible layout-wise: - \badcode + \code auto a = QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage, buffer); auto b = QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage, someOtherBuffer, 256); \endcode */ bool QRhiShaderResourceBinding::isLayoutCompatible(const QRhiShaderResourceBinding &other) const { - return d.binding == other.d.binding && d.stage == other.d.stage && d.type == other.d.type; + // everything that goes into a VkDescriptorSetLayoutBinding must match + return d.binding == other.d.binding + && d.stage == other.d.stage + && d.type == other.d.type + && d.arraySize() == other.d.arraySize(); } /*! @@ -3106,6 +5519,9 @@ bool QRhiShaderResourceBinding::isLayoutCompatible(const QRhiShaderResourceBindi suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note If the size of \a buf exceeds the limit reported for + QRhi::MaxUniformBufferRange, unexpected errors may occur. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBuffer( int binding, StageFlags stage, QRhiBuffer *buf) @@ -3140,9 +5556,12 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBuffer( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note If \a size exceeds the limit reported for QRhi::MaxUniformBufferRange, + unexpected errors may occur. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBuffer( - int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size) + int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size) { Q_ASSERT(size > 0); QRhiShaderResourceBinding b; @@ -3174,9 +5593,12 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBuffer( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note If \a size exceeds the limit reported for QRhi::MaxUniformBufferRange, + unexpected errors may occur. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBufferWithDynamicOffset( - int binding, StageFlags stage, QRhiBuffer *buf, int size) + int binding, StageFlags stage, QRhiBuffer *buf, quint32 size) { Q_ASSERT(size > 0); QRhiShaderResourceBinding b; @@ -3205,6 +5627,14 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::uniformBufferWithDynamicOff together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + \note A shader may not be able to consume more than 16 textures/samplers, + depending on the underlying graphics API. This hard limit must be kept in + mind in renderer design. This does not apply to texture arrays which + consume a single binding point (shader register) and can contain 256-2048 + textures, depending on the underlying graphics API. Arrays of textures (see + sampledTextures()) are however no different in this regard than using the + same number of individual textures. + \sa sampledTextures() */ QRhiShaderResourceBinding QRhiShaderResourceBinding::sampledTexture( @@ -3215,8 +5645,7 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::sampledTexture( b.d.stage = stage; b.d.type = SampledTexture; b.d.u.stex.count = 1; - b.d.u.stex.texSamplers[0].tex = tex; - b.d.u.stex.texSamplers[0].sampler = sampler; + b.d.u.stex.texSamplers[0] = { tex, sampler }; return b; } @@ -3267,12 +5696,128 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::sampledTextures( if (texSamplers) b.d.u.stex.texSamplers[i] = texSamplers[i]; else - b.d.u.stex.texSamplers[i] = {}; + b.d.u.stex.texSamplers[i] = { nullptr, nullptr }; + } + return b; +} + +/*! + \return a shader resource binding for the given binding number, pipeline + stages, and texture specified by \a binding, \a stage, \a tex. + + \note This function is equivalent to calling textures() with a + \c count of 1. + + \note \a tex can be null. It is valid to create a + QRhiShaderResourceBindings with unspecified resources, but such an object + cannot be used with QRhiCommandBuffer::setShaderResources(). It is however + suitable for creating pipelines. Such a pipeline must then always be used + together with another, layout compatible QRhiShaderResourceBindings with + resources present passed to QRhiCommandBuffer::setShaderResources(). + + This creates a binding for a separate texture (image) object, whereas + sampledTexture() is suitable for combined image samplers. In + Vulkan-compatible GLSL code separate textures are declared as \c texture2D + as opposed to \c sampler2D: \c{layout(binding = 1) uniform texture2D tex;} + + \note A shader may not be able to consume more than 16 textures, depending + on the underlying graphics API. This hard limit must be kept in mind in + renderer design. This does not apply to texture arrays which consume a + single binding point (shader register) and can contain 256-2048 textures, + depending on the underlying graphics API. Arrays of textures (see + sampledTextures()) are however no different in this regard than using the + same number of individual textures. + + \sa textures(), sampler() + */ +QRhiShaderResourceBinding QRhiShaderResourceBinding::texture(int binding, StageFlags stage, QRhiTexture *tex) +{ + QRhiShaderResourceBinding b; + b.d.binding = binding; + b.d.stage = stage; + b.d.type = Texture; + b.d.u.stex.count = 1; + b.d.u.stex.texSamplers[0] = { tex, nullptr }; + return b; +} + +/*! + \return a shader resource binding for the given binding number, pipeline + stages, and the array of (separate) textures specified by \a binding, \a + stage, \a count, and \a tex. + + \note \a count must be at least 1, and not larger than 16. + + \note When \a count is 1, this function is equivalent to texture(). + + \warning All elements of the array must be specified. + + \note \a tex can be null. It is valid to create a + QRhiShaderResourceBindings with unspecified resources, but such an object + cannot be used with QRhiCommandBuffer::setShaderResources(). It is however + suitable for creating pipelines. Such a pipeline must then always be used + together with another, layout compatible QRhiShaderResourceBindings with + resources present passed to QRhiCommandBuffer::setShaderResources(). + + \sa texture(), sampler() + */ +QRhiShaderResourceBinding QRhiShaderResourceBinding::textures(int binding, StageFlags stage, int count, QRhiTexture **tex) +{ + Q_ASSERT(count >= 1 && count <= Data::MAX_TEX_SAMPLER_ARRAY_SIZE); + QRhiShaderResourceBinding b; + b.d.binding = binding; + b.d.stage = stage; + b.d.type = Texture; + b.d.u.stex.count = count; + for (int i = 0; i < count; ++i) { + if (tex) + b.d.u.stex.texSamplers[i] = { tex[i], nullptr }; + else + b.d.u.stex.texSamplers[i] = { nullptr, nullptr }; } return b; } /*! + \return a shader resource binding for the given binding number, pipeline + stages, and sampler specified by \a binding, \a stage, \a sampler. + + \note \a sampler can be null. It is valid to create a + QRhiShaderResourceBindings with unspecified resources, but such an object + cannot be used with QRhiCommandBuffer::setShaderResources(). It is however + suitable for creating pipelines. Such a pipeline must then always be used + together with another, layout compatible QRhiShaderResourceBindings with + resources present passed to QRhiCommandBuffer::setShaderResources(). + + Arrays of separate samplers are not supported. + + This creates a binding for a separate sampler object, whereas + sampledTexture() is suitable for combined image samplers. In + Vulkan-compatible GLSL code separate samplers are declared as \c sampler + as opposed to \c sampler2D: \c{layout(binding = 2) uniform sampler samp;} + + With both a \c texture2D and \c sampler present, they can be used together + to sample the texture: \c{fragColor = texture(sampler2D(tex, samp), + texcoord);}. + + \note A shader may not be able to consume more than 16 samplers, depending + on the underlying graphics API. This hard limit must be kept in mind in + renderer design. + + \sa texture() + */ +QRhiShaderResourceBinding QRhiShaderResourceBinding::sampler(int binding, StageFlags stage, QRhiSampler *sampler) +{ + QRhiShaderResourceBinding b; + b.d.binding = binding; + b.d.stage = stage; + b.d.type = Sampler; + b.d.u.stex.count = 1; + b.d.u.stex.texSamplers[0] = { nullptr, sampler }; + return b; +} + +/*! \return a shader resource binding for a read-only storage image with the given \a binding number and pipeline \a stage. The image load operations will have access to all layers of the specified \a level. (so if the texture @@ -3287,6 +5832,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::sampledTextures( pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Image load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageLoad( int binding, StageFlags stage, QRhiTexture *tex, int level) @@ -3315,6 +5867,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageLoad( pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Image load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageStore( int binding, StageFlags stage, QRhiTexture *tex, int level) @@ -3343,6 +5902,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageStore( pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Image load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageLoadStore( int binding, StageFlags stage, QRhiTexture *tex, int level) @@ -3369,6 +5935,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::imageLoadStore( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoad( int binding, StageFlags stage, QRhiBuffer *buf) @@ -3397,9 +5970,16 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoad( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoad( - int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size) + int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size) { Q_ASSERT(size > 0); QRhiShaderResourceBinding b; @@ -3425,6 +6005,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoad( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferStore( int binding, StageFlags stage, QRhiBuffer *buf) @@ -3453,9 +6040,16 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferStore( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferStore( - int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size) + int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size) { Q_ASSERT(size > 0); QRhiShaderResourceBinding b; @@ -3481,6 +6075,13 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferStore( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoadStore( int binding, StageFlags stage, QRhiBuffer *buf) @@ -3509,9 +6110,16 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoadStore( suitable for creating pipelines. Such a pipeline must then always be used together with another, layout compatible QRhiShaderResourceBindings with resources present passed to QRhiCommandBuffer::setShaderResources(). + + \note Buffer load/store is only guaranteed to be available within a compute + pipeline. While some backends may support using these resources in a + graphics pipeline as well, this is not universally supported, and even when + it is, unexpected problems may arise when it comes to barriers and + synchronization. Therefore, avoid using such resources with shaders other + than compute. */ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoadStore( - int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size) + int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size) { Q_ASSERT(size > 0); QRhiShaderResourceBinding b; @@ -3536,8 +6144,8 @@ QRhiShaderResourceBinding QRhiShaderResourceBinding::bufferLoadStore( */ bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept { - const QRhiShaderResourceBinding::Data *da = a.data(); - const QRhiShaderResourceBinding::Data *db = b.data(); + const QRhiShaderResourceBinding::Data *da = QRhiImplementation::shaderResourceBindingData(a); + const QRhiShaderResourceBinding::Data *db = QRhiImplementation::shaderResourceBindingData(b); if (da == db) return true; @@ -3570,10 +6178,20 @@ bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBind } } break; + case QRhiShaderResourceBinding::Texture: + if (da->u.stex.count != db->u.stex.count) + return false; + for (int i = 0; i < da->u.stex.count; ++i) { + if (da->u.stex.texSamplers[i].tex != db->u.stex.texSamplers[i].tex) + return false; + } + break; + case QRhiShaderResourceBinding::Sampler: + if (da->u.stex.texSamplers[0].sampler != db->u.stex.texSamplers[0].sampler) + return false; + break; case QRhiShaderResourceBinding::ImageLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageLoadStore: if (da->u.simage.tex != db->u.simage.tex || da->u.simage.level != db->u.simage.level) @@ -3582,9 +6200,7 @@ bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBind } break; case QRhiShaderResourceBinding::BufferLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferLoadStore: if (da->u.sbuf.buf != db->u.sbuf.buf || da->u.sbuf.offset != db->u.sbuf.offset @@ -3594,8 +6210,7 @@ bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBind } break; default: - Q_UNREACHABLE(); - return false; + Q_UNREACHABLE_RETURN(false); } return true; @@ -3619,41 +6234,44 @@ bool operator!=(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBind */ size_t qHash(const QRhiShaderResourceBinding &b, size_t seed) noexcept { - const QRhiShaderResourceBinding::Data *d = b.data(); - size_t h = uint(d->binding) ^ uint(d->stage) ^ uint(d->type) ^ seed; + const QRhiShaderResourceBinding::Data *d = QRhiImplementation::shaderResourceBindingData(b); + QtPrivate::QHashCombine hash; + seed = hash(seed, d->binding); + seed = hash(seed, d->stage); + seed = hash(seed, d->type); switch (d->type) { case QRhiShaderResourceBinding::UniformBuffer: - h ^= qHash(reinterpret_cast<quintptr>(d->u.ubuf.buf)); + seed = hash(seed, reinterpret_cast<quintptr>(d->u.ubuf.buf)); break; case QRhiShaderResourceBinding::SampledTexture: - h ^= qHash(reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].tex)); - h ^= qHash(reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].sampler)); + seed = hash(seed, reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].tex)); + seed = hash(seed, reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].sampler)); + break; + case QRhiShaderResourceBinding::Texture: + seed = hash(seed, reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].tex)); + break; + case QRhiShaderResourceBinding::Sampler: + seed = hash(seed, reinterpret_cast<quintptr>(d->u.stex.texSamplers[0].sampler)); break; case QRhiShaderResourceBinding::ImageLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageLoadStore: - h ^= qHash(reinterpret_cast<quintptr>(d->u.simage.tex)); + seed = hash(seed, reinterpret_cast<quintptr>(d->u.simage.tex)); break; case QRhiShaderResourceBinding::BufferLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferLoadStore: - h ^= qHash(reinterpret_cast<quintptr>(d->u.sbuf.buf)); - break; - default: + seed = hash(seed, reinterpret_cast<quintptr>(d->u.sbuf.buf)); break; } - return h; + return seed; } #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiShaderResourceBinding &b) { QDebugStateSaver saver(dbg); - const QRhiShaderResourceBinding::Data *d = b.data(); + const QRhiShaderResourceBinding::Data *d = QRhiImplementation::shaderResourceBindingData(b); dbg.nospace() << "QRhiShaderResourceBinding(" << "binding=" << d->binding << " stage=" << d->stage @@ -3675,6 +6293,18 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBinding &b) } dbg.nospace() << ')'; break; + case QRhiShaderResourceBinding::Texture: + dbg.nospace() << " Textures(" + << "count=" << d->u.stex.count; + for (int i = 0; i < d->u.stex.count; ++i) + dbg.nospace() << " texture=" << d->u.stex.texSamplers[i].tex; + dbg.nospace() << ')'; + break; + case QRhiShaderResourceBinding::Sampler: + dbg.nospace() << " Sampler(" + << " sampler=" << d->u.stex.texSamplers[0].sampler + << ')'; + break; case QRhiShaderResourceBinding::ImageLoad: dbg.nospace() << " ImageLoad(" << "texture=" << d->u.simage.tex @@ -3736,18 +6366,44 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) /*! \class QRhiGraphicsPipeline - \internal \inmodule QtGui + \since 6.6 \brief Graphics pipeline state resource. + Represents a graphics pipeline. What exactly this map to in the underlying + native graphics API, varies. Where there is a concept of pipeline objects, + for example with Vulkan, the QRhi backend will create such an object upon + calling create(). Elsewhere, for example with OpenGL, the + QRhiGraphicsPipeline may merely collect the various state, and create()'s + main task is to set up the corresponding shader program, but deferring + looking at any of the requested state to a later point. + + As with all QRhiResource subclasses, the two-phased initialization pattern + applies: setting any values via the setters, for example setDepthTest(), is + only effective after calling create(). Avoid changing any values once the + QRhiGraphicsPipeline has been initialized via create(). To change some + state, set the new value and call create() again. However, that will + effectively release all underlying native resources and create new ones. As + a result, it may be a heavy, expensive operation. Rather, prefer creating + multiple pipelines with the different states, and + \l{QRhiCommandBuffer::setGraphicsPipeline()}{switch between them} when + recording the render pass. + \note Setting the shader stages is mandatory. There must be at least one stage, and there must be a vertex stage. \note Setting the shader resource bindings is mandatory. The referenced QRhiShaderResourceBindings must already have create() called on it by the time create() is called. Associating with a QRhiShaderResourceBindings that - has no bindings is also valid, as long as no shader in any stage expects - any resources. + has no bindings is also valid, as long as no shader in any stage expects any + resources. Using a QRhiShaderResourceBindings object that does not specify + any actual resources (i.e., the buffers, textures, etc. for the binding + points are set to \nullptr) is valid as well, as long as a + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings, that specifies resources for all the bindings, + is going to be set via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} when + recording the render pass. \note Setting the render pass descriptor is mandatory. To obtain a QRhiRenderPassDescriptor that can be passed to setRenderPassDescriptor(), @@ -3760,20 +6416,49 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) render target's color and depth stencil attachments. \note The depth test, depth write, and stencil test are disabled by - default. + default. The face culling mode defaults to no culling. \note stencilReadMask() and stencilWriteMask() apply to both faces. They both default to 0xFF. - */ -/*! - \fn void QRhiGraphicsPipeline::setTargetBlends(const QList<TargetBlend> &blends) + \section2 Example usage + + All settings of a graphics pipeline have defaults which might be suitable + to many applications. Therefore a minimal example of creating a graphics + pipeline could be the following. This assumes that the vertex shader takes + a single \c{vec3 position} input at the input location 0. With the + QRhiShaderResourceBindings and QRhiRenderPassDescriptor objects, plus the + QShader collections for the vertex and fragment stages, a pipeline could be + created like this: - Sets the blend specification for color attachments. Each element in \a - blends corresponds to a color attachment of the render target. + \code + QRhiShaderResourceBindings *srb; + QRhiRenderPassDescriptor *rpDesc; + QShader vs, fs; + // ... + + QRhiVertexInputLayout inputLayout; + inputLayout.setBindings({ { 3 * sizeof(float) } }); + inputLayout.setAttributes({ { 0, 0, QRhiVertexInputAttribute::Float3, 0 } }); - By default no blends are set, which is a shortcut to disabling blending and - enabling color write for all four channels. + QRhiGraphicsPipeline *ps = rhi->newGraphicsPipeline(); + ps->setShaderStages({ { QRhiShaderStage::Vertex, vs }, { QRhiShaderStage::Fragment, fs } }); + ps->setVertexInputLayout(inputLayout); + ps->setShaderResourceBindings(srb); + ps->setRenderPassDescriptor(rpDesc); + if (!ps->create()) { error(); } + \endcode + + The above code creates a pipeline object that uses the defaults for many + settings and states. For example, it will use a \l Triangles topology, no + backface culling, blending is disabled but color write is enabled for all + four channels, depth test/write are disabled, stencil operations are + disabled. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiCommandBuffer, QRhi */ /*! @@ -3818,6 +6503,9 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) \value Lines \value LineStrip \value Points + + \value Patches (only available if QRhi::Tessellation is supported, and + requires the tessellation stages to be present in the pipeline) */ /*! @@ -3912,21 +6600,102 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) */ /*! - \class QRhiGraphicsPipeline::TargetBlend - \internal + \enum QRhiGraphicsPipeline::PolygonMode + \brief Specifies the polygon rasterization mode + + Polygon Mode (Triangle Fill Mode in Metal, Fill Mode in D3D) specifies + the fill mode used when rasterizing polygons. Polygons may be drawn as + solids (Fill), or as a wire mesh (Line). + + Support for non-fill polygon modes is optional and is indicated by the + QRhi::NonFillPolygonMode feature. With OpenGL ES and some Vulkan + implementations the feature will likely be reported as unsupported, which + then means values other than Fill cannot be used. + + \value Fill The interior of the polygon is filled (default) + \value Line Boundary edges of the polygon are drawn as line segments. + */ + +/*! + \struct QRhiGraphicsPipeline::TargetBlend \inmodule QtGui + \since 6.6 \brief Describes the blend state for one color attachment. Defaults to color write enabled, blending disabled. The blend values are set up for pre-multiplied alpha (One, OneMinusSrcAlpha, One, - OneMinusSrcAlpha) by default. + OneMinusSrcAlpha) by default. This means that to get the alpha blending + mode Qt Quick uses, it is enough to set the \c enable flag to true while + leaving other values at their defaults. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \class QRhiGraphicsPipeline::StencilOpState - \internal + \variable QRhiGraphicsPipeline::TargetBlend::colorWrite + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::enable + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::srcColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::dstColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::opColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::srcAlpha + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::dstAlpha + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::opAlpha + */ + +/*! + \struct QRhiGraphicsPipeline::StencilOpState \inmodule QtGui + \since 6.6 \brief Describes the stencil operation state. + + The default-constructed StencilOpState has the following set: + \list + \li failOp - \l Keep + \li depthFailOp - \l Keep + \li passOp - \l Keep + \li compareOp \l Always + \endlist + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::failOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::depthFailOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::passOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::compareOp */ /*! @@ -3946,7 +6715,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const } /*! - \fn bool QRhiGraphicsPipeline::create() + \fn virtual bool QRhiGraphicsPipeline::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -3954,23 +6723,132 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const \return \c true when successful, \c false when a graphics operation failed. Regardless of the return value, calling destroy() is always safe. + + \note This may be, depending on the underlying graphics API, an expensive + operation, especially when shaders get compiled/optimized from source or + from an intermediate bytecode format to the GPU's own instruction set. + Where applicable, the QRhi backend automatically sets up the relevant + non-persistent facilities to accelerate this, for example the Vulkan + backend automatically creates a \c VkPipelineCache to improve data reuse + during the lifetime of the application. + + \note Drivers may also employ various persistent (disk-based) caching + strategies for shader and pipeline data, which is hidden to and is outside + of Qt's control. In some cases, depending on the graphics API and the QRhi + backend, there are facilities within QRhi for manually managing such a + cache, allowing the retrieval of a serializable blob that can then be + reloaded in the future runs of the application to ensure faster pipeline + creation times. See QRhi::pipelineCacheData() and + QRhi::setPipelineCacheData() for details. Note also that when working with + a QRhi instance managed by a higher level Qt framework, such as Qt Quick, + it is possible that such disk-based caching is taken care of automatically, + for example QQuickWindow uses a disk-based pipeline cache by default (which + comes in addition to any driver-level caching). + */ + +/*! + \fn QRhiGraphicsPipeline::Flags QRhiGraphicsPipeline::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiGraphicsPipeline::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiGraphicsPipeline::Topology QRhiGraphicsPipeline::topology() const + \return the currently set primitive topology. + */ + +/*! + \fn void QRhiGraphicsPipeline::setTopology(Topology t) + Sets the primitive topology \a t. + */ + +/*! + \fn QRhiGraphicsPipeline::CullMode QRhiGraphicsPipeline::cullMode() const + \return the currently set face culling mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setCullMode(CullMode mode) + Sets the specified face culling \a mode. + */ + +/*! + \fn QRhiGraphicsPipeline::FrontFace QRhiGraphicsPipeline::frontFace() const + \return the currently set front face mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setFrontFace(FrontFace f) + Sets the front face mode \a f. + */ + +/*! + \fn void QRhiGraphicsPipeline::setTargetBlends(std::initializer_list<TargetBlend> list) + + Sets the \a list of render target blend settings. This is a list because + when multiple render targets are used (i.e., a QRhiTextureRenderTarget with + more than one QRhiColorAttachment), there needs to be a TargetBlend + structure per render target (color attachment). + + By default there is one default-constructed TargetBlend set. + + \sa QRhi::MaxColorAttachments + */ + +/*! + \fn template<typename InputIterator> void QRhiGraphicsPipeline::setTargetBlends(InputIterator first, InputIterator last) + Sets the list of render target blend settings from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::cbeginTargetBlends() const + \return a const iterator pointing to the first item in the render target blend setting list. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::cendTargetBlends() const + \return a const iterator pointing just after the last item in the render target blend setting list. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::targetBlendAt(qsizetype index) const + \return the render target blend setting at the specified \a index. + */ + +/*! + \fn qsizetype QRhiGraphicsPipeline::targetBlendCount() const + \return the number of render target blend settings. + */ + +/*! + \fn bool QRhiGraphicsPipeline::hasDepthTest() const + \return true if depth testing is enabled. */ /*! \fn void QRhiGraphicsPipeline::setDepthTest(bool enable) - Enables or disables depth testing. Both depth test and the writing out of - depth data are disabled by default. + Enables or disables depth testing based on \a enable. Both depth test and + the writing out of depth data are disabled by default. \sa setDepthWrite() */ /*! + \fn bool QRhiGraphicsPipeline::hasDepthWrite() const + \return true if depth write is enabled. + */ + +/*! \fn void QRhiGraphicsPipeline::setDepthWrite(bool enable) - Controls the writing out of depth data into the depth buffer. By default - this is disabled. Depth write is typically enabled together with the depth - test. + Controls the writing out of depth data into the depth buffer based on + \a enable. By default this is disabled. Depth write is typically enabled + together with the depth test. \note Enabling depth write without having depth testing enabled may not lead to the desired result, and should be avoided. @@ -3979,9 +6857,233 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const */ /*! + \fn QRhiGraphicsPipeline::CompareOp QRhiGraphicsPipeline::depthOp() const + \return the depth comparison function. + */ + +/*! + \fn void QRhiGraphicsPipeline::setDepthOp(CompareOp op) + Sets the depth comparison function \a op. + */ + +/*! + \fn bool QRhiGraphicsPipeline::hasStencilTest() const + \return true if stencil testing is enabled. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilTest(bool enable) + Enables or disables stencil tests based on \a enable. + By default this is disabled. + */ + +/*! + \fn QRhiGraphicsPipeline::StencilOpState QRhiGraphicsPipeline::stencilFront() const + \return the current stencil test state for front faces. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilFront(const StencilOpState &state) + Sets the stencil test \a state for front faces. + */ + +/*! + \fn QRhiGraphicsPipeline::StencilOpState QRhiGraphicsPipeline::stencilBack() const + \return the current stencil test state for back faces. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilBack(const StencilOpState &state) + Sets the stencil test \a state for back faces. + */ + +/*! + \fn quint32 QRhiGraphicsPipeline::stencilReadMask() const + \return the currrent stencil read mask. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilReadMask(quint32 mask) + Sets the stencil read \a mask. The default value is 0xFF. + */ + +/*! + \fn quint32 QRhiGraphicsPipeline::stencilWriteMask() const + \return the current stencil write mask. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilWriteMask(quint32 mask) + Sets the stencil write \a mask. The default value is 0xFF. + */ + +/*! + \fn int QRhiGraphicsPipeline::sampleCount() const + \return the currently set sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiGraphicsPipeline::setSampleCount(int s) + + Sets the sample count. Typical values for \a s are 1, 4, or 8. The pipeline + must always be compatible with the render target, i.e. the sample counts + must match. + + \sa QRhi::supportedSampleCounts() + */ + +/*! + \fn float QRhiGraphicsPipeline::lineWidth() const + \return the currently set line width. The default is 1.0f. + */ + +/*! + \fn void QRhiGraphicsPipeline::setLineWidth(float width) + + Sets the line \a width. If the QRhi::WideLines feature is reported as + unsupported at runtime, values other than 1.0f are ignored. + */ + +/*! + \fn int QRhiGraphicsPipeline::depthBias() const + \return the currently set depth bias. + */ + +/*! + \fn void QRhiGraphicsPipeline::setDepthBias(int bias) + Sets the depth \a bias. The default value is 0. + */ + +/*! + \fn float QRhiGraphicsPipeline::slopeScaledDepthBias() const + \return the currently set slope scaled depth bias. + */ + +/*! + \fn void QRhiGraphicsPipeline::setSlopeScaledDepthBias(float bias) + Sets the slope scaled depth \a bias. The default value is 0. + */ + +/*! + \fn void QRhiGraphicsPipeline::setShaderStages(std::initializer_list<QRhiShaderStage> list) + Sets the \a list of shader stages. + */ + +/*! + \fn template<typename InputIterator> void QRhiGraphicsPipeline::setShaderStages(InputIterator first, InputIterator last) + Sets the list of shader stages from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::cbeginShaderStages() const + \return a const iterator pointing to the first item in the shader stage list. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::cendShaderStages() const + \return a const iterator pointing just after the last item in the shader stage list. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::shaderStageAt(qsizetype index) const + \return the shader stage at the specified \a index. + */ + +/*! + \fn qsizetype QRhiGraphicsPipeline::shaderStageCount() const + \return the number of shader stages in this pipeline. + */ + +/*! + \fn QRhiVertexInputLayout QRhiGraphicsPipeline::vertexInputLayout() const + \return the currently set vertex input layout specification. + */ + +/*! + \fn void QRhiGraphicsPipeline::setVertexInputLayout(const QRhiVertexInputLayout &layout) + Specifies the vertex input \a layout. + */ + +/*! + \fn QRhiShaderResourceBindings *QRhiGraphicsPipeline::shaderResourceBindings() const + \return the currently associated QRhiShaderResourceBindings object. + */ + +/*! + \fn void QRhiGraphicsPipeline::setShaderResourceBindings(QRhiShaderResourceBindings *srb) + + Associates with \a srb describing the resource binding layout and the + resources (QRhiBuffer, QRhiTexture) themselves. The latter is optional, + because only the layout matters during pipeline creation. Therefore, the \a + srb passed in here can leave the actual buffer or texture objects + unspecified (\nullptr) as long as there is another, + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings bound via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} before + recording the draw calls. + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiGraphicsPipeline::renderPassDescriptor() const + \return the currently set QRhiRenderPassDescriptor. + */ + +/*! + \fn void QRhiGraphicsPipeline::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + Associates with the specified QRhiRenderPassDescriptor \a desc. + */ + +/*! + \fn int QRhiGraphicsPipeline::patchControlPointCount() const + \return the currently set patch control point count. + */ + +/*! + \fn void QRhiGraphicsPipeline::setPatchControlPointCount(int count) + + Sets the number of patch control points to \a count. The default value is + 3. This is used only when the topology is set to \l Patches. + */ + +/*! + \fn QRhiGraphicsPipeline::PolygonMode QRhiGraphicsPipeline::polygonMode() const + \return the polygon mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setPolygonMode(PolygonMode mode) + Sets the polygon \a mode. The default is Fill. + + \sa QRhi::NonFillPolygonMode + */ + +/*! + \fn int QRhiGraphicsPipeline::multiViewCount() const + \return the view count. The default is 0, indicating no multiview rendering. + \since 6.7 + */ + +/*! + \fn void QRhiGraphicsPipeline::setMultiViewCount(int count) + Sets the view \a count for multiview rendering. The default is 0, + indicating no multiview rendering. + \a count must be 2 or larger to trigger multiview rendering. + + Multiview is only available when the \l{QRhi::MultiView}{MultiView feature} + is reported as supported. The render target must be a 2D texture array, and + the color attachment for the render target must have the same \a count set. + + See QRhiColorAttachment::setMultiViewCount() for further details on + multiview rendering. + + \since 6.7 + \sa QRhi::MultiView, QRhiColorAttachment::setMultiViewCount() + */ + +/*! \class QRhiSwapChain - \internal \inmodule QtGui + \since 6.6 \brief Swapchain resource. A swapchain enables presenting rendering results to a surface. A swapchain @@ -3991,7 +7093,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const Below is a typical pattern for creating and managing a swapchain and some associated resources in order to render onto a QWindow: - \badcode + \code void init() { sc = rhi->newSwapChain(); @@ -4037,7 +7139,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const Releasing the swapchain must happen while the QWindow and the underlying native window is fully up and running. Building on the previous example: - \badcode + \code void releaseSwapChain() { if (hasSwapChain) { @@ -4069,7 +7171,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const events. QExposeEvent is a loosely specified event that is sent whenever a window gets mapped, obscured, and resized, depending on the platform. - \badcode + \code void Window::exposeEvent(QExposeEvent *) { // initialize and start rendering when the window becomes usable for graphics purposes @@ -4115,6 +7217,9 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const command at the end of a frame. For OpenGL, it is necessary to request the appropriate sample count also via QSurfaceFormat, by calling QSurfaceFormat::setDefaultFormat() before initializing the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -4135,12 +7240,13 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const with premultiplied alpha. In that case the behavior with this flag set is expected to be equivalent to SurfaceHasPreMulAlpha. - \value sRGB Requests to pick an sRGB format for the swapchain and/or its - render target views, where applicable. Note that this implies that sRGB - framebuffer update and blending will get enabled for all content targeting - this swapchain, and opting out is not possible. For OpenGL, set - \l{QSurfaceFormat::sRGBColorSpace}{sRGBColorSpace} on the QSurfaceFormat of - the QWindow in addition. + \value sRGB Requests to pick an sRGB format for the swapchain's color + buffers and/or render target views, where applicable. Note that this + implies that sRGB framebuffer update and blending will get enabled for all + content targeting this swapchain, and opting out is not possible. For + OpenGL, set \l{QSurfaceFormat::sRGBColorSpace}{sRGBColorSpace} on the + QSurfaceFormat of the QWindow in addition. Applicable only when the + swapchain format is set to QRhiSwapChain::SDR. \value UsedAsTransferSource Indicates the swapchain will be used as the source of a readback in QRhiResourceUpdateBatch::readBackTexture(). @@ -4169,6 +7275,39 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const */ /*! + \enum QRhiSwapChain::Format + Describes the swapchain format. The default format is SDR. + + This enum is used with + \l{QRhiSwapChain::isFormatSupported()}{isFormatSupported()} to check + upfront if creating the swapchain with the given format is supported by the + platform and the window's associated screen, and with + \l{QRhiSwapChain::setFormat()}{setFormat()} + to set the requested format in the swapchain before calling + \l{QRhiSwapChain::createOrResize()}{createOrResize()} for the first time. + + \value SDR 8-bit RGBA or BGRA, depending on the backend and platform. With + OpenGL ES in particular, it could happen that the platform provides less + than 8 bits (e.g. due to EGL and the QSurfaceFormat choosing a 565 or 444 + format - this is outside the control of QRhi). Standard dynamic range. May + be combined with setting the QRhiSwapChain::sRGB flag. + + \value HDRExtendedSrgbLinear 16-bit float RGBA, high dynamic range, + extended linear sRGB (scRGB) color space. This involves Rec. 709 primaries + (same as SDR/sRGB) and linear colors. Conversion to the display's native + color space (such as, HDR10) is performed by the windowing system. On + Windows this is the canonical color space of the system compositor, and is + the recommended format for HDR swapchains in general on desktop platforms. + + \value HDR10 10-bit unsigned int RGB or BGR with 2 bit alpha, high dynamic + range, HDR10 (Rec. 2020) color space with an ST2084 PQ transfer function. + + \value HDRExtendedDisplayP3Linear 16-bit float RGBA, high dynamic range, + extended linear Display P3 color space. The primary choice for HDR on + platforms such as iOS and VisionOS. + */ + +/*! \internal */ QRhiSwapChain::QRhiSwapChain(QRhiImplementation *rhi) @@ -4211,7 +7350,7 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! - \fn QSize QRhiSwapChain::surfacePixelSize() + \fn virtual QSize QRhiSwapChain::surfacePixelSize() = 0 \return The size of the window's associated surface or layer. @@ -4219,13 +7358,13 @@ QRhiResource::Type QRhiSwapChain::resourceType() const QWindow::devicePixelRatio()}. With some graphics APIs and windowing system interfaces (for example, Vulkan) there is a theoretical possibility for a surface to assume a size different from the associated window. To support - these cases, rendering logic must always base size-derived calculations + these cases, \b{rendering logic must always base size-derived calculations (such as, viewports) on the size reported from QRhiSwapChain, and never on - the size queried from QWindow. + the size queried from QWindow}. - \note Can also be called before createOrResize(), if at least window() is - already set) This in combination with currentPixelSize() allows to detect - when a swapchain needs to be resized. However, watch out for the fact that + \note \b{Can also be called before createOrResize(), if at least window() is + already set. This in combination with currentPixelSize() allows to detect + when a swapchain needs to be resized.} However, watch out for the fact that the size of the underlying native object (surface, layer, or similar) is "live", so whenever this function is called, it returns the latest value reported by the underlying implementation, without any atomicity guarantee. @@ -4246,17 +7385,58 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! - \fn QRhiCommandBuffer *QRhiSwapChain::currentFrameCommandBuffer() + \fn virtual bool QRhiSwapChain::isFormatSupported(Format f) = 0 - \return a command buffer on which rendering commands can be recorded. Only - valid within a QRhi::beginFrame() - QRhi::endFrame() block where - beginFrame() was called with this swapchain. + \return true if the given swapchain format \a f is supported. SDR is always + supported. - \note the value must not be cached and reused between frames + \note Can be called independently of createOrResize(), but window() must + already be set. Calling without the window set may lead to unexpected + results depending on the backend and platform (most likely false for any + HDR format), because HDR format support is usually tied to the output + (screen) to which the swapchain's associated window belongs at any given + time. If the result is true for a HDR format, then creating the swapchain + with that format is expected to succeed as long as the window is not moved + to another screen in the meantime. + + The main use of this function is to call it before the first + createOrResize() after the window is already set. This allow the QRhi + backends to perform platform or windowing system specific queries to + determine if the window (and the screen it is on) is capable of true HDR + output with the specified format. + + When the format is reported as supported, call setFormat() to set the + requested format and call createOrResize(). Be aware of the consequences + however: successfully requesting a HDR format will involve having to deal + with a different color space, possibly doing white level correction for + non-HDR-aware content, adjusting tonemapping methods, adjusting offscreen + render target settings, etc. + + \sa setFormat() + */ + +/*! + \fn virtual QRhiCommandBuffer *QRhiSwapChain::currentFrameCommandBuffer() = 0 + + \return a command buffer on which rendering commands and resource updates + can be recorded within a \l{QRhi::beginFrame()}{beginFrame} - + \l{QRhi::endFrame()}{endFrame} block, assuming beginFrame() was called with + this swapchain. + + \note The returned object is valid also after endFrame(), up until the next + beginFrame(), but the returned command buffer should not be used to record + any commands then. Rather, it can be used to query data collected during + the frame (or previous frames), for example by calling + \l{QRhiCommandBuffer::lastCompletedGpuTime()}{lastCompletedGpuTime()}. + + \note The value must not be cached and reused between frames. The caller + should not hold on to the returned object once + \l{QRhi::beginFrame()}{beginFrame()} is called again. Instead, the command + buffer object should be queried again by calling this function. */ /*! - \fn QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget() + \fn virtual QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget() = 0 \return a render target that can used with beginPass() in order to render the swapchain's current backbuffer. Only valid within a @@ -4267,7 +7447,35 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! - \fn bool QRhiSwapChain::createOrResize() + \enum QRhiSwapChain::StereoTargetBuffer + Selects the backbuffer to use with a stereoscopic swapchain. + + \value LeftBuffer + \value RightBuffer + */ + +/*! + \return a render target that can be used with beginPass() in order to + render to the swapchain's left or right backbuffer. This overload should be + used only with stereoscopic rendering, that is, when the associated QWindow + is backed by two color buffers, one for each eye, instead of just one. + + When stereoscopic rendering is not supported, the return value will be + the default target. It is supported by all hardware backends except for Metal, in + combination with \l QSurfaceFormat::StereoBuffers, assuming it is supported + by the graphics and display driver stack at run time. Metal and Null backends + are going to return the default render target from this overload. + + \note the value must not be cached and reused between frames + */ +QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + Q_UNUSED(targetBuffer); + return currentFrameRenderTarget(); +} + +/*! + \fn virtual bool QRhiSwapChain::createOrResize() = 0 Creates the swapchain if not already done and resizes the swapchain buffers to match the current size of the targeted surface. Call this whenever the @@ -4283,9 +7491,352 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! + \fn QWindow *QRhiSwapChain::window() const + \return the currently set window. + */ + +/*! + \fn void QRhiSwapChain::setWindow(QWindow *window) + Sets the \a window. + */ + +/*! + \fn QRhiSwapChainProxyData QRhiSwapChain::proxyData() const + \return the currently set proxy data. + */ + +/*! + \fn void QRhiSwapChain::setProxyData(const QRhiSwapChainProxyData &d) + Sets the proxy data \a d. + + \sa QRhi::updateSwapChainProxyData() + */ + +/*! + \fn QRhiSwapChain::Flags QRhiSwapChain::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiSwapChain::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiSwapChain::Format QRhiSwapChain::format() const + \return the currently set format. + */ + +/*! + \fn void QRhiSwapChain::setFormat(Format f) + Sets the format \a f. + + Avoid setting formats that are reported as unsupported from + isFormatSupported(). Note that support for a given format may depend on the + screen the swapchain's associated window is opened on. On some platforms, + such as Windows and macOS, for HDR output to work it is necessary to have + HDR output enabled in the display settings. + + See isFormatSupported(), \l QRhiSwapChainHdrInfo, and \l Format for more + information on high dynamic range output. + */ + +/*! + \fn QRhiRenderBuffer *QRhiSwapChain::depthStencil() const + \return the currently associated renderbuffer for depth-stencil. + */ + +/*! + \fn void QRhiSwapChain::setDepthStencil(QRhiRenderBuffer *ds) + Sets the renderbuffer \a ds for use as a depth-stencil buffer. + */ + +/*! + \fn int QRhiSwapChain::sampleCount() const + \return the currently set sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiSwapChain::setSampleCount(int samples) + + Sets the sample count. Common values for \a samples are 1 (no MSAA), 4 (4x + MSAA), or 8 (8x MSAA). + + \sa QRhi::supportedSampleCounts() + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiSwapChain::renderPassDescriptor() const + \return the currently associated QRhiRenderPassDescriptor object. + */ + +/*! + \fn void QRhiSwapChain::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + Associates with the QRhiRenderPassDescriptor \a desc. + */ + +/*! + \fn virtual QRhiRenderPassDescriptor *QRhiSwapChain::newCompatibleRenderPassDescriptor() = 0; + + \return a new QRhiRenderPassDescriptor that is compatible with this swapchain. + + The returned value is used in two ways: it can be passed to + setRenderPassDescriptor() and + QRhiGraphicsPipeline::setRenderPassDescriptor(). A render pass descriptor + describes the attachments (color, depth/stencil) and the load/store + behavior that can be affected by flags(). A QRhiGraphicsPipeline can only + be used in combination with a swapchain that has a + \l{QRhiRenderPassDescriptor::isCompatible()}{compatible} + QRhiRenderPassDescriptor set. + + \sa createOrResize() + */ + +/*! + \struct QRhiSwapChainHdrInfo + \inmodule QtGui + \since 6.6 + + \brief Describes the high dynamic range related information of the + swapchain's associated output. + + To perform HDR-compatible tonemapping, where the target range is not [0,1], + one often needs to know the maximum luminance of the display the + swapchain's window is associated with. While this is often made + user-configurable (think brightness, gamma and similar settings in games), + it can be highly useful to set defaults based on the values reported by the + display itself, thus providing a decent starting point. + + There are some problems however: the information is exposed in different + forms on different platforms, whereas with cross-platform graphics APIs + there is often no associated solution at all, because managing such + information is not in the scope of the API (and may rather be retrievable + via other platform-specific means, if any). + + With Metal on macOS/iOS, there is no luminance values exposed in the + platform APIs. Instead, the maximum color component value, that would be + 1.0 in a non-HDR setup, is provided. The \c limitsType field indicates what + kind of information is available. It is then up to the clients of QRhi to + access the correct data from the \c limits union and use it as they see + fit. + + With an API like Vulkan, where there is no way to get such information, the + values are always the built-in defaults. + + Therefore, the struct returned from QRhiSwapChain::hdrInfo() contains + either some hard-coded defaults or real values received from an API such as + DXGI (IDXGIOutput6) or Cocoa (NSScreen). When no platform queries are + available (or needs using platform facilities out of scope for QRhi), the + hard-coded defaults are a maximum luminance of 1000 nits and an SDR white + level of 200. + + The struct also exposes the presumed luminance behavior of the platform and + its compositor, to indicate what a color component value of 1.0 is treated + as in a HDR color buffer. In some cases it will be necessary to perform + color correction of non-HDR content composited with HDR content. To enable + this, the SDR white level is queried from the system on some platforms + (Windows) and exposed here. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiSwapChain::hdrInfo() + */ + +/*! + \enum QRhiSwapChainHdrInfo::LimitsType + + \value LuminanceInNits Indicates that the \l limits union has its + \c luminanceInNits struct set + + \value ColorComponentValue Indicates that the \l limits union has its + \c colorComponentValue struct set +*/ + +/*! + \enum QRhiSwapChainHdrInfo::LuminanceBehavior + + \value SceneReferred Indicates that the color value of 1.0 is interpreted + as 80 nits. This is the behavior of HDR-enabled windows with the Windows + compositor. See + \l{https://learn.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range}{this + page} for more information on HDR on Windows. + + \value DisplayReferred Indicates that the color value of 1.0 is interpreted + as the value of the SDR white. (which can be e.g. 200 nits, but will vary + depending on screen brightness) This is the behavior of HDR-enabled windows + on Apple platforms. See + \l{https://developer.apple.com/documentation/metal/hdr_content/displaying_hdr_content_in_a_metal_layer}{this + page} for more information on Apple's EDR system. +*/ + +/*! + \variable QRhiSwapChainHdrInfo::limitsType + + With Metal on macOS/iOS, there is no luminance values exposed in the + platform APIs. Instead, the maximum color component value, that would be + 1.0 in a non-HDR setup, is provided. This value indicates what kind of + information is available in \l limits. + + \sa QRhiSwapChain::hdrInfo() +*/ + +/*! + \variable QRhiSwapChainHdrInfo::limits + + Contains the actual values queried from the graphics API or the platform. + The type of data is indicated by \l limitsType. This is therefore a union. + There are currently two options: + + Luminance values in nits: + + \code + struct { + float minLuminance; + float maxLuminance; + } luminanceInNits; + \endcode + + On Windows the minimum and maximum luminance depends on the screen + brightness. While not relevant for desktops, on laptops the screen + brightness may change at any time. Increasing brightness implies decreased + maximum luminance. In addition, the results may also be dependent on the + HDR Content Brightness set in Windows Settings' System/Display/HDR view, + if there is such a setting. + + Note however that the changes made to the laptop screen's brightness or in + the system settings while the application is running are not necessarily + reflected in the returned values, meaning calling hdrInfo() again may still + return the same luminance range as before for the rest of the process' + lifetime. The exact behavior is up to DXGI and Qt has no control over it. + + \note The Windows compositor works in scene-referred mode for HDR content. + A color component value of 1.0 corresponds to a luminance of 80 nits. When + rendering non-HDR content (e.g. 2D UI elements), the correction of the + white level is often necessary. (e.g., outputting the fragment color (1, 1, + 1) will likely lead to showing a shade of white that is too dim on-screen) + See \l sdrWhiteLevel. + + For macOS/iOS, the current maximum and potential maximum color + component values are provided: + + \code + struct { + float maxColorComponentValue; + float maxPotentialColorComponentValue; + } colorComponentValue; + \endcode + + The value may depend on the screen brightness, which on laptops means that + the result may change in the next call to hdrInfo() if the brightness was + changed in the meantime. The maximum screen brightness implies a maximum + color value of 1.0. + + \note Apple's EDR is display-referred. 1.0 corresponds to a luminance level + of SDR white (e.g. 200 nits), the value of which varies based on the screen + brightness and possibly other settings. The exact luminance value for that, + or the maximum luminance of the display, are not exposed to the + applications. + + \note It has been observed that the color component values are not set to + the correct larger-than-1 value right away on startup on some macOS + systems, but the values tend to change during or after the first frame. + + \sa QRhiSwapChain::hdrInfo() +*/ + +/*! + \variable QRhiSwapChainHdrInfo::luminanceBehavior + + Describes the platform's presumed behavior with regards to color values. + + \sa sdrWhiteLevel + */ + +/*! + \variable QRhiSwapChainHdrInfo::sdrWhiteLevel + + On Windows this is the dynamic SDR white level in nits. The value is + dependent on the screen brightness (on laptops), and the SDR or HDR Content + Brightness settings in the Windows settings' System/Display/HDR view. + + To perform white level correction for non-HDR (SDR) content, such as 2D UI + elemenents, multiply the final color with sdrWhiteLevel / 80.0 whenever + \l luminanceBehavior is SceneReferred. (assuming Windows and a linear + extended sRGB (scRGB) color space) + + On other platforms the value is always a pre-defined value, 200. This may + not match the system's actual SDR white level, but the value of this + variable is not relevant in practice when the \l luminanceBehavior is + DisplayReferred, because then the color component value of 1.0 refers to + the SDR white by default. + + \sa luminanceBehavior +*/ + +/*! + \return the HDR information for the associated display. + + Do not assume that this is a cheap operation. Depending on the platform, + this function makes various platform queries which may have a performance + impact. + + \note Can be called before createOrResize() as long as the window is + \l{setWindow()}{set}. + + \note What happens when moving a window with an initialized swapchain + between displays (HDR to HDR with different characteristics, HDR to SDR, + etc.) is not currently well-defined and depends heavily on the windowing + system and compositor, with potentially varying behavior between platforms. + Currently QRhi only guarantees that hdrInfo() returns valid data, if + available, for the display to which the swapchain's associated window + belonged at the time of createOrResize(). + + \sa QRhiSwapChainHdrInfo + */ +QRhiSwapChainHdrInfo QRhiSwapChain::hdrInfo() +{ + QRhiSwapChainHdrInfo info; + info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; + info.limits.luminanceInNits.minLuminance = 0.0f; + info.limits.luminanceInNits.maxLuminance = 1000.0f; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; + info.sdrWhiteLevel = 200.0f; + return info; +} + +#ifndef QT_NO_DEBUG_STREAM +QDebug operator<<(QDebug dbg, const QRhiSwapChainHdrInfo &info) +{ + QDebugStateSaver saver(dbg); + dbg.nospace() << "QRhiSwapChainHdrInfo("; + switch (info.limitsType) { + case QRhiSwapChainHdrInfo::LuminanceInNits: + dbg.nospace() << " minLuminance=" << info.limits.luminanceInNits.minLuminance + << " maxLuminance=" << info.limits.luminanceInNits.maxLuminance; + break; + case QRhiSwapChainHdrInfo::ColorComponentValue: + dbg.nospace() << " maxColorComponentValue=" << info.limits.colorComponentValue.maxColorComponentValue; + dbg.nospace() << " maxPotentialColorComponentValue=" << info.limits.colorComponentValue.maxPotentialColorComponentValue; + break; + } + switch (info.luminanceBehavior) { + case QRhiSwapChainHdrInfo::SceneReferred: + dbg.nospace() << " scene-referred, SDR white level=" << info.sdrWhiteLevel; + break; + case QRhiSwapChainHdrInfo::DisplayReferred: + dbg.nospace() << " display-referred"; + break; + } + dbg.nospace() << ')'; + return dbg; +} +#endif + +/*! \class QRhiComputePipeline - \internal \inmodule QtGui + \since 6.6 \brief Compute pipeline state resource. \note Setting the shader resource bindings is mandatory. The referenced @@ -4293,6 +7844,9 @@ QRhiResource::Type QRhiSwapChain::resourceType() const time create() is called. \note Setting the shader is mandatory. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -4322,15 +7876,59 @@ QRhiComputePipeline::QRhiComputePipeline(QRhiImplementation *rhi) } /*! + \fn QRhiComputePipeline::Flags QRhiComputePipeline::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiComputePipeline::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiShaderStage QRhiComputePipeline::shaderStage() const + \return the currently set shader. + */ + +/*! + \fn void QRhiComputePipeline::setShaderStage(const QRhiShaderStage &stage) + + Sets the shader to use. \a stage can only refer to the + \l{QRhiShaderStage::Compute}{compute stage}. + */ + +/*! + \fn QRhiShaderResourceBindings *QRhiComputePipeline::shaderResourceBindings() const + \return the currently associated QRhiShaderResourceBindings object. + */ + +/*! + \fn void QRhiComputePipeline::setShaderResourceBindings(QRhiShaderResourceBindings *srb) + + Associates with \a srb describing the resource binding layout and the + resources (QRhiBuffer, QRhiTexture) themselves. The latter is optional. As + with graphics pipelines, the \a srb passed in here can leave the actual + buffer or texture objects unspecified (\nullptr) as long as there is + another, + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings bound via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} before + recording the dispatch call. + */ + +/*! \class QRhiCommandBuffer - \internal \inmodule QtGui + \since 6.6 \brief Command buffer resource. Not creatable by applications at the moment. The only ways to obtain a valid QRhiCommandBuffer are to get it from the targeted swapchain via QRhiSwapChain::currentFrameCommandBuffer(), or, in case of rendering completely offscreen, initializing one via QRhi::beginOffscreenFrame(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -4391,7 +7989,7 @@ QRhiResource::Type QRhiCommandBuffer::resourceType() const return CommandBuffer; } -static const char *resourceTypeStr(QRhiResource *res) +static const char *resourceTypeStr(const QRhiResource *res) { switch (res->resourceType()) { case QRhiResource::Buffer: @@ -4404,8 +8002,8 @@ static const char *resourceTypeStr(QRhiResource *res) return "RenderBuffer"; case QRhiResource::RenderPassDescriptor: return "RenderPassDescriptor"; - case QRhiResource::RenderTarget: - return "RenderTarget"; + case QRhiResource::SwapChainRenderTarget: + return "SwapChainRenderTarget"; case QRhiResource::TextureRenderTarget: return "TextureRenderTarget"; case QRhiResource::ShaderResourceBindings: @@ -4418,11 +8016,9 @@ static const char *resourceTypeStr(QRhiResource *res) return "ComputePipeline"; case QRhiResource::CommandBuffer: return "CommandBuffer"; - default: - Q_UNREACHABLE(); - break; } - return ""; + + Q_UNREACHABLE_RETURN(""); } QRhiImplementation::~QRhiImplementation() @@ -4439,11 +8035,21 @@ QRhiImplementation::~QRhiImplementation() // release builds: opt-in static bool leakCheck = qEnvironmentVariableIntValue("QT_RHI_LEAK_CHECK"); #endif - if (leakCheck && !resources.isEmpty()) { - qWarning("QRhi %p going down with %d unreleased resources that own native graphics objects. This is not nice.", - q, int(resources.count())); - for (QRhiResource *res : qAsConst(resources)) { - qWarning(" %s resource %p (%s)", resourceTypeStr(res), res, res->m_objectName.constData()); + if (!resources.isEmpty()) { + if (leakCheck) { + qWarning("QRhi %p going down with %d unreleased resources that own native graphics objects. This is not nice.", + q, int(resources.size())); + } + for (auto it = resources.cbegin(), end = resources.cend(); it != end; ++it) { + QRhiResource *res = it.key(); + const bool ownsNativeResources = it.value(); + if (leakCheck && ownsNativeResources) + qWarning(" %s resource %p (%s)", resourceTypeStr(res), res, res->m_objectName.constData()); + + // Null out the resource's rhi pointer. This is why it makes sense to do null + // checks in the destroy() implementations of the various resource types. It + // allows to survive in bad applications that somehow manage to destroy a + // resource of a QRhi after the QRhi itself. res->m_rhi = nullptr; } } @@ -4621,6 +8227,10 @@ void QRhiImplementation::textureFormatInfo(QRhiTexture::Format format, const QSi bpc = 4; break; + case QRhiTexture::RGB10A2: + bpc = 4; + break; + case QRhiTexture::D16: bpc = 2; break; @@ -4643,21 +8253,15 @@ void QRhiImplementation::textureFormatInfo(QRhiTexture::Format format, const QSi *bytesPerPixel = bpc; } -// Approximate because it excludes subresource alignment or multisampling. -quint32 QRhiImplementation::approxByteSizeForTexture(QRhiTexture::Format format, const QSize &baseSize, int depth, - int mipCount, int layerCount) +bool QRhiImplementation::isStencilSupportingFormat(QRhiTexture::Format format) const { - quint32 approxSize = 0; - for (int level = 0; level < mipCount; ++level) { - quint32 byteSize = 0; - const QSize size(qFloor(qreal(qMax(1, baseSize.width() >> level))), - qFloor(qreal(qMax(1, baseSize.height() >> level)))); - textureFormatInfo(format, size, nullptr, &byteSize, nullptr); - approxSize += byteSize; + switch (format) { + case QRhiTexture::D24S8: + return true; + default: + break; } - approxSize *= depth; // 3D texture depth or 1 otherwise - approxSize *= uint(layerCount); // 6 for cubemaps or 1 otherwise - return approxSize; + return false; } bool QRhiImplementation::sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps) @@ -4701,7 +8305,7 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin const int CHECKED_BINDINGS_COUNT = 64; bool bindingSeen[CHECKED_BINDINGS_COUNT] = {}; for (auto it = srb->cbeginBindings(), end = srb->cendBindings(); it != end; ++it) { - const int binding = it->data()->binding; + const int binding = shaderResourceBindingData(*it)->binding; if (binding >= CHECKED_BINDINGS_COUNT) continue; if (binding < 0) { @@ -4709,7 +8313,7 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin bindingsOk = false; continue; } - switch (it->data()->type) { + switch (shaderResourceBindingData(*it)->type) { case QRhiShaderResourceBinding::UniformBuffer: if (!bindingSeen[binding]) { bindingSeen[binding] = true; @@ -4726,10 +8330,24 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin bindingsOk = false; } break; + case QRhiShaderResourceBinding::Texture: + if (!bindingSeen[binding]) { + bindingSeen[binding] = true; + } else { + qWarning("Texture duplicates an existing binding number %d", binding); + bindingsOk = false; + } + break; + case QRhiShaderResourceBinding::Sampler: + if (!bindingSeen[binding]) { + bindingSeen[binding] = true; + } else { + qWarning("Sampler duplicates an existing binding number %d", binding); + bindingsOk = false; + } + break; case QRhiShaderResourceBinding::ImageLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::ImageLoadStore: if (!bindingSeen[binding]) { bindingSeen[binding] = true; @@ -4739,9 +8357,7 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin } break; case QRhiShaderResourceBinding::BufferLoad: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferStore: - Q_FALLTHROUGH(); case QRhiShaderResourceBinding::BufferLoadStore: if (!bindingSeen[binding]) { bindingSeen[binding] = true; @@ -4751,7 +8367,7 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin } break; default: - qWarning("Unknown binding type %d", int(it->data()->type)); + qWarning("Unknown binding type %d", int(shaderResourceBindingData(*it)->type)); bindingsOk = false; break; } @@ -4767,6 +8383,41 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin return true; } +int QRhiImplementation::effectiveSampleCount(int sampleCount) const +{ + // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. + const int s = qBound(1, sampleCount, 64); + const QList<int> supported = supportedSampleCounts(); + int result = 1; + + // Stay compatible with Qt 5 in that requesting an unsupported sample count + // is not an error (although we still do a categorized debug print about + // this), and rather a supported value, preferably a close one, not just 1, + // is used instead. This is actually deviating from Qt 5 as that performs a + // clamping only and does not handle cases such as when sample count 2 is + // not supported but 4 is. (OpenGL handles things like that gracefully, + // other APIs may not, so improve this by picking the next largest, or in + // absence of that, the largest value; this with the goal to not reduce + // quality by rather picking a larger-than-requested value than a smaller one) + + for (int i = 0, ie = supported.count(); i != ie; ++i) { + // assumes the 'supported' list is sorted + if (supported[i] >= s) { + result = supported[i]; + break; + } + } + + if (result != s) { + if (result == 1 && !supported.isEmpty()) + result = supported.last(); + qCDebug(QRHI_LOG_INFO, "Attempted to set unsupported sample count %d, using %d instead", + sampleCount, result); + } + + return result; +} + /*! \internal */ @@ -4782,30 +8433,64 @@ QRhi::~QRhi() if (!d) return; + runCleanup(); + qDeleteAll(d->pendingDeleteResources); d->pendingDeleteResources.clear(); - runCleanup(); - d->destroy(); delete d; } +void QRhiImplementation::prepareForCreate(QRhi *rhi, QRhi::Implementation impl, QRhi::Flags flags) +{ + q = rhi; + + // Play nice with QSG_INFO since that is still the most commonly used + // way to get graphics info printed from Qt Quick apps, and the Quick + // scenegraph is our primary user. + if (qEnvironmentVariableIsSet("QSG_INFO")) + const_cast<QLoggingCategory &>(QRHI_LOG_INFO()).setEnabled(QtDebugMsg, true); + + debugMarkers = flags.testFlag(QRhi::EnableDebugMarkers); + + implType = impl; + implThread = QThread::currentThread(); +} + /*! - \return a new QRhi instance with a backend for the graphics API specified by \a impl. + \return a new QRhi instance with a backend for the graphics API specified + by \a impl with the specified \a flags. \a params must point to an instance of one of the backend-specific subclasses of QRhiInitParams, such as, QRhiVulkanInitParams, - QRhiMetalInitParams, QRhiD3D11InitParams, QRhiGles2InitParams. See these - classes for examples on creating a QRhi. - - \a flags is optional. It is used to enable profile and debug related - features that are potentially expensive and should only be used during - development. + QRhiMetalInitParams, QRhiD3D11InitParams, QRhiD3D12InitParams, + QRhiGles2InitParams. See these classes for examples on creating a QRhi. + + QRhi by design does not implement any fallback logic: if the specified API + cannot be initialized, create() will fail, with warnings printed on the + debug output by the backends. The clients of QRhi, for example Qt Quick, + may however provide additional logic that allow falling back to an API + different than what was requested, depending on the platform. If the + intention is just to test if initialization would succeed when calling + create() at later point, it is preferable to use probe() instead of + create(), because with some backends probing can be implemented in a more + lightweight manner as opposed to create(), which performs full + initialization of the infrastructure and is wasteful if that QRhi instance + is then thrown immediately away. + + \a importDevice allows using an already existing graphics device, without + QRhi creating its own. When not null, this parameter must point to an + instance of one of the subclasses of QRhiNativeHandles: + QRhiVulkanNativeHandles, QRhiD3D11NativeHandles, QRhiD3D12NativeHandles, + QRhiMetalNativeHandles, QRhiGles2NativeHandles. The exact details and + semantics depend on the backand and the underlying graphics API. + + \sa probe() */ QRhi *QRhi::create(Implementation impl, QRhiInitParams *params, Flags flags, QRhiNativeHandles *importDevice) { - QScopedPointer<QRhi> r(new QRhi); + std::unique_ptr<QRhi> r(new QRhi); switch (impl) { case Null: @@ -4840,7 +8525,7 @@ QRhi *QRhi::create(Implementation impl, QRhiInitParams *params, Flags flags, QRh break; #endif case Metal: -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) +#if QT_CONFIG(metal) r->d = new QRhiMetal(static_cast<QRhiMetalInitParams *>(params), static_cast<QRhiMetalNativeHandles *>(importDevice)); break; @@ -4848,35 +8533,116 @@ QRhi *QRhi::create(Implementation impl, QRhiInitParams *params, Flags flags, QRh qWarning("This platform has no Metal support"); break; #endif - default: + case D3D12: +#ifdef Q_OS_WIN +#ifdef QRHI_D3D12_AVAILABLE + r->d = new QRhiD3D12(static_cast<QRhiD3D12InitParams *>(params), + static_cast<QRhiD3D12NativeHandles *>(importDevice)); + break; +#else + qWarning("Qt was built without Direct3D 12 support. " + "This is likely due to having ancient SDK headers (such as d3d12.h) in the Qt build environment. " + "Rebuild Qt with an SDK supporting D3D12 features introduced in Windows 10 version 1703, " + "or use an MSVC build as those typically are built with more up-to-date SDKs."); + break; +#endif +#else + qWarning("This platform has no Direct3D 12 support"); break; +#endif } if (r->d) { - r->d->q = r.data(); + r->d->prepareForCreate(r.get(), impl, flags); + if (r->d->create(flags)) + return r.release(); + } - if (flags.testFlag(EnableProfiling)) { - QRhiProfilerPrivate *profD = QRhiProfilerPrivate::get(&r->d->profiler); - profD->rhiDWhenEnabled = r->d; - const_cast<QLoggingCategory &>(QRHI_LOG_INFO()).setEnabled(QtDebugMsg, true); - } + return nullptr; +} + +/*! + \return true if create() can be expected to succeed when called the given + \a impl and \a params. - // Play nice with QSG_INFO since that is still the most commonly used - // way to get graphics info printed from Qt Quick apps, and the Quick - // scenegraph is our primary user. - if (qEnvironmentVariableIsSet("QSG_INFO")) - const_cast<QLoggingCategory &>(QRHI_LOG_INFO()).setEnabled(QtDebugMsg, true); + For some backends this is equivalent to calling create(), checking its + return value, and then destroying the resulting QRhi. - r->d->debugMarkers = flags.testFlag(EnableDebugMarkers); + For others, in particular with Metal, there may be a specific probing + implementation, which allows testing in a more lightweight manner without + polluting the debug output with warnings upon failures. - if (r->d->create(flags)) { - r->d->implType = impl; - r->d->implThread = QThread::currentThread(); - return r.take(); - } + \sa create() + */ +bool QRhi::probe(QRhi::Implementation impl, QRhiInitParams *params) +{ + bool ok = false; + + // The only place currently where this makes sense is Metal, where the API + // is simple enough so that a special probing function - doing nothing but + // a MTLCreateSystemDefaultDevice - is reasonable. Elsewhere, just call + // create() and then drop the result. + + if (impl == Metal) { +#if QT_CONFIG(metal) + ok = QRhiMetal::probe(static_cast<QRhiMetalInitParams *>(params)); +#endif + } else { + QRhi *rhi = create(impl, params); + ok = rhi != nullptr; + delete rhi; } + return ok; +} - return nullptr; +/*! + \struct QRhiSwapChainProxyData + \inmodule QtGui + \since 6.6 + + \brief Opaque data describing native objects needed to set up a swapchain. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhi::updateSwapChainProxyData() + */ + +/*! + Generates and returns a QRhiSwapChainProxyData struct containing opaque + data specific to the backend and graphics API specified by \a impl. \a + window is the QWindow a swapchain is targeting. + + The returned struct can be passed to QRhiSwapChain::setProxyData(). This + makes sense in threaded rendering systems: this static function is expected + to be called on the \b{main (gui) thread}, unlike all QRhi operations, then + transferred to the thread working with the QRhi and QRhiSwapChain and passed + on to the swapchain. This allows doing native platform queries that are + only safe to be called on the main thread, for example to query the + CAMetalLayer from a NSView, and then passing on the data to the + QRhiSwapChain living on the rendering thread. With the Metal example, doing + the view.layer access on a dedicated rendering thread causes a warning in + the Xcode Thread Checker. With the data proxy mechanism, this is avoided. + + When threads are not involved, generating and passing on the + QRhiSwapChainProxyData is not required: backends are guaranteed to be able + to query whatever is needed on their own, and if everything lives on the + main (gui) thread, that should be sufficient. + + \note \a impl should match what the QRhi is created with. For example, + calling with QRhi::Metal on a non-Apple platform will not generate any + useful data. + */ +QRhiSwapChainProxyData QRhi::updateSwapChainProxyData(QRhi::Implementation impl, QWindow *window) +{ +#if QT_CONFIG(metal) + if (impl == Metal) + return QRhiMetal::updateSwapChainProxyData(window); +#else + Q_UNUSED(impl); + Q_UNUSED(window); +#endif + return {}; } /*! @@ -4888,11 +8654,12 @@ QRhi::Implementation QRhi::backend() const } /*! - \return the backend type as string for this QRhi. + \return a friendly name for the backend \a impl, usually the name of the 3D + API in use. */ -const char *QRhi::backendName() const +const char *QRhi::backendName(Implementation impl) { - switch (d->implType) { + switch (impl) { case QRhi::Null: return "Null"; case QRhi::Vulkan: @@ -4903,16 +8670,28 @@ const char *QRhi::backendName() const return "D3D11"; case QRhi::Metal: return "Metal"; - default: - return "Unknown"; + case QRhi::D3D12: + return "D3D12"; } + + Q_UNREACHABLE_RETURN("Unknown"); +} + +/*! + \return the backend type as string for this QRhi. + */ +const char *QRhi::backendName() const +{ + return backendName(d->implType); } /*! \enum QRhiDriverInfo::DeviceType - Specifies the graphics device's type, when the information is available. In - practice this is only applicable with Vulkan and Metal. With others the - value will always be UnknownDevice. + Specifies the graphics device's type, when the information is available. + + In practice this is only applicable with Vulkan and Metal. With Direct 3D + 11 and 12, using an adapter with the software flag set leads to the value + \c CpuDevice. Otherwise, and with OpenGL, the value is always UnknownDevice. \value UnknownDevice \value IntegratedDevice @@ -4924,9 +8703,8 @@ const char *QRhi::backendName() const /*! \struct QRhiDriverInfo - \internal \inmodule QtGui - \since 6.1 + \since 6.6 \brief Describes the physical device, adapter, or graphics API implementation that is used by an initialized QRhi. @@ -4938,8 +8716,35 @@ const char *QRhi::backendName() const \c{GL_VERSION}. The deviceId is always 0 for OpenGL. vendorId is always 0 for OpenGL and Metal. deviceType is always UnknownDevice for OpenGL and Direct 3D. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiDriverInfo::deviceName + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::deviceId + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::vendorId + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::deviceType + + \sa QRhi::driverInfo(), QRhiDriverInfo::DeviceType +*/ + #ifndef QT_NO_DEBUG_STREAM static inline const char *deviceTypeStr(QRhiDriverInfo::DeviceType type) { @@ -4956,9 +8761,9 @@ static inline const char *deviceTypeStr(QRhiDriverInfo::DeviceType type) return "Virtual"; case QRhiDriverInfo::CpuDevice: return "Cpu"; - default: - return ""; } + + Q_UNREACHABLE_RETURN(nullptr); } QDebug operator<<(QDebug dbg, const QRhiDriverInfo &info) { @@ -5007,6 +8812,33 @@ void QRhi::addCleanupCallback(const CleanupCallback &callback) } /*! + \overload + + Registers \a callback to be invoked either when the QRhi is destroyed or + when runCleanup() is called. This overload takes an opaque pointer, \a key, + that is used to ensure that a given callback is registered (and so called) + only once. + + \sa removeCleanupCallback() + */ +void QRhi::addCleanupCallback(const void *key, const CleanupCallback &callback) +{ + d->addCleanupCallback(key, callback); +} + +/*! + Deregisters the callback with \a key. If no cleanup callback was registered + with \a key, the function does nothing. Callbacks registered without a key + cannot be removed. + + \sa addCleanupCallback() + */ +void QRhi::removeCleanupCallback(const void *key) +{ + d->removeCleanupCallback(key); +} + +/*! Invokes all registered cleanup functions. The list of cleanup callbacks it then cleared. Normally destroying the QRhi does this automatically, but sometimes it can be useful to trigger cleanup in order to release all @@ -5016,16 +8848,21 @@ void QRhi::addCleanupCallback(const CleanupCallback &callback) */ void QRhi::runCleanup() { - for (const CleanupCallback &f : qAsConst(d->cleanupCallbacks)) + for (const CleanupCallback &f : std::as_const(d->cleanupCallbacks)) f(this); d->cleanupCallbacks.clear(); + + for (auto it = d->keyedCleanupCallbacks.cbegin(), end = d->keyedCleanupCallbacks.cend(); it != end; ++it) + it.value()(this); + + d->keyedCleanupCallbacks.clear(); } /*! \class QRhiResourceUpdateBatch - \internal \inmodule QtGui + \since 6.6 \brief Records upload and copy type of operations. With QRhi it is no longer possible to perform copy type of operations at @@ -5041,6 +8878,9 @@ void QRhi::runCleanup() To get an available, empty batch from the pool, call QRhi::nextResourceUpdateBatch(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -5084,29 +8924,26 @@ void QRhiResourceUpdateBatch::release() that is then merged into another when starting to first render pass later on: - \badcode - void init() - { - ... - initialUpdates = rhi->nextResourceUpdateBatch(); - initialUpdates->uploadStaticBuffer(vbuf, vertexData); - initialUpdates->uploadStaticBuffer(ibuf, indexData); - ... - } + \code + void init() + { + initialUpdates = rhi->nextResourceUpdateBatch(); + initialUpdates->uploadStaticBuffer(vbuf, vertexData); + initialUpdates->uploadStaticBuffer(ibuf, indexData); + // ... + } - void render() - { - ... - QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); - if (initialUpdates) { - resUpdates->merge(initialUpdates); - initialUpdates->release(); - initialUpdates = nullptr; + void render() + { + QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); + if (initialUpdates) { + resUpdates->merge(initialUpdates); + initialUpdates->release(); + initialUpdates = nullptr; + } + // resUpdates->updateDynamicBuffer(...); + cb->beginPass(rt, clearCol, clearDs, resUpdates); } - resUpdates->updateDynamicBuffer(...); - ... - cb->beginPass(rt, clearCol, clearDs, resUpdates); - } \endcode */ void QRhiResourceUpdateBatch::merge(QRhiResourceUpdateBatch *other) @@ -5148,10 +8985,10 @@ bool QRhiResourceUpdateBatch::hasOptimalCapacity() const \note QRhi transparently manages double buffering in order to prevent stalling the graphics pipeline. The fact that a QRhiBuffer may have - multiple native underneath can be safely ignored when using the QRhi and - QRhiResourceUpdateBatch. + multiple native buffer objects underneath can be safely ignored when using + the QRhi and QRhiResourceUpdateBatch. */ -void QRhiResourceUpdateBatch::updateDynamicBuffer(QRhiBuffer *buf, int offset, int size, const void *data) +void QRhiResourceUpdateBatch::updateDynamicBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) { if (size > 0) { const int idx = d->activeBufferOpCount++; @@ -5171,7 +9008,7 @@ void QRhiResourceUpdateBatch::updateDynamicBuffer(QRhiBuffer *buf, int offset, i are specified by \a data which must have at least \a size bytes available. \a data can safely be destroyed or changed once this function returns. */ -void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, int offset, int size, const void *data) +void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) { if (size > 0) { const int idx = d->activeBufferOpCount++; @@ -5183,6 +9020,8 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, int offset, in } /*! + \overload + Enqueues updating the entire QRhiBuffer \a buf created with the type QRhiBuffer::Immutable or QRhiBuffer::Static. */ @@ -5204,7 +9043,7 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, const void *da A readback is asynchronous. \a result contains a callback that is invoked when the operation has completed. The data is provided in - QRhiBufferReadbackResult::data. Upon successful completion that QByteArray + QRhiReadbackResult::data. Upon successful completion that QByteArray will have a size equal to \a size. On failure the QByteArray will be empty. \note Reading buffers with a usage different than QRhiBuffer::UniformBuffer @@ -5221,7 +9060,7 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, const void *da \sa readBackTexture(), QRhi::isFeatureSupported(), QRhi::resourceLimit() */ -void QRhiResourceUpdateBatch::readBackBuffer(QRhiBuffer *buf, int offset, int size, QRhiBufferReadbackResult *result) +void QRhiResourceUpdateBatch::readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result) { const int idx = d->activeBufferOpCount++; if (idx < d->bufferOps.size()) @@ -5268,6 +9107,11 @@ void QRhiResourceUpdateBatch::uploadTexture(QRhiTexture *tex, const QImage &imag \note The source texture \a src must be created with QRhiTexture::UsedAsTransferSource. + + \note The format of the textures must match. With most graphics + APIs the data is copied as-is without any format conversions. If + \a dst and \a src are created with different formats, unspecified + issues may arise. */ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc) { @@ -5290,10 +9134,10 @@ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, co application. Therefore, \a result provides not just the data but also a callback as operations on the batch are asynchronous by nature: - \badcode - beginFrame(sc); - beginPass - ... + \code + rhi->beginFrame(swapchain); + cb->beginPass(swapchain->currentFrameRenderTarget(), colorClear, dsClear); + // ... QRhiReadbackResult *rbResult = new QRhiReadbackResult; rbResult->completed = [rbResult] { { @@ -5304,11 +9148,11 @@ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, co } delete rbResult; }; - u = nextResourceUpdateBatch(); + QRhiResourceUpdateBatch *u = nextResourceUpdateBatch(); QRhiReadbackDescription rb; // no texture -> uses the current backbuffer of sc u->readBackTexture(rb, rbResult); - endPass(u); - endFrame(sc); + cb->endPass(u); + rhi->endFrame(swapchain); \endcode \note The texture must be created with QRhiTexture::UsedAsTransferSource. @@ -5334,8 +9178,8 @@ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, co QRhi::MaxAsyncReadbackFrames. A single readback operation copies one mip level of one layer (cubemap face - or 3D slice) at a time. The level and layer are specified by the respective - fields in \a rb. + or 3D slice or texture array element) at a time. The level and layer are + specified by the respective fields in \a rb. \sa readBackBuffer(), QRhi::resourceLimit() */ @@ -5385,11 +9229,43 @@ void QRhiResourceUpdateBatch::generateMips(QRhiTexture *tex) \note Can be called outside beginFrame() - endFrame() as well since a batch instance just collects data on its own, it does not perform any operations. - \warning The maximum number of batches is 64. When this limit is reached, - the function will return null until a batch is returned to the pool. + Due to not being tied to a frame being recorded, the following sequence is + valid for example: + + \code + rhi->beginFrame(swapchain); + QRhiResourceUpdateBatch *u = rhi->nextResourceUpdateBatch(); + u->uploadStaticBuffer(buf, data); + // ... do not commit the batch + rhi->endFrame(); + // u stays valid (assuming buf stays valid as well) + rhi->beginFrame(swapchain); + swapchain->currentFrameCommandBuffer()->resourceUpdate(u); + // ... draw with buf + rhi->endFrame(); + \endcode + + \warning The maximum number of batches per QRhi is 64. When this limit is + reached, the function will return null until a batch is returned to the + pool. */ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() { + // By default we prefer spreading out the utilization of the 64 batches as + // much as possible, meaning we won't pick the first one even if it's free, + // but prefer picking one after the last picked one. Relevant due to how + // QVLA and QRhiBufferData allocations behind the bufferOps are reused; in + // typical Qt Quick scenes this leads to a form of (eventually) seeding all + // the 64 resource batches with buffer operation data allocations which are + // then reused in subsequent frames. This comes at the expense of using + // more memory, but has proven good results when (CPU) profiling typical + // Quick/Quick3D apps. + // + // Prefering memory over performance means that we always pick the first + // free batch, and triggering the aggressive deallocating of all backing + // memory (see trimOpLists) before returning it. + static const bool preferMemoryOverPerformance = qEnvironmentVariableIntValue("QT_RHI_MINIMIZE_POOLS"); + auto nextFreeBatch = [this]() -> QRhiResourceUpdateBatch * { auto isFree = [this](int i) -> QRhiResourceUpdateBatch * { const quint64 mask = 1ULL << quint64(i); @@ -5397,7 +9273,8 @@ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() d->resUpdPoolMap |= mask; QRhiResourceUpdateBatch *u = d->resUpdPool[i]; QRhiResourceUpdateBatchPrivate::get(u)->poolIndex = i; - d->lastResUpdIdx = i; + if (!preferMemoryOverPerformance) + d->lastResUpdIdx = i; return u; } return nullptr; @@ -5416,7 +9293,7 @@ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() QRhiResourceUpdateBatch *u = nextFreeBatch(); if (!u) { - const int oldSize = d->resUpdPool.count(); + const int oldSize = d->resUpdPool.size(); const int newSize = oldSize + qMin(4, qMax(0, 64 - oldSize)); d->resUpdPool.resize(newSize); for (int i = oldSize; i < newSize; ++i) @@ -5426,6 +9303,9 @@ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() qWarning("Resource update batch pool exhausted (max is 64)"); } + if (preferMemoryOverPerformance && u) + u->d->trimOpLists(); + return u; } @@ -5439,6 +9319,16 @@ void QRhiResourceUpdateBatchPrivate::free() const quint64 mask = 1ULL << quint64(poolIndex); rhi->resUpdPoolMap &= ~mask; poolIndex = -1; + + // textureOps is cleared, to not keep the potentially large image pixel + // data alive, but it is expected that the container keeps the list alloc + // at least. Only trimOpList() goes for the more aggressive route with squeeze. + textureOps.clear(); + + // bufferOps is not touched, to allow reusing allocations (incl. in the + // elements' QRhiBufferData) as much as possible when this batch is used + // again in the future, which is important for performance, in particular + // with Qt Quick. } void QRhiResourceUpdateBatchPrivate::merge(QRhiResourceUpdateBatchPrivate *other) @@ -5466,19 +9356,28 @@ bool QRhiResourceUpdateBatchPrivate::hasOptimalCapacity() const void QRhiResourceUpdateBatchPrivate::trimOpLists() { - Q_ASSERT(poolIndex == -1); // must not be in use + // Unlike free(), this is expected to aggressively deallocate all memory + // used by both the buffer and texture operation lists. (i.e. using + // squeeze() to only keep the stack prealloc of the QVLAs) + // + // This (e.g. just the destruction of bufferOps elements) may have a + // non-negligible performance impact e.g. with Qt Quick with scenes where + // there are lots of buffer operations per frame. activeBufferOpCount = 0; bufferOps.clear(); + bufferOps.squeeze(); activeTextureOpCount = 0; textureOps.clear(); + textureOps.squeeze(); } /*! - Sometimes committing resource updates is necessary without starting a - render pass. Not often needed, updates should typically be passed to - beginPass (or endPass, in case of readbacks) instead. + Sometimes committing resource updates is necessary or just more convenient + without starting a render pass. Calling this function with \a + resourceUpdates is an alternative to passing \a resourceUpdates to a + beginPass() call (or endPass(), which would be typical in case of readbacks). \note Cannot be called inside a pass. */ @@ -5517,6 +9416,29 @@ void QRhiCommandBuffer::resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates) called inside a pass. Also, with the exception of setGraphicsPipeline(), they expect to have a pipeline set already on the command buffer. Unspecified issues may arise otherwise, depending on the backend. + + If \a rt is a QRhiTextureRenderTarget, beginPass() performs a check to see + if the texture and renderbuffer objects referenced from the render target + are up-to-date. This is similar to what setShaderResources() does for + QRhiShaderResourceBindings. If any of the attachments had been rebuilt + since QRhiTextureRenderTarget::create(), an implicit call to create() is + made on \a rt. Therefore, if \a rt has a QRhiTexture color attachment \c + texture, and one needs to make the texture a different size, the following + is then valid: + \code + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ { texture } }); + rt->create(); + // ... + texture->setPixelSize(new_size); + texture->create(); + cb->beginPass(rt, colorClear, dsClear); // this is ok, no explicit rt->create() is required before + \endcode + + \a flags allow controlling certain advanced functionality. One commonly used + flag is \c ExternalContents. This should be specified whenever + beginExternal() will be called within the pass started by this function. + + \sa endPass(), BeginPassFlags */ void QRhiCommandBuffer::beginPass(QRhiRenderTarget *rt, const QColor &colorClearValue, @@ -5532,6 +9454,8 @@ void QRhiCommandBuffer::beginPass(QRhiRenderTarget *rt, \a resourceUpdates, when not null, specifies a resource update batch that is to be committed and then released. + + \sa beginPass() */ void QRhiCommandBuffer::endPass(QRhiResourceUpdateBatch *resourceUpdates) { @@ -5550,9 +9474,12 @@ void QRhiCommandBuffer::endPass(QRhiResourceUpdateBatch *resourceUpdates) \note This function can only be called inside a render pass, meaning between a beginPass() and endPass() call. + + \note The new graphics pipeline \a ps must be a valid pointer. */ void QRhiCommandBuffer::setGraphicsPipeline(QRhiGraphicsPipeline *ps) { + Q_ASSERT(ps != nullptr); m_rhi->setGraphicsPipeline(this, ps); } @@ -5577,6 +9504,13 @@ void QRhiCommandBuffer::setGraphicsPipeline(QRhiGraphicsPipeline *ps) srb. In this case setShaderResources() must be called even if \a srb is the same as in the last call. + When \a srb is not null, the QRhiShaderResourceBindings object the pipeline + was built with in create() is guaranteed to be not accessed in any form. In + fact, it does not need to be valid even at this point: destroying the + pipeline's associated srb after create() and instead explicitly specifying + another, \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout + compatible} one in every setShaderResources() call is valid. + \a dynamicOffsets allows specifying buffer offsets for uniform buffers that were associated with \a srb via QRhiShaderResourceBinding::uniformBufferWithDynamicOffset(). This is @@ -5642,7 +9576,7 @@ void QRhiCommandBuffer::setShaderResources(QRhiShaderResourceBindings *srb, floats for position (so 5 floats per vertex: x, y, r, g, b). A QRhiGraphicsPipeline for this shader can then be created using the input layout: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 5 * sizeof(float) } @@ -5655,11 +9589,10 @@ void QRhiCommandBuffer::setShaderResources(QRhiShaderResourceBindings *srb, Here there is one buffer binding (binding number 0), with two inputs referencing it. When recording the pass, once the pipeline is set, the - vertex bindings can be specified simply like the following (using C++11 - initializer syntax), assuming vbuf is the QRhiBuffer with all the - interleaved position+color data: + vertex bindings can be specified simply like the following, assuming vbuf + is the QRhiBuffer with all the interleaved position+color data: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBinding(vbuf, 0); cb->setVertexInput(0, 1, &vbufBinding); \endcode @@ -5799,8 +9732,9 @@ void QRhiCommandBuffer::drawIndexed(quint32 indexCount, } /*! - Records a named debug group on the command buffer. This is shown in - graphics debugging tools such as \l{https://renderdoc.org/}{RenderDoc} and + Records a named debug group on the command buffer with the specified \a + name. This is shown in graphics debugging tools such as + \l{https://renderdoc.org/}{RenderDoc} and \l{https://developer.apple.com/xcode/}{XCode}. The end of the grouping is indicated by debugMarkEnd(). @@ -5858,6 +9792,8 @@ void QRhiCommandBuffer::debugMarkMsg(const QByteArray &msg) \note Compute is only available when the \l{QRhi::Compute}{Compute} feature is reported as supported. + + \a flags is not currently used. */ void QRhiCommandBuffer::beginComputePass(QRhiResourceUpdateBatch *resourceUpdates, BeginPassFlags flags) { @@ -5941,9 +9877,9 @@ const QRhiNativeHandles *QRhiCommandBuffer::nativeHandles() called when the pass recording was started with specifying QRhiCommandBuffer::ExternalContent. - With Vulkan or Metal one can query the native command buffer or encoder - objects via nativeHandles() and enqueue commands to them. With OpenGL or - Direct3D 11 the (device) context can be retrieved from + With Vulkan, Metal, or Direct3D 12 one can query the native command buffer + or encoder objects via nativeHandles() and enqueue commands to them. With + OpenGL or Direct3D 11 the (device) context can be retrieved from QRhi::nativeHandles(). However, this must never be done without ensuring the QRhiCommandBuffer's state stays up-to-date. Hence the requirement for wrapping any externally added command recording between beginExternal() and @@ -5991,6 +9927,72 @@ void QRhiCommandBuffer::endExternal() } /*! + \return the last available timestamp, in seconds, when + \l QRhi::EnableTimestamps was enabled when creating the QRhi. The value + indicates the elapsed time on the GPU during the last completed frame. + + \note Do not expect results other than 0 when the QRhi::Timestamps feature + is not reported as supported, or when QRhi::EnableTimestamps was not passed + to QRhi::create(). There are exceptions to this, because with some graphics + APIs (Metal) timings are available without having to perform extra + operations (timestamp queries), but portable applications should always + consciously opt-in to timestamp collection when they know it is needed, and + call this function accordingly. + + Care must be exercised with the interpretation of the value, as its + precision and granularity is often not controlled by Qt, and depends on the + underlying graphics API and its implementation. In particular, comparing + the values between different graphics APIs and hardware is discouraged and + may be meaningless. + + When the frame was recorded with \l{QRhi::beginFrame()}{beginFrame()} and + \l{QRhi::endFrame()}{endFrame()}, i.e., with a swapchain, the timing values + will likely become available asynchronously. The returned value may + therefore be 0 (e.g., for the first 1-2 frames) or the last known value + referring to some previous frame. The value my also + become 0 again under certain conditions, such as when resizing the window. + It can be expected that the most up-to-date available value is retrieved in + beginFrame() and becomes queriable via this function once beginFrame() + returns. + + \note Do not assume that the value refers to the previous + (\c{currently_recorded - 1}) frame. It may refer to \c{currently_recorded - + 2} or \c{currently_recorded - 3} as well. The exact behavior may depend on + the graphics API and its implementation. + + On the other hand, with offscreen frames the returned value is up-to-date + once \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} returns, because + offscreen frames reduce GPU pipelining and wait the the commands to be + complete. + + \note This means that, unlike with swapchain frames, with offscreen frames + the returned value is guaranteed to refer to the frame that has just been + submitted and completed. (assuming this function is called after + endOffscreenFrame() but before the next beginOffscreenFrame()) + + Watch out for the consequences of GPU frequency scaling and GPU clock + changes, depending on the platform. For example, on Windows the returned + timing may vary in a quite wide range between frames with modern graphics + cards, even when submitting frames with a similar, or the same workload. + This is out of scope for Qt to control and solve, generally speaking. + However, the D3D12 backend automatically calls + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-id3d12device-setstablepowerstate}{ID3D12Device::SetStablePowerState()} + whenever the environment variable \c QT_D3D_STABLE_POWER_STATE is set to a + non-zero value. This can greatly stabilize the result. It can also have a + non-insignificant effect on the CPU-side timings measured via QElapsedTimer + for example, especially when offscreen frames are involved. + + \note Do not and never ship applications to production with + \c QT_D3D_STABLE_POWER_STATE set. See the Windows API documentation for details. + + \sa QRhi::Timestamps, QRhi::EnableTimestamps + */ +double QRhiCommandBuffer::lastCompletedGpuTime() +{ + return m_rhi->lastCompletedGpuTime(this); +} + +/*! \return the value (typically an offset) \a v aligned to the uniform buffer alignment given by by ubufAlignment(). */ @@ -6003,7 +10005,7 @@ int QRhi::ubufAligned(int v) const /*! \return the number of mip levels for a given \a size. */ -int QRhi::mipLevelsForSize(const QSize &size) const +int QRhi::mipLevelsForSize(const QSize &size) { return qFloor(std::log2(qMax(size.width(), size.height()))) + 1; } @@ -6012,7 +10014,7 @@ int QRhi::mipLevelsForSize(const QSize &size) const \return the texture image size for a given \a mipLevel, calculated based on the level 0 size given in \a baseLevelSize. */ -QSize QRhi::sizeForMipLevel(int mipLevel, const QSize &baseLevelSize) const +QSize QRhi::sizeForMipLevel(int mipLevel, const QSize &baseLevelSize) { const int w = qMax(1, baseLevelSize.width() >> mipLevel); const int h = qMax(1, baseLevelSize.height() >> mipLevel); @@ -6126,7 +10128,8 @@ int QRhi::resourceLimit(ResourceLimit limit) const for the device, context, and similar concepts used by the backend. Cast to QRhiVulkanNativeHandles, QRhiD3D11NativeHandles, - QRhiGles2NativeHandles, QRhiMetalNativeHandles as appropriate. + QRhiD3D12NativeHandles, QRhiGles2NativeHandles, or QRhiMetalNativeHandles + as appropriate. \note No ownership is transferred, neither for the returned pointer nor for any native objects. @@ -6158,18 +10161,6 @@ bool QRhi::makeThreadLocalNativeContextCurrent() } /*! - \return the associated QRhiProfiler instance. - - An instance is always available for each QRhi, but it is not very useful - without EnableProfiling because no data is collected without setting the - flag upon creation. - */ -QRhiProfiler *QRhi::profiler() -{ - return &d->profiler; -} - -/*! Attempts to release resources in the backend's caches. This can include both CPU and GPU resources. Only memory and resources that can be recreated automatically are in scope. As an example, if the backend's @@ -6239,13 +10230,15 @@ bool QRhi::isDeviceLost() const } /*! - \return a binary \a data blob with data collected from the + \return a binary data blob with data collected from the QRhiGraphicsPipeline and QRhiComputePipeline successfully created during the lifetime of this QRhi. By saving and then, in subsequent runs of the same application, reloading the cache data, pipeline and shader creation times can potentially be - accelerated. + reduced. What exactly the cache and its serialized version includes is not + specified, is always specific to the backend used, and in some cases also + dependent on the particular implementation of the graphics API. When the PipelineCacheDataLoadSave is reported as unsupported, the returned QByteArray is empty. @@ -6254,15 +10247,25 @@ bool QRhi::isDeviceLost() const create(), the returned QByteArray may be empty, even when the PipelineCacheDataLoadSave feature is supported. - When the returned data is non-empty, it is always specific to the QRhi - backend, the graphics device, and the driver implementation in use. QRhi + When the returned data is non-empty, it is always specific to the Qt + version and QRhi backend. In addition, in some cases there is a strong + dependency to the graphics device and the exact driver version used. QRhi takes care of adding the appropriate header and safeguards that ensure that - the data can always be passed safely to setPipelineCacheData(). + the data can always be passed safely to setPipelineCacheData(), therefore + attempting to load data from a run on another version of a driver will be + handled safely and gracefully. \note Calling releaseCachedResources() may, depending on the backend, clear the pipeline data collected. A subsequent call to this function may then not return any data. + See EnablePipelineCacheDataSave for further details about this feature. + + \note Minimize the number of calls to this function. Retrieving the blob is + not always a cheap operation, and therefore this function should only be + called at a low frequency, ideally only once e.g. when closing the + application. + \sa setPipelineCacheData(), create(), isFeatureSupported() */ QByteArray QRhi::pipelineCacheData() @@ -6276,13 +10279,14 @@ QByteArray QRhi::pipelineCacheData() When the PipelineCacheDataLoadSave is reported as unsupported, the function is safe to call, but has no effect. - The blob returned by pipelineCacheData() is always specific to a QRhi - backend, a graphics device, and a given version of the graphics driver. - QRhi takes care of adding the appropriate header and safeguards that ensure - that the data can always be passed safely to this function. If there is a - mismatch, e.g. because the driver has been upgraded to a newer version, or - because the data was generated from a different QRhi backend, a warning is - printed and \a data is safely ignored. + The blob returned by pipelineCacheData() is always specific to the Qt + version, the QRhi backend, and, in some cases, also to the graphics device, + and a given version of the graphics driver. QRhi takes care of adding the + appropriate header and safeguards that ensure that the data can always be + passed safely to this function. If there is a mismatch, e.g. because the + driver has been upgraded to a newer version, or because the data was + generated from a different QRhi backend, a warning is printed and \a data + is safely ignored. With Vulkan, this maps directly to VkPipelineCache. Calling this function creates a new Vulkan pipeline cache object, with its initial data sourced @@ -6290,11 +10294,32 @@ QByteArray QRhi::pipelineCacheData() created QRhiGraphicsPipeline and QRhiComputePipeline objects, thus accelerating, potentially, the pipeline creation. + With other APIs there is no real pipeline cache, but they may provide a + cache with bytecode from shader compilations (D3D) or program binaries + (OpenGL). In applications that perform a lot of shader compilation from + source at run time this can provide a significant boost in subsequent runs + if the "pipeline cache" is pre-seeded from an earlier run using this + function. + \note QRhi cannot give any guarantees that \a data has an effect on the pipeline and shader creation performance. With APIs like Vulkan, it is up to the driver to decide if \a data is used for some purpose, or if it is ignored. + See EnablePipelineCacheDataSave for further details about this feature. + + \note This mechanism offered by QRhi is independent of the drivers' own + internal caching mechanism, if any. This means that, depending on the + graphics API and its implementation, the exact effects of retrieving and + then reloading \a data are not predictable. Improved performance may not be + visible at all in case other caching mechanisms outside of Qt's control are + already active. + + \note Minimize the number of calls to this function. Loading the blob is + not always a cheap operation, and therefore this function should only be + called at a low frequency, ideally only once e.g. when starting the + application. + \sa pipelineCacheData(), isFeatureSupported() */ void QRhi::setPipelineCacheData(const QByteArray &data) @@ -6303,6 +10328,137 @@ void QRhi::setPipelineCacheData(const QByteArray &data) } /*! + \struct QRhiStats + \inmodule QtGui + \since 6.6 + + \brief Statistics provided from the underlying memory allocator. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiStats::totalPipelineCreationTime + + The total time in milliseconds spent in graphics and compute pipeline + creation, which usually involves shader compilation or cache lookups, and + potentially expensive processing. + + \note The value should not be compared between different backends since the + concept of "pipelines" and what exactly happens under the hood during, for + instance, a call to QRhiGraphicsPipeline::create(), differ greatly between + graphics APIs and their implementations. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::blockCount + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::allocCount + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::usedBytes + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::unusedBytes + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::totalUsageBytes + + Valid only with D3D12 currently. Matches IDXGIAdapter3::QueryVideoMemoryInfo(). + + \sa QRhi::statistics() +*/ + +#ifndef QT_NO_DEBUG_STREAM +QDebug operator<<(QDebug dbg, const QRhiStats &info) +{ + QDebugStateSaver saver(dbg); + dbg.nospace() << "QRhiStats(" + << "totalPipelineCreationTime=" << info.totalPipelineCreationTime + << " blockCount=" << info.blockCount + << " allocCount=" << info.allocCount + << " usedBytes=" << info.usedBytes + << " unusedBytes=" << info.unusedBytes + << " totalUsageBytes=" << info.totalUsageBytes + << ')'; + return dbg; +} +#endif + +/*! + Gathers and returns statistics about the timings and allocations of + graphics resources. + + Data about memory allocations is only available with some backends, where + such operations are under Qt's control. With graphics APIs where there is + no lower level control over resource memory allocations, this will never be + supported and all relevant fields in the results are 0. + + With Vulkan in particular, the values are valid always, and are queried + from the underlying memory allocator library. This gives an insight into + the memory requirements of the active buffers and textures. + + The same is true for Direct 3D 12. In addition to the memory allocator + library's statistics, here the result also includes a \c totalUsageBytes + field which reports the total size including additional resources that are + not under the memory allocator library's control (swapchain buffers, + descriptor heaps, etc.), as reported by DXGI. + + The values correspond to all types of memory used, combined. (i.e. video + + system in case of a discreet GPU) + + Additional data, such as the total time in milliseconds spent in graphics + and compute pipeline creation (which usually involves shader compilation or + cache lookups, and potentially expensive processing) is available with most + backends. + + \note The elapsed times for operations such as pipeline creation may be + affected by various factors. The results should not be compared between + different backends since the concept of "pipelines" and what exactly + happens under the hood during, for instance, a call to + QRhiGraphicsPipeline::create(), differ greatly between graphics APIs and + their implementations. + + \note Additionally, many drivers will likely employ various caching + strategies for shaders, programs, pipelines. (independently of Qt's own + similar facilities, such as setPipelineCacheData() or the OpenGL-specific + program binary disk cache). Because such internal behavior is transparent + to the API client, Qt and QRhi have no knowledge or control over the exact + caching strategy, persistency, invalidation of the cached data, etc. When + reading timings, such as the time spent on pipeline creation, the potential + presence and unspecified behavior of driver-level caching mechanisms should + be kept in mind. + */ +QRhiStats QRhi::statistics() const +{ + return d->statistics(); +} + +/*! \return a new graphics pipeline resource. \sa QRhiResource::destroy() @@ -6351,7 +10507,7 @@ QRhiShaderResourceBindings *QRhi::newShaderResourceBindings() */ QRhiBuffer *QRhi::newBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, - int size) + quint32 size) { return d->createBuffer(type, usage, size); } @@ -6386,14 +10542,21 @@ QRhiRenderBuffer *QRhi::newRenderBuffer(QRhiRenderBuffer::Type type, } /*! - \return a new texture with the specified \a format, \a pixelSize, \a + \return a new 1D or 2D texture with the specified \a format, \a pixelSize, \a sampleCount, and \a flags. + A 1D texture array must have QRhiTexture::OneDimensional set in \a flags. This + function will implicitly set this flag if the \a pixelSize height is 0. + \note \a format specifies the requested internal and external format, meaning the data to be uploaded to the texture will need to be in a compatible format, while the native texture may (but is not guaranteed to, in case of OpenGL at least) use this format internally. + \note 1D textures are only functional when the OneDimensionalTextures feature is + reported as supported at run time. Further, mipmaps on 1D textures are only + functional when the OneDimensionalTextureMipmaps feature is reported at run time. + \sa QRhiResource::destroy() */ QRhiTexture *QRhi::newTexture(QRhiTexture::Format format, @@ -6401,19 +10564,32 @@ QRhiTexture *QRhi::newTexture(QRhiTexture::Format format, int sampleCount, QRhiTexture::Flags flags) { - return d->createTexture(format, pixelSize, 1, sampleCount, flags); + if (pixelSize.height() == 0) + flags |= QRhiTexture::OneDimensional; + + return d->createTexture(format, pixelSize, 1, 0, sampleCount, flags); } /*! - \return a new texture with the specified \a format, \a width, \a height, \a - depth, \a sampleCount, and \a flags. + \return a new 1D, 2D or 3D texture with the specified \a format, \a width, \a + height, \a depth, \a sampleCount, and \a flags. This overload is suitable for 3D textures because it allows specifying \a depth. A 3D texture must have QRhiTexture::ThreeDimensional set in \a flags, but using this overload that can be omitted because the flag is set - implicitly whenever \a depth is greater than 0. For 2D and cube textures \a + implicitly whenever \a depth is greater than 0. For 1D, 2D and cube textures \a depth should be set to 0. + A 1D texture must have QRhiTexture::OneDimensional set in \a flags. This overload + will implicitly set this flag if both \a height and \a depth are 0. + + \note 3D textures are only functional when the ThreeDimensionalTextures + feature is reported as supported at run time. + + \note 1D textures are only functional when the OneDimensionalTextures feature is + reported as supported at run time. Further, mipmaps on 1D textures are only + functional when the OneDimensionalTextureMipmaps feature is reported at run time. + \overload */ QRhiTexture *QRhi::newTexture(QRhiTexture::Format format, @@ -6424,7 +10600,51 @@ QRhiTexture *QRhi::newTexture(QRhiTexture::Format format, if (depth > 0) flags |= QRhiTexture::ThreeDimensional; - return d->createTexture(format, QSize(width, height), depth, sampleCount, flags); + if (height == 0 && depth == 0) + flags |= QRhiTexture::OneDimensional; + + return d->createTexture(format, QSize(width, height), depth, 0, sampleCount, flags); +} + +/*! + \return a new 1D or 2D texture array with the specified \a format, \a arraySize, + \a pixelSize, \a sampleCount, and \a flags. + + This function implicitly sets QRhiTexture::TextureArray in \a flags. + + A 1D texture array must have QRhiTexture::OneDimensional set in \a flags. This + function will implicitly set this flag if the \a pixelSize height is 0. + + \note Do not confuse texture arrays with arrays of textures. A QRhiTexture + created by this function is usable with 1D or 2D array samplers in the shader, for + example: \c{layout(binding = 1) uniform sampler2DArray texArr;}. Arrays of + textures refers to a list of textures that are exposed to the shader via + QRhiShaderResourceBinding::sampledTextures() and a count > 1, and declared + in the shader for example like this: \c{layout(binding = 1) uniform + sampler2D textures[4];} + + \note This is only functional when the TextureArrays feature is reported as + supported at run time. + + \note 1D textures are only functional when the OneDimensionalTextures feature is + reported as supported at run time. Further, mipmaps on 1D textures are only + functional when the OneDimensionalTextureMipmaps feature is reported at run time. + + + \sa newTexture() + */ +QRhiTexture *QRhi::newTextureArray(QRhiTexture::Format format, + int arraySize, + const QSize &pixelSize, + int sampleCount, + QRhiTexture::Flags flags) +{ + flags |= QRhiTexture::TextureArray; + + if (pixelSize.height() == 0) + flags |= QRhiTexture::OneDimensional; + + return d->createTexture(format, pixelSize, 1, arraySize, sampleCount, flags); } /*! @@ -6432,6 +10652,14 @@ QRhiTexture *QRhi::newTexture(QRhiTexture::Format format, minification filter \a minFilter, mipmapping mode \a mipmapMode, and the addressing (wrap) modes \a addressU, \a addressV, and \a addressW. + \note Setting \a mipmapMode to a value other than \c None implies that + images for all relevant mip levels will be provided either via + \l{QRhiResourceUpdateBatch::uploadTexture()}{texture uploads} or by calling + \l{QRhiResourceUpdateBatch::generateMips()}{generateMips()} on the texture + that is used with this sampler. Attempting to use the sampler with a + texture that has no data for all relevant mip levels will lead to rendering + errors, with the exact behavior dependent on the underlying graphics API. + \sa QRhiResource::destroy() */ QRhiSampler *QRhi::newSampler(QRhiSampler::Filter magFilter, @@ -6600,7 +10828,7 @@ bool QRhi::isRecordingFrame() const C++ containers and other types. It may also be similar to what an OpenGL or Direct 3D 11 implementation performs internally for certain type of objects. - In practice, such double (or tripple) buffering resources is realized in + In practice, such double (or triple) buffering resources is realized in the Vulkan, Metal, and similar QRhi backends by having a fixed number of native resource (such as, VkBuffer) \c slots behind a QRhiResource. That can then be indexed by a frame slot index running 0, 1, .., @@ -6608,7 +10836,7 @@ bool QRhi::isRecordingFrame() const All this is managed transparently to the users of QRhi. However, applications that integrate rendering done directly with the graphics API - may want to perform a similar double or tripple buffering of their own + may want to perform a similar double or triple buffering of their own graphics resources. That is then most easily achieved by knowing the values of the maximum number of in-flight frames (retrievable via resourceLimit()) and the current frame (slot) index (returned by this function). @@ -6635,26 +10863,29 @@ int QRhi::currentFrameSlot() const beginOffscreenFrame, endOffscreenFrame, beginFrame, ...) is possible too but it does reduce parallelism so it should be done only infrequently. - Offscreen frames do not let the CPU - potentially - generate another frame + Offscreen frames do not let the CPU potentially generate another frame while the GPU is still processing the previous one. This has the side effect that if readbacks are scheduled, the results are guaranteed to be available once endOffscreenFrame() returns. That is not the case with - frames targeting a swapchain. + frames targeting a swapchain: there the GPU is potentially better utilized, + but working with readback operations needs more care from the application + because endFrame(), unlike endOffscreenFrame(), does not guarantee that the + results from the readback are available at that point. The skeleton of rendering a frame without a swapchain and then reading the frame contents back could look like the following: - \badcode - QRhiReadbackResult rbResult; - QRhiCommandBuffer *cb; - beginOffscreenFrame(&cb); - beginPass - ... - u = nextResourceUpdateBatch(); - u->readBackTexture(rb, &rbResult); - endPass(u); - endOffscreenFrame(); - // image data available in rbResult + \code + QRhiReadbackResult rbResult; + QRhiCommandBuffer *cb; + rhi->beginOffscreenFrame(&cb); + cb->beginPass(rt, colorClear, dsClear); + // ... + u = nextResourceUpdateBatch(); + u->readBackTexture(rb, &rbResult); + cb->endPass(u); + rhi->endOffscreenFrame(); + // image data available in rbResult \endcode \sa endOffscreenFrame(), beginFrame() @@ -6672,7 +10903,9 @@ QRhi::FrameOpResult QRhi::beginOffscreenFrame(QRhiCommandBuffer **cb, BeginFrame } /*! - Ends and waits for the offscreen frame. + Ends, submits, and waits for the offscreen frame. + + \a flags is not currently used. \sa beginOffscreenFrame() */ @@ -6713,6 +10946,9 @@ QRhi::FrameOpResult QRhi::finish() With some backend this list of supported values is fixed in advance, while with some others the (physical) device properties indicate what is supported at run time. + + \sa QRhiRenderBuffer::setSampleCount(), QRhiTexture::setSampleCount(), + QRhiGraphicsPipeline::setSampleCount(), QRhiSwapChain::setSampleCount() */ QList<int> QRhi::supportedSampleCounts() const { @@ -6734,7 +10970,7 @@ int QRhi::ubufAlignment() const return d->ubufAlignment(); } -static QBasicAtomicInteger<QRhiGlobalObjectIdGenerator::Type> counter = Q_BASIC_ATOMIC_INITIALIZER(0); +Q_CONSTINIT static QBasicAtomicInteger<QRhiGlobalObjectIdGenerator::Type> counter = Q_BASIC_ATOMIC_INITIALIZER(0); QRhiGlobalObjectIdGenerator::Type QRhiGlobalObjectIdGenerator::newId() { @@ -6835,13 +11071,18 @@ QRhiPassResourceTracker::BufferStage QRhiPassResourceTracker::toPassTrackerBuffe // pick the earlier stage (as this is going to be dstAccessMask) if (stages.testFlag(QRhiShaderResourceBinding::VertexStage)) return QRhiPassResourceTracker::BufVertexStage; + if (stages.testFlag(QRhiShaderResourceBinding::TessellationControlStage)) + return QRhiPassResourceTracker::BufTCStage; + if (stages.testFlag(QRhiShaderResourceBinding::TessellationEvaluationStage)) + return QRhiPassResourceTracker::BufTEStage; if (stages.testFlag(QRhiShaderResourceBinding::FragmentStage)) return QRhiPassResourceTracker::BufFragmentStage; if (stages.testFlag(QRhiShaderResourceBinding::ComputeStage)) return QRhiPassResourceTracker::BufComputeStage; + if (stages.testFlag(QRhiShaderResourceBinding::GeometryStage)) + return QRhiPassResourceTracker::BufGeometryStage; - Q_UNREACHABLE(); - return QRhiPassResourceTracker::BufVertexStage; + Q_UNREACHABLE_RETURN(QRhiPassResourceTracker::BufVertexStage); } QRhiPassResourceTracker::TextureStage QRhiPassResourceTracker::toPassTrackerTextureStage(QRhiShaderResourceBinding::StageFlags stages) @@ -6849,13 +11090,18 @@ QRhiPassResourceTracker::TextureStage QRhiPassResourceTracker::toPassTrackerText // pick the earlier stage (as this is going to be dstAccessMask) if (stages.testFlag(QRhiShaderResourceBinding::VertexStage)) return QRhiPassResourceTracker::TexVertexStage; + if (stages.testFlag(QRhiShaderResourceBinding::TessellationControlStage)) + return QRhiPassResourceTracker::TexTCStage; + if (stages.testFlag(QRhiShaderResourceBinding::TessellationEvaluationStage)) + return QRhiPassResourceTracker::TexTEStage; if (stages.testFlag(QRhiShaderResourceBinding::FragmentStage)) return QRhiPassResourceTracker::TexFragmentStage; if (stages.testFlag(QRhiShaderResourceBinding::ComputeStage)) return QRhiPassResourceTracker::TexComputeStage; + if (stages.testFlag(QRhiShaderResourceBinding::GeometryStage)) + return QRhiPassResourceTracker::TexGeometryStage; - Q_UNREACHABLE(); - return QRhiPassResourceTracker::TexVertexStage; + Q_UNREACHABLE_RETURN(QRhiPassResourceTracker::TexVertexStage); } QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhi.h b/src/gui/rhi/qrhi.h new file mode 100644 index 0000000000..d20b7e00d1 --- /dev/null +++ b/src/gui/rhi/qrhi.h @@ -0,0 +1,2026 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHI_H +#define QRHI_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qsize.h> +#include <QtCore/qlist.h> +#include <QtCore/qvarlengtharray.h> +#include <QtCore/qthread.h> +#include <QtGui/qmatrix4x4.h> +#include <QtGui/qcolor.h> +#include <QtGui/qimage.h> +#include <functional> +#include <array> + +#include <rhi/qshader.h> + +QT_BEGIN_NAMESPACE + +class QWindow; +class QRhi; +class QRhiImplementation; +class QRhiBuffer; +class QRhiRenderBuffer; +class QRhiTexture; +class QRhiSampler; +class QRhiCommandBuffer; +class QRhiResourceUpdateBatch; +class QRhiResourceUpdateBatchPrivate; +class QRhiSwapChain; + +class Q_GUI_EXPORT QRhiDepthStencilClearValue +{ +public: + QRhiDepthStencilClearValue() = default; + QRhiDepthStencilClearValue(float d, quint32 s); + + float depthClearValue() const { return m_d; } + void setDepthClearValue(float d) { m_d = d; } + + quint32 stencilClearValue() const { return m_s; } + void setStencilClearValue(quint32 s) { m_s = s; } + +private: + float m_d = 1.0f; + quint32 m_s = 0; + + friend bool operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + { + return a.m_d == b.m_d && a.m_s == b.m_s; + } + + friend bool operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiDepthStencilClearValue &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_d); + seed = hash(seed, v.m_s); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiDepthStencilClearValue, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDepthStencilClearValue &); +#endif + +class Q_GUI_EXPORT QRhiViewport +{ +public: + QRhiViewport() = default; + QRhiViewport(float x, float y, float w, float h, float minDepth = 0.0f, float maxDepth = 1.0f); + + std::array<float, 4> viewport() const { return m_rect; } + void setViewport(float x, float y, float w, float h) { + m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + } + + float minDepth() const { return m_minDepth; } + void setMinDepth(float minDepth) { m_minDepth = minDepth; } + + float maxDepth() const { return m_maxDepth; } + void setMaxDepth(float maxDepth) { m_maxDepth = maxDepth; } + +private: + std::array<float, 4> m_rect { { 0.0f, 0.0f, 0.0f, 0.0f } }; + float m_minDepth = 0.0f; + float m_maxDepth = 1.0f; + + friend bool operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept + { + return a.m_rect == b.m_rect + && a.m_minDepth == b.m_minDepth + && a.m_maxDepth == b.m_maxDepth; + } + + friend bool operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiViewport &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_rect[0]); + seed = hash(seed, v.m_rect[1]); + seed = hash(seed, v.m_rect[2]); + seed = hash(seed, v.m_rect[3]); + seed = hash(seed, v.m_minDepth); + seed = hash(seed, v.m_maxDepth); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiViewport, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiViewport &); +#endif + +class Q_GUI_EXPORT QRhiScissor +{ +public: + QRhiScissor() = default; + QRhiScissor(int x, int y, int w, int h); + + std::array<int, 4> scissor() const { return m_rect; } + void setScissor(int x, int y, int w, int h) { + m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + } + +private: + std::array<int, 4> m_rect { { 0, 0, 0, 0 } }; + + friend bool operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept + { + return a.m_rect == b.m_rect; + } + + friend bool operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiScissor &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_rect[0]); + seed = hash(seed, v.m_rect[1]); + seed = hash(seed, v.m_rect[2]); + seed = hash(seed, v.m_rect[3]); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiScissor, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiScissor &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputBinding +{ +public: + enum Classification { + PerVertex, + PerInstance + }; + + QRhiVertexInputBinding() = default; + QRhiVertexInputBinding(quint32 stride, Classification cls = PerVertex, quint32 stepRate = 1); + + quint32 stride() const { return m_stride; } + void setStride(quint32 s) { m_stride = s; } + + Classification classification() const { return m_classification; } + void setClassification(Classification c) { m_classification = c; } + + quint32 instanceStepRate() const { return m_instanceStepRate; } + void setInstanceStepRate(quint32 rate) { m_instanceStepRate = rate; } + +private: + quint32 m_stride = 0; + Classification m_classification = PerVertex; + quint32 m_instanceStepRate = 1; + + friend bool operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + { + return a.m_stride == b.m_stride + && a.m_classification == b.m_classification + && a.m_instanceStepRate == b.m_instanceStepRate; + } + + friend bool operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputBinding &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_stride); + seed = hash(seed, v.m_classification); + seed = hash(seed, v.m_instanceStepRate); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiVertexInputBinding, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputBinding &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputAttribute +{ +public: + enum Format { + Float4, + Float3, + Float2, + Float, + UNormByte4, + UNormByte2, + UNormByte, + UInt4, + UInt3, + UInt2, + UInt, + SInt4, + SInt3, + SInt2, + SInt, + Half4, + Half3, + Half2, + Half, + UShort4, + UShort3, + UShort2, + UShort, + SShort4, + SShort3, + SShort2, + SShort, + }; + + QRhiVertexInputAttribute() = default; + QRhiVertexInputAttribute(int binding, int location, Format format, quint32 offset, int matrixSlice = -1); + + int binding() const { return m_binding; } + void setBinding(int b) { m_binding = b; } + + int location() const { return m_location; } + void setLocation(int loc) { m_location = loc; } + + Format format() const { return m_format; } + void setFormat(Format f) { m_format = f; } + + quint32 offset() const { return m_offset; } + void setOffset(quint32 ofs) { m_offset = ofs; } + + int matrixSlice() const { return m_matrixSlice; } + void setMatrixSlice(int slice) { m_matrixSlice = slice; } + +private: + int m_binding = 0; + int m_location = 0; + Format m_format = Float4; + quint32 m_offset = 0; + int m_matrixSlice = -1; + + friend bool operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + { + return a.m_binding == b.m_binding + && a.m_location == b.m_location + && a.m_format == b.m_format + && a.m_offset == b.m_offset; + // matrixSlice excluded intentionally + } + + friend bool operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputAttribute &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_binding); + seed = hash(seed, v.m_location); + seed = hash(seed, v.m_format); + seed = hash(seed, v.m_offset); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiVertexInputAttribute, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputAttribute &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputLayout +{ +public: + QRhiVertexInputLayout() = default; + + void setBindings(std::initializer_list<QRhiVertexInputBinding> list) { m_bindings = list; } + template<typename InputIterator> + void setBindings(InputIterator first, InputIterator last) + { + m_bindings.clear(); + std::copy(first, last, std::back_inserter(m_bindings)); + } + const QRhiVertexInputBinding *cbeginBindings() const { return m_bindings.cbegin(); } + const QRhiVertexInputBinding *cendBindings() const { return m_bindings.cend(); } + const QRhiVertexInputBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } + qsizetype bindingCount() const { return m_bindings.count(); } + + void setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) { m_attributes = list; } + template<typename InputIterator> + void setAttributes(InputIterator first, InputIterator last) + { + m_attributes.clear(); + std::copy(first, last, std::back_inserter(m_attributes)); + } + const QRhiVertexInputAttribute *cbeginAttributes() const { return m_attributes.cbegin(); } + const QRhiVertexInputAttribute *cendAttributes() const { return m_attributes.cend(); } + const QRhiVertexInputAttribute *attributeAt(qsizetype index) const { return &m_attributes.at(index); } + qsizetype attributeCount() const { return m_attributes.count(); } + +private: + QVarLengthArray<QRhiVertexInputBinding, 8> m_bindings; + QVarLengthArray<QRhiVertexInputAttribute, 8> m_attributes; + + friend bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + { + return a.m_bindings == b.m_bindings && a.m_attributes == b.m_attributes; + } + + friend bool operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputLayout &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_bindings); + seed = hash(seed, v.m_attributes); + return seed; + } + + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); +}; + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); +#endif + +class Q_GUI_EXPORT QRhiShaderStage +{ +public: + enum Type { + Vertex, + TessellationControl, + TessellationEvaluation, + Geometry, + Fragment, + Compute + }; + + QRhiShaderStage() = default; + QRhiShaderStage(Type type, const QShader &shader, + QShader::Variant v = QShader::StandardShader); + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + QShader shader() const { return m_shader; } + void setShader(const QShader &s) { m_shader = s; } + + QShader::Variant shaderVariant() const { return m_shaderVariant; } + void setShaderVariant(QShader::Variant v) { m_shaderVariant = v; } + +private: + Type m_type = Vertex; + QShader m_shader; + QShader::Variant m_shaderVariant = QShader::StandardShader; + + friend bool operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + { + return a.m_type == b.m_type + && a.m_shader == b.m_shader + && a.m_shaderVariant == b.m_shaderVariant; + } + + friend bool operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiShaderStage &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_type); + seed = hash(seed, v.m_shader); + seed = hash(seed, v.m_shaderVariant); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiShaderStage, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderStage &); +#endif + +using QRhiGraphicsShaderStage = QRhiShaderStage; + +class Q_GUI_EXPORT QRhiShaderResourceBinding +{ +public: + enum Type { + UniformBuffer, + SampledTexture, + Texture, + Sampler, + ImageLoad, + ImageStore, + ImageLoadStore, + BufferLoad, + BufferStore, + BufferLoadStore + }; + + enum StageFlag { + VertexStage = 1 << 0, + TessellationControlStage = 1 << 1, + TessellationEvaluationStage = 1 << 2, + GeometryStage = 1 << 3, + FragmentStage = 1 << 4, + ComputeStage = 1 << 5 + }; + Q_DECLARE_FLAGS(StageFlags, StageFlag) + + QRhiShaderResourceBinding() = default; + + bool isLayoutCompatible(const QRhiShaderResourceBinding &other) const; + + static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding uniformBufferWithDynamicOffset(int binding, StageFlags stage, QRhiBuffer *buf, quint32 size); + + static QRhiShaderResourceBinding sampledTexture(int binding, StageFlags stage, QRhiTexture *tex, QRhiSampler *sampler); + + struct TextureAndSampler { + QRhiTexture *tex; + QRhiSampler *sampler; + }; + static QRhiShaderResourceBinding sampledTextures(int binding, StageFlags stage, int count, const TextureAndSampler *texSamplers); + + static QRhiShaderResourceBinding texture(int binding, StageFlags stage, QRhiTexture *tex); + static QRhiShaderResourceBinding textures(int binding, StageFlags stage, int count, QRhiTexture **tex); + static QRhiShaderResourceBinding sampler(int binding, StageFlags stage, QRhiSampler *sampler); + + static QRhiShaderResourceBinding imageLoad(int binding, StageFlags stage, QRhiTexture *tex, int level); + static QRhiShaderResourceBinding imageStore(int binding, StageFlags stage, QRhiTexture *tex, int level); + static QRhiShaderResourceBinding imageLoadStore(int binding, StageFlags stage, QRhiTexture *tex, int level); + + static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + + struct Data + { + int binding; + QRhiShaderResourceBinding::StageFlags stage; + QRhiShaderResourceBinding::Type type; + struct UniformBufferData { + QRhiBuffer *buf; + quint32 offset; + quint32 maybeSize; + bool hasDynamicOffset; + }; + static constexpr int MAX_TEX_SAMPLER_ARRAY_SIZE = 16; + struct TextureAndOrSamplerData { + int count; + TextureAndSampler texSamplers[MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct StorageImageData { + QRhiTexture *tex; + int level; + }; + struct StorageBufferData { + QRhiBuffer *buf; + quint32 offset; + quint32 maybeSize; + }; + union { + UniformBufferData ubuf; + TextureAndOrSamplerData stex; + StorageImageData simage; + StorageBufferData sbuf; + } u; + + int arraySize() const + { + return type == QRhiShaderResourceBinding::SampledTexture || type == QRhiShaderResourceBinding::Texture + ? u.stex.count + : 1; + } + + template<typename Output> + Output serialize(Output dst) const + { + // must write out exactly LAYOUT_DESC_ENTRIES_PER_BINDING elements here + *dst++ = quint32(binding); + *dst++ = quint32(stage); + *dst++ = quint32(type); + *dst++ = quint32(arraySize()); + return dst; + } + }; + + static constexpr int LAYOUT_DESC_ENTRIES_PER_BINDING = 4; + + template<typename Output> + static void serializeLayoutDescription(const QRhiShaderResourceBinding *first, + const QRhiShaderResourceBinding *last, + Output dst) + { + while (first != last) { + dst = first->d.serialize(dst); + ++first; + } + } + +private: + Data d; + friend class QRhiImplementation; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBinding::StageFlags) + +Q_DECLARE_TYPEINFO(QRhiShaderResourceBinding, Q_PRIMITIVE_TYPE); + +Q_GUI_EXPORT bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; +Q_GUI_EXPORT bool operator!=(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; +Q_GUI_EXPORT size_t qHash(const QRhiShaderResourceBinding &b, size_t seed = 0) noexcept; +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBinding &); +#endif + +class Q_GUI_EXPORT QRhiColorAttachment +{ +public: + QRhiColorAttachment() = default; + QRhiColorAttachment(QRhiTexture *texture); + QRhiColorAttachment(QRhiRenderBuffer *renderBuffer); + + QRhiTexture *texture() const { return m_texture; } + void setTexture(QRhiTexture *tex) { m_texture = tex; } + + QRhiRenderBuffer *renderBuffer() const { return m_renderBuffer; } + void setRenderBuffer(QRhiRenderBuffer *rb) { m_renderBuffer = rb; } + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + + QRhiTexture *resolveTexture() const { return m_resolveTexture; } + void setResolveTexture(QRhiTexture *tex) { m_resolveTexture = tex; } + + int resolveLayer() const { return m_resolveLayer; } + void setResolveLayer(int layer) { m_resolveLayer = layer; } + + int resolveLevel() const { return m_resolveLevel; } + void setResolveLevel(int level) { m_resolveLevel = level; } + + int multiViewCount() const { return m_multiViewCount; } + void setMultiViewCount(int count) { m_multiViewCount = count; } + +private: + QRhiTexture *m_texture = nullptr; + QRhiRenderBuffer *m_renderBuffer = nullptr; + int m_layer = 0; + int m_level = 0; + QRhiTexture *m_resolveTexture = nullptr; + int m_resolveLayer = 0; + int m_resolveLevel = 0; + int m_multiViewCount = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiColorAttachment, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureRenderTargetDescription +{ +public: + QRhiTextureRenderTargetDescription() = default; + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment); + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiRenderBuffer *depthStencilBuffer); + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiTexture *depthTexture); + + void setColorAttachments(std::initializer_list<QRhiColorAttachment> list) { m_colorAttachments = list; } + template<typename InputIterator> + void setColorAttachments(InputIterator first, InputIterator last) + { + m_colorAttachments.clear(); + std::copy(first, last, std::back_inserter(m_colorAttachments)); + } + const QRhiColorAttachment *cbeginColorAttachments() const { return m_colorAttachments.cbegin(); } + const QRhiColorAttachment *cendColorAttachments() const { return m_colorAttachments.cend(); } + const QRhiColorAttachment *colorAttachmentAt(qsizetype index) const { return &m_colorAttachments.at(index); } + qsizetype colorAttachmentCount() const { return m_colorAttachments.count(); } + + QRhiRenderBuffer *depthStencilBuffer() const { return m_depthStencilBuffer; } + void setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) { m_depthStencilBuffer = renderBuffer; } + + QRhiTexture *depthTexture() const { return m_depthTexture; } + void setDepthTexture(QRhiTexture *texture) { m_depthTexture = texture; } + + QRhiTexture *depthResolveTexture() const { return m_depthResolveTexture; } + void setDepthResolveTexture(QRhiTexture *tex) { m_depthResolveTexture = tex; } + +private: + QVarLengthArray<QRhiColorAttachment, 8> m_colorAttachments; + QRhiRenderBuffer *m_depthStencilBuffer = nullptr; + QRhiTexture *m_depthTexture = nullptr; + QRhiTexture *m_depthResolveTexture = nullptr; +}; + +class Q_GUI_EXPORT QRhiTextureSubresourceUploadDescription +{ +public: + QRhiTextureSubresourceUploadDescription() = default; + explicit QRhiTextureSubresourceUploadDescription(const QImage &image); + QRhiTextureSubresourceUploadDescription(const void *data, quint32 size); + explicit QRhiTextureSubresourceUploadDescription(const QByteArray &data); + + QImage image() const { return m_image; } + void setImage(const QImage &image) { m_image = image; } + + QByteArray data() const { return m_data; } + void setData(const QByteArray &data) { m_data = data; } + + quint32 dataStride() const { return m_dataStride; } + void setDataStride(quint32 stride) { m_dataStride = stride; } + + QPoint destinationTopLeft() const { return m_destinationTopLeft; } + void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } + + QSize sourceSize() const { return m_sourceSize; } + void setSourceSize(const QSize &size) { m_sourceSize = size; } + + QPoint sourceTopLeft() const { return m_sourceTopLeft; } + void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } + +private: + QImage m_image; + QByteArray m_data; + quint32 m_dataStride = 0; + QPoint m_destinationTopLeft; + QSize m_sourceSize; + QPoint m_sourceTopLeft; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureSubresourceUploadDescription, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureUploadEntry +{ +public: + QRhiTextureUploadEntry() = default; + QRhiTextureUploadEntry(int layer, int level, const QRhiTextureSubresourceUploadDescription &desc); + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + + QRhiTextureSubresourceUploadDescription description() const { return m_desc; } + void setDescription(const QRhiTextureSubresourceUploadDescription &desc) { m_desc = desc; } + +private: + int m_layer = 0; + int m_level = 0; + QRhiTextureSubresourceUploadDescription m_desc; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureUploadEntry, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureUploadDescription +{ +public: + QRhiTextureUploadDescription() = default; + QRhiTextureUploadDescription(const QRhiTextureUploadEntry &entry); + QRhiTextureUploadDescription(std::initializer_list<QRhiTextureUploadEntry> list); + + void setEntries(std::initializer_list<QRhiTextureUploadEntry> list) { m_entries = list; } + template<typename InputIterator> + void setEntries(InputIterator first, InputIterator last) + { + m_entries.clear(); + std::copy(first, last, std::back_inserter(m_entries)); + } + const QRhiTextureUploadEntry *cbeginEntries() const { return m_entries.cbegin(); } + const QRhiTextureUploadEntry *cendEntries() const { return m_entries.cend(); } + const QRhiTextureUploadEntry *entryAt(qsizetype index) const { return &m_entries.at(index); } + qsizetype entryCount() const { return m_entries.count(); } + +private: + QVarLengthArray<QRhiTextureUploadEntry, 16> m_entries; +}; + +class Q_GUI_EXPORT QRhiTextureCopyDescription +{ +public: + QRhiTextureCopyDescription() = default; + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int sourceLayer() const { return m_sourceLayer; } + void setSourceLayer(int layer) { m_sourceLayer = layer; } + + int sourceLevel() const { return m_sourceLevel; } + void setSourceLevel(int level) { m_sourceLevel = level; } + + QPoint sourceTopLeft() const { return m_sourceTopLeft; } + void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } + + int destinationLayer() const { return m_destinationLayer; } + void setDestinationLayer(int layer) { m_destinationLayer = layer; } + + int destinationLevel() const { return m_destinationLevel; } + void setDestinationLevel(int level) { m_destinationLevel = level; } + + QPoint destinationTopLeft() const { return m_destinationTopLeft; } + void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } + +private: + QSize m_pixelSize; + int m_sourceLayer = 0; + int m_sourceLevel = 0; + QPoint m_sourceTopLeft; + int m_destinationLayer = 0; + int m_destinationLevel = 0; + QPoint m_destinationTopLeft; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureCopyDescription, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiReadbackDescription +{ +public: + QRhiReadbackDescription() = default; + QRhiReadbackDescription(QRhiTexture *texture); + + QRhiTexture *texture() const { return m_texture; } + void setTexture(QRhiTexture *tex) { m_texture = tex; } + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + +private: + QRhiTexture *m_texture = nullptr; + int m_layer = 0; + int m_level = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiReadbackDescription, Q_RELOCATABLE_TYPE); + +struct Q_GUI_EXPORT QRhiNativeHandles +{ +}; + +class Q_GUI_EXPORT QRhiResource +{ +public: + enum Type { + Buffer, + Texture, + Sampler, + RenderBuffer, + RenderPassDescriptor, + SwapChainRenderTarget, + TextureRenderTarget, + ShaderResourceBindings, + GraphicsPipeline, + SwapChain, + ComputePipeline, + CommandBuffer + }; + + virtual ~QRhiResource(); + + virtual Type resourceType() const = 0; + + virtual void destroy() = 0; + + void deleteLater(); + + QByteArray name() const; + void setName(const QByteArray &name); + + quint64 globalResourceId() const; + + QRhi *rhi() const; + +protected: + QRhiResource(QRhiImplementation *rhi); + Q_DISABLE_COPY(QRhiResource) + friend class QRhiImplementation; + QRhiImplementation *m_rhi = nullptr; + quint64 m_id; + QByteArray m_objectName; +}; + +class Q_GUI_EXPORT QRhiBuffer : public QRhiResource +{ +public: + enum Type { + Immutable, + Static, + Dynamic + }; + + enum UsageFlag { + VertexBuffer = 1 << 0, + IndexBuffer = 1 << 1, + UniformBuffer = 1 << 2, + StorageBuffer = 1 << 3 + }; + Q_DECLARE_FLAGS(UsageFlags, UsageFlag) + + struct NativeBuffer { + const void *objects[3]; + int slotCount; + }; + + QRhiResource::Type resourceType() const override; + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + UsageFlags usage() const { return m_usage; } + void setUsage(UsageFlags u) { m_usage = u; } + + quint32 size() const { return m_size; } + void setSize(quint32 sz) { m_size = sz; } + + virtual bool create() = 0; + + virtual NativeBuffer nativeBuffer(); + + virtual char *beginFullDynamicBufferUpdateForCurrentFrame(); + virtual void endFullDynamicBufferUpdateForCurrentFrame(); + +protected: + QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, quint32 size_); + Type m_type; + UsageFlags m_usage; + quint32 m_size; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiBuffer::UsageFlags) + +class Q_GUI_EXPORT QRhiTexture : public QRhiResource +{ +public: + enum Flag { + RenderTarget = 1 << 0, + CubeMap = 1 << 2, + MipMapped = 1 << 3, + sRGB = 1 << 4, + UsedAsTransferSource = 1 << 5, + UsedWithGenerateMips = 1 << 6, + UsedWithLoadStore = 1 << 7, + UsedAsCompressedAtlas = 1 << 8, + ExternalOES = 1 << 9, + ThreeDimensional = 1 << 10, + TextureRectangleGL = 1 << 11, + TextureArray = 1 << 12, + OneDimensional = 1 << 13 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Format { + UnknownFormat, + + RGBA8, + BGRA8, + R8, + RG8, + R16, + RG16, + RED_OR_ALPHA8, + + RGBA16F, + RGBA32F, + R16F, + R32F, + + RGB10A2, + + D16, + D24, + D24S8, + D32F, + + BC1, + BC2, + BC3, + BC4, + BC5, + BC6H, + BC7, + + ETC2_RGB8, + ETC2_RGB8A1, + ETC2_RGBA8, + + ASTC_4x4, + ASTC_5x4, + ASTC_5x5, + ASTC_6x5, + ASTC_6x6, + ASTC_8x5, + ASTC_8x6, + ASTC_8x8, + ASTC_10x5, + ASTC_10x6, + ASTC_10x8, + ASTC_10x10, + ASTC_12x10, + ASTC_12x12 + }; + + struct NativeTexture { + quint64 object; + int layout; // or state + }; + + QRhiResource::Type resourceType() const override; + + Format format() const { return m_format; } + void setFormat(Format fmt) { m_format = fmt; } + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int depth() const { return m_depth; } + void setDepth(int depth) { m_depth = depth; } + + int arraySize() const { return m_arraySize; } + void setArraySize(int arraySize) { m_arraySize = arraySize; } + + int arrayRangeStart() const { return m_arrayRangeStart; } + int arrayRangeLength() const { return m_arrayRangeLength; } + void setArrayRange(int startIndex, int count) + { + m_arrayRangeStart = startIndex; + m_arrayRangeLength = count; + } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + struct ViewFormat { + QRhiTexture::Format format; + bool srgb; + }; + ViewFormat readViewFormat() const { return m_readViewFormat; } + void setReadViewFormat(const ViewFormat &fmt) { m_readViewFormat = fmt; } + ViewFormat writeViewFormat() const { return m_writeViewFormat; } + void setWriteViewFormat(const ViewFormat &fmt) { m_writeViewFormat = fmt; } + + virtual bool create() = 0; + virtual NativeTexture nativeTexture(); + virtual bool createFrom(NativeTexture src); + virtual void setNativeLayout(int layout); + +protected: + QRhiTexture(QRhiImplementation *rhi, Format format_, const QSize &pixelSize_, int depth_, + int arraySize_, int sampleCount_, Flags flags_); + Format m_format; + QSize m_pixelSize; + int m_depth; + int m_arraySize; + int m_sampleCount; + Flags m_flags; + int m_arrayRangeStart = -1; + int m_arrayRangeLength = -1; + ViewFormat m_readViewFormat = { UnknownFormat, false }; + ViewFormat m_writeViewFormat = { UnknownFormat, false }; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTexture::Flags) + +class Q_GUI_EXPORT QRhiSampler : public QRhiResource +{ +public: + enum Filter { + None, + Nearest, + Linear + }; + + enum AddressMode { + Repeat, + ClampToEdge, + Mirror, + }; + + enum CompareOp { + Never, + Less, + Equal, + LessOrEqual, + Greater, + NotEqual, + GreaterOrEqual, + Always + }; + + QRhiResource::Type resourceType() const override; + + Filter magFilter() const { return m_magFilter; } + void setMagFilter(Filter f) { m_magFilter = f; } + + Filter minFilter() const { return m_minFilter; } + void setMinFilter(Filter f) { m_minFilter = f; } + + Filter mipmapMode() const { return m_mipmapMode; } + void setMipmapMode(Filter f) { m_mipmapMode = f; } + + AddressMode addressU() const { return m_addressU; } + void setAddressU(AddressMode mode) { m_addressU = mode; } + + AddressMode addressV() const { return m_addressV; } + void setAddressV(AddressMode mode) { m_addressV = mode; } + + AddressMode addressW() const { return m_addressW; } + void setAddressW(AddressMode mode) { m_addressW = mode; } + + CompareOp textureCompareOp() const { return m_compareOp; } + void setTextureCompareOp(CompareOp op) { m_compareOp = op; } + + virtual bool create() = 0; + +protected: + QRhiSampler(QRhiImplementation *rhi, + Filter magFilter_, Filter minFilter_, Filter mipmapMode_, + AddressMode u_, AddressMode v_, AddressMode w_); + Filter m_magFilter; + Filter m_minFilter; + Filter m_mipmapMode; + AddressMode m_addressU; + AddressMode m_addressV; + AddressMode m_addressW; + CompareOp m_compareOp; +}; + +class Q_GUI_EXPORT QRhiRenderBuffer : public QRhiResource +{ +public: + enum Type { + DepthStencil, + Color + }; + + enum Flag { + UsedWithSwapChainOnly = 1 << 0 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + struct NativeRenderBuffer { + quint64 object; + }; + + QRhiResource::Type resourceType() const override; + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + virtual bool create() = 0; + virtual bool createFrom(NativeRenderBuffer src); + + virtual QRhiTexture::Format backingFormat() const = 0; + +protected: + QRhiRenderBuffer(QRhiImplementation *rhi, Type type_, const QSize &pixelSize_, + int sampleCount_, Flags flags_, QRhiTexture::Format backingFormatHint_); + Type m_type; + QSize m_pixelSize; + int m_sampleCount; + Flags m_flags; + QRhiTexture::Format m_backingFormatHint; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiRenderBuffer::Flags) + +class Q_GUI_EXPORT QRhiRenderPassDescriptor : public QRhiResource +{ +public: + QRhiResource::Type resourceType() const override; + + virtual bool isCompatible(const QRhiRenderPassDescriptor *other) const = 0; + virtual const QRhiNativeHandles *nativeHandles(); + + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const = 0; + + virtual QVector<quint32> serializedFormat() const = 0; + +protected: + QRhiRenderPassDescriptor(QRhiImplementation *rhi); +}; + +class Q_GUI_EXPORT QRhiRenderTarget : public QRhiResource +{ +public: + virtual QSize pixelSize() const = 0; + virtual float devicePixelRatio() const = 0; + virtual int sampleCount() const = 0; + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + +protected: + QRhiRenderTarget(QRhiImplementation *rhi); + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; +}; + +class Q_GUI_EXPORT QRhiSwapChainRenderTarget : public QRhiRenderTarget +{ +public: + QRhiResource::Type resourceType() const override; + QRhiSwapChain *swapChain() const { return m_swapchain; } + +protected: + QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain_); + QRhiSwapChain *m_swapchain; +}; + +class Q_GUI_EXPORT QRhiTextureRenderTarget : public QRhiRenderTarget +{ +public: + enum Flag { + PreserveColorContents = 1 << 0, + PreserveDepthStencilContents = 1 << 1, + DoNotStoreDepthStencilContents = 1 << 2 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QRhiResource::Type resourceType() const override; + + QRhiTextureRenderTargetDescription description() const { return m_desc; } + void setDescription(const QRhiTextureRenderTargetDescription &desc) { m_desc = desc; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; + + virtual bool create() = 0; + +protected: + QRhiTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc_, Flags flags_); + QRhiTextureRenderTargetDescription m_desc; + Flags m_flags; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTextureRenderTarget::Flags) + +class Q_GUI_EXPORT QRhiShaderResourceBindings : public QRhiResource +{ +public: + QRhiResource::Type resourceType() const override; + + void setBindings(std::initializer_list<QRhiShaderResourceBinding> list) { m_bindings = list; } + template<typename InputIterator> + void setBindings(InputIterator first, InputIterator last) + { + m_bindings.clear(); + std::copy(first, last, std::back_inserter(m_bindings)); + } + const QRhiShaderResourceBinding *cbeginBindings() const { return m_bindings.cbegin(); } + const QRhiShaderResourceBinding *cendBindings() const { return m_bindings.cend(); } + const QRhiShaderResourceBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } + qsizetype bindingCount() const { return m_bindings.count(); } + + bool isLayoutCompatible(const QRhiShaderResourceBindings *other) const; + + QVector<quint32> serializedLayoutDescription() const { return m_layoutDesc; } + + virtual bool create() = 0; + + enum UpdateFlag { + BindingsAreSorted = 0x01 + }; + Q_DECLARE_FLAGS(UpdateFlags, UpdateFlag) + + virtual void updateResources(UpdateFlags flags = {}) = 0; + +protected: + static constexpr int BINDING_PREALLOC = 12; + QRhiShaderResourceBindings(QRhiImplementation *rhi); + QVarLengthArray<QRhiShaderResourceBinding, BINDING_PREALLOC> m_bindings; + size_t m_layoutDescHash = 0; + // Intentionally not using QVLA for m_layoutDesc: clients like Qt Quick are much + // better served with an implicitly shared container here, because they will likely + // throw this directly into structs serving as cache keys. + QVector<quint32> m_layoutDesc; + friend class QRhiImplementation; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); +#endif +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBindings::UpdateFlags) + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); +#endif + +// The proper name. Until it gets rolled out universally, have the better name +// as a typedef. Eventually it should be reversed (the old name being a typedef +// to the new one). +using QRhiShaderResourceBindingSet = QRhiShaderResourceBindings; + +class Q_GUI_EXPORT QRhiGraphicsPipeline : public QRhiResource +{ +public: + enum Flag { + UsesBlendConstants = 1 << 0, + UsesStencilRef = 1 << 1, + UsesScissor = 1 << 2, + CompileShadersWithDebugInfo = 1 << 3 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Topology { + Triangles, + TriangleStrip, + TriangleFan, + Lines, + LineStrip, + Points, + Patches + }; + + enum CullMode { + None, + Front, + Back + }; + + enum FrontFace { + CCW, + CW + }; + + enum ColorMaskComponent { + R = 1 << 0, + G = 1 << 1, + B = 1 << 2, + A = 1 << 3 + }; + Q_DECLARE_FLAGS(ColorMask, ColorMaskComponent) + + enum BlendFactor { + Zero, + One, + SrcColor, + OneMinusSrcColor, + DstColor, + OneMinusDstColor, + SrcAlpha, + OneMinusSrcAlpha, + DstAlpha, + OneMinusDstAlpha, + ConstantColor, + OneMinusConstantColor, + ConstantAlpha, + OneMinusConstantAlpha, + SrcAlphaSaturate, + Src1Color, + OneMinusSrc1Color, + Src1Alpha, + OneMinusSrc1Alpha + }; + + enum BlendOp { + Add, + Subtract, + ReverseSubtract, + Min, + Max + }; + + struct TargetBlend { + ColorMask colorWrite = ColorMask(0xF); // R | G | B | A + bool enable = false; + BlendFactor srcColor = One; + BlendFactor dstColor = OneMinusSrcAlpha; + BlendOp opColor = Add; + BlendFactor srcAlpha = One; + BlendFactor dstAlpha = OneMinusSrcAlpha; + BlendOp opAlpha = Add; + }; + + enum CompareOp { + Never, + Less, + Equal, + LessOrEqual, + Greater, + NotEqual, + GreaterOrEqual, + Always + }; + + enum StencilOp { + StencilZero, + Keep, + Replace, + IncrementAndClamp, + DecrementAndClamp, + Invert, + IncrementAndWrap, + DecrementAndWrap + }; + + struct StencilOpState { + StencilOp failOp = Keep; + StencilOp depthFailOp = Keep; + StencilOp passOp = Keep; + CompareOp compareOp = Always; + }; + + enum PolygonMode { + Fill, + Line + }; + + QRhiResource::Type resourceType() const override; + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + Topology topology() const { return m_topology; } + void setTopology(Topology t) { m_topology = t; } + + CullMode cullMode() const { return m_cullMode; } + void setCullMode(CullMode mode) { m_cullMode = mode; } + + FrontFace frontFace() const { return m_frontFace; } + void setFrontFace(FrontFace f) { m_frontFace = f; } + + void setTargetBlends(std::initializer_list<TargetBlend> list) { m_targetBlends = list; } + template<typename InputIterator> + void setTargetBlends(InputIterator first, InputIterator last) + { + m_targetBlends.clear(); + std::copy(first, last, std::back_inserter(m_targetBlends)); + } + const TargetBlend *cbeginTargetBlends() const { return m_targetBlends.cbegin(); } + const TargetBlend *cendTargetBlends() const { return m_targetBlends.cend(); } + const TargetBlend *targetBlendAt(qsizetype index) const { return &m_targetBlends.at(index); } + qsizetype targetBlendCount() const { return m_targetBlends.count(); } + + bool hasDepthTest() const { return m_depthTest; } + void setDepthTest(bool enable) { m_depthTest = enable; } + + bool hasDepthWrite() const { return m_depthWrite; } + void setDepthWrite(bool enable) { m_depthWrite = enable; } + + CompareOp depthOp() const { return m_depthOp; } + void setDepthOp(CompareOp op) { m_depthOp = op; } + + bool hasStencilTest() const { return m_stencilTest; } + void setStencilTest(bool enable) { m_stencilTest = enable; } + + StencilOpState stencilFront() const { return m_stencilFront; } + void setStencilFront(const StencilOpState &state) { m_stencilFront = state; } + + StencilOpState stencilBack() const { return m_stencilBack; } + void setStencilBack(const StencilOpState &state) { m_stencilBack = state; } + + quint32 stencilReadMask() const { return m_stencilReadMask; } + void setStencilReadMask(quint32 mask) { m_stencilReadMask = mask; } + + quint32 stencilWriteMask() const { return m_stencilWriteMask; } + void setStencilWriteMask(quint32 mask) { m_stencilWriteMask = mask; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + float lineWidth() const { return m_lineWidth; } + void setLineWidth(float width) { m_lineWidth = width; } + + int depthBias() const { return m_depthBias; } + void setDepthBias(int bias) { m_depthBias = bias; } + + float slopeScaledDepthBias() const { return m_slopeScaledDepthBias; } + void setSlopeScaledDepthBias(float bias) { m_slopeScaledDepthBias = bias; } + + void setShaderStages(std::initializer_list<QRhiShaderStage> list) { m_shaderStages = list; } + template<typename InputIterator> + void setShaderStages(InputIterator first, InputIterator last) + { + m_shaderStages.clear(); + std::copy(first, last, std::back_inserter(m_shaderStages)); + } + const QRhiShaderStage *cbeginShaderStages() const { return m_shaderStages.cbegin(); } + const QRhiShaderStage *cendShaderStages() const { return m_shaderStages.cend(); } + const QRhiShaderStage *shaderStageAt(qsizetype index) const { return &m_shaderStages.at(index); } + qsizetype shaderStageCount() const { return m_shaderStages.count(); } + + QRhiVertexInputLayout vertexInputLayout() const { return m_vertexInputLayout; } + void setVertexInputLayout(const QRhiVertexInputLayout &layout) { m_vertexInputLayout = layout; } + + QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } + void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + + int patchControlPointCount() const { return m_patchControlPointCount; } + void setPatchControlPointCount(int count) { m_patchControlPointCount = count; } + + PolygonMode polygonMode() const {return m_polygonMode; } + void setPolygonMode(PolygonMode mode) {m_polygonMode = mode; } + + int multiViewCount() const { return m_multiViewCount; } + void setMultiViewCount(int count) { m_multiViewCount = count; } + + virtual bool create() = 0; + +protected: + QRhiGraphicsPipeline(QRhiImplementation *rhi); + Flags m_flags; + Topology m_topology = Triangles; + CullMode m_cullMode = None; + FrontFace m_frontFace = CCW; + QVarLengthArray<TargetBlend, 8> m_targetBlends; + bool m_depthTest = false; + bool m_depthWrite = false; + CompareOp m_depthOp = Less; + bool m_stencilTest = false; + StencilOpState m_stencilFront; + StencilOpState m_stencilBack; + quint32 m_stencilReadMask = 0xFF; + quint32 m_stencilWriteMask = 0xFF; + int m_sampleCount = 1; + float m_lineWidth = 1.0f; + int m_depthBias = 0; + float m_slopeScaledDepthBias = 0.0f; + int m_patchControlPointCount = 3; + PolygonMode m_polygonMode = Fill; + int m_multiViewCount = 0; + QVarLengthArray<QRhiShaderStage, 4> m_shaderStages; + QRhiVertexInputLayout m_vertexInputLayout; + QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::Flags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::ColorMask) +Q_DECLARE_TYPEINFO(QRhiGraphicsPipeline::TargetBlend, Q_RELOCATABLE_TYPE); + +struct QRhiSwapChainHdrInfo +{ + enum LimitsType { + LuminanceInNits, + ColorComponentValue + }; + + enum LuminanceBehavior { + SceneReferred, + DisplayReferred + }; + + LimitsType limitsType; + union { + struct { + float minLuminance; + float maxLuminance; + } luminanceInNits; + struct { + float maxColorComponentValue; + float maxPotentialColorComponentValue; + } colorComponentValue; + } limits; + LuminanceBehavior luminanceBehavior; + float sdrWhiteLevel; +}; + +Q_DECLARE_TYPEINFO(QRhiSwapChainHdrInfo, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiSwapChainHdrInfo &); +#endif + +struct QRhiSwapChainProxyData +{ + void *reserved[2] = {}; +}; + +class Q_GUI_EXPORT QRhiSwapChain : public QRhiResource +{ +public: + enum Flag { + SurfaceHasPreMulAlpha = 1 << 0, + SurfaceHasNonPreMulAlpha = 1 << 1, + sRGB = 1 << 2, + UsedAsTransferSource = 1 << 3, + NoVSync = 1 << 4, + MinimalBufferCount = 1 << 5 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Format { + SDR, + HDRExtendedSrgbLinear, + HDR10, + HDRExtendedDisplayP3Linear + }; + + enum StereoTargetBuffer { + LeftBuffer, + RightBuffer + }; + + QRhiResource::Type resourceType() const override; + + QWindow *window() const { return m_window; } + void setWindow(QWindow *window) { m_window = window; } + + QRhiSwapChainProxyData proxyData() const { return m_proxyData; } + void setProxyData(const QRhiSwapChainProxyData &d) { m_proxyData = d; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + Format format() const { return m_format; } + void setFormat(Format f) { m_format = f; } + + QRhiRenderBuffer *depthStencil() const { return m_depthStencil; } + void setDepthStencil(QRhiRenderBuffer *ds) { m_depthStencil = ds; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int samples) { m_sampleCount = samples; } + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + + QSize currentPixelSize() const { return m_currentPixelSize; } + + virtual QRhiCommandBuffer *currentFrameCommandBuffer() = 0; + virtual QRhiRenderTarget *currentFrameRenderTarget() = 0; + virtual QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer); + virtual QSize surfacePixelSize() = 0; + virtual bool isFormatSupported(Format f) = 0; + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; + virtual bool createOrResize() = 0; + virtual QRhiSwapChainHdrInfo hdrInfo(); + +protected: + QRhiSwapChain(QRhiImplementation *rhi); + QWindow *m_window = nullptr; + Flags m_flags; + Format m_format = SDR; + QRhiRenderBuffer *m_depthStencil = nullptr; + int m_sampleCount = 1; + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; + QSize m_currentPixelSize; + QRhiSwapChainProxyData m_proxyData; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiSwapChain::Flags) + +class Q_GUI_EXPORT QRhiComputePipeline : public QRhiResource +{ +public: + enum Flag { + CompileShadersWithDebugInfo = 1 << 0 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QRhiResource::Type resourceType() const override; + virtual bool create() = 0; + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + QRhiShaderStage shaderStage() const { return m_shaderStage; } + void setShaderStage(const QRhiShaderStage &stage) { m_shaderStage = stage; } + + QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } + void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } + +protected: + QRhiComputePipeline(QRhiImplementation *rhi); + Flags m_flags; + QRhiShaderStage m_shaderStage; + QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiComputePipeline::Flags) + +class Q_GUI_EXPORT QRhiCommandBuffer : public QRhiResource +{ +public: + enum IndexFormat { + IndexUInt16, + IndexUInt32 + }; + + enum BeginPassFlag { + ExternalContent = 0x01, + DoNotTrackResourcesForCompute = 0x02 + }; + Q_DECLARE_FLAGS(BeginPassFlags, BeginPassFlag) + + QRhiResource::Type resourceType() const override; + + void resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates); + + void beginPass(QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates = nullptr, + BeginPassFlags flags = {}); + void endPass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); + + void setGraphicsPipeline(QRhiGraphicsPipeline *ps); + using DynamicOffset = QPair<int, quint32>; // binding, offset + void setShaderResources(QRhiShaderResourceBindings *srb = nullptr, + int dynamicOffsetCount = 0, + const DynamicOffset *dynamicOffsets = nullptr); + using VertexInput = QPair<QRhiBuffer *, quint32>; // buffer, offset + void setVertexInput(int startBinding, int bindingCount, const VertexInput *bindings, + QRhiBuffer *indexBuf = nullptr, quint32 indexOffset = 0, + IndexFormat indexFormat = IndexUInt16); + + void setViewport(const QRhiViewport &viewport); + void setScissor(const QRhiScissor &scissor); + void setBlendConstants(const QColor &c); + void setStencilRef(quint32 refValue); + + void draw(quint32 vertexCount, + quint32 instanceCount = 1, + quint32 firstVertex = 0, + quint32 firstInstance = 0); + + void drawIndexed(quint32 indexCount, + quint32 instanceCount = 1, + quint32 firstIndex = 0, + qint32 vertexOffset = 0, + quint32 firstInstance = 0); + + void debugMarkBegin(const QByteArray &name); + void debugMarkEnd(); + void debugMarkMsg(const QByteArray &msg); + + void beginComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr, BeginPassFlags flags = {}); + void endComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); + void setComputePipeline(QRhiComputePipeline *ps); + void dispatch(int x, int y, int z); + + const QRhiNativeHandles *nativeHandles(); + void beginExternal(); + void endExternal(); + + double lastCompletedGpuTime(); + +protected: + QRhiCommandBuffer(QRhiImplementation *rhi); +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiCommandBuffer::BeginPassFlags) + +struct Q_GUI_EXPORT QRhiReadbackResult +{ + std::function<void()> completed = nullptr; + QRhiTexture::Format format; + QSize pixelSize; + QByteArray data; +}; + +class Q_GUI_EXPORT QRhiResourceUpdateBatch +{ +public: + ~QRhiResourceUpdateBatch(); + + void release(); + + void merge(QRhiResourceUpdateBatch *other); + bool hasOptimalCapacity() const; + + void updateDynamicBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); + void uploadStaticBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); + void uploadStaticBuffer(QRhiBuffer *buf, const void *data); + void readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result); + void uploadTexture(QRhiTexture *tex, const QRhiTextureUploadDescription &desc); + void uploadTexture(QRhiTexture *tex, const QImage &image); + void copyTexture(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc = QRhiTextureCopyDescription()); + void readBackTexture(const QRhiReadbackDescription &rb, QRhiReadbackResult *result); + void generateMips(QRhiTexture *tex); + +private: + QRhiResourceUpdateBatch(QRhiImplementation *rhi); + Q_DISABLE_COPY(QRhiResourceUpdateBatch) + QRhiResourceUpdateBatchPrivate *d; + friend class QRhiResourceUpdateBatchPrivate; + friend class QRhi; +}; + +struct Q_GUI_EXPORT QRhiDriverInfo +{ + enum DeviceType { + UnknownDevice, + IntegratedDevice, + DiscreteDevice, + ExternalDevice, + VirtualDevice, + CpuDevice + }; + + QByteArray deviceName; + quint64 deviceId = 0; + quint64 vendorId = 0; + DeviceType deviceType = UnknownDevice; +}; + +Q_DECLARE_TYPEINFO(QRhiDriverInfo, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDriverInfo &); +#endif + +struct Q_GUI_EXPORT QRhiStats +{ + qint64 totalPipelineCreationTime = 0; + // Vulkan or D3D12 memory allocator statistics + quint32 blockCount = 0; + quint32 allocCount = 0; + quint64 usedBytes = 0; + quint64 unusedBytes = 0; + // D3D12 only, from IDXGIAdapter3::QueryVideoMemoryInfo(), incl. all resources + quint64 totalUsageBytes = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiStats, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiStats &); +#endif + +struct Q_GUI_EXPORT QRhiInitParams +{ +}; + +class Q_GUI_EXPORT QRhi +{ +public: + enum Implementation { + Null, + Vulkan, + OpenGLES2, + D3D11, + Metal, + D3D12 + }; + + enum Flag { + EnableDebugMarkers = 1 << 0, + PreferSoftwareRenderer = 1 << 1, + EnablePipelineCacheDataSave = 1 << 2, + EnableTimestamps = 1 << 3, + SuppressSmokeTestWarnings = 1 << 4 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum FrameOpResult { + FrameOpSuccess = 0, + FrameOpError, + FrameOpSwapChainOutOfDate, + FrameOpDeviceLost + }; + + enum Feature { + MultisampleTexture = 1, + MultisampleRenderBuffer, + DebugMarkers, + Timestamps, + Instancing, + CustomInstanceStepRate, + PrimitiveRestart, + NonDynamicUniformBuffers, + NonFourAlignedEffectiveIndexBufferOffset, + NPOTTextureRepeat, + RedOrAlpha8IsRed, + ElementIndexUint, + Compute, + WideLines, + VertexShaderPointSize, + BaseVertex, + BaseInstance, + TriangleFanTopology, + ReadBackNonUniformBuffer, + ReadBackNonBaseMipLevel, + TexelFetch, + RenderToNonBaseMipLevel, + IntAttributes, + ScreenSpaceDerivatives, + ReadBackAnyTextureFormat, + PipelineCacheDataLoadSave, + ImageDataStride, + RenderBufferImport, + ThreeDimensionalTextures, + RenderTo3DTextureSlice, + TextureArrays, + Tessellation, + GeometryShader, + TextureArrayRange, + NonFillPolygonMode, + OneDimensionalTextures, + OneDimensionalTextureMipmaps, + HalfAttributes, + RenderToOneDimensionalTexture, + ThreeDimensionalTextureMipmaps, + MultiView, + TextureViewFormat, + ResolveDepthStencil + }; + + enum BeginFrameFlag { + }; + Q_DECLARE_FLAGS(BeginFrameFlags, BeginFrameFlag) + + enum EndFrameFlag { + SkipPresent = 1 << 0 + }; + Q_DECLARE_FLAGS(EndFrameFlags, EndFrameFlag) + + enum ResourceLimit { + TextureSizeMin = 1, + TextureSizeMax, + MaxColorAttachments, + FramesInFlight, + MaxAsyncReadbackFrames, + MaxThreadGroupsPerDimension, + MaxThreadsPerThreadGroup, + MaxThreadGroupX, + MaxThreadGroupY, + MaxThreadGroupZ, + TextureArraySizeMax, + MaxUniformBufferRange, + MaxVertexInputs, + MaxVertexOutputs + }; + + ~QRhi(); + + static QRhi *create(Implementation impl, + QRhiInitParams *params, + Flags flags = {}, + QRhiNativeHandles *importDevice = nullptr); + static bool probe(Implementation impl, QRhiInitParams *params); + + Implementation backend() const; + const char *backendName() const; + static const char *backendName(Implementation impl); + QRhiDriverInfo driverInfo() const; + QThread *thread() const; + + using CleanupCallback = std::function<void(QRhi *)>; + void addCleanupCallback(const CleanupCallback &callback); + void addCleanupCallback(const void *key, const CleanupCallback &callback); + void removeCleanupCallback(const void *key); + void runCleanup(); + + QRhiGraphicsPipeline *newGraphicsPipeline(); + QRhiComputePipeline *newComputePipeline(); + QRhiShaderResourceBindings *newShaderResourceBindings(); + + QRhiBuffer *newBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size); + + QRhiRenderBuffer *newRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount = 1, + QRhiRenderBuffer::Flags flags = {}, + QRhiTexture::Format backingFormatHint = QRhiTexture::UnknownFormat); + + QRhiTexture *newTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiTexture *newTexture(QRhiTexture::Format format, + int width, int height, int depth, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiTexture *newTextureArray(QRhiTexture::Format format, + int arraySize, + const QSize &pixelSize, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiSampler *newSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler::AddressMode addressU, + QRhiSampler::AddressMode addressV, + QRhiSampler::AddressMode addressW = QRhiSampler::Repeat); + + QRhiTextureRenderTarget *newTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags = {}); + + QRhiSwapChain *newSwapChain(); + FrameOpResult beginFrame(QRhiSwapChain *swapChain, BeginFrameFlags flags = {}); + FrameOpResult endFrame(QRhiSwapChain *swapChain, EndFrameFlags flags = {}); + bool isRecordingFrame() const; + int currentFrameSlot() const; + + FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, BeginFrameFlags flags = {}); + FrameOpResult endOffscreenFrame(EndFrameFlags flags = {}); + + QRhi::FrameOpResult finish(); + + QRhiResourceUpdateBatch *nextResourceUpdateBatch(); + + QList<int> supportedSampleCounts() const; + + int ubufAlignment() const; + int ubufAligned(int v) const; + + static int mipLevelsForSize(const QSize &size); + static QSize sizeForMipLevel(int mipLevel, const QSize &baseLevelSize); + + bool isYUpInFramebuffer() const; + bool isYUpInNDC() const; + bool isClipDepthZeroToOne() const; + + QMatrix4x4 clipSpaceCorrMatrix() const; + + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags = {}) const; + bool isFeatureSupported(QRhi::Feature feature) const; + int resourceLimit(ResourceLimit limit) const; + + const QRhiNativeHandles *nativeHandles(); + bool makeThreadLocalNativeContextCurrent(); + + static constexpr int MAX_MIP_LEVELS = 16; // -> max width or height is 65536 + + void releaseCachedResources(); + + bool isDeviceLost() const; + + QByteArray pipelineCacheData(); + void setPipelineCacheData(const QByteArray &data); + + QRhiStats statistics() const; + + static QRhiSwapChainProxyData updateSwapChainProxyData(Implementation impl, QWindow *window); + +protected: + QRhi(); + +private: + Q_DISABLE_COPY(QRhi) + QRhiImplementation *d = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::Flags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::BeginFrameFlags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::EndFrameFlags) + +QT_END_NAMESPACE + +#include <rhi/qrhi_platform.h> + +#endif diff --git a/src/gui/rhi/qrhi_p.h b/src/gui/rhi/qrhi_p.h index c2a09b7edf..b5429372a8 100644 --- a/src/gui/rhi/qrhi_p.h +++ b/src/gui/rhi/qrhi_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHI_H -#define QRHI_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHI_P_H +#define QRHI_P_H // // W A R N I N G @@ -51,1633 +15,803 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <QSize> -#include <QMatrix4x4> -#include <QList> -#include <QVarLengthArray> -#include <QThread> -#include <QColor> -#include <QImage> -#include <functional> -#include <array> -#include <private/qshader_p.h> +#include <rhi/qrhi.h> +#include <QBitArray> +#include <QAtomicInt> +#include <QElapsedTimer> +#include <QLoggingCategory> +#include <QtCore/qset.h> +#include <QtCore/qvarlengtharray.h> QT_BEGIN_NAMESPACE -class QWindow; -class QRhiImplementation; -class QRhiBuffer; -class QRhiRenderBuffer; -class QRhiTexture; -class QRhiSampler; -class QRhiCommandBuffer; -class QRhiResourceUpdateBatch; -class QRhiResourceUpdateBatchPrivate; -class QRhiProfiler; - -class Q_GUI_EXPORT QRhiDepthStencilClearValue -{ -public: - QRhiDepthStencilClearValue() = default; - QRhiDepthStencilClearValue(float d, quint32 s); - - float depthClearValue() const { return m_d; } - void setDepthClearValue(float d) { m_d = d; } - - quint32 stencilClearValue() const { return m_s; } - void setStencilClearValue(quint32 s) { m_s = s; } - -private: - float m_d = 1.0f; - quint32 m_s = 0; -}; - -Q_DECLARE_TYPEINFO(QRhiDepthStencilClearValue, Q_RELOCATABLE_TYPE); +#define QRHI_RES(t, x) static_cast<t *>(x) +#define QRHI_RES_RHI(t) t *rhiD = static_cast<t *>(m_rhi) -Q_GUI_EXPORT bool operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiDepthStencilClearValue &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDepthStencilClearValue &); -#endif +Q_DECLARE_LOGGING_CATEGORY(QRHI_LOG_INFO) -class Q_GUI_EXPORT QRhiViewport +class QRhiImplementation { public: - QRhiViewport() = default; - QRhiViewport(float x, float y, float w, float h, float minDepth = 0.0f, float maxDepth = 1.0f); - - std::array<float, 4> viewport() const { return m_rect; } - void setViewport(float x, float y, float w, float h) { - m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; - } - - float minDepth() const { return m_minDepth; } - void setMinDepth(float minDepth) { m_minDepth = minDepth; } - - float maxDepth() const { return m_maxDepth; } - void setMaxDepth(float maxDepth) { m_maxDepth = maxDepth; } + virtual ~QRhiImplementation(); -private: - std::array<float, 4> m_rect { { 0.0f, 0.0f, 0.0f, 0.0f } }; - float m_minDepth = 0.0f; - float m_maxDepth = 1.0f; -}; - -Q_DECLARE_TYPEINFO(QRhiViewport, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiViewport &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiViewport &); -#endif - -class Q_GUI_EXPORT QRhiScissor -{ -public: - QRhiScissor() = default; - QRhiScissor(int x, int y, int w, int h); + virtual bool create(QRhi::Flags flags) = 0; + virtual void destroy() = 0; - std::array<int, 4> scissor() const { return m_rect; } - void setScissor(int x, int y, int w, int h) { - m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + virtual QRhiGraphicsPipeline *createGraphicsPipeline() = 0; + virtual QRhiComputePipeline *createComputePipeline() = 0; + virtual QRhiShaderResourceBindings *createShaderResourceBindings() = 0; + virtual QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) = 0; + virtual QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) = 0; + virtual QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) = 0; + virtual QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) = 0; + + virtual QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) = 0; + + virtual QRhiSwapChain *createSwapChain() = 0; + virtual QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) = 0; + virtual QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) = 0; + virtual QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) = 0; + virtual QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) = 0; + virtual QRhi::FrameOpResult finish() = 0; + + virtual void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + + virtual void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) = 0; + virtual void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + + virtual void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) = 0; + + virtual void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) = 0; + + virtual void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) = 0; + + virtual void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) = 0; + virtual void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) = 0; + virtual void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) = 0; + virtual void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) = 0; + + virtual void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) = 0; + virtual void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) = 0; + + virtual void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) = 0; + virtual void debugMarkEnd(QRhiCommandBuffer *cb) = 0; + virtual void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) = 0; + + virtual void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) = 0; + virtual void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + virtual void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) = 0; + virtual void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) = 0; + + virtual const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) = 0; + virtual void beginExternal(QRhiCommandBuffer *cb) = 0; + virtual void endExternal(QRhiCommandBuffer *cb) = 0; + virtual double lastCompletedGpuTime(QRhiCommandBuffer *cb) = 0; + + virtual QList<int> supportedSampleCounts() const = 0; + virtual int ubufAlignment() const = 0; + virtual bool isYUpInFramebuffer() const = 0; + virtual bool isYUpInNDC() const = 0; + virtual bool isClipDepthZeroToOne() const = 0; + virtual QMatrix4x4 clipSpaceCorrMatrix() const = 0; + virtual bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const = 0; + virtual bool isFeatureSupported(QRhi::Feature feature) const = 0; + virtual int resourceLimit(QRhi::ResourceLimit limit) const = 0; + virtual const QRhiNativeHandles *nativeHandles() = 0; + virtual QRhiDriverInfo driverInfo() const = 0; + virtual QRhiStats statistics() = 0; + virtual bool makeThreadLocalNativeContextCurrent() = 0; + virtual void releaseCachedResources() = 0; + virtual bool isDeviceLost() const = 0; + + virtual QByteArray pipelineCacheData() = 0; + virtual void setPipelineCacheData(const QByteArray &data) = 0; + + void prepareForCreate(QRhi *rhi, QRhi::Implementation impl, QRhi::Flags flags); + + bool isCompressedFormat(QRhiTexture::Format format) const; + void compressedFormatInfo(QRhiTexture::Format format, const QSize &size, + quint32 *bpl, quint32 *byteSize, + QSize *blockDim) const; + void textureFormatInfo(QRhiTexture::Format format, const QSize &size, + quint32 *bpl, quint32 *byteSize, quint32 *bytesPerPixel) const; + bool isStencilSupportingFormat(QRhiTexture::Format format) const; + + void registerResource(QRhiResource *res, bool ownsNativeResources = true) + { + // The ownsNativeResources is relevant for the (graphics resource) leak + // check in ~QRhiImplementation; when false, the registration's sole + // purpose is to automatically null out the resource's m_rhi pointer in + // case the rhi goes away first. (which should not happen in + // well-written applications but we try to be graceful) + resources.insert(res, ownsNativeResources); } -private: - std::array<int, 4> m_rect { { 0, 0, 0, 0 } }; -}; - -Q_DECLARE_TYPEINFO(QRhiScissor, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiScissor &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiScissor &); -#endif - -class Q_GUI_EXPORT QRhiVertexInputBinding -{ -public: - enum Classification { - PerVertex, - PerInstance - }; - - QRhiVertexInputBinding() = default; - QRhiVertexInputBinding(quint32 stride, Classification cls = PerVertex, int stepRate = 1); - - quint32 stride() const { return m_stride; } - void setStride(quint32 s) { m_stride = s; } - - Classification classification() const { return m_classification; } - void setClassification(Classification c) { m_classification = c; } - - int instanceStepRate() const { return m_instanceStepRate; } - void setInstanceStepRate(int rate) { m_instanceStepRate = rate; } - -private: - quint32 m_stride = 0; - Classification m_classification = PerVertex; - int m_instanceStepRate = 1; -}; - -Q_DECLARE_TYPEINFO(QRhiVertexInputBinding, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiVertexInputBinding &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputBinding &); -#endif - -class Q_GUI_EXPORT QRhiVertexInputAttribute -{ -public: - enum Format { - Float4, - Float3, - Float2, - Float, - UNormByte4, - UNormByte2, - UNormByte, - UInt4, - UInt3, - UInt2, - UInt, - SInt4, - SInt3, - SInt2, - SInt - }; - - QRhiVertexInputAttribute() = default; - QRhiVertexInputAttribute(int binding, int location, Format format, quint32 offset, int matrixSlice = -1); - - int binding() const { return m_binding; } - void setBinding(int b) { m_binding = b; } - - int location() const { return m_location; } - void setLocation(int loc) { m_location = loc; } - - Format format() const { return m_format; } - void setFormat(Format f) { m_format = f; } - - quint32 offset() const { return m_offset; } - void setOffset(quint32 ofs) { m_offset = ofs; } - - int matrixSlice() const { return m_matrixSlice; } - void setMatrixSlice(int slice) { m_matrixSlice = slice; } - -private: - int m_binding = 0; - int m_location = 0; - Format m_format = Float4; - quint32 m_offset = 0; - int m_matrixSlice = -1; -}; - -Q_DECLARE_TYPEINFO(QRhiVertexInputAttribute, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiVertexInputAttribute &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputAttribute &); -#endif - -class Q_GUI_EXPORT QRhiVertexInputLayout -{ -public: - QRhiVertexInputLayout() = default; - - void setBindings(std::initializer_list<QRhiVertexInputBinding> list) { m_bindings = list; } - template<typename InputIterator> - void setBindings(InputIterator first, InputIterator last) + void unregisterResource(QRhiResource *res) { - m_bindings.clear(); - std::copy(first, last, std::back_inserter(m_bindings)); + resources.remove(res); } - const QRhiVertexInputBinding *cbeginBindings() const { return m_bindings.cbegin(); } - const QRhiVertexInputBinding *cendBindings() const { return m_bindings.cend(); } - const QRhiVertexInputBinding *bindingAt(int index) const { return &m_bindings.at(index); } - void setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) { m_attributes = list; } - template<typename InputIterator> - void setAttributes(InputIterator first, InputIterator last) + void addDeleteLater(QRhiResource *res) { - m_attributes.clear(); - std::copy(first, last, std::back_inserter(m_attributes)); + if (inFrame) + pendingDeleteResources.insert(res); + else + delete res; } - const QRhiVertexInputAttribute *cbeginAttributes() const { return m_attributes.cbegin(); } - const QRhiVertexInputAttribute *cendAttributes() const { return m_attributes.cend(); } - -private: - QVarLengthArray<QRhiVertexInputBinding, 8> m_bindings; - QVarLengthArray<QRhiVertexInputAttribute, 8> m_attributes; - - friend Q_GUI_EXPORT bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept; - friend Q_GUI_EXPORT size_t qHash(const QRhiVertexInputLayout &v, size_t seed) noexcept; - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); -}; - -Q_DECLARE_TYPEINFO(QRhiVertexInputLayout, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiVertexInputLayout &v, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); -#endif - -class Q_GUI_EXPORT QRhiShaderStage -{ -public: - enum Type { - Vertex, - Fragment, - Compute - }; - - QRhiShaderStage() = default; - QRhiShaderStage(Type type, const QShader &shader, - QShader::Variant v = QShader::StandardShader); - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - QShader shader() const { return m_shader; } - void setShader(const QShader &s) { m_shader = s; } - - QShader::Variant shaderVariant() const { return m_shaderVariant; } - void setShaderVariant(QShader::Variant v) { m_shaderVariant = v; } - -private: - Type m_type = Vertex; - QShader m_shader; - QShader::Variant m_shaderVariant = QShader::StandardShader; -}; -Q_DECLARE_TYPEINFO(QRhiShaderStage, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiShaderStage &s, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderStage &); -#endif - -using QRhiGraphicsShaderStage = QRhiShaderStage; - -class Q_GUI_EXPORT QRhiShaderResourceBinding -{ -public: - enum Type { - UniformBuffer, - SampledTexture, - ImageLoad, - ImageStore, - ImageLoadStore, - BufferLoad, - BufferStore, - BufferLoadStore - }; - - enum StageFlag { - VertexStage = 1 << 0, - FragmentStage = 1 << 1, - ComputeStage = 1 << 2 - }; - Q_DECLARE_FLAGS(StageFlags, StageFlag) - - QRhiShaderResourceBinding() = default; - - bool isLayoutCompatible(const QRhiShaderResourceBinding &other) const; - - static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size); - static QRhiShaderResourceBinding uniformBufferWithDynamicOffset(int binding, StageFlags stage, QRhiBuffer *buf, int size); - - static QRhiShaderResourceBinding sampledTexture(int binding, StageFlags stage, QRhiTexture *tex, QRhiSampler *sampler); - - struct TextureAndSampler { - QRhiTexture *tex; - QRhiSampler *sampler; - }; - static QRhiShaderResourceBinding sampledTextures(int binding, StageFlags stage, int count, const TextureAndSampler *texSamplers); - - static QRhiShaderResourceBinding imageLoad(int binding, StageFlags stage, QRhiTexture *tex, int level); - static QRhiShaderResourceBinding imageStore(int binding, StageFlags stage, QRhiTexture *tex, int level); - static QRhiShaderResourceBinding imageLoadStore(int binding, StageFlags stage, QRhiTexture *tex, int level); - - static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size); - static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size); - static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf, int offset, int size); - - struct Data + void addCleanupCallback(const QRhi::CleanupCallback &callback) { - int binding; - QRhiShaderResourceBinding::StageFlags stage; - QRhiShaderResourceBinding::Type type; - struct UniformBufferData { - QRhiBuffer *buf; - int offset; - int maybeSize; - bool hasDynamicOffset; - }; - static const int MAX_TEX_SAMPLER_ARRAY_SIZE = 16; - struct SampledTextureData { - int count; - TextureAndSampler texSamplers[MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct StorageImageData { - QRhiTexture *tex; - int level; - }; - struct StorageBufferData { - QRhiBuffer *buf; - int offset; - int maybeSize; - }; - union { - UniformBufferData ubuf; - SampledTextureData stex; - StorageImageData simage; - StorageBufferData sbuf; - } u; - }; - - Data *data() { return &d; } - const Data *data() const { return &d; } - -private: - Data d; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBinding::StageFlags) - -Q_DECLARE_TYPEINFO(QRhiShaderResourceBinding, Q_PRIMITIVE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiShaderResourceBinding &b, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBinding &); -#endif - -class Q_GUI_EXPORT QRhiColorAttachment -{ -public: - QRhiColorAttachment() = default; - QRhiColorAttachment(QRhiTexture *texture); - QRhiColorAttachment(QRhiRenderBuffer *renderBuffer); - - QRhiTexture *texture() const { return m_texture; } - void setTexture(QRhiTexture *tex) { m_texture = tex; } - - QRhiRenderBuffer *renderBuffer() const { return m_renderBuffer; } - void setRenderBuffer(QRhiRenderBuffer *rb) { m_renderBuffer = rb; } - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } - - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } - - QRhiTexture *resolveTexture() const { return m_resolveTexture; } - void setResolveTexture(QRhiTexture *tex) { m_resolveTexture = tex; } - - int resolveLayer() const { return m_resolveLayer; } - void setResolveLayer(int layer) { m_resolveLayer = layer; } - - int resolveLevel() const { return m_resolveLevel; } - void setResolveLevel(int level) { m_resolveLevel = level; } - -private: - QRhiTexture *m_texture = nullptr; - QRhiRenderBuffer *m_renderBuffer = nullptr; - int m_layer = 0; - int m_level = 0; - QRhiTexture *m_resolveTexture = nullptr; - int m_resolveLayer = 0; - int m_resolveLevel = 0; -}; - -Q_DECLARE_TYPEINFO(QRhiColorAttachment, Q_RELOCATABLE_TYPE); + cleanupCallbacks.append(callback); + } -class Q_GUI_EXPORT QRhiTextureRenderTargetDescription -{ -public: - QRhiTextureRenderTargetDescription() = default; - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment); - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiRenderBuffer *depthStencilBuffer); - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiTexture *depthTexture); - - void setColorAttachments(std::initializer_list<QRhiColorAttachment> list) { m_colorAttachments = list; } - template<typename InputIterator> - void setColorAttachments(InputIterator first, InputIterator last) + void addCleanupCallback(const void *key, const QRhi::CleanupCallback &callback) { - m_colorAttachments.clear(); - std::copy(first, last, std::back_inserter(m_colorAttachments)); + keyedCleanupCallbacks[key] = callback; } - const QRhiColorAttachment *cbeginColorAttachments() const { return m_colorAttachments.cbegin(); } - const QRhiColorAttachment *cendColorAttachments() const { return m_colorAttachments.cend(); } - const QRhiColorAttachment *colorAttachmentAt(int index) const { return &m_colorAttachments.at(index); } - - QRhiRenderBuffer *depthStencilBuffer() const { return m_depthStencilBuffer; } - void setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) { m_depthStencilBuffer = renderBuffer; } - - QRhiTexture *depthTexture() const { return m_depthTexture; } - void setDepthTexture(QRhiTexture *texture) { m_depthTexture = texture; } - -private: - QVarLengthArray<QRhiColorAttachment, 8> m_colorAttachments; - QRhiRenderBuffer *m_depthStencilBuffer = nullptr; - QRhiTexture *m_depthTexture = nullptr; -}; - -Q_DECLARE_TYPEINFO(QRhiTextureRenderTargetDescription, Q_RELOCATABLE_TYPE); -class Q_GUI_EXPORT QRhiTextureSubresourceUploadDescription -{ -public: - QRhiTextureSubresourceUploadDescription() = default; - explicit QRhiTextureSubresourceUploadDescription(const QImage &image); - QRhiTextureSubresourceUploadDescription(const void *data, int size); - explicit QRhiTextureSubresourceUploadDescription(const QByteArray &data); - - QImage image() const { return m_image; } - void setImage(const QImage &image) { m_image = image; } - - QByteArray data() const { return m_data; } - void setData(const QByteArray &data) { m_data = data; } - - quint32 dataStride() const { return m_dataStride; } - void setDataStride(quint32 stride) { m_dataStride = stride; } - - QPoint destinationTopLeft() const { return m_destinationTopLeft; } - void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } - - QSize sourceSize() const { return m_sourceSize; } - void setSourceSize(const QSize &size) { m_sourceSize = size; } - - QPoint sourceTopLeft() const { return m_sourceTopLeft; } - void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } - -private: - QImage m_image; - QByteArray m_data; - quint32 m_dataStride = 0; - QPoint m_destinationTopLeft; - QSize m_sourceSize; - QPoint m_sourceTopLeft; -}; - -Q_DECLARE_TYPEINFO(QRhiTextureSubresourceUploadDescription, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QRhiTextureUploadEntry -{ -public: - QRhiTextureUploadEntry() = default; - QRhiTextureUploadEntry(int layer, int level, const QRhiTextureSubresourceUploadDescription &desc); - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } - - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } - - QRhiTextureSubresourceUploadDescription description() const { return m_desc; } - void setDescription(const QRhiTextureSubresourceUploadDescription &desc) { m_desc = desc; } - -private: - int m_layer = 0; - int m_level = 0; - QRhiTextureSubresourceUploadDescription m_desc; -}; + void removeCleanupCallback(const void *key) + { + keyedCleanupCallbacks.remove(key); + } -Q_DECLARE_TYPEINFO(QRhiTextureUploadEntry, Q_RELOCATABLE_TYPE); + bool sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps); + bool sanityCheckShaderResourceBindings(QRhiShaderResourceBindings *srb); + void updateLayoutDesc(QRhiShaderResourceBindings *srb); -class Q_GUI_EXPORT QRhiTextureUploadDescription -{ -public: - QRhiTextureUploadDescription() = default; - QRhiTextureUploadDescription(const QRhiTextureUploadEntry &entry); - QRhiTextureUploadDescription(std::initializer_list<QRhiTextureUploadEntry> list); + quint32 pipelineCacheRhiId() const + { + const quint32 ver = (QT_VERSION_MAJOR << 16) | (QT_VERSION_MINOR << 8) | (QT_VERSION_PATCH); + return (quint32(implType) << 24) | ver; + } - void setEntries(std::initializer_list<QRhiTextureUploadEntry> list) { m_entries = list; } - template<typename InputIterator> - void setEntries(InputIterator first, InputIterator last) + void pipelineCreationStart() { - m_entries.clear(); - std::copy(first, last, std::back_inserter(m_entries)); + pipelineCreationTimer.start(); } - const QRhiTextureUploadEntry *cbeginEntries() const { return m_entries.cbegin(); } - const QRhiTextureUploadEntry *cendEntries() const { return m_entries.cend(); } -private: - QVarLengthArray<QRhiTextureUploadEntry, 16> m_entries; -}; + void pipelineCreationEnd() + { + accumulatedPipelineCreationTime += pipelineCreationTimer.elapsed(); + } -Q_DECLARE_TYPEINFO(QRhiTextureUploadDescription, Q_RELOCATABLE_TYPE); + qint64 totalPipelineCreationTime() const + { + return accumulatedPipelineCreationTime; + } -class Q_GUI_EXPORT QRhiTextureCopyDescription -{ -public: - QRhiTextureCopyDescription() = default; + QRhiVertexInputAttribute::Format shaderDescVariableFormatToVertexInputFormat(QShaderDescription::VariableType type) const; + quint32 byteSizePerVertexForVertexInputFormat(QRhiVertexInputAttribute::Format format) const; - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + static const QRhiShaderResourceBinding::Data *shaderResourceBindingData(const QRhiShaderResourceBinding &binding) + { + return &binding.d; + } - int sourceLayer() const { return m_sourceLayer; } - void setSourceLayer(int layer) { m_sourceLayer = layer; } + static QRhiShaderResourceBinding::Data *shaderResourceBindingData(QRhiShaderResourceBinding &binding) + { + return &binding.d; + } - int sourceLevel() const { return m_sourceLevel; } - void setSourceLevel(int level) { m_sourceLevel = level; } + static bool sortedBindingLessThan(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) + { + return a.d.binding < b.d.binding; + } - QPoint sourceTopLeft() const { return m_sourceTopLeft; } - void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } + int effectiveSampleCount(int sampleCount) const; - int destinationLayer() const { return m_destinationLayer; } - void setDestinationLayer(int layer) { m_destinationLayer = layer; } + QRhi *q; - int destinationLevel() const { return m_destinationLevel; } - void setDestinationLevel(int level) { m_destinationLevel = level; } + static const int MAX_SHADER_CACHE_ENTRIES = 128; - QPoint destinationTopLeft() const { return m_destinationTopLeft; } - void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } + bool debugMarkers = false; + int currentFrameSlot = 0; // for vk, mtl, and similar. unused by gl and d3d11. + bool inFrame = false; private: - QSize m_pixelSize; - int m_sourceLayer = 0; - int m_sourceLevel = 0; - QPoint m_sourceTopLeft; - int m_destinationLayer = 0; - int m_destinationLevel = 0; - QPoint m_destinationTopLeft; -}; - -Q_DECLARE_TYPEINFO(QRhiTextureCopyDescription, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QRhiReadbackDescription -{ -public: - QRhiReadbackDescription() = default; - QRhiReadbackDescription(QRhiTexture *texture); - - QRhiTexture *texture() const { return m_texture; } - void setTexture(QRhiTexture *tex) { m_texture = tex; } - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } + QRhi::Implementation implType; + QThread *implThread; + QVarLengthArray<QRhiResourceUpdateBatch *, 4> resUpdPool; + quint64 resUpdPoolMap = 0; + int lastResUpdIdx = -1; + QHash<QRhiResource *, bool> resources; + QSet<QRhiResource *> pendingDeleteResources; + QVarLengthArray<QRhi::CleanupCallback, 4> cleanupCallbacks; + QHash<const void *, QRhi::CleanupCallback> keyedCleanupCallbacks; + QElapsedTimer pipelineCreationTimer; + qint64 accumulatedPipelineCreationTime = 0; - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } - -private: - QRhiTexture *m_texture = nullptr; - int m_layer = 0; - int m_level = 0; + friend class QRhi; + friend class QRhiResourceUpdateBatchPrivate; }; -Q_DECLARE_TYPEINFO(QRhiReadbackDescription, Q_RELOCATABLE_TYPE); - -struct Q_GUI_EXPORT QRhiNativeHandles +enum QRhiTargetRectBoundMode { + UnBounded, + Bounded }; -class Q_GUI_EXPORT QRhiResource +template<QRhiTargetRectBoundMode boundingMode, typename T, size_t N> +bool qrhi_toTopLeftRenderTargetRect(const QSize &outputSize, const std::array<T, N> &r, + T *x, T *y, T *w, T *h) { -public: - enum Type { - Buffer, - Texture, - Sampler, - RenderBuffer, - RenderPassDescriptor, - RenderTarget, - TextureRenderTarget, - ShaderResourceBindings, - GraphicsPipeline, - SwapChain, - ComputePipeline, - CommandBuffer - }; - - virtual ~QRhiResource(); + // x,y are bottom-left in QRhiScissor and QRhiViewport but top-left in + // Vulkan/Metal/D3D. Our input is an OpenGL-style scissor rect where both + // negative x or y, and partly or completely out of bounds rects are + // allowed. The only thing the input here cannot have is a negative width + // or height. We must handle all other input gracefully, clamping to a zero + // width or height rect in the worst case, and ensuring the resulting rect + // is inside the rendertarget's bounds because some APIs' validation/debug + // layers are allergic to out of bounds scissor rects. - virtual Type resourceType() const = 0; + const T outputWidth = outputSize.width(); + const T outputHeight = outputSize.height(); + const T inputWidth = r[2]; + const T inputHeight = r[3]; - virtual void destroy() = 0; + if (inputWidth < 0 || inputHeight < 0) + return false; - void deleteLater(); + *x = r[0]; + *y = outputHeight - (r[1] + inputHeight); + *w = inputWidth; + *h = inputHeight; - QByteArray name() const; - void setName(const QByteArray &name); + if (boundingMode == Bounded) { + const T widthOffset = *x < 0 ? -*x : 0; + const T heightOffset = *y < 0 ? -*y : 0; + *w = *x < outputWidth ? qMax<T>(0, inputWidth - widthOffset) : 0; + *h = *y < outputHeight ? qMax<T>(0, inputHeight - heightOffset) : 0; - quint64 globalResourceId() const; + if (outputWidth > 0) + *x = qBound<T>(0, *x, outputWidth - 1); + if (outputHeight > 0) + *y = qBound<T>(0, *y, outputHeight - 1); -protected: - QRhiResource(QRhiImplementation *rhi); - Q_DISABLE_COPY(QRhiResource) - friend class QRhiImplementation; - QRhiImplementation *m_rhi = nullptr; - quint64 m_id; - QByteArray m_objectName; -}; + if (*x + *w > outputWidth) + *w = qMax<T>(0, outputWidth - *x); + if (*y + *h > outputHeight) + *h = qMax<T>(0, outputHeight - *y); + } + return true; +} -class Q_GUI_EXPORT QRhiBuffer : public QRhiResource +struct QRhiBufferDataPrivate { -public: - enum Type { - Immutable, - Static, - Dynamic - }; - - enum UsageFlag { - VertexBuffer = 1 << 0, - IndexBuffer = 1 << 1, - UniformBuffer = 1 << 2, - StorageBuffer = 1 << 3 - }; - Q_DECLARE_FLAGS(UsageFlags, UsageFlag) - - struct NativeBuffer { - const void *objects[3]; - int slotCount; - }; - - QRhiResource::Type resourceType() const override; - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - UsageFlags usage() const { return m_usage; } - void setUsage(UsageFlags u) { m_usage = u; } - - int size() const { return m_size; } - void setSize(int sz) { m_size = sz; } - - virtual bool create() = 0; - - virtual NativeBuffer nativeBuffer(); - - virtual char *beginFullDynamicBufferUpdateForCurrentFrame(); - virtual void endFullDynamicBufferUpdateForCurrentFrame(); - -protected: - QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, int size_); - Type m_type; - UsageFlags m_usage; - int m_size; + Q_DISABLE_COPY_MOVE(QRhiBufferDataPrivate) + QRhiBufferDataPrivate() { } + ~QRhiBufferDataPrivate() { delete[] largeData; } + int ref = 1; + quint32 size = 0; + quint32 largeAlloc = 0; + char *largeData = nullptr; + static constexpr quint32 SMALL_DATA_SIZE = 1024; + char data[SMALL_DATA_SIZE]; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiBuffer::UsageFlags) - -class Q_GUI_EXPORT QRhiTexture : public QRhiResource +// no detach-with-contents, no atomic refcount, no shrink +class QRhiBufferData { public: - enum Flag { - RenderTarget = 1 << 0, - CubeMap = 1 << 2, - MipMapped = 1 << 3, - sRGB = 1 << 4, - UsedAsTransferSource = 1 << 5, - UsedWithGenerateMips = 1 << 6, - UsedWithLoadStore = 1 << 7, - UsedAsCompressedAtlas = 1 << 8, - ExternalOES = 1 << 9, - ThreeDimensional = 1 << 10 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum Format { - UnknownFormat, - - RGBA8, - BGRA8, - R8, - RG8, - R16, - RG16, - RED_OR_ALPHA8, - - RGBA16F, - RGBA32F, - R16F, - R32F, - - D16, - D24, - D24S8, - D32F, - - BC1, - BC2, - BC3, - BC4, - BC5, - BC6H, - BC7, - - ETC2_RGB8, - ETC2_RGB8A1, - ETC2_RGBA8, - - ASTC_4x4, - ASTC_5x4, - ASTC_5x5, - ASTC_6x5, - ASTC_6x6, - ASTC_8x5, - ASTC_8x6, - ASTC_8x8, - ASTC_10x5, - ASTC_10x6, - ASTC_10x8, - ASTC_10x10, - ASTC_12x10, - ASTC_12x12 - }; - - struct NativeTexture { - quint64 object; - int layout; - }; - - QRhiResource::Type resourceType() const override; - - Format format() const { return m_format; } - void setFormat(Format fmt) { m_format = fmt; } - - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } - - int depth() const { return m_depth; } - void setDepth(int depth) { m_depth = depth; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } - - virtual bool create() = 0; - virtual NativeTexture nativeTexture(); - virtual bool createFrom(NativeTexture src); - virtual void setNativeLayout(int layout); - -protected: - QRhiTexture(QRhiImplementation *rhi, Format format_, const QSize &pixelSize_, int depth_, - int sampleCount_, Flags flags_); - Format m_format; - QSize m_pixelSize; - int m_depth; - int m_sampleCount; - Flags m_flags; + QRhiBufferData() = default; + ~QRhiBufferData() + { + if (d && !--d->ref) + delete d; + } + QRhiBufferData(const QRhiBufferData &other) + : d(other.d) + { + if (d) + d->ref += 1; + } + QRhiBufferData &operator=(const QRhiBufferData &other) + { + if (d == other.d) + return *this; + if (other.d) + other.d->ref += 1; + if (d && !--d->ref) + delete d; + d = other.d; + return *this; + } + const char *constData() const + { + return d->size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE ? d->data : d->largeData; + } + quint32 size() const + { + return d->size; + } + void assign(const char *s, quint32 size) + { + if (!d) { + d = new QRhiBufferDataPrivate; + } else if (d->ref != 1) { + d->ref -= 1; + d = new QRhiBufferDataPrivate; + } + d->size = size; + if (size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE) { + memcpy(d->data, s, size); + } else { + if (d->largeAlloc < size) { + delete[] d->largeData; + d->largeAlloc = size; + d->largeData = new char[size]; + } + memcpy(d->largeData, s, size); + } + } +private: + QRhiBufferDataPrivate *d = nullptr; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTexture::Flags) - -class Q_GUI_EXPORT QRhiSampler : public QRhiResource -{ -public: - enum Filter { - None, - Nearest, - Linear - }; - - enum AddressMode { - Repeat, - ClampToEdge, - Mirror, - }; - - enum CompareOp { - Never, - Less, - Equal, - LessOrEqual, - Greater, - NotEqual, - GreaterOrEqual, - Always - }; - - QRhiResource::Type resourceType() const override; - - Filter magFilter() const { return m_magFilter; } - void setMagFilter(Filter f) { m_magFilter = f; } - - Filter minFilter() const { return m_minFilter; } - void setMinFilter(Filter f) { m_minFilter = f; } - - Filter mipmapMode() const { return m_mipmapMode; } - void setMipmapMode(Filter f) { m_mipmapMode = f; } - - AddressMode addressU() const { return m_addressU; } - void setAddressU(AddressMode mode) { m_addressU = mode; } - - AddressMode addressV() const { return m_addressV; } - void setAddressV(AddressMode mode) { m_addressV = mode; } - - AddressMode addressW() const { return m_addressW; } - void setAddressW(AddressMode mode) { m_addressW = mode; } +Q_DECLARE_TYPEINFO(QRhiBufferData, Q_RELOCATABLE_TYPE); - CompareOp textureCompareOp() const { return m_compareOp; } - void setTextureCompareOp(CompareOp op) { m_compareOp = op; } - - virtual bool create() = 0; - -protected: - QRhiSampler(QRhiImplementation *rhi, - Filter magFilter_, Filter minFilter_, Filter mipmapMode_, - AddressMode u_, AddressMode v_, AddressMode w_); - Filter m_magFilter; - Filter m_minFilter; - Filter m_mipmapMode; - AddressMode m_addressU; - AddressMode m_addressV; - AddressMode m_addressW; - CompareOp m_compareOp; -}; - -class Q_GUI_EXPORT QRhiRenderBuffer : public QRhiResource +class QRhiResourceUpdateBatchPrivate { public: - enum Type { - DepthStencil, - Color - }; - - enum Flag { - UsedWithSwapChainOnly = 1 << 0 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - struct NativeRenderBuffer { - quint64 object; - }; - - QRhiResource::Type resourceType() const override; - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } - - Flags flags() const { return m_flags; } - void setFlags(Flags h) { m_flags = h; } - - virtual bool create() = 0; - virtual bool createFrom(NativeRenderBuffer src); - - virtual QRhiTexture::Format backingFormat() const = 0; + struct BufferOp { + enum Type { + DynamicUpdate, + StaticUpload, + Read + }; + Type type; + QRhiBuffer *buf; + quint32 offset; + QRhiBufferData data; + quint32 readSize; + QRhiReadbackResult *result; + + static BufferOp dynamicUpdate(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + BufferOp op = {}; + op.type = DynamicUpdate; + op.buf = buf; + op.offset = offset; + const int effectiveSize = size ? size : buf->size(); + op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); + return op; + } + + static void changeToDynamicUpdate(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + op->type = DynamicUpdate; + op->buf = buf; + op->offset = offset; + const int effectiveSize = size ? size : buf->size(); + op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); + } + + static BufferOp staticUpload(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + BufferOp op = {}; + op.type = StaticUpload; + op.buf = buf; + op.offset = offset; + const int effectiveSize = size ? size : buf->size(); + op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); + return op; + } + + static void changeToStaticUpload(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + op->type = StaticUpload; + op->buf = buf; + op->offset = offset; + const int effectiveSize = size ? size : buf->size(); + op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); + } + + static BufferOp read(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result) + { + BufferOp op = {}; + op.type = Read; + op.buf = buf; + op.offset = offset; + op.readSize = size; + op.result = result; + return op; + } + }; + + struct TextureOp { + enum Type { + Upload, + Copy, + Read, + GenMips + }; + Type type; + QRhiTexture *dst; + // Specifying multiple uploads for a subresource must be supported. + // In the backend this can then end up, where applicable, as a + // single, batched copy operation with only one set of barriers. + // This helps when doing for example glyph cache fills. + using MipLevelUploadList = std::array<QVector<QRhiTextureSubresourceUploadDescription>, QRhi::MAX_MIP_LEVELS>; + QVarLengthArray<MipLevelUploadList, 6> subresDesc; + QRhiTexture *src; + QRhiTextureCopyDescription desc; + QRhiReadbackDescription rb; + QRhiReadbackResult *result; + + static TextureOp upload(QRhiTexture *tex, const QRhiTextureUploadDescription &desc) + { + TextureOp op = {}; + op.type = Upload; + op.dst = tex; + int maxLayer = -1; + for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) { + if (it->layer() > maxLayer) + maxLayer = it->layer(); + } + op.subresDesc.resize(maxLayer + 1); + for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) + op.subresDesc[it->layer()][it->level()].append(it->description()); + return op; + } + + static TextureOp copy(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc) + { + TextureOp op = {}; + op.type = Copy; + op.dst = dst; + op.src = src; + op.desc = desc; + return op; + } + + static TextureOp read(const QRhiReadbackDescription &rb, QRhiReadbackResult *result) + { + TextureOp op = {}; + op.type = Read; + op.rb = rb; + op.result = result; + return op; + } + + static TextureOp genMips(QRhiTexture *tex) + { + TextureOp op = {}; + op.type = GenMips; + op.dst = tex; + return op; + } + }; + + int activeBufferOpCount = 0; // this is the real number of used elements in bufferOps, not bufferOps.count() + static const int BUFFER_OPS_STATIC_ALLOC = 1024; + QVarLengthArray<BufferOp, BUFFER_OPS_STATIC_ALLOC> bufferOps; + + int activeTextureOpCount = 0; // this is the real number of used elements in textureOps, not textureOps.count() + static const int TEXTURE_OPS_STATIC_ALLOC = 256; + QVarLengthArray<TextureOp, TEXTURE_OPS_STATIC_ALLOC> textureOps; + + QRhiResourceUpdateBatch *q = nullptr; + QRhiImplementation *rhi = nullptr; + int poolIndex = -1; + + void free(); + void merge(QRhiResourceUpdateBatchPrivate *other); + bool hasOptimalCapacity() const; + void trimOpLists(); -protected: - QRhiRenderBuffer(QRhiImplementation *rhi, Type type_, const QSize &pixelSize_, - int sampleCount_, Flags flags_, QRhiTexture::Format backingFormatHint_); - Type m_type; - QSize m_pixelSize; - int m_sampleCount; - Flags m_flags; - QRhiTexture::Format m_backingFormatHint; + static QRhiResourceUpdateBatchPrivate *get(QRhiResourceUpdateBatch *b) { return b->d; } }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiRenderBuffer::Flags) - -class Q_GUI_EXPORT QRhiRenderPassDescriptor : public QRhiResource +template<typename T> +struct QRhiBatchedBindings { -public: - QRhiResource::Type resourceType() const override; - - virtual bool isCompatible(const QRhiRenderPassDescriptor *other) const = 0; - virtual const QRhiNativeHandles *nativeHandles(); - - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const = 0; - -protected: - QRhiRenderPassDescriptor(QRhiImplementation *rhi); -}; + void feed(int binding, T resource) { // binding must be strictly increasing + if (curBinding == -1 || binding > curBinding + 1) { + finish(); + curBatch.startBinding = binding; + curBatch.resources.clear(); + curBatch.resources.append(resource); + } else { + Q_ASSERT(binding == curBinding + 1); + curBatch.resources.append(resource); + } + curBinding = binding; + } -class Q_GUI_EXPORT QRhiRenderTarget : public QRhiResource -{ -public: - QRhiResource::Type resourceType() const override; + bool finish() { + if (!curBatch.resources.isEmpty()) + batches.append(curBatch); + return !batches.isEmpty(); + } - virtual QSize pixelSize() const = 0; - virtual float devicePixelRatio() const = 0; - virtual int sampleCount() const = 0; + void clear() { + batches.clear(); + curBatch.resources.clear(); + curBinding = -1; + } - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + struct Batch { + uint startBinding; + QVarLengthArray<T, 4> resources; -protected: - QRhiRenderTarget(QRhiImplementation *rhi); - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; -}; + bool operator==(const Batch &other) const + { + return startBinding == other.startBinding && resources == other.resources; + } -class Q_GUI_EXPORT QRhiTextureRenderTarget : public QRhiRenderTarget -{ -public: - enum Flag { - PreserveColorContents = 1 << 0, - PreserveDepthStencilContents = 1 << 1 + bool operator!=(const Batch &other) const + { + return !operator==(other); + } }; - Q_DECLARE_FLAGS(Flags, Flag) - - QRhiResource::Type resourceType() const override; - QRhiTextureRenderTargetDescription description() const { return m_desc; } - void setDescription(const QRhiTextureRenderTargetDescription &desc) { m_desc = desc; } + QVarLengthArray<Batch, 4> batches; // sorted by startBinding - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; + bool operator==(const QRhiBatchedBindings<T> &other) const + { + return batches == other.batches; + } - virtual bool create() = 0; + bool operator!=(const QRhiBatchedBindings<T> &other) const + { + return !operator==(other); + } -protected: - QRhiTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc_, Flags flags_); - QRhiTextureRenderTargetDescription m_desc; - Flags m_flags; +private: + Batch curBatch; + int curBinding = -1; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTextureRenderTarget::Flags) - -class Q_GUI_EXPORT QRhiShaderResourceBindings : public QRhiResource +class QRhiGlobalObjectIdGenerator { public: - QRhiResource::Type resourceType() const override; - - void setBindings(std::initializer_list<QRhiShaderResourceBinding> list) { m_bindings = list; } - - template<typename InputIterator> - void setBindings(InputIterator first, InputIterator last) - { - m_bindings.clear(); - std::copy(first, last, std::back_inserter(m_bindings)); - } - - const QRhiShaderResourceBinding *cbeginBindings() const { return m_bindings.cbegin(); } - const QRhiShaderResourceBinding *cendBindings() const { return m_bindings.cend(); } - - bool isLayoutCompatible(const QRhiShaderResourceBindings *other) const; - - virtual bool create() = 0; - -protected: - QRhiShaderResourceBindings(QRhiImplementation *rhi); - QVarLengthArray<QRhiShaderResourceBinding, 16> m_bindings; - uint m_layoutDescHash = 0; - QVarLengthArray<uint, 16 * 3> m_layoutDesc; - friend class QRhiImplementation; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); +#ifdef Q_ATOMIC_INT64_IS_SUPPORTED + using Type = quint64; +#else + using Type = quint32; #endif + static Type newId(); }; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); -#endif - -class Q_GUI_EXPORT QRhiGraphicsPipeline : public QRhiResource +class QRhiPassResourceTracker { public: - enum Flag { - UsesBlendConstants = 1 << 0, - UsesStencilRef = 1 << 1, - UsesScissor = 1 << 2, - CompileShadersWithDebugInfo = 1 << 3 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum Topology { - Triangles, - TriangleStrip, - TriangleFan, - Lines, - LineStrip, - Points - }; + bool isEmpty() const; + void reset(); - enum CullMode { - None, - Front, - Back + struct UsageState { + int layout; + int access; + int stage; }; - enum FrontFace { - CCW, - CW + enum BufferStage { + BufVertexInputStage, + BufVertexStage, + BufTCStage, + BufTEStage, + BufFragmentStage, + BufComputeStage, + BufGeometryStage }; - enum ColorMaskComponent { - R = 1 << 0, - G = 1 << 1, - B = 1 << 2, - A = 1 << 3 - }; - Q_DECLARE_FLAGS(ColorMask, ColorMaskComponent) - - enum BlendFactor { - Zero, - One, - SrcColor, - OneMinusSrcColor, - DstColor, - OneMinusDstColor, - SrcAlpha, - OneMinusSrcAlpha, - DstAlpha, - OneMinusDstAlpha, - ConstantColor, - OneMinusConstantColor, - ConstantAlpha, - OneMinusConstantAlpha, - SrcAlphaSaturate, - Src1Color, - OneMinusSrc1Color, - Src1Alpha, - OneMinusSrc1Alpha + enum BufferAccess { + BufVertexInput, + BufIndexRead, + BufUniformRead, + BufStorageLoad, + BufStorageStore, + BufStorageLoadStore }; - enum BlendOp { - Add, - Subtract, - ReverseSubtract, - Min, - Max - }; + void registerBuffer(QRhiBuffer *buf, int slot, BufferAccess *access, BufferStage *stage, + const UsageState &state); - struct TargetBlend { - ColorMask colorWrite = ColorMask(0xF); // R | G | B | A - bool enable = false; - BlendFactor srcColor = One; - BlendFactor dstColor = OneMinusSrcAlpha; - BlendOp opColor = Add; - BlendFactor srcAlpha = One; - BlendFactor dstAlpha = OneMinusSrcAlpha; - BlendOp opAlpha = Add; + enum TextureStage { + TexVertexStage, + TexTCStage, + TexTEStage, + TexFragmentStage, + TexColorOutputStage, + TexDepthOutputStage, + TexComputeStage, + TexGeometryStage }; - enum CompareOp { - Never, - Less, - Equal, - LessOrEqual, - Greater, - NotEqual, - GreaterOrEqual, - Always + enum TextureAccess { + TexSample, + TexColorOutput, + TexDepthOutput, + TexStorageLoad, + TexStorageStore, + TexStorageLoadStore }; - enum StencilOp { - StencilZero, - Keep, - Replace, - IncrementAndClamp, - DecrementAndClamp, - Invert, - IncrementAndWrap, - DecrementAndWrap - }; + void registerTexture(QRhiTexture *tex, TextureAccess *access, TextureStage *stage, + const UsageState &state); - struct StencilOpState { - StencilOp failOp = Keep; - StencilOp depthFailOp = Keep; - StencilOp passOp = Keep; - CompareOp compareOp = Always; + struct Buffer { + int slot; + BufferAccess access; + BufferStage stage; + UsageState stateAtPassBegin; }; - QRhiResource::Type resourceType() const override; - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - Topology topology() const { return m_topology; } - void setTopology(Topology t) { m_topology = t; } - - CullMode cullMode() const { return m_cullMode; } - void setCullMode(CullMode mode) { m_cullMode = mode; } - - FrontFace frontFace() const { return m_frontFace; } - void setFrontFace(FrontFace f) { m_frontFace = f; } - - void setTargetBlends(std::initializer_list<TargetBlend> list) { m_targetBlends = list; } - template<typename InputIterator> - void setTargetBlends(InputIterator first, InputIterator last) - { - m_targetBlends.clear(); - std::copy(first, last, std::back_inserter(m_targetBlends)); - } - const TargetBlend *cbeginTargetBlends() const { return m_targetBlends.cbegin(); } - const TargetBlend *cendTargetBlends() const { return m_targetBlends.cend(); } - - bool hasDepthTest() const { return m_depthTest; } - void setDepthTest(bool enable) { m_depthTest = enable; } - - bool hasDepthWrite() const { return m_depthWrite; } - void setDepthWrite(bool enable) { m_depthWrite = enable; } - - CompareOp depthOp() const { return m_depthOp; } - void setDepthOp(CompareOp op) { m_depthOp = op; } - - bool hasStencilTest() const { return m_stencilTest; } - void setStencilTest(bool enable) { m_stencilTest = enable; } - - StencilOpState stencilFront() const { return m_stencilFront; } - void setStencilFront(const StencilOpState &state) { m_stencilFront = state; } - - StencilOpState stencilBack() const { return m_stencilBack; } - void setStencilBack(const StencilOpState &state) { m_stencilBack = state; } - - quint32 stencilReadMask() const { return m_stencilReadMask; } - void setStencilReadMask(quint32 mask) { m_stencilReadMask = mask; } - - quint32 stencilWriteMask() const { return m_stencilWriteMask; } - void setStencilWriteMask(quint32 mask) { m_stencilWriteMask = mask; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } - - float lineWidth() const { return m_lineWidth; } - void setLineWidth(float width) { m_lineWidth = width; } + using BufferIterator = QHash<QRhiBuffer *, Buffer>::const_iterator; + BufferIterator cbeginBuffers() const { return m_buffers.cbegin(); } + BufferIterator cendBuffers() const { return m_buffers.cend(); } - int depthBias() const { return m_depthBias; } - void setDepthBias(int bias) { m_depthBias = bias; } - - float slopeScaledDepthBias() const { return m_slopeScaledDepthBias; } - void setSlopeScaledDepthBias(float bias) { m_slopeScaledDepthBias = bias; } - - void setShaderStages(std::initializer_list<QRhiShaderStage> list) { m_shaderStages = list; } - template<typename InputIterator> - void setShaderStages(InputIterator first, InputIterator last) - { - m_shaderStages.clear(); - std::copy(first, last, std::back_inserter(m_shaderStages)); - } - const QRhiShaderStage *cbeginShaderStages() const { return m_shaderStages.cbegin(); } - const QRhiShaderStage *cendShaderStages() const { return m_shaderStages.cend(); } - - QRhiVertexInputLayout vertexInputLayout() const { return m_vertexInputLayout; } - void setVertexInputLayout(const QRhiVertexInputLayout &layout) { m_vertexInputLayout = layout; } - - QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } - void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } - - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } - - virtual bool create() = 0; - -protected: - QRhiGraphicsPipeline(QRhiImplementation *rhi); - Flags m_flags; - Topology m_topology = Triangles; - CullMode m_cullMode = None; - FrontFace m_frontFace = CCW; - QVarLengthArray<TargetBlend, 8> m_targetBlends; - bool m_depthTest = false; - bool m_depthWrite = false; - CompareOp m_depthOp = Less; - bool m_stencilTest = false; - StencilOpState m_stencilFront; - StencilOpState m_stencilBack; - quint32 m_stencilReadMask = 0xFF; - quint32 m_stencilWriteMask = 0xFF; - int m_sampleCount = 1; - float m_lineWidth = 1.0f; - int m_depthBias = 0; - float m_slopeScaledDepthBias = 0.0f; - QVarLengthArray<QRhiShaderStage, 4> m_shaderStages; - QRhiVertexInputLayout m_vertexInputLayout; - QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::Flags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::ColorMask) -Q_DECLARE_TYPEINFO(QRhiGraphicsPipeline::TargetBlend, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QRhiSwapChain : public QRhiResource -{ -public: - enum Flag { - SurfaceHasPreMulAlpha = 1 << 0, - SurfaceHasNonPreMulAlpha = 1 << 1, - sRGB = 1 << 2, - UsedAsTransferSource = 1 << 3, - NoVSync = 1 << 4, - MinimalBufferCount = 1 << 5 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - QRhiResource::Type resourceType() const override; - - QWindow *window() const { return m_window; } - void setWindow(QWindow *window) { m_window = window; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - QRhiRenderBuffer *depthStencil() const { return m_depthStencil; } - void setDepthStencil(QRhiRenderBuffer *ds) { m_depthStencil = ds; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int samples) { m_sampleCount = samples; } - - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } - - QSize currentPixelSize() const { return m_currentPixelSize; } - - virtual QRhiCommandBuffer *currentFrameCommandBuffer() = 0; - virtual QRhiRenderTarget *currentFrameRenderTarget() = 0; - virtual QSize surfacePixelSize() = 0; - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; - virtual bool createOrResize() = 0; - -protected: - QRhiSwapChain(QRhiImplementation *rhi); - QWindow *m_window = nullptr; - Flags m_flags; - QRhiRenderBuffer *m_depthStencil = nullptr; - int m_sampleCount = 1; - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; - QSize m_currentPixelSize; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiSwapChain::Flags) - -class Q_GUI_EXPORT QRhiComputePipeline : public QRhiResource -{ -public: - enum Flag { - CompileShadersWithDebugInfo = 1 << 0 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - QRhiResource::Type resourceType() const override; - virtual bool create() = 0; - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - QRhiShaderStage shaderStage() const { return m_shaderStage; } - void setShaderStage(const QRhiShaderStage &stage) { m_shaderStage = stage; } - - QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } - void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } - -protected: - QRhiComputePipeline(QRhiImplementation *rhi); - Flags m_flags; - QRhiShaderStage m_shaderStage; - QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiComputePipeline::Flags) - -class Q_GUI_EXPORT QRhiCommandBuffer : public QRhiResource -{ -public: - enum IndexFormat { - IndexUInt16, - IndexUInt32 + struct Texture { + TextureAccess access; + TextureStage stage; + UsageState stateAtPassBegin; }; - enum BeginPassFlag { - ExternalContent = 0x01, - DoNotTrackResourcesForCompute = 0x02 - }; - Q_DECLARE_FLAGS(BeginPassFlags, BeginPassFlag) - - QRhiResource::Type resourceType() const override; - - void resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates); - - void beginPass(QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates = nullptr, - BeginPassFlags flags = {}); - void endPass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); - - void setGraphicsPipeline(QRhiGraphicsPipeline *ps); - using DynamicOffset = QPair<int, quint32>; // binding, offset - void setShaderResources(QRhiShaderResourceBindings *srb = nullptr, - int dynamicOffsetCount = 0, - const DynamicOffset *dynamicOffsets = nullptr); - using VertexInput = QPair<QRhiBuffer *, quint32>; // buffer, offset - void setVertexInput(int startBinding, int bindingCount, const VertexInput *bindings, - QRhiBuffer *indexBuf = nullptr, quint32 indexOffset = 0, - IndexFormat indexFormat = IndexUInt16); - - void setViewport(const QRhiViewport &viewport); - void setScissor(const QRhiScissor &scissor); - void setBlendConstants(const QColor &c); - void setStencilRef(quint32 refValue); - - void draw(quint32 vertexCount, - quint32 instanceCount = 1, - quint32 firstVertex = 0, - quint32 firstInstance = 0); - - void drawIndexed(quint32 indexCount, - quint32 instanceCount = 1, - quint32 firstIndex = 0, - qint32 vertexOffset = 0, - quint32 firstInstance = 0); - - void debugMarkBegin(const QByteArray &name); - void debugMarkEnd(); - void debugMarkMsg(const QByteArray &msg); - - void beginComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr, BeginPassFlags flags = {}); - void endComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); - void setComputePipeline(QRhiComputePipeline *ps); - void dispatch(int x, int y, int z); - - const QRhiNativeHandles *nativeHandles(); - void beginExternal(); - void endExternal(); - -protected: - QRhiCommandBuffer(QRhiImplementation *rhi); -}; + using TextureIterator = QHash<QRhiTexture *, Texture>::const_iterator; + TextureIterator cbeginTextures() const { return m_textures.cbegin(); } + TextureIterator cendTextures() const { return m_textures.cend(); } -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiCommandBuffer::BeginPassFlags) - -struct Q_GUI_EXPORT QRhiReadbackResult -{ - std::function<void()> completed = nullptr; - QRhiTexture::Format format; - QSize pixelSize; - QByteArray data; -}; // non-movable due to the std::function - -struct Q_GUI_EXPORT QRhiBufferReadbackResult -{ - std::function<void()> completed = nullptr; - QByteArray data; -}; - -class Q_GUI_EXPORT QRhiResourceUpdateBatch -{ -public: - ~QRhiResourceUpdateBatch(); - - void release(); - - void merge(QRhiResourceUpdateBatch *other); - bool hasOptimalCapacity() const; - - void updateDynamicBuffer(QRhiBuffer *buf, int offset, int size, const void *data); - void uploadStaticBuffer(QRhiBuffer *buf, int offset, int size, const void *data); - void uploadStaticBuffer(QRhiBuffer *buf, const void *data); - void readBackBuffer(QRhiBuffer *buf, int offset, int size, QRhiBufferReadbackResult *result); - void uploadTexture(QRhiTexture *tex, const QRhiTextureUploadDescription &desc); - void uploadTexture(QRhiTexture *tex, const QImage &image); - void copyTexture(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc = QRhiTextureCopyDescription()); - void readBackTexture(const QRhiReadbackDescription &rb, QRhiReadbackResult *result); - void generateMips(QRhiTexture *tex); + static BufferStage toPassTrackerBufferStage(QRhiShaderResourceBinding::StageFlags stages); + static TextureStage toPassTrackerTextureStage(QRhiShaderResourceBinding::StageFlags stages); private: - QRhiResourceUpdateBatch(QRhiImplementation *rhi); - Q_DISABLE_COPY(QRhiResourceUpdateBatch) - QRhiResourceUpdateBatchPrivate *d; - friend class QRhiResourceUpdateBatchPrivate; - friend class QRhi; + QHash<QRhiBuffer *, Buffer> m_buffers; + QHash<QRhiTexture *, Texture> m_textures; }; -struct Q_GUI_EXPORT QRhiDriverInfo -{ - enum DeviceType { - UnknownDevice, - IntegratedDevice, - DiscreteDevice, - ExternalDevice, - VirtualDevice, - CpuDevice - }; +Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Buffer, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Texture, Q_RELOCATABLE_TYPE); - QByteArray deviceName; - quint64 deviceId = 0; - quint64 vendorId = 0; - DeviceType deviceType = UnknownDevice; -}; - -Q_DECLARE_TYPEINFO(QRhiDriverInfo, Q_RELOCATABLE_TYPE); - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDriverInfo &); -#endif - -struct Q_GUI_EXPORT QRhiInitParams -{ -}; - -class Q_GUI_EXPORT QRhi +template<typename T, int GROW = 1024> +class QRhiBackendCommandList { public: - enum Implementation { - Null, - Vulkan, - OpenGLES2, - D3D11, - Metal - }; - - enum Flag { - EnableProfiling = 1 << 0, - EnableDebugMarkers = 1 << 1, - PreferSoftwareRenderer = 1 << 2, - EnablePipelineCacheDataSave = 1 << 3 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum FrameOpResult { - FrameOpSuccess = 0, - FrameOpError, - FrameOpSwapChainOutOfDate, - FrameOpDeviceLost - }; - - enum Feature { - MultisampleTexture = 1, - MultisampleRenderBuffer, - DebugMarkers, - Timestamps, - Instancing, - CustomInstanceStepRate, - PrimitiveRestart, - NonDynamicUniformBuffers, - NonFourAlignedEffectiveIndexBufferOffset, - NPOTTextureRepeat, - RedOrAlpha8IsRed, - ElementIndexUint, - Compute, - WideLines, - VertexShaderPointSize, - BaseVertex, - BaseInstance, - TriangleFanTopology, - ReadBackNonUniformBuffer, - ReadBackNonBaseMipLevel, - TexelFetch, - RenderToNonBaseMipLevel, - IntAttributes, - ScreenSpaceDerivatives, - ReadBackAnyTextureFormat, - PipelineCacheDataLoadSave, - ImageDataStride, - RenderBufferImport, - ThreeDimensionalTextures, - RenderTo3DTextureSlice - }; - - enum BeginFrameFlag { - }; - Q_DECLARE_FLAGS(BeginFrameFlags, BeginFrameFlag) - - enum EndFrameFlag { - SkipPresent = 1 << 0 - }; - Q_DECLARE_FLAGS(EndFrameFlags, EndFrameFlag) - - enum ResourceLimit { - TextureSizeMin = 1, - TextureSizeMax, - MaxColorAttachments, - FramesInFlight, - MaxAsyncReadbackFrames, - MaxThreadGroupsPerDimension, - MaxThreadsPerThreadGroup, - MaxThreadGroupX, - MaxThreadGroupY, - MaxThreadGroupZ - }; - - ~QRhi(); - - static QRhi *create(Implementation impl, - QRhiInitParams *params, - Flags flags = {}, - QRhiNativeHandles *importDevice = nullptr); - - Implementation backend() const; - const char *backendName() const; - QRhiDriverInfo driverInfo() const; - QThread *thread() const; - - using CleanupCallback = std::function<void(QRhi *)>; - void addCleanupCallback(const CleanupCallback &callback); - void runCleanup(); - - QRhiGraphicsPipeline *newGraphicsPipeline(); - QRhiComputePipeline *newComputePipeline(); - QRhiShaderResourceBindings *newShaderResourceBindings(); - - QRhiBuffer *newBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size); - - QRhiRenderBuffer *newRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount = 1, - QRhiRenderBuffer::Flags flags = {}, - QRhiTexture::Format backingFormatHint = QRhiTexture::UnknownFormat); - - QRhiTexture *newTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int sampleCount = 1, - QRhiTexture::Flags flags = {}); - - QRhiTexture *newTexture(QRhiTexture::Format format, - int width, int height, int depth, - int sampleCount = 1, - QRhiTexture::Flags flags = {}); - - QRhiSampler *newSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler::AddressMode addressU, - QRhiSampler::AddressMode addressV, - QRhiSampler::AddressMode addressW = QRhiSampler::Repeat); - - QRhiTextureRenderTarget *newTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags = {}); - - QRhiSwapChain *newSwapChain(); - FrameOpResult beginFrame(QRhiSwapChain *swapChain, BeginFrameFlags flags = {}); - FrameOpResult endFrame(QRhiSwapChain *swapChain, EndFrameFlags flags = {}); - bool isRecordingFrame() const; - int currentFrameSlot() const; - - FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, BeginFrameFlags flags = {}); - FrameOpResult endOffscreenFrame(EndFrameFlags flags = {}); - - QRhi::FrameOpResult finish(); - - QRhiResourceUpdateBatch *nextResourceUpdateBatch(); - - QList<int> supportedSampleCounts() const; - - int ubufAlignment() const; - int ubufAligned(int v) const; - - int mipLevelsForSize(const QSize &size) const; - QSize sizeForMipLevel(int mipLevel, const QSize &baseLevelSize) const; - - bool isYUpInFramebuffer() const; - bool isYUpInNDC() const; - bool isClipDepthZeroToOne() const; - - QMatrix4x4 clipSpaceCorrMatrix() const; - - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags = {}) const; - bool isFeatureSupported(QRhi::Feature feature) const; - int resourceLimit(ResourceLimit limit) const; - - const QRhiNativeHandles *nativeHandles(); - bool makeThreadLocalNativeContextCurrent(); - - QRhiProfiler *profiler(); - - static const int MAX_MIP_LEVELS = 16; // a width and/or height of 65536 should be enough for everyone - - void releaseCachedResources(); - - bool isDeviceLost() const; - - QByteArray pipelineCacheData(); - void setPipelineCacheData(const QByteArray &data); - -protected: - QRhi(); - + QRhiBackendCommandList() = default; + ~QRhiBackendCommandList() { delete[] v; } + inline void reset() { p = 0; } + inline bool isEmpty() const { return p == 0; } + inline T &get() { + if (p == a) { + a += GROW; + T *nv = new T[a]; + if (v) { + memcpy(nv, v, p * sizeof(T)); + delete[] v; + } + v = nv; + } + return v[p++]; + } + inline void unget() { --p; } + inline T *cbegin() const { return v; } + inline T *cend() const { return v + p; } + inline T *begin() { return v; } + inline T *end() { return v + p; } private: - Q_DISABLE_COPY(QRhi) - QRhiImplementation *d = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::Flags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::BeginFrameFlags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::EndFrameFlags) + Q_DISABLE_COPY(QRhiBackendCommandList) + T *v = nullptr; + int a = 0; + int p = 0; +}; + +struct QRhiRenderTargetAttachmentTracker +{ + struct ResId { quint64 id; uint generation; }; + using ResIdList = QVarLengthArray<ResId, 8 * 2 + 1>; // color, resolve, ds + + template<typename TexType, typename RenderBufferType> + static void updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst); + + template<typename TexType, typename RenderBufferType> + static bool isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList); +}; + +inline bool operator==(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) +{ + return a.id == b.id && a.generation == b.generation; +} + +inline bool operator!=(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) +{ + return !(a == b); +} + +template<typename TexType, typename RenderBufferType> +void QRhiRenderTargetAttachmentTracker::updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst) +{ + const bool hasDepthStencil = desc.depthStencilBuffer() || desc.depthTexture(); + dst->resize(desc.colorAttachmentCount() * 2 + (hasDepthStencil ? 1 : 0)); + int n = 0; + for (auto it = desc.cbeginColorAttachments(), itEnd = desc.cendColorAttachments(); it != itEnd; ++it, ++n) { + const QRhiColorAttachment &colorAtt(*it); + if (colorAtt.texture()) { + TexType *texD = QRHI_RES(TexType, colorAtt.texture()); + (*dst)[n] = { texD->globalResourceId(), texD->generation }; + } else if (colorAtt.renderBuffer()) { + RenderBufferType *rbD = QRHI_RES(RenderBufferType, colorAtt.renderBuffer()); + (*dst)[n] = { rbD->globalResourceId(), rbD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } + ++n; + if (colorAtt.resolveTexture()) { + TexType *texD = QRHI_RES(TexType, colorAtt.resolveTexture()); + (*dst)[n] = { texD->globalResourceId(), texD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } + } + if (hasDepthStencil) { + if (desc.depthTexture()) { + TexType *depthTexD = QRHI_RES(TexType, desc.depthTexture()); + (*dst)[n] = { depthTexD->globalResourceId(), depthTexD->generation }; + } else if (desc.depthStencilBuffer()) { + RenderBufferType *depthRbD = QRHI_RES(RenderBufferType, desc.depthStencilBuffer()); + (*dst)[n] = { depthRbD->globalResourceId(), depthRbD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } + } +} + +template<typename TexType, typename RenderBufferType> +bool QRhiRenderTargetAttachmentTracker::isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList) +{ + // Just as setShaderResources() recognizes if an srb's referenced + // resources have been rebuilt (got a create() since the srb's + // create()), we should do the same for the textures and renderbuffers + // referenced from the rendertarget. It is not uncommon that a texture + // or ds buffer gets resized due to following a window size in some + // form, which involves a create() on them. It is then nice if the + // render target auto-rebuilds in beginPass(). + + ResIdList resIdList; + updateResIdList<TexType, RenderBufferType>(desc, &resIdList); + return resIdList == currentResIdList; +} + +template<typename T> +inline T *qrhi_objectFromProxyData(QRhiSwapChainProxyData *pd, QWindow *window, QRhi::Implementation impl, uint objectIndex) +{ + Q_ASSERT(objectIndex < std::size(pd->reserved)); + if (!pd->reserved[objectIndex]) // // was not set, no other choice, do it here, whatever thread this is + *pd = QRhi::updateSwapChainProxyData(impl, window); + return static_cast<T *>(pd->reserved[objectIndex]); +} QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhi_p_p.h b/src/gui/rhi/qrhi_p_p.h deleted file mode 100644 index 89dd7b8e44..0000000000 --- a/src/gui/rhi/qrhi_p_p.h +++ /dev/null @@ -1,713 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHI_P_H -#define QRHI_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhi_p.h" -#include "qrhiprofiler_p_p.h" -#include <QBitArray> -#include <QAtomicInt> -#include <QLoggingCategory> - -QT_BEGIN_NAMESPACE - -#define QRHI_RES(t, x) static_cast<t *>(x) -#define QRHI_RES_RHI(t) t *rhiD = static_cast<t *>(m_rhi) -#define QRHI_PROF QRhiProfilerPrivate *rhiP = m_rhi->profilerPrivateOrNull() -#define QRHI_PROF_F(f) for (bool qrhip_enabled = rhiP != nullptr; qrhip_enabled; qrhip_enabled = false) rhiP->f - -Q_DECLARE_LOGGING_CATEGORY(QRHI_LOG_INFO) - -class QRhiImplementation -{ -public: - virtual ~QRhiImplementation(); - - virtual bool create(QRhi::Flags flags) = 0; - virtual void destroy() = 0; - - virtual QRhiGraphicsPipeline *createGraphicsPipeline() = 0; - virtual QRhiComputePipeline *createComputePipeline() = 0; - virtual QRhiShaderResourceBindings *createShaderResourceBindings() = 0; - virtual QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) = 0; - virtual QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) = 0; - virtual QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) = 0; - virtual QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) = 0; - - virtual QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) = 0; - - virtual QRhiSwapChain *createSwapChain() = 0; - virtual QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) = 0; - virtual QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) = 0; - virtual QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) = 0; - virtual QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) = 0; - virtual QRhi::FrameOpResult finish() = 0; - - virtual void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - - virtual void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) = 0; - virtual void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - - virtual void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) = 0; - - virtual void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) = 0; - - virtual void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) = 0; - - virtual void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) = 0; - virtual void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) = 0; - virtual void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) = 0; - virtual void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) = 0; - - virtual void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) = 0; - virtual void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) = 0; - - virtual void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) = 0; - virtual void debugMarkEnd(QRhiCommandBuffer *cb) = 0; - virtual void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) = 0; - - virtual void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) = 0; - virtual void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - virtual void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) = 0; - virtual void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) = 0; - - virtual const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) = 0; - virtual void beginExternal(QRhiCommandBuffer *cb) = 0; - virtual void endExternal(QRhiCommandBuffer *cb) = 0; - - virtual QList<int> supportedSampleCounts() const = 0; - virtual int ubufAlignment() const = 0; - virtual bool isYUpInFramebuffer() const = 0; - virtual bool isYUpInNDC() const = 0; - virtual bool isClipDepthZeroToOne() const = 0; - virtual QMatrix4x4 clipSpaceCorrMatrix() const = 0; - virtual bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const = 0; - virtual bool isFeatureSupported(QRhi::Feature feature) const = 0; - virtual int resourceLimit(QRhi::ResourceLimit limit) const = 0; - virtual const QRhiNativeHandles *nativeHandles() = 0; - virtual QRhiDriverInfo driverInfo() const = 0; - virtual void sendVMemStatsToProfiler() = 0; - virtual bool makeThreadLocalNativeContextCurrent() = 0; - virtual void releaseCachedResources() = 0; - virtual bool isDeviceLost() const = 0; - - virtual QByteArray pipelineCacheData() = 0; - virtual void setPipelineCacheData(const QByteArray &data) = 0; - - bool isCompressedFormat(QRhiTexture::Format format) const; - void compressedFormatInfo(QRhiTexture::Format format, const QSize &size, - quint32 *bpl, quint32 *byteSize, - QSize *blockDim) const; - void textureFormatInfo(QRhiTexture::Format format, const QSize &size, - quint32 *bpl, quint32 *byteSize, quint32 *bytesPerPixel) const; - quint32 approxByteSizeForTexture(QRhiTexture::Format format, const QSize &baseSize, int depth, - int mipCount, int layerCount); - - QRhiProfilerPrivate *profilerPrivateOrNull() - { - // return null when QRhi::EnableProfiling was not set - QRhiProfilerPrivate *p = QRhiProfilerPrivate::get(&profiler); - return p->rhiDWhenEnabled ? p : nullptr; - } - - // only really care about resources that own native graphics resources underneath - void registerResource(QRhiResource *res) - { - resources.insert(res); - } - - void unregisterResource(QRhiResource *res) - { - resources.remove(res); - } - - QSet<QRhiResource *> activeResources() const - { - return resources; - } - - void addDeleteLater(QRhiResource *res) - { - if (inFrame) - pendingDeleteResources.insert(res); - else - delete res; - } - - void addCleanupCallback(const QRhi::CleanupCallback &callback) - { - cleanupCallbacks.append(callback); - } - - bool sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps); - bool sanityCheckShaderResourceBindings(QRhiShaderResourceBindings *srb); - void updateLayoutDesc(QRhiShaderResourceBindings *srb); - - quint32 pipelineCacheRhiId() const - { - const quint32 ver = (QT_VERSION_MAJOR << 16) | (QT_VERSION_MINOR << 8) | (QT_VERSION_PATCH); - return (quint32(implType) << 24) | ver; - } - - QRhi *q; - - static const int MAX_SHADER_CACHE_ENTRIES = 128; - - bool debugMarkers = false; - int currentFrameSlot = 0; // for vk, mtl, and similar. unused by gl and d3d11. - bool inFrame = false; - -private: - QRhi::Implementation implType; - QThread *implThread; - QRhiProfiler profiler; - QVarLengthArray<QRhiResourceUpdateBatch *, 4> resUpdPool; - quint64 resUpdPoolMap = 0; - int lastResUpdIdx = -1; - QSet<QRhiResource *> resources; - QSet<QRhiResource *> pendingDeleteResources; - QVarLengthArray<QRhi::CleanupCallback, 4> cleanupCallbacks; - - friend class QRhi; - friend class QRhiResourceUpdateBatchPrivate; -}; - -template<typename T, size_t N> -bool qrhi_toTopLeftRenderTargetRect(const QSize &outputSize, const std::array<T, N> &r, - T *x, T *y, T *w, T *h) -{ - // x,y are bottom-left in QRhiScissor and QRhiViewport but top-left in - // Vulkan/Metal/D3D. Our input is an OpenGL-style scissor rect where both - // negative x or y, and partly or completely out of bounds rects are - // allowed. The only thing the input here cannot have is a negative width - // or height. We must handle all other input gracefully, clamping to a zero - // width or height rect in the worst case, and ensuring the resulting rect - // is inside the rendertarget's bounds because some APIs' validation/debug - // layers are allergic to out of bounds scissor or viewport rects. - - const T outputWidth = outputSize.width(); - const T outputHeight = outputSize.height(); - const T inputWidth = r[2]; - const T inputHeight = r[3]; - - if (inputWidth < 0 || inputHeight < 0) - return false; - - *x = r[0]; - *y = outputHeight - (r[1] + inputHeight); - - const T widthOffset = *x < 0 ? -*x : 0; - const T heightOffset = *y < 0 ? -*y : 0; - *w = *x < outputWidth ? qMax<T>(0, inputWidth - widthOffset) : 0; - *h = *y < outputHeight ? qMax<T>(0, inputHeight - heightOffset) : 0; - - *x = qBound<T>(0, *x, outputWidth - 1); - *y = qBound<T>(0, *y, outputHeight - 1); - - if (*x + *w > outputWidth) - *w = qMax<T>(0, outputWidth - *x); - if (*y + *h > outputHeight) - *h = qMax<T>(0, outputHeight - *y); - - return true; -} - -struct QRhiBufferDataPrivate -{ - Q_DISABLE_COPY_MOVE(QRhiBufferDataPrivate) - QRhiBufferDataPrivate() { } - ~QRhiBufferDataPrivate() { delete[] largeData; } - int ref = 1; - int size = 0; - int largeAlloc = 0; - char *largeData = nullptr; - static constexpr int SMALL_DATA_SIZE = 1024; - char data[SMALL_DATA_SIZE]; -}; - -// no detach-with-contents, no atomic refcount, no shrink -class QRhiBufferData -{ -public: - QRhiBufferData() = default; - ~QRhiBufferData() - { - if (d && !--d->ref) - delete d; - } - QRhiBufferData(const QRhiBufferData &other) - : d(other.d) - { - if (d) - d->ref += 1; - } - QRhiBufferData &operator=(const QRhiBufferData &other) - { - if (d == other.d) - return *this; - if (other.d) - other.d->ref += 1; - if (d && !--d->ref) - delete d; - d = other.d; - return *this; - } - const char *constData() const - { - return d->size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE ? d->data : d->largeData; - } - int size() const - { - return d->size; - } - void assign(const char *s, int size) - { - if (!d) { - d = new QRhiBufferDataPrivate; - } else if (d->ref != 1) { - d->ref -= 1; - d = new QRhiBufferDataPrivate; - } - d->size = size; - if (size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE) { - memcpy(d->data, s, size); - } else { - if (d->largeAlloc < size) { - delete[] d->largeData; - d->largeAlloc = size; - d->largeData = new char[size]; - } - memcpy(d->largeData, s, size); - } - } -private: - QRhiBufferDataPrivate *d = nullptr; -}; - -Q_DECLARE_TYPEINFO(QRhiBufferData, Q_RELOCATABLE_TYPE); - -class QRhiResourceUpdateBatchPrivate -{ -public: - struct BufferOp { - enum Type { - DynamicUpdate, - StaticUpload, - Read - }; - Type type; - QRhiBuffer *buf; - int offset; - QRhiBufferData data; - int readSize; - QRhiBufferReadbackResult *result; - - static BufferOp dynamicUpdate(QRhiBuffer *buf, int offset, int size, const void *data) - { - BufferOp op = {}; - op.type = DynamicUpdate; - op.buf = buf; - op.offset = offset; - const int effectiveSize = size ? size : buf->size(); - op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); - return op; - } - - static void changeToDynamicUpdate(BufferOp *op, QRhiBuffer *buf, int offset, int size, const void *data) - { - op->type = DynamicUpdate; - op->buf = buf; - op->offset = offset; - const int effectiveSize = size ? size : buf->size(); - op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); - } - - static BufferOp staticUpload(QRhiBuffer *buf, int offset, int size, const void *data) - { - BufferOp op = {}; - op.type = StaticUpload; - op.buf = buf; - op.offset = offset; - const int effectiveSize = size ? size : buf->size(); - op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); - return op; - } - - static void changeToStaticUpload(BufferOp *op, QRhiBuffer *buf, int offset, int size, const void *data) - { - op->type = StaticUpload; - op->buf = buf; - op->offset = offset; - const int effectiveSize = size ? size : buf->size(); - op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); - } - - static BufferOp read(QRhiBuffer *buf, int offset, int size, QRhiBufferReadbackResult *result) - { - BufferOp op = {}; - op.type = Read; - op.buf = buf; - op.offset = offset; - op.readSize = size; - op.result = result; - return op; - } - }; - - struct TextureOp { - enum Type { - Upload, - Copy, - Read, - GenMips - }; - Type type; - QRhiTexture *dst; - // Specifying multiple uploads for a subresource must be supported. - // In the backend this can then end up, where applicable, as a - // single, batched copy operation with only one set of barriers. - // This helps when doing for example glyph cache fills. - using MipLevelUploadList = std::array<QVector<QRhiTextureSubresourceUploadDescription>, QRhi::MAX_MIP_LEVELS>; - QVarLengthArray<MipLevelUploadList, 6> subresDesc; - QRhiTexture *src; - QRhiTextureCopyDescription desc; - QRhiReadbackDescription rb; - QRhiReadbackResult *result; - - static TextureOp upload(QRhiTexture *tex, const QRhiTextureUploadDescription &desc) - { - TextureOp op = {}; - op.type = Upload; - op.dst = tex; - int maxLayer = -1; - for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) { - if (it->layer() > maxLayer) - maxLayer = it->layer(); - } - op.subresDesc.resize(maxLayer + 1); - for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) - op.subresDesc[it->layer()][it->level()].append(it->description()); - return op; - } - - static TextureOp copy(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc) - { - TextureOp op = {}; - op.type = Copy; - op.dst = dst; - op.src = src; - op.desc = desc; - return op; - } - - static TextureOp read(const QRhiReadbackDescription &rb, QRhiReadbackResult *result) - { - TextureOp op = {}; - op.type = Read; - op.rb = rb; - op.result = result; - return op; - } - - static TextureOp genMips(QRhiTexture *tex) - { - TextureOp op = {}; - op.type = GenMips; - op.dst = tex; - return op; - } - }; - - int activeBufferOpCount = 0; // this is the real number of used elements in bufferOps, not bufferOps.count() - static const int BUFFER_OPS_STATIC_ALLOC = 1024; - QVarLengthArray<BufferOp, BUFFER_OPS_STATIC_ALLOC> bufferOps; - - int activeTextureOpCount = 0; // this is the real number of used elements in textureOps, not textureOps.count() - static const int TEXTURE_OPS_STATIC_ALLOC = 256; - QVarLengthArray<TextureOp, TEXTURE_OPS_STATIC_ALLOC> textureOps; - - QRhiResourceUpdateBatch *q = nullptr; - QRhiImplementation *rhi = nullptr; - int poolIndex = -1; - - void free(); - void merge(QRhiResourceUpdateBatchPrivate *other); - bool hasOptimalCapacity() const; - void trimOpLists(); - - static QRhiResourceUpdateBatchPrivate *get(QRhiResourceUpdateBatch *b) { return b->d; } -}; - -Q_DECLARE_TYPEINFO(QRhiResourceUpdateBatchPrivate::BufferOp, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiResourceUpdateBatchPrivate::TextureOp, Q_RELOCATABLE_TYPE); - -template<typename T> -struct QRhiBatchedBindings -{ - void feed(int binding, T resource) { // binding must be strictly increasing - if (curBinding == -1 || binding > curBinding + 1) { - finish(); - curBatch.startBinding = binding; - curBatch.resources.clear(); - curBatch.resources.append(resource); - } else { - Q_ASSERT(binding == curBinding + 1); - curBatch.resources.append(resource); - } - curBinding = binding; - } - - bool finish() { - if (!curBatch.resources.isEmpty()) - batches.append(curBatch); - return !batches.isEmpty(); - } - - void clear() { - batches.clear(); - curBatch.resources.clear(); - curBinding = -1; - } - - struct Batch { - uint startBinding; - QVarLengthArray<T, 4> resources; - - bool operator==(const Batch &other) const - { - return startBinding == other.startBinding && resources == other.resources; - } - - bool operator!=(const Batch &other) const - { - return !operator==(other); - } - }; - - QVarLengthArray<Batch, 4> batches; // sorted by startBinding - - bool operator==(const QRhiBatchedBindings<T> &other) const - { - return batches == other.batches; - } - - bool operator!=(const QRhiBatchedBindings<T> &other) const - { - return !operator==(other); - } - -private: - Batch curBatch; - int curBinding = -1; -}; - -class QRhiGlobalObjectIdGenerator -{ -public: -#ifdef Q_ATOMIC_INT64_IS_SUPPORTED - using Type = quint64; -#else - using Type = quint32; -#endif - static Type newId(); -}; - -class QRhiPassResourceTracker -{ -public: - bool isEmpty() const; - void reset(); - - struct UsageState { - int layout; - int access; - int stage; - }; - - enum BufferStage { - BufVertexInputStage, - BufVertexStage, - BufFragmentStage, - BufComputeStage - }; - - enum BufferAccess { - BufVertexInput, - BufIndexRead, - BufUniformRead, - BufStorageLoad, - BufStorageStore, - BufStorageLoadStore - }; - - void registerBuffer(QRhiBuffer *buf, int slot, BufferAccess *access, BufferStage *stage, - const UsageState &state); - - enum TextureStage { - TexVertexStage, - TexFragmentStage, - TexColorOutputStage, - TexDepthOutputStage, - TexComputeStage - }; - - enum TextureAccess { - TexSample, - TexColorOutput, - TexDepthOutput, - TexStorageLoad, - TexStorageStore, - TexStorageLoadStore - }; - - void registerTexture(QRhiTexture *tex, TextureAccess *access, TextureStage *stage, - const UsageState &state); - - struct Buffer { - int slot; - BufferAccess access; - BufferStage stage; - UsageState stateAtPassBegin; - }; - - using BufferIterator = QHash<QRhiBuffer *, Buffer>::const_iterator; - BufferIterator cbeginBuffers() const { return m_buffers.cbegin(); } - BufferIterator cendBuffers() const { return m_buffers.cend(); } - - struct Texture { - TextureAccess access; - TextureStage stage; - UsageState stateAtPassBegin; - }; - - using TextureIterator = QHash<QRhiTexture *, Texture>::const_iterator; - TextureIterator cbeginTextures() const { return m_textures.cbegin(); } - TextureIterator cendTextures() const { return m_textures.cend(); } - - static BufferStage toPassTrackerBufferStage(QRhiShaderResourceBinding::StageFlags stages); - static TextureStage toPassTrackerTextureStage(QRhiShaderResourceBinding::StageFlags stages); - -private: - QHash<QRhiBuffer *, Buffer> m_buffers; - QHash<QRhiTexture *, Texture> m_textures; -}; - -Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Buffer, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Texture, Q_RELOCATABLE_TYPE); - -template<typename T, int GROW = 1024> -class QRhiBackendCommandList -{ -public: - QRhiBackendCommandList() = default; - ~QRhiBackendCommandList() { delete[] v; } - inline void reset() { p = 0; } - inline bool isEmpty() const { return p == 0; } - inline T &get() { - if (p == a) { - a += GROW; - T *nv = new T[a]; - if (v) { - memcpy(nv, v, p * sizeof(T)); - delete[] v; - } - v = nv; - } - return v[p++]; - } - inline void unget() { --p; } - inline T *cbegin() const { return v; } - inline T *cend() const { return v + p; } - inline T *begin() { return v; } - inline T *end() { return v + p; } -private: - Q_DISABLE_COPY(QRhiBackendCommandList) - T *v = nullptr; - int a = 0; - int p = 0; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhi_platform.h b/src/gui/rhi/qrhi_platform.h new file mode 100644 index 0000000000..e7be522c52 --- /dev/null +++ b/src/gui/rhi/qrhi_platform.h @@ -0,0 +1,175 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHIPLATFORM_H +#define QRHIPLATFORM_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <rhi/qrhi.h> + +#if QT_CONFIG(opengl) +#include <QtGui/qsurfaceformat.h> +#endif + +#if QT_CONFIG(vulkan) +#include <QtGui/qvulkaninstance.h> +#endif + +#if QT_CONFIG(metal) || defined(Q_QDOC) +Q_FORWARD_DECLARE_OBJC_CLASS(MTLDevice); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandQueue); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandBuffer); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLRenderCommandEncoder); +#endif + +QT_BEGIN_NAMESPACE + +struct Q_GUI_EXPORT QRhiNullInitParams : public QRhiInitParams +{ +}; + +struct Q_GUI_EXPORT QRhiNullNativeHandles : public QRhiNativeHandles +{ +}; + +#if QT_CONFIG(opengl) || defined(Q_QDOC) + +class QOpenGLContext; +class QOffscreenSurface; +class QSurface; +class QWindow; + +struct Q_GUI_EXPORT QRhiGles2InitParams : public QRhiInitParams +{ + QRhiGles2InitParams(); + + QSurfaceFormat format; + QSurface *fallbackSurface = nullptr; + QWindow *window = nullptr; + QOpenGLContext *shareContext = nullptr; + + static QOffscreenSurface *newFallbackSurface(const QSurfaceFormat &format = QSurfaceFormat::defaultFormat()); +}; + +struct Q_GUI_EXPORT QRhiGles2NativeHandles : public QRhiNativeHandles +{ + QOpenGLContext *context = nullptr; +}; + +#endif // opengl/qdoc + +#if (QT_CONFIG(vulkan) && __has_include(<vulkan/vulkan.h>)) || defined(Q_QDOC) + +struct Q_GUI_EXPORT QRhiVulkanInitParams : public QRhiInitParams +{ + QVulkanInstance *inst = nullptr; + QWindow *window = nullptr; + QByteArrayList deviceExtensions; + + static QByteArrayList preferredInstanceExtensions(); + static QByteArrayList preferredExtensionsForImportedDevice(); +}; + +struct Q_GUI_EXPORT QRhiVulkanNativeHandles : public QRhiNativeHandles +{ + // to import a physical device (always required) + VkPhysicalDevice physDev = VK_NULL_HANDLE; + // to import a device and queue + VkDevice dev = VK_NULL_HANDLE; + quint32 gfxQueueFamilyIdx = 0; + quint32 gfxQueueIdx = 0; + // and optionally, the mem allocator + void *vmemAllocator = nullptr; + + // only for querying (rhi->nativeHandles()) + VkQueue gfxQueue = VK_NULL_HANDLE; + QVulkanInstance *inst = nullptr; +}; + +struct Q_GUI_EXPORT QRhiVulkanCommandBufferNativeHandles : public QRhiNativeHandles +{ + VkCommandBuffer commandBuffer = VK_NULL_HANDLE; +}; + +struct Q_GUI_EXPORT QRhiVulkanRenderPassNativeHandles : public QRhiNativeHandles +{ + VkRenderPass renderPass = VK_NULL_HANDLE; +}; + +#endif // vulkan/qdoc + +#if defined(Q_OS_WIN) || defined(Q_QDOC) + +// no d3d includes here, to prevent precompiled header mess due to COM, hence the void pointers + +struct Q_GUI_EXPORT QRhiD3D11InitParams : public QRhiInitParams +{ + bool enableDebugLayer = false; +}; + +struct Q_GUI_EXPORT QRhiD3D11NativeHandles : public QRhiNativeHandles +{ + // to import a device and a context + void *dev = nullptr; + void *context = nullptr; + // alternatively, to specify the device feature level and/or the adapter to use + int featureLevel = 0; + quint32 adapterLuidLow = 0; + qint32 adapterLuidHigh = 0; +}; + +struct Q_GUI_EXPORT QRhiD3D12InitParams : public QRhiInitParams +{ + bool enableDebugLayer = false; +}; + +struct Q_GUI_EXPORT QRhiD3D12NativeHandles : public QRhiNativeHandles +{ + // to import a device + void *dev = nullptr; + int minimumFeatureLevel = 0; + // to just specify the adapter to use, set these and leave dev set to null + quint32 adapterLuidLow = 0; + qint32 adapterLuidHigh = 0; + // in addition, can specify the command queue to use + void *commandQueue = nullptr; +}; + +struct Q_GUI_EXPORT QRhiD3D12CommandBufferNativeHandles : public QRhiNativeHandles +{ + void *commandList = nullptr; // ID3D12GraphicsCommandList1 +}; + +#endif // WIN/QDOC + +#if QT_CONFIG(metal) || defined(Q_QDOC) + +struct Q_GUI_EXPORT QRhiMetalInitParams : public QRhiInitParams +{ +}; + +struct Q_GUI_EXPORT QRhiMetalNativeHandles : public QRhiNativeHandles +{ + MTLDevice *dev = nullptr; + MTLCommandQueue *cmdQueue = nullptr; +}; + +struct Q_GUI_EXPORT QRhiMetalCommandBufferNativeHandles : public QRhiNativeHandles +{ + MTLCommandBuffer *commandBuffer = nullptr; + MTLRenderCommandEncoder *encoder = nullptr; +}; + +#endif // MACOS/IOS/QDOC + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qrhid3d11.cpp b/src/gui/rhi/qrhid3d11.cpp index 6b4be76fef..b09baf57b2 100644 --- a/src/gui/rhi/qrhid3d11.cpp +++ b/src/gui/rhi/qrhid3d11.cpp @@ -1,58 +1,22 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhid3d11_p_p.h" -#include "qshader_p.h" -#include "cs_tdr_p.h" +// Copyright (C) 2019 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhid3d11_p.h" +#include "qshader.h" +#include "vs_test_p.h" #include <QWindow> -#include <QOperatingSystemVersion> #include <qmath.h> -#include <private/qsystemlibrary_p.h> - -#include <d3dcompiler.h> -#include <comdef.h> +#include <QtCore/qcryptographichash.h> +#include <QtCore/private/qsystemerror_p.h> +#include "qrhid3dhelpers_p.h" QT_BEGIN_NAMESPACE +using namespace Qt::StringLiterals; + /* - Direct3D 11 backend. Provides a double-buffered flip model (FLIP_DISCARD) - swapchain. Textures and "static" buffers are USAGE_DEFAULT, leaving it to + Direct3D 11 backend. Provides a double-buffered flip model swapchain. + Textures and "static" buffers are USAGE_DEFAULT, leaving it to UpdateSubResource to upload the data in any way it sees fit. "Dynamic" buffers are USAGE_DYNAMIC and updating is done by mapping with WRITE_DISCARD. (so here QRhiBuffer keeps a copy of the buffer contents and all of it is @@ -62,10 +26,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiD3D11InitParams - \internal \inmodule QtGui + \since 6.6 \brief Direct3D 11 specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A D3D11-based QRhi needs no special parameters for initialization. If desired, enableDebugLayer can be set to \c true to enable the Direct3D debug layer. This can be useful during development, but should be avoided @@ -78,9 +45,7 @@ QT_BEGIN_NAMESPACE \endcode \note QRhiSwapChain should only be used in combination with QWindow - instances that have their surface type set to QSurface::OpenGLSurface. - There are currently no Direct3D specifics in the Windows platform support - of Qt and therefore there is no separate QSurface type available. + instances that have their surface type set to QSurface::Direct3DSurface. \section2 Working with existing Direct3D 11 devices @@ -106,16 +71,71 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiD3D11InitParams::enableDebugLayer + + When set to true, a debug device is created, assuming the debug layer is + available. The default value is false. +*/ + +/*! \class QRhiD3D11NativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the D3D device and device context used by the QRhi. \note The class uses \c{void *} as the type since including the COM-based \c{d3d11.h} headers is not acceptable here. The actual types are \c{ID3D11Device *} and \c{ID3D11DeviceContext *}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiD3D11NativeHandles::dev + + Points to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nn-d3d11-id3d11device}{ID3D11Device} + or left set to \nullptr if no existing device is to be imported. + + \note When importing a device, both the device and the device context must be set to valid objects. +*/ + +/*! + \variable QRhiD3D11NativeHandles::context + + Points to a \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nn-d3d11-id3d11devicecontext}{ID3D11DeviceContext} + or left set to \nullptr if no existing device context is to be imported. + + \note When importing a device, both the device and the device context must be set to valid objects. +*/ + +/*! + \variable QRhiD3D11NativeHandles::featureLevel + + Specifies the feature level passed to + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nf-d3d11-d3d11createdevice}{D3D11CreateDevice()}. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. When not set, the default rules outlined in the D3D + documentation apply. +*/ + +/*! + \variable QRhiD3D11NativeHandles::adapterLuidLow + + The low part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D11NativeHandles::adapterLuidHigh + + The high part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + // help mingw with its ancient sdk headers #ifndef DXGI_ADAPTER_FLAG_SOFTWARE #define DXGI_ADAPTER_FLAG_SOFTWARE 2 @@ -125,20 +145,20 @@ QT_BEGIN_NAMESPACE #define D3D11_1_UAV_SLOT_COUNT 64 #endif +#ifndef D3D11_VS_INPUT_REGISTER_COUNT +#define D3D11_VS_INPUT_REGISTER_COUNT 32 +#endif + QRhiD3D11::QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *importParams) - : ofr(this), - deviceCurse(this) + : ofr(this) { debugLayer = params->enableDebugLayer; - deviceCurse.framesToActivate = params->framesUntilKillingDeviceViaTdr; - deviceCurse.permanent = params->repeatDeviceKill; - if (importParams) { if (importParams->dev && importParams->context) { dev = reinterpret_cast<ID3D11Device *>(importParams->dev); ID3D11DeviceContext *ctx = reinterpret_cast<ID3D11DeviceContext *>(importParams->context); - if (SUCCEEDED(ctx->QueryInterface(IID_ID3D11DeviceContext1, reinterpret_cast<void **>(&context)))) { + if (SUCCEEDED(ctx->QueryInterface(__uuidof(ID3D11DeviceContext1), reinterpret_cast<void **>(&context)))) { // get rid of the ref added by QueryInterface ctx->Release(); importedDeviceAndContext = true; @@ -152,15 +172,6 @@ QRhiD3D11::QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *import } } -static QString comErrorMessage(HRESULT hr) -{ - const _com_error comError(hr); - QString result = QLatin1String("Error 0x") + QString::number(ulong(hr), 16); - if (const wchar_t *msg = comError.ErrorMessage()) - result += QLatin1String(": ") + QString::fromWCharArray(msg); - return result; -} - template <class Int> inline Int aligned(Int v, Int byteAlign) { @@ -170,28 +181,10 @@ inline Int aligned(Int v, Int byteAlign) static IDXGIFactory1 *createDXGIFactory2() { IDXGIFactory1 *result = nullptr; - if (QOperatingSystemVersion::current() > QOperatingSystemVersion::Windows7) { - using PtrCreateDXGIFactory2 = HRESULT (WINAPI *)(UINT, REFIID, void **); - QSystemLibrary dxgilib(QStringLiteral("dxgi")); - if (auto createDXGIFactory2 = reinterpret_cast<PtrCreateDXGIFactory2>(dxgilib.resolve("CreateDXGIFactory2"))) { - const HRESULT hr = createDXGIFactory2(0, IID_IDXGIFactory2, reinterpret_cast<void **>(&result)); - if (FAILED(hr)) { - qWarning("CreateDXGIFactory2() failed to create DXGI factory: %s", qPrintable(comErrorMessage(hr))); - result = nullptr; - } - } else { - qWarning("Unable to resolve CreateDXGIFactory2()"); - } - } - return result; -} - -static IDXGIFactory1 *createDXGIFactory1() -{ - IDXGIFactory1 *result = nullptr; - const HRESULT hr = CreateDXGIFactory1(IID_IDXGIFactory1, reinterpret_cast<void **>(&result)); + const HRESULT hr = CreateDXGIFactory2(0, __uuidof(IDXGIFactory2), reinterpret_cast<void **>(&result)); if (FAILED(hr)) { - qWarning("CreateDXGIFactory1() failed to create DXGI factory: %s", qPrintable(comErrorMessage(hr))); + qWarning("CreateDXGIFactory2() failed to create DXGI factory: %s", + qPrintable(QSystemError::windowsComString(hr))); result = nullptr; } return result; @@ -199,28 +192,39 @@ static IDXGIFactory1 *createDXGIFactory1() bool QRhiD3D11::create(QRhi::Flags flags) { - Q_UNUSED(flags); + rhiFlags = flags; uint devFlags = 0; if (debugLayer) devFlags |= D3D11_CREATE_DEVICE_DEBUG; dxgiFactory = createDXGIFactory2(); - if (dxgiFactory != nullptr) { - hasDxgi2 = true; - supportsFlipDiscardSwapchain = QOperatingSystemVersion::current() >= QOperatingSystemVersion::Windows10 - && !qEnvironmentVariableIntValue("QT_D3D_NO_FLIP"); - } else { - dxgiFactory = createDXGIFactory1(); - hasDxgi2 = false; - supportsFlipDiscardSwapchain = false; + if (!dxgiFactory) + return false; + + // For a FLIP_* swapchain Present(0, 0) is not necessarily + // sufficient to get non-blocking behavior, try using ALLOW_TEARING + // when available. + supportsAllowTearing = false; + IDXGIFactory5 *factory5 = nullptr; + if (SUCCEEDED(dxgiFactory->QueryInterface(__uuidof(IDXGIFactory5), reinterpret_cast<void **>(&factory5)))) { + BOOL allowTearing = false; + if (SUCCEEDED(factory5->CheckFeatureSupport(DXGI_FEATURE_PRESENT_ALLOW_TEARING, &allowTearing, sizeof(allowTearing)))) + supportsAllowTearing = allowTearing; + factory5->Release(); } - if (dxgiFactory == nullptr) - return false; + if (qEnvironmentVariableIntValue("QT_D3D_FLIP_DISCARD")) + qWarning("The default swap effect is FLIP_DISCARD, QT_D3D_FLIP_DISCARD is now ignored"); - qCDebug(QRHI_LOG_INFO, "DXGI 1.2 = %s, FLIP_DISCARD swapchain supported = %s", - hasDxgi2 ? "true" : "false", supportsFlipDiscardSwapchain ? "true" : "false"); + // Support for flip model swapchains is required now (since we are + // targeting Windows 10+), but the option for using the old model is still + // there. (some features are not supported then, however) + useLegacySwapchainModel = qEnvironmentVariableIntValue("QT_D3D_NO_FLIP"); + + qCDebug(QRHI_LOG_INFO, "FLIP_* swapchain supported = true, ALLOW_TEARING supported = %s, use legacy (non-FLIP) model = %s", + supportsAllowTearing ? "true" : "false", + useLegacySwapchainModel ? "true" : "false"); if (!importedDeviceAndContext) { IDXGIAdapter1 *adapter; @@ -255,7 +259,7 @@ bool QRhiD3D11::create(QRhi::Flags flags) } } - IDXGIAdapter1 *adapterToUse = nullptr; + activeAdapter = nullptr; for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { DXGI_ADAPTER_DESC1 desc; adapter->GetDesc1(&desc); @@ -266,18 +270,16 @@ bool QRhiD3D11::create(QRhi::Flags flags) desc.VendorId, desc.DeviceId, desc.Flags); - if (!adapterToUse && (requestedAdapterIndex < 0 || requestedAdapterIndex == adapterIndex)) { - adapterToUse = adapter; + if (!activeAdapter && (requestedAdapterIndex < 0 || requestedAdapterIndex == adapterIndex)) { + activeAdapter = adapter; adapterLuid = desc.AdapterLuid; - driverInfoStruct.deviceName = name.toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); qCDebug(QRHI_LOG_INFO, " using this adapter"); } else { adapter->Release(); } } - if (!adapterToUse) { + if (!activeAdapter) { qWarning("No adapter"); return false; } @@ -292,7 +294,7 @@ bool QRhiD3D11::create(QRhi::Flags flags) } ID3D11DeviceContext *ctx = nullptr; - HRESULT hr = D3D11CreateDevice(adapterToUse, D3D_DRIVER_TYPE_UNKNOWN, nullptr, devFlags, + HRESULT hr = D3D11CreateDevice(activeAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr, devFlags, requestFeatureLevels ? requestedFeatureLevels.constData() : nullptr, requestFeatureLevels ? requestedFeatureLevels.count() : 0, D3D11_SDK_VERSION, @@ -302,36 +304,75 @@ bool QRhiD3D11::create(QRhi::Flags flags) qCDebug(QRHI_LOG_INFO, "Debug layer was requested but is not available. " "Attempting to create D3D11 device without it."); devFlags &= ~D3D11_CREATE_DEVICE_DEBUG; - hr = D3D11CreateDevice(adapterToUse, D3D_DRIVER_TYPE_UNKNOWN, nullptr, devFlags, + hr = D3D11CreateDevice(activeAdapter, D3D_DRIVER_TYPE_UNKNOWN, nullptr, devFlags, requestFeatureLevels ? requestedFeatureLevels.constData() : nullptr, requestFeatureLevels ? requestedFeatureLevels.count() : 0, D3D11_SDK_VERSION, &dev, &featureLevel, &ctx); } - adapterToUse->Release(); if (FAILED(hr)) { - qWarning("Failed to create D3D11 device and context: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create D3D11 device and context: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - if (SUCCEEDED(ctx->QueryInterface(IID_ID3D11DeviceContext1, reinterpret_cast<void **>(&context)))) { - ctx->Release(); - } else { + + const bool supports11_1 = SUCCEEDED(ctx->QueryInterface(__uuidof(ID3D11DeviceContext1), reinterpret_cast<void **>(&context))); + ctx->Release(); + if (!supports11_1) { qWarning("ID3D11DeviceContext1 not supported"); return false; } + + // Test if creating a Shader Model 5.0 vertex shader works; we want to + // fail already in create() if that's not the case. + ID3D11VertexShader *testShader = nullptr; + if (SUCCEEDED(dev->CreateVertexShader(g_testVertexShader, sizeof(g_testVertexShader), nullptr, &testShader))) { + testShader->Release(); + } else { + static const char *msg = "D3D11 smoke test: Failed to create vertex shader"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); + return false; + } + + D3D11_FEATURE_DATA_D3D11_OPTIONS features = {}; + if (SUCCEEDED(dev->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &features, sizeof(features)))) { + // The D3D _runtime_ may be 11.1, but the underlying _driver_ may + // still not support this D3D_FEATURE_LEVEL_11_1 feature. (e.g. + // because it only does 11_0) + if (!features.ConstantBufferOffsetting) { + static const char *msg = "D3D11 smoke test: Constant buffer offsetting is not supported by the driver"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); + return false; + } + } else { + static const char *msg = "D3D11 smoke test: Failed to query D3D11_FEATURE_D3D11_OPTIONS"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); + return false; + } } else { Q_ASSERT(dev && context); featureLevel = dev->GetFeatureLevel(); IDXGIDevice *dxgiDev = nullptr; - if (SUCCEEDED(dev->QueryInterface(IID_IDXGIDevice, reinterpret_cast<void **>(&dxgiDev)))) { + if (SUCCEEDED(dev->QueryInterface(__uuidof(IDXGIDevice), reinterpret_cast<void **>(&dxgiDev)))) { IDXGIAdapter *adapter = nullptr; if (SUCCEEDED(dxgiDev->GetAdapter(&adapter))) { - DXGI_ADAPTER_DESC desc; - adapter->GetDesc(&desc); - adapterLuid = desc.AdapterLuid; - driverInfoStruct.deviceName = QString::fromUtf16(reinterpret_cast<char16_t *>(desc.Description)).toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + IDXGIAdapter1 *adapter1 = nullptr; + if (SUCCEEDED(adapter->QueryInterface(__uuidof(IDXGIAdapter1), reinterpret_cast<void **>(&adapter1)))) { + DXGI_ADAPTER_DESC1 desc; + adapter1->GetDesc1(&desc); + adapterLuid = desc.AdapterLuid; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); + adapter1->Release(); + } adapter->Release(); } dxgiDev->Release(); @@ -339,7 +380,7 @@ bool QRhiD3D11::create(QRhi::Flags flags) qCDebug(QRHI_LOG_INFO, "Using imported device %p", dev); } - if (FAILED(context->QueryInterface(IID_ID3DUserDefinedAnnotation, reinterpret_cast<void **>(&annotations)))) + if (FAILED(context->QueryInterface(__uuidof(ID3DUserDefinedAnnotation), reinterpret_cast<void **>(&annotations)))) annotations = nullptr; deviceLost = false; @@ -350,9 +391,6 @@ bool QRhiD3D11::create(QRhi::Flags flags) nativeHandlesStruct.adapterLuidLow = adapterLuid.LowPart; nativeHandlesStruct.adapterLuidHigh = adapterLuid.HighPart; - if (deviceCurse.framesToActivate > 0) - deviceCurse.initResources(); - return true; } @@ -370,7 +408,16 @@ void QRhiD3D11::destroy() clearShaderCache(); - deviceCurse.releaseResources(); + if (ofr.tsDisjointQuery) { + ofr.tsDisjointQuery->Release(); + ofr.tsDisjointQuery = nullptr; + } + for (int i = 0; i < 2; ++i) { + if (ofr.tsQueries[i]) { + ofr.tsQueries[i]->Release(); + ofr.tsQueries[i] = nullptr; + } + } if (annotations) { annotations->Release(); @@ -388,6 +435,16 @@ void QRhiD3D11::destroy() } } + if (dcompDevice) { + dcompDevice->Release(); + dcompDevice = nullptr; + } + + if (activeAdapter) { + activeAdapter->Release(); + activeAdapter = nullptr; + } + if (dxgiFactory) { dxgiFactory->Release(); dxgiFactory = nullptr; @@ -398,7 +455,7 @@ void QRhiD3D11::reportLiveObjects(ID3D11Device *device) { // this works only when params.enableDebugLayer was true ID3D11Debug *debug; - if (SUCCEEDED(device->QueryInterface(IID_ID3D11Debug, reinterpret_cast<void **>(&debug)))) { + if (SUCCEEDED(device->QueryInterface(__uuidof(ID3D11Debug), reinterpret_cast<void **>(&debug)))) { debug->ReportLiveDeviceObjects(D3D11_RLDO_DETAIL); debug->Release(); } @@ -409,19 +466,13 @@ QList<int> QRhiD3D11::supportedSampleCounts() const return { 1, 2, 4, 8 }; } -DXGI_SAMPLE_DESC QRhiD3D11::effectiveSampleCount(int sampleCount) const +DXGI_SAMPLE_DESC QRhiD3D11::effectiveSampleDesc(int sampleCount) const { DXGI_SAMPLE_DESC desc; desc.Count = 1; desc.Quality = 0; - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - int s = qBound(1, sampleCount, 64); - - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return desc; - } + const int s = effectiveSampleCount(sampleCount); desc.Count = UINT(s); if (s > 1) @@ -437,7 +488,7 @@ QRhiSwapChain *QRhiD3D11::createSwapChain() return new QD3D11SwapChain(this); } -QRhiBuffer *QRhiD3D11::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, int size) +QRhiBuffer *QRhiD3D11::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) { return new QD3D11Buffer(this, type, usage, size); } @@ -541,7 +592,7 @@ bool QRhiD3D11::isFeatureSupported(QRhi::Feature feature) const case QRhi::ReadBackAnyTextureFormat: return true; case QRhi::PipelineCacheDataLoadSave: - return false; + return true; case QRhi::ImageDataStride: return true; case QRhi::RenderBufferImport: @@ -550,6 +601,32 @@ bool QRhiD3D11::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::RenderTo3DTextureSlice: return true; + case QRhi::TextureArrays: + return true; + case QRhi::Tessellation: + return true; + case QRhi::GeometryShader: + return true; + case QRhi::TextureArrayRange: + return true; + case QRhi::NonFillPolygonMode: + return true; + case QRhi::OneDimensionalTextures: + return true; + case QRhi::OneDimensionalTextureMipmaps: + return true; + case QRhi::HalfAttributes: + return true; + case QRhi::RenderToOneDimensionalTexture: + return true; + case QRhi::ThreeDimensionalTextureMipmaps: + return true; + case QRhi::MultiView: + return false; + case QRhi::TextureViewFormat: + return false; // because we use fully typed formats for textures and relaxed casting is a D3D12 thing + case QRhi::ResolveDepthStencil: + return false; default: Q_UNREACHABLE(); return false; @@ -583,6 +660,14 @@ int QRhiD3D11::resourceLimit(QRhi::ResourceLimit limit) const return D3D11_CS_THREAD_GROUP_MAX_Y; case QRhi::MaxThreadGroupZ: return D3D11_CS_THREAD_GROUP_MAX_Z; + case QRhi::TextureArraySizeMax: + return D3D11_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION; + case QRhi::MaxUniformBufferRange: + return 65536; + case QRhi::MaxVertexInputs: + return D3D11_VS_INPUT_REGISTER_COUNT; + case QRhi::MaxVertexOutputs: + return D3D11_VS_OUTPUT_REGISTER_COUNT; default: Q_UNREACHABLE(); return 0; @@ -599,9 +684,11 @@ QRhiDriverInfo QRhiD3D11::driverInfo() const return driverInfoStruct; } -void QRhiD3D11::sendVMemStatsToProfiler() +QRhiStats QRhiD3D11::statistics() { - // nothing to do here + QRhiStats result; + result.totalPipelineCreationTime = totalPipelineCreationTime(); + return result; } bool QRhiD3D11::makeThreadLocalNativeContextCurrent() @@ -613,6 +700,7 @@ bool QRhiD3D11::makeThreadLocalNativeContextCurrent() void QRhiD3D11::releaseCachedResources() { clearShaderCache(); + m_bytecodeCache.clear(); } bool QRhiD3D11::isDeviceLost() const @@ -620,14 +708,159 @@ bool QRhiD3D11::isDeviceLost() const return deviceLost; } +struct QD3D11PipelineCacheDataHeader +{ + quint32 rhiId; + quint32 arch; + // no need for driver specifics + quint32 count; + quint32 dataSize; +}; + QByteArray QRhiD3D11::pipelineCacheData() { - return QByteArray(); + QByteArray data; + if (m_bytecodeCache.isEmpty()) + return data; + + QD3D11PipelineCacheDataHeader header; + memset(&header, 0, sizeof(header)); + header.rhiId = pipelineCacheRhiId(); + header.arch = quint32(sizeof(void*)); + header.count = m_bytecodeCache.count(); + + const size_t dataOffset = sizeof(header); + size_t dataSize = 0; + for (auto it = m_bytecodeCache.cbegin(), end = m_bytecodeCache.cend(); it != end; ++it) { + BytecodeCacheKey key = it.key(); + QByteArray bytecode = it.value(); + dataSize += + sizeof(quint32) + key.sourceHash.size() + + sizeof(quint32) + key.target.size() + + sizeof(quint32) + key.entryPoint.size() + + sizeof(quint32) // compileFlags + + sizeof(quint32) + bytecode.size(); + } + + QByteArray buf(dataOffset + dataSize, Qt::Uninitialized); + char *p = buf.data() + dataOffset; + for (auto it = m_bytecodeCache.cbegin(), end = m_bytecodeCache.cend(); it != end; ++it) { + BytecodeCacheKey key = it.key(); + QByteArray bytecode = it.value(); + + quint32 i = key.sourceHash.size(); + memcpy(p, &i, 4); + p += 4; + memcpy(p, key.sourceHash.constData(), key.sourceHash.size()); + p += key.sourceHash.size(); + + i = key.target.size(); + memcpy(p, &i, 4); + p += 4; + memcpy(p, key.target.constData(), key.target.size()); + p += key.target.size(); + + i = key.entryPoint.size(); + memcpy(p, &i, 4); + p += 4; + memcpy(p, key.entryPoint.constData(), key.entryPoint.size()); + p += key.entryPoint.size(); + + quint32 f = key.compileFlags; + memcpy(p, &f, 4); + p += 4; + + i = bytecode.size(); + memcpy(p, &i, 4); + p += 4; + memcpy(p, bytecode.constData(), bytecode.size()); + p += bytecode.size(); + } + Q_ASSERT(p == buf.data() + dataOffset + dataSize); + + header.dataSize = quint32(dataSize); + memcpy(buf.data(), &header, sizeof(header)); + + return buf; } void QRhiD3D11::setPipelineCacheData(const QByteArray &data) { - Q_UNUSED(data); + if (data.isEmpty()) + return; + + const size_t headerSize = sizeof(QD3D11PipelineCacheDataHeader); + if (data.size() < qsizetype(headerSize)) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (header incomplete)"); + return; + } + const size_t dataOffset = headerSize; + QD3D11PipelineCacheDataHeader header; + memcpy(&header, data.constData(), headerSize); + + const quint32 rhiId = pipelineCacheRhiId(); + if (header.rhiId != rhiId) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", + rhiId, header.rhiId); + return; + } + const quint32 arch = quint32(sizeof(void*)); + if (header.arch != arch) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Architecture does not match (%u, %u)", + arch, header.arch); + return; + } + if (header.count == 0) + return; + + if (data.size() < qsizetype(dataOffset + header.dataSize)) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (data incomplete)"); + return; + } + + m_bytecodeCache.clear(); + + const char *p = data.constData() + dataOffset; + for (quint32 i = 0; i < header.count; ++i) { + quint32 len = 0; + memcpy(&len, p, 4); + p += 4; + QByteArray sourceHash(len, Qt::Uninitialized); + memcpy(sourceHash.data(), p, len); + p += len; + + memcpy(&len, p, 4); + p += 4; + QByteArray target(len, Qt::Uninitialized); + memcpy(target.data(), p, len); + p += len; + + memcpy(&len, p, 4); + p += 4; + QByteArray entryPoint(len, Qt::Uninitialized); + memcpy(entryPoint.data(), p, len); + p += len; + + quint32 flags; + memcpy(&flags, p, 4); + p += 4; + + memcpy(&len, p, 4); + p += 4; + QByteArray bytecode(len, Qt::Uninitialized); + memcpy(bytecode.data(), p, len); + p += len; + + BytecodeCacheKey cacheKey; + cacheKey.sourceHash = sourceHash; + cacheKey.target = target; + cacheKey.entryPoint = entryPoint; + cacheKey.compileFlags = flags; + + m_bytecodeCache.insert(cacheKey, bytecode); + } + + qCDebug(QRHI_LOG_INFO, "Seeded bytecode cache with %d shaders", int(m_bytecodeCache.count())); } QRhiRenderBuffer *QRhiD3D11::createRenderBuffer(QRhiRenderBuffer::Type type, const QSize &pixelSize, @@ -638,10 +871,10 @@ QRhiRenderBuffer *QRhiD3D11::createRenderBuffer(QRhiRenderBuffer::Type type, con } QRhiTexture *QRhiD3D11::createTexture(QRhiTexture::Format format, - const QSize &pixelSize, int depth, + const QSize &pixelSize, int depth, int arraySize, int sampleCount, QRhiTexture::Flags flags) { - return new QD3D11Texture(this, format, pixelSize, depth, sampleCount, flags); + return new QD3D11Texture(this, format, pixelSize, depth, arraySize, sampleCount, flags); } QRhiSampler *QRhiD3D11::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, @@ -690,10 +923,13 @@ void QRhiD3D11::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline } } -static const int RBM_SUPPORTED_STAGES = 3; +static const int RBM_SUPPORTED_STAGES = 6; static const int RBM_VERTEX = 0; -static const int RBM_FRAGMENT = 1; -static const int RBM_COMPUTE = 2; +static const int RBM_HULL = 1; +static const int RBM_DOMAIN = 2; +static const int RBM_GEOMETRY = 3; +static const int RBM_FRAGMENT = 4; +static const int RBM_COMPUTE = 5; void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBindings *srb, int dynamicOffsetCount, @@ -715,7 +951,7 @@ void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind bool srbUpdate = false; for (int i = 0, ie = srbD->sortedBindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->sortedBindings.at(i).data(); + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings.at(i)); QD3D11ShaderResourceBindings::BoundResourceData &bd(srbD->boundResourceData[i]); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: @@ -734,8 +970,10 @@ void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind } break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; if (bd.stex.count != data->count) { bd.stex.count = data->count; srbUpdate = true; @@ -743,16 +981,24 @@ void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind for (int elem = 0; elem < data->count; ++elem) { QD3D11Texture *texD = QRHI_RES(QD3D11Texture, data->texSamplers[elem].tex); QD3D11Sampler *samplerD = QRHI_RES(QD3D11Sampler, data->texSamplers[elem].sampler); - if (texD->generation != bd.stex.d[elem].texGeneration - || texD->m_id != bd.stex.d[elem].texId - || samplerD->generation != bd.stex.d[elem].samplerGeneration - || samplerD->m_id != bd.stex.d[elem].samplerId) + // We use the same code path for both combined and separate + // images and samplers, so tex or sampler (but not both) can be + // null here. + Q_ASSERT(texD || samplerD); + const quint64 texId = texD ? texD->m_id : 0; + const uint texGen = texD ? texD->generation : 0; + const quint64 samplerId = samplerD ? samplerD->m_id : 0; + const uint samplerGen = samplerD ? samplerD->generation : 0; + if (texGen != bd.stex.d[elem].texGeneration + || texId != bd.stex.d[elem].texId + || samplerGen != bd.stex.d[elem].samplerGeneration + || samplerId != bd.stex.d[elem].samplerId) { srbUpdate = true; - bd.stex.d[elem].texId = texD->m_id; - bd.stex.d[elem].texGeneration = texD->generation; - bd.stex.d[elem].samplerId = samplerD->m_id; - bd.stex.d[elem].samplerGeneration = samplerD->generation; + bd.stex.d[elem].texId = texId; + bd.stex.d[elem].texGeneration = texGen; + bd.stex.d[elem].samplerId = samplerId; + bd.stex.d[elem].samplerGeneration = samplerGen; } } } @@ -792,6 +1038,9 @@ void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind memset(resBindMaps, 0, sizeof(resBindMaps)); if (gfxPsD) { resBindMaps[RBM_VERTEX] = &gfxPsD->vs.nativeResourceBindingMap; + resBindMaps[RBM_HULL] = &gfxPsD->hs.nativeResourceBindingMap; + resBindMaps[RBM_DOMAIN] = &gfxPsD->ds.nativeResourceBindingMap; + resBindMaps[RBM_GEOMETRY] = &gfxPsD->gs.nativeResourceBindingMap; resBindMaps[RBM_FRAGMENT] = &gfxPsD->fs.nativeResourceBindingMap; } else { resBindMaps[RBM_COMPUTE] = &compPsD->cs.nativeResourceBindingMap; @@ -826,8 +1075,8 @@ void QRhiD3D11::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind for (int i = 0; i < dynamicOffsetCount; ++i) { const QRhiCommandBuffer::DynamicOffset &dynOfs(dynamicOffsets[i]); const uint binding = uint(dynOfs.first); - Q_ASSERT(aligned(dynOfs.second, quint32(256)) == dynOfs.second); - const uint offsetInConstants = dynOfs.second / 16; + Q_ASSERT(aligned(dynOfs.second, 256u) == dynOfs.second); + const quint32 offsetInConstants = dynOfs.second / 16; *p++ = binding; *p++ = offsetInConstants; } @@ -918,7 +1167,7 @@ void QRhiD3D11::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) // d3d expects top-left, QRhiViewport is bottom-left float x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, viewport.viewport(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<UnBounded>(outputSize, viewport.viewport(), &x, &y, &w, &h)) return; QD3D11CommandBuffer::Command &cmd(cbD->commands.get()); @@ -940,7 +1189,7 @@ void QRhiD3D11::setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) // d3d expects top-left, QRhiScissor is bottom-left int x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, scissor.scissor(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, scissor.scissor(), &x, &y, &w, &h)) return; QD3D11CommandBuffer::Command &cmd(cbD->commands.get()); @@ -1048,7 +1297,6 @@ const QRhiNativeHandles *QRhiD3D11::nativeHandles(QRhiCommandBuffer *cb) void QRhiD3D11::beginExternal(QRhiCommandBuffer *cb) { QD3D11CommandBuffer *cbD = QRHI_RES(QD3D11CommandBuffer, cb); - // no timestampSwapChain, in order to avoid timestamp mess executeCommandBuffer(cbD); cbD->resetCommands(); } @@ -1065,6 +1313,25 @@ void QRhiD3D11::endExternal(QRhiCommandBuffer *cb) } } +double QRhiD3D11::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QD3D11CommandBuffer *cbD = QRHI_RES(QD3D11CommandBuffer, cb); + return cbD->lastGpuTime; +} + +static inline QD3D11RenderTargetData *rtData(QRhiRenderTarget *rt) +{ + switch (rt->resourceType()) { + case QRhiResource::SwapChainRenderTarget: + return &QRHI_RES(QD3D11SwapChainRenderTarget, rt)->d; + case QRhiResource::TextureRenderTarget: + return &QRHI_RES(QD3D11TextureRenderTarget, rt)->d; + default: + Q_UNREACHABLE(); + return nullptr; + } +} + QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); @@ -1072,32 +1339,6 @@ QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF QD3D11SwapChain *swapChainD = QRHI_RES(QD3D11SwapChain, swapChain); contextState.currentSwapChain = swapChainD; const int currentFrameSlot = swapChainD->currentFrameSlot; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - - if (swapChainD->timestampActive[currentFrameSlot]) { - ID3D11Query *tsDisjoint = swapChainD->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = swapChainD->timestampQuery[tsIdx]; - ID3D11Query *tsEnd = swapChainD->timestampQuery[tsIdx + 1]; - quint64 timestamps[2]; - D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; - bool ok = true; - ok &= context->GetData(tsDisjoint, &dj, sizeof(dj), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - ok &= context->GetData(tsEnd, ×tamps[1], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - // this above is often not ready, not even in frame_where_recorded+2, - // not clear why. so make the whole thing async and do not touch the - // queries until they are finally all available in frame this+2 or - // this+4 or ... - ok &= context->GetData(tsStart, ×tamps[0], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - if (ok) { - if (!dj.Disjoint && dj.Frequency) { - const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; - // finally got a value, just report it, the profiler cares about min/max/avg anyway - QRHI_PROF_F(swapChainFrameGpuTime(swapChain, elapsedMs)); - } - swapChainD->timestampActive[currentFrameSlot] = false; - } // else leave timestampActive set to true, will retry in a subsequent beginFrame - } swapChainD->cb.resetState(); @@ -1105,10 +1346,24 @@ QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF swapChainD->msaaRtv[currentFrameSlot] : swapChainD->backBufferRtv; swapChainD->rt.d.dsv = swapChainD->ds ? swapChainD->ds->dsv : nullptr; - QRHI_PROF_F(beginSwapChainFrame(swapChain)); - finishActiveReadbacks(); + if (swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]) { + double elapsedSec = 0; + if (swapChainD->timestamps.tryQueryTimestamps(swapChainD->currentTimestampPairIndex, context, &elapsedSec)) + swapChainD->cb.lastGpuTime = elapsedSec; + } + + ID3D11Query *tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + ID3D11Query *tsDisjoint = swapChainD->timestamps.disjointQuery[swapChainD->currentTimestampPairIndex]; + const bool recordTimestamps = tsStart && tsDisjoint && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + + QD3D11CommandBuffer::Command &cmd(swapChainD->cb.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::BeginFrame; + cmd.args.beginFrame.tsQuery = recordTimestamps ? tsStart : nullptr; + cmd.args.beginFrame.tsDisjointQuery = recordTimestamps ? tsDisjoint : nullptr; + cmd.args.beginFrame.swapchainData = rtData(&swapChainD->rt); + return QRhi::FrameOpSuccess; } @@ -1118,17 +1373,13 @@ QRhi::FrameOpResult QRhiD3D11::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame Q_ASSERT(contextState.currentSwapChain = swapChainD); const int currentFrameSlot = swapChainD->currentFrameSlot; - ID3D11Query *tsDisjoint = swapChainD->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = swapChainD->timestampQuery[tsIdx]; - ID3D11Query *tsEnd = swapChainD->timestampQuery[tsIdx + 1]; - const bool recordTimestamps = tsDisjoint && tsStart && tsEnd && !swapChainD->timestampActive[currentFrameSlot]; + QD3D11CommandBuffer::Command &cmd(swapChainD->cb.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::EndFrame; + cmd.args.endFrame.tsQuery = nullptr; // done later manually, see below + cmd.args.endFrame.tsDisjointQuery = nullptr; // send all commands to the context - if (recordTimestamps) - executeCommandBuffer(&swapChainD->cb, swapChainD); - else - executeCommandBuffer(&swapChainD->cb); + executeCommandBuffer(&swapChainD->cb); if (swapChainD->sampleDesc.Count > 1) { context->ResolveSubresource(swapChainD->backBufferTex, 0, @@ -1136,29 +1387,39 @@ QRhi::FrameOpResult QRhiD3D11::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame swapChainD->colorFormat); } - // this is here because we want to include the time spent on the resolve as well + // this is here because we want to include the time spent on the ResolveSubresource as well + ID3D11Query *tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + ID3D11Query *tsDisjoint = swapChainD->timestamps.disjointQuery[swapChainD->currentTimestampPairIndex]; + const bool recordTimestamps = tsEnd && tsDisjoint && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; if (recordTimestamps) { context->End(tsEnd); context->End(tsDisjoint); - swapChainD->timestampActive[currentFrameSlot] = true; + swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex] = true; + swapChainD->currentTimestampPairIndex = (swapChainD->currentTimestampPairIndex + 1) % QD3D11SwapChainTimestamps::TIMESTAMP_PAIRS; } - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - // this must be done before the Present - QRHI_PROF_F(endSwapChainFrame(swapChain, swapChainD->frameCount + 1)); - if (!flags.testFlag(QRhi::SkipPresent)) { - const UINT presentFlags = 0; + UINT presentFlags = 0; + if (swapChainD->swapInterval == 0 && (swapChainD->swapChainFlags & DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING)) + presentFlags |= DXGI_PRESENT_ALLOW_TEARING; + if (!swapChainD->swapChain) { + qWarning("Failed to present: IDXGISwapChain is unavailable"); + return QRhi::FrameOpError; + } HRESULT hr = swapChainD->swapChain->Present(swapChainD->swapInterval, presentFlags); if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { qWarning("Device loss detected in Present()"); deviceLost = true; return QRhi::FrameOpDeviceLost; } else if (FAILED(hr)) { - qWarning("Failed to present: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to present: %s", + qPrintable(QSystemError::windowsComString(hr))); return QRhi::FrameOpError; } + if (dcompDevice && swapChainD->dcompTarget && swapChainD->dcompVisual) + dcompDevice->Commit(); + // move on to the next buffer swapChainD->currentFrameSlot = (swapChainD->currentFrameSlot + 1) % QD3D11SwapChain::BUFFER_COUNT; } else { @@ -1168,19 +1429,6 @@ QRhi::FrameOpResult QRhiD3D11::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame swapChainD->frameCount += 1; contextState.currentSwapChain = nullptr; - if (deviceCurse.framesToActivate > 0) { - deviceCurse.framesLeft -= 1; - if (deviceCurse.framesLeft == 0) { - deviceCurse.framesLeft = deviceCurse.framesToActivate; - if (!deviceCurse.permanent) - deviceCurse.framesToActivate = -1; - - deviceCurse.activate(); - } else if (deviceCurse.framesLeft % 100 == 0) { - qDebug("Impending doom: %d frames left", deviceCurse.framesLeft); - } - } - return QRhi::FrameOpSuccess; } @@ -1192,6 +1440,36 @@ QRhi::FrameOpResult QRhiD3D11::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: ofr.cbWrapper.resetState(); *cb = &ofr.cbWrapper; + if (rhiFlags.testFlag(QRhi::EnableTimestamps)) { + D3D11_QUERY_DESC queryDesc = {}; + if (!ofr.tsDisjointQuery) { + queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; + HRESULT hr = dev->CreateQuery(&queryDesc, &ofr.tsDisjointQuery); + if (FAILED(hr)) { + qWarning("Failed to create timestamp disjoint query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + } + queryDesc.Query = D3D11_QUERY_TIMESTAMP; + for (int i = 0; i < 2; ++i) { + if (!ofr.tsQueries[i]) { + HRESULT hr = dev->CreateQuery(&queryDesc, &ofr.tsQueries[i]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + } + } + } + + QD3D11CommandBuffer::Command &cmd(ofr.cbWrapper.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::BeginFrame; + cmd.args.beginFrame.tsQuery = ofr.tsQueries[0] ? ofr.tsQueries[0] : nullptr; + cmd.args.beginFrame.tsDisjointQuery = ofr.tsDisjointQuery ? ofr.tsDisjointQuery : nullptr; + cmd.args.beginFrame.swapchainData = nullptr; + return QRhi::FrameOpSuccess; } @@ -1200,10 +1478,41 @@ QRhi::FrameOpResult QRhiD3D11::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_UNUSED(flags); ofr.active = false; + QD3D11CommandBuffer::Command &cmd(ofr.cbWrapper.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::EndFrame; + cmd.args.endFrame.tsQuery = ofr.tsQueries[1] ? ofr.tsQueries[1] : nullptr; + cmd.args.endFrame.tsDisjointQuery = ofr.tsDisjointQuery ? ofr.tsDisjointQuery : nullptr; + executeCommandBuffer(&ofr.cbWrapper); + context->Flush(); finishActiveReadbacks(); + if (ofr.tsQueries[0]) { + quint64 timestamps[2]; + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; + HRESULT hr; + bool ok = true; + do { + hr = context->GetData(ofr.tsDisjointQuery, &dj, sizeof(dj), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + do { + hr = context->GetData(ofr.tsQueries[1], ×tamps[1], sizeof(quint64), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + do { + hr = context->GetData(ofr.tsQueries[0], ×tamps[0], sizeof(quint64), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + if (ok) { + if (!dj.Disjoint && dj.Frequency) { + const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; + ofr.cbWrapper.lastGpuTime = elapsedMs / 1000.0; + } + } + } + return QRhi::FrameOpSuccess; } @@ -1235,12 +1544,15 @@ static inline DXGI_FORMAT toD3DTextureFormat(QRhiTexture::Format format, QRhiTex case QRhiTexture::R32F: return DXGI_FORMAT_R32_FLOAT; + case QRhiTexture::RGB10A2: + return DXGI_FORMAT_R10G10B10A2_UNORM; + case QRhiTexture::D16: return DXGI_FORMAT_R16_TYPELESS; case QRhiTexture::D24: - return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + return DXGI_FORMAT_R24G8_TYPELESS; case QRhiTexture::D24S8: - return DXGI_FORMAT_D24_UNORM_S8_UINT; + return DXGI_FORMAT_R24G8_TYPELESS; case QRhiTexture::D32F: return DXGI_FORMAT_R32_TYPELESS; @@ -1288,7 +1600,7 @@ static inline DXGI_FORMAT toD3DTextureFormat(QRhiTexture::Format format, QRhiTex } } -static inline QRhiTexture::Format colorTextureFormatFromDxgiFormat(DXGI_FORMAT format, QRhiTexture::Flags *flags) +static inline QRhiTexture::Format swapchainReadbackTextureFormat(DXGI_FORMAT format, QRhiTexture::Flags *flags) { switch (format) { case DXGI_FORMAT_R8G8B8A8_UNORM: @@ -1303,16 +1615,14 @@ static inline QRhiTexture::Format colorTextureFormatFromDxgiFormat(DXGI_FORMAT f if (flags) (*flags) |= QRhiTexture::sRGB; return QRhiTexture::BGRA8; - case DXGI_FORMAT_R8_UNORM: - return QRhiTexture::R8; - case DXGI_FORMAT_R8G8_UNORM: - return QRhiTexture::RG8; - case DXGI_FORMAT_R16_UNORM: - return QRhiTexture::R16; - case DXGI_FORMAT_R16G16_UNORM: - return QRhiTexture::RG16; - default: // this cannot assert, must warn and return unknown - qWarning("DXGI_FORMAT %d is not a recognized uncompressed color format", format); + case DXGI_FORMAT_R16G16B16A16_FLOAT: + return QRhiTexture::RGBA16F; + case DXGI_FORMAT_R32G32B32A32_FLOAT: + return QRhiTexture::RGBA32F; + case DXGI_FORMAT_R10G10B10A2_UNORM: + return QRhiTexture::RGB10A2; + default: + qWarning("DXGI_FORMAT %d cannot be read back", format); break; } return QRhiTexture::UnknownFormat; @@ -1343,7 +1653,7 @@ QRhi::FrameOpResult QRhiD3D11::finish() } else { Q_ASSERT(contextState.currentSwapChain); Q_ASSERT(contextState.currentSwapChain->cb.recordingPass == QD3D11CommandBuffer::NoPass); - executeCommandBuffer(&contextState.currentSwapChain->cb); // no timestampSwapChain, in order to avoid timestamp mess + executeCommandBuffer(&contextState.currentSwapChain->cb); contextState.currentSwapChain->cb.resetCommands(); } } @@ -1437,7 +1747,6 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate { QD3D11CommandBuffer *cbD = QRHI_RES(QD3D11CommandBuffer, cb); QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); for (int opIdx = 0; opIdx < ud->activeBufferOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::BufferOp &u(ud->bufferOps[opIdx]); @@ -1460,10 +1769,10 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate // since the ID3D11Buffer's size is rounded up to be a multiple of 256 // while the data we have has the original size. D3D11_BOX box; - box.left = UINT(u.offset); + box.left = u.offset; box.top = box.front = 0; box.back = box.bottom = 1; - box.right = UINT(u.offset + u.data.size()); // no -1: right, bottom, back are exclusive, see D3D11_BOX doc + box.right = u.offset + u.data.size(); // no -1: right, bottom, back are exclusive, see D3D11_BOX doc cmd.args.updateSubRes.hasDstBox = true; cmd.args.updateSubRes.dstBox = box; } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::Read) { @@ -1471,22 +1780,23 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate if (bufD->m_type == QRhiBuffer::Dynamic) { u.result->data.resize(u.readSize); memcpy(u.result->data.data(), bufD->dynBuf + u.offset, size_t(u.readSize)); + if (u.result->completed) + u.result->completed(); } else { BufferReadback readback; readback.result = u.result; readback.byteSize = u.readSize; - D3D11_BUFFER_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_BUFFER_DESC desc = {}; desc.ByteWidth = readback.byteSize; desc.Usage = D3D11_USAGE_STAGING; desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ; HRESULT hr = dev->CreateBuffer(&desc, nullptr, &readback.stagingBuf); if (FAILED(hr)) { - qWarning("Failed to create buffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create buffer: %s", + qPrintable(QSystemError::windowsComString(hr))); continue; } - QRHI_PROF_F(newReadbackBuffer(qint64(qintptr(readback.stagingBuf)), bufD, readback.byteSize)); QD3D11CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QD3D11CommandBuffer::Command::CopySubRes; @@ -1499,16 +1809,14 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate cmd.args.copySubRes.srcSubRes = 0; cmd.args.copySubRes.hasSrcBox = true; D3D11_BOX box; - box.left = UINT(u.offset); + box.left = u.offset; box.top = box.front = 0; box.back = box.bottom = 1; - box.right = UINT(u.offset + u.readSize); + box.right = u.offset + u.readSize; cmd.args.copySubRes.srcBox = box; activeBufferReadbacks.append(readback); } - if (u.result->completed) - u.result->completed(); } } for (int opIdx = 0; opIdx < ud->activeTextureOpCount; ++opIdx) { @@ -1517,7 +1825,7 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate QD3D11Texture *texD = QRHI_RES(QD3D11Texture, u.dst); for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) enqueueSubresUpload(texD, cbD, layer, level, subresDesc); } } @@ -1594,7 +1902,7 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate src = swapChainD->backBufferTex; dxgiFormat = swapChainD->colorFormat; pixelSize = swapChainD->pixelSize; - format = colorTextureFormatFromDxgiFormat(dxgiFormat, nullptr); + format = swapchainReadbackTextureFormat(dxgiFormat, nullptr); if (format == QRhiTexture::UnknownFormat) continue; } @@ -1602,8 +1910,7 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate quint32 bpl = 0; textureFormatInfo(format, pixelSize, &bpl, &byteSize, nullptr); - D3D11_TEXTURE2D_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_TEXTURE2D_DESC desc = {}; desc.Width = UINT(pixelSize.width()); desc.Height = UINT(pixelSize.height()); desc.MipLevels = 1; @@ -1615,12 +1922,10 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate ID3D11Texture2D *stagingTex; HRESULT hr = dev->CreateTexture2D(&desc, nullptr, &stagingTex); if (FAILED(hr)) { - qWarning("Failed to create readback staging texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create readback staging texture: %s", + qPrintable(QSystemError::windowsComString(hr))); return; } - QRHI_PROF_F(newReadbackBuffer(qint64(qintptr(stagingTex)), - texD ? static_cast<QRhiResource *>(texD) : static_cast<QRhiResource *>(swapChainD), - byteSize)); QD3D11CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QD3D11CommandBuffer::Command::CopySubRes; @@ -1632,8 +1937,7 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate cmd.args.copySubRes.src = src; cmd.args.copySubRes.srcSubRes = subres; if (is3D) { - D3D11_BOX srcBox; - memset(&srcBox, 0, sizeof(srcBox)); + D3D11_BOX srcBox = {}; srcBox.front = UINT(u.rb.layer()); srcBox.right = desc.Width; // exclusive srcBox.bottom = desc.Height; @@ -1665,7 +1969,6 @@ void QRhiD3D11::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate void QRhiD3D11::finishActiveReadbacks() { QVarLengthArray<std::function<void()>, 4> completedCallbacks; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); for (int i = activeTextureReadbacks.count() - 1; i >= 0; --i) { const QRhiD3D11::TextureReadback &readback(activeTextureReadbacks[i]); @@ -1687,11 +1990,11 @@ void QRhiD3D11::finishActiveReadbacks() } context->Unmap(readback.stagingTex, 0); } else { - qWarning("Failed to map readback staging texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to map readback staging texture: %s", + qPrintable(QSystemError::windowsComString(hr))); } readback.stagingTex->Release(); - QRHI_PROF_F(releaseReadbackBuffer(qint64(qintptr(readback.stagingTex)))); if (readback.result->completed) completedCallbacks.append(readback.result->completed); @@ -1709,11 +2012,11 @@ void QRhiD3D11::finishActiveReadbacks() memcpy(readback.result->data.data(), mp.pData, readback.byteSize); context->Unmap(readback.stagingBuf, 0); } else { - qWarning("Failed to map readback staging texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to map readback staging texture: %s", + qPrintable(QSystemError::windowsComString(hr))); } readback.stagingBuf->Release(); - QRHI_PROF_F(releaseReadbackBuffer(qint64(qintptr(readback.stagingBuf)))); if (readback.result->completed) completedCallbacks.append(readback.result->completed); @@ -1725,19 +2028,6 @@ void QRhiD3D11::finishActiveReadbacks() f(); } -static inline QD3D11RenderTargetData *rtData(QRhiRenderTarget *rt) -{ - switch (rt->resourceType()) { - case QRhiResource::RenderTarget: - return &QRHI_RES(QD3D11ReferenceRenderTarget, rt)->d; - case QRhiResource::TextureRenderTarget: - return &QRHI_RES(QD3D11TextureRenderTarget, rt)->d; - default: - Q_UNREACHABLE(); - return nullptr; - } -} - void QRhiD3D11::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) { Q_ASSERT(QRHI_RES(QD3D11CommandBuffer, cb)->recordingPass == QD3D11CommandBuffer::NoPass); @@ -1765,6 +2055,8 @@ void QRhiD3D11::beginPass(QRhiCommandBuffer *cb, QD3D11TextureRenderTarget *rtTex = QRHI_RES(QD3D11TextureRenderTarget, rt); wantsColorClear = !rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents); wantsDsClear = !rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents); + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QD3D11Texture, QD3D11RenderBuffer>(rtTex->description(), rtD->currentResIdList)) + rtTex->create(); } cbD->commands.get().cmd = QD3D11CommandBuffer::Command::ResetShaderResources; @@ -1792,7 +2084,7 @@ void QRhiD3D11::beginPass(QRhiCommandBuffer *cb, cbD->recordingPass = QD3D11CommandBuffer::RenderPass; cbD->currentTarget = rt; - cbD->resetCachedShaderResourceState(); + cbD->resetCachedState(); } void QRhiD3D11::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -1854,6 +2146,8 @@ void QRhiD3D11::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resource cmd.args.resolveSubRes.srcSubRes = D3D11CalcSubresource(0, UINT(colorAtt.layer()), 1); cmd.args.resolveSubRes.format = dstTexD->dxgiFormat; } + if (rtTex->m_desc.depthResolveTexture()) + qWarning("Resolving multisample depth-stencil buffers is not supported with D3D"); } cbD->recordingPass = QD3D11CommandBuffer::NoPass; @@ -1878,7 +2172,7 @@ void QRhiD3D11::beginComputePass(QRhiCommandBuffer *cb, cbD->recordingPass = QD3D11CommandBuffer::ComputePass; - cbD->resetCachedShaderResourceState(); + cbD->resetCachedState(); } void QRhiD3D11::endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -1928,7 +2222,7 @@ static inline QPair<int, int> mapBinding(int binding, { const QShader::NativeResourceBindingMap *map = nativeResourceBindingMaps[stageIndex]; if (!map || map->isEmpty()) - return { binding, binding }; // old QShader versions do not have this map, assume 1:1 mapping then + return { binding, binding }; // assume 1:1 mapping auto it = map->constFind(binding); if (it != map->cend()) @@ -1943,31 +2237,21 @@ static inline QPair<int, int> mapBinding(int binding, void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[]) { - srbD->vsubufs.clear(); - srbD->vsubuforigbindings.clear(); - srbD->vsubufoffsets.clear(); - srbD->vsubufsizes.clear(); - - srbD->fsubufs.clear(); - srbD->fsubuforigbindings.clear(); - srbD->fsubufoffsets.clear(); - srbD->fsubufsizes.clear(); + srbD->vsUniformBufferBatches.clear(); + srbD->hsUniformBufferBatches.clear(); + srbD->dsUniformBufferBatches.clear(); + srbD->gsUniformBufferBatches.clear(); + srbD->fsUniformBufferBatches.clear(); + srbD->csUniformBufferBatches.clear(); - srbD->csubufs.clear(); - srbD->csubuforigbindings.clear(); - srbD->csubufoffsets.clear(); - srbD->csubufsizes.clear(); + srbD->vsSamplerBatches.clear(); + srbD->hsSamplerBatches.clear(); + srbD->dsSamplerBatches.clear(); + srbD->gsSamplerBatches.clear(); + srbD->fsSamplerBatches.clear(); + srbD->csSamplerBatches.clear(); - srbD->vssamplers.clear(); - srbD->vsshaderresources.clear(); - - srbD->fssamplers.clear(); - srbD->fsshaderresources.clear(); - - srbD->cssamplers.clear(); - srbD->csshaderresources.clear(); - - srbD->csUAVs.clear(); + srbD->csUavBatches.clear(); struct Stage { struct Buffer { @@ -1993,16 +2277,40 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, QVarLengthArray<Texture, 8> textures; QVarLengthArray<Sampler, 8> samplers; QVarLengthArray<Uav, 8> uavs; + void buildBufferBatches(QD3D11ShaderResourceBindings::StageUniformBufferBatches &batches) const + { + for (const Buffer &buf : buffers) { + batches.ubufs.feed(buf.breg, buf.buffer); + batches.ubuforigbindings.feed(buf.breg, UINT(buf.binding)); + batches.ubufoffsets.feed(buf.breg, buf.offsetInConstants); + batches.ubufsizes.feed(buf.breg, buf.sizeInConstants); + } + batches.finish(); + } + void buildSamplerBatches(QD3D11ShaderResourceBindings::StageSamplerBatches &batches) const + { + for (const Texture &t : textures) + batches.shaderresources.feed(t.treg, t.srv); + for (const Sampler &s : samplers) + batches.samplers.feed(s.sreg, s.sampler); + batches.finish(); + } + void buildUavBatches(QD3D11ShaderResourceBindings::StageUavBatches &batches) const + { + for (const Stage::Uav &u : uavs) + batches.uavs.feed(u.ureg, u.uav); + batches.finish(); + } } res[RBM_SUPPORTED_STAGES]; for (int i = 0, ie = srbD->sortedBindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->sortedBindings.at(i).data(); + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings.at(i)); QD3D11ShaderResourceBindings::BoundResourceData &bd(srbD->boundResourceData[i]); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: { QD3D11Buffer *bufD = QRHI_RES(QD3D11Buffer, b->u.ubuf.buf); - Q_ASSERT(aligned(b->u.ubuf.offset, 256) == b->u.ubuf.offset); + Q_ASSERT(aligned(b->u.ubuf.offset, 256u) == b->u.ubuf.offset); bd.ubuf.id = bufD->m_id; bd.ubuf.generation = bufD->generation; // Dynamic ubuf offsets are not considered here, those are baked in @@ -2011,16 +2319,31 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, // Metal) are different in this respect since those do not store // per-srb vsubufoffsets etc. data so life's a bit easier for them. // But here we have to defer baking in the dynamic offset. - const uint offsetInConstants = uint(b->u.ubuf.offset) / 16; + const quint32 offsetInConstants = b->u.ubuf.offset / 16; // size must be 16 mult. (in constants, i.e. multiple of 256 bytes). // We can round up if needed since the buffers's actual size // (ByteWidth) is always a multiple of 256. - const uint sizeInConstants = uint(aligned(b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size, 256) / 16); + const quint32 sizeInConstants = aligned(b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size, 256u) / 16; if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_VERTEX, nativeResourceBindingMaps); if (nativeBinding.first >= 0) res[RBM_VERTEX].buffers.append({ b->binding, nativeBinding.first, bufD->buffer, offsetInConstants, sizeInConstants }); } + if (b->stage.testFlag(QRhiShaderResourceBinding::TessellationControlStage)) { + QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_HULL, nativeResourceBindingMaps); + if (nativeBinding.first >= 0) + res[RBM_HULL].buffers.append({ b->binding, nativeBinding.first, bufD->buffer, offsetInConstants, sizeInConstants }); + } + if (b->stage.testFlag(QRhiShaderResourceBinding::TessellationEvaluationStage)) { + QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_DOMAIN, nativeResourceBindingMaps); + if (nativeBinding.first >= 0) + res[RBM_DOMAIN].buffers.append({ b->binding, nativeBinding.first, bufD->buffer, offsetInConstants, sizeInConstants }); + } + if (b->stage.testFlag(QRhiShaderResourceBinding::GeometryStage)) { + QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_GEOMETRY, nativeResourceBindingMaps); + if (nativeBinding.first >= 0) + res[RBM_GEOMETRY].buffers.append({ b->binding, nativeBinding.first, bufD->buffer, offsetInConstants, sizeInConstants }); + } if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_FRAGMENT, nativeResourceBindingMaps); if (nativeBinding.first >= 0) @@ -2034,10 +2357,15 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, } break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; bd.stex.count = data->count; const QPair<int, int> nativeBindingVert = mapBinding(b->binding, RBM_VERTEX, nativeResourceBindingMaps); + const QPair<int, int> nativeBindingHull = mapBinding(b->binding, RBM_HULL, nativeResourceBindingMaps); + const QPair<int, int> nativeBindingDomain = mapBinding(b->binding, RBM_DOMAIN, nativeResourceBindingMaps); + const QPair<int, int> nativeBindingGeom = mapBinding(b->binding, RBM_GEOMETRY, nativeResourceBindingMaps); const QPair<int, int> nativeBindingFrag = mapBinding(b->binding, RBM_FRAGMENT, nativeResourceBindingMaps); const QPair<int, int> nativeBindingComp = mapBinding(b->binding, RBM_COMPUTE, nativeResourceBindingMaps); // if SPIR-V binding b is mapped to tN and sN in HLSL, and it @@ -2046,27 +2374,61 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, for (int elem = 0; elem < data->count; ++elem) { QD3D11Texture *texD = QRHI_RES(QD3D11Texture, data->texSamplers[elem].tex); QD3D11Sampler *samplerD = QRHI_RES(QD3D11Sampler, data->texSamplers[elem].sampler); - bd.stex.d[elem].texId = texD->m_id; - bd.stex.d[elem].texGeneration = texD->generation; - bd.stex.d[elem].samplerId = samplerD->m_id; - bd.stex.d[elem].samplerGeneration = samplerD->generation; + bd.stex.d[elem].texId = texD ? texD->m_id : 0; + bd.stex.d[elem].texGeneration = texD ? texD->generation : 0; + bd.stex.d[elem].samplerId = samplerD ? samplerD->m_id : 0; + bd.stex.d[elem].samplerGeneration = samplerD ? samplerD->generation : 0; + // Must handle all three cases (combined, separate, separate): + // first = texture binding, second = sampler binding + // first = texture binding + // first = sampler binding if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { - if (nativeBindingVert.first >= 0 && nativeBindingVert.second >= 0) { + const int samplerBinding = texD && samplerD ? nativeBindingVert.second + : (samplerD ? nativeBindingVert.first : -1); + if (nativeBindingVert.first >= 0 && texD) res[RBM_VERTEX].textures.append({ nativeBindingVert.first + elem, texD->srv }); - res[RBM_VERTEX].samplers.append({ nativeBindingVert.second + elem, samplerD->samplerState }); - } + if (samplerBinding >= 0) + res[RBM_VERTEX].samplers.append({ samplerBinding + elem, samplerD->samplerState }); + } + if (b->stage.testFlag(QRhiShaderResourceBinding::TessellationControlStage)) { + const int samplerBinding = texD && samplerD ? nativeBindingHull.second + : (samplerD ? nativeBindingHull.first : -1); + if (nativeBindingHull.first >= 0 && texD) + res[RBM_HULL].textures.append({ nativeBindingHull.first + elem, texD->srv }); + if (samplerBinding >= 0) + res[RBM_HULL].samplers.append({ samplerBinding + elem, samplerD->samplerState }); + } + if (b->stage.testFlag(QRhiShaderResourceBinding::TessellationEvaluationStage)) { + const int samplerBinding = texD && samplerD ? nativeBindingDomain.second + : (samplerD ? nativeBindingDomain.first : -1); + if (nativeBindingDomain.first >= 0 && texD) + res[RBM_DOMAIN].textures.append({ nativeBindingDomain.first + elem, texD->srv }); + if (samplerBinding >= 0) + res[RBM_DOMAIN].samplers.append({ samplerBinding + elem, samplerD->samplerState }); + } + if (b->stage.testFlag(QRhiShaderResourceBinding::GeometryStage)) { + const int samplerBinding = texD && samplerD ? nativeBindingGeom.second + : (samplerD ? nativeBindingGeom.first : -1); + if (nativeBindingGeom.first >= 0 && texD) + res[RBM_GEOMETRY].textures.append({ nativeBindingGeom.first + elem, texD->srv }); + if (samplerBinding >= 0) + res[RBM_GEOMETRY].samplers.append({ samplerBinding + elem, samplerD->samplerState }); } if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { - if (nativeBindingFrag.first >= 0 && nativeBindingFrag.second >= 0) { + const int samplerBinding = texD && samplerD ? nativeBindingFrag.second + : (samplerD ? nativeBindingFrag.first : -1); + if (nativeBindingFrag.first >= 0 && texD) res[RBM_FRAGMENT].textures.append({ nativeBindingFrag.first + elem, texD->srv }); - res[RBM_FRAGMENT].samplers.append({ nativeBindingFrag.second + elem, samplerD->samplerState }); - } + if (samplerBinding >= 0) + res[RBM_FRAGMENT].samplers.append({ samplerBinding + elem, samplerD->samplerState }); } if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { - if (nativeBindingComp.first >= 0 && nativeBindingComp.second >= 0) { + const int samplerBinding = texD && samplerD ? nativeBindingComp.second + : (samplerD ? nativeBindingComp.first : -1); + if (nativeBindingComp.first >= 0 && texD) res[RBM_COMPUTE].textures.append({ nativeBindingComp.first + elem, texD->srv }); - res[RBM_COMPUTE].samplers.append({ nativeBindingComp.second + elem, samplerD->samplerState }); - } + if (samplerBinding >= 0) + res[RBM_COMPUTE].samplers.append({ samplerBinding + elem, samplerD->samplerState }); } } } @@ -2100,7 +2462,7 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { QPair<int, int> nativeBinding = mapBinding(b->binding, RBM_COMPUTE, nativeResourceBindingMaps); if (nativeBinding.first >= 0) { - ID3D11UnorderedAccessView *uav = bufD->unorderedAccessView(); + ID3D11UnorderedAccessView *uav = bufD->unorderedAccessView(b->u.sbuf.offset); if (uav) res[RBM_COMPUTE].uavs.append({ nativeBinding.first, uav }); } @@ -2134,63 +2496,21 @@ void QRhiD3D11::updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, }); } - for (const Stage::Buffer &buf : qAsConst(res[RBM_VERTEX].buffers)) { - srbD->vsubufs.feed(buf.breg, buf.buffer); - srbD->vsubuforigbindings.feed(buf.breg, UINT(buf.binding)); - srbD->vsubufoffsets.feed(buf.breg, buf.offsetInConstants); - srbD->vsubufsizes.feed(buf.breg, buf.sizeInConstants); - } - srbD->vsubufsPresent = srbD->vsubufs.finish(); - srbD->vsubuforigbindings.finish(); - srbD->vsubufoffsets.finish(); - srbD->vsubufsizes.finish(); - - for (const Stage::Buffer &buf : qAsConst(res[RBM_FRAGMENT].buffers)) { - srbD->fsubufs.feed(buf.breg, buf.buffer); - srbD->fsubuforigbindings.feed(buf.breg, UINT(buf.binding)); - srbD->fsubufoffsets.feed(buf.breg, buf.offsetInConstants); - srbD->fsubufsizes.feed(buf.breg, buf.sizeInConstants); - } - srbD->fsubufsPresent = srbD->fsubufs.finish(); - srbD->fsubuforigbindings.finish(); - srbD->fsubufoffsets.finish(); - srbD->fsubufsizes.finish(); - - for (const Stage::Buffer &buf : qAsConst(res[RBM_COMPUTE].buffers)) { - srbD->csubufs.feed(buf.breg, buf.buffer); - srbD->csubuforigbindings.feed(buf.breg, UINT(buf.binding)); - srbD->csubufoffsets.feed(buf.breg, buf.offsetInConstants); - srbD->csubufsizes.feed(buf.breg, buf.sizeInConstants); - } - srbD->csubufsPresent = srbD->csubufs.finish(); - srbD->csubuforigbindings.finish(); - srbD->csubufoffsets.finish(); - srbD->csubufsizes.finish(); - - for (const Stage::Texture &t : qAsConst(res[RBM_VERTEX].textures)) - srbD->vsshaderresources.feed(t.treg, t.srv); - for (const Stage::Sampler &s : qAsConst(res[RBM_VERTEX].samplers)) - srbD->vssamplers.feed(s.sreg, s.sampler); - srbD->vssamplersPresent = srbD->vssamplers.finish(); - srbD->vsshaderresources.finish(); - - for (const Stage::Texture &t : qAsConst(res[RBM_FRAGMENT].textures)) - srbD->fsshaderresources.feed(t.treg, t.srv); - for (const Stage::Sampler &s : qAsConst(res[RBM_FRAGMENT].samplers)) - srbD->fssamplers.feed(s.sreg, s.sampler); - srbD->fssamplersPresent = srbD->fssamplers.finish(); - srbD->fsshaderresources.finish(); - - for (const Stage::Texture &t : qAsConst(res[RBM_COMPUTE].textures)) - srbD->csshaderresources.feed(t.treg, t.srv); - for (const Stage::Sampler &s : qAsConst(res[RBM_COMPUTE].samplers)) - srbD->cssamplers.feed(s.sreg, s.sampler); - srbD->cssamplersPresent = srbD->cssamplers.finish(); - srbD->csshaderresources.finish(); - - for (const Stage::Uav &u : qAsConst(res[RBM_COMPUTE].uavs)) - srbD->csUAVs.feed(u.ureg, u.uav); - srbD->csUAVsPresent = srbD->csUAVs.finish(); + res[RBM_VERTEX].buildBufferBatches(srbD->vsUniformBufferBatches); + res[RBM_HULL].buildBufferBatches(srbD->hsUniformBufferBatches); + res[RBM_DOMAIN].buildBufferBatches(srbD->dsUniformBufferBatches); + res[RBM_GEOMETRY].buildBufferBatches(srbD->gsUniformBufferBatches); + res[RBM_FRAGMENT].buildBufferBatches(srbD->fsUniformBufferBatches); + res[RBM_COMPUTE].buildBufferBatches(srbD->csUniformBufferBatches); + + res[RBM_VERTEX].buildSamplerBatches(srbD->vsSamplerBatches); + res[RBM_HULL].buildSamplerBatches(srbD->hsSamplerBatches); + res[RBM_DOMAIN].buildSamplerBatches(srbD->dsSamplerBatches); + res[RBM_GEOMETRY].buildSamplerBatches(srbD->gsSamplerBatches); + res[RBM_FRAGMENT].buildSamplerBatches(srbD->fsSamplerBatches); + res[RBM_COMPUTE].buildSamplerBatches(srbD->csSamplerBatches); + + res[RBM_COMPUTE].buildUavBatches(srbD->csUavBatches); } void QRhiD3D11::executeBufferHostWrites(QD3D11Buffer *bufD) @@ -2203,17 +2523,18 @@ void QRhiD3D11::executeBufferHostWrites(QD3D11Buffer *bufD) D3D11_MAPPED_SUBRESOURCE mp; HRESULT hr = context->Map(bufD->buffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mp); if (SUCCEEDED(hr)) { - memcpy(mp.pData, bufD->dynBuf, size_t(bufD->m_size)); + memcpy(mp.pData, bufD->dynBuf, bufD->m_size); context->Unmap(bufD->buffer, 0); } else { - qWarning("Failed to map buffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to map buffer: %s", + qPrintable(QSystemError::windowsComString(hr))); } } static void applyDynamicOffsets(UINT *offsets, int batchIndex, - QRhiBatchedBindings<UINT> *originalBindings, - QRhiBatchedBindings<UINT> *staticOffsets, + const QRhiBatchedBindings<UINT> *originalBindings, + const QRhiBatchedBindings<UINT> *staticOffsets, const uint *dynOfsPairs, int dynOfsPairCount) { const int count = staticOffsets->batches[batchIndex].resources.count(); @@ -2244,162 +2565,92 @@ static inline uint clampedResourceCount(uint startSlot, int countSlots, uint max return countSlots; } +#define SETUBUFBATCH(stagePrefixL, stagePrefixU) \ + if (srbD->stagePrefixL##UniformBufferBatches.present) { \ + const QD3D11ShaderResourceBindings::StageUniformBufferBatches &batches(srbD->stagePrefixL##UniformBufferBatches); \ + for (int i = 0, ie = batches.ubufs.batches.count(); i != ie; ++i) { \ + const uint count = clampedResourceCount(batches.ubufs.batches[i].startBinding, \ + batches.ubufs.batches[i].resources.count(), \ + D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, \ + #stagePrefixU " cbuf"); \ + if (count) { \ + if (!dynOfsPairCount) { \ + context->stagePrefixU##SetConstantBuffers1(batches.ubufs.batches[i].startBinding, \ + count, \ + batches.ubufs.batches[i].resources.constData(), \ + batches.ubufoffsets.batches[i].resources.constData(), \ + batches.ubufsizes.batches[i].resources.constData()); \ + } else { \ + applyDynamicOffsets(offsets, i, \ + &batches.ubuforigbindings, &batches.ubufoffsets, \ + dynOfsPairs, dynOfsPairCount); \ + context->stagePrefixU##SetConstantBuffers1(batches.ubufs.batches[i].startBinding, \ + count, \ + batches.ubufs.batches[i].resources.constData(), \ + offsets, \ + batches.ubufsizes.batches[i].resources.constData()); \ + } \ + } \ + } \ + } + +#define SETSAMPLERBATCH(stagePrefixL, stagePrefixU) \ + if (srbD->stagePrefixL##SamplerBatches.present) { \ + for (const auto &batch : srbD->stagePrefixL##SamplerBatches.samplers.batches) { \ + const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), \ + D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT, #stagePrefixU " sampler"); \ + if (count) \ + context->stagePrefixU##SetSamplers(batch.startBinding, count, batch.resources.constData()); \ + } \ + for (const auto &batch : srbD->stagePrefixL##SamplerBatches.shaderresources.batches) { \ + const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), \ + D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT, #stagePrefixU " SRV"); \ + if (count) { \ + context->stagePrefixU##SetShaderResources(batch.startBinding, count, batch.resources.constData()); \ + contextState.stagePrefixL##HighestActiveSrvBinding = qMax(contextState.stagePrefixL##HighestActiveSrvBinding, \ + int(batch.startBinding + count) - 1); \ + } \ + } \ + } + +#define SETUAVBATCH(stagePrefixL, stagePrefixU) \ + if (srbD->stagePrefixL##UavBatches.present) { \ + for (const auto &batch : srbD->stagePrefixL##UavBatches.uavs.batches) { \ + const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), \ + D3D11_1_UAV_SLOT_COUNT, #stagePrefixU " UAV"); \ + if (count) { \ + context->stagePrefixU##SetUnorderedAccessViews(batch.startBinding, \ + count, \ + batch.resources.constData(), \ + nullptr); \ + contextState.stagePrefixL##HighestActiveUavBinding = qMax(contextState.stagePrefixL##HighestActiveUavBinding, \ + int(batch.startBinding + count) - 1); \ + } \ + } \ + } + void QRhiD3D11::bindShaderResources(QD3D11ShaderResourceBindings *srbD, const uint *dynOfsPairs, int dynOfsPairCount, bool offsetOnlyChange) { UINT offsets[QD3D11CommandBuffer::MAX_DYNAMIC_OFFSET_COUNT]; - if (srbD->vsubufsPresent) { - for (int i = 0, ie = srbD->vsubufs.batches.count(); i != ie; ++i) { - const uint count = clampedResourceCount(srbD->vsubufs.batches[i].startBinding, - srbD->vsubufs.batches[i].resources.count(), - D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, - "VS cbuf"); - if (count) { - if (!dynOfsPairCount) { - context->VSSetConstantBuffers1(srbD->vsubufs.batches[i].startBinding, - count, - srbD->vsubufs.batches[i].resources.constData(), - srbD->vsubufoffsets.batches[i].resources.constData(), - srbD->vsubufsizes.batches[i].resources.constData()); - } else { - applyDynamicOffsets(offsets, i, &srbD->vsubuforigbindings, &srbD->vsubufoffsets, - dynOfsPairs, dynOfsPairCount); - context->VSSetConstantBuffers1(srbD->vsubufs.batches[i].startBinding, - count, - srbD->vsubufs.batches[i].resources.constData(), - offsets, - srbD->vsubufsizes.batches[i].resources.constData()); - } - } - } - } - - if (srbD->fsubufsPresent) { - for (int i = 0, ie = srbD->fsubufs.batches.count(); i != ie; ++i) { - const uint count = clampedResourceCount(srbD->fsubufs.batches[i].startBinding, - srbD->fsubufs.batches[i].resources.count(), - D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, - "PS cbuf"); - if (count) { - if (!dynOfsPairCount) { - context->PSSetConstantBuffers1(srbD->fsubufs.batches[i].startBinding, - count, - srbD->fsubufs.batches[i].resources.constData(), - srbD->fsubufoffsets.batches[i].resources.constData(), - srbD->fsubufsizes.batches[i].resources.constData()); - } else { - applyDynamicOffsets(offsets, i, &srbD->fsubuforigbindings, &srbD->fsubufoffsets, - dynOfsPairs, dynOfsPairCount); - context->PSSetConstantBuffers1(srbD->fsubufs.batches[i].startBinding, - count, - srbD->fsubufs.batches[i].resources.constData(), - offsets, - srbD->fsubufsizes.batches[i].resources.constData()); - } - } - } - } - - if (srbD->csubufsPresent) { - for (int i = 0, ie = srbD->csubufs.batches.count(); i != ie; ++i) { - const uint count = clampedResourceCount(srbD->csubufs.batches[i].startBinding, - srbD->csubufs.batches[i].resources.count(), - D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT, - "CS cbuf"); - if (count) { - if (!dynOfsPairCount) { - context->CSSetConstantBuffers1(srbD->csubufs.batches[i].startBinding, - count, - srbD->csubufs.batches[i].resources.constData(), - srbD->csubufoffsets.batches[i].resources.constData(), - srbD->csubufsizes.batches[i].resources.constData()); - } else { - applyDynamicOffsets(offsets, i, &srbD->csubuforigbindings, &srbD->csubufoffsets, - dynOfsPairs, dynOfsPairCount); - context->CSSetConstantBuffers1(srbD->csubufs.batches[i].startBinding, - count, - srbD->csubufs.batches[i].resources.constData(), - offsets, - srbD->csubufsizes.batches[i].resources.constData()); - } - } - } - } + SETUBUFBATCH(vs, VS) + SETUBUFBATCH(hs, HS) + SETUBUFBATCH(ds, DS) + SETUBUFBATCH(gs, GS) + SETUBUFBATCH(fs, PS) + SETUBUFBATCH(cs, CS) if (!offsetOnlyChange) { - if (srbD->vssamplersPresent) { - for (const auto &batch : srbD->vssamplers.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT, "VS sampler"); - if (count) - context->VSSetSamplers(batch.startBinding, count, batch.resources.constData()); - } - - for (const auto &batch : srbD->vsshaderresources.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT, "VS SRV"); - if (count) { - context->VSSetShaderResources(batch.startBinding, count, batch.resources.constData()); - contextState.vsHighestActiveSrvBinding = qMax(contextState.vsHighestActiveSrvBinding, - int(batch.startBinding + count) - 1); - } - } - } - - if (srbD->fssamplersPresent) { - for (const auto &batch : srbD->fssamplers.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT, "PS sampler"); - if (count) - context->PSSetSamplers(batch.startBinding, count, batch.resources.constData()); - } - - for (const auto &batch : srbD->fsshaderresources.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT, "PS SRV"); - if (count) { - context->PSSetShaderResources(batch.startBinding, count, batch.resources.constData()); - contextState.fsHighestActiveSrvBinding = qMax(contextState.fsHighestActiveSrvBinding, - int(batch.startBinding + count) - 1); - } - } - } - - if (srbD->cssamplersPresent) { - for (const auto &batch : srbD->cssamplers.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT, "CS sampler"); - if (count) - context->CSSetSamplers(batch.startBinding, count, batch.resources.constData()); - } - - for (const auto &batch : srbD->csshaderresources.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_COMMONSHADER_INPUT_RESOURCE_SLOT_COUNT, "CS SRV"); - if (count) { - context->CSSetShaderResources(batch.startBinding, count, batch.resources.constData()); - contextState.csHighestActiveSrvBinding = qMax(contextState.csHighestActiveSrvBinding, - int(batch.startBinding + count) - 1); - } - } - } + SETSAMPLERBATCH(vs, VS) + SETSAMPLERBATCH(hs, HS) + SETSAMPLERBATCH(ds, DS) + SETSAMPLERBATCH(gs, GS) + SETSAMPLERBATCH(fs, PS) + SETSAMPLERBATCH(cs, CS) - if (srbD->csUAVsPresent) { - for (const auto &batch : srbD->csUAVs.batches) { - const uint count = clampedResourceCount(batch.startBinding, batch.resources.count(), - D3D11_1_UAV_SLOT_COUNT, "CS UAV"); - if (count) { - context->CSSetUnorderedAccessViews(batch.startBinding, - count, - batch.resources.constData(), - nullptr); - contextState.csHighestActiveUavBinding = qMax(contextState.csHighestActiveUavBinding, - int(batch.startBinding + count) - 1); - } - } - } + SETUAVBATCH(cs, CS) } } @@ -2428,6 +2679,9 @@ void QRhiD3D11::resetShaderResources() } int nullsrvCount = qMax(contextState.vsHighestActiveSrvBinding, contextState.fsHighestActiveSrvBinding); + nullsrvCount = qMax(nullsrvCount, contextState.hsHighestActiveSrvBinding); + nullsrvCount = qMax(nullsrvCount, contextState.dsHighestActiveSrvBinding); + nullsrvCount = qMax(nullsrvCount, contextState.gsHighestActiveSrvBinding); nullsrvCount = qMax(nullsrvCount, contextState.csHighestActiveSrvBinding); nullsrvCount += 1; if (nullsrvCount > 0) { @@ -2439,6 +2693,18 @@ void QRhiD3D11::resetShaderResources() context->VSSetShaderResources(0, UINT(contextState.vsHighestActiveSrvBinding + 1), nullsrvs.constData()); contextState.vsHighestActiveSrvBinding = -1; } + if (contextState.hsHighestActiveSrvBinding >= 0) { + context->HSSetShaderResources(0, UINT(contextState.hsHighestActiveSrvBinding + 1), nullsrvs.constData()); + contextState.hsHighestActiveSrvBinding = -1; + } + if (contextState.dsHighestActiveSrvBinding >= 0) { + context->DSSetShaderResources(0, UINT(contextState.dsHighestActiveSrvBinding + 1), nullsrvs.constData()); + contextState.dsHighestActiveSrvBinding = -1; + } + if (contextState.gsHighestActiveSrvBinding >= 0) { + context->GSSetShaderResources(0, UINT(contextState.gsHighestActiveSrvBinding + 1), nullsrvs.constData()); + contextState.gsHighestActiveSrvBinding = -1; + } if (contextState.fsHighestActiveSrvBinding >= 0) { context->PSSetShaderResources(0, UINT(contextState.fsHighestActiveSrvBinding + 1), nullsrvs.constData()); contextState.fsHighestActiveSrvBinding = -1; @@ -2460,31 +2726,52 @@ void QRhiD3D11::resetShaderResources() } } -void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain *timestampSwapChain) +#define SETSHADER(StageL, StageU) \ + if (psD->StageL.shader) { \ + context->StageU##SetShader(psD->StageL.shader, nullptr, 0); \ + currentShaderMask |= StageU##MaskBit; \ + } else if (currentShaderMask & StageU##MaskBit) { \ + context->StageU##SetShader(nullptr, nullptr, 0); \ + currentShaderMask &= ~StageU##MaskBit; \ + } + +void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD) { quint32 stencilRef = 0; float blendConstants[] = { 1, 1, 1, 1 }; - - if (timestampSwapChain) { - const int currentFrameSlot = timestampSwapChain->currentFrameSlot; - ID3D11Query *tsDisjoint = timestampSwapChain->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = timestampSwapChain->timestampQuery[tsIdx]; - if (tsDisjoint && tsStart && !timestampSwapChain->timestampActive[currentFrameSlot]) { - // The timestamps seem to include vsync time with Present(1), except - // when running on a non-primary gpu. This is not ideal. So try working - // it around by issuing a semi-fake OMSetRenderTargets early and - // writing the first timestamp only afterwards. - context->Begin(tsDisjoint); - QD3D11RenderTargetData *rtD = rtData(×tampSwapChain->rt); - context->OMSetRenderTargets(UINT(rtD->colorAttCount), rtD->colorAttCount ? rtD->rtv : nullptr, rtD->dsv); - context->End(tsStart); // just record a timestamp, no Begin needed - } - } + enum ActiveShaderMask { + VSMaskBit = 0x01, + HSMaskBit = 0x02, + DSMaskBit = 0x04, + GSMaskBit = 0x08, + PSMaskBit = 0x10 + }; + int currentShaderMask = 0xFF; for (auto it = cbD->commands.cbegin(), end = cbD->commands.cend(); it != end; ++it) { const QD3D11CommandBuffer::Command &cmd(*it); switch (cmd.cmd) { + case QD3D11CommandBuffer::Command::BeginFrame: + if (cmd.args.beginFrame.tsDisjointQuery) + context->Begin(cmd.args.beginFrame.tsDisjointQuery); + if (cmd.args.beginFrame.tsQuery) { + if (cmd.args.beginFrame.swapchainData) { + // The timestamps seem to include vsync time with Present(1), except + // when running on a non-primary gpu. This is not ideal. So try working + // it around by issuing a semi-fake OMSetRenderTargets early and + // writing the first timestamp only afterwards. + QD3D11RenderTargetData *rtD = cmd.args.beginFrame.swapchainData; + context->OMSetRenderTargets(UINT(rtD->colorAttCount), rtD->colorAttCount ? rtD->rtv : nullptr, rtD->dsv); + } + context->End(cmd.args.beginFrame.tsQuery); // no Begin() for D3D11_QUERY_TIMESTAMP + } + break; + case QD3D11CommandBuffer::Command::EndFrame: + if (cmd.args.endFrame.tsQuery) + context->End(cmd.args.endFrame.tsQuery); + if (cmd.args.endFrame.tsDisjointQuery) + context->End(cmd.args.endFrame.tsDisjointQuery); + break; case QD3D11CommandBuffer::Command::ResetShaderResources: resetShaderResources(); break; @@ -2552,8 +2839,11 @@ void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain * case QD3D11CommandBuffer::Command::BindGraphicsPipeline: { QD3D11GraphicsPipeline *psD = cmd.args.bindGraphicsPipeline.ps; - context->VSSetShader(psD->vs.shader, nullptr, 0); - context->PSSetShader(psD->fs.shader, nullptr, 0); + SETSHADER(vs, VS) + SETSHADER(hs, HS) + SETSHADER(ds, DS) + SETSHADER(gs, GS) + SETSHADER(fs, PS) context->IASetPrimitiveTopology(psD->d3dTopology); context->IASetInputLayout(psD->inputLayout); // may be null, that's ok context->OMSetDepthStencilState(psD->dsState, stencilRef); @@ -2639,7 +2929,7 @@ void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain * } } -QD3D11Buffer::QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size) +QD3D11Buffer::QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) : QRhiBuffer(rhi, type, usage, size) { } @@ -2660,15 +2950,13 @@ void QD3D11Buffer::destroy() delete[] dynBuf; dynBuf = nullptr; - if (uav) { - uav->Release(); - uav = nullptr; - } + for (auto it = uavs.begin(), end = uavs.end(); it != end; ++it) + it.value()->Release(); + uavs.clear(); QRHI_RES_RHI(QRhiD3D11); - QRHI_PROF; - QRHI_PROF_F(releaseBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) + rhiD->unregisterResource(this); } static inline uint toD3DBufferUsage(QRhiBuffer::UsageFlags usage) @@ -2700,12 +2988,11 @@ bool QD3D11Buffer::create() return false; } - const int nonZeroSize = m_size <= 0 ? 256 : m_size; - const int roundedSize = aligned(nonZeroSize, m_usage.testFlag(QRhiBuffer::UniformBuffer) ? 256 : 4); + const quint32 nonZeroSize = m_size <= 0 ? 256 : m_size; + const quint32 roundedSize = aligned(nonZeroSize, m_usage.testFlag(QRhiBuffer::UniformBuffer) ? 256u : 4u); - D3D11_BUFFER_DESC desc; - memset(&desc, 0, sizeof(desc)); - desc.ByteWidth = UINT(roundedSize); + D3D11_BUFFER_DESC desc = {}; + desc.ByteWidth = roundedSize; desc.Usage = m_type == Dynamic ? D3D11_USAGE_DYNAMIC : D3D11_USAGE_DEFAULT; desc.BindFlags = toD3DBufferUsage(m_usage); desc.CPUAccessFlags = m_type == Dynamic ? D3D11_CPU_ACCESS_WRITE : 0; @@ -2714,7 +3001,8 @@ bool QD3D11Buffer::create() QRHI_RES_RHI(QRhiD3D11); HRESULT hr = rhiD->dev->CreateBuffer(&desc, nullptr, &buffer); if (FAILED(hr)) { - qWarning("Failed to create buffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create buffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } @@ -2726,9 +3014,6 @@ bool QD3D11Buffer::create() if (!m_objectName.isEmpty()) buffer->SetPrivateData(WKPDID_D3DDebugObjectName, UINT(m_objectName.size()), m_objectName.constData()); - QRHI_PROF; - QRHI_PROF_F(newBuffer(this, quint32(roundedSize), m_type == Dynamic ? 2 : 1, m_type == Dynamic ? 1 : 0)); - generation += 1; rhiD->registerResource(this); return true; @@ -2756,7 +3041,8 @@ char *QD3D11Buffer::beginFullDynamicBufferUpdateForCurrentFrame() QRHI_RES_RHI(QRhiD3D11); HRESULT hr = rhiD->context->Map(buffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mp); if (FAILED(hr)) { - qWarning("Failed to map buffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to map buffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return nullptr; } return static_cast<char *>(mp.pData); @@ -2768,27 +3054,30 @@ void QD3D11Buffer::endFullDynamicBufferUpdateForCurrentFrame() rhiD->context->Unmap(buffer, 0); } -ID3D11UnorderedAccessView *QD3D11Buffer::unorderedAccessView() +ID3D11UnorderedAccessView *QD3D11Buffer::unorderedAccessView(quint32 offset) { - if (uav) - return uav; + auto it = uavs.find(offset); + if (it != uavs.end()) + return it.value(); // SPIRV-Cross generated HLSL uses RWByteAddressBuffer - D3D11_UNORDERED_ACCESS_VIEW_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_UNORDERED_ACCESS_VIEW_DESC desc = {}; desc.Format = DXGI_FORMAT_R32_TYPELESS; desc.ViewDimension = D3D11_UAV_DIMENSION_BUFFER; - desc.Buffer.FirstElement = 0; - desc.Buffer.NumElements = UINT(aligned(m_size, 4) / 4); + desc.Buffer.FirstElement = offset / 4u; + desc.Buffer.NumElements = aligned(m_size - offset, 4u) / 4u; desc.Buffer.Flags = D3D11_BUFFER_UAV_FLAG_RAW; QRHI_RES_RHI(QRhiD3D11); + ID3D11UnorderedAccessView *uav = nullptr; HRESULT hr = rhiD->dev->CreateUnorderedAccessView(buffer, &desc, &uav); if (FAILED(hr)) { - qWarning("Failed to create UAV: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create UAV: %s", + qPrintable(QSystemError::windowsComString(hr))); return nullptr; } + uavs[offset] = uav; return uav; } @@ -2823,9 +3112,8 @@ void QD3D11RenderBuffer::destroy() tex = nullptr; QRHI_RES_RHI(QRhiD3D11); - QRHI_PROF; - QRHI_PROF_F(releaseRenderBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11RenderBuffer::create() @@ -2837,10 +3125,9 @@ bool QD3D11RenderBuffer::create() return false; QRHI_RES_RHI(QRhiD3D11); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); - D3D11_TEXTURE2D_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_TEXTURE2D_DESC desc = {}; desc.Width = UINT(m_pixelSize.width()); desc.Height = UINT(m_pixelSize.height()); desc.MipLevels = 1; @@ -2855,17 +3142,18 @@ bool QD3D11RenderBuffer::create() desc.BindFlags = D3D11_BIND_RENDER_TARGET; HRESULT hr = rhiD->dev->CreateTexture2D(&desc, nullptr, &tex); if (FAILED(hr)) { - qWarning("Failed to create color renderbuffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create color renderbuffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; - memset(&rtvDesc, 0, sizeof(rtvDesc)); + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc = {}; rtvDesc.Format = dxgiFormat; rtvDesc.ViewDimension = desc.SampleDesc.Count > 1 ? D3D11_RTV_DIMENSION_TEXTURE2DMS : D3D11_RTV_DIMENSION_TEXTURE2D; hr = rhiD->dev->CreateRenderTargetView(tex, &rtvDesc, &rtv); if (FAILED(hr)) { - qWarning("Failed to create rtv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create rtv: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } } else if (m_type == DepthStencil) { @@ -2874,17 +3162,18 @@ bool QD3D11RenderBuffer::create() desc.BindFlags = D3D11_BIND_DEPTH_STENCIL; HRESULT hr = rhiD->dev->CreateTexture2D(&desc, nullptr, &tex); if (FAILED(hr)) { - qWarning("Failed to create depth-stencil buffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create depth-stencil buffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; - memset(&dsvDesc, 0, sizeof(dsvDesc)); + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc = {}; dsvDesc.Format = dxgiFormat; dsvDesc.ViewDimension = desc.SampleDesc.Count > 1 ? D3D11_DSV_DIMENSION_TEXTURE2DMS : D3D11_DSV_DIMENSION_TEXTURE2D; hr = rhiD->dev->CreateDepthStencilView(tex, &dsvDesc, &dsv); if (FAILED(hr)) { - qWarning("Failed to create dsv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create dsv: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } } else { @@ -2894,9 +3183,7 @@ bool QD3D11RenderBuffer::create() if (!m_objectName.isEmpty()) tex->SetPrivateData(WKPDID_D3DDebugObjectName, UINT(m_objectName.size()), m_objectName.constData()); - QRHI_PROF; - QRHI_PROF_F(newRenderBuffer(this, false, false, int(sampleDesc.Count))); - + generation += 1; rhiD->registerResource(this); return true; } @@ -2910,8 +3197,8 @@ QRhiTexture::Format QD3D11RenderBuffer::backingFormat() const } QD3D11Texture::QD3D11Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags) - : QRhiTexture(rhi, format, pixelSize, depth, sampleCount, flags) + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags) { for (int i = 0; i < QRhi::MAX_MIP_LEVELS; ++i) perLevelViews[i] = nullptr; @@ -2924,7 +3211,7 @@ QD3D11Texture::~QD3D11Texture() void QD3D11Texture::destroy() { - if (!tex && !tex3D) + if (!tex && !tex3D && !tex1D) return; if (srv) { @@ -2944,15 +3231,17 @@ void QD3D11Texture::destroy() tex->Release(); if (tex3D) tex3D->Release(); + if (tex1D) + tex1D->Release(); } tex = nullptr; tex3D = nullptr; + tex1D = nullptr; QRHI_RES_RHI(QRhiD3D11); - QRHI_PROF; - QRHI_PROF_F(releaseTexture(this)); - rhiD->unregisterResource(this); + if (rhiD) + rhiD->unregisterResource(this); } static inline DXGI_FORMAT toD3DDepthTextureSRVFormat(QRhiTexture::Format format) @@ -2978,7 +3267,7 @@ static inline DXGI_FORMAT toD3DDepthTextureDSVFormat(QRhiTexture::Format format) case QRhiTexture::Format::D16: return DXGI_FORMAT_D16_UNORM; case QRhiTexture::Format::D24: - return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + return DXGI_FORMAT_D24_UNORM_S8_UINT; case QRhiTexture::Format::D24S8: return DXGI_FORMAT_D24_UNORM_S8_UINT; case QRhiTexture::Format::D32F: @@ -2991,19 +3280,23 @@ static inline DXGI_FORMAT toD3DDepthTextureDSVFormat(QRhiTexture::Format format) bool QD3D11Texture::prepareCreate(QSize *adjustedSize) { - if (tex || tex3D) + if (tex || tex3D || tex1D) destroy(); - const QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; const bool isDepth = isDepthTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); const bool hasMipMaps = m_flags.testFlag(MipMapped); + const bool is1D = m_flags.testFlag(OneDimensional); + + const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); QRHI_RES_RHI(QRhiD3D11); dxgiFormat = toD3DTextureFormat(m_format, m_flags); mipLevelCount = uint(hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); if (sampleDesc.Count > 1) { if (isCube) { qWarning("Cubemap texture cannot be multisample"); @@ -3026,11 +3319,30 @@ bool QD3D11Texture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both cube and 3D"); return false; } - m_depth = qMax(1, m_depth); + if (isArray && is3D) { + qWarning("Texture cannot be both array and 3D"); + return false; + } + if (isCube && is1D) { + qWarning("Texture cannot be both cube and 1D"); + return false; + } + if (is1D && is3D) { + qWarning("Texture cannot be both 1D and 3D"); + return false; + } if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } + if (m_arraySize > 0 && !isArray) { + qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); + return false; + } + if (m_arraySize < 1 && isArray) { + qWarning("Texture is an array but array size is %d", m_arraySize); + return false; + } if (adjustedSize) *adjustedSize = size; @@ -3044,28 +3356,68 @@ bool QD3D11Texture::finishCreate() const bool isDepth = isDepthTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); + const bool is1D = m_flags.testFlag(OneDimensional); - D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; - memset(&srvDesc, 0, sizeof(srvDesc)); + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc = {}; srvDesc.Format = isDepth ? toD3DDepthTextureSRVFormat(m_format) : dxgiFormat; if (isCube) { srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; srvDesc.TextureCube.MipLevels = mipLevelCount; } else { - if (sampleDesc.Count > 1) { - srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DMS; - } else if (is3D) { - srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE3D; - srvDesc.Texture3D.MipLevels = mipLevelCount; + if (is1D) { + if (isArray) { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE1DARRAY; + srvDesc.Texture1DArray.MipLevels = mipLevelCount; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture1DArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture1DArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture1DArray.FirstArraySlice = 0; + srvDesc.Texture1DArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } else { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE1D; + srvDesc.Texture1D.MipLevels = mipLevelCount; + } + } else if (isArray) { + if (sampleDesc.Count > 1) { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DMSARRAY; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture2DMSArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture2DMSArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture2DMSArray.FirstArraySlice = 0; + srvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } else { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MipLevels = mipLevelCount; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture2DArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture2DArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture2DArray.FirstArraySlice = 0; + srvDesc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } } else { - srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; - srvDesc.Texture2D.MipLevels = mipLevelCount; + if (sampleDesc.Count > 1) { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DMS; + } else if (is3D) { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE3D; + srvDesc.Texture3D.MipLevels = mipLevelCount; + } else { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MipLevels = mipLevelCount; + } } } HRESULT hr = rhiD->dev->CreateShaderResourceView(textureResource(), &srvDesc, &srv); if (FAILED(hr)) { - qWarning("Failed to create srv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create srv: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } @@ -3082,6 +3434,8 @@ bool QD3D11Texture::create() const bool isDepth = isDepthTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); + const bool is1D = m_flags.testFlag(OneDimensional); uint bindFlags = D3D11_BIND_SHADER_RESOURCE; uint miscFlags = isCube ? D3D11_RESOURCE_MISC_TEXTURECUBE : 0; @@ -3103,13 +3457,31 @@ bool QD3D11Texture::create() bindFlags |= D3D11_BIND_UNORDERED_ACCESS; QRHI_RES_RHI(QRhiD3D11); - if (!is3D) { - D3D11_TEXTURE2D_DESC desc; - memset(&desc, 0, sizeof(desc)); + if (is1D) { + D3D11_TEXTURE1D_DESC desc = {}; + desc.Width = UINT(size.width()); + desc.MipLevels = mipLevelCount; + desc.ArraySize = isArray ? UINT(qMax(0, m_arraySize)) : 1; + desc.Format = dxgiFormat; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = bindFlags; + desc.MiscFlags = miscFlags; + + HRESULT hr = rhiD->dev->CreateTexture1D(&desc, nullptr, &tex1D); + if (FAILED(hr)) { + qWarning("Failed to create 1D texture: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + if (!m_objectName.isEmpty()) + tex->SetPrivateData(WKPDID_D3DDebugObjectName, UINT(m_objectName.size()), + m_objectName.constData()); + } else if (!is3D) { + D3D11_TEXTURE2D_DESC desc = {}; desc.Width = UINT(size.width()); desc.Height = UINT(size.height()); desc.MipLevels = mipLevelCount; - desc.ArraySize = isCube ? 6 : 1; + desc.ArraySize = isCube ? 6 : (isArray ? UINT(qMax(0, m_arraySize)) : 1); desc.Format = dxgiFormat; desc.SampleDesc = sampleDesc; desc.Usage = D3D11_USAGE_DEFAULT; @@ -3118,17 +3490,17 @@ bool QD3D11Texture::create() HRESULT hr = rhiD->dev->CreateTexture2D(&desc, nullptr, &tex); if (FAILED(hr)) { - qWarning("Failed to create 2D texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create 2D texture: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } if (!m_objectName.isEmpty()) tex->SetPrivateData(WKPDID_D3DDebugObjectName, UINT(m_objectName.size()), m_objectName.constData()); } else { - D3D11_TEXTURE3D_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_TEXTURE3D_DESC desc = {}; desc.Width = UINT(size.width()); desc.Height = UINT(size.height()); - desc.Depth = UINT(m_depth); + desc.Depth = UINT(qMax(1, m_depth)); desc.MipLevels = mipLevelCount; desc.Format = dxgiFormat; desc.Usage = D3D11_USAGE_DEFAULT; @@ -3137,7 +3509,8 @@ bool QD3D11Texture::create() HRESULT hr = rhiD->dev->CreateTexture3D(&desc, nullptr, &tex3D); if (FAILED(hr)) { - qWarning("Failed to create 3D texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create 3D texture: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } if (!m_objectName.isEmpty()) @@ -3147,9 +3520,6 @@ bool QD3D11Texture::create() if (!finishCreate()) return false; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, true, int(mipLevelCount), isCube ? 6 : 1, int(sampleDesc.Count))); - owns = true; rhiD->registerResource(this); return true; @@ -3165,15 +3535,14 @@ bool QD3D11Texture::createFrom(QRhiTexture::NativeTexture src) if (m_flags.testFlag(ThreeDimensional)) tex3D = reinterpret_cast<ID3D11Texture3D *>(src.object); + else if (m_flags.testFlags(OneDimensional)) + tex1D = reinterpret_cast<ID3D11Texture1D *>(src.object); else tex = reinterpret_cast<ID3D11Texture2D *>(src.object); if (!finishCreate()) return false; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, false, int(mipLevelCount), m_flags.testFlag(CubeMap) ? 6 : 1, int(sampleDesc.Count))); - owns = false; QRHI_RES_RHI(QRhiD3D11); rhiD->registerResource(this); @@ -3191,15 +3560,20 @@ ID3D11UnorderedAccessView *QD3D11Texture::unorderedAccessViewForLevel(int level) return perLevelViews[level]; const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); - D3D11_UNORDERED_ACCESS_VIEW_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_UNORDERED_ACCESS_VIEW_DESC desc = {}; desc.Format = dxgiFormat; if (isCube) { desc.ViewDimension = D3D11_UAV_DIMENSION_TEXTURE2DARRAY; desc.Texture2DArray.MipSlice = UINT(level); desc.Texture2DArray.FirstArraySlice = 0; desc.Texture2DArray.ArraySize = 6; + } else if (isArray) { + desc.ViewDimension = D3D11_UAV_DIMENSION_TEXTURE2DARRAY; + desc.Texture2DArray.MipSlice = UINT(level); + desc.Texture2DArray.FirstArraySlice = 0; + desc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); } else if (is3D) { desc.ViewDimension = D3D11_UAV_DIMENSION_TEXTURE3D; desc.Texture3D.MipSlice = UINT(level); @@ -3212,7 +3586,8 @@ ID3D11UnorderedAccessView *QD3D11Texture::unorderedAccessViewForLevel(int level) ID3D11UnorderedAccessView *uav = nullptr; HRESULT hr = rhiD->dev->CreateUnorderedAccessView(textureResource(), &desc, &uav); if (FAILED(hr)) { - qWarning("Failed to create UAV: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create UAV: %s", + qPrintable(QSystemError::windowsComString(hr))); return nullptr; } @@ -3240,7 +3615,8 @@ void QD3D11Sampler::destroy() samplerState = nullptr; QRHI_RES_RHI(QRhiD3D11); - rhiD->unregisterResource(this); + if (rhiD) + rhiD->unregisterResource(this); } static inline D3D11_FILTER toD3DFilter(QRhiSampler::Filter minFilter, QRhiSampler::Filter magFilter, QRhiSampler::Filter mipFilter) @@ -3320,8 +3696,7 @@ bool QD3D11Sampler::create() if (samplerState) destroy(); - D3D11_SAMPLER_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_SAMPLER_DESC desc = {}; desc.Filter = toD3DFilter(m_minFilter, m_magFilter, m_mipmapMode); if (m_compareOp != Never) desc.Filter = D3D11_FILTER(desc.Filter | 0x80); @@ -3335,7 +3710,8 @@ bool QD3D11Sampler::create() QRHI_RES_RHI(QRhiD3D11); HRESULT hr = rhiD->dev->CreateSamplerState(&desc, &samplerState); if (FAILED(hr)) { - qWarning("Failed to create sampler state: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create sampler state: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } @@ -3357,7 +3733,9 @@ QD3D11RenderPassDescriptor::~QD3D11RenderPassDescriptor() void QD3D11RenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -3368,36 +3746,44 @@ bool QD3D11RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *ot QRhiRenderPassDescriptor *QD3D11RenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } -QD3D11ReferenceRenderTarget::QD3D11ReferenceRenderTarget(QRhiImplementation *rhi) - : QRhiRenderTarget(rhi), +QVector<quint32> QD3D11RenderPassDescriptor::serializedFormat() const +{ + return {}; +} + +QD3D11SwapChainRenderTarget::QD3D11SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain), d(rhi) { } -QD3D11ReferenceRenderTarget::~QD3D11ReferenceRenderTarget() +QD3D11SwapChainRenderTarget::~QD3D11SwapChainRenderTarget() { destroy(); } -void QD3D11ReferenceRenderTarget::destroy() +void QD3D11SwapChainRenderTarget::destroy() { // nothing to do here } -QSize QD3D11ReferenceRenderTarget::pixelSize() const +QSize QD3D11SwapChainRenderTarget::pixelSize() const { return d.pixelSize; } -float QD3D11ReferenceRenderTarget::devicePixelRatio() const +float QD3D11SwapChainRenderTarget::devicePixelRatio() const { return d.dpr; } -int QD3D11ReferenceRenderTarget::sampleCount() const +int QD3D11SwapChainRenderTarget::sampleCount() const { return d.sampleCount; } @@ -3421,8 +3807,6 @@ QD3D11TextureRenderTarget::~QD3D11TextureRenderTarget() void QD3D11TextureRenderTarget::destroy() { - QRHI_RES_RHI(QRhiD3D11); - if (!rtv[0] && !dsv) return; @@ -3440,12 +3824,17 @@ void QD3D11TextureRenderTarget::destroy() } } - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } QRhiRenderPassDescriptor *QD3D11TextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } bool QD3D11TextureRenderTarget::create() @@ -3453,8 +3842,7 @@ bool QD3D11TextureRenderTarget::create() if (rtv[0] || dsv) destroy(); - const bool hasColorAttachments = m_desc.cbeginColorAttachments() != m_desc.cendColorAttachments(); - Q_ASSERT(hasColorAttachments || m_desc.depthTexture()); + Q_ASSERT(m_desc.colorAttachmentCount() > 0 || m_desc.depthTexture()); Q_ASSERT(!m_desc.depthStencilBuffer() || !m_desc.depthTexture()); const bool hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); @@ -3470,14 +3858,34 @@ bool QD3D11TextureRenderTarget::create() Q_ASSERT(texture || rb); if (texture) { QD3D11Texture *texD = QRHI_RES(QD3D11Texture, texture); - D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; - memset(&rtvDesc, 0, sizeof(rtvDesc)); + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc = {}; rtvDesc.Format = toD3DTextureFormat(texD->format(), texD->flags()); if (texD->flags().testFlag(QRhiTexture::CubeMap)) { rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); rtvDesc.Texture2DArray.ArraySize = 1; + } else if (texD->flags().testFlag(QRhiTexture::OneDimensional)) { + if (texD->flags().testFlag(QRhiTexture::TextureArray)) { + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE1DARRAY; + rtvDesc.Texture1DArray.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture1DArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture1DArray.ArraySize = 1; + } else { + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE1D; + rtvDesc.Texture1D.MipSlice = UINT(colorAtt.level()); + } + } else if (texD->flags().testFlag(QRhiTexture::TextureArray)) { + if (texD->sampleDesc.Count > 1) { + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DMSARRAY; + rtvDesc.Texture2DMSArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture2DMSArray.ArraySize = 1; + } else { + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture2DArray.ArraySize = 1; + } } else if (texD->flags().testFlag(QRhiTexture::ThreeDimensional)) { rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE3D; rtvDesc.Texture3D.MipSlice = UINT(colorAtt.level()); @@ -3493,7 +3901,8 @@ bool QD3D11TextureRenderTarget::create() } HRESULT hr = rhiD->dev->CreateRenderTargetView(texD->textureResource(), &rtvDesc, &rtv[attIndex]); if (FAILED(hr)) { - qWarning("Failed to create rtv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create rtv: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } ownsRtv[attIndex] = true; @@ -3517,14 +3926,35 @@ bool QD3D11TextureRenderTarget::create() if (m_desc.depthTexture()) { ownsDsv = true; QD3D11Texture *depthTexD = QRHI_RES(QD3D11Texture, m_desc.depthTexture()); - D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; - memset(&dsvDesc, 0, sizeof(dsvDesc)); + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc = {}; dsvDesc.Format = toD3DDepthTextureDSVFormat(depthTexD->format()); dsvDesc.ViewDimension = depthTexD->sampleDesc.Count > 1 ? D3D11_DSV_DIMENSION_TEXTURE2DMS : D3D11_DSV_DIMENSION_TEXTURE2D; + if (depthTexD->flags().testFlag(QRhiTexture::TextureArray)) { + if (depthTexD->sampleDesc.Count > 1) { + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DMSARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DMSArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DMSArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DMSArray.FirstArraySlice = 0; + dsvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } else { + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DArray.FirstArraySlice = 0; + dsvDesc.Texture2DArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } + } HRESULT hr = rhiD->dev->CreateDepthStencilView(depthTexD->tex, &dsvDesc, &dsv); if (FAILED(hr)) { - qWarning("Failed to create dsv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create dsv: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } if (d.colorAttCount == 0) { @@ -3551,12 +3981,17 @@ bool QD3D11TextureRenderTarget::create() d.dsv = dsv; d.rp = QRHI_RES(QD3D11RenderPassDescriptor, m_renderPassDesc); + QRhiRenderTargetAttachmentTracker::updateResIdList<QD3D11Texture, QD3D11RenderBuffer>(m_desc, &d.currentResIdList); + rhiD->registerResource(this); return true; } QSize QD3D11TextureRenderTarget::pixelSize() const { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QD3D11Texture, QD3D11RenderBuffer>(m_desc, d.currentResIdList)) + const_cast<QD3D11TextureRenderTarget *>(this)->create(); + return d.pixelSize; } @@ -3584,6 +4019,10 @@ void QD3D11ShaderResourceBindings::destroy() { sortedBindings.clear(); boundResourceData.clear(); + + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11ShaderResourceBindings::create() @@ -3598,11 +4037,7 @@ bool QD3D11ShaderResourceBindings::create() rhiD->updateLayoutDesc(this); std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); - std::sort(sortedBindings.begin(), sortedBindings.end(), - [](const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) - { - return a.data()->binding < b.data()->binding; - }); + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); boundResourceData.resize(sortedBindings.count()); @@ -3611,7 +4046,7 @@ bool QD3D11ShaderResourceBindings::create() hasDynamicOffset = false; for (const QRhiShaderResourceBinding &b : sortedBindings) { - const QRhiShaderResourceBinding::Data *bd = b.data(); + const QRhiShaderResourceBinding::Data *bd = QRhiImplementation::shaderResourceBindingData(b); if (bd->type == QRhiShaderResourceBinding::UniformBuffer && bd->u.ubuf.hasDynamicOffset) { hasDynamicOffset = true; break; @@ -3619,9 +4054,24 @@ bool QD3D11ShaderResourceBindings::create() } generation += 1; + rhiD->registerResource(this, false); return true; } +void QD3D11ShaderResourceBindings::updateResources(UpdateFlags flags) +{ + sortedBindings.clear(); + std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); + if (!flags.testFlag(BindingsAreSorted)) + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); + + Q_ASSERT(boundResourceData.count() == sortedBindings.count()); + for (BoundResourceData &bd : boundResourceData) + memset(&bd, 0, sizeof(BoundResourceData)); + + generation += 1; +} + QD3D11GraphicsPipeline::QD3D11GraphicsPipeline(QRhiImplementation *rhi) : QRhiGraphicsPipeline(rhi) { @@ -3632,10 +4082,18 @@ QD3D11GraphicsPipeline::~QD3D11GraphicsPipeline() destroy(); } -void QD3D11GraphicsPipeline::destroy() +template<typename T> +inline void releasePipelineShader(T &s) { - QRHI_RES_RHI(QRhiD3D11); + if (s.shader) { + s.shader->Release(); + s.shader = nullptr; + } + s.nativeResourceBindingMap.clear(); +} +void QD3D11GraphicsPipeline::destroy() +{ if (!dsState) return; @@ -3657,19 +4115,15 @@ void QD3D11GraphicsPipeline::destroy() rastState = nullptr; } - if (vs.shader) { - vs.shader->Release(); - vs.shader = nullptr; - } - vs.nativeResourceBindingMap.clear(); + releasePipelineShader(vs); + releasePipelineShader(hs); + releasePipelineShader(ds); + releasePipelineShader(gs); + releasePipelineShader(fs); - if (fs.shader) { - fs.shader->Release(); - fs.shader = nullptr; - } - fs.nativeResourceBindingMap.clear(); - - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } static inline D3D11_CULL_MODE toD3DCullMode(QRhiGraphicsPipeline::CullMode c) @@ -3687,6 +4141,19 @@ static inline D3D11_CULL_MODE toD3DCullMode(QRhiGraphicsPipeline::CullMode c) } } +static inline D3D11_FILL_MODE toD3DFillMode(QRhiGraphicsPipeline::PolygonMode mode) +{ + switch (mode) { + case QRhiGraphicsPipeline::Fill: + return D3D11_FILL_SOLID; + case QRhiGraphicsPipeline::Line: + return D3D11_FILL_WIREFRAME; + default: + Q_UNREACHABLE(); + return D3D11_FILL_SOLID; + } +} + static inline D3D11_COMPARISON_FUNC toD3DCompareOp(QRhiGraphicsPipeline::CompareOp op) { switch (op) { @@ -3770,13 +4237,37 @@ static inline DXGI_FORMAT toD3DAttributeFormat(QRhiVertexInputAttribute::Format return DXGI_FORMAT_R32G32_SINT; case QRhiVertexInputAttribute::SInt: return DXGI_FORMAT_R32_SINT; + case QRhiVertexInputAttribute::Half4: + // Note: D3D does not support half3. Pass through half3 as half4. + case QRhiVertexInputAttribute::Half3: + return DXGI_FORMAT_R16G16B16A16_FLOAT; + case QRhiVertexInputAttribute::Half2: + return DXGI_FORMAT_R16G16_FLOAT; + case QRhiVertexInputAttribute::Half: + return DXGI_FORMAT_R16_FLOAT; + case QRhiVertexInputAttribute::UShort4: + // Note: D3D does not support UShort3. Pass through UShort3 as UShort4. + case QRhiVertexInputAttribute::UShort3: + return DXGI_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort2: + return DXGI_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return DXGI_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + // Note: D3D does not support SShort3. Pass through SShort3 as SShort4. + case QRhiVertexInputAttribute::SShort3: + return DXGI_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort2: + return DXGI_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return DXGI_FORMAT_R16_SINT; default: Q_UNREACHABLE(); return DXGI_FORMAT_R32G32B32A32_FLOAT; } } -static inline D3D11_PRIMITIVE_TOPOLOGY toD3DTopology(QRhiGraphicsPipeline::Topology t) +static inline D3D11_PRIMITIVE_TOPOLOGY toD3DTopology(QRhiGraphicsPipeline::Topology t, int patchControlPointCount) { switch (t) { case QRhiGraphicsPipeline::Triangles: @@ -3789,6 +4280,9 @@ static inline D3D11_PRIMITIVE_TOPOLOGY toD3DTopology(QRhiGraphicsPipeline::Topol return D3D11_PRIMITIVE_TOPOLOGY_LINESTRIP; case QRhiGraphicsPipeline::Points: return D3D11_PRIMITIVE_TOPOLOGY_POINTLIST; + case QRhiGraphicsPipeline::Patches: + Q_ASSERT(patchControlPointCount >= 1 && patchControlPointCount <= 32); + return D3D11_PRIMITIVE_TOPOLOGY(D3D11_PRIMITIVE_TOPOLOGY_1_CONTROL_POINT_PATCHLIST + (patchControlPointCount - 1)); default: Q_UNREACHABLE(); return D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; @@ -3879,20 +4373,16 @@ static inline D3D11_BLEND_OP toD3DBlendOp(QRhiGraphicsPipeline::BlendOp op) } } -static pD3DCompile resolveD3DCompile() +static inline QByteArray sourceHash(const QByteArray &source) { - for (const wchar_t *libraryName : {L"D3DCompiler_47", L"D3DCompiler_43"}) { - QSystemLibrary library(libraryName); - if (library.load()) { - if (auto symbol = library.resolve("D3DCompile")) - return reinterpret_cast<pD3DCompile>(symbol); - } - } - return nullptr; + // taken from the GL backend, use the same mechanism to get a key + QCryptographicHash keyBuilder(QCryptographicHash::Sha1); + keyBuilder.addData(source); + return keyBuilder.result().toHex(); } -static QByteArray compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, UINT flags, - QString *error, QShaderKey *usedShaderKey) +QByteArray QRhiD3D11::compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, uint flags, + QString *error, QShaderKey *usedShaderKey) { QShaderKey key = { QShader::DxbcShader, 50, shaderVariant }; QShaderCode dxbc = shader.shader(key); @@ -3909,6 +4399,9 @@ static QByteArray compileHlslShaderSource(const QShader &shader, QShader::Varian return QByteArray(); } + if (usedShaderKey) + *usedShaderKey = key; + const char *target; switch (shader.stage()) { case QShader::VertexStage: @@ -3934,7 +4427,18 @@ static QByteArray compileHlslShaderSource(const QShader &shader, QShader::Varian return QByteArray(); } - static const pD3DCompile d3dCompile = resolveD3DCompile(); + BytecodeCacheKey cacheKey; + if (rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) { + cacheKey.sourceHash = sourceHash(hlslSource.shader()); + cacheKey.target = target; + cacheKey.entryPoint = hlslSource.entryPoint(); + cacheKey.compileFlags = flags; + auto cacheIt = m_bytecodeCache.constFind(cacheKey); + if (cacheIt != m_bytecodeCache.constEnd()) + return cacheIt.value(); + } + + static const pD3DCompile d3dCompile = QRhiD3D::resolveD3DCompile(); if (d3dCompile == nullptr) { qWarning("Unable to resolve function D3DCompile()"); return QByteArray(); @@ -3955,13 +4459,14 @@ static QByteArray compileHlslShaderSource(const QShader &shader, QShader::Varian return QByteArray(); } - if (usedShaderKey) - *usedShaderKey = key; - QByteArray result; result.resize(int(bytecode->GetBufferSize())); memcpy(result.data(), bytecode->GetBufferPointer(), size_t(result.size())); bytecode->Release(); + + if (rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + m_bytecodeCache.insert(cacheKey, result); + return result; } @@ -3971,27 +4476,27 @@ bool QD3D11GraphicsPipeline::create() destroy(); QRHI_RES_RHI(QRhiD3D11); + rhiD->pipelineCreationStart(); if (!rhiD->sanityCheckGraphicsPipeline(this)) return false; - D3D11_RASTERIZER_DESC rastDesc; - memset(&rastDesc, 0, sizeof(rastDesc)); - rastDesc.FillMode = D3D11_FILL_SOLID; + D3D11_RASTERIZER_DESC rastDesc = {}; + rastDesc.FillMode = toD3DFillMode(m_polygonMode); rastDesc.CullMode = toD3DCullMode(m_cullMode); rastDesc.FrontCounterClockwise = m_frontFace == CCW; rastDesc.DepthBias = m_depthBias; rastDesc.SlopeScaledDepthBias = m_slopeScaledDepthBias; rastDesc.DepthClipEnable = true; rastDesc.ScissorEnable = m_flags.testFlag(UsesScissor); - rastDesc.MultisampleEnable = rhiD->effectiveSampleCount(m_sampleCount).Count > 1; + rastDesc.MultisampleEnable = rhiD->effectiveSampleDesc(m_sampleCount).Count > 1; HRESULT hr = rhiD->dev->CreateRasterizerState(&rastDesc, &rastState); if (FAILED(hr)) { - qWarning("Failed to create rasterizer state: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create rasterizer state: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_DEPTH_STENCIL_DESC dsDesc; - memset(&dsDesc, 0, sizeof(dsDesc)); + D3D11_DEPTH_STENCIL_DESC dsDesc = {}; dsDesc.DepthEnable = m_depthTest; dsDesc.DepthWriteMask = m_depthWrite ? D3D11_DEPTH_WRITE_MASK_ALL : D3D11_DEPTH_WRITE_MASK_ZERO; dsDesc.DepthFunc = toD3DCompareOp(m_depthOp); @@ -4010,17 +4515,16 @@ bool QD3D11GraphicsPipeline::create() } hr = rhiD->dev->CreateDepthStencilState(&dsDesc, &dsState); if (FAILED(hr)) { - qWarning("Failed to create depth-stencil state: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create depth-stencil state: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_BLEND_DESC blendDesc; - memset(&blendDesc, 0, sizeof(blendDesc)); + D3D11_BLEND_DESC blendDesc = {}; blendDesc.IndependentBlendEnable = m_targetBlends.count() > 1; for (int i = 0, ie = m_targetBlends.count(); i != ie; ++i) { const QRhiGraphicsPipeline::TargetBlend &b(m_targetBlends[i]); - D3D11_RENDER_TARGET_BLEND_DESC blend; - memset(&blend, 0, sizeof(blend)); + D3D11_RENDER_TARGET_BLEND_DESC blend = {}; blend.BlendEnable = b.enable; blend.SrcBlend = toD3DBlendFactor(b.srcColor, true); blend.DestBlend = toD3DBlendFactor(b.dstColor, true); @@ -4032,19 +4536,19 @@ bool QD3D11GraphicsPipeline::create() blendDesc.RenderTarget[i] = blend; } if (m_targetBlends.isEmpty()) { - D3D11_RENDER_TARGET_BLEND_DESC blend; - memset(&blend, 0, sizeof(blend)); + D3D11_RENDER_TARGET_BLEND_DESC blend = {}; blend.RenderTargetWriteMask = D3D11_COLOR_WRITE_ENABLE_ALL; blendDesc.RenderTarget[0] = blend; } hr = rhiD->dev->CreateBlendState(&blendDesc, &blendState); if (FAILED(hr)) { - qWarning("Failed to create blend state: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create blend state: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } QByteArray vsByteCode; - for (const QRhiShaderStage &shaderStage : qAsConst(m_shaderStages)) { + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { auto cacheIt = rhiD->m_shaderCache.constFind(shaderStage); if (cacheIt != rhiD->m_shaderCache.constEnd()) { switch (shaderStage.type()) { @@ -4054,6 +4558,21 @@ bool QD3D11GraphicsPipeline::create() vsByteCode = cacheIt->bytecode; vs.nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; break; + case QRhiShaderStage::TessellationControl: + hs.shader = static_cast<ID3D11HullShader *>(cacheIt->s); + hs.shader->AddRef(); + hs.nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; + break; + case QRhiShaderStage::TessellationEvaluation: + ds.shader = static_cast<ID3D11DomainShader *>(cacheIt->s); + ds.shader->AddRef(); + ds.nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; + break; + case QRhiShaderStage::Geometry: + gs.shader = static_cast<ID3D11GeometryShader *>(cacheIt->s); + gs.shader->AddRef(); + gs.nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; + break; case QRhiShaderStage::Fragment: fs.shader = static_cast<ID3D11PixelShader *>(cacheIt->s); fs.shader->AddRef(); @@ -4069,8 +4588,8 @@ bool QD3D11GraphicsPipeline::create() if (m_flags.testFlag(CompileShadersWithDebugInfo)) compileFlags |= D3DCOMPILE_DEBUG; - const QByteArray bytecode = compileHlslShaderSource(shaderStage.shader(), shaderStage.shaderVariant(), compileFlags, - &error, &shaderKey); + const QByteArray bytecode = rhiD->compileHlslShaderSource(shaderStage.shader(), shaderStage.shaderVariant(), compileFlags, + &error, &shaderKey); if (bytecode.isEmpty()) { qWarning("HLSL shader compilation failed: %s", qPrintable(error)); return false; @@ -4085,23 +4604,56 @@ bool QD3D11GraphicsPipeline::create() case QRhiShaderStage::Vertex: hr = rhiD->dev->CreateVertexShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &vs.shader); if (FAILED(hr)) { - qWarning("Failed to create vertex shader: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create vertex shader: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } vsByteCode = bytecode; - if (const QShader::NativeResourceBindingMap *map = shaderStage.shader().nativeResourceBindingMap(shaderKey)) - vs.nativeResourceBindingMap = *map; + vs.nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); rhiD->m_shaderCache.insert(shaderStage, QRhiD3D11::Shader(vs.shader, bytecode, vs.nativeResourceBindingMap)); vs.shader->AddRef(); break; + case QRhiShaderStage::TessellationControl: + hr = rhiD->dev->CreateHullShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &hs.shader); + if (FAILED(hr)) { + qWarning("Failed to create hull shader: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + hs.nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); + rhiD->m_shaderCache.insert(shaderStage, QRhiD3D11::Shader(hs.shader, bytecode, hs.nativeResourceBindingMap)); + hs.shader->AddRef(); + break; + case QRhiShaderStage::TessellationEvaluation: + hr = rhiD->dev->CreateDomainShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &ds.shader); + if (FAILED(hr)) { + qWarning("Failed to create domain shader: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + ds.nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); + rhiD->m_shaderCache.insert(shaderStage, QRhiD3D11::Shader(ds.shader, bytecode, ds.nativeResourceBindingMap)); + ds.shader->AddRef(); + break; + case QRhiShaderStage::Geometry: + hr = rhiD->dev->CreateGeometryShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &gs.shader); + if (FAILED(hr)) { + qWarning("Failed to create geometry shader: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + gs.nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); + rhiD->m_shaderCache.insert(shaderStage, QRhiD3D11::Shader(gs.shader, bytecode, gs.nativeResourceBindingMap)); + gs.shader->AddRef(); + break; case QRhiShaderStage::Fragment: hr = rhiD->dev->CreatePixelShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &fs.shader); if (FAILED(hr)) { - qWarning("Failed to create pixel shader: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create pixel shader: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - if (const QShader::NativeResourceBindingMap *map = shaderStage.shader().nativeResourceBindingMap(shaderKey)) - fs.nativeResourceBindingMap = *map; + fs.nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); rhiD->m_shaderCache.insert(shaderStage, QRhiD3D11::Shader(fs.shader, bytecode, fs.nativeResourceBindingMap)); fs.shader->AddRef(); break; @@ -4111,7 +4663,7 @@ bool QD3D11GraphicsPipeline::create() } } - d3dTopology = toD3DTopology(m_topology); + d3dTopology = toD3DTopology(m_topology, m_patchControlPointCount); if (!vsByteCode.isEmpty()) { QByteArrayList matrixSliceSemantics; @@ -4119,8 +4671,7 @@ bool QD3D11GraphicsPipeline::create() for (auto it = m_vertexInputLayout.cbeginAttributes(), itEnd = m_vertexInputLayout.cendAttributes(); it != itEnd; ++it) { - D3D11_INPUT_ELEMENT_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_INPUT_ELEMENT_DESC desc = {}; // The output from SPIRV-Cross uses TEXCOORD<location> as the // semantic, except for matrices that are unrolled into consecutive // vec2/3/4s attributes and need TEXCOORD<location>_ as @@ -4143,7 +4694,7 @@ bool QD3D11GraphicsPipeline::create() const QRhiVertexInputBinding *inputBinding = m_vertexInputLayout.bindingAt(it->binding()); if (inputBinding->classification() == QRhiVertexInputBinding::PerInstance) { desc.InputSlotClass = D3D11_INPUT_PER_INSTANCE_DATA; - desc.InstanceDataStepRate = UINT(inputBinding->instanceStepRate()); + desc.InstanceDataStepRate = inputBinding->instanceStepRate(); } else { desc.InputSlotClass = D3D11_INPUT_PER_VERTEX_DATA; } @@ -4153,12 +4704,14 @@ bool QD3D11GraphicsPipeline::create() hr = rhiD->dev->CreateInputLayout(inputDescs.constData(), UINT(inputDescs.count()), vsByteCode, SIZE_T(vsByteCode.size()), &inputLayout); if (FAILED(hr)) { - qWarning("Failed to create input layout: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create input layout: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } } // else leave inputLayout set to nullptr; that's valid and it avoids a debug layer warning about an input layout with 0 elements } + rhiD->pipelineCreationEnd(); generation += 1; rhiD->registerResource(this); return true; @@ -4176,8 +4729,6 @@ QD3D11ComputePipeline::~QD3D11ComputePipeline() void QD3D11ComputePipeline::destroy() { - QRHI_RES_RHI(QRhiD3D11); - if (!cs.shader) return; @@ -4185,7 +4736,9 @@ void QD3D11ComputePipeline::destroy() cs.shader = nullptr; cs.nativeResourceBindingMap.clear(); - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11ComputePipeline::create() @@ -4194,6 +4747,7 @@ bool QD3D11ComputePipeline::create() destroy(); QRHI_RES_RHI(QRhiD3D11); + rhiD->pipelineCreationStart(); auto cacheIt = rhiD->m_shaderCache.constFind(m_shaderStage); if (cacheIt != rhiD->m_shaderCache.constEnd()) { @@ -4206,8 +4760,8 @@ bool QD3D11ComputePipeline::create() if (m_flags.testFlag(CompileShadersWithDebugInfo)) compileFlags |= D3DCOMPILE_DEBUG; - const QByteArray bytecode = compileHlslShaderSource(m_shaderStage.shader(), m_shaderStage.shaderVariant(), compileFlags, - &error, &shaderKey); + const QByteArray bytecode = rhiD->compileHlslShaderSource(m_shaderStage.shader(), m_shaderStage.shaderVariant(), compileFlags, + &error, &shaderKey); if (bytecode.isEmpty()) { qWarning("HLSL compute shader compilation failed: %s", qPrintable(error)); return false; @@ -4215,12 +4769,12 @@ bool QD3D11ComputePipeline::create() HRESULT hr = rhiD->dev->CreateComputeShader(bytecode.constData(), SIZE_T(bytecode.size()), nullptr, &cs.shader); if (FAILED(hr)) { - qWarning("Failed to create compute shader: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create compute shader: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - if (const QShader::NativeResourceBindingMap *map = m_shaderStage.shader().nativeResourceBindingMap(shaderKey)) - cs.nativeResourceBindingMap = *map; + cs.nativeResourceBindingMap = m_shaderStage.shader().nativeResourceBindingMap(shaderKey); if (rhiD->m_shaderCache.count() >= QRhiD3D11::MAX_SHADER_CACHE_ENTRIES) rhiD->clearShaderCache(); @@ -4230,6 +4784,7 @@ bool QD3D11ComputePipeline::create() cs.shader->AddRef(); + rhiD->pipelineCreationEnd(); generation += 1; rhiD->registerResource(this); return true; @@ -4251,20 +4806,93 @@ void QD3D11CommandBuffer::destroy() // nothing to do here } +bool QD3D11SwapChainTimestamps::prepare(QRhiD3D11 *rhiD) +{ + // Creates the query objects if not yet done, but otherwise calling this + // function is expected to be a no-op. + + D3D11_QUERY_DESC queryDesc = {}; + for (int i = 0; i < TIMESTAMP_PAIRS; ++i) { + if (!disjointQuery[i]) { + queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; + HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, &disjointQuery[i]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp disjoint query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + queryDesc.Query = D3D11_QUERY_TIMESTAMP; + for (int j = 0; j < 2; ++j) { + const int idx = 2 * i + j; + if (!query[idx]) { + HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, &query[idx]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + } + } + return true; +} + +void QD3D11SwapChainTimestamps::destroy() +{ + for (int i = 0; i < TIMESTAMP_PAIRS; ++i) { + active[i] = false; + if (disjointQuery[i]) { + disjointQuery[i]->Release(); + disjointQuery[i] = nullptr; + } + for (int j = 0; j < 2; ++j) { + const int idx = TIMESTAMP_PAIRS * i + j; + if (query[idx]) { + query[idx]->Release(); + query[idx] = nullptr; + } + } + } +} + +bool QD3D11SwapChainTimestamps::tryQueryTimestamps(int pairIndex, ID3D11DeviceContext *context, double *elapsedSec) +{ + bool result = false; + if (!active[pairIndex]) + return result; + + ID3D11Query *tsDisjoint = disjointQuery[pairIndex]; + ID3D11Query *tsStart = query[pairIndex * 2]; + ID3D11Query *tsEnd = query[pairIndex * 2 + 1]; + quint64 timestamps[2]; + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; + + bool ok = true; + ok &= context->GetData(tsDisjoint, &dj, sizeof(dj), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + ok &= context->GetData(tsEnd, ×tamps[1], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + ok &= context->GetData(tsStart, ×tamps[0], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + + if (ok) { + if (!dj.Disjoint && dj.Frequency) { + const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; + *elapsedSec = elapsedMs / 1000.0; + result = true; + } + active[pairIndex] = false; + } // else leave active set, will retry in a subsequent beginFrame + + return result; +} + QD3D11SwapChain::QD3D11SwapChain(QRhiImplementation *rhi) - : QRhiSwapChain(rhi), - rt(rhi), - cb(rhi) + : QRhiSwapChain(rhi), rt(rhi, this), rtRight(rhi, this), cb(rhi) { backBufferTex = nullptr; backBufferRtv = nullptr; for (int i = 0; i < BUFFER_COUNT; ++i) { msaaTex[i] = nullptr; msaaRtv[i] = nullptr; - timestampActive[i] = false; - timestampDisjointQuery[i] = nullptr; - timestampQuery[2 * i] = nullptr; - timestampQuery[2 * i + 1] = nullptr; } } @@ -4279,6 +4907,10 @@ void QD3D11SwapChain::releaseBuffers() backBufferRtv->Release(); backBufferRtv = nullptr; } + if (backBufferRtvRight) { + backBufferRtvRight->Release(); + backBufferRtvRight = nullptr; + } if (backBufferTex) { backBufferTex->Release(); backBufferTex = nullptr; @@ -4302,28 +4934,28 @@ void QD3D11SwapChain::destroy() releaseBuffers(); - for (int i = 0; i < BUFFER_COUNT; ++i) { - if (timestampDisjointQuery[i]) { - timestampDisjointQuery[i]->Release(); - timestampDisjointQuery[i] = nullptr; - } - for (int j = 0; j < 2; ++j) { - const int idx = BUFFER_COUNT * i + j; - if (timestampQuery[idx]) { - timestampQuery[idx]->Release(); - timestampQuery[idx] = nullptr; - } - } - } + timestamps.destroy(); swapChain->Release(); swapChain = nullptr; - QRHI_PROF; - QRHI_PROF_F(releaseSwapChain(this)); + if (dcompVisual) { + dcompVisual->Release(); + dcompVisual = nullptr; + } + + if (dcompTarget) { + dcompTarget->Release(); + dcompTarget = nullptr; + } QRHI_RES_RHI(QRhiD3D11); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->unregisterResource(this); + // See Deferred Destruction Issues with Flip Presentation Swap Chains in + // https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nf-d3d11-id3d11devicecontext-flush + rhiD->context->Flush(); + } } QRhiCommandBuffer *QD3D11SwapChain::currentFrameCommandBuffer() @@ -4336,22 +4968,67 @@ QRhiRenderTarget *QD3D11SwapChain::currentFrameRenderTarget() return &rt; } +QRhiRenderTarget *QD3D11SwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + return targetBuffer == StereoTargetBuffer::LeftBuffer? &rt: &rtRight; +} + QSize QD3D11SwapChain::surfacePixelSize() { Q_ASSERT(m_window); return m_window->size() * m_window->devicePixelRatio(); } +bool QD3D11SwapChain::isFormatSupported(Format f) +{ + if (f == SDR) + return true; + + if (!m_window) { + qWarning("Attempted to call isFormatSupported() without a window set"); + return false; + } + + QRHI_RES_RHI(QRhiD3D11); + DXGI_OUTPUT_DESC1 desc1; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) { + if (desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) + return f == QRhiSwapChain::HDRExtendedSrgbLinear || f == QRhiSwapChain::HDR10; + } + + return false; +} + +QRhiSwapChainHdrInfo QD3D11SwapChain::hdrInfo() +{ + QRhiSwapChainHdrInfo info = QRhiSwapChain::hdrInfo(); + // Must use m_window, not window, given this may be called before createOrResize(). + if (m_window) { + QRHI_RES_RHI(QRhiD3D11); + DXGI_OUTPUT_DESC1 hdrOutputDesc; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { + info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; + info.limits.luminanceInNits.minLuminance = hdrOutputDesc.MinLuminance; + info.limits.luminanceInNits.maxLuminance = hdrOutputDesc.MaxLuminance; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; // 1.0 = 80 nits + info.sdrWhiteLevel = QRhiD3D::sdrWhiteLevelInNits(hdrOutputDesc); + } + } + return info; +} + QRhiRenderPassDescriptor *QD3D11SwapChain::newCompatibleRenderPassDescriptor() { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } bool QD3D11SwapChain::newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, ID3D11Texture2D **tex, ID3D11RenderTargetView **rtv) const { - D3D11_TEXTURE2D_DESC desc; - memset(&desc, 0, sizeof(desc)); + D3D11_TEXTURE2D_DESC desc = {}; desc.Width = UINT(size.width()); desc.Height = UINT(size.height()); desc.MipLevels = 1; @@ -4364,17 +5041,18 @@ bool QD3D11SwapChain::newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI QRHI_RES_RHI(QRhiD3D11); HRESULT hr = rhiD->dev->CreateTexture2D(&desc, nullptr, tex); if (FAILED(hr)) { - qWarning("Failed to create color buffer texture: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create color buffer texture: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; - memset(&rtvDesc, 0, sizeof(rtvDesc)); + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc = {}; rtvDesc.Format = format; rtvDesc.ViewDimension = sampleDesc.Count > 1 ? D3D11_RTV_DIMENSION_TEXTURE2DMS : D3D11_RTV_DIMENSION_TEXTURE2D; hr = rhiD->dev->CreateRenderTargetView(*tex, &rtvDesc, rtv); if (FAILED(hr)) { - qWarning("Failed to create color buffer rtv: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create color buffer rtv: %s", + qPrintable(QSystemError::windowsComString(hr))); (*tex)->Release(); *tex = nullptr; return false; @@ -4383,6 +5061,19 @@ bool QD3D11SwapChain::newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI return true; } +bool QRhiD3D11::ensureDirectCompositionDevice() +{ + if (dcompDevice) + return true; + + qCDebug(QRHI_LOG_INFO, "Creating Direct Composition device (needed for semi-transparent windows)"); + dcompDevice = QRhiD3D::createDirectCompositionDevice(); + return dcompDevice ? true : false; +} + +static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; +static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; + bool QD3D11SwapChain::createOrResize() { // Can be called multiple times due to window resizes - that is not the @@ -4390,6 +5081,7 @@ bool QD3D11SwapChain::createOrResize() // resize the buffers then. const bool needsRegistration = !window || window != m_window; + const bool stereo = m_window->format().stereo(); // except if the window actually changes if (window && window != m_window) @@ -4402,101 +5094,183 @@ bool QD3D11SwapChain::createOrResize() if (pixelSize.isEmpty()) return false; - colorFormat = DXGI_FORMAT_R8G8B8A8_UNORM; - const DXGI_FORMAT srgbAdjustedFormat = m_flags.testFlag(sRGB) ? - DXGI_FORMAT_R8G8B8A8_UNORM_SRGB : DXGI_FORMAT_R8G8B8A8_UNORM; - - const UINT swapChainFlags = 0; + HWND hwnd = reinterpret_cast<HWND>(window->winId()); + HRESULT hr; QRHI_RES_RHI(QRhiD3D11); - bool useFlipDiscard = rhiD->hasDxgi2 && rhiD->supportsFlipDiscardSwapchain; + + if (m_flags.testFlag(SurfaceHasPreMulAlpha) || m_flags.testFlag(SurfaceHasNonPreMulAlpha)) { + if (!rhiD->useLegacySwapchainModel && rhiD->ensureDirectCompositionDevice()) { + if (!dcompTarget) { + hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, false, &dcompTarget); + if (FAILED(hr)) { + qWarning("Failed to create Direct Compsition target for the window: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } + if (dcompTarget && !dcompVisual) { + hr = rhiD->dcompDevice->CreateVisual(&dcompVisual); + if (FAILED(hr)) { + qWarning("Failed to create DirectComposition visual: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } + } + // simple consistency check + if (window->requestedFormat().alphaBufferSize() <= 0) + qWarning("Swapchain says surface has alpha but the window has no alphaBufferSize set. " + "This may lead to problems."); + } + + swapInterval = m_flags.testFlag(QRhiSwapChain::NoVSync) ? 0 : 1; + swapChainFlags = 0; + + // A non-flip swapchain can do Present(0) as expected without + // ALLOW_TEARING, and ALLOW_TEARING is not compatible with it at all so the + // flag must not be set then. Whereas for flip we should use it, if + // supported, to get better results for 'unthrottled' presentation. + if (swapInterval == 0 && rhiD->supportsAllowTearing) + swapChainFlags |= DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING; + if (!swapChain) { - HWND hwnd = reinterpret_cast<HWND>(window->winId()); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); - - // Take a shortcut for alpha: our QWindow is OpenGLSurface so whatever - // the platform plugin does to enable transparency for OpenGL window - // will be sufficient for us too on the legacy (DISCARD) path. For - // FLIP_DISCARD we'd need to use DirectComposition (create a - // IDCompositionDevice/Target/Visual), avoid that for now. - if (m_flags.testFlag(SurfaceHasPreMulAlpha) || m_flags.testFlag(SurfaceHasNonPreMulAlpha)) { - useFlipDiscard = false; - if (window->requestedFormat().alphaBufferSize() <= 0) - qWarning("Swapchain says surface has alpha but the window has no alphaBufferSize set. " - "This may lead to problems."); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); + colorFormat = DEFAULT_FORMAT; + srgbAdjustedColorFormat = m_flags.testFlag(sRGB) ? DEFAULT_SRGB_FORMAT : DEFAULT_FORMAT; + + DXGI_COLOR_SPACE_TYPE hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G22_NONE_P709; // SDR + DXGI_OUTPUT_DESC1 hdrOutputDesc; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { + // https://docs.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range + if (hdrOutputDesc.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) { + switch (m_format) { + case HDRExtendedSrgbLinear: + colorFormat = DXGI_FORMAT_R16G16B16A16_FLOAT; + hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G10_NONE_P709; + srgbAdjustedColorFormat = colorFormat; + break; + case HDR10: + colorFormat = DXGI_FORMAT_R10G10B10A2_UNORM; + hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020; + srgbAdjustedColorFormat = colorFormat; + break; + default: + break; + } + } else { + // This happens also when Use HDR is set to Off in the Windows + // Display settings. Show a helpful warning, but continue with the + // default non-HDR format. + qWarning("The output associated with the window is not HDR capable " + "(or Use HDR is Off in the Display Settings), ignoring HDR format request"); + } } - HRESULT hr; - if (useFlipDiscard) { - // We use FLIP_DISCARD which implies a buffer count of 2 (as opposed to the - // old DISCARD with back buffer count == 1). This makes no difference for - // the rest of the stuff except that automatic MSAA is unsupported and - // needs to be implemented via a custom multisample render target and an - // explicit resolve. - - DXGI_SWAP_CHAIN_DESC1 desc; - memset(&desc, 0, sizeof(desc)); - desc.Width = UINT(pixelSize.width()); - desc.Height = UINT(pixelSize.height()); - desc.Format = colorFormat; - desc.SampleDesc.Count = 1; - desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; - desc.BufferCount = BUFFER_COUNT; + // We use a FLIP model swapchain which implies a buffer count of 2 + // (as opposed to the old DISCARD with back buffer count == 1). + // This makes no difference for the rest of the stuff except that + // automatic MSAA is unsupported and needs to be implemented via a + // custom multisample render target and an explicit resolve. + + DXGI_SWAP_CHAIN_DESC1 desc = {}; + desc.Width = UINT(pixelSize.width()); + desc.Height = UINT(pixelSize.height()); + desc.Format = colorFormat; + desc.SampleDesc.Count = 1; + desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; + desc.BufferCount = BUFFER_COUNT; + desc.Flags = swapChainFlags; + desc.Scaling = rhiD->useLegacySwapchainModel ? DXGI_SCALING_STRETCH : DXGI_SCALING_NONE; + desc.SwapEffect = rhiD->useLegacySwapchainModel ? DXGI_SWAP_EFFECT_DISCARD : DXGI_SWAP_EFFECT_FLIP_DISCARD; + desc.Stereo = stereo; + + if (dcompVisual) { + // With DirectComposition setting AlphaMode to STRAIGHT fails the + // swapchain creation, whereas the result seems to be identical + // with any of the other values, including IGNORE. (?) + desc.AlphaMode = DXGI_ALPHA_MODE_PREMULTIPLIED; + + // DirectComposition has its own limitations, cannot use + // SCALING_NONE. So with semi-transparency requested we are forced + // to SCALING_STRETCH. desc.Scaling = DXGI_SCALING_STRETCH; - desc.SwapEffect = DXGI_SWAP_EFFECT(4); // DXGI_SWAP_EFFECT_FLIP_DISCARD - // Do not bother with AlphaMode, if won't work unless we go through - // DirectComposition. Instead, we just take the other (DISCARD) - // path for now when alpha is requested. - desc.Flags = swapChainFlags; - - IDXGISwapChain1 *sc1; - hr = static_cast<IDXGIFactory2 *>(rhiD->dxgiFactory)->CreateSwapChainForHwnd(rhiD->dev, hwnd, &desc, - nullptr, nullptr, &sc1); - if (SUCCEEDED(hr)) - swapChain = sc1; - } else { - // Windows 7 for instance. Use DISCARD mode. Regardless, keep on - // using our manual resolve for symmetry with the FLIP_DISCARD code - // path when MSAA is requested. - - DXGI_SWAP_CHAIN_DESC desc; - memset(&desc, 0, sizeof(desc)); - desc.BufferDesc.Width = UINT(pixelSize.width()); - desc.BufferDesc.Height = UINT(pixelSize.height()); - desc.BufferDesc.RefreshRate.Numerator = 60; - desc.BufferDesc.RefreshRate.Denominator = 1; - desc.BufferDesc.Format = colorFormat; - desc.SampleDesc.Count = 1; - desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; - desc.BufferCount = 1; - desc.OutputWindow = hwnd; - desc.Windowed = true; - desc.SwapEffect = DXGI_SWAP_EFFECT_DISCARD; - desc.Flags = swapChainFlags; - - hr = rhiD->dxgiFactory->CreateSwapChain(rhiD->dev, &desc, &swapChain); + } + + IDXGIFactory2 *fac = static_cast<IDXGIFactory2 *>(rhiD->dxgiFactory); + IDXGISwapChain1 *sc1; + + if (dcompVisual) + hr = fac->CreateSwapChainForComposition(rhiD->dev, &desc, nullptr, &sc1); + else + hr = fac->CreateSwapChainForHwnd(rhiD->dev, hwnd, &desc, nullptr, nullptr, &sc1); + + // If failed and we tried a HDR format, then try with SDR. This + // matches other backends, such as Vulkan where if the format is + // not supported, the default one is used instead. + if (FAILED(hr) && m_format != SDR) { + colorFormat = DEFAULT_FORMAT; + desc.Format = DEFAULT_FORMAT; + if (dcompVisual) + hr = fac->CreateSwapChainForComposition(rhiD->dev, &desc, nullptr, &sc1); + else + hr = fac->CreateSwapChainForHwnd(rhiD->dev, hwnd, &desc, nullptr, nullptr, &sc1); + } + + if (SUCCEEDED(hr)) { + swapChain = sc1; + if (m_format != SDR) { + IDXGISwapChain3 *sc3 = nullptr; + if (SUCCEEDED(sc1->QueryInterface(__uuidof(IDXGISwapChain3), reinterpret_cast<void **>(&sc3)))) { + hr = sc3->SetColorSpace1(hdrColorSpace); + if (FAILED(hr)) + qWarning("Failed to set color space on swapchain: %s", + qPrintable(QSystemError::windowsComString(hr))); + sc3->Release(); + } else { + qWarning("IDXGISwapChain3 not available, HDR swapchain will not work as expected"); + } + } + if (dcompVisual) { + hr = dcompVisual->SetContent(sc1); + if (SUCCEEDED(hr)) { + hr = dcompTarget->SetRoot(dcompVisual); + if (FAILED(hr)) { + qWarning("Failed to associate Direct Composition visual with the target: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } else { + qWarning("Failed to set content for Direct Composition visual: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } else { + // disable Alt+Enter; not relevant when using DirectComposition + rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); + } } if (FAILED(hr)) { - qWarning("Failed to create D3D11 swapchain: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create D3D11 swapchain: %s" + " (Width=%u Height=%u Format=%u SampleCount=%u BufferCount=%u Scaling=%u SwapEffect=%u Stereo=%u)", + qPrintable(QSystemError::windowsComString(hr)), + desc.Width, desc.Height, UINT(desc.Format), desc.SampleDesc.Count, + desc.BufferCount, UINT(desc.Scaling), UINT(desc.SwapEffect), UINT(desc.Stereo)); return false; } - rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); } else { releaseBuffers(); - const UINT count = useFlipDiscard ? BUFFER_COUNT : 1; - HRESULT hr = swapChain->ResizeBuffers(count, UINT(pixelSize.width()), UINT(pixelSize.height()), - colorFormat, swapChainFlags); + // flip model -> buffer count is the real buffer count, not 1 like with the legacy modes + hr = swapChain->ResizeBuffers(UINT(BUFFER_COUNT), UINT(pixelSize.width()), UINT(pixelSize.height()), + colorFormat, swapChainFlags); if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { qWarning("Device loss detected in ResizeBuffers()"); rhiD->deviceLost = true; return false; } else if (FAILED(hr)) { - qWarning("Failed to resize D3D11 swapchain: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to resize D3D11 swapchain: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } } - // This looks odd (for FLIP_DISCARD, esp. compared with backends for Vulkan + // This looks odd (for FLIP_*, esp. compared with backends for Vulkan // & co.) but the backbuffer is always at index 0, with magic underneath. // Some explanation from // https://docs.microsoft.com/en-us/windows/win32/direct3ddxgi/dxgi-1-4-improvements @@ -4510,25 +5284,39 @@ bool QD3D11SwapChain::createOrResize() // swapchain." // So just query index 0 once (per resize) and be done with it. - HRESULT hr = swapChain->GetBuffer(0, IID_ID3D11Texture2D, reinterpret_cast<void **>(&backBufferTex)); + hr = swapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), reinterpret_cast<void **>(&backBufferTex)); if (FAILED(hr)) { - qWarning("Failed to query swapchain backbuffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to query swapchain backbuffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } - D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; - memset(&rtvDesc, 0, sizeof(rtvDesc)); - rtvDesc.Format = srgbAdjustedFormat; + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = srgbAdjustedColorFormat; rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; hr = rhiD->dev->CreateRenderTargetView(backBufferTex, &rtvDesc, &backBufferRtv); if (FAILED(hr)) { - qWarning("Failed to create rtv for swapchain backbuffer: %s", qPrintable(comErrorMessage(hr))); + qWarning("Failed to create rtv for swapchain backbuffer: %s", + qPrintable(QSystemError::windowsComString(hr))); return false; } + if (stereo) { + // Create a second render target view for the right eye + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.FirstArraySlice = 1; + rtvDesc.Texture2DArray.ArraySize = 1; + hr = rhiD->dev->CreateRenderTargetView(backBufferTex, &rtvDesc, &backBufferRtvRight); + if (FAILED(hr)) { + qWarning("Failed to create rtv for swapchain backbuffer (right eye): %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + // Try to reduce stalls by having a dedicated MSAA texture per swapchain buffer. for (int i = 0; i < BUFFER_COUNT; ++i) { if (sampleDesc.Count > 1) { - if (!newColorBuffer(pixelSize, srgbAdjustedFormat, sampleDesc, &msaaTex[i], &msaaRtv[i])) + if (!newColorBuffer(pixelSize, srgbAdjustedColorFormat, sampleDesc, &msaaTex[i], &msaaRtv[i])) return false; } } @@ -4553,9 +5341,9 @@ bool QD3D11SwapChain::createOrResize() currentFrameSlot = 0; frameCount = 0; ds = m_depthStencil ? QRHI_RES(QD3D11RenderBuffer, m_depthStencil) : nullptr; - swapInterval = m_flags.testFlag(QRhiSwapChain::NoVSync) ? 0 : 1; - QD3D11ReferenceRenderTarget *rtD = QRHI_RES(QD3D11ReferenceRenderTarget, &rt); + rt.setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget + QD3D11SwapChainRenderTarget *rtD = QRHI_RES(QD3D11SwapChainRenderTarget, &rt); rtD->d.rp = QRHI_RES(QD3D11RenderPassDescriptor, m_renderPassDesc); rtD->d.pixelSize = pixelSize; rtD->d.dpr = float(window->devicePixelRatio()); @@ -4563,32 +5351,20 @@ bool QD3D11SwapChain::createOrResize() rtD->d.colorAttCount = 1; rtD->d.dsAttCount = m_depthStencil ? 1 : 0; - QRHI_PROF; - QRHI_PROF_F(resizeSwapChain(this, BUFFER_COUNT, sampleDesc.Count > 1 ? BUFFER_COUNT : 0, int(sampleDesc.Count))); - if (rhiP) { - D3D11_QUERY_DESC queryDesc; - memset(&queryDesc, 0, sizeof(queryDesc)); - for (int i = 0; i < BUFFER_COUNT; ++i) { - if (!timestampDisjointQuery[i]) { - queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; - HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, ×tampDisjointQuery[i]); - if (FAILED(hr)) { - qWarning("Failed to create timestamp disjoint query: %s", qPrintable(comErrorMessage(hr))); - break; - } - } - queryDesc.Query = D3D11_QUERY_TIMESTAMP; - for (int j = 0; j < 2; ++j) { - const int idx = BUFFER_COUNT * i + j; // one pair per buffer (frame) - if (!timestampQuery[idx]) { - HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, ×tampQuery[idx]); - if (FAILED(hr)) { - qWarning("Failed to create timestamp query: %s", qPrintable(comErrorMessage(hr))); - break; - } - } - } - } + if (stereo) { + rtD = QRHI_RES(QD3D11SwapChainRenderTarget, &rtRight); + rtD->d.rp = QRHI_RES(QD3D11RenderPassDescriptor, m_renderPassDesc); + rtD->d.pixelSize = pixelSize; + rtD->d.dpr = float(window->devicePixelRatio()); + rtD->d.sampleCount = int(sampleDesc.Count); + rtD->d.colorAttCount = 1; + rtD->d.dsAttCount = m_depthStencil ? 1 : 0; + rtD->d.rtv[0] = backBufferRtvRight; + rtD->d.dsv = ds ? ds->dsv : nullptr; + } + + if (rhiD->rhiFlags.testFlag(QRhi::EnableTimestamps)) { + timestamps.prepare(rhiD); // timestamp queries are optional so we can go on even if they failed } @@ -4598,34 +5374,4 @@ bool QD3D11SwapChain::createOrResize() return true; } -void QRhiD3D11::DeviceCurse::initResources() -{ - framesLeft = framesToActivate; - - HRESULT hr = q->dev->CreateComputeShader(g_killDeviceByTimingOut, sizeof(g_killDeviceByTimingOut), nullptr, &cs); - if (FAILED(hr)) { - qWarning("Failed to create compute shader: %s", qPrintable(comErrorMessage(hr))); - return; - } -} - -void QRhiD3D11::DeviceCurse::releaseResources() -{ - if (cs) { - cs->Release(); - cs = nullptr; - } -} - -void QRhiD3D11::DeviceCurse::activate() -{ - if (!cs) - return; - - qDebug("Activating Curse. Goodbye Cruel World."); - - q->context->CSSetShader(cs, nullptr, 0); - q->context->Dispatch(256, 1, 1); -} - QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhid3d11_p.h b/src/gui/rhi/qrhid3d11_p.h index ed26e9af16..7644748407 100644 --- a/src/gui/rhi/qrhid3d11_p.h +++ b/src/gui/rhi/qrhid3d11_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHID3D11_H -#define QRHID3D11_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHID3D11_P_H +#define QRHID3D11_P_H // // W A R N I N G @@ -51,31 +15,852 @@ // We mean it. // -#include <private/qrhi_p.h> +#include "qrhi_p.h" +#include <rhi/qshaderdescription.h> +#include <QWindow> -// no d3d includes here, to prevent precompiled header mess due to COM +#include <d3d11_1.h> +#include <dxgi1_6.h> +#include <dcomp.h> QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiD3D11InitParams : public QRhiInitParams +class QRhiD3D11; + +struct QD3D11Buffer : public QRhiBuffer +{ + QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QD3D11Buffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + ID3D11UnorderedAccessView *unorderedAccessView(quint32 offset); + + ID3D11Buffer *buffer = nullptr; + char *dynBuf = nullptr; + bool hasPendingDynamicUpdates = false; + QHash<quint32, ID3D11UnorderedAccessView *> uavs; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11RenderBuffer : public QRhiRenderBuffer +{ + QD3D11RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QD3D11RenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + ID3D11Texture2D *tex = nullptr; + ID3D11DepthStencilView *dsv = nullptr; + ID3D11RenderTargetView *rtv = nullptr; + DXGI_FORMAT dxgiFormat; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11Texture : public QRhiTexture +{ + QD3D11Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QD3D11Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + ID3D11UnorderedAccessView *unorderedAccessViewForLevel(int level); + ID3D11Resource *textureResource() const + { + if (tex) + return tex; + else if (tex1D) + return tex1D; + return tex3D; + } + + ID3D11Texture2D *tex = nullptr; + ID3D11Texture3D *tex3D = nullptr; + ID3D11Texture1D *tex1D = nullptr; + bool owns = true; + ID3D11ShaderResourceView *srv = nullptr; + DXGI_FORMAT dxgiFormat; + uint mipLevelCount = 0; + DXGI_SAMPLE_DESC sampleDesc; + ID3D11UnorderedAccessView *perLevelViews[QRhi::MAX_MIP_LEVELS]; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11Sampler : public QRhiSampler +{ + QD3D11Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QD3D11Sampler(); + void destroy() override; + bool create() override; + + ID3D11SamplerState *samplerState = nullptr; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QD3D11RenderPassDescriptor(QRhiImplementation *rhi); + ~QD3D11RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QD3D11RenderTargetData +{ + QD3D11RenderTargetData(QRhiImplementation *) + { + for (int i = 0; i < MAX_COLOR_ATTACHMENTS; ++i) + rtv[i] = nullptr; + } + + QD3D11RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + + static const int MAX_COLOR_ATTACHMENTS = 8; + ID3D11RenderTargetView *rtv[MAX_COLOR_ATTACHMENTS]; + ID3D11DepthStencilView *dsv = nullptr; + + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; +}; + +struct QD3D11SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QD3D11SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QD3D11SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QD3D11RenderTargetData d; +}; + +struct QD3D11TextureRenderTarget : public QRhiTextureRenderTarget +{ + QD3D11TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QD3D11TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QD3D11RenderTargetData d; + bool ownsRtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; + ID3D11RenderTargetView *rtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; + bool ownsDsv = false; + ID3D11DepthStencilView *dsv = nullptr; + friend class QRhiD3D11; +}; + +struct QD3D11ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QD3D11ShaderResourceBindings(QRhiImplementation *rhi); + ~QD3D11ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + bool hasDynamicOffset = false; + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + uint generation = 0; + + // Keep track of the generation number of each referenced QRhi* to be able + // to detect that the batched bindings are out of date. + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData; + + struct StageUniformBufferBatches { + bool present = false; + QRhiBatchedBindings<ID3D11Buffer *> ubufs; + QRhiBatchedBindings<UINT> ubuforigbindings; + QRhiBatchedBindings<UINT> ubufoffsets; + QRhiBatchedBindings<UINT> ubufsizes; + void finish() { + present = ubufs.finish(); + ubuforigbindings.finish(); + ubufoffsets.finish(); + ubufsizes.finish(); + } + void clear() { + ubufs.clear(); + ubuforigbindings.clear(); + ubufoffsets.clear(); + ubufsizes.clear(); + } + }; + + struct StageSamplerBatches { + bool present = false; + QRhiBatchedBindings<ID3D11SamplerState *> samplers; + QRhiBatchedBindings<ID3D11ShaderResourceView *> shaderresources; + void finish() { + present = samplers.finish(); + shaderresources.finish(); + } + void clear() { + samplers.clear(); + shaderresources.clear(); + } + }; + + struct StageUavBatches { + bool present = false; + QRhiBatchedBindings<ID3D11UnorderedAccessView *> uavs; + void finish() { + present = uavs.finish(); + } + void clear() { + uavs.clear(); + } + }; + + StageUniformBufferBatches vsUniformBufferBatches; + StageUniformBufferBatches hsUniformBufferBatches; + StageUniformBufferBatches dsUniformBufferBatches; + StageUniformBufferBatches gsUniformBufferBatches; + StageUniformBufferBatches fsUniformBufferBatches; + StageUniformBufferBatches csUniformBufferBatches; + + StageSamplerBatches vsSamplerBatches; + StageSamplerBatches hsSamplerBatches; + StageSamplerBatches dsSamplerBatches; + StageSamplerBatches gsSamplerBatches; + StageSamplerBatches fsSamplerBatches; + StageSamplerBatches csSamplerBatches; + + StageUavBatches csUavBatches; + + friend class QRhiD3D11; +}; + +Q_DECLARE_TYPEINFO(QD3D11ShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); + +struct QD3D11GraphicsPipeline : public QRhiGraphicsPipeline +{ + QD3D11GraphicsPipeline(QRhiImplementation *rhi); + ~QD3D11GraphicsPipeline(); + void destroy() override; + bool create() override; + + ID3D11DepthStencilState *dsState = nullptr; + ID3D11BlendState *blendState = nullptr; + struct { + ID3D11VertexShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } vs; + struct { + ID3D11HullShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } hs; + struct { + ID3D11DomainShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } ds; + struct { + ID3D11GeometryShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } gs; + struct { + ID3D11PixelShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } fs; + ID3D11InputLayout *inputLayout = nullptr; + D3D11_PRIMITIVE_TOPOLOGY d3dTopology = D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; + ID3D11RasterizerState *rastState = nullptr; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11ComputePipeline : public QRhiComputePipeline +{ + QD3D11ComputePipeline(QRhiImplementation *rhi); + ~QD3D11ComputePipeline(); + void destroy() override; + bool create() override; + + struct { + ID3D11ComputeShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } cs; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11SwapChain; + +struct QD3D11CommandBuffer : public QRhiCommandBuffer +{ + QD3D11CommandBuffer(QRhiImplementation *rhi); + ~QD3D11CommandBuffer(); + void destroy() override; + + // these must be kept at a reasonably low value otherwise sizeof Command explodes + static const int MAX_DYNAMIC_OFFSET_COUNT = 8; + static const int MAX_VERTEX_BUFFER_BINDING_COUNT = 8; + + struct Command { + enum Cmd { + BeginFrame, + EndFrame, + ResetShaderResources, + SetRenderTarget, + Clear, + Viewport, + Scissor, + BindVertexBuffers, + BindIndexBuffer, + BindGraphicsPipeline, + BindShaderResources, + StencilRef, + BlendConstants, + Draw, + DrawIndexed, + UpdateSubRes, + CopySubRes, + ResolveSubRes, + GenMip, + DebugMarkBegin, + DebugMarkEnd, + DebugMarkMsg, + BindComputePipeline, + Dispatch + }; + enum ClearFlag { Color = 1, Depth = 2, Stencil = 4 }; + Cmd cmd; + + // QRhi*/QD3D11* references should be kept at minimum (so no + // QRhiTexture/Buffer/etc. pointers). + union Args { + struct { + ID3D11Query *tsQuery; + ID3D11Query *tsDisjointQuery; + QD3D11RenderTargetData *swapchainData; + } beginFrame; + struct { + ID3D11Query *tsQuery; + ID3D11Query *tsDisjointQuery; + } endFrame; + struct { + QRhiRenderTarget *rt; + } setRenderTarget; + struct { + QRhiRenderTarget *rt; + int mask; + float c[4]; + float d; + quint32 s; + } clear; + struct { + float x, y, w, h; + float d0, d1; + } viewport; + struct { + int x, y, w, h; + } scissor; + struct { + int startSlot; + int slotCount; + ID3D11Buffer *buffers[MAX_VERTEX_BUFFER_BINDING_COUNT]; + UINT offsets[MAX_VERTEX_BUFFER_BINDING_COUNT]; + UINT strides[MAX_VERTEX_BUFFER_BINDING_COUNT]; + } bindVertexBuffers; + struct { + ID3D11Buffer *buffer; + quint32 offset; + DXGI_FORMAT format; + } bindIndexBuffer; + struct { + QD3D11GraphicsPipeline *ps; + } bindGraphicsPipeline; + struct { + QD3D11ShaderResourceBindings *srb; + bool offsetOnlyChange; + int dynamicOffsetCount; + uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offsetInConstants + } bindShaderResources; + struct { + QD3D11GraphicsPipeline *ps; + quint32 ref; + } stencilRef; + struct { + QD3D11GraphicsPipeline *ps; + float c[4]; + } blendConstants; + struct { + QD3D11GraphicsPipeline *ps; + quint32 vertexCount; + quint32 instanceCount; + quint32 firstVertex; + quint32 firstInstance; + } draw; + struct { + QD3D11GraphicsPipeline *ps; + quint32 indexCount; + quint32 instanceCount; + quint32 firstIndex; + qint32 vertexOffset; + quint32 firstInstance; + } drawIndexed; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + bool hasDstBox; + D3D11_BOX dstBox; + const void *src; // must come from retain*() + UINT srcRowPitch; + } updateSubRes; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + UINT dstX; + UINT dstY; + UINT dstZ; + ID3D11Resource *src; + UINT srcSubRes; + bool hasSrcBox; + D3D11_BOX srcBox; + } copySubRes; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + ID3D11Resource *src; + UINT srcSubRes; + DXGI_FORMAT format; + } resolveSubRes; + struct { + ID3D11ShaderResourceView *srv; + } genMip; + struct { + char s[64]; + } debugMark; + struct { + QD3D11ComputePipeline *ps; + } bindComputePipeline; + struct { + UINT x; + UINT y; + UINT z; + } dispatch; + } args; + }; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + QRhiBackendCommandList<Command> commands; + PassType recordingPass; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + ID3D11Buffer *currentIndexBuffer; + quint32 currentIndexOffset; + DXGI_FORMAT currentIndexFormat; + ID3D11Buffer *currentVertexBuffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + quint32 currentVertexOffsets[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + + QVarLengthArray<QByteArray, 4> dataRetainPool; + QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; + QVarLengthArray<QImage, 4> imageRetainPool; + + // relies heavily on implicit sharing (no copies of the actual data will be made) + const uchar *retainData(const QByteArray &data) { + dataRetainPool.append(data); + return reinterpret_cast<const uchar *>(dataRetainPool.last().constData()); + } + const uchar *retainBufferData(const QRhiBufferData &data) { + bufferDataRetainPool.append(data); + return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); + } + const uchar *retainImage(const QImage &image) { + imageRetainPool.append(image); + return imageRetainPool.last().constBits(); + } + void resetCommands() { + commands.reset(); + dataRetainPool.clear(); + bufferDataRetainPool.clear(); + imageRetainPool.clear(); + } + void resetState() { + recordingPass = NoPass; + // do not zero lastGpuTime + currentTarget = nullptr; + resetCommands(); + resetCachedState(); + } + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentIndexBuffer = nullptr; + currentIndexOffset = 0; + currentIndexFormat = DXGI_FORMAT_R16_UINT; + memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); + memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); + } +}; + +struct QD3D11SwapChainTimestamps +{ + static const int TIMESTAMP_PAIRS = 2; + + bool active[TIMESTAMP_PAIRS] = {}; + ID3D11Query *disjointQuery[TIMESTAMP_PAIRS] = {}; + ID3D11Query *query[TIMESTAMP_PAIRS * 2] = {}; + + bool prepare(QRhiD3D11 *rhiD); + void destroy(); + bool tryQueryTimestamps(int idx, ID3D11DeviceContext *context, double *elapsedSec); +}; + +struct QD3D11SwapChain : public QRhiSwapChain { - bool enableDebugLayer = false; + QD3D11SwapChain(QRhiImplementation *rhi); + ~QD3D11SwapChain(); + void destroy() override; - int framesUntilKillingDeviceViaTdr = -1; - bool repeatDeviceKill = false; + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + QRhiSwapChainHdrInfo hdrInfo() override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void releaseBuffers(); + bool newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, + ID3D11Texture2D **tex, ID3D11RenderTargetView **rtv) const; + + QWindow *window = nullptr; + QSize pixelSize; + QD3D11SwapChainRenderTarget rt; + QD3D11SwapChainRenderTarget rtRight; + QD3D11CommandBuffer cb; + DXGI_FORMAT colorFormat; + DXGI_FORMAT srgbAdjustedColorFormat; + IDXGISwapChain *swapChain = nullptr; + UINT swapChainFlags = 0; + ID3D11Texture2D *backBufferTex; + ID3D11RenderTargetView *backBufferRtv; + ID3D11RenderTargetView *backBufferRtvRight = nullptr; + static const int BUFFER_COUNT = 2; + ID3D11Texture2D *msaaTex[BUFFER_COUNT]; + ID3D11RenderTargetView *msaaRtv[BUFFER_COUNT]; + DXGI_SAMPLE_DESC sampleDesc; + int currentFrameSlot = 0; + int frameCount = 0; + QD3D11RenderBuffer *ds = nullptr; + UINT swapInterval = 1; + IDCompositionTarget *dcompTarget = nullptr; + IDCompositionVisual *dcompVisual = nullptr; + QD3D11SwapChainTimestamps timestamps; + int currentTimestampPairIndex = 0; }; -struct Q_GUI_EXPORT QRhiD3D11NativeHandles : public QRhiNativeHandles +class QRhiD3D11 : public QRhiImplementation { - // to import a device and a context - void *dev = nullptr; - void *context = nullptr; - // alternatively, to specify the device feature level and/or the adapter to use - int featureLevel = 0; - quint32 adapterLuidLow = 0; - qint32 adapterLuidHigh = 0; +public: + QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void enqueueSubresUpload(QD3D11Texture *texD, QD3D11CommandBuffer *cbD, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, + const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[]); + void executeBufferHostWrites(QD3D11Buffer *bufD); + void bindShaderResources(QD3D11ShaderResourceBindings *srbD, + const uint *dynOfsPairs, int dynOfsPairCount, + bool offsetOnlyChange); + void resetShaderResources(); + void executeCommandBuffer(QD3D11CommandBuffer *cbD); + DXGI_SAMPLE_DESC effectiveSampleDesc(int sampleCount) const; + void finishActiveReadbacks(); + void reportLiveObjects(ID3D11Device *device); + void clearShaderCache(); + QByteArray compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, uint flags, + QString *error, QShaderKey *usedShaderKey); + bool ensureDirectCompositionDevice(); + + QRhi::Flags rhiFlags; + bool debugLayer = false; + bool importedDeviceAndContext = false; + ID3D11Device *dev = nullptr; + ID3D11DeviceContext1 *context = nullptr; + D3D_FEATURE_LEVEL featureLevel = D3D_FEATURE_LEVEL(0); + LUID adapterLuid = {}; + ID3DUserDefinedAnnotation *annotations = nullptr; + IDXGIAdapter1 *activeAdapter = nullptr; + IDXGIFactory1 *dxgiFactory = nullptr; + IDCompositionDevice *dcompDevice = nullptr; + bool supportsAllowTearing = false; + bool useLegacySwapchainModel = false; + bool deviceLost = false; + QRhiD3D11NativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + + struct { + int vsHighestActiveVertexBufferBinding = -1; + bool vsHasIndexBufferBound = false; + int vsHighestActiveSrvBinding = -1; + int hsHighestActiveSrvBinding = -1; + int dsHighestActiveSrvBinding = -1; + int gsHighestActiveSrvBinding = -1; + int fsHighestActiveSrvBinding = -1; + int csHighestActiveSrvBinding = -1; + int csHighestActiveUavBinding = -1; + QD3D11SwapChain *currentSwapChain = nullptr; + } contextState; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } + bool active = false; + QD3D11CommandBuffer cbWrapper; + ID3D11Query *tsQueries[2] = {}; + ID3D11Query *tsDisjointQuery = nullptr; + } ofr; + + struct TextureReadback { + QRhiReadbackDescription desc; + QRhiReadbackResult *result; + ID3D11Texture2D *stagingTex; + quint32 byteSize; + quint32 bpl; + QSize pixelSize; + QRhiTexture::Format format; + }; + QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; + struct BufferReadback { + QRhiReadbackResult *result; + quint32 byteSize; + ID3D11Buffer *stagingBuf; + }; + QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; + + struct Shader { + Shader() = default; + Shader(IUnknown *s, const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) + : s(s), bytecode(bytecode), nativeResourceBindingMap(rbm) { } + IUnknown *s; + QByteArray bytecode; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + }; + QHash<QRhiShaderStage, Shader> m_shaderCache; + + // This is what gets exposed as the "pipeline cache", not that that concept + // applies anyway. Here we are just storing the DX bytecode for a shader so + // we can skip the HLSL->DXBC compilation when the QShader has HLSL source + // code and the same shader source has already been compiled before. + // m_shaderCache seemingly does the same, but this here does not care about + // the ID3D11*Shader, this is just about the bytecode and about allowing + // the data to be serialized to persistent storage and then reloaded in + // future runs of the app, or when creating another QRhi, etc. + struct BytecodeCacheKey { + QByteArray sourceHash; + QByteArray target; + QByteArray entryPoint; + uint compileFlags; + }; + QHash<BytecodeCacheKey, QByteArray> m_bytecodeCache; }; +Q_DECLARE_TYPEINFO(QRhiD3D11::TextureReadback, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiD3D11::BufferReadback, Q_RELOCATABLE_TYPE); + +inline bool operator==(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept +{ + return a.sourceHash == b.sourceHash + && a.target == b.target + && a.entryPoint == b.entryPoint + && a.compileFlags == b.compileFlags; +} + +inline bool operator!=(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept +{ + return !(a == b); +} + +inline size_t qHash(const QRhiD3D11::BytecodeCacheKey &k, size_t seed = 0) noexcept +{ + return qHash(k.sourceHash, seed) ^ qHash(k.target) ^ qHash(k.entryPoint) ^ k.compileFlags; +} + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhid3d11_p_p.h b/src/gui/rhi/qrhid3d11_p_p.h deleted file mode 100644 index 0695259612..0000000000 --- a/src/gui/rhi/qrhid3d11_p_p.h +++ /dev/null @@ -1,788 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHID3D11_P_H -#define QRHID3D11_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhid3d11_p.h" -#include "qrhi_p_p.h" -#include "qshaderdescription_p.h" -#include <QWindow> - -#include <d3d11_1.h> -#include <dxgi1_3.h> - -QT_BEGIN_NAMESPACE - -struct QD3D11Buffer : public QRhiBuffer -{ - QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size); - ~QD3D11Buffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - ID3D11UnorderedAccessView *unorderedAccessView(); - - ID3D11Buffer *buffer = nullptr; - char *dynBuf = nullptr; - bool hasPendingDynamicUpdates = false; - ID3D11UnorderedAccessView *uav = nullptr; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11RenderBuffer : public QRhiRenderBuffer -{ - QD3D11RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QD3D11RenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - ID3D11Texture2D *tex = nullptr; - ID3D11DepthStencilView *dsv = nullptr; - ID3D11RenderTargetView *rtv = nullptr; - DXGI_FORMAT dxgiFormat; - DXGI_SAMPLE_DESC sampleDesc; - friend class QRhiD3D11; -}; - -struct QD3D11Texture : public QRhiTexture -{ - QD3D11Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags); - ~QD3D11Texture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - bool finishCreate(); - ID3D11UnorderedAccessView *unorderedAccessViewForLevel(int level); - ID3D11Resource *textureResource() const - { - if (tex) - return tex; - return tex3D; - } - - ID3D11Texture2D *tex = nullptr; - ID3D11Texture3D *tex3D = nullptr; - bool owns = true; - ID3D11ShaderResourceView *srv = nullptr; - DXGI_FORMAT dxgiFormat; - uint mipLevelCount = 0; - DXGI_SAMPLE_DESC sampleDesc; - ID3D11UnorderedAccessView *perLevelViews[QRhi::MAX_MIP_LEVELS]; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11Sampler : public QRhiSampler -{ - QD3D11Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QD3D11Sampler(); - void destroy() override; - bool create() override; - - ID3D11SamplerState *samplerState = nullptr; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11RenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QD3D11RenderPassDescriptor(QRhiImplementation *rhi); - ~QD3D11RenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; -}; - -struct QD3D11RenderTargetData -{ - QD3D11RenderTargetData(QRhiImplementation *) - { - for (int i = 0; i < MAX_COLOR_ATTACHMENTS; ++i) - rtv[i] = nullptr; - } - - QD3D11RenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - - static const int MAX_COLOR_ATTACHMENTS = 8; - ID3D11RenderTargetView *rtv[MAX_COLOR_ATTACHMENTS]; - ID3D11DepthStencilView *dsv = nullptr; -}; - -struct QD3D11ReferenceRenderTarget : public QRhiRenderTarget -{ - QD3D11ReferenceRenderTarget(QRhiImplementation *rhi); - ~QD3D11ReferenceRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QD3D11RenderTargetData d; -}; - -struct QD3D11TextureRenderTarget : public QRhiTextureRenderTarget -{ - QD3D11TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QD3D11TextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QD3D11RenderTargetData d; - bool ownsRtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; - ID3D11RenderTargetView *rtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; - bool ownsDsv = false; - ID3D11DepthStencilView *dsv = nullptr; - friend class QRhiD3D11; -}; - -struct QD3D11ShaderResourceBindings : public QRhiShaderResourceBindings -{ - QD3D11ShaderResourceBindings(QRhiImplementation *rhi); - ~QD3D11ShaderResourceBindings(); - void destroy() override; - bool create() override; - - bool hasDynamicOffset = false; - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - uint generation = 0; - - // Keep track of the generation number of each referenced QRhi* to be able - // to detect that the batched bindings are out of date. - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData; - - bool vsubufsPresent = false; - bool fsubufsPresent = false; - bool csubufsPresent = false; - bool vssamplersPresent = false; - bool fssamplersPresent = false; - bool cssamplersPresent = false; - bool csUAVsPresent = false; - - QRhiBatchedBindings<ID3D11Buffer *> vsubufs; - QRhiBatchedBindings<UINT> vsubuforigbindings; - QRhiBatchedBindings<UINT> vsubufoffsets; - QRhiBatchedBindings<UINT> vsubufsizes; - - QRhiBatchedBindings<ID3D11Buffer *> fsubufs; - QRhiBatchedBindings<UINT> fsubuforigbindings; - QRhiBatchedBindings<UINT> fsubufoffsets; - QRhiBatchedBindings<UINT> fsubufsizes; - - QRhiBatchedBindings<ID3D11Buffer *> csubufs; - QRhiBatchedBindings<UINT> csubuforigbindings; - QRhiBatchedBindings<UINT> csubufoffsets; - QRhiBatchedBindings<UINT> csubufsizes; - - QRhiBatchedBindings<ID3D11SamplerState *> vssamplers; - QRhiBatchedBindings<ID3D11ShaderResourceView *> vsshaderresources; - - QRhiBatchedBindings<ID3D11SamplerState *> fssamplers; - QRhiBatchedBindings<ID3D11ShaderResourceView *> fsshaderresources; - - QRhiBatchedBindings<ID3D11SamplerState *> cssamplers; - QRhiBatchedBindings<ID3D11ShaderResourceView *> csshaderresources; - - QRhiBatchedBindings<ID3D11UnorderedAccessView *> csUAVs; - - friend class QRhiD3D11; -}; - -Q_DECLARE_TYPEINFO(QD3D11ShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); - -struct QD3D11GraphicsPipeline : public QRhiGraphicsPipeline -{ - QD3D11GraphicsPipeline(QRhiImplementation *rhi); - ~QD3D11GraphicsPipeline(); - void destroy() override; - bool create() override; - - ID3D11DepthStencilState *dsState = nullptr; - ID3D11BlendState *blendState = nullptr; - struct { - ID3D11VertexShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } vs; - struct { - ID3D11PixelShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } fs; - ID3D11InputLayout *inputLayout = nullptr; - D3D11_PRIMITIVE_TOPOLOGY d3dTopology = D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; - ID3D11RasterizerState *rastState = nullptr; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11ComputePipeline : public QRhiComputePipeline -{ - QD3D11ComputePipeline(QRhiImplementation *rhi); - ~QD3D11ComputePipeline(); - void destroy() override; - bool create() override; - - struct { - ID3D11ComputeShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } cs; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11SwapChain; - -struct QD3D11CommandBuffer : public QRhiCommandBuffer -{ - QD3D11CommandBuffer(QRhiImplementation *rhi); - ~QD3D11CommandBuffer(); - void destroy() override; - - // these must be kept at a reasonably low value otherwise sizeof Command explodes - static const int MAX_DYNAMIC_OFFSET_COUNT = 8; - static const int MAX_VERTEX_BUFFER_BINDING_COUNT = 8; - - struct Command { - enum Cmd { - ResetShaderResources, - SetRenderTarget, - Clear, - Viewport, - Scissor, - BindVertexBuffers, - BindIndexBuffer, - BindGraphicsPipeline, - BindShaderResources, - StencilRef, - BlendConstants, - Draw, - DrawIndexed, - UpdateSubRes, - CopySubRes, - ResolveSubRes, - GenMip, - DebugMarkBegin, - DebugMarkEnd, - DebugMarkMsg, - BindComputePipeline, - Dispatch - }; - enum ClearFlag { Color = 1, Depth = 2, Stencil = 4 }; - Cmd cmd; - - // QRhi*/QD3D11* references should be kept at minimum (so no - // QRhiTexture/Buffer/etc. pointers). - union Args { - struct { - QRhiRenderTarget *rt; - } setRenderTarget; - struct { - QRhiRenderTarget *rt; - int mask; - float c[4]; - float d; - quint32 s; - } clear; - struct { - float x, y, w, h; - float d0, d1; - } viewport; - struct { - int x, y, w, h; - } scissor; - struct { - int startSlot; - int slotCount; - ID3D11Buffer *buffers[MAX_VERTEX_BUFFER_BINDING_COUNT]; - UINT offsets[MAX_VERTEX_BUFFER_BINDING_COUNT]; - UINT strides[MAX_VERTEX_BUFFER_BINDING_COUNT]; - } bindVertexBuffers; - struct { - ID3D11Buffer *buffer; - quint32 offset; - DXGI_FORMAT format; - } bindIndexBuffer; - struct { - QD3D11GraphicsPipeline *ps; - } bindGraphicsPipeline; - struct { - QD3D11ShaderResourceBindings *srb; - bool offsetOnlyChange; - int dynamicOffsetCount; - uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offsetInConstants - } bindShaderResources; - struct { - QD3D11GraphicsPipeline *ps; - quint32 ref; - } stencilRef; - struct { - QD3D11GraphicsPipeline *ps; - float c[4]; - } blendConstants; - struct { - QD3D11GraphicsPipeline *ps; - quint32 vertexCount; - quint32 instanceCount; - quint32 firstVertex; - quint32 firstInstance; - } draw; - struct { - QD3D11GraphicsPipeline *ps; - quint32 indexCount; - quint32 instanceCount; - quint32 firstIndex; - qint32 vertexOffset; - quint32 firstInstance; - } drawIndexed; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - bool hasDstBox; - D3D11_BOX dstBox; - const void *src; // must come from retain*() - UINT srcRowPitch; - } updateSubRes; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - UINT dstX; - UINT dstY; - UINT dstZ; - ID3D11Resource *src; - UINT srcSubRes; - bool hasSrcBox; - D3D11_BOX srcBox; - } copySubRes; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - ID3D11Resource *src; - UINT srcSubRes; - DXGI_FORMAT format; - } resolveSubRes; - struct { - ID3D11ShaderResourceView *srv; - } genMip; - struct { - char s[64]; - } debugMark; - struct { - QD3D11ComputePipeline *ps; - } bindComputePipeline; - struct { - UINT x; - UINT y; - UINT z; - } dispatch; - } args; - }; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - QRhiBackendCommandList<Command> commands; - PassType recordingPass; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - ID3D11Buffer *currentIndexBuffer; - quint32 currentIndexOffset; - DXGI_FORMAT currentIndexFormat; - ID3D11Buffer *currentVertexBuffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - quint32 currentVertexOffsets[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - - QVarLengthArray<QByteArray, 4> dataRetainPool; - QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; - QVarLengthArray<QImage, 4> imageRetainPool; - - // relies heavily on implicit sharing (no copies of the actual data will be made) - const uchar *retainData(const QByteArray &data) { - dataRetainPool.append(data); - return reinterpret_cast<const uchar *>(dataRetainPool.last().constData()); - } - const uchar *retainBufferData(const QRhiBufferData &data) { - bufferDataRetainPool.append(data); - return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); - } - const uchar *retainImage(const QImage &image) { - imageRetainPool.append(image); - return imageRetainPool.last().constBits(); - } - void resetCommands() { - commands.reset(); - dataRetainPool.clear(); - bufferDataRetainPool.clear(); - imageRetainPool.clear(); - } - void resetState() { - recordingPass = NoPass; - currentTarget = nullptr; - resetCommands(); - resetCachedState(); - } - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - resetCachedShaderResourceState(); - } - void resetCachedShaderResourceState() { - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - currentIndexBuffer = nullptr; - currentIndexOffset = 0; - currentIndexFormat = DXGI_FORMAT_R16_UINT; - memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); - memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); - } -}; - -struct QD3D11SwapChain : public QRhiSwapChain -{ - QD3D11SwapChain(QRhiImplementation *rhi); - ~QD3D11SwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - void releaseBuffers(); - bool newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, - ID3D11Texture2D **tex, ID3D11RenderTargetView **rtv) const; - - QWindow *window = nullptr; - QSize pixelSize; - QD3D11ReferenceRenderTarget rt; - QD3D11CommandBuffer cb; - DXGI_FORMAT colorFormat; - IDXGISwapChain *swapChain = nullptr; - static const int BUFFER_COUNT = 2; - ID3D11Texture2D *backBufferTex; - ID3D11RenderTargetView *backBufferRtv; - ID3D11Texture2D *msaaTex[BUFFER_COUNT]; - ID3D11RenderTargetView *msaaRtv[BUFFER_COUNT]; - DXGI_SAMPLE_DESC sampleDesc; - int currentFrameSlot = 0; - int frameCount = 0; - QD3D11RenderBuffer *ds = nullptr; - bool timestampActive[BUFFER_COUNT]; - ID3D11Query *timestampDisjointQuery[BUFFER_COUNT]; - ID3D11Query *timestampQuery[BUFFER_COUNT * 2]; - UINT swapInterval = 1; -}; - -class QRhiD3D11 : public QRhiImplementation -{ -public: - QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *importDevice = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - void sendVMemStatsToProfiler() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void enqueueSubresUpload(QD3D11Texture *texD, QD3D11CommandBuffer *cbD, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, - const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[]); - void executeBufferHostWrites(QD3D11Buffer *bufD); - void bindShaderResources(QD3D11ShaderResourceBindings *srbD, - const uint *dynOfsPairs, int dynOfsPairCount, - bool offsetOnlyChange); - void resetShaderResources(); - void executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain *timestampSwapChain = nullptr); - DXGI_SAMPLE_DESC effectiveSampleCount(int sampleCount) const; - void finishActiveReadbacks(); - void reportLiveObjects(ID3D11Device *device); - void clearShaderCache(); - - bool debugLayer = false; - bool importedDeviceAndContext = false; - ID3D11Device *dev = nullptr; - ID3D11DeviceContext1 *context = nullptr; - D3D_FEATURE_LEVEL featureLevel = D3D_FEATURE_LEVEL(0); - LUID adapterLuid = {}; - ID3DUserDefinedAnnotation *annotations = nullptr; - IDXGIFactory1 *dxgiFactory = nullptr; - bool hasDxgi2 = false; - bool supportsFlipDiscardSwapchain = false; - bool deviceLost = false; - QRhiD3D11NativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - - struct { - int vsHighestActiveVertexBufferBinding = -1; - bool vsHasIndexBufferBound = false; - int vsHighestActiveSrvBinding = -1; - int fsHighestActiveSrvBinding = -1; - int csHighestActiveSrvBinding = -1; - int csHighestActiveUavBinding = -1; - QD3D11SwapChain *currentSwapChain = nullptr; - } contextState; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } - bool active = false; - QD3D11CommandBuffer cbWrapper; - } ofr; - - struct TextureReadback { - QRhiReadbackDescription desc; - QRhiReadbackResult *result; - ID3D11Texture2D *stagingTex; - quint32 byteSize; - quint32 bpl; - QSize pixelSize; - QRhiTexture::Format format; - }; - QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; - struct BufferReadback { - QRhiBufferReadbackResult *result; - quint32 byteSize; - ID3D11Buffer *stagingBuf; - }; - QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; - - struct Shader { - Shader() = default; - Shader(IUnknown *s, const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) - : s(s), bytecode(bytecode), nativeResourceBindingMap(rbm) { } - IUnknown *s; - QByteArray bytecode; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - }; - QHash<QRhiShaderStage, Shader> m_shaderCache; - - struct DeviceCurse { - DeviceCurse(QRhiD3D11 *impl) : q(impl) { } - QRhiD3D11 *q; - int framesToActivate = -1; - bool permanent = false; - int framesLeft = 0; - ID3D11ComputeShader *cs = nullptr; - - void initResources(); - void releaseResources(); - void activate(); - } deviceCurse; -}; - -Q_DECLARE_TYPEINFO(QRhiD3D11::TextureReadback, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiD3D11::BufferReadback, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhid3d12.cpp b/src/gui/rhi/qrhid3d12.cpp new file mode 100644 index 0000000000..0f176c683d --- /dev/null +++ b/src/gui/rhi/qrhid3d12.cpp @@ -0,0 +1,6569 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhid3d12_p.h" +#include <qmath.h> +#include <QtCore/private/qsystemerror_p.h> +#include <comdef.h> +#include "qrhid3dhelpers_p.h" +#include "cs_mipmap_p.h" + +#if __has_include(<pix.h>) +#include <pix.h> +#define QRHI_D3D12_HAS_OLD_PIX +#endif + +#ifdef __ID3D12Device2_INTERFACE_DEFINED__ + +QT_BEGIN_NAMESPACE + +/* + Direct 3D 12 backend. +*/ + +/*! + \class QRhiD3D12InitParams + \inmodule QtGui + \brief Direct3D 12 specific initialization parameters. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + A D3D12-based QRhi needs no special parameters for initialization. If + desired, enableDebugLayer can be set to \c true to enable the Direct3D + debug layer. This can be useful during development, but should be avoided + in production builds. + + \badcode + QRhiD3D12InitParams params; + params.enableDebugLayer = true; + rhi = QRhi::create(QRhi::D3D12, ¶ms); + \endcode + + \note QRhiSwapChain should only be used in combination with QWindow + instances that have their surface type set to QSurface::Direct3DSurface. + + \section2 Working with existing Direct3D 12 devices + + When interoperating with another graphics engine, it may be necessary to + get a QRhi instance that uses the same Direct3D device. This can be + achieved by passing a pointer to a QRhiD3D12NativeHandles to + QRhi::create(). QRhi does not take ownership of any of the external + objects. + + Sometimes, for example when using QRhi in combination with OpenXR, one will + want to specify which adapter to use, and optionally, which feature level + to request on the device, while leaving the device creation to QRhi. This + is achieved by leaving the device pointer set to null, while specifying the + adapter LUID and feature level. + + Optionally the ID3D12CommandQueue can be specified as well, by setting \c + commandQueue to a non-null value. + */ + +/*! + \variable QRhiD3D12InitParams::enableDebugLayer + + When set to true, the debug layer is enabled, if installed and available. + The default value is false. +*/ + +/*! + \class QRhiD3D12NativeHandles + \inmodule QtGui + \brief Holds the D3D12 device used by the QRhi. + + \note The class uses \c{void *} as the type since including the COM-based + \c{d3d12.h} headers is not acceptable here. The actual types are + \c{ID3D12Device *} and \c{ID3D12CommandQueue *}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiD3D12NativeHandles::dev + + Points to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nn-d3d12-id3d12device}{ID3D12Device} + or left set to \nullptr if no existing device is to be imported. +*/ + +/*! + \variable QRhiD3D12NativeHandles::minimumFeatureLevel + + Specifies the \b minimum feature level passed to + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-d3d12createdevice}{D3D12CreateDevice()}. + When not set, \c{D3D_FEATURE_LEVEL_11_0} is used. See + \l{https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-levels}{this + page} for details. + + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::adapterLuidLow + + The low part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::adapterLuidHigh + + The high part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::commandQueue + + When set, must point to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nn-d3d12-id3d12commandqueue}{ID3D12CommandQueue}. + It allows to optionally import a command queue as well, in addition to a + device. +*/ + +/*! + \class QRhiD3D12CommandBufferNativeHandles + \inmodule QtGui + \brief Holds the ID3D12GraphicsCommandList1 object that is backing a QRhiCommandBuffer. + + \note The command list object is only guaranteed to be valid, and + in recording state, while recording a frame. That is, between a + \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} or + \l{QRhi::beginOffscreenFrame()}{beginOffscreenFrame()} - + \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} pair. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiD3D12CommandBufferNativeHandles::commandList +*/ + +// https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-levels +static const D3D_FEATURE_LEVEL MIN_FEATURE_LEVEL = D3D_FEATURE_LEVEL_11_0; + +QRhiD3D12::QRhiD3D12(QRhiD3D12InitParams *params, QRhiD3D12NativeHandles *importParams) +{ + debugLayer = params->enableDebugLayer; + if (importParams) { + if (importParams->dev) { + ID3D12Device *d3d12Device = reinterpret_cast<ID3D12Device *>(importParams->dev); + if (SUCCEEDED(d3d12Device->QueryInterface(__uuidof(ID3D12Device2), reinterpret_cast<void **>(&dev)))) { + // get rid of the ref added by QueryInterface + d3d12Device->Release(); + importedDevice = true; + } else { + qWarning("ID3D12Device2 not supported, cannot import device"); + } + } + if (importParams->commandQueue) { + cmdQueue = reinterpret_cast<ID3D12CommandQueue *>(importParams->commandQueue); + importedCommandQueue = true; + } + minimumFeatureLevel = D3D_FEATURE_LEVEL(importParams->minimumFeatureLevel); + adapterLuid.LowPart = importParams->adapterLuidLow; + adapterLuid.HighPart = importParams->adapterLuidHigh; + } +} + +template <class Int> +inline Int aligned(Int v, Int byteAlign) +{ + return (v + byteAlign - 1) & ~(byteAlign - 1); +} + +static inline UINT calcSubresource(UINT mipSlice, UINT arraySlice, UINT mipLevels) +{ + return mipSlice + arraySlice * mipLevels; +} + +static inline QD3D12RenderTargetData *rtData(QRhiRenderTarget *rt) +{ + switch (rt->resourceType()) { + case QRhiResource::SwapChainRenderTarget: + return &QRHI_RES(QD3D12SwapChainRenderTarget, rt)->d; + case QRhiResource::TextureRenderTarget: + return &QRHI_RES(QD3D12TextureRenderTarget, rt)->d; + break; + default: + break; + } + Q_UNREACHABLE_RETURN(nullptr); +} + +bool QRhiD3D12::create(QRhi::Flags flags) +{ + rhiFlags = flags; + + UINT factoryFlags = 0; + if (debugLayer) + factoryFlags |= DXGI_CREATE_FACTORY_DEBUG; + HRESULT hr = CreateDXGIFactory2(factoryFlags, __uuidof(IDXGIFactory2), reinterpret_cast<void **>(&dxgiFactory)); + if (FAILED(hr)) { + // retry without debug, if it was requested (to match D3D11 backend behavior) + if (debugLayer) { + qCDebug(QRHI_LOG_INFO, "Debug layer was requested but is not available. " + "Attempting to create DXGIFactory2 without it."); + factoryFlags &= ~DXGI_CREATE_FACTORY_DEBUG; + hr = CreateDXGIFactory2(factoryFlags, __uuidof(IDXGIFactory2), reinterpret_cast<void **>(&dxgiFactory)); + } + if (SUCCEEDED(hr)) { + debugLayer = false; + } else { + qWarning("CreateDXGIFactory2() failed to create DXGI factory: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + + supportsAllowTearing = false; + IDXGIFactory5 *factory5 = nullptr; + if (SUCCEEDED(dxgiFactory->QueryInterface(__uuidof(IDXGIFactory5), reinterpret_cast<void **>(&factory5)))) { + BOOL allowTearing = false; + if (SUCCEEDED(factory5->CheckFeatureSupport(DXGI_FEATURE_PRESENT_ALLOW_TEARING, &allowTearing, sizeof(allowTearing)))) + supportsAllowTearing = allowTearing; + factory5->Release(); + } + + if (debugLayer) { + ID3D12Debug1 *debug = nullptr; + if (SUCCEEDED(D3D12GetDebugInterface(__uuidof(ID3D12Debug1), reinterpret_cast<void **>(&debug)))) { + qCDebug(QRHI_LOG_INFO, "Enabling D3D12 debug layer"); + debug->EnableDebugLayer(); + debug->Release(); + } + } + + if (!importedDevice) { + IDXGIAdapter1 *adapter; + int requestedAdapterIndex = -1; + if (qEnvironmentVariableIsSet("QT_D3D_ADAPTER_INDEX")) + requestedAdapterIndex = qEnvironmentVariableIntValue("QT_D3D_ADAPTER_INDEX"); + + // The importParams may specify an adapter by the luid, take that into account. + if (requestedAdapterIndex < 0 && (adapterLuid.LowPart || adapterLuid.HighPart)) { + for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { + DXGI_ADAPTER_DESC1 desc; + adapter->GetDesc1(&desc); + adapter->Release(); + if (desc.AdapterLuid.LowPart == adapterLuid.LowPart + && desc.AdapterLuid.HighPart == adapterLuid.HighPart) + { + requestedAdapterIndex = adapterIndex; + break; + } + } + } + + if (requestedAdapterIndex < 0 && flags.testFlag(QRhi::PreferSoftwareRenderer)) { + for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { + DXGI_ADAPTER_DESC1 desc; + adapter->GetDesc1(&desc); + adapter->Release(); + if (desc.Flags & DXGI_ADAPTER_FLAG_SOFTWARE) { + requestedAdapterIndex = adapterIndex; + break; + } + } + } + + activeAdapter = nullptr; + for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { + DXGI_ADAPTER_DESC1 desc; + adapter->GetDesc1(&desc); + const QString name = QString::fromUtf16(reinterpret_cast<char16_t *>(desc.Description)); + qCDebug(QRHI_LOG_INFO, "Adapter %d: '%s' (vendor 0x%X device 0x%X flags 0x%X)", + adapterIndex, + qPrintable(name), + desc.VendorId, + desc.DeviceId, + desc.Flags); + if (!activeAdapter && (requestedAdapterIndex < 0 || requestedAdapterIndex == adapterIndex)) { + activeAdapter = adapter; + adapterLuid = desc.AdapterLuid; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); + qCDebug(QRHI_LOG_INFO, " using this adapter"); + } else { + adapter->Release(); + } + } + if (!activeAdapter) { + qWarning("No adapter"); + return false; + } + + if (minimumFeatureLevel == 0) + minimumFeatureLevel = MIN_FEATURE_LEVEL; + + hr = D3D12CreateDevice(activeAdapter, + minimumFeatureLevel, + __uuidof(ID3D12Device2), + reinterpret_cast<void **>(&dev)); + if (FAILED(hr)) { + qWarning("Failed to create D3D12 device: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } else { + Q_ASSERT(dev); + // cannot just get a IDXGIDevice from the ID3D12Device anymore, look up the adapter instead + adapterLuid = dev->GetAdapterLuid(); + IDXGIAdapter1 *adapter; + for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { + DXGI_ADAPTER_DESC1 desc; + adapter->GetDesc1(&desc); + if (desc.AdapterLuid.LowPart == adapterLuid.LowPart + && desc.AdapterLuid.HighPart == adapterLuid.HighPart) + { + activeAdapter = adapter; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); + break; + } else { + adapter->Release(); + } + } + if (!activeAdapter) { + qWarning("No adapter"); + return false; + } + qCDebug(QRHI_LOG_INFO, "Using imported device %p", dev); + } + + if (debugLayer) { + ID3D12InfoQueue *infoQueue; + if (SUCCEEDED(dev->QueryInterface(__uuidof(ID3D12InfoQueue), reinterpret_cast<void **>(&infoQueue)))) { + if (qEnvironmentVariableIntValue("QT_D3D_DEBUG_BREAK")) { + infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_CORRUPTION, true); + infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_ERROR, true); + infoQueue->SetBreakOnSeverity(D3D12_MESSAGE_SEVERITY_WARNING, true); + } + D3D12_INFO_QUEUE_FILTER filter = {}; + D3D12_MESSAGE_ID suppressedMessages[2] = { + // there is no way of knowing the clear color upfront + D3D12_MESSAGE_ID_CLEARRENDERTARGETVIEW_MISMATCHINGCLEARVALUE, + // we have no control over viewport and scissor rects + D3D12_MESSAGE_ID_DRAW_EMPTY_SCISSOR_RECTANGLE + }; + filter.DenyList.NumIDs = 2; + filter.DenyList.pIDList = suppressedMessages; + // Setting the filter would enable Info messages (e.g. about + // resource creation) which we don't need. + D3D12_MESSAGE_SEVERITY infoSev = D3D12_MESSAGE_SEVERITY_INFO; + filter.DenyList.NumSeverities = 1; + filter.DenyList.pSeverityList = &infoSev; + infoQueue->PushStorageFilter(&filter); + infoQueue->Release(); + } + } + + if (!importedCommandQueue) { + D3D12_COMMAND_QUEUE_DESC queueDesc = {}; + queueDesc.Type = D3D12_COMMAND_LIST_TYPE_DIRECT; + queueDesc.Priority = D3D12_COMMAND_QUEUE_PRIORITY_NORMAL; + hr = dev->CreateCommandQueue(&queueDesc, __uuidof(ID3D12CommandQueue), reinterpret_cast<void **>(&cmdQueue)); + if (FAILED(hr)) { + qWarning("Failed to create command queue: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + + hr = dev->CreateFence(0, D3D12_FENCE_FLAG_NONE, __uuidof(ID3D12Fence), reinterpret_cast<void **>(&fullFence)); + if (FAILED(hr)) { + qWarning("Failed to create fence: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + fullFenceEvent = CreateEvent(nullptr, FALSE, FALSE, nullptr); + fullFenceCounter = 0; + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + hr = dev->CreateCommandAllocator(D3D12_COMMAND_LIST_TYPE_DIRECT, + __uuidof(ID3D12CommandAllocator), + reinterpret_cast<void **>(&cmdAllocators[i])); + if (FAILED(hr)) { + qWarning("Failed to create command allocator: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + + if (!vma.create(dev, activeAdapter)) { + qWarning("Failed to initialize graphics memory suballocator"); + return false; + } + + if (!rtvPool.create(dev, D3D12_DESCRIPTOR_HEAP_TYPE_RTV, "main RTV pool")) { + qWarning("Could not create RTV pool"); + return false; + } + + if (!dsvPool.create(dev, D3D12_DESCRIPTOR_HEAP_TYPE_DSV, "main DSV pool")) { + qWarning("Could not create DSV pool"); + return false; + } + + if (!cbvSrvUavPool.create(dev, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, "main CBV-SRV-UAV pool")) { + qWarning("Could not create CBV-SRV-UAV pool"); + return false; + } + + resourcePool.create("main resource pool"); + pipelinePool.create("main pipeline pool"); + rootSignaturePool.create("main root signature pool"); + releaseQueue.create(&resourcePool, &pipelinePool, &rootSignaturePool); + barrierGen.create(&resourcePool); + + if (!samplerMgr.create(dev)) { + qWarning("Could not create sampler pool and shader-visible sampler heap"); + return false; + } + + if (!mipmapGen.create(this)) { + qWarning("Could not initialize mipmap generator"); + return false; + } + + const qint32 smallStagingSize = aligned(SMALL_STAGING_AREA_BYTES_PER_FRAME, QD3D12StagingArea::ALIGNMENT); + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (!smallStagingAreas[i].create(this, smallStagingSize, D3D12_HEAP_TYPE_UPLOAD)) { + qWarning("Could not create host-visible staging area"); + return false; + } + QString decoratedName = QLatin1String("Small staging area buffer/"); + decoratedName += QString::number(i); + smallStagingAreas[i].mem.buffer->SetName(reinterpret_cast<LPCWSTR>(decoratedName.utf16())); + } + + if (!shaderVisibleCbvSrvUavHeap.create(dev, + D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, + SHADER_VISIBLE_CBV_SRV_UAV_HEAP_PER_FRAME_START_SIZE)) + { + qWarning("Could not create first shader-visible CBV/SRV/UAV heap"); + return false; + } + + if (flags.testFlag(QRhi::EnableTimestamps)) { + static bool wantsStablePowerState = qEnvironmentVariableIntValue("QT_D3D_STABLE_POWER_STATE"); + // + // https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-id3d12device-setstablepowerstate + // + // NB! This is a _global_ setting, affecting other processes (and 3D + // APIs such as Vulkan), as long as this application is running. Hence + // making it an env.var. for now. Never enable it in production. But + // extremely useful for the GPU timings with NVIDIA at least; the + // timestamps become stable and smooth, making the number readable and + // actually useful e.g. in Quick 3D's DebugView when this is enabled. + // (otherwise the number's all over the place) + // + // See also + // https://developer.nvidia.com/blog/advanced-api-performance-setstablepowerstate/ + // for possible other approaches. + // + if (wantsStablePowerState) + dev->SetStablePowerState(TRUE); + + hr = cmdQueue->GetTimestampFrequency(×tampTicksPerSecond); + if (FAILED(hr)) { + qWarning("Failed to query timestamp frequency: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + if (!timestampQueryHeap.create(dev, QD3D12_FRAMES_IN_FLIGHT * 2, D3D12_QUERY_HEAP_TYPE_TIMESTAMP)) { + qWarning("Failed to create timestamp query pool"); + return false; + } + const quint32 readbackBufSize = QD3D12_FRAMES_IN_FLIGHT * 2 * sizeof(quint64); + if (!timestampReadbackArea.create(this, readbackBufSize, D3D12_HEAP_TYPE_READBACK)) { + qWarning("Failed to create timestamp readback buffer"); + return false; + } + timestampReadbackArea.mem.buffer->SetName(L"Timestamp readback buffer"); + memset(timestampReadbackArea.mem.p, 0, readbackBufSize); + } + + caps = {}; + D3D12_FEATURE_DATA_D3D12_OPTIONS3 options3 = {}; + if (SUCCEEDED(dev->CheckFeatureSupport(D3D12_FEATURE_D3D12_OPTIONS3, &options3, sizeof(options3)))) { + caps.multiView = options3.ViewInstancingTier != D3D12_VIEW_INSTANCING_TIER_NOT_SUPPORTED; + // https://microsoft.github.io/DirectX-Specs/d3d/RelaxedCasting.html + caps.textureViewFormat = options3.CastingFullyTypedFormatSupported; + } + + deviceLost = false; + offscreenActive = false; + + nativeHandlesStruct.dev = dev; + nativeHandlesStruct.minimumFeatureLevel = minimumFeatureLevel; + nativeHandlesStruct.adapterLuidLow = adapterLuid.LowPart; + nativeHandlesStruct.adapterLuidHigh = adapterLuid.HighPart; + nativeHandlesStruct.commandQueue = cmdQueue; + + return true; +} + +void QRhiD3D12::destroy() +{ + if (!deviceLost && fullFence && fullFenceEvent) + waitGpu(); + + releaseQueue.releaseAll(); + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (offscreenCb[i]) { + if (offscreenCb[i]->cmdList) + offscreenCb[i]->cmdList->Release(); + delete offscreenCb[i]; + offscreenCb[i] = nullptr; + } + } + + timestampQueryHeap.destroy(); + timestampReadbackArea.destroy(); + + shaderVisibleCbvSrvUavHeap.destroy(); + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) + smallStagingAreas[i].destroy(); + + mipmapGen.destroy(); + samplerMgr.destroy(); + resourcePool.destroy(); + pipelinePool.destroy(); + rootSignaturePool.destroy(); + rtvPool.destroy(); + dsvPool.destroy(); + cbvSrvUavPool.destroy(); + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (cmdAllocators[i]) { + cmdAllocators[i]->Release(); + cmdAllocators[i] = nullptr; + } + } + + if (fullFenceEvent) { + CloseHandle(fullFenceEvent); + fullFenceEvent = nullptr; + } + + if (fullFence) { + fullFence->Release(); + fullFence = nullptr; + } + + if (!importedCommandQueue) { + if (cmdQueue) { + cmdQueue->Release(); + cmdQueue = nullptr; + } + } + + vma.destroy(); + + if (!importedDevice) { + if (dev) { + dev->Release(); + dev = nullptr; + } + } + + if (dcompDevice) { + dcompDevice->Release(); + dcompDevice = nullptr; + } + + if (activeAdapter) { + activeAdapter->Release(); + activeAdapter = nullptr; + } + + if (dxgiFactory) { + dxgiFactory->Release(); + dxgiFactory = nullptr; + } +} + +QList<int> QRhiD3D12::supportedSampleCounts() const +{ + return { 1, 2, 4, 8 }; +} + +QRhiSwapChain *QRhiD3D12::createSwapChain() +{ + return new QD3D12SwapChain(this); +} + +QRhiBuffer *QRhiD3D12::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) +{ + return new QD3D12Buffer(this, type, usage, size); +} + +int QRhiD3D12::ubufAlignment() const +{ + return D3D12_CONSTANT_BUFFER_DATA_PLACEMENT_ALIGNMENT; // 256 +} + +bool QRhiD3D12::isYUpInFramebuffer() const +{ + return false; +} + +bool QRhiD3D12::isYUpInNDC() const +{ + return true; +} + +bool QRhiD3D12::isClipDepthZeroToOne() const +{ + return true; +} + +QMatrix4x4 QRhiD3D12::clipSpaceCorrMatrix() const +{ + // Like with Vulkan, but Y is already good. + + static QMatrix4x4 m; + if (m.isIdentity()) { + // NB the ctor takes row-major + m = QMatrix4x4(1.0f, 0.0f, 0.0f, 0.0f, + 0.0f, 1.0f, 0.0f, 0.0f, + 0.0f, 0.0f, 0.5f, 0.5f, + 0.0f, 0.0f, 0.0f, 1.0f); + } + return m; +} + +bool QRhiD3D12::isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const +{ + Q_UNUSED(flags); + + if (format >= QRhiTexture::ETC2_RGB8 && format <= QRhiTexture::ASTC_12x12) + return false; + + return true; +} + +bool QRhiD3D12::isFeatureSupported(QRhi::Feature feature) const +{ + switch (feature) { + case QRhi::MultisampleTexture: + return true; + case QRhi::MultisampleRenderBuffer: + return true; + case QRhi::DebugMarkers: +#ifdef QRHI_D3D12_HAS_OLD_PIX + return true; +#else + return false; +#endif + case QRhi::Timestamps: + return true; + case QRhi::Instancing: + return true; + case QRhi::CustomInstanceStepRate: + return true; + case QRhi::PrimitiveRestart: + return true; + case QRhi::NonDynamicUniformBuffers: + return false; + case QRhi::NonFourAlignedEffectiveIndexBufferOffset: + return true; + case QRhi::NPOTTextureRepeat: + return true; + case QRhi::RedOrAlpha8IsRed: + return true; + case QRhi::ElementIndexUint: + return true; + case QRhi::Compute: + return true; + case QRhi::WideLines: + return false; + case QRhi::VertexShaderPointSize: + return false; + case QRhi::BaseVertex: + return true; + case QRhi::BaseInstance: + return true; + case QRhi::TriangleFanTopology: + return false; + case QRhi::ReadBackNonUniformBuffer: + return true; + case QRhi::ReadBackNonBaseMipLevel: + return true; + case QRhi::TexelFetch: + return true; + case QRhi::RenderToNonBaseMipLevel: + return true; + case QRhi::IntAttributes: + return true; + case QRhi::ScreenSpaceDerivatives: + return true; + case QRhi::ReadBackAnyTextureFormat: + return true; + case QRhi::PipelineCacheDataLoadSave: + return false; // ### + case QRhi::ImageDataStride: + return true; + case QRhi::RenderBufferImport: + return false; + case QRhi::ThreeDimensionalTextures: + return true; + case QRhi::RenderTo3DTextureSlice: + return true; + case QRhi::TextureArrays: + return true; + case QRhi::Tessellation: + return true; + case QRhi::GeometryShader: + return true; + case QRhi::TextureArrayRange: + return true; + case QRhi::NonFillPolygonMode: + return true; + case QRhi::OneDimensionalTextures: + return true; + case QRhi::OneDimensionalTextureMipmaps: + return false; // we generate mipmaps ourselves with compute and this is not implemented + case QRhi::HalfAttributes: + return true; + case QRhi::RenderToOneDimensionalTexture: + return true; + case QRhi::ThreeDimensionalTextureMipmaps: + return false; // we generate mipmaps ourselves with compute and this is not implemented + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return caps.textureViewFormat; + case QRhi::ResolveDepthStencil: + // there is no Multisample Resolve support for depth/stencil formats + // https://learn.microsoft.com/en-us/windows/win32/direct3ddxgi/hardware-support-for-direct3d-12-1-formats + return false; + } + return false; +} + +int QRhiD3D12::resourceLimit(QRhi::ResourceLimit limit) const +{ + switch (limit) { + case QRhi::TextureSizeMin: + return 1; + case QRhi::TextureSizeMax: + return 16384; + case QRhi::MaxColorAttachments: + return 8; + case QRhi::FramesInFlight: + return QD3D12_FRAMES_IN_FLIGHT; + case QRhi::MaxAsyncReadbackFrames: + return QD3D12_FRAMES_IN_FLIGHT; + case QRhi::MaxThreadGroupsPerDimension: + return 65535; + case QRhi::MaxThreadsPerThreadGroup: + return 1024; + case QRhi::MaxThreadGroupX: + return 1024; + case QRhi::MaxThreadGroupY: + return 1024; + case QRhi::MaxThreadGroupZ: + return 1024; + case QRhi::TextureArraySizeMax: + return 2048; + case QRhi::MaxUniformBufferRange: + return 65536; + case QRhi::MaxVertexInputs: + return 32; + case QRhi::MaxVertexOutputs: + return 32; + } + return 0; +} + +const QRhiNativeHandles *QRhiD3D12::nativeHandles() +{ + return &nativeHandlesStruct; +} + +QRhiDriverInfo QRhiD3D12::driverInfo() const +{ + return driverInfoStruct; +} + +QRhiStats QRhiD3D12::statistics() +{ + QRhiStats result; + result.totalPipelineCreationTime = totalPipelineCreationTime(); + + D3D12MA::Budget budgets[2]; // [gpu, system] with discreet GPU or [shared, nothing] with UMA + vma.getBudget(&budgets[0], &budgets[1]); + for (int i = 0; i < 2; ++i) { + const D3D12MA::Statistics &stats(budgets[i].Stats); + result.blockCount += stats.BlockCount; + result.allocCount += stats.AllocationCount; + result.usedBytes += stats.AllocationBytes; + result.unusedBytes += stats.BlockBytes - stats.AllocationBytes; + result.totalUsageBytes += budgets[i].UsageBytes; + } + + return result; +} + +bool QRhiD3D12::makeThreadLocalNativeContextCurrent() +{ + // not applicable + return false; +} + +void QRhiD3D12::releaseCachedResources() +{ + shaderBytecodeCache.data.clear(); +} + +bool QRhiD3D12::isDeviceLost() const +{ + return deviceLost; +} + +QByteArray QRhiD3D12::pipelineCacheData() +{ + return {}; +} + +void QRhiD3D12::setPipelineCacheData(const QByteArray &data) +{ + Q_UNUSED(data); +} + +QRhiRenderBuffer *QRhiD3D12::createRenderBuffer(QRhiRenderBuffer::Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) +{ + return new QD3D12RenderBuffer(this, type, pixelSize, sampleCount, flags, backingFormatHint); +} + +QRhiTexture *QRhiD3D12::createTexture(QRhiTexture::Format format, + const QSize &pixelSize, int depth, int arraySize, + int sampleCount, QRhiTexture::Flags flags) +{ + return new QD3D12Texture(this, format, pixelSize, depth, arraySize, sampleCount, flags); +} + +QRhiSampler *QRhiD3D12::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler::AddressMode u, QRhiSampler::AddressMode v, QRhiSampler::AddressMode w) +{ + return new QD3D12Sampler(this, magFilter, minFilter, mipmapMode, u, v, w); +} + +QRhiTextureRenderTarget *QRhiD3D12::createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) +{ + return new QD3D12TextureRenderTarget(this, desc, flags); +} + +QRhiGraphicsPipeline *QRhiD3D12::createGraphicsPipeline() +{ + return new QD3D12GraphicsPipeline(this); +} + +QRhiComputePipeline *QRhiD3D12::createComputePipeline() +{ + return new QD3D12ComputePipeline(this); +} + +QRhiShaderResourceBindings *QRhiD3D12::createShaderResourceBindings() +{ + return new QD3D12ShaderResourceBindings(this); +} + +void QRhiD3D12::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline *ps) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + QD3D12GraphicsPipeline *psD = QRHI_RES(QD3D12GraphicsPipeline, ps); + const bool pipelineChanged = cbD->currentGraphicsPipeline != psD || cbD->currentPipelineGeneration != psD->generation; + + if (pipelineChanged) { + cbD->currentGraphicsPipeline = psD; + cbD->currentComputePipeline = nullptr; + cbD->currentPipelineGeneration = psD->generation; + + if (QD3D12Pipeline *pipeline = pipelinePool.lookupRef(psD->handle)) { + Q_ASSERT(pipeline->type == QD3D12Pipeline::Graphics); + cbD->cmdList->SetPipelineState(pipeline->pso); + if (QD3D12RootSignature *rs = rootSignaturePool.lookupRef(psD->rootSigHandle)) + cbD->cmdList->SetGraphicsRootSignature(rs->rootSig); + } + + cbD->cmdList->IASetPrimitiveTopology(psD->topology); + + if (psD->viewInstanceMask) + cbD->cmdList->SetViewInstanceMask(psD->viewInstanceMask); + } +} + +void QD3D12CommandBuffer::visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int, + int binding, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) +{ + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, d.buf); + quint32 offset = d.offset; + if (d.hasDynamicOffset) { + for (int i = 0; i < dynamicOffsetCount; ++i) { + const QRhiCommandBuffer::DynamicOffset &dynOfs(dynamicOffsets[i]); + if (dynOfs.first == binding) { + Q_ASSERT(aligned(dynOfs.second, 256u) == dynOfs.second); + offset += dynOfs.second; + } + } + } + QRHI_RES_RHI(QRhiD3D12); + visitorData.cbufs[s].append({ bufD->handles[rhiD->currentFrameSlot], offset }); +} + +void QD3D12CommandBuffer::visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int) +{ + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, d.tex); + visitorData.srvs[s].append(texD->srv); +} + +void QD3D12CommandBuffer::visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int) +{ + QD3D12Sampler *samplerD = QRHI_RES(QD3D12Sampler, d.sampler); + visitorData.samplers[s].append(samplerD->lookupOrCreateShaderVisibleDescriptor()); +} + +void QD3D12CommandBuffer::visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp, + int) +{ + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, d.buf); + // SPIRV-Cross generated HLSL uses RWByteAddressBuffer + D3D12_UNORDERED_ACCESS_VIEW_DESC uavDesc = {}; + uavDesc.Format = DXGI_FORMAT_R32_TYPELESS; + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_BUFFER; + uavDesc.Buffer.FirstElement = d.offset / 4; + uavDesc.Buffer.NumElements = aligned(bufD->m_size - d.offset, 4u) / 4; + uavDesc.Buffer.Flags = D3D12_BUFFER_UAV_FLAG_RAW; + visitorData.uavs[s].append({ bufD->handles[0], uavDesc }); +} + +void QD3D12CommandBuffer::visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp, + int) +{ + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, d.tex); + const bool isCube = texD->m_flags.testFlag(QRhiTexture::CubeMap); + const bool isArray = texD->m_flags.testFlag(QRhiTexture::TextureArray); + const bool is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + D3D12_UNORDERED_ACCESS_VIEW_DESC uavDesc = {}; + uavDesc.Format = texD->rtFormat; + if (isCube) { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY; + uavDesc.Texture2DArray.MipSlice = UINT(d.level); + uavDesc.Texture2DArray.FirstArraySlice = 0; + uavDesc.Texture2DArray.ArraySize = 6; + } else if (isArray) { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY; + uavDesc.Texture2DArray.MipSlice = UINT(d.level); + uavDesc.Texture2DArray.FirstArraySlice = 0; + uavDesc.Texture2DArray.ArraySize = UINT(qMax(0, texD->m_arraySize)); + } else if (is3D) { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE3D; + uavDesc.Texture3D.MipSlice = UINT(d.level); + } else { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2D; + uavDesc.Texture2D.MipSlice = UINT(d.level); + } + visitorData.uavs[s].append({ texD->handle, uavDesc }); +} + +void QRhiD3D12::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass != QD3D12CommandBuffer::NoPass); + QD3D12GraphicsPipeline *gfxPsD = QRHI_RES(QD3D12GraphicsPipeline, cbD->currentGraphicsPipeline); + QD3D12ComputePipeline *compPsD = QRHI_RES(QD3D12ComputePipeline, cbD->currentComputePipeline); + + if (!srb) { + if (gfxPsD) + srb = gfxPsD->m_shaderResourceBindings; + else + srb = compPsD->m_shaderResourceBindings; + } + + QD3D12ShaderResourceBindings *srbD = QRHI_RES(QD3D12ShaderResourceBindings, srb); + + for (int i = 0, ie = srbD->m_bindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings[i]); + switch (b->type) { + case QRhiShaderResourceBinding::UniformBuffer: + { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, b->u.ubuf.buf); + Q_ASSERT(bufD->m_usage.testFlag(QRhiBuffer::UniformBuffer)); + Q_ASSERT(bufD->m_type == QRhiBuffer::Dynamic); + bufD->executeHostWritesForFrameSlot(currentFrameSlot); + } + break; + case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: + { + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; + for (int elem = 0; elem < data->count; ++elem) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, data->texSamplers[elem].tex); + QD3D12Sampler *samplerD = QRHI_RES(QD3D12Sampler, data->texSamplers[elem].sampler); + // We use the same code path for both combined and separate + // images and samplers, so tex or sampler (but not both) can be + // null here. + Q_ASSERT(texD || samplerD); + if (texD) { + UINT state = 0; + if (b->stage == QRhiShaderResourceBinding::FragmentStage) { + state = D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE; + } else if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { + state = D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE | D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE; + } else { + state = D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE; + } + barrierGen.addTransitionBarrier(texD->handle, D3D12_RESOURCE_STATES(state)); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + } + } + break; + case QRhiShaderResourceBinding::ImageLoad: + case QRhiShaderResourceBinding::ImageStore: + case QRhiShaderResourceBinding::ImageLoadStore: + { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, b->u.simage.tex); + if (QD3D12Resource *res = resourcePool.lookupRef(texD->handle)) { + if (res->uavUsage) { + if (res->uavUsage & QD3D12Resource::UavUsageWrite) { + // RaW or WaW + barrierGen.enqueueUavBarrier(cbD, texD->handle); + } else { + if (b->type == QRhiShaderResourceBinding::ImageStore + || b->type == QRhiShaderResourceBinding::ImageLoadStore) + { + // WaR or WaW + barrierGen.enqueueUavBarrier(cbD, texD->handle); + } + } + } + res->uavUsage = 0; + if (b->type == QRhiShaderResourceBinding::ImageLoad || b->type == QRhiShaderResourceBinding::ImageLoadStore) + res->uavUsage |= QD3D12Resource::UavUsageRead; + if (b->type == QRhiShaderResourceBinding::ImageStore || b->type == QRhiShaderResourceBinding::ImageLoadStore) + res->uavUsage |= QD3D12Resource::UavUsageWrite; + barrierGen.addTransitionBarrier(texD->handle, D3D12_RESOURCE_STATE_UNORDERED_ACCESS); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + } + break; + case QRhiShaderResourceBinding::BufferLoad: + case QRhiShaderResourceBinding::BufferStore: + case QRhiShaderResourceBinding::BufferLoadStore: + { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, b->u.sbuf.buf); + Q_ASSERT(bufD->m_usage.testFlag(QRhiBuffer::StorageBuffer)); + Q_ASSERT(bufD->m_type != QRhiBuffer::Dynamic); + if (QD3D12Resource *res = resourcePool.lookupRef(bufD->handles[0])) { + if (res->uavUsage) { + if (res->uavUsage & QD3D12Resource::UavUsageWrite) { + // RaW or WaW + barrierGen.enqueueUavBarrier(cbD, bufD->handles[0]); + } else { + if (b->type == QRhiShaderResourceBinding::BufferStore + || b->type == QRhiShaderResourceBinding::BufferLoadStore) + { + // WaR or WaW + barrierGen.enqueueUavBarrier(cbD, bufD->handles[0]); + } + } + } + res->uavUsage = 0; + if (b->type == QRhiShaderResourceBinding::BufferLoad || b->type == QRhiShaderResourceBinding::BufferLoadStore) + res->uavUsage |= QD3D12Resource::UavUsageRead; + if (b->type == QRhiShaderResourceBinding::BufferStore || b->type == QRhiShaderResourceBinding::BufferLoadStore) + res->uavUsage |= QD3D12Resource::UavUsageWrite; + barrierGen.addTransitionBarrier(bufD->handles[0], D3D12_RESOURCE_STATE_UNORDERED_ACCESS); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + } + break; + } + } + + const bool srbChanged = gfxPsD ? (cbD->currentGraphicsSrb != srb) : (cbD->currentComputeSrb != srb); + const bool srbRebuilt = cbD->currentSrbGeneration != srbD->generation; + + if (srbChanged || srbRebuilt || srbD->hasDynamicOffset) { + const QD3D12ShaderStageData *stageData = gfxPsD ? gfxPsD->stageData.data() : &compPsD->stageData; + + // The order of root parameters must match + // QD3D12ShaderResourceBindings::createRootSignature(), meaning the + // logic below must mirror that function (uniform buffers first etc.) + + QD3D12ShaderResourceVisitor visitor(srbD, stageData, gfxPsD ? 5 : 1); + + QD3D12CommandBuffer::VisitorData &visitorData(cbD->visitorData); + visitorData = {}; + + using namespace std::placeholders; + visitor.uniformBuffer = std::bind(&QD3D12CommandBuffer::visitUniformBuffer, cbD, _1, _2, _3, _4, dynamicOffsetCount, dynamicOffsets); + visitor.texture = std::bind(&QD3D12CommandBuffer::visitTexture, cbD, _1, _2, _3); + visitor.sampler = std::bind(&QD3D12CommandBuffer::visitSampler, cbD, _1, _2, _3); + visitor.storageBuffer = std::bind(&QD3D12CommandBuffer::visitStorageBuffer, cbD, _1, _2, _3, _4); + visitor.storageImage = std::bind(&QD3D12CommandBuffer::visitStorageImage, cbD, _1, _2, _3, _4); + + visitor.visit(); + + quint32 cbvSrvUavCount = 0; + for (int s = 0; s < 6; ++s) { + // CBs use root constant buffer views, no need to count them here + cbvSrvUavCount += visitorData.srvs[s].count(); + cbvSrvUavCount += visitorData.uavs[s].count(); + } + + bool gotNewHeap = false; + if (!ensureShaderVisibleDescriptorHeapCapacity(&shaderVisibleCbvSrvUavHeap, + D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, + currentFrameSlot, + cbvSrvUavCount, + &gotNewHeap)) + { + return; + } + if (gotNewHeap) { + qCDebug(QRHI_LOG_INFO, "Created new shader-visible CBV/SRV/UAV descriptor heap," + " per-frame slice size is now %u," + " if this happens frequently then that's not great.", + shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[0].capacity); + bindShaderVisibleHeaps(cbD); + } + + int rootParamIndex = 0; + for (int s = 0; s < 6; ++s) { + if (!visitorData.cbufs[s].isEmpty()) { + for (int i = 0, count = visitorData.cbufs[s].count(); i < count; ++i) { + const auto &cbuf(visitorData.cbufs[s][i]); + if (QD3D12Resource *res = resourcePool.lookupRef(cbuf.first)) { + quint32 offset = cbuf.second; + D3D12_GPU_VIRTUAL_ADDRESS gpuAddr = res->resource->GetGPUVirtualAddress() + offset; + if (cbD->currentGraphicsPipeline) + cbD->cmdList->SetGraphicsRootConstantBufferView(rootParamIndex, gpuAddr); + else + cbD->cmdList->SetComputeRootConstantBufferView(rootParamIndex, gpuAddr); + } + rootParamIndex += 1; + } + } + } + for (int s = 0; s < 6; ++s) { + if (!visitorData.srvs[s].isEmpty()) { + QD3D12DescriptorHeap &gpuSrvHeap(shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[currentFrameSlot]); + QD3D12Descriptor startDesc = gpuSrvHeap.get(visitorData.srvs[s].count()); + for (int i = 0, count = visitorData.srvs[s].count(); i < count; ++i) { + const auto &srv(visitorData.srvs[s][i]); + dev->CopyDescriptorsSimple(1, gpuSrvHeap.incremented(startDesc, i).cpuHandle, srv.cpuHandle, + D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV); + } + + if (cbD->currentGraphicsPipeline) + cbD->cmdList->SetGraphicsRootDescriptorTable(rootParamIndex, startDesc.gpuHandle); + else if (cbD->currentComputePipeline) + cbD->cmdList->SetComputeRootDescriptorTable(rootParamIndex, startDesc.gpuHandle); + + rootParamIndex += 1; + } + } + for (int s = 0; s < 6; ++s) { + // Samplers are one parameter / descriptor table each, and the + // descriptor is from the shader visible sampler heap already. + for (const QD3D12Descriptor &samplerDescriptor : visitorData.samplers[s]) { + if (cbD->currentGraphicsPipeline) + cbD->cmdList->SetGraphicsRootDescriptorTable(rootParamIndex, samplerDescriptor.gpuHandle); + else if (cbD->currentComputePipeline) + cbD->cmdList->SetComputeRootDescriptorTable(rootParamIndex, samplerDescriptor.gpuHandle); + + rootParamIndex += 1; + } + } + for (int s = 0; s < 6; ++s) { + if (!visitorData.uavs[s].isEmpty()) { + QD3D12DescriptorHeap &gpuUavHeap(shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[currentFrameSlot]); + QD3D12Descriptor startDesc = gpuUavHeap.get(visitorData.uavs[s].count()); + for (int i = 0, count = visitorData.uavs[s].count(); i < count; ++i) { + const auto &uav(visitorData.uavs[s][i]); + if (QD3D12Resource *res = resourcePool.lookupRef(uav.first)) { + dev->CreateUnorderedAccessView(res->resource, nullptr, &uav.second, + gpuUavHeap.incremented(startDesc, i).cpuHandle); + } else { + dev->CreateUnorderedAccessView(nullptr, nullptr, nullptr, + gpuUavHeap.incremented(startDesc, i).cpuHandle); + } + } + + if (cbD->currentGraphicsPipeline) + cbD->cmdList->SetGraphicsRootDescriptorTable(rootParamIndex, startDesc.gpuHandle); + else if (cbD->currentComputePipeline) + cbD->cmdList->SetComputeRootDescriptorTable(rootParamIndex, startDesc.gpuHandle); + + rootParamIndex += 1; + } + } + + if (gfxPsD) { + cbD->currentGraphicsSrb = srb; + cbD->currentComputeSrb = nullptr; + } else { + cbD->currentGraphicsSrb = nullptr; + cbD->currentComputeSrb = srb; + } + cbD->currentSrbGeneration = srbD->generation; + } +} + +void QRhiD3D12::setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, QRhiCommandBuffer::IndexFormat indexFormat) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + + bool needsBindVBuf = false; + for (int i = 0; i < bindingCount; ++i) { + const int inputSlot = startBinding + i; + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, bindings[i].first); + Q_ASSERT(bufD->m_usage.testFlag(QRhiBuffer::VertexBuffer)); + const bool isDynamic = bufD->m_type == QRhiBuffer::Dynamic; + if (isDynamic) + bufD->executeHostWritesForFrameSlot(currentFrameSlot); + + if (cbD->currentVertexBuffers[inputSlot] != bufD->handles[isDynamic ? currentFrameSlot : 0] + || cbD->currentVertexOffsets[inputSlot] != bindings[i].second) + { + needsBindVBuf = true; + cbD->currentVertexBuffers[inputSlot] = bufD->handles[isDynamic ? currentFrameSlot : 0]; + cbD->currentVertexOffsets[inputSlot] = bindings[i].second; + } + } + + if (needsBindVBuf) { + QVarLengthArray<D3D12_VERTEX_BUFFER_VIEW, 4> vbv; + vbv.reserve(bindingCount); + + QD3D12GraphicsPipeline *psD = cbD->currentGraphicsPipeline; + const QRhiVertexInputLayout &inputLayout(psD->m_vertexInputLayout); + const int inputBindingCount = inputLayout.cendBindings() - inputLayout.cbeginBindings(); + + for (int i = 0, ie = qMin(bindingCount, inputBindingCount); i != ie; ++i) { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, bindings[i].first); + const QD3D12ObjectHandle handle = bufD->handles[bufD->m_type == QRhiBuffer::Dynamic ? currentFrameSlot : 0]; + const quint32 offset = bindings[i].second; + const quint32 stride = inputLayout.bindingAt(i)->stride(); + + if (bufD->m_type != QRhiBuffer::Dynamic) { + barrierGen.addTransitionBarrier(handle, D3D12_RESOURCE_STATE_VERTEX_AND_CONSTANT_BUFFER); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + + if (QD3D12Resource *res = resourcePool.lookupRef(handle)) { + vbv.append({ + res->resource->GetGPUVirtualAddress() + offset, + UINT(res->desc.Width - offset), + stride + }); + } + } + + cbD->cmdList->IASetVertexBuffers(UINT(startBinding), vbv.count(), vbv.constData()); + } + + if (indexBuf) { + QD3D12Buffer *ibufD = QRHI_RES(QD3D12Buffer, indexBuf); + Q_ASSERT(ibufD->m_usage.testFlag(QRhiBuffer::IndexBuffer)); + const bool isDynamic = ibufD->m_type == QRhiBuffer::Dynamic; + if (isDynamic) + ibufD->executeHostWritesForFrameSlot(currentFrameSlot); + + const DXGI_FORMAT dxgiFormat = indexFormat == QRhiCommandBuffer::IndexUInt16 ? DXGI_FORMAT_R16_UINT + : DXGI_FORMAT_R32_UINT; + if (cbD->currentIndexBuffer != ibufD->handles[isDynamic ? currentFrameSlot : 0] + || cbD->currentIndexOffset != indexOffset + || cbD->currentIndexFormat != dxgiFormat) + { + cbD->currentIndexBuffer = ibufD->handles[isDynamic ? currentFrameSlot : 0]; + cbD->currentIndexOffset = indexOffset; + cbD->currentIndexFormat = dxgiFormat; + + if (ibufD->m_type != QRhiBuffer::Dynamic) { + barrierGen.addTransitionBarrier(cbD->currentIndexBuffer, D3D12_RESOURCE_STATE_INDEX_BUFFER); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + + if (QD3D12Resource *res = resourcePool.lookupRef(cbD->currentIndexBuffer)) { + const D3D12_INDEX_BUFFER_VIEW ibv = { + res->resource->GetGPUVirtualAddress() + indexOffset, + UINT(res->desc.Width - indexOffset), + dxgiFormat + }; + cbD->cmdList->IASetIndexBuffer(&ibv); + } + } + } +} + +void QRhiD3D12::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + Q_ASSERT(cbD->currentTarget); + const QSize outputSize = cbD->currentTarget->pixelSize(); + + // D3D expects top-left, QRhiViewport is bottom-left + float x, y, w, h; + if (!qrhi_toTopLeftRenderTargetRect<UnBounded>(outputSize, viewport.viewport(), &x, &y, &w, &h)) + return; + + D3D12_VIEWPORT v; + v.TopLeftX = x; + v.TopLeftY = y; + v.Width = w; + v.Height = h; + v.MinDepth = viewport.minDepth(); + v.MaxDepth = viewport.maxDepth(); + cbD->cmdList->RSSetViewports(1, &v); + + if (cbD->currentGraphicsPipeline + && !cbD->currentGraphicsPipeline->flags().testFlag(QRhiGraphicsPipeline::UsesScissor)) + { + qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, viewport.viewport(), &x, &y, &w, &h); + D3D12_RECT r; + r.left = x; + r.top = y; + // right and bottom are exclusive + r.right = x + w; + r.bottom = y + h; + cbD->cmdList->RSSetScissorRects(1, &r); + } +} + +void QRhiD3D12::setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + Q_ASSERT(cbD->currentTarget); + const QSize outputSize = cbD->currentTarget->pixelSize(); + + // D3D expects top-left, QRhiScissor is bottom-left + int x, y, w, h; + if (!qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, scissor.scissor(), &x, &y, &w, &h)) + return; + + D3D12_RECT r; + r.left = x; + r.top = y; + // right and bottom are exclusive + r.right = x + w; + r.bottom = y + h; + cbD->cmdList->RSSetScissorRects(1, &r); +} + +void QRhiD3D12::setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + float v[4] = { c.redF(), c.greenF(), c.blueF(), c.alphaF() }; + cbD->cmdList->OMSetBlendFactor(v); +} + +void QRhiD3D12::setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + cbD->cmdList->OMSetStencilRef(refValue); +} + +void QRhiD3D12::draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + cbD->cmdList->DrawInstanced(vertexCount, instanceCount, firstVertex, firstInstance); +} + +void QRhiD3D12::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, qint32 vertexOffset, quint32 firstInstance) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + cbD->cmdList->DrawIndexedInstanced(indexCount, instanceCount, + firstIndex, vertexOffset, + firstInstance); +} + +void QRhiD3D12::debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) +{ + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXBeginEvent(cbD->cmdList, PIX_COLOR_DEFAULT, reinterpret_cast<LPCWSTR>(QString::fromLatin1(name).utf16())); +#else + Q_UNUSED(cbD); + Q_UNUSED(name); +#endif +} + +void QRhiD3D12::debugMarkEnd(QRhiCommandBuffer *cb) +{ + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXEndEvent(cbD->cmdList); +#else + Q_UNUSED(cbD); +#endif +} + +void QRhiD3D12::debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) +{ + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXSetMarker(cbD->cmdList, PIX_COLOR_DEFAULT, reinterpret_cast<LPCWSTR>(QString::fromLatin1(msg).utf16())); +#else + Q_UNUSED(cbD); + Q_UNUSED(msg); +#endif +} + +const QRhiNativeHandles *QRhiD3D12::nativeHandles(QRhiCommandBuffer *cb) +{ + return QRHI_RES(QD3D12CommandBuffer, cb)->nativeHandles(); +} + +void QRhiD3D12::beginExternal(QRhiCommandBuffer *cb) +{ + Q_UNUSED(cb); +} + +void QRhiD3D12::endExternal(QRhiCommandBuffer *cb) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + cbD->resetPerPassState(); + bindShaderVisibleHeaps(cbD); + if (cbD->currentTarget) { // could be compute, no rendertarget then + QD3D12RenderTargetData *rtD = rtData(cbD->currentTarget); + cbD->cmdList->OMSetRenderTargets(UINT(rtD->colorAttCount), + rtD->rtv, + TRUE, + rtD->dsAttCount ? &rtD->dsv : nullptr); + } +} + +double QRhiD3D12::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + return cbD->lastGpuTime; +} + +static void calculateGpuTime(QD3D12CommandBuffer *cbD, + int timestampPairStartIndex, + const quint8 *readbackBufPtr, + quint64 timestampTicksPerSecond) +{ + const size_t byteOffset = timestampPairStartIndex * sizeof(quint64); + const quint64 *p = reinterpret_cast<const quint64 *>(readbackBufPtr + byteOffset); + const quint64 startTime = *p++; + const quint64 endTime = *p; + if (startTime < endTime) { + const quint64 ticks = endTime - startTime; + const double timeSec = ticks / double(timestampTicksPerSecond); + cbD->lastGpuTime = timeSec; + } +} + +QRhi::FrameOpResult QRhiD3D12::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) +{ + Q_UNUSED(flags); + + QD3D12SwapChain *swapChainD = QRHI_RES(QD3D12SwapChain, swapChain); + currentSwapChain = swapChainD; + currentFrameSlot = swapChainD->currentFrameSlot; + QD3D12SwapChain::FrameResources &fr(swapChainD->frameRes[currentFrameSlot]); + + // We could do smarter things but mirror the Vulkan backend for now: Make + // sure the previous commands for this same frame slot have finished. Do + // this also for any other swapchain's commands with the same frame slot. + // While this reduces concurrency in render-to-swapchain-A, + // render-to-swapchain-B, repeat kind of scenarios, it keeps resource usage + // safe: swapchain A starting its frame 0, followed by swapchain B starting + // its own frame 0 will make B wait for A's frame 0 commands. If a resource + // is written in B's frame or when B checks for pending resource releases, + // that won't mess up A's in-flight commands (as they are guaranteed not to + // be in flight anymore). With Qt Quick this situation cannot happen anyway + // by design (one QRhi per window). + for (QD3D12SwapChain *sc : std::as_const(swapchains)) + sc->waitCommandCompletionForFrameSlot(currentFrameSlot); // note: swapChainD->currentFrameSlot, not sc's + + HRESULT hr = cmdAllocators[currentFrameSlot]->Reset(); + if (FAILED(hr)) { + qWarning("Failed to reset command allocator: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + if (!startCommandListForCurrentFrameSlot(&fr.cmdList)) + return QRhi::FrameOpError; + + QD3D12CommandBuffer *cbD = &swapChainD->cbWrapper; + cbD->cmdList = fr.cmdList; + + swapChainD->rtWrapper.d.rtv[0] = swapChainD->sampleDesc.Count > 1 + ? swapChainD->msaaRtvs[swapChainD->currentBackBufferIndex].cpuHandle + : swapChainD->rtvs[swapChainD->currentBackBufferIndex].cpuHandle; + + swapChainD->rtWrapper.d.dsv = swapChainD->ds ? swapChainD->ds->dsv.cpuHandle + : D3D12_CPU_DESCRIPTOR_HANDLE { 0 }; + + if (swapChainD->stereo) { + swapChainD->rtWrapperRight.d.rtv[0] = swapChainD->sampleDesc.Count > 1 + ? swapChainD->msaaRtvs[swapChainD->currentBackBufferIndex].cpuHandle + : swapChainD->rtvsRight[swapChainD->currentBackBufferIndex].cpuHandle; + + swapChainD->rtWrapperRight.d.dsv = + swapChainD->ds ? swapChainD->ds->dsv.cpuHandle : D3D12_CPU_DESCRIPTOR_HANDLE{ 0 }; + } + + + // Time to release things that are marked for currentFrameSlot since due to + // the wait above we know that the previous commands on the GPU for this + // slot must have finished already. + releaseQueue.executeDeferredReleases(currentFrameSlot); + + // Full reset of the command buffer data. + cbD->resetState(); + + // Move the head back to zero for the per-frame shader-visible descriptor heap work areas. + shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[currentFrameSlot].head = 0; + // Same for the small staging area. + smallStagingAreas[currentFrameSlot].head = 0; + + bindShaderVisibleHeaps(cbD); + + finishActiveReadbacks(); // last, in case the readback-completed callback issues rhi calls + + if (timestampQueryHeap.isValid() && timestampTicksPerSecond) { + // Read the timestamps for the previous frame for this slot. (the + // ResolveQuery() should have completed by now due to the wait above) + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + calculateGpuTime(cbD, + timestampPairStartIndex, + timestampReadbackArea.mem.p, + timestampTicksPerSecond); + // Write the start timestamp for this frame for this slot. + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex); + } + + return QRhi::FrameOpSuccess; +} + +QRhi::FrameOpResult QRhiD3D12::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) +{ + QD3D12SwapChain *swapChainD = QRHI_RES(QD3D12SwapChain, swapChain); + Q_ASSERT(currentSwapChain == swapChainD); + QD3D12CommandBuffer *cbD = &swapChainD->cbWrapper; + + QD3D12ObjectHandle backBufferResourceHandle = swapChainD->colorBuffers[swapChainD->currentBackBufferIndex]; + if (swapChainD->sampleDesc.Count > 1) { + QD3D12ObjectHandle msaaBackBufferResourceHandle = swapChainD->msaaBuffers[swapChainD->currentBackBufferIndex]; + barrierGen.addTransitionBarrier(msaaBackBufferResourceHandle, D3D12_RESOURCE_STATE_RESOLVE_SOURCE); + barrierGen.addTransitionBarrier(backBufferResourceHandle, D3D12_RESOURCE_STATE_RESOLVE_DEST); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + const QD3D12Resource *src = resourcePool.lookupRef(msaaBackBufferResourceHandle); + const QD3D12Resource *dst = resourcePool.lookupRef(backBufferResourceHandle); + if (src && dst) + cbD->cmdList->ResolveSubresource(dst->resource, 0, src->resource, 0, swapChainD->colorFormat); + } + + barrierGen.addTransitionBarrier(backBufferResourceHandle, D3D12_RESOURCE_STATE_PRESENT); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + + if (timestampQueryHeap.isValid()) { + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex + 1); + cbD->cmdList->ResolveQueryData(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex, + 2, + timestampReadbackArea.mem.buffer, + timestampPairStartIndex * sizeof(quint64)); + } + + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; + HRESULT hr = cmdList->Close(); + if (FAILED(hr)) { + qWarning("Failed to close command list: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + ID3D12CommandList *execList[] = { cmdList }; + cmdQueue->ExecuteCommandLists(1, execList); + + if (!flags.testFlag(QRhi::SkipPresent)) { + UINT presentFlags = 0; + if (swapChainD->swapInterval == 0 + && (swapChainD->swapChainFlags & DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING)) + { + presentFlags |= DXGI_PRESENT_ALLOW_TEARING; + } + if (!swapChainD->swapChain) { + qWarning("Failed to present, no swapchain"); + return QRhi::FrameOpError; + } + HRESULT hr = swapChainD->swapChain->Present(swapChainD->swapInterval, presentFlags); + if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { + qWarning("Device loss detected in Present()"); + deviceLost = true; + return QRhi::FrameOpDeviceLost; + } else if (FAILED(hr)) { + qWarning("Failed to present: %s", qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + if (dcompDevice && swapChainD->dcompTarget && swapChainD->dcompVisual) + dcompDevice->Commit(); + } + + swapChainD->addCommandCompletionSignalForCurrentFrameSlot(); + + // NB! The deferred-release mechanism here differs from the older QRhi + // backends. There is no lastActiveFrameSlot tracking. Instead, + // currentFrameSlot is written to the registered entries now, and so the + // resources will get released in the frames_in_flight'th beginFrame() + // counting starting from now. + releaseQueue.activatePendingDeferredReleaseRequests(currentFrameSlot); + + if (!flags.testFlag(QRhi::SkipPresent)) { + // Only move to the next slot if we presented. Otherwise will block and + // wait for completion in the next beginFrame already, but SkipPresent + // should be infrequent anyway. + swapChainD->currentFrameSlot = (swapChainD->currentFrameSlot + 1) % QD3D12_FRAMES_IN_FLIGHT; + swapChainD->currentBackBufferIndex = swapChainD->swapChain->GetCurrentBackBufferIndex(); + } + + currentSwapChain = nullptr; + return QRhi::FrameOpSuccess; +} + +QRhi::FrameOpResult QRhiD3D12::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) +{ + Q_UNUSED(flags); + + // Switch to the next slot manually. Swapchains do not know about this + // which is good. So for example an onscreen, onscreen, offscreen, + // onscreen, onscreen, onscreen sequence of frames leads to 0, 1, 0, 0, 1, + // 0. (no strict alternation anymore) But this is not different from what + // happens when multiple swapchains are involved. Offscreen frames are + // synchronous anyway in the sense that they wait for execution to complete + // in endOffscreenFrame, so no resources used in that frame are busy + // anymore in the next frame. + + currentFrameSlot = (currentFrameSlot + 1) % QD3D12_FRAMES_IN_FLIGHT; + + for (QD3D12SwapChain *sc : std::as_const(swapchains)) + sc->waitCommandCompletionForFrameSlot(currentFrameSlot); // note: not sc's currentFrameSlot + + if (!offscreenCb[currentFrameSlot]) + offscreenCb[currentFrameSlot] = new QD3D12CommandBuffer(this); + QD3D12CommandBuffer *cbD = offscreenCb[currentFrameSlot]; + if (!startCommandListForCurrentFrameSlot(&cbD->cmdList)) + return QRhi::FrameOpError; + + releaseQueue.executeDeferredReleases(currentFrameSlot); + cbD->resetState(); + shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[currentFrameSlot].head = 0; + smallStagingAreas[currentFrameSlot].head = 0; + + bindShaderVisibleHeaps(cbD); + + if (timestampQueryHeap.isValid() && timestampTicksPerSecond) { + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT); + } + + offscreenActive = true; + *cb = cbD; + + return QRhi::FrameOpSuccess; +} + +QRhi::FrameOpResult QRhiD3D12::endOffscreenFrame(QRhi::EndFrameFlags flags) +{ + Q_UNUSED(flags); + Q_ASSERT(offscreenActive); + offscreenActive = false; + + QD3D12CommandBuffer *cbD = offscreenCb[currentFrameSlot]; + if (timestampQueryHeap.isValid()) { + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex + 1); + cbD->cmdList->ResolveQueryData(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex, + 2, + timestampReadbackArea.mem.buffer, + timestampPairStartIndex * sizeof(quint64)); + } + + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; + HRESULT hr = cmdList->Close(); + if (FAILED(hr)) { + qWarning("Failed to close command list: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + ID3D12CommandList *execList[] = { cmdList }; + cmdQueue->ExecuteCommandLists(1, execList); + + releaseQueue.activatePendingDeferredReleaseRequests(currentFrameSlot); + + // wait for completion + waitGpu(); + + // Here we know that executing the host-side reads for this (or any + // previous) frame is safe since we waited for completion above. + finishActiveReadbacks(true); + + // the timestamp query results should be available too, given the wait + if (timestampQueryHeap.isValid()) { + calculateGpuTime(cbD, + currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT, + timestampReadbackArea.mem.p, + timestampTicksPerSecond); + } + + return QRhi::FrameOpSuccess; +} + +QRhi::FrameOpResult QRhiD3D12::finish() +{ + if (!inFrame) + return QRhi::FrameOpSuccess; + + QD3D12CommandBuffer *cbD = nullptr; + if (offscreenActive) { + Q_ASSERT(!currentSwapChain); + cbD = offscreenCb[currentFrameSlot]; + } else { + Q_ASSERT(currentSwapChain); + cbD = ¤tSwapChain->cbWrapper; + } + if (!cbD) + return QRhi::FrameOpError; + + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::NoPass); + + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; + HRESULT hr = cmdList->Close(); + if (FAILED(hr)) { + qWarning("Failed to close command list: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + ID3D12CommandList *execList[] = { cmdList }; + cmdQueue->ExecuteCommandLists(1, execList); + + releaseQueue.activatePendingDeferredReleaseRequests(currentFrameSlot); + + // full blocking wait for everything, frame slots do not matter now + waitGpu(); + + hr = cmdAllocators[currentFrameSlot]->Reset(); + if (FAILED(hr)) { + qWarning("Failed to reset command allocator: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + + if (!startCommandListForCurrentFrameSlot(&cmdList)) + return QRhi::FrameOpError; + + cbD->resetState(); + + shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[currentFrameSlot].head = 0; + smallStagingAreas[currentFrameSlot].head = 0; + + bindShaderVisibleHeaps(cbD); + + releaseQueue.executeDeferredReleases(currentFrameSlot); + + finishActiveReadbacks(true); + + return QRhi::FrameOpSuccess; +} + +void QRhiD3D12::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::NoPass); + enqueueResourceUpdates(cbD, resourceUpdates); +} + +void QRhiD3D12::beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::NoPass); + + if (resourceUpdates) + enqueueResourceUpdates(cbD, resourceUpdates); + + QD3D12RenderTargetData *rtD = rtData(rt); + bool wantsColorClear = true; + bool wantsDsClear = true; + if (rt->resourceType() == QRhiRenderTarget::TextureRenderTarget) { + QD3D12TextureRenderTarget *rtTex = QRHI_RES(QD3D12TextureRenderTarget, rt); + wantsColorClear = !rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents); + wantsDsClear = !rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents); + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QD3D12Texture, QD3D12RenderBuffer>(rtTex->description(), rtD->currentResIdList)) + rtTex->create(); + + for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); it != itEnd; ++it) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, it->texture()); + QD3D12Texture *resolveTexD = QRHI_RES(QD3D12Texture, it->resolveTexture()); + QD3D12RenderBuffer *rbD = QRHI_RES(QD3D12RenderBuffer, it->renderBuffer()); + if (texD) + barrierGen.addTransitionBarrier(texD->handle, D3D12_RESOURCE_STATE_RENDER_TARGET); + else if (rbD) + barrierGen.addTransitionBarrier(rbD->handle, D3D12_RESOURCE_STATE_RENDER_TARGET); + if (resolveTexD) + barrierGen.addTransitionBarrier(resolveTexD->handle, D3D12_RESOURCE_STATE_RENDER_TARGET); + } + if (rtTex->m_desc.depthStencilBuffer()) { + QD3D12RenderBuffer *rbD = QRHI_RES(QD3D12RenderBuffer, rtTex->m_desc.depthStencilBuffer()); + Q_ASSERT(rbD->m_type == QRhiRenderBuffer::DepthStencil); + barrierGen.addTransitionBarrier(rbD->handle, D3D12_RESOURCE_STATE_DEPTH_WRITE); + } else if (rtTex->m_desc.depthTexture()) { + QD3D12Texture *depthTexD = QRHI_RES(QD3D12Texture, rtTex->m_desc.depthTexture()); + barrierGen.addTransitionBarrier(depthTexD->handle, D3D12_RESOURCE_STATE_DEPTH_WRITE); + } + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } else { + Q_ASSERT(currentSwapChain); + barrierGen.addTransitionBarrier(currentSwapChain->sampleDesc.Count > 1 + ? currentSwapChain->msaaBuffers[currentSwapChain->currentBackBufferIndex] + : currentSwapChain->colorBuffers[currentSwapChain->currentBackBufferIndex], + D3D12_RESOURCE_STATE_RENDER_TARGET); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + } + + cbD->cmdList->OMSetRenderTargets(UINT(rtD->colorAttCount), + rtD->rtv, + TRUE, + rtD->dsAttCount ? &rtD->dsv : nullptr); + + if (rtD->colorAttCount && wantsColorClear) { + float clearColor[4] = { + colorClearValue.redF(), + colorClearValue.greenF(), + colorClearValue.blueF(), + colorClearValue.alphaF() + }; + for (int i = 0; i < rtD->colorAttCount; ++i) + cbD->cmdList->ClearRenderTargetView(rtD->rtv[i], clearColor, 0, nullptr); + } + if (rtD->dsAttCount && wantsDsClear) { + cbD->cmdList->ClearDepthStencilView(rtD->dsv, + D3D12_CLEAR_FLAGS(D3D12_CLEAR_FLAG_DEPTH | D3D12_CLEAR_FLAG_STENCIL), + depthStencilClearValue.depthClearValue(), + UINT8(depthStencilClearValue.stencilClearValue()), + 0, + nullptr); + } + + cbD->recordingPass = QD3D12CommandBuffer::RenderPass; + cbD->currentTarget = rt; + + cbD->resetPerPassState(); +} + +void QRhiD3D12::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::RenderPass); + + if (cbD->currentTarget->resourceType() == QRhiResource::TextureRenderTarget) { + QD3D12TextureRenderTarget *rtTex = QRHI_RES(QD3D12TextureRenderTarget, cbD->currentTarget); + for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); + it != itEnd; ++it) + { + const QRhiColorAttachment &colorAtt(*it); + if (!colorAtt.resolveTexture()) + continue; + + QD3D12Texture *dstTexD = QRHI_RES(QD3D12Texture, colorAtt.resolveTexture()); + QD3D12Resource *dstRes = resourcePool.lookupRef(dstTexD->handle); + if (!dstRes) + continue; + + QD3D12Texture *srcTexD = QRHI_RES(QD3D12Texture, colorAtt.texture()); + QD3D12RenderBuffer *srcRbD = QRHI_RES(QD3D12RenderBuffer, colorAtt.renderBuffer()); + Q_ASSERT(srcTexD || srcRbD); + QD3D12Resource *srcRes = resourcePool.lookupRef(srcTexD ? srcTexD->handle : srcRbD->handle); + if (!srcRes) + continue; + + if (srcTexD) { + if (srcTexD->dxgiFormat != dstTexD->dxgiFormat) { + qWarning("Resolve source (%d) and destination (%d) formats do not match", + int(srcTexD->dxgiFormat), int(dstTexD->dxgiFormat)); + continue; + } + if (srcTexD->sampleDesc.Count <= 1) { + qWarning("Cannot resolve a non-multisample texture"); + continue; + } + if (srcTexD->m_pixelSize != dstTexD->m_pixelSize) { + qWarning("Resolve source and destination sizes do not match"); + continue; + } + } else { + if (srcRbD->dxgiFormat != dstTexD->dxgiFormat) { + qWarning("Resolve source (%d) and destination (%d) formats do not match", + int(srcRbD->dxgiFormat), int(dstTexD->dxgiFormat)); + continue; + } + if (srcRbD->m_pixelSize != dstTexD->m_pixelSize) { + qWarning("Resolve source and destination sizes do not match"); + continue; + } + } + + barrierGen.addTransitionBarrier(srcTexD ? srcTexD->handle : srcRbD->handle, D3D12_RESOURCE_STATE_RESOLVE_SOURCE); + barrierGen.addTransitionBarrier(dstTexD->handle, D3D12_RESOURCE_STATE_RESOLVE_DEST); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + + const UINT resolveCount = colorAtt.multiViewCount() >= 2 ? colorAtt.multiViewCount() : 1; + for (UINT resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + const UINT srcSubresource = calcSubresource(0, UINT(colorAtt.layer()) + resolveIdx, 1); + const UINT dstSubresource = calcSubresource(UINT(colorAtt.resolveLevel()), + UINT(colorAtt.resolveLayer()) + resolveIdx, + dstTexD->mipLevelCount); + cbD->cmdList->ResolveSubresource(dstRes->resource, dstSubresource, + srcRes->resource, srcSubresource, + dstTexD->dxgiFormat); + } + } + if (rtTex->m_desc.depthResolveTexture()) + qWarning("Resolving multisample depth-stencil buffers is not supported with D3D"); + } + + cbD->recordingPass = QD3D12CommandBuffer::NoPass; + cbD->currentTarget = nullptr; + + if (resourceUpdates) + enqueueResourceUpdates(cbD, resourceUpdates); +} + +void QRhiD3D12::beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::NoPass); + + if (resourceUpdates) + enqueueResourceUpdates(cbD, resourceUpdates); + + cbD->recordingPass = QD3D12CommandBuffer::ComputePass; + + cbD->resetPerPassState(); +} + +void QRhiD3D12::endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::ComputePass); + + cbD->recordingPass = QD3D12CommandBuffer::NoPass; + + if (resourceUpdates) + enqueueResourceUpdates(cbD, resourceUpdates); +} + +void QRhiD3D12::setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::ComputePass); + QD3D12ComputePipeline *psD = QRHI_RES(QD3D12ComputePipeline, ps); + const bool pipelineChanged = cbD->currentComputePipeline != psD || cbD->currentPipelineGeneration != psD->generation; + + if (pipelineChanged) { + cbD->currentGraphicsPipeline = nullptr; + cbD->currentComputePipeline = psD; + cbD->currentPipelineGeneration = psD->generation; + + if (QD3D12Pipeline *pipeline = pipelinePool.lookupRef(psD->handle)) { + Q_ASSERT(pipeline->type == QD3D12Pipeline::Compute); + cbD->cmdList->SetPipelineState(pipeline->pso); + if (QD3D12RootSignature *rs = rootSignaturePool.lookupRef(psD->rootSigHandle)) + cbD->cmdList->SetComputeRootSignature(rs->rootSig); + } + } +} + +void QRhiD3D12::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::ComputePass); + cbD->cmdList->Dispatch(UINT(x), UINT(y), UINT(z)); +} + +bool QD3D12DescriptorHeap::create(ID3D12Device *device, + quint32 descriptorCount, + D3D12_DESCRIPTOR_HEAP_TYPE heapType, + D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags) +{ + head = 0; + capacity = descriptorCount; + this->heapType = heapType; + this->heapFlags = heapFlags; + + D3D12_DESCRIPTOR_HEAP_DESC heapDesc = {}; + heapDesc.Type = heapType; + heapDesc.NumDescriptors = capacity; + heapDesc.Flags = D3D12_DESCRIPTOR_HEAP_FLAGS(heapFlags); + + HRESULT hr = device->CreateDescriptorHeap(&heapDesc, __uuidof(ID3D12DescriptorHeap), reinterpret_cast<void **>(&heap)); + if (FAILED(hr)) { + qWarning("Failed to create descriptor heap: %s", qPrintable(QSystemError::windowsComString(hr))); + heap = nullptr; + capacity = descriptorByteSize = 0; + return false; + } + + descriptorByteSize = device->GetDescriptorHandleIncrementSize(heapType); + heapStart.cpuHandle = heap->GetCPUDescriptorHandleForHeapStart(); + if (heapFlags & D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE) + heapStart.gpuHandle = heap->GetGPUDescriptorHandleForHeapStart(); + + return true; +} + +void QD3D12DescriptorHeap::createWithExisting(const QD3D12DescriptorHeap &other, + quint32 offsetInDescriptors, + quint32 descriptorCount) +{ + heap = nullptr; + head = 0; + capacity = descriptorCount; + heapType = other.heapType; + heapFlags = other.heapFlags; + descriptorByteSize = other.descriptorByteSize; + heapStart = incremented(other.heapStart, offsetInDescriptors); +} + +void QD3D12DescriptorHeap::destroy() +{ + if (heap) { + heap->Release(); + heap = nullptr; + } + capacity = 0; +} + +void QD3D12DescriptorHeap::destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue) +{ + if (heap) { + releaseQueue->deferredReleaseDescriptorHeap(heap); + heap = nullptr; + } + capacity = 0; +} + +QD3D12Descriptor QD3D12DescriptorHeap::get(quint32 count) +{ + Q_ASSERT(count > 0); + if (head + count > capacity) { + qWarning("Cannot get %u descriptors as that would exceed capacity %u", count, capacity); + return {}; + } + head += count; + return at(head - count); +} + +QD3D12Descriptor QD3D12DescriptorHeap::at(quint32 index) const +{ + const quint32 startOffset = index * descriptorByteSize; + QD3D12Descriptor result; + result.cpuHandle.ptr = heapStart.cpuHandle.ptr + startOffset; + if (heapStart.gpuHandle.ptr != 0) + result.gpuHandle.ptr = heapStart.gpuHandle.ptr + startOffset; + return result; +} + +bool QD3D12CpuDescriptorPool::create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE heapType, const char *debugName) +{ + QD3D12DescriptorHeap firstHeap; + if (!firstHeap.create(device, DESCRIPTORS_PER_HEAP, heapType, D3D12_DESCRIPTOR_HEAP_FLAG_NONE)) + return false; + heaps.append(HeapWithMap::init(firstHeap, DESCRIPTORS_PER_HEAP)); + descriptorByteSize = heaps[0].heap.descriptorByteSize; + this->device = device; + this->debugName = debugName; + return true; +} + +void QD3D12CpuDescriptorPool::destroy() +{ +#ifndef QT_NO_DEBUG + // debug builds: just do it always + static bool leakCheck = true; +#else + // release builds: opt-in + static bool leakCheck = qEnvironmentVariableIntValue("QT_RHI_LEAK_CHECK"); +#endif + if (leakCheck) { + for (HeapWithMap &heap : heaps) { + const int leakedDescriptorCount = heap.map.count(true); + if (leakedDescriptorCount > 0) { + qWarning("QD3D12CpuDescriptorPool::destroy(): " + "Heap %p for descriptor pool %p '%s' has %d unreleased descriptors", + &heap.heap, this, debugName, leakedDescriptorCount); + } + } + } + for (HeapWithMap &heap : heaps) + heap.heap.destroy(); + heaps.clear(); +} + +QD3D12Descriptor QD3D12CpuDescriptorPool::allocate(quint32 count) +{ + Q_ASSERT(count > 0 && count <= DESCRIPTORS_PER_HEAP); + + HeapWithMap &last(heaps.last()); + if (last.heap.head + count <= last.heap.capacity) { + quint32 firstIndex = last.heap.head; + for (quint32 i = 0; i < count; ++i) + last.map.setBit(firstIndex + i); + return last.heap.get(count); + } + + for (HeapWithMap &heap : heaps) { + quint32 freeCount = 0; + for (quint32 i = 0; i < DESCRIPTORS_PER_HEAP; ++i) { + if (heap.map.testBit(i)) { + freeCount = 0; + } else { + freeCount += 1; + if (freeCount == count) { + quint32 firstIndex = i - (freeCount - 1); + for (quint32 j = 0; j < count; ++j) { + heap.map.setBit(firstIndex + j); + return heap.heap.at(firstIndex); + } + } + } + } + } + + QD3D12DescriptorHeap newHeap; + if (!newHeap.create(device, DESCRIPTORS_PER_HEAP, last.heap.heapType, last.heap.heapFlags)) + return {}; + + heaps.append(HeapWithMap::init(newHeap, DESCRIPTORS_PER_HEAP)); + + for (quint32 i = 0; i < count; ++i) + heaps.last().map.setBit(i); + + return heaps.last().heap.get(count); +} + +void QD3D12CpuDescriptorPool::release(const QD3D12Descriptor &descriptor, quint32 count) +{ + Q_ASSERT(count > 0 && count <= DESCRIPTORS_PER_HEAP); + if (!descriptor.isValid()) + return; + + const SIZE_T addr = descriptor.cpuHandle.ptr; + for (HeapWithMap &heap : heaps) { + const SIZE_T begin = heap.heap.heapStart.cpuHandle.ptr; + const SIZE_T end = begin + heap.heap.descriptorByteSize * heap.heap.capacity; + if (addr >= begin && addr < end) { + quint32 firstIndex = (addr - begin) / heap.heap.descriptorByteSize; + for (quint32 i = 0; i < count; ++i) + heap.map.setBit(firstIndex + i, false); + return; + } + } + + qWarning("QD3D12CpuDescriptorPool::release: Descriptor with address %llu is not in any heap", + quint64(descriptor.cpuHandle.ptr)); +} + +bool QD3D12QueryHeap::create(ID3D12Device *device, + quint32 queryCount, + D3D12_QUERY_HEAP_TYPE heapType) +{ + capacity = queryCount; + + D3D12_QUERY_HEAP_DESC heapDesc = {}; + heapDesc.Type = heapType; + heapDesc.Count = capacity; + + HRESULT hr = device->CreateQueryHeap(&heapDesc, __uuidof(ID3D12QueryHeap), reinterpret_cast<void **>(&heap)); + if (FAILED(hr)) { + qWarning("Failed to create query heap: %s", qPrintable(QSystemError::windowsComString(hr))); + heap = nullptr; + capacity = 0; + return false; + } + + return true; +} + +void QD3D12QueryHeap::destroy() +{ + if (heap) { + heap->Release(); + heap = nullptr; + } + capacity = 0; +} + +bool QD3D12StagingArea::create(QRhiD3D12 *rhi, quint32 capacity, D3D12_HEAP_TYPE heapType) +{ + Q_ASSERT(heapType == D3D12_HEAP_TYPE_UPLOAD || heapType == D3D12_HEAP_TYPE_READBACK); + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER; + resourceDesc.Width = capacity; + resourceDesc.Height = 1; + resourceDesc.DepthOrArraySize = 1; + resourceDesc.MipLevels = 1; + resourceDesc.Format = DXGI_FORMAT_UNKNOWN; + resourceDesc.SampleDesc = { 1, 0 }; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR; + resourceDesc.Flags = D3D12_RESOURCE_FLAG_NONE; + UINT state = heapType == D3D12_HEAP_TYPE_UPLOAD ? D3D12_RESOURCE_STATE_GENERIC_READ : D3D12_RESOURCE_STATE_COPY_DEST; + HRESULT hr = rhi->vma.createResource(heapType, + &resourceDesc, + D3D12_RESOURCE_STATES(state), + nullptr, + &allocation, + __uuidof(ID3D12Resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) { + qWarning("Failed to create buffer for staging area: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + void *p = nullptr; + hr = resource->Map(0, nullptr, &p); + if (FAILED(hr)) { + qWarning("Failed to map buffer for staging area: %s", + qPrintable(QSystemError::windowsComString(hr))); + destroy(); + return false; + } + + mem.p = static_cast<quint8 *>(p); + mem.gpuAddr = resource->GetGPUVirtualAddress(); + mem.buffer = resource; + mem.bufferOffset = 0; + + this->capacity = capacity; + head = 0; + + return true; +} + +void QD3D12StagingArea::destroy() +{ + if (resource) { + resource->Release(); + resource = nullptr; + } + if (allocation) { + allocation->Release(); + allocation = nullptr; + } + mem = {}; +} + +void QD3D12StagingArea::destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue) +{ + if (resource) + releaseQueue->deferredReleaseResourceAndAllocation(resource, allocation); + mem = {}; +} + +QD3D12StagingArea::Allocation QD3D12StagingArea::get(quint32 byteSize) +{ + const quint32 allocSize = aligned(byteSize, ALIGNMENT); + if (head + allocSize > capacity) { + qWarning("Failed to allocate %u (%u) bytes from staging area of size %u with %u bytes left", + allocSize, byteSize, capacity, remainingCapacity()); + return {}; + } + const quint32 offset = head; + head += allocSize; + return { + mem.p + offset, + mem.gpuAddr + offset, + mem.buffer, + offset + }; +} + +// Can be called inside and outside of begin-endFrame. Removes from the pool +// and releases the underlying native resource only in the frames_in_flight'th +// beginFrame() counted starting from the next endFrame(). +void QD3D12ReleaseQueue::deferredReleaseResource(const QD3D12ObjectHandle &handle) +{ + DeferredReleaseEntry e; + e.handle = handle; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseResourceWithViews(const QD3D12ObjectHandle &handle, + QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::Resource; + e.handle = handle; + e.poolForViews = pool; + e.viewsStart = viewsStart; + e.viewCount = viewCount; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleasePipeline(const QD3D12ObjectHandle &handle) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::Pipeline; + e.handle = handle; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseRootSignature(const QD3D12ObjectHandle &handle) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::RootSignature; + e.handle = handle; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseCallback(std::function<void(void*)> callback, void *userData) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::Callback; + e.callback = callback; + e.callbackUserData = userData; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseResourceAndAllocation(ID3D12Resource *resource, + D3D12MA::Allocation *allocation) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::ResourceAndAllocation; + e.resourceAndAllocation = { resource, allocation }; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseDescriptorHeap(ID3D12DescriptorHeap *heap) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::DescriptorHeap; + e.descriptorHeap = heap; + queue.append(e); +} + +void QD3D12ReleaseQueue::deferredReleaseViews(QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount) +{ + DeferredReleaseEntry e; + e.type = DeferredReleaseEntry::Views; + e.poolForViews = pool; + e.viewsStart = viewsStart; + e.viewCount = viewCount; + queue.append(e); +} + +void QD3D12ReleaseQueue::activatePendingDeferredReleaseRequests(int frameSlot) +{ + for (DeferredReleaseEntry &e : queue) { + if (!e.frameSlotToBeReleasedIn.has_value()) + e.frameSlotToBeReleasedIn = frameSlot; + } +} + +void QD3D12ReleaseQueue::executeDeferredReleases(int frameSlot, bool forced) +{ + for (int i = queue.count() - 1; i >= 0; --i) { + const DeferredReleaseEntry &e(queue[i]); + if (forced || (e.frameSlotToBeReleasedIn.has_value() && e.frameSlotToBeReleasedIn.value() == frameSlot)) { + switch (e.type) { + case DeferredReleaseEntry::Resource: + resourcePool->remove(e.handle); + if (e.poolForViews && e.viewsStart.isValid() && e.viewCount > 0) + e.poolForViews->release(e.viewsStart, e.viewCount); + break; + case DeferredReleaseEntry::Pipeline: + pipelinePool->remove(e.handle); + break; + case DeferredReleaseEntry::RootSignature: + rootSignaturePool->remove(e.handle); + break; + case DeferredReleaseEntry::Callback: + e.callback(e.callbackUserData); + break; + case DeferredReleaseEntry::ResourceAndAllocation: + // order matters: resource first, then the allocation (which + // may be null) + e.resourceAndAllocation.first->Release(); + if (e.resourceAndAllocation.second) + e.resourceAndAllocation.second->Release(); + break; + case DeferredReleaseEntry::DescriptorHeap: + e.descriptorHeap->Release(); + break; + case DeferredReleaseEntry::Views: + e.poolForViews->release(e.viewsStart, e.viewCount); + break; + } + queue.removeAt(i); + } + } +} + +void QD3D12ReleaseQueue::releaseAll() +{ + executeDeferredReleases(0, true); +} + +void QD3D12ResourceBarrierGenerator::addTransitionBarrier(const QD3D12ObjectHandle &resourceHandle, + D3D12_RESOURCE_STATES stateAfter) +{ + if (QD3D12Resource *res = resourcePool->lookupRef(resourceHandle)) { + if (stateAfter != res->state) { + transitionResourceBarriers.append({ resourceHandle, res->state, stateAfter }); + res->state = stateAfter; + } + } +} + +void QD3D12ResourceBarrierGenerator::enqueueBufferedTransitionBarriers(QD3D12CommandBuffer *cbD) +{ + QVarLengthArray<D3D12_RESOURCE_BARRIER, PREALLOC> barriers; + for (const TransitionResourceBarrier &trb : transitionResourceBarriers) { + if (QD3D12Resource *res = resourcePool->lookupRef(trb.resourceHandle)) { + D3D12_RESOURCE_BARRIER barrier = {}; + barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_TRANSITION; + barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE; + barrier.Transition.pResource = res->resource; + barrier.Transition.Subresource = D3D12_RESOURCE_BARRIER_ALL_SUBRESOURCES; + barrier.Transition.StateBefore = trb.stateBefore; + barrier.Transition.StateAfter = trb.stateAfter; + barriers.append(barrier); + } + } + transitionResourceBarriers.clear(); + if (!barriers.isEmpty()) + cbD->cmdList->ResourceBarrier(barriers.count(), barriers.constData()); +} + +void QD3D12ResourceBarrierGenerator::enqueueSubresourceTransitionBarrier(QD3D12CommandBuffer *cbD, + const QD3D12ObjectHandle &resourceHandle, + UINT subresource, + D3D12_RESOURCE_STATES stateBefore, + D3D12_RESOURCE_STATES stateAfter) +{ + if (QD3D12Resource *res = resourcePool->lookupRef(resourceHandle)) { + D3D12_RESOURCE_BARRIER barrier = {}; + barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_TRANSITION; + barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE; + barrier.Transition.pResource = res->resource; + barrier.Transition.Subresource = subresource; + barrier.Transition.StateBefore = stateBefore; + barrier.Transition.StateAfter = stateAfter; + cbD->cmdList->ResourceBarrier(1, &barrier); + } +} + +void QD3D12ResourceBarrierGenerator::enqueueUavBarrier(QD3D12CommandBuffer *cbD, + const QD3D12ObjectHandle &resourceHandle) +{ + if (QD3D12Resource *res = resourcePool->lookupRef(resourceHandle)) { + D3D12_RESOURCE_BARRIER barrier = {}; + barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_UAV; + barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE; + barrier.UAV.pResource = res->resource; + cbD->cmdList->ResourceBarrier(1, &barrier); + } +} + +void QD3D12ShaderBytecodeCache::insertWithCapacityLimit(const QRhiShaderStage &key, const Shader &s) +{ + if (data.count() >= QRhiD3D12::MAX_SHADER_CACHE_ENTRIES) + data.clear(); + data.insert(key, s); +} + +bool QD3D12ShaderVisibleDescriptorHeap::create(ID3D12Device *device, + D3D12_DESCRIPTOR_HEAP_TYPE type, + quint32 perFrameDescriptorCount) +{ + Q_ASSERT(type == D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV || type == D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER); + + quint32 size = perFrameDescriptorCount * QD3D12_FRAMES_IN_FLIGHT; + + // https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-support + const quint32 CBV_SRV_UAV_MAX = 1000000; + const quint32 SAMPLER_MAX = 2048; + if (type == D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV) + size = qMin(size, CBV_SRV_UAV_MAX); + else if (type == D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER) + size = qMin(size, SAMPLER_MAX); + + if (!heap.create(device, size, type, D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE)) { + qWarning("Failed to create shader-visible descriptor heap of size %u", size); + return false; + } + + perFrameDescriptorCount = size / QD3D12_FRAMES_IN_FLIGHT; + quint32 currentOffsetInDescriptors = 0; + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + perFrameHeapSlice[i].createWithExisting(heap, currentOffsetInDescriptors, perFrameDescriptorCount); + currentOffsetInDescriptors += perFrameDescriptorCount; + } + + return true; +} + +void QD3D12ShaderVisibleDescriptorHeap::destroy() +{ + heap.destroy(); +} + +void QD3D12ShaderVisibleDescriptorHeap::destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue) +{ + heap.destroyWithDeferredRelease(releaseQueue); +} + +static inline QPair<int, int> mapBinding(int binding, const QShader::NativeResourceBindingMap &map) +{ + if (map.isEmpty()) + return { binding, binding }; // assume 1:1 mapping + + auto it = map.constFind(binding); + if (it != map.cend()) + return *it; + + // Hitting this path is normal too. It is not given that the resource is + // present in the shaders for all the stages specified by the visibility + // mask in the QRhiShaderResourceBinding. + return { -1, -1 }; +} + +void QD3D12ShaderResourceVisitor::visit() +{ + for (int bindingIdx = 0, bindingCount = srb->m_bindings.count(); bindingIdx != bindingCount; ++bindingIdx) { + const QRhiShaderResourceBinding &b(srb->m_bindings[bindingIdx]); + const QRhiShaderResourceBinding::Data *bd = QRhiImplementation::shaderResourceBindingData(b); + + for (int stageIdx = 0; stageIdx < stageCount; ++stageIdx) { + const QD3D12ShaderStageData *sd = &stageData[stageIdx]; + if (!sd->valid) + continue; + + if (!bd->stage.testFlag(qd3d12_stageToSrb(sd->stage))) + continue; + + switch (bd->type) { + case QRhiShaderResourceBinding::UniformBuffer: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && uniformBuffer) + uniformBuffer(sd->stage, bd->u.ubuf, shaderRegister, bd->binding); + } + break; + case QRhiShaderResourceBinding::SampledTexture: + { + Q_ASSERT(bd->u.stex.count > 0); + const int textureBaseShaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + const int samplerBaseShaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).second; + for (int i = 0; i < bd->u.stex.count; ++i) { + if (textureBaseShaderRegister >= 0 && texture) + texture(sd->stage, bd->u.stex.texSamplers[i], textureBaseShaderRegister + i); + if (samplerBaseShaderRegister >= 0 && sampler) + sampler(sd->stage, bd->u.stex.texSamplers[i], samplerBaseShaderRegister + i); + } + } + break; + case QRhiShaderResourceBinding::Texture: + { + Q_ASSERT(bd->u.stex.count > 0); + const int baseShaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (baseShaderRegister >= 0 && texture) { + for (int i = 0; i < bd->u.stex.count; ++i) + texture(sd->stage, bd->u.stex.texSamplers[i], baseShaderRegister + i); + } + } + break; + case QRhiShaderResourceBinding::Sampler: + { + Q_ASSERT(bd->u.stex.count > 0); + const int baseShaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (baseShaderRegister >= 0 && sampler) { + for (int i = 0; i < bd->u.stex.count; ++i) + sampler(sd->stage, bd->u.stex.texSamplers[i], baseShaderRegister + i); + } + } + break; + case QRhiShaderResourceBinding::ImageLoad: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageImage) + storageImage(sd->stage, bd->u.simage, Load, shaderRegister); + } + break; + case QRhiShaderResourceBinding::ImageStore: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageImage) + storageImage(sd->stage, bd->u.simage, Store, shaderRegister); + } + break; + case QRhiShaderResourceBinding::ImageLoadStore: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageImage) + storageImage(sd->stage, bd->u.simage, LoadStore, shaderRegister); + } + break; + case QRhiShaderResourceBinding::BufferLoad: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageBuffer) + storageBuffer(sd->stage, bd->u.sbuf, Load, shaderRegister); + } + break; + case QRhiShaderResourceBinding::BufferStore: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageBuffer) + storageBuffer(sd->stage, bd->u.sbuf, Store, shaderRegister); + } + break; + case QRhiShaderResourceBinding::BufferLoadStore: + { + const int shaderRegister = mapBinding(bd->binding, sd->nativeResourceBindingMap).first; + if (shaderRegister >= 0 && storageBuffer) + storageBuffer(sd->stage, bd->u.sbuf, LoadStore, shaderRegister); + } + break; + } + } + } +} + +bool QD3D12SamplerManager::create(ID3D12Device *device) +{ + // This does not need to be per-frame slot, just grab space for MAX_SAMPLERS samplers. + if (!shaderVisibleSamplerHeap.create(device, + D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, + MAX_SAMPLERS / QD3D12_FRAMES_IN_FLIGHT)) + { + qWarning("Could not create shader-visible SAMPLER heap"); + return false; + } + + this->device = device; + return true; +} + +void QD3D12SamplerManager::destroy() +{ + if (device) { + shaderVisibleSamplerHeap.destroy(); + device = nullptr; + } +} + +QD3D12Descriptor QD3D12SamplerManager::getShaderVisibleDescriptor(const D3D12_SAMPLER_DESC &desc) +{ + auto it = gpuMap.constFind({desc}); + if (it != gpuMap.cend()) + return *it; + + QD3D12Descriptor descriptor = shaderVisibleSamplerHeap.heap.get(1); + if (descriptor.isValid()) { + device->CreateSampler(&desc, descriptor.cpuHandle); + gpuMap.insert({desc}, descriptor); + } else { + qWarning("Out of shader-visible SAMPLER descriptor heap space," + " this should not happen, maximum number of unique samplers is %u", + shaderVisibleSamplerHeap.heap.capacity); + } + + return descriptor; +} + +bool QD3D12MipmapGenerator::create(QRhiD3D12 *rhiD) +{ + this->rhiD = rhiD; + + D3D12_ROOT_PARAMETER1 rootParams[3] = {}; + D3D12_DESCRIPTOR_RANGE1 descriptorRanges[2] = {}; + + // b0 + rootParams[0].ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV; + rootParams[0].ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL; + rootParams[0].Descriptor.Flags = D3D12_ROOT_DESCRIPTOR_FLAG_DATA_STATIC; + + // t0 + descriptorRanges[0].RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_SRV; + descriptorRanges[0].NumDescriptors = 1; + descriptorRanges[0].Flags = D3D12_DESCRIPTOR_RANGE_FLAG_DATA_VOLATILE; + rootParams[1].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + rootParams[1].ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL; + rootParams[1].DescriptorTable.NumDescriptorRanges = 1; + rootParams[1].DescriptorTable.pDescriptorRanges = &descriptorRanges[0]; + + // u0..3 + descriptorRanges[1].RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_UAV; + descriptorRanges[1].NumDescriptors = 4; + rootParams[2].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + rootParams[2].ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL; + rootParams[2].DescriptorTable.NumDescriptorRanges = 1; + rootParams[2].DescriptorTable.pDescriptorRanges = &descriptorRanges[1]; + + // s0 + D3D12_STATIC_SAMPLER_DESC samplerDesc = {}; + samplerDesc.Filter = D3D12_FILTER_MIN_MAG_MIP_LINEAR; + samplerDesc.AddressU = D3D12_TEXTURE_ADDRESS_MODE_CLAMP; + samplerDesc.AddressV = D3D12_TEXTURE_ADDRESS_MODE_CLAMP; + samplerDesc.AddressW = D3D12_TEXTURE_ADDRESS_MODE_CLAMP; + samplerDesc.MaxLOD = 10000.0f; + samplerDesc.ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL; + + D3D12_VERSIONED_ROOT_SIGNATURE_DESC rsDesc = {}; + rsDesc.Version = D3D_ROOT_SIGNATURE_VERSION_1_1; + rsDesc.Desc_1_1.NumParameters = 3; + rsDesc.Desc_1_1.pParameters = rootParams; + rsDesc.Desc_1_1.NumStaticSamplers = 1; + rsDesc.Desc_1_1.pStaticSamplers = &samplerDesc; + + ID3DBlob *signature = nullptr; + HRESULT hr = D3D12SerializeVersionedRootSignature(&rsDesc, &signature, nullptr); + if (FAILED(hr)) { + qWarning("Failed to serialize root signature: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + ID3D12RootSignature *rootSig = nullptr; + hr = rhiD->dev->CreateRootSignature(0, + signature->GetBufferPointer(), + signature->GetBufferSize(), + __uuidof(ID3D12RootSignature), + reinterpret_cast<void **>(&rootSig)); + signature->Release(); + if (FAILED(hr)) { + qWarning("Failed to create root signature: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + + rootSigHandle = QD3D12RootSignature::addToPool(&rhiD->rootSignaturePool, rootSig); + + D3D12_COMPUTE_PIPELINE_STATE_DESC psoDesc = {}; + psoDesc.pRootSignature = rootSig; + psoDesc.CS.pShaderBytecode = g_csMipmap; + psoDesc.CS.BytecodeLength = sizeof(g_csMipmap); + ID3D12PipelineState *pso = nullptr; + hr = rhiD->dev->CreateComputePipelineState(&psoDesc, + __uuidof(ID3D12PipelineState), + reinterpret_cast<void **>(&pso)); + if (FAILED(hr)) { + qWarning("Failed to create compute pipeline state: %s", + qPrintable(QSystemError::windowsComString(hr))); + rhiD->rootSignaturePool.remove(rootSigHandle); + rootSigHandle = {}; + return false; + } + + pipelineHandle = QD3D12Pipeline::addToPool(&rhiD->pipelinePool, QD3D12Pipeline::Compute, pso); + + return true; +} + +void QD3D12MipmapGenerator::destroy() +{ + rhiD->pipelinePool.remove(pipelineHandle); + pipelineHandle = {}; + rhiD->rootSignaturePool.remove(rootSigHandle); + rootSigHandle = {}; +} + +void QD3D12MipmapGenerator::generate(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &textureHandle) +{ + QD3D12Pipeline *pipeline = rhiD->pipelinePool.lookupRef(pipelineHandle); + if (!pipeline) + return; + QD3D12RootSignature *rootSig = rhiD->rootSignaturePool.lookupRef(rootSigHandle); + if (!rootSig) + return; + QD3D12Resource *res = rhiD->resourcePool.lookupRef(textureHandle); + if (!res) + return; + + const quint32 mipLevelCount = res->desc.MipLevels; + if (mipLevelCount < 2) + return; + + if (res->desc.SampleDesc.Count > 1) { + qWarning("Cannot generate mipmaps for MSAA texture"); + return; + } + + const bool is1D = res->desc.Dimension == D3D12_RESOURCE_DIMENSION_TEXTURE1D; + if (is1D) { + qWarning("Cannot generate mipmaps for 1D texture"); + return; + } + + const bool is3D = res->desc.Dimension == D3D12_RESOURCE_DIMENSION_TEXTURE3D; + const bool isCubeOrArray = res->desc.Dimension == D3D12_RESOURCE_DIMENSION_TEXTURE2D + && res->desc.DepthOrArraySize > 1; + const quint32 layerCount = isCubeOrArray ? res->desc.DepthOrArraySize : 1; + + if (is3D) { + // ### needs its own shader and maybe a different solution + qWarning("3D texture mipmapping is not implemented for D3D12 atm"); + return; + } + + rhiD->barrierGen.addTransitionBarrier(textureHandle, D3D12_RESOURCE_STATE_UNORDERED_ACCESS); + rhiD->barrierGen.enqueueBufferedTransitionBarriers(cbD); + + cbD->cmdList->SetPipelineState(pipeline->pso); + cbD->cmdList->SetComputeRootSignature(rootSig->rootSig); + + const quint32 descriptorByteSize = rhiD->shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[rhiD->currentFrameSlot].descriptorByteSize; + + struct CBufData { + quint32 srcMipLevel; + quint32 numMipLevels; + float texelWidth; + float texelHeight; + }; + + const quint32 allocSize = QD3D12StagingArea::allocSizeForArray(sizeof(CBufData), mipLevelCount * layerCount); + std::optional<QD3D12StagingArea> ownStagingArea; + if (rhiD->smallStagingAreas[rhiD->currentFrameSlot].remainingCapacity() < allocSize) { + ownStagingArea = QD3D12StagingArea(); + if (!ownStagingArea->create(rhiD, allocSize, D3D12_HEAP_TYPE_UPLOAD)) { + qWarning("Could not create staging area for mipmap generation"); + return; + } + } + QD3D12StagingArea *workArea = ownStagingArea.has_value() + ? &ownStagingArea.value() + : &rhiD->smallStagingAreas[rhiD->currentFrameSlot]; + + bool gotNewHeap = false; + if (!rhiD->ensureShaderVisibleDescriptorHeapCapacity(&rhiD->shaderVisibleCbvSrvUavHeap, + D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, + rhiD->currentFrameSlot, + (1 + 4) * mipLevelCount * layerCount, + &gotNewHeap)) + { + qWarning("Could not ensure enough space in descriptor heap for mipmap generation"); + return; + } + if (gotNewHeap) + rhiD->bindShaderVisibleHeaps(cbD); + + for (quint32 layer = 0; layer < layerCount; ++layer) { + for (quint32 level = 0; level < mipLevelCount ;) { + UINT subresource = calcSubresource(level, layer, res->desc.MipLevels); + rhiD->barrierGen.enqueueSubresourceTransitionBarrier(cbD, textureHandle, subresource, + D3D12_RESOURCE_STATE_UNORDERED_ACCESS, + D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE); + + quint32 levelPlusOneMipWidth = res->desc.Width >> (level + 1); + quint32 levelPlusOneMipHeight = res->desc.Height >> (level + 1); + const quint32 dw = levelPlusOneMipWidth == 1 ? levelPlusOneMipHeight : levelPlusOneMipWidth; + const quint32 dh = levelPlusOneMipHeight == 1 ? levelPlusOneMipWidth : levelPlusOneMipHeight; + // number of times the size can be halved while still resulting in an even dimension + const quint32 additionalMips = qCountTrailingZeroBits(dw | dh); + const quint32 numGenMips = qMin(1u + qMin(3u, additionalMips), res->desc.MipLevels - level); + levelPlusOneMipWidth = qMax(1u, levelPlusOneMipWidth); + levelPlusOneMipHeight = qMax(1u, levelPlusOneMipHeight); + + CBufData cbufData = { + level, + numGenMips, + 1.0f / float(levelPlusOneMipWidth), + 1.0f / float(levelPlusOneMipHeight) + }; + + QD3D12StagingArea::Allocation cbuf = workArea->get(sizeof(cbufData)); + memcpy(cbuf.p, &cbufData, sizeof(cbufData)); + cbD->cmdList->SetComputeRootConstantBufferView(0, cbuf.gpuAddr); + + QD3D12Descriptor srv = rhiD->shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[rhiD->currentFrameSlot].get(1); + D3D12_SHADER_RESOURCE_VIEW_DESC srvDesc = {}; + srvDesc.Format = res->desc.Format; + srvDesc.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING; + if (isCubeOrArray) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MostDetailedMip = level; + srvDesc.Texture2DArray.MipLevels = 1; + srvDesc.Texture2DArray.FirstArraySlice = layer; + srvDesc.Texture2DArray.ArraySize = 1; + } else if (is3D) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D; + srvDesc.Texture3D.MostDetailedMip = level; + srvDesc.Texture3D.MipLevels = 1; + } else { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MostDetailedMip = level; + srvDesc.Texture2D.MipLevels = 1; + } + rhiD->dev->CreateShaderResourceView(res->resource, &srvDesc, srv.cpuHandle); + cbD->cmdList->SetComputeRootDescriptorTable(1, srv.gpuHandle); + + QD3D12Descriptor uavStart = rhiD->shaderVisibleCbvSrvUavHeap.perFrameHeapSlice[rhiD->currentFrameSlot].get(4); + D3D12_CPU_DESCRIPTOR_HANDLE uavCpuHandle = uavStart.cpuHandle; + // if level is N, then need UAVs for levels N+1, ..., N+4 + for (quint32 uavIdx = 0; uavIdx < 4; ++uavIdx) { + const quint32 uavMipLevel = qMin(level + 1u + uavIdx, res->desc.MipLevels - 1u); + D3D12_UNORDERED_ACCESS_VIEW_DESC uavDesc = {}; + uavDesc.Format = res->desc.Format; + if (isCubeOrArray) { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY; + uavDesc.Texture2DArray.MipSlice = uavMipLevel; + uavDesc.Texture2DArray.FirstArraySlice = layer; + uavDesc.Texture2DArray.ArraySize = 1; + } else if (is3D) { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE3D; + uavDesc.Texture3D.MipSlice = uavMipLevel; + uavDesc.Texture3D.FirstWSlice = 0; // depth etc. not implemented yet + uavDesc.Texture3D.WSize = 1; + } else { + uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2D; + uavDesc.Texture2D.MipSlice = uavMipLevel; + } + rhiD->dev->CreateUnorderedAccessView(res->resource, nullptr, &uavDesc, uavCpuHandle); + uavCpuHandle.ptr += descriptorByteSize; + } + cbD->cmdList->SetComputeRootDescriptorTable(2, uavStart.gpuHandle); + + cbD->cmdList->Dispatch(levelPlusOneMipWidth, levelPlusOneMipHeight, 1); + + rhiD->barrierGen.enqueueUavBarrier(cbD, textureHandle); + rhiD->barrierGen.enqueueSubresourceTransitionBarrier(cbD, textureHandle, subresource, + D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE, + D3D12_RESOURCE_STATE_UNORDERED_ACCESS); + + level += numGenMips; + } + } + + if (ownStagingArea.has_value()) + ownStagingArea->destroyWithDeferredRelease(&rhiD->releaseQueue); +} + +bool QD3D12MemoryAllocator::create(ID3D12Device *device, IDXGIAdapter1 *adapter) +{ + this->device = device; + + // We can function with and without D3D12MA: CreateCommittedResource is + // just fine for our purposes and not any complicated API-wise; the memory + // allocator is interesting for efficiency mainly since it can suballocate + // instead of making everything a committed resource allocation. + + static bool disableMA = qEnvironmentVariableIntValue("QT_D3D_NO_SUBALLOC"); + if (disableMA) + return true; + + DXGI_ADAPTER_DESC1 desc; + adapter->GetDesc1(&desc); + if (desc.Flags & DXGI_ADAPTER_FLAG_SOFTWARE) + return true; + + D3D12MA::ALLOCATOR_DESC allocatorDesc = {}; + allocatorDesc.pDevice = device; + allocatorDesc.pAdapter = adapter; + // A QRhi is supposed to be used from one single thread only. Disable + // the allocator's own mutexes. This may give a performance boost. + allocatorDesc.Flags = D3D12MA::ALLOCATOR_FLAG_SINGLETHREADED; + HRESULT hr = D3D12MA::CreateAllocator(&allocatorDesc, &allocator); + if (FAILED(hr)) { + qWarning("Failed to initialize D3D12 Memory Allocator: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + return true; +} + +void QD3D12MemoryAllocator::destroy() +{ + if (allocator) { + allocator->Release(); + allocator = nullptr; + } +} + +HRESULT QD3D12MemoryAllocator::createResource(D3D12_HEAP_TYPE heapType, + const D3D12_RESOURCE_DESC *resourceDesc, + D3D12_RESOURCE_STATES initialState, + const D3D12_CLEAR_VALUE *optimizedClearValue, + D3D12MA::Allocation **maybeAllocation, + REFIID riidResource, + void **ppvResource) +{ + if (allocator) { + D3D12MA::ALLOCATION_DESC allocDesc = {}; + allocDesc.HeapType = heapType; + return allocator->CreateResource(&allocDesc, + resourceDesc, + initialState, + optimizedClearValue, + maybeAllocation, + riidResource, + ppvResource); + } else { + *maybeAllocation = nullptr; + D3D12_HEAP_PROPERTIES heapProps = {}; + heapProps.Type = heapType; + return device->CreateCommittedResource(&heapProps, + D3D12_HEAP_FLAG_NONE, + resourceDesc, + initialState, + optimizedClearValue, + riidResource, + ppvResource); + } +} + +void QD3D12MemoryAllocator::getBudget(D3D12MA::Budget *localBudget, D3D12MA::Budget *nonLocalBudget) +{ + if (allocator) { + allocator->GetBudget(localBudget, nonLocalBudget); + } else { + *localBudget = {}; + *nonLocalBudget = {}; + } +} + +void QRhiD3D12::waitGpu() +{ + fullFenceCounter += 1u; + if (SUCCEEDED(cmdQueue->Signal(fullFence, fullFenceCounter))) { + if (SUCCEEDED(fullFence->SetEventOnCompletion(fullFenceCounter, fullFenceEvent))) + WaitForSingleObject(fullFenceEvent, INFINITE); + } +} + +DXGI_SAMPLE_DESC QRhiD3D12::effectiveSampleDesc(int sampleCount, DXGI_FORMAT format) const +{ + DXGI_SAMPLE_DESC desc; + desc.Count = 1; + desc.Quality = 0; + + const int s = effectiveSampleCount(sampleCount); + + if (s > 1) { + D3D12_FEATURE_DATA_MULTISAMPLE_QUALITY_LEVELS msaaInfo = {}; + msaaInfo.Format = format; + msaaInfo.SampleCount = UINT(s); + if (SUCCEEDED(dev->CheckFeatureSupport(D3D12_FEATURE_MULTISAMPLE_QUALITY_LEVELS, &msaaInfo, sizeof(msaaInfo)))) { + if (msaaInfo.NumQualityLevels > 0) { + desc.Count = UINT(s); + desc.Quality = msaaInfo.NumQualityLevels - 1; + } else { + qWarning("No quality levels for multisampling with sample count %d", s); + } + } + } + + return desc; +} + +bool QRhiD3D12::startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList1 **cmdList) +{ + ID3D12CommandAllocator *cmdAlloc = cmdAllocators[currentFrameSlot]; + if (!*cmdList) { + HRESULT hr = dev->CreateCommandList(0, + D3D12_COMMAND_LIST_TYPE_DIRECT, + cmdAlloc, + nullptr, + __uuidof(ID3D12GraphicsCommandList1), + reinterpret_cast<void **>(cmdList)); + if (FAILED(hr)) { + qWarning("Failed to create command list: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } else { + HRESULT hr = (*cmdList)->Reset(cmdAlloc, nullptr); + if (FAILED(hr)) { + qWarning("Failed to reset command list: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + return true; +} + +static inline QRhiTexture::Format swapchainReadbackTextureFormat(DXGI_FORMAT format, QRhiTexture::Flags *flags) +{ + switch (format) { + case DXGI_FORMAT_R8G8B8A8_UNORM: + return QRhiTexture::RGBA8; + case DXGI_FORMAT_R8G8B8A8_UNORM_SRGB: + if (flags) + (*flags) |= QRhiTexture::sRGB; + return QRhiTexture::RGBA8; + case DXGI_FORMAT_B8G8R8A8_UNORM: + return QRhiTexture::BGRA8; + case DXGI_FORMAT_B8G8R8A8_UNORM_SRGB: + if (flags) + (*flags) |= QRhiTexture::sRGB; + return QRhiTexture::BGRA8; + case DXGI_FORMAT_R16G16B16A16_FLOAT: + return QRhiTexture::RGBA16F; + case DXGI_FORMAT_R32G32B32A32_FLOAT: + return QRhiTexture::RGBA32F; + case DXGI_FORMAT_R10G10B10A2_UNORM: + return QRhiTexture::RGB10A2; + default: + qWarning("DXGI_FORMAT %d cannot be read back", format); + break; + } + return QRhiTexture::UnknownFormat; +} + +void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates) +{ + QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); + + for (int opIdx = 0; opIdx < ud->activeBufferOpCount; ++opIdx) { + const QRhiResourceUpdateBatchPrivate::BufferOp &u(ud->bufferOps[opIdx]); + if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::DynamicUpdate) { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, u.buf); + Q_ASSERT(bufD->m_type == QRhiBuffer::Dynamic); + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (u.offset == 0 && u.data.size() == bufD->m_size) + bufD->pendingHostWrites[i].clear(); + bufD->pendingHostWrites[i].append({ u.offset, u.data }); + } + } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::StaticUpload) { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, u.buf); + Q_ASSERT(bufD->m_type != QRhiBuffer::Dynamic); + Q_ASSERT(u.offset + u.data.size() <= bufD->m_size); + + // The general approach to staging upload data is to first try + // using the per-frame "small" staging area, which is a very simple + // linear allocator; if that's not big enough then create a + // dedicated StagingArea and then deferred-release it to make sure + // if stays alive while the frame is possibly still in flight. + + QD3D12StagingArea::Allocation stagingAlloc; + const quint32 allocSize = QD3D12StagingArea::allocSizeForArray(bufD->m_size, 1); + if (smallStagingAreas[currentFrameSlot].remainingCapacity() >= allocSize) + stagingAlloc = smallStagingAreas[currentFrameSlot].get(bufD->m_size); + + std::optional<QD3D12StagingArea> ownStagingArea; + if (!stagingAlloc.isValid()) { + ownStagingArea = QD3D12StagingArea(); + if (!ownStagingArea->create(this, allocSize, D3D12_HEAP_TYPE_UPLOAD)) + continue; + stagingAlloc = ownStagingArea->get(allocSize); + if (!stagingAlloc.isValid()) { + ownStagingArea->destroy(); + continue; + } + } + + memcpy(stagingAlloc.p + u.offset, u.data.constData(), u.data.size()); + + barrierGen.addTransitionBarrier(bufD->handles[0], D3D12_RESOURCE_STATE_COPY_DEST); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + + if (QD3D12Resource *res = resourcePool.lookupRef(bufD->handles[0])) { + cbD->cmdList->CopyBufferRegion(res->resource, + u.offset, + stagingAlloc.buffer, + stagingAlloc.bufferOffset + u.offset, + u.data.size()); + } + + if (ownStagingArea.has_value()) + ownStagingArea->destroyWithDeferredRelease(&releaseQueue); + } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::Read) { + QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, u.buf); + if (bufD->m_type == QRhiBuffer::Dynamic) { + bufD->executeHostWritesForFrameSlot(currentFrameSlot); + if (QD3D12Resource *res = resourcePool.lookupRef(bufD->handles[currentFrameSlot])) { + Q_ASSERT(res->cpuMapPtr); + u.result->data.resize(u.readSize); + memcpy(u.result->data.data(), reinterpret_cast<char *>(res->cpuMapPtr) + u.offset, u.readSize); + } + if (u.result->completed) + u.result->completed(); + } else { + QD3D12Readback readback; + readback.frameSlot = currentFrameSlot; + readback.result = u.result; + readback.byteSize = u.readSize; + const quint32 allocSize = aligned(u.readSize, QD3D12StagingArea::ALIGNMENT); + if (!readback.staging.create(this, allocSize, D3D12_HEAP_TYPE_READBACK)) { + if (u.result->completed) + u.result->completed(); + continue; + } + QD3D12StagingArea::Allocation stagingAlloc = readback.staging.get(u.readSize); + if (!stagingAlloc.isValid()) { + readback.staging.destroy(); + if (u.result->completed) + u.result->completed(); + continue; + } + Q_ASSERT(stagingAlloc.bufferOffset == 0); + barrierGen.addTransitionBarrier(bufD->handles[0], D3D12_RESOURCE_STATE_COPY_SOURCE); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + if (QD3D12Resource *res = resourcePool.lookupRef(bufD->handles[0])) { + cbD->cmdList->CopyBufferRegion(stagingAlloc.buffer, 0, res->resource, u.offset, u.readSize); + activeReadbacks.append(readback); + } else { + readback.staging.destroy(); + if (u.result->completed) + u.result->completed(); + } + } + } + } + + for (int opIdx = 0; opIdx < ud->activeTextureOpCount; ++opIdx) { + const QRhiResourceUpdateBatchPrivate::TextureOp &u(ud->textureOps[opIdx]); + if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Upload) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, u.dst); + const bool is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + QD3D12Resource *res = resourcePool.lookupRef(texD->handle); + if (!res) + continue; + barrierGen.addTransitionBarrier(texD->handle, D3D12_RESOURCE_STATE_COPY_DEST); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { + for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) { + D3D12_SUBRESOURCE_FOOTPRINT footprint = {}; + footprint.Format = res->desc.Format; + footprint.Depth = 1; + quint32 totalBytes = 0; + + const QSize subresSize = subresDesc.sourceSize().isEmpty() ? q->sizeForMipLevel(level, texD->m_pixelSize) + : subresDesc.sourceSize(); + const QPoint srcPos = subresDesc.sourceTopLeft(); + QPoint dstPos = subresDesc.destinationTopLeft(); + + if (!subresDesc.image().isNull()) { + const QImage img = subresDesc.image(); + const int bpl = img.bytesPerLine(); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + totalBytes = footprint.RowPitch * img.height(); + } else if (!subresDesc.data().isEmpty() && isCompressedFormat(texD->m_format)) { + QSize blockDim; + quint32 bpl = 0; + compressedFormatInfo(texD->m_format, subresSize, &bpl, nullptr, &blockDim); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + const int rowCount = aligned(subresSize.height(), blockDim.height()) / blockDim.height(); + totalBytes = footprint.RowPitch * rowCount; + } else if (!subresDesc.data().isEmpty()) { + quint32 bpl = 0; + if (subresDesc.dataStride()) + bpl = subresDesc.dataStride(); + else + textureFormatInfo(texD->m_format, subresSize, &bpl, nullptr, nullptr); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + totalBytes = footprint.RowPitch * subresSize.height(); + } else { + qWarning("Invalid texture upload for %p layer=%d mip=%d", texD, layer, level); + continue; + } + + const quint32 allocSize = QD3D12StagingArea::allocSizeForArray(totalBytes, 1); + QD3D12StagingArea::Allocation stagingAlloc; + if (smallStagingAreas[currentFrameSlot].remainingCapacity() >= allocSize) + stagingAlloc = smallStagingAreas[currentFrameSlot].get(allocSize); + + std::optional<QD3D12StagingArea> ownStagingArea; + if (!stagingAlloc.isValid()) { + ownStagingArea = QD3D12StagingArea(); + if (!ownStagingArea->create(this, allocSize, D3D12_HEAP_TYPE_UPLOAD)) + continue; + stagingAlloc = ownStagingArea->get(allocSize); + if (!stagingAlloc.isValid()) { + ownStagingArea->destroy(); + continue; + } + } + + D3D12_TEXTURE_COPY_LOCATION dst; + dst.pResource = res->resource; + dst.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + dst.SubresourceIndex = calcSubresource(UINT(level), is3D ? 0u : UINT(layer), texD->mipLevelCount); + D3D12_TEXTURE_COPY_LOCATION src; + src.pResource = stagingAlloc.buffer; + src.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT; + src.PlacedFootprint.Offset = stagingAlloc.bufferOffset; + + D3D12_BOX srcBox; // back, right, bottom are exclusive + + if (!subresDesc.image().isNull()) { + const QImage img = subresDesc.image(); + const int bpc = qMax(1, img.depth() / 8); + const int bpl = img.bytesPerLine(); + + QSize size = subresDesc.sourceSize().isEmpty() ? img.size() : subresDesc.sourceSize(); + size.setWidth(qMin(size.width(), img.width() - srcPos.x())); + size.setHeight(qMin(size.height(), img.height() - srcPos.y())); + + footprint.Width = size.width(); + footprint.Height = size.height(); + + srcBox.left = 0; + srcBox.top = 0; + srcBox.right = UINT(size.width()); + srcBox.bottom = UINT(size.height()); + srcBox.front = 0; + srcBox.back = 1; + + const uchar *imgPtr = img.constBits(); + const quint32 lineBytes = size.width() * bpc; + for (int y = 0, h = size.height(); y < h; ++y) { + memcpy(stagingAlloc.p + y * footprint.RowPitch, + imgPtr + srcPos.x() * bpc + (y + srcPos.y()) * bpl, + lineBytes); + } + } else if (!subresDesc.data().isEmpty() && isCompressedFormat(texD->m_format)) { + QSize blockDim; + quint32 bpl = 0; + compressedFormatInfo(texD->m_format, subresSize, &bpl, nullptr, &blockDim); + // x and y must be multiples of the block width and height + dstPos.setX(aligned(dstPos.x(), blockDim.width())); + dstPos.setY(aligned(dstPos.y(), blockDim.height())); + + srcBox.left = 0; + srcBox.top = 0; + // width and height must be multiples of the block width and height + srcBox.right = aligned(subresSize.width(), blockDim.width()); + srcBox.bottom = aligned(subresSize.height(), blockDim.height()); + + srcBox.front = 0; + srcBox.back = 1; + + footprint.Width = aligned(subresSize.width(), blockDim.width()); + footprint.Height = aligned(subresSize.height(), blockDim.height()); + + const quint32 copyBytes = qMin(bpl, footprint.RowPitch); + const QByteArray imgData = subresDesc.data(); + const char *imgPtr = imgData.constData(); + const int rowCount = aligned(subresSize.height(), blockDim.height()) / blockDim.height(); + for (int y = 0; y < rowCount; ++y) + memcpy(stagingAlloc.p + y * footprint.RowPitch, imgPtr + y * bpl, copyBytes); + } else if (!subresDesc.data().isEmpty()) { + srcBox.left = 0; + srcBox.top = 0; + srcBox.right = subresSize.width(); + srcBox.bottom = subresSize.height(); + srcBox.front = 0; + srcBox.back = 1; + + footprint.Width = subresSize.width(); + footprint.Height = subresSize.height(); + + quint32 bpl = 0; + if (subresDesc.dataStride()) + bpl = subresDesc.dataStride(); + else + textureFormatInfo(texD->m_format, subresSize, &bpl, nullptr, nullptr); + + const quint32 copyBytes = qMin(bpl, footprint.RowPitch); + const QByteArray data = subresDesc.data(); + const char *imgPtr = data.constData(); + for (int y = 0, h = subresSize.height(); y < h; ++y) + memcpy(stagingAlloc.p + y * footprint.RowPitch, imgPtr + y * bpl, copyBytes); + } + + src.PlacedFootprint.Footprint = footprint; + + cbD->cmdList->CopyTextureRegion(&dst, + UINT(dstPos.x()), + UINT(dstPos.y()), + is3D ? UINT(layer) : 0u, + &src, + &srcBox); + + if (ownStagingArea.has_value()) + ownStagingArea->destroyWithDeferredRelease(&releaseQueue); + } + } + } + } else if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Copy) { + Q_ASSERT(u.src && u.dst); + QD3D12Texture *srcD = QRHI_RES(QD3D12Texture, u.src); + QD3D12Texture *dstD = QRHI_RES(QD3D12Texture, u.dst); + const bool srcIs3D = srcD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + const bool dstIs3D = dstD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + QD3D12Resource *srcRes = resourcePool.lookupRef(srcD->handle); + QD3D12Resource *dstRes = resourcePool.lookupRef(dstD->handle); + if (!srcRes || !dstRes) + continue; + + barrierGen.addTransitionBarrier(srcD->handle, D3D12_RESOURCE_STATE_COPY_SOURCE); + barrierGen.addTransitionBarrier(dstD->handle, D3D12_RESOURCE_STATE_COPY_DEST); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + + const UINT srcSubresource = calcSubresource(UINT(u.desc.sourceLevel()), + srcIs3D ? 0u : UINT(u.desc.sourceLayer()), + srcD->mipLevelCount); + const UINT dstSubresource = calcSubresource(UINT(u.desc.destinationLevel()), + dstIs3D ? 0u : UINT(u.desc.destinationLayer()), + dstD->mipLevelCount); + const QPoint dp = u.desc.destinationTopLeft(); + const QSize mipSize = q->sizeForMipLevel(u.desc.sourceLevel(), srcD->m_pixelSize); + const QSize copySize = u.desc.pixelSize().isEmpty() ? mipSize : u.desc.pixelSize(); + const QPoint sp = u.desc.sourceTopLeft(); + + D3D12_BOX srcBox; + srcBox.left = UINT(sp.x()); + srcBox.top = UINT(sp.y()); + srcBox.front = srcIs3D ? UINT(u.desc.sourceLayer()) : 0u; + // back, right, bottom are exclusive + srcBox.right = srcBox.left + UINT(copySize.width()); + srcBox.bottom = srcBox.top + UINT(copySize.height()); + srcBox.back = srcBox.front + 1; + + D3D12_TEXTURE_COPY_LOCATION src; + src.pResource = srcRes->resource; + src.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + src.SubresourceIndex = srcSubresource; + D3D12_TEXTURE_COPY_LOCATION dst; + dst.pResource = dstRes->resource; + dst.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + dst.SubresourceIndex = dstSubresource; + + cbD->cmdList->CopyTextureRegion(&dst, + UINT(dp.x()), + UINT(dp.y()), + dstIs3D ? UINT(u.desc.destinationLayer()) : 0u, + &src, + &srcBox); + } else if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Read) { + QD3D12Readback readback; + readback.frameSlot = currentFrameSlot; + readback.result = u.result; + + QD3D12ObjectHandle srcHandle; + bool is3D = false; + if (u.rb.texture()) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, u.rb.texture()); + if (texD->sampleDesc.Count > 1) { + qWarning("Multisample texture cannot be read back"); + continue; + } + is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + readback.pixelSize = q->sizeForMipLevel(u.rb.level(), texD->m_pixelSize); + readback.format = texD->m_format; + srcHandle = texD->handle; + } else { + Q_ASSERT(currentSwapChain); + readback.pixelSize = currentSwapChain->pixelSize; + readback.format = swapchainReadbackTextureFormat(currentSwapChain->colorFormat, nullptr); + if (readback.format == QRhiTexture::UnknownFormat) + continue; + srcHandle = currentSwapChain->colorBuffers[currentSwapChain->currentBackBufferIndex]; + } + + textureFormatInfo(readback.format, + readback.pixelSize, + &readback.bytesPerLine, + &readback.byteSize, + nullptr); + + QD3D12Resource *srcRes = resourcePool.lookupRef(srcHandle); + if (!srcRes) + continue; + + const UINT subresource = calcSubresource(UINT(u.rb.level()), + is3D ? 0u : UINT(u.rb.layer()), + srcRes->desc.MipLevels); + D3D12_PLACED_SUBRESOURCE_FOOTPRINT layout; + // totalBytes is what we get from D3D, with the 256 aligned stride, + // readback.byteSize is the final result that's not relevant here yet + UINT64 totalBytes = 0; + dev->GetCopyableFootprints(&srcRes->desc, subresource, 1, 0, + &layout, nullptr, nullptr, &totalBytes); + readback.stagingRowPitch = layout.Footprint.RowPitch; + + const quint32 allocSize = aligned<quint32>(totalBytes, QD3D12StagingArea::ALIGNMENT); + if (!readback.staging.create(this, allocSize, D3D12_HEAP_TYPE_READBACK)) { + if (u.result->completed) + u.result->completed(); + continue; + } + QD3D12StagingArea::Allocation stagingAlloc = readback.staging.get(totalBytes); + if (!stagingAlloc.isValid()) { + readback.staging.destroy(); + if (u.result->completed) + u.result->completed(); + continue; + } + Q_ASSERT(stagingAlloc.bufferOffset == 0); + + barrierGen.addTransitionBarrier(srcHandle, D3D12_RESOURCE_STATE_COPY_SOURCE); + barrierGen.enqueueBufferedTransitionBarriers(cbD); + + D3D12_TEXTURE_COPY_LOCATION dst; + dst.pResource = stagingAlloc.buffer; + dst.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT; + dst.PlacedFootprint.Offset = 0; + dst.PlacedFootprint.Footprint = layout.Footprint; + + D3D12_TEXTURE_COPY_LOCATION src; + src.pResource = srcRes->resource; + src.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + src.SubresourceIndex = subresource; + + D3D12_BOX srcBox = {}; + if (is3D) { + srcBox.front = UINT(u.rb.layer()); + srcBox.back = srcBox.front + 1; + srcBox.right = readback.pixelSize.width(); // exclusive + srcBox.bottom = readback.pixelSize.height(); + } + cbD->cmdList->CopyTextureRegion(&dst, 0, 0, 0, &src, is3D ? &srcBox : nullptr); + activeReadbacks.append(readback); + } else if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::GenMips) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, u.dst); + Q_ASSERT(texD->flags().testFlag(QRhiTexture::UsedWithGenerateMips)); + mipmapGen.generate(cbD, texD->handle); + } + } + + ud->free(); +} + +void QRhiD3D12::finishActiveReadbacks(bool forced) +{ + QVarLengthArray<std::function<void()>, 4> completedCallbacks; + + for (int i = activeReadbacks.size() - 1; i >= 0; --i) { + QD3D12Readback &readback(activeReadbacks[i]); + if (forced || currentFrameSlot == readback.frameSlot || readback.frameSlot < 0) { + readback.result->format = readback.format; + readback.result->pixelSize = readback.pixelSize; + readback.result->data.resize(int(readback.byteSize)); + + if (readback.format != QRhiTexture::UnknownFormat) { + quint8 *dstPtr = reinterpret_cast<quint8 *>(readback.result->data.data()); + const quint8 *srcPtr = readback.staging.mem.p; + const quint32 lineSize = qMin(readback.bytesPerLine, readback.stagingRowPitch); + for (int y = 0, h = readback.pixelSize.height(); y < h; ++y) + memcpy(dstPtr + y * readback.bytesPerLine, srcPtr + y * readback.stagingRowPitch, lineSize); + } else { + memcpy(readback.result->data.data(), readback.staging.mem.p, readback.byteSize); + } + + readback.staging.destroy(); + + if (readback.result->completed) + completedCallbacks.append(readback.result->completed); + + activeReadbacks.removeLast(); + } + } + + for (auto f : completedCallbacks) + f(); +} + +bool QRhiD3D12::ensureShaderVisibleDescriptorHeapCapacity(QD3D12ShaderVisibleDescriptorHeap *h, + D3D12_DESCRIPTOR_HEAP_TYPE type, + int frameSlot, + quint32 neededDescriptorCount, + bool *gotNew) +{ + // Gets a new heap if needed. Note that the capacity we get is clamped + // automatically (e.g. to 1 million, or 2048 for samplers), so * 2 does not + // mean we can grow indefinitely, then again even using the same size would + // work (because we what we are after here is a new heap for the rest of + // the commands, not affecting what's already recorded). + if (h->perFrameHeapSlice[frameSlot].remainingCapacity() < neededDescriptorCount) { + const quint32 newPerFrameSize = qMax(h->perFrameHeapSlice[frameSlot].capacity * 2, + neededDescriptorCount); + QD3D12ShaderVisibleDescriptorHeap newHeap; + if (!newHeap.create(dev, type, newPerFrameSize)) { + qWarning("Could not create new shader-visible descriptor heap"); + return false; + } + h->destroyWithDeferredRelease(&releaseQueue); + *h = newHeap; + *gotNew = true; + } + return true; +} + +void QRhiD3D12::bindShaderVisibleHeaps(QD3D12CommandBuffer *cbD) +{ + ID3D12DescriptorHeap *heaps[] = { + shaderVisibleCbvSrvUavHeap.heap.heap, + samplerMgr.shaderVisibleSamplerHeap.heap.heap + }; + cbD->cmdList->SetDescriptorHeaps(2, heaps); +} + +QD3D12Buffer::QD3D12Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) + : QRhiBuffer(rhi, type, usage, size) +{ +} + +QD3D12Buffer::~QD3D12Buffer() +{ + destroy(); +} + +void QD3D12Buffer::destroy() +{ + if (handles[0].isNull()) + return; + + QRHI_RES_RHI(QRhiD3D12); + + // destroy() implementations, unlike other functions, are expected to test + // for m_rhi (rhiD) being null, to allow surviving in case one attempts to + // destroy a (leaked) resource after the QRhi. + // + // If there is no QRhi anymore, we do not deferred-release but that's fine + // since the QRhi already released everything that was in the resourcePool. + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (rhiD) + rhiD->releaseQueue.deferredReleaseResource(handles[i]); + handles[i] = {}; + pendingHostWrites[i].clear(); + } + + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12Buffer::create() +{ + if (!handles[0].isNull()) + destroy(); + + if (m_usage.testFlag(QRhiBuffer::UniformBuffer) && m_type != Dynamic) { + qWarning("UniformBuffer must always be Dynamic"); + return false; + } + + if (m_usage.testFlag(QRhiBuffer::StorageBuffer) && m_type == Dynamic) { + qWarning("StorageBuffer cannot be combined with Dynamic"); + return false; + } + + const quint32 nonZeroSize = m_size <= 0 ? 256 : m_size; + const quint32 roundedSize = aligned(nonZeroSize, m_usage.testFlag(QRhiBuffer::UniformBuffer) ? 256u : 4u); + + UINT resourceFlags = D3D12_RESOURCE_FLAG_NONE; + if (m_usage.testFlag(QRhiBuffer::StorageBuffer)) + resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS; + + QRHI_RES_RHI(QRhiD3D12); + HRESULT hr = 0; + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + if (i == 0 || m_type == Dynamic) { + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_BUFFER; + resourceDesc.Width = roundedSize; + resourceDesc.Height = 1; + resourceDesc.DepthOrArraySize = 1; + resourceDesc.MipLevels = 1; + resourceDesc.Format = DXGI_FORMAT_UNKNOWN; + resourceDesc.SampleDesc = { 1, 0 }; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_ROW_MAJOR; + resourceDesc.Flags = D3D12_RESOURCE_FLAGS(resourceFlags); + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + // Dynamic == host (CPU) visible + D3D12_HEAP_TYPE heapType = m_type == Dynamic + ? D3D12_HEAP_TYPE_UPLOAD + : D3D12_HEAP_TYPE_DEFAULT; + D3D12_RESOURCE_STATES resourceState = m_type == Dynamic + ? D3D12_RESOURCE_STATE_GENERIC_READ + : D3D12_RESOURCE_STATE_COMMON; + hr = rhiD->vma.createResource(heapType, + &resourceDesc, + resourceState, + nullptr, + &allocation, + __uuidof(resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) + break; + if (!m_objectName.isEmpty()) { + QString decoratedName = QString::fromUtf8(m_objectName); + if (m_type == Dynamic) { + decoratedName += QLatin1Char('/'); + decoratedName += QString::number(i); + } + resource->SetName(reinterpret_cast<LPCWSTR>(decoratedName.utf16())); + } + void *cpuMemPtr = nullptr; + if (m_type == Dynamic) { + // will be mapped for ever on the CPU, this makes future host write operations very simple + hr = resource->Map(0, nullptr, &cpuMemPtr); + if (FAILED(hr)) { + qWarning("Map() failed to dynamic buffer"); + resource->Release(); + if (allocation) + allocation->Release(); + break; + } + } + handles[i] = QD3D12Resource::addToPool(&rhiD->resourcePool, + resource, + resourceState, + allocation, + cpuMemPtr); + } + } + if (FAILED(hr)) { + qWarning("Failed to create buffer: '%s' Type was %d, size was %u, using D3D12MA was %d.", + qPrintable(QSystemError::windowsComString(hr)), + int(m_type), + roundedSize, + int(rhiD->vma.isUsingD3D12MA())); + return false; + } + + rhiD->registerResource(this); + return true; +} + +QRhiBuffer::NativeBuffer QD3D12Buffer::nativeBuffer() +{ + NativeBuffer b; + Q_ASSERT(sizeof(b.objects) / sizeof(b.objects[0]) >= size_t(QD3D12_FRAMES_IN_FLIGHT)); + QRHI_RES_RHI(QRhiD3D12); + if (m_type == Dynamic) { + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + executeHostWritesForFrameSlot(i); + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handles[i])) + b.objects[i] = res->resource; + else + b.objects[i] = nullptr; + } + b.slotCount = QD3D12_FRAMES_IN_FLIGHT; + return b; + } + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handles[0])) + b.objects[0] = res->resource; + else + b.objects[0] = nullptr; + b.slotCount = 1; + return b; +} + +char *QD3D12Buffer::beginFullDynamicBufferUpdateForCurrentFrame() +{ + // Shortcut the entire buffer update mechanism and allow the client to do + // the host writes directly to the buffer. This will lead to unexpected + // results when combined with QRhiResourceUpdateBatch-based updates for the + // buffer, but provides a fast path for dynamic buffers that have all their + // content changed in every frame. + + Q_ASSERT(m_type == Dynamic); + QRHI_RES_RHI(QRhiD3D12); + Q_ASSERT(rhiD->inFrame); + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handles[rhiD->currentFrameSlot])) + return static_cast<char *>(res->cpuMapPtr); + + return nullptr; +} + +void QD3D12Buffer::endFullDynamicBufferUpdateForCurrentFrame() +{ + // nothing to do here +} + +void QD3D12Buffer::executeHostWritesForFrameSlot(int frameSlot) +{ + if (pendingHostWrites[frameSlot].isEmpty()) + return; + + Q_ASSERT(m_type == QRhiBuffer::Dynamic); + QRHI_RES_RHI(QRhiD3D12); + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handles[frameSlot])) { + Q_ASSERT(res->cpuMapPtr); + for (const QD3D12Buffer::HostWrite &u : std::as_const(pendingHostWrites[frameSlot])) + memcpy(static_cast<char *>(res->cpuMapPtr) + u.offset, u.data.constData(), u.data.size()); + } + pendingHostWrites[frameSlot].clear(); +} + +static inline DXGI_FORMAT toD3DTextureFormat(QRhiTexture::Format format, QRhiTexture::Flags flags) +{ + const bool srgb = flags.testFlag(QRhiTexture::sRGB); + switch (format) { + case QRhiTexture::RGBA8: + return srgb ? DXGI_FORMAT_R8G8B8A8_UNORM_SRGB : DXGI_FORMAT_R8G8B8A8_UNORM; + case QRhiTexture::BGRA8: + return srgb ? DXGI_FORMAT_B8G8R8A8_UNORM_SRGB : DXGI_FORMAT_B8G8R8A8_UNORM; + case QRhiTexture::R8: + return DXGI_FORMAT_R8_UNORM; + case QRhiTexture::RG8: + return DXGI_FORMAT_R8G8_UNORM; + case QRhiTexture::R16: + return DXGI_FORMAT_R16_UNORM; + case QRhiTexture::RG16: + return DXGI_FORMAT_R16G16_UNORM; + case QRhiTexture::RED_OR_ALPHA8: + return DXGI_FORMAT_R8_UNORM; + + case QRhiTexture::RGBA16F: + return DXGI_FORMAT_R16G16B16A16_FLOAT; + case QRhiTexture::RGBA32F: + return DXGI_FORMAT_R32G32B32A32_FLOAT; + case QRhiTexture::R16F: + return DXGI_FORMAT_R16_FLOAT; + case QRhiTexture::R32F: + return DXGI_FORMAT_R32_FLOAT; + + case QRhiTexture::RGB10A2: + return DXGI_FORMAT_R10G10B10A2_UNORM; + + case QRhiTexture::D16: + return DXGI_FORMAT_R16_TYPELESS; + case QRhiTexture::D24: + return DXGI_FORMAT_R24G8_TYPELESS; + case QRhiTexture::D24S8: + return DXGI_FORMAT_R24G8_TYPELESS; + case QRhiTexture::D32F: + return DXGI_FORMAT_R32_TYPELESS; + + case QRhiTexture::BC1: + return srgb ? DXGI_FORMAT_BC1_UNORM_SRGB : DXGI_FORMAT_BC1_UNORM; + case QRhiTexture::BC2: + return srgb ? DXGI_FORMAT_BC2_UNORM_SRGB : DXGI_FORMAT_BC2_UNORM; + case QRhiTexture::BC3: + return srgb ? DXGI_FORMAT_BC3_UNORM_SRGB : DXGI_FORMAT_BC3_UNORM; + case QRhiTexture::BC4: + return DXGI_FORMAT_BC4_UNORM; + case QRhiTexture::BC5: + return DXGI_FORMAT_BC5_UNORM; + case QRhiTexture::BC6H: + return DXGI_FORMAT_BC6H_UF16; + case QRhiTexture::BC7: + return srgb ? DXGI_FORMAT_BC7_UNORM_SRGB : DXGI_FORMAT_BC7_UNORM; + + case QRhiTexture::ETC2_RGB8: + case QRhiTexture::ETC2_RGB8A1: + case QRhiTexture::ETC2_RGBA8: + qWarning("QRhiD3D12 does not support ETC2 textures"); + return DXGI_FORMAT_R8G8B8A8_UNORM; + + case QRhiTexture::ASTC_4x4: + case QRhiTexture::ASTC_5x4: + case QRhiTexture::ASTC_5x5: + case QRhiTexture::ASTC_6x5: + case QRhiTexture::ASTC_6x6: + case QRhiTexture::ASTC_8x5: + case QRhiTexture::ASTC_8x6: + case QRhiTexture::ASTC_8x8: + case QRhiTexture::ASTC_10x5: + case QRhiTexture::ASTC_10x6: + case QRhiTexture::ASTC_10x8: + case QRhiTexture::ASTC_10x10: + case QRhiTexture::ASTC_12x10: + case QRhiTexture::ASTC_12x12: + qWarning("QRhiD3D12 does not support ASTC textures"); + return DXGI_FORMAT_R8G8B8A8_UNORM; + + default: + break; + } + return DXGI_FORMAT_R8G8B8A8_UNORM; +} + +QD3D12RenderBuffer::QD3D12RenderBuffer(QRhiImplementation *rhi, + Type type, + const QSize &pixelSize, + int sampleCount, + Flags flags, + QRhiTexture::Format backingFormatHint) + : QRhiRenderBuffer(rhi, type, pixelSize, sampleCount, flags, backingFormatHint) +{ +} + +QD3D12RenderBuffer::~QD3D12RenderBuffer() +{ + destroy(); +} + +void QD3D12RenderBuffer::destroy() +{ + if (handle.isNull()) + return; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) { + if (rtv.isValid()) + rhiD->releaseQueue.deferredReleaseResourceWithViews(handle, &rhiD->rtvPool, rtv, 1); + else if (dsv.isValid()) + rhiD->releaseQueue.deferredReleaseResourceWithViews(handle, &rhiD->dsvPool, dsv, 1); + } + + handle = {}; + rtv = {}; + dsv = {}; + + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12RenderBuffer::create() +{ + if (!handle.isNull()) + destroy(); + + if (m_pixelSize.isEmpty()) + return false; + + QRHI_RES_RHI(QRhiD3D12); + + switch (m_type) { + case QRhiRenderBuffer::Color: + { + dxgiFormat = toD3DTextureFormat(backingFormat(), {}); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; + resourceDesc.Width = UINT64(m_pixelSize.width()); + resourceDesc.Height = UINT(m_pixelSize.height()); + resourceDesc.DepthOrArraySize = 1; + resourceDesc.MipLevels = 1; + resourceDesc.Format = dxgiFormat; + resourceDesc.SampleDesc = sampleDesc; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN; + resourceDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET; + D3D12_CLEAR_VALUE clearValue = {}; + clearValue.Format = dxgiFormat; + // have a separate allocation and resource object (meaning both will need its own Release()) + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + HRESULT hr = rhiD->vma.createResource(D3D12_HEAP_TYPE_DEFAULT, + &resourceDesc, + D3D12_RESOURCE_STATE_RENDER_TARGET, + &clearValue, + &allocation, + __uuidof(ID3D12Resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) { + qWarning("Failed to create color buffer: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + handle = QD3D12Resource::addToPool(&rhiD->resourcePool, resource, D3D12_RESOURCE_STATE_RENDER_TARGET, allocation); + rtv = rhiD->rtvPool.allocate(1); + if (!rtv.isValid()) + return false; + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = dxgiFormat; + rtvDesc.ViewDimension = sampleDesc.Count > 1 ? D3D12_RTV_DIMENSION_TEXTURE2DMS + : D3D12_RTV_DIMENSION_TEXTURE2D; + rhiD->dev->CreateRenderTargetView(resource, &rtvDesc, rtv.cpuHandle); + } + break; + case QRhiRenderBuffer::DepthStencil: + { + dxgiFormat = DS_FORMAT; + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; + resourceDesc.Width = UINT64(m_pixelSize.width()); + resourceDesc.Height = UINT(m_pixelSize.height()); + resourceDesc.DepthOrArraySize = 1; + resourceDesc.MipLevels = 1; + resourceDesc.Format = dxgiFormat; + resourceDesc.SampleDesc = sampleDesc; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN; + resourceDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL; + if (m_flags.testFlag(UsedWithSwapChainOnly)) + resourceDesc.Flags |= D3D12_RESOURCE_FLAG_DENY_SHADER_RESOURCE; + D3D12_CLEAR_VALUE clearValue = {}; + clearValue.Format = dxgiFormat; + clearValue.DepthStencil.Depth = 1.0f; + clearValue.DepthStencil.Stencil = 0; + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + HRESULT hr = rhiD->vma.createResource(D3D12_HEAP_TYPE_DEFAULT, + &resourceDesc, + D3D12_RESOURCE_STATE_DEPTH_WRITE, + &clearValue, + &allocation, + __uuidof(ID3D12Resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) { + qWarning("Failed to create depth-stencil buffer: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + handle = QD3D12Resource::addToPool(&rhiD->resourcePool, resource, D3D12_RESOURCE_STATE_DEPTH_WRITE, allocation); + dsv = rhiD->dsvPool.allocate(1); + if (!dsv.isValid()) + return false; + D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc = {}; + dsvDesc.Format = dxgiFormat; + dsvDesc.ViewDimension = sampleDesc.Count > 1 ? D3D12_DSV_DIMENSION_TEXTURE2DMS + : D3D12_DSV_DIMENSION_TEXTURE2D; + rhiD->dev->CreateDepthStencilView(resource, &dsvDesc, dsv.cpuHandle); + } + break; + } + + if (!m_objectName.isEmpty()) { + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handle)) { + const QString name = QString::fromUtf8(m_objectName); + res->resource->SetName(reinterpret_cast<LPCWSTR>(name.utf16())); + } + } + + generation += 1; + rhiD->registerResource(this); + return true; +} + +QRhiTexture::Format QD3D12RenderBuffer::backingFormat() const +{ + if (m_backingFormatHint != QRhiTexture::UnknownFormat) + return m_backingFormatHint; + else + return m_type == Color ? QRhiTexture::RGBA8 : QRhiTexture::UnknownFormat; +} + +QD3D12Texture::QD3D12Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags) +{ +} + +QD3D12Texture::~QD3D12Texture() +{ + destroy(); +} + +void QD3D12Texture::destroy() +{ + if (handle.isNull()) + return; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->releaseQueue.deferredReleaseResourceWithViews(handle, &rhiD->cbvSrvUavPool, srv, 1); + + handle = {}; + srv = {}; + + if (rhiD) + rhiD->unregisterResource(this); +} + +static inline DXGI_FORMAT toD3DDepthTextureSRVFormat(QRhiTexture::Format format) +{ + switch (format) { + case QRhiTexture::Format::D16: + return DXGI_FORMAT_R16_FLOAT; + case QRhiTexture::Format::D24: + return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + case QRhiTexture::Format::D24S8: + return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + case QRhiTexture::Format::D32F: + return DXGI_FORMAT_R32_FLOAT; + default: + break; + } + Q_UNREACHABLE_RETURN(DXGI_FORMAT_R32_FLOAT); +} + +static inline DXGI_FORMAT toD3DDepthTextureDSVFormat(QRhiTexture::Format format) +{ + // here the result cannot be typeless + switch (format) { + case QRhiTexture::Format::D16: + return DXGI_FORMAT_D16_UNORM; + case QRhiTexture::Format::D24: + return DXGI_FORMAT_D24_UNORM_S8_UINT; + case QRhiTexture::Format::D24S8: + return DXGI_FORMAT_D24_UNORM_S8_UINT; + case QRhiTexture::Format::D32F: + return DXGI_FORMAT_D32_FLOAT; + default: + break; + } + Q_UNREACHABLE_RETURN(DXGI_FORMAT_D32_FLOAT); +} + +static inline bool isDepthTextureFormat(QRhiTexture::Format format) +{ + switch (format) { + case QRhiTexture::Format::D16: + case QRhiTexture::Format::D24: + case QRhiTexture::Format::D24S8: + case QRhiTexture::Format::D32F: + return true; + default: + return false; + } +} + +bool QD3D12Texture::prepareCreate(QSize *adjustedSize) +{ + if (!handle.isNull()) + destroy(); + + const bool isDepth = isDepthTextureFormat(m_format); + const bool isCube = m_flags.testFlag(CubeMap); + const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); + const bool hasMipMaps = m_flags.testFlag(MipMapped); + const bool is1D = m_flags.testFlag(OneDimensional); + + const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); + + dxgiFormat = toD3DTextureFormat(m_format, m_flags); + if (isDepth) { + srvFormat = toD3DDepthTextureSRVFormat(m_format); + rtFormat = toD3DDepthTextureDSVFormat(m_format); + } else { + srvFormat = dxgiFormat; + rtFormat = dxgiFormat; + } + if (m_writeViewFormat.format != UnknownFormat) { + if (isDepth) + rtFormat = toD3DDepthTextureDSVFormat(m_writeViewFormat.format); + else + rtFormat = toD3DTextureFormat(m_writeViewFormat.format, m_writeViewFormat.srgb ? sRGB : Flags()); + } + if (m_readViewFormat.format != UnknownFormat) { + if (isDepth) + srvFormat = toD3DDepthTextureSRVFormat(m_readViewFormat.format); + else + srvFormat = toD3DTextureFormat(m_readViewFormat.format, m_readViewFormat.srgb ? sRGB : Flags()); + } + + QRHI_RES_RHI(QRhiD3D12); + mipLevelCount = uint(hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); + if (sampleDesc.Count > 1) { + if (isCube) { + qWarning("Cubemap texture cannot be multisample"); + return false; + } + if (is3D) { + qWarning("3D texture cannot be multisample"); + return false; + } + if (hasMipMaps) { + qWarning("Multisample texture cannot have mipmaps"); + return false; + } + } + if (isDepth && hasMipMaps) { + qWarning("Depth texture cannot have mipmaps"); + return false; + } + if (isCube && is3D) { + qWarning("Texture cannot be both cube and 3D"); + return false; + } + if (isArray && is3D) { + qWarning("Texture cannot be both array and 3D"); + return false; + } + if (isCube && is1D) { + qWarning("Texture cannot be both cube and 1D"); + return false; + } + if (is1D && is3D) { + qWarning("Texture cannot be both 1D and 3D"); + return false; + } + if (m_depth > 1 && !is3D) { + qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); + return false; + } + if (m_arraySize > 0 && !isArray) { + qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); + return false; + } + if (m_arraySize < 1 && isArray) { + qWarning("Texture is an array but array size is %d", m_arraySize); + return false; + } + + if (adjustedSize) + *adjustedSize = size; + + return true; +} + +bool QD3D12Texture::finishCreate() +{ + QRHI_RES_RHI(QRhiD3D12); + const bool isCube = m_flags.testFlag(CubeMap); + const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); + const bool is1D = m_flags.testFlag(OneDimensional); + + D3D12_SHADER_RESOURCE_VIEW_DESC srvDesc = {}; + srvDesc.Format = srvFormat; + srvDesc.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING; + + if (isCube) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURECUBE; + srvDesc.TextureCube.MipLevels = mipLevelCount; + } else { + if (is1D) { + if (isArray) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1DARRAY; + srvDesc.Texture1DArray.MipLevels = mipLevelCount; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture1DArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture1DArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture1DArray.FirstArraySlice = 0; + srvDesc.Texture1DArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } else { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1D; + srvDesc.Texture1D.MipLevels = mipLevelCount; + } + } else if (isArray) { + if (sampleDesc.Count > 1) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DMSARRAY; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture2DMSArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture2DMSArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture2DMSArray.FirstArraySlice = 0; + srvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } else { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MipLevels = mipLevelCount; + if (m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + srvDesc.Texture2DArray.FirstArraySlice = UINT(m_arrayRangeStart); + srvDesc.Texture2DArray.ArraySize = UINT(m_arrayRangeLength); + } else { + srvDesc.Texture2DArray.FirstArraySlice = 0; + srvDesc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); + } + } + } else { + if (sampleDesc.Count > 1) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DMS; + } else if (is3D) { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D; + srvDesc.Texture3D.MipLevels = mipLevelCount; + } else { + srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MipLevels = mipLevelCount; + } + } + } + + srv = rhiD->cbvSrvUavPool.allocate(1); + if (!srv.isValid()) + return false; + + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handle)) { + rhiD->dev->CreateShaderResourceView(res->resource, &srvDesc, srv.cpuHandle); + if (!m_objectName.isEmpty()) { + const QString name = QString::fromUtf8(m_objectName); + res->resource->SetName(reinterpret_cast<LPCWSTR>(name.utf16())); + } + } else { + return false; + } + + generation += 1; + return true; +} + +bool QD3D12Texture::create() +{ + QSize size; + if (!prepareCreate(&size)) + return false; + + const bool isDepth = isDepthTextureFormat(m_format); + const bool isCube = m_flags.testFlag(CubeMap); + const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); + const bool is1D = m_flags.testFlag(OneDimensional); + + QRHI_RES_RHI(QRhiD3D12); + + bool needsOptimizedClearValueSpecified = false; + UINT resourceFlags = 0; + if (m_flags.testFlag(RenderTarget) || sampleDesc.Count > 1) { + if (isDepth) + resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL; + else + resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET; + needsOptimizedClearValueSpecified = true; + } + if (m_flags.testFlag(UsedWithGenerateMips)) { + if (isDepth) { + qWarning("Depth texture cannot have mipmaps generated"); + return false; + } + resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS; + } + if (m_flags.testFlag(UsedWithLoadStore)) + resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS; + + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = is1D ? D3D12_RESOURCE_DIMENSION_TEXTURE1D + : (is3D ? D3D12_RESOURCE_DIMENSION_TEXTURE3D + : D3D12_RESOURCE_DIMENSION_TEXTURE2D); + resourceDesc.Width = UINT64(size.width()); + resourceDesc.Height = UINT(size.height()); + resourceDesc.DepthOrArraySize = isCube ? 6 + : (isArray ? UINT(qMax(0, m_arraySize)) + : (is3D ? qMax(1, m_depth) + : 1)); + resourceDesc.MipLevels = mipLevelCount; + resourceDesc.Format = dxgiFormat; + resourceDesc.SampleDesc = sampleDesc; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN; + resourceDesc.Flags = D3D12_RESOURCE_FLAGS(resourceFlags); + D3D12_CLEAR_VALUE clearValue = {}; + clearValue.Format = dxgiFormat; + if (isDepth) { + clearValue.Format = toD3DDepthTextureDSVFormat(m_format); + clearValue.DepthStencil.Depth = 1.0f; + clearValue.DepthStencil.Stencil = 0; + } + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + HRESULT hr = rhiD->vma.createResource(D3D12_HEAP_TYPE_DEFAULT, + &resourceDesc, + D3D12_RESOURCE_STATE_COMMON, + needsOptimizedClearValueSpecified ? &clearValue : nullptr, + &allocation, + __uuidof(ID3D12Resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) { + qWarning("Failed to create texture: '%s'" + " Dim was %d Size was %ux%u Depth/ArraySize was %u MipLevels was %u Format was %d Sample count was %d", + qPrintable(QSystemError::windowsComString(hr)), + int(resourceDesc.Dimension), + uint(resourceDesc.Width), + uint(resourceDesc.Height), + uint(resourceDesc.DepthOrArraySize), + uint(resourceDesc.MipLevels), + int(resourceDesc.Format), + int(resourceDesc.SampleDesc.Count)); + return false; + } + + handle = QD3D12Resource::addToPool(&rhiD->resourcePool, resource, D3D12_RESOURCE_STATE_COMMON, allocation); + + if (!finishCreate()) + return false; + + rhiD->registerResource(this); + return true; +} + +bool QD3D12Texture::createFrom(QRhiTexture::NativeTexture src) +{ + if (!src.object) + return false; + + if (!prepareCreate()) + return false; + + ID3D12Resource *resource = reinterpret_cast<ID3D12Resource *>(src.object); + D3D12_RESOURCE_STATES state = D3D12_RESOURCE_STATES(src.layout); + + QRHI_RES_RHI(QRhiD3D12); + handle = QD3D12Resource::addNonOwningToPool(&rhiD->resourcePool, resource, state); + + if (!finishCreate()) + return false; + + rhiD->registerResource(this); + return true; +} + +QRhiTexture::NativeTexture QD3D12Texture::nativeTexture() +{ + QRHI_RES_RHI(QRhiD3D12); + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handle)) + return { quint64(res->resource), int(res->state) }; + + return {}; +} + +void QD3D12Texture::setNativeLayout(int layout) +{ + QRHI_RES_RHI(QRhiD3D12); + if (QD3D12Resource *res = rhiD->resourcePool.lookupRef(handle)) + res->state = D3D12_RESOURCE_STATES(layout); +} + +QD3D12Sampler::QD3D12Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w) + : QRhiSampler(rhi, magFilter, minFilter, mipmapMode, u, v, w) +{ +} + +QD3D12Sampler::~QD3D12Sampler() +{ + destroy(); +} + +void QD3D12Sampler::destroy() +{ + shaderVisibleDescriptor = {}; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); +} + +static inline D3D12_FILTER toD3DFilter(QRhiSampler::Filter minFilter, QRhiSampler::Filter magFilter, QRhiSampler::Filter mipFilter) +{ + if (minFilter == QRhiSampler::Nearest) { + if (magFilter == QRhiSampler::Nearest) { + if (mipFilter == QRhiSampler::Linear) + return D3D12_FILTER_MIN_MAG_POINT_MIP_LINEAR; + else + return D3D12_FILTER_MIN_MAG_MIP_POINT; + } else { + if (mipFilter == QRhiSampler::Linear) + return D3D12_FILTER_MIN_POINT_MAG_MIP_LINEAR; + else + return D3D12_FILTER_MIN_POINT_MAG_LINEAR_MIP_POINT; + } + } else { + if (magFilter == QRhiSampler::Nearest) { + if (mipFilter == QRhiSampler::Linear) + return D3D12_FILTER_MIN_LINEAR_MAG_POINT_MIP_LINEAR; + else + return D3D12_FILTER_MIN_LINEAR_MAG_MIP_POINT; + } else { + if (mipFilter == QRhiSampler::Linear) + return D3D12_FILTER_MIN_MAG_MIP_LINEAR; + else + return D3D12_FILTER_MIN_MAG_LINEAR_MIP_POINT; + } + } + Q_UNREACHABLE_RETURN(D3D12_FILTER_MIN_MAG_MIP_LINEAR); +} + +static inline D3D12_TEXTURE_ADDRESS_MODE toD3DAddressMode(QRhiSampler::AddressMode m) +{ + switch (m) { + case QRhiSampler::Repeat: + return D3D12_TEXTURE_ADDRESS_MODE_WRAP; + case QRhiSampler::ClampToEdge: + return D3D12_TEXTURE_ADDRESS_MODE_CLAMP; + case QRhiSampler::Mirror: + return D3D12_TEXTURE_ADDRESS_MODE_MIRROR; + } + Q_UNREACHABLE_RETURN(D3D12_TEXTURE_ADDRESS_MODE_CLAMP); +} + +static inline D3D12_COMPARISON_FUNC toD3DTextureComparisonFunc(QRhiSampler::CompareOp op) +{ + switch (op) { + case QRhiSampler::Never: + return D3D12_COMPARISON_FUNC_NEVER; + case QRhiSampler::Less: + return D3D12_COMPARISON_FUNC_LESS; + case QRhiSampler::Equal: + return D3D12_COMPARISON_FUNC_EQUAL; + case QRhiSampler::LessOrEqual: + return D3D12_COMPARISON_FUNC_LESS_EQUAL; + case QRhiSampler::Greater: + return D3D12_COMPARISON_FUNC_GREATER; + case QRhiSampler::NotEqual: + return D3D12_COMPARISON_FUNC_NOT_EQUAL; + case QRhiSampler::GreaterOrEqual: + return D3D12_COMPARISON_FUNC_GREATER_EQUAL; + case QRhiSampler::Always: + return D3D12_COMPARISON_FUNC_ALWAYS; + } + Q_UNREACHABLE_RETURN(D3D12_COMPARISON_FUNC_NEVER); +} + +bool QD3D12Sampler::create() +{ + desc = {}; + desc.Filter = toD3DFilter(m_minFilter, m_magFilter, m_mipmapMode); + if (m_compareOp != Never) + desc.Filter = D3D12_FILTER(desc.Filter | 0x80); + desc.AddressU = toD3DAddressMode(m_addressU); + desc.AddressV = toD3DAddressMode(m_addressV); + desc.AddressW = toD3DAddressMode(m_addressW); + desc.MaxAnisotropy = 1.0f; + desc.ComparisonFunc = toD3DTextureComparisonFunc(m_compareOp); + desc.MaxLOD = m_mipmapMode == None ? 0.0f : 10000.0f; + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(this, false); + return true; +} + +QD3D12Descriptor QD3D12Sampler::lookupOrCreateShaderVisibleDescriptor() +{ + if (!shaderVisibleDescriptor.isValid()) { + QRHI_RES_RHI(QRhiD3D12); + shaderVisibleDescriptor = rhiD->samplerMgr.getShaderVisibleDescriptor(desc); + } + return shaderVisibleDescriptor; +} + +QD3D12TextureRenderTarget::QD3D12TextureRenderTarget(QRhiImplementation *rhi, + const QRhiTextureRenderTargetDescription &desc, + Flags flags) + : QRhiTextureRenderTarget(rhi, desc, flags), + d(rhi) +{ +} + +QD3D12TextureRenderTarget::~QD3D12TextureRenderTarget() +{ + destroy(); +} + +void QD3D12TextureRenderTarget::destroy() +{ + if (!rtv[0].isValid() && !dsv.isValid()) + return; + + QRHI_RES_RHI(QRhiD3D12); + if (dsv.isValid()) { + if (ownsDsv && rhiD) + rhiD->releaseQueue.deferredReleaseViews(&rhiD->dsvPool, dsv, 1); + dsv = {}; + } + + for (int i = 0; i < QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS; ++i) { + if (rtv[i].isValid()) { + if (ownsRtv[i] && rhiD) + rhiD->releaseQueue.deferredReleaseViews(&rhiD->rtvPool, rtv[i], 1); + rtv[i] = {}; + } + } + + if (rhiD) + rhiD->unregisterResource(this); +} + +QRhiRenderPassDescriptor *QD3D12TextureRenderTarget::newCompatibleRenderPassDescriptor() +{ + // not yet built so cannot rely on data computed in create() + + QD3D12RenderPassDescriptor *rpD = new QD3D12RenderPassDescriptor(m_rhi); + + rpD->colorAttachmentCount = 0; + for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, it->texture()); + QD3D12RenderBuffer *rbD = QRHI_RES(QD3D12RenderBuffer, it->renderBuffer()); + if (texD) + rpD->colorFormat[rpD->colorAttachmentCount] = texD->rtFormat; + else if (rbD) + rpD->colorFormat[rpD->colorAttachmentCount] = rbD->dxgiFormat; + rpD->colorAttachmentCount += 1; + } + + rpD->hasDepthStencil = false; + if (m_desc.depthStencilBuffer()) { + rpD->hasDepthStencil = true; + rpD->dsFormat = QD3D12RenderBuffer::DS_FORMAT; + } else if (m_desc.depthTexture()) { + QD3D12Texture *depthTexD = QRHI_RES(QD3D12Texture, m_desc.depthTexture()); + rpD->hasDepthStencil = true; + rpD->dsFormat = toD3DDepthTextureDSVFormat(depthTexD->format()); // cannot be a typeless format + } + + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); + return rpD; +} + +bool QD3D12TextureRenderTarget::create() +{ + if (rtv[0].isValid() || dsv.isValid()) + destroy(); + + QRHI_RES_RHI(QRhiD3D12); + Q_ASSERT(m_desc.colorAttachmentCount() > 0 || m_desc.depthTexture()); + Q_ASSERT(!m_desc.depthStencilBuffer() || !m_desc.depthTexture()); + const bool hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); + d.colorAttCount = 0; + int attIndex = 0; + + for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it, ++attIndex) { + d.colorAttCount += 1; + const QRhiColorAttachment &colorAtt(*it); + QRhiTexture *texture = colorAtt.texture(); + QRhiRenderBuffer *rb = colorAtt.renderBuffer(); + Q_ASSERT(texture || rb); + if (texture) { + QD3D12Texture *texD = QRHI_RES(QD3D12Texture, texture); + QD3D12Resource *res = rhiD->resourcePool.lookupRef(texD->handle); + if (!res) { + qWarning("Could not look up texture handle for render target"); + return false; + } + const bool isMultiView = it->multiViewCount() >= 2; + UINT layerCount = isMultiView ? UINT(it->multiViewCount()) : 1; + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = texD->rtFormat; + if (texD->flags().testFlag(QRhiTexture::CubeMap)) { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture2DArray.ArraySize = layerCount; + } else if (texD->flags().testFlag(QRhiTexture::OneDimensional)) { + if (texD->flags().testFlag(QRhiTexture::TextureArray)) { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE1DARRAY; + rtvDesc.Texture1DArray.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture1DArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture1DArray.ArraySize = layerCount; + } else { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE1D; + rtvDesc.Texture1D.MipSlice = UINT(colorAtt.level()); + } + } else if (texD->flags().testFlag(QRhiTexture::TextureArray)) { + if (texD->sampleDesc.Count > 1) { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DMSARRAY; + rtvDesc.Texture2DMSArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture2DMSArray.ArraySize = layerCount; + } else { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); + rtvDesc.Texture2DArray.ArraySize = layerCount; + } + } else if (texD->flags().testFlag(QRhiTexture::ThreeDimensional)) { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE3D; + rtvDesc.Texture3D.MipSlice = UINT(colorAtt.level()); + rtvDesc.Texture3D.FirstWSlice = UINT(colorAtt.layer()); + rtvDesc.Texture3D.WSize = layerCount; + } else { + if (texD->sampleDesc.Count > 1) { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DMS; + } else { + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2D; + rtvDesc.Texture2D.MipSlice = UINT(colorAtt.level()); + } + } + rtv[attIndex] = rhiD->rtvPool.allocate(1); + if (!rtv[attIndex].isValid()) { + qWarning("Failed to allocate RTV for texture render target"); + return false; + } + rhiD->dev->CreateRenderTargetView(res->resource, &rtvDesc, rtv[attIndex].cpuHandle); + ownsRtv[attIndex] = true; + if (attIndex == 0) { + d.pixelSize = rhiD->q->sizeForMipLevel(colorAtt.level(), texD->pixelSize()); + d.sampleCount = int(texD->sampleDesc.Count); + } + } else if (rb) { + QD3D12RenderBuffer *rbD = QRHI_RES(QD3D12RenderBuffer, rb); + ownsRtv[attIndex] = false; + rtv[attIndex] = rbD->rtv; + if (attIndex == 0) { + d.pixelSize = rbD->pixelSize(); + d.sampleCount = int(rbD->sampleDesc.Count); + } + } + } + + d.dpr = 1; + + if (hasDepthStencil) { + if (m_desc.depthTexture()) { + ownsDsv = true; + QD3D12Texture *depthTexD = QRHI_RES(QD3D12Texture, m_desc.depthTexture()); + QD3D12Resource *res = rhiD->resourcePool.lookupRef(depthTexD->handle); + if (!res) { + qWarning("Could not look up depth texture handle"); + return false; + } + D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc = {}; + dsvDesc.Format = depthTexD->rtFormat; + dsvDesc.ViewDimension = depthTexD->sampleDesc.Count > 1 ? D3D12_DSV_DIMENSION_TEXTURE2DMS + : D3D12_DSV_DIMENSION_TEXTURE2D; + if (depthTexD->flags().testFlag(QRhiTexture::TextureArray)) { + if (depthTexD->sampleDesc.Count > 1) { + dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DMSARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DMSArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DMSArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DMSArray.FirstArraySlice = 0; + dsvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } else { + dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DArray.FirstArraySlice = 0; + dsvDesc.Texture2DArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } + } + dsv = rhiD->dsvPool.allocate(1); + if (!dsv.isValid()) { + qWarning("Failed to allocate DSV for texture render target"); + return false; + } + rhiD->dev->CreateDepthStencilView(res->resource, &dsvDesc, dsv.cpuHandle); + if (d.colorAttCount == 0) { + d.pixelSize = depthTexD->pixelSize(); + d.sampleCount = int(depthTexD->sampleDesc.Count); + } + } else { + ownsDsv = false; + QD3D12RenderBuffer *depthRbD = QRHI_RES(QD3D12RenderBuffer, m_desc.depthStencilBuffer()); + dsv = depthRbD->dsv; + if (d.colorAttCount == 0) { + d.pixelSize = m_desc.depthStencilBuffer()->pixelSize(); + d.sampleCount = int(depthRbD->sampleDesc.Count); + } + } + d.dsAttCount = 1; + } else { + d.dsAttCount = 0; + } + + D3D12_CPU_DESCRIPTOR_HANDLE nullDescHandle = { 0 }; + for (int i = 0; i < QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS; ++i) + d.rtv[i] = i < d.colorAttCount ? rtv[i].cpuHandle : nullDescHandle; + d.dsv = dsv.cpuHandle; + d.rp = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); + + QRhiRenderTargetAttachmentTracker::updateResIdList<QD3D12Texture, QD3D12RenderBuffer>(m_desc, &d.currentResIdList); + + rhiD->registerResource(this); + return true; +} + +QSize QD3D12TextureRenderTarget::pixelSize() const +{ + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QD3D12Texture, QD3D12RenderBuffer>(m_desc, d.currentResIdList)) + const_cast<QD3D12TextureRenderTarget *>(this)->create(); + + return d.pixelSize; +} + +float QD3D12TextureRenderTarget::devicePixelRatio() const +{ + return d.dpr; +} + +int QD3D12TextureRenderTarget::sampleCount() const +{ + return d.sampleCount; +} + +QD3D12ShaderResourceBindings::QD3D12ShaderResourceBindings(QRhiImplementation *rhi) + : QRhiShaderResourceBindings(rhi) +{ +} + +QD3D12ShaderResourceBindings::~QD3D12ShaderResourceBindings() +{ + destroy(); +} + +void QD3D12ShaderResourceBindings::destroy() +{ + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12ShaderResourceBindings::create() +{ + QRHI_RES_RHI(QRhiD3D12); + if (!rhiD->sanityCheckShaderResourceBindings(this)) + return false; + + rhiD->updateLayoutDesc(this); + + hasDynamicOffset = false; + for (const QRhiShaderResourceBinding &b : std::as_const(m_bindings)) { + const QRhiShaderResourceBinding::Data *bd = QRhiImplementation::shaderResourceBindingData(b); + if (bd->type == QRhiShaderResourceBinding::UniformBuffer && bd->u.ubuf.hasDynamicOffset) { + hasDynamicOffset = true; + break; + } + } + + // The root signature is not part of the srb. Unintuitive, but the shader + // translation pipeline ties our hands: as long as the per-shader (so per + // stage!) nativeResourceBindingMap exist, meaning f.ex. that a SPIR-V + // combined image sampler binding X passed in here may map to the tY and sY + // HLSL registers, where Y is known only once the mapping table from the + // shader is looked up. Creating a root parameters at this stage is + // therefore impossible. + + generation += 1; + rhiD->registerResource(this, false); + return true; +} + +void QD3D12ShaderResourceBindings::updateResources(UpdateFlags flags) +{ + Q_UNUSED(flags); + generation += 1; +} + +// Accessing the QRhiBuffer/Texture/Sampler resources must be avoided in the +// callbacks; that would only be possible if the srb had those specified, and +// that's not required at the time of srb and pipeline create() time, and +// createRootSignature is called from the pipeline create(). + +void QD3D12ShaderResourceBindings::visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &, + int shaderRegister, + int) +{ + D3D12_ROOT_PARAMETER1 rootParam = {}; + rootParam.ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV; + rootParam.ShaderVisibility = qd3d12_stageToVisibility(s); + rootParam.Descriptor.ShaderRegister = shaderRegister; + rootParam.Descriptor.Flags = D3D12_ROOT_DESCRIPTOR_FLAG_DATA_STATIC; + visitorData.cbParams[s].append(rootParam); +} + +void QD3D12ShaderResourceBindings::visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &, + int shaderRegister) +{ + D3D12_DESCRIPTOR_RANGE1 range = {}; + range.RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_SRV; + range.NumDescriptors = 1; + range.BaseShaderRegister = shaderRegister; + range.OffsetInDescriptorsFromTableStart = visitorData.currentSrvRangeOffset[s]; + visitorData.currentSrvRangeOffset[s] += 1; + visitorData.srvRanges[s].append(range); + if (visitorData.srvRanges[s].count() == 1) { + visitorData.srvTables[s].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + visitorData.srvTables[s].ShaderVisibility = qd3d12_stageToVisibility(s); + } +} + +void QD3D12ShaderResourceBindings::visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &, + int shaderRegister) +{ + // Unlike SRVs and UAVs, samplers are handled so that each sampler becomes + // a root parameter with its own descriptor table. + + int &rangeStoreIdx(visitorData.samplerRangeHeads[s]); + if (rangeStoreIdx == 16) { + qWarning("Sampler count in QD3D12Stage %d exceeds the limit of 16, this is disallowed by QRhi", s); + return; + } + D3D12_DESCRIPTOR_RANGE1 range = {}; + range.RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_SAMPLER; + range.NumDescriptors = 1; + range.BaseShaderRegister = shaderRegister; + visitorData.samplerRanges[s][rangeStoreIdx] = range; + D3D12_ROOT_PARAMETER1 param = {}; + param.ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + param.ShaderVisibility = qd3d12_stageToVisibility(s); + param.DescriptorTable.NumDescriptorRanges = 1; + param.DescriptorTable.pDescriptorRanges = &visitorData.samplerRanges[s][rangeStoreIdx]; + rangeStoreIdx += 1; + visitorData.samplerTables[s].append(param); +} + +void QD3D12ShaderResourceBindings::visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &, + QD3D12ShaderResourceVisitor::StorageOp, + int shaderRegister) +{ + D3D12_DESCRIPTOR_RANGE1 range = {}; + range.RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_UAV; + range.NumDescriptors = 1; + range.BaseShaderRegister = shaderRegister; + range.OffsetInDescriptorsFromTableStart = visitorData.currentUavRangeOffset[s]; + visitorData.currentUavRangeOffset[s] += 1; + visitorData.uavRanges[s].append(range); + if (visitorData.uavRanges[s].count() == 1) { + visitorData.uavTables[s].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + visitorData.uavTables[s].ShaderVisibility = qd3d12_stageToVisibility(s); + } +} + +void QD3D12ShaderResourceBindings::visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &, + QD3D12ShaderResourceVisitor::StorageOp, + int shaderRegister) +{ + D3D12_DESCRIPTOR_RANGE1 range = {}; + range.RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_UAV; + range.NumDescriptors = 1; + range.BaseShaderRegister = shaderRegister; + range.OffsetInDescriptorsFromTableStart = visitorData.currentUavRangeOffset[s]; + visitorData.currentUavRangeOffset[s] += 1; + visitorData.uavRanges[s].append(range); + if (visitorData.uavRanges[s].count() == 1) { + visitorData.uavTables[s].ParameterType = D3D12_ROOT_PARAMETER_TYPE_DESCRIPTOR_TABLE; + visitorData.uavTables[s].ShaderVisibility = qd3d12_stageToVisibility(s); + } +} + +QD3D12ObjectHandle QD3D12ShaderResourceBindings::createRootSignature(const QD3D12ShaderStageData *stageData, + int stageCount) +{ + QRHI_RES_RHI(QRhiD3D12); + + // It's not just that the root signature has to be tied to the pipeline + // (cannot just freely create it like e.g. with Vulkan where one just + // creates a descriptor layout 1:1 with the QRhiShaderResourceBindings' + // data), due to not knowing the shader-specific resource binding mapping + // tables at the point of srb creation, but each shader stage may have a + // different mapping table. (ugh!) + // + // Hence we set up everything per-stage, even if it means the root + // signature gets unnecessarily big. (note that the magic is in the + // ShaderVisibility: even though the register range is the same in the + // descriptor tables, the visibility is different) + + QD3D12ShaderResourceVisitor visitor(this, stageData, stageCount); + + visitorData = {}; + + using namespace std::placeholders; + visitor.uniformBuffer = std::bind(&QD3D12ShaderResourceBindings::visitUniformBuffer, this, _1, _2, _3, _4); + visitor.texture = std::bind(&QD3D12ShaderResourceBindings::visitTexture, this, _1, _2, _3); + visitor.sampler = std::bind(&QD3D12ShaderResourceBindings::visitSampler, this, _1, _2, _3); + visitor.storageBuffer = std::bind(&QD3D12ShaderResourceBindings::visitStorageBuffer, this, _1, _2, _3, _4); + visitor.storageImage = std::bind(&QD3D12ShaderResourceBindings::visitStorageImage, this, _1, _2, _3, _4); + + visitor.visit(); + + // The maximum size of a root signature is 256 bytes, where a descriptor + // table is 4, a root descriptor (e.g. CBV) is 8. We have 5 stages at most + // (or 1 with compute) and a separate descriptor table for SRVs (-> + // textures) and UAVs (-> storage buffers and images) per stage, plus each + // uniform buffer counts as a CBV in the stages it is visible. + // + // Due to the limited maximum size of a shader-visible sampler heap (2048) + // and the potential costly switching of descriptor heaps, each sampler is + // declared as a separate root parameter / descriptor table (meaning that + // two samplers in the same stage are two parameters and two tables, not + // just one). QRhi documents a hard limit of 16 on texture/sampler bindings + // in a shader (matching D3D11), so we can hopefully get away with this. + // + // This means that e.g. a vertex+fragment shader with a uniform buffer + // visible in both and one texture+sampler in the fragment shader would + // consume 2*8 + 4 + 4 = 24 bytes. This also implies that clients + // specifying the minimal stage bit mask for each entry in + // QRhiShaderResourceBindings are ideal for this backend since it helps + // reducing the chance of hitting the size limit. + + QVarLengthArray<D3D12_ROOT_PARAMETER1, 4> rootParams; + for (int s = 0; s < 6; ++s) { + if (!visitorData.cbParams[s].isEmpty()) + rootParams.append(visitorData.cbParams[s].constData(), visitorData.cbParams[s].count()); + } + for (int s = 0; s < 6; ++s) { + if (!visitorData.srvRanges[s].isEmpty()) { + visitorData.srvTables[s].DescriptorTable.NumDescriptorRanges = visitorData.srvRanges[s].count(); + visitorData.srvTables[s].DescriptorTable.pDescriptorRanges = visitorData.srvRanges[s].constData(); + rootParams.append(visitorData.srvTables[s]); + } + } + for (int s = 0; s < 6; ++s) { + if (!visitorData.samplerTables[s].isEmpty()) + rootParams.append(visitorData.samplerTables[s].constData(), visitorData.samplerTables[s].count()); + } + for (int s = 0; s < 6; ++s) { + if (!visitorData.uavRanges[s].isEmpty()) { + visitorData.uavTables[s].DescriptorTable.NumDescriptorRanges = visitorData.uavRanges[s].count(); + visitorData.uavTables[s].DescriptorTable.pDescriptorRanges = visitorData.uavRanges[s].constData(); + rootParams.append(visitorData.uavTables[s]); + } + } + + D3D12_VERSIONED_ROOT_SIGNATURE_DESC rsDesc = {}; + rsDesc.Version = D3D_ROOT_SIGNATURE_VERSION_1_1; + if (!rootParams.isEmpty()) { + rsDesc.Desc_1_1.NumParameters = rootParams.count(); + rsDesc.Desc_1_1.pParameters = rootParams.constData(); + } + + UINT rsFlags = 0; + for (int stageIdx = 0; stageIdx < stageCount; ++stageIdx) { + if (stageData[stageIdx].valid && stageData[stageIdx].stage == VS) + rsFlags |= D3D12_ROOT_SIGNATURE_FLAG_ALLOW_INPUT_ASSEMBLER_INPUT_LAYOUT; + } + rsDesc.Desc_1_1.Flags = D3D12_ROOT_SIGNATURE_FLAGS(rsFlags); + + ID3DBlob *signature = nullptr; + HRESULT hr = D3D12SerializeVersionedRootSignature(&rsDesc, &signature, nullptr); + if (FAILED(hr)) { + qWarning("Failed to serialize root signature: %s", qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + ID3D12RootSignature *rootSig = nullptr; + hr = rhiD->dev->CreateRootSignature(0, + signature->GetBufferPointer(), + signature->GetBufferSize(), + __uuidof(ID3D12RootSignature), + reinterpret_cast<void **>(&rootSig)); + signature->Release(); + if (FAILED(hr)) { + qWarning("Failed to create root signature: %s", qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + + return QD3D12RootSignature::addToPool(&rhiD->rootSignaturePool, rootSig); +} + +// For shader model < 6.0 we do the same as the D3D11 backend: use the old +// compiler (D3DCompile) to generate DXBC, just as qsb does (when -c is passed) +// by invoking fxc, not dxc. For SM >= 6.0 we have to use the new compiler and +// work with DXIL. And that involves IDxcCompiler and needs the presence of +// dxcompiler.dll and dxil.dll at runtime. Plus there's a chance we have +// ancient SDK headers when not using MSVC. So this is heavily optional, +// meaning support for dxc can be disabled both at build time (no dxcapi.h) and +// at run time (no DLLs). + +static inline void makeHlslTargetString(char target[7], const char stage[3], int version) +{ + const int smMajor = version / 10; + const int smMinor = version % 10; + target[0] = stage[0]; + target[1] = stage[1]; + target[2] = '_'; + target[3] = '0' + smMajor; + target[4] = '_'; + target[5] = '0' + smMinor; + target[6] = '\0'; +} + +enum class HlslCompileFlag +{ + WithDebugInfo = 0x01 +}; + +static QByteArray legacyCompile(const QShaderCode &hlslSource, const char *target, int flags, QString *error) +{ + static const pD3DCompile d3dCompile = QRhiD3D::resolveD3DCompile(); + if (!d3dCompile) { + qWarning("Unable to resolve function D3DCompile()"); + return QByteArray(); + } + + ID3DBlob *bytecode = nullptr; + ID3DBlob *errors = nullptr; + UINT d3dCompileFlags = 0; + if (flags & int(HlslCompileFlag::WithDebugInfo)) + d3dCompileFlags |= D3DCOMPILE_DEBUG; + + HRESULT hr = d3dCompile(hlslSource.shader().constData(), SIZE_T(hlslSource.shader().size()), + nullptr, nullptr, nullptr, + hlslSource.entryPoint().constData(), target, d3dCompileFlags, 0, &bytecode, &errors); + if (FAILED(hr) || !bytecode) { + qWarning("HLSL shader compilation failed: 0x%x", uint(hr)); + if (errors) { + *error = QString::fromUtf8(static_cast<const char *>(errors->GetBufferPointer()), + int(errors->GetBufferSize())); + errors->Release(); + } + return QByteArray(); + } + + QByteArray result; + result.resize(int(bytecode->GetBufferSize())); + memcpy(result.data(), bytecode->GetBufferPointer(), size_t(result.size())); + bytecode->Release(); + return result; +} + +#ifdef QRHI_D3D12_HAS_DXC + +#ifndef DXC_CP_UTF8 +#define DXC_CP_UTF8 65001 +#endif + +#ifndef DXC_ARG_DEBUG +#define DXC_ARG_DEBUG L"-Zi" +#endif + +static QByteArray dxcCompile(const QShaderCode &hlslSource, const char *target, int flags, QString *error) +{ + static std::pair<IDxcCompiler *, IDxcLibrary *> dxc = QRhiD3D::createDxcCompiler(); + IDxcCompiler *compiler = dxc.first; + if (!compiler) { + qWarning("Unable to instantiate IDxcCompiler. Likely no dxcompiler.dll and dxil.dll present. " + "Use windeployqt or try https://github.com/microsoft/DirectXShaderCompiler/releases"); + return QByteArray(); + } + IDxcLibrary *library = dxc.second; + if (!library) + return QByteArray(); + + IDxcBlobEncoding *sourceBlob = nullptr; + HRESULT hr = library->CreateBlobWithEncodingOnHeapCopy(hlslSource.shader().constData(), + UINT32(hlslSource.shader().size()), + DXC_CP_UTF8, + &sourceBlob); + if (FAILED(hr)) { + qWarning("Failed to create source blob for dxc: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + return QByteArray(); + } + + const QString entryPointStr = QString::fromLatin1(hlslSource.entryPoint()); + const QString targetStr = QString::fromLatin1(target); + + QVarLengthArray<LPCWSTR, 4> argPtrs; + QString debugArg; + if (flags & int(HlslCompileFlag::WithDebugInfo)) { + debugArg = QString::fromUtf16(reinterpret_cast<const char16_t *>(DXC_ARG_DEBUG)); + argPtrs.append(reinterpret_cast<LPCWSTR>(debugArg.utf16())); + } + + IDxcOperationResult *result = nullptr; + hr = compiler->Compile(sourceBlob, + nullptr, + reinterpret_cast<LPCWSTR>(entryPointStr.utf16()), + reinterpret_cast<LPCWSTR>(targetStr.utf16()), + argPtrs.data(), argPtrs.count(), + nullptr, 0, + nullptr, + &result); + sourceBlob->Release(); + if (SUCCEEDED(hr)) + result->GetStatus(&hr); + if (FAILED(hr)) { + qWarning("HLSL shader compilation failed: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + if (result) { + IDxcBlobEncoding *errorsBlob = nullptr; + if (SUCCEEDED(result->GetErrorBuffer(&errorsBlob))) { + if (errorsBlob) { + *error = QString::fromUtf8(static_cast<const char *>(errorsBlob->GetBufferPointer()), + int(errorsBlob->GetBufferSize())); + errorsBlob->Release(); + } + } + } + return QByteArray(); + } + + IDxcBlob *bytecode = nullptr; + if FAILED(result->GetResult(&bytecode)) { + qWarning("No result from IDxcCompiler: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + return QByteArray(); + } + + QByteArray ba; + ba.resize(int(bytecode->GetBufferSize())); + memcpy(ba.data(), bytecode->GetBufferPointer(), size_t(ba.size())); + bytecode->Release(); + return ba; +} + +#endif // QRHI_D3D12_HAS_DXC + +static QByteArray compileHlslShaderSource(const QShader &shader, + QShader::Variant shaderVariant, + int flags, + QString *error, + QShaderKey *usedShaderKey) +{ + // look for SM 6.7, 6.6, .., 5.0 + const int shaderModelMax = 67; + for (int sm = shaderModelMax; sm >= 50; --sm) { + for (QShader::Source type : { QShader::DxilShader, QShader::DxbcShader }) { + QShaderKey key = { type, sm, shaderVariant }; + QShaderCode intermediateBytecodeShader = shader.shader(key); + if (!intermediateBytecodeShader.shader().isEmpty()) { + if (usedShaderKey) + *usedShaderKey = key; + return intermediateBytecodeShader.shader(); + } + } + } + + QShaderCode hlslSource; + QShaderKey key; + for (int sm = shaderModelMax; sm >= 50; --sm) { + key = { QShader::HlslShader, sm, shaderVariant }; + hlslSource = shader.shader(key); + if (!hlslSource.shader().isEmpty()) + break; + } + + if (hlslSource.shader().isEmpty()) { + qWarning() << "No HLSL (shader model 6.7..5.0) code found in baked shader" << shader; + return QByteArray(); + } + + if (usedShaderKey) + *usedShaderKey = key; + + char target[7]; + switch (shader.stage()) { + case QShader::VertexStage: + makeHlslTargetString(target, "vs", key.sourceVersion().version()); + break; + case QShader::TessellationControlStage: + makeHlslTargetString(target, "hs", key.sourceVersion().version()); + break; + case QShader::TessellationEvaluationStage: + makeHlslTargetString(target, "ds", key.sourceVersion().version()); + break; + case QShader::GeometryStage: + makeHlslTargetString(target, "gs", key.sourceVersion().version()); + break; + case QShader::FragmentStage: + makeHlslTargetString(target, "ps", key.sourceVersion().version()); + break; + case QShader::ComputeStage: + makeHlslTargetString(target, "cs", key.sourceVersion().version()); + break; + } + + if (key.sourceVersion().version() >= 60) { +#ifdef QRHI_D3D12_HAS_DXC + return dxcCompile(hlslSource, target, flags, error); +#else + qWarning("Attempted to runtime-compile HLSL source code for shader model >= 6.0 " + "but the Qt build has no support for DXC. " + "Rebuild Qt with a recent Windows SDK or switch to an MSVC build."); +#endif + } + + return legacyCompile(hlslSource, target, flags, error); +} + +static inline UINT8 toD3DColorWriteMask(QRhiGraphicsPipeline::ColorMask c) +{ + UINT8 f = 0; + if (c.testFlag(QRhiGraphicsPipeline::R)) + f |= D3D12_COLOR_WRITE_ENABLE_RED; + if (c.testFlag(QRhiGraphicsPipeline::G)) + f |= D3D12_COLOR_WRITE_ENABLE_GREEN; + if (c.testFlag(QRhiGraphicsPipeline::B)) + f |= D3D12_COLOR_WRITE_ENABLE_BLUE; + if (c.testFlag(QRhiGraphicsPipeline::A)) + f |= D3D12_COLOR_WRITE_ENABLE_ALPHA; + return f; +} + +static inline D3D12_BLEND toD3DBlendFactor(QRhiGraphicsPipeline::BlendFactor f, bool rgb) +{ + // SrcBlendAlpha and DstBlendAlpha do not accept *_COLOR. With other APIs + // this is handled internally (so that e.g. VK_BLEND_FACTOR_SRC_COLOR is + // accepted and is in effect equivalent to VK_BLEND_FACTOR_SRC_ALPHA when + // set as an alpha src/dest factor), but for D3D we have to take care of it + // ourselves. Hence the rgb argument. + + switch (f) { + case QRhiGraphicsPipeline::Zero: + return D3D12_BLEND_ZERO; + case QRhiGraphicsPipeline::One: + return D3D12_BLEND_ONE; + case QRhiGraphicsPipeline::SrcColor: + return rgb ? D3D12_BLEND_SRC_COLOR : D3D12_BLEND_SRC_ALPHA; + case QRhiGraphicsPipeline::OneMinusSrcColor: + return rgb ? D3D12_BLEND_INV_SRC_COLOR : D3D12_BLEND_INV_SRC_ALPHA; + case QRhiGraphicsPipeline::DstColor: + return rgb ? D3D12_BLEND_DEST_COLOR : D3D12_BLEND_DEST_ALPHA; + case QRhiGraphicsPipeline::OneMinusDstColor: + return rgb ? D3D12_BLEND_INV_DEST_COLOR : D3D12_BLEND_INV_DEST_ALPHA; + case QRhiGraphicsPipeline::SrcAlpha: + return D3D12_BLEND_SRC_ALPHA; + case QRhiGraphicsPipeline::OneMinusSrcAlpha: + return D3D12_BLEND_INV_SRC_ALPHA; + case QRhiGraphicsPipeline::DstAlpha: + return D3D12_BLEND_DEST_ALPHA; + case QRhiGraphicsPipeline::OneMinusDstAlpha: + return D3D12_BLEND_INV_DEST_ALPHA; + case QRhiGraphicsPipeline::ConstantColor: + case QRhiGraphicsPipeline::ConstantAlpha: + return D3D12_BLEND_BLEND_FACTOR; + case QRhiGraphicsPipeline::OneMinusConstantColor: + case QRhiGraphicsPipeline::OneMinusConstantAlpha: + return D3D12_BLEND_INV_BLEND_FACTOR; + case QRhiGraphicsPipeline::SrcAlphaSaturate: + return D3D12_BLEND_SRC_ALPHA_SAT; + case QRhiGraphicsPipeline::Src1Color: + return rgb ? D3D12_BLEND_SRC1_COLOR : D3D12_BLEND_SRC1_ALPHA; + case QRhiGraphicsPipeline::OneMinusSrc1Color: + return rgb ? D3D12_BLEND_INV_SRC1_COLOR : D3D12_BLEND_INV_SRC1_ALPHA; + case QRhiGraphicsPipeline::Src1Alpha: + return D3D12_BLEND_SRC1_ALPHA; + case QRhiGraphicsPipeline::OneMinusSrc1Alpha: + return D3D12_BLEND_INV_SRC1_ALPHA; + } + Q_UNREACHABLE_RETURN(D3D12_BLEND_ZERO); +} + +static inline D3D12_BLEND_OP toD3DBlendOp(QRhiGraphicsPipeline::BlendOp op) +{ + switch (op) { + case QRhiGraphicsPipeline::Add: + return D3D12_BLEND_OP_ADD; + case QRhiGraphicsPipeline::Subtract: + return D3D12_BLEND_OP_SUBTRACT; + case QRhiGraphicsPipeline::ReverseSubtract: + return D3D12_BLEND_OP_REV_SUBTRACT; + case QRhiGraphicsPipeline::Min: + return D3D12_BLEND_OP_MIN; + case QRhiGraphicsPipeline::Max: + return D3D12_BLEND_OP_MAX; + } + Q_UNREACHABLE_RETURN(D3D12_BLEND_OP_ADD); +} + +static inline D3D12_CULL_MODE toD3DCullMode(QRhiGraphicsPipeline::CullMode c) +{ + switch (c) { + case QRhiGraphicsPipeline::None: + return D3D12_CULL_MODE_NONE; + case QRhiGraphicsPipeline::Front: + return D3D12_CULL_MODE_FRONT; + case QRhiGraphicsPipeline::Back: + return D3D12_CULL_MODE_BACK; + } + Q_UNREACHABLE_RETURN(D3D12_CULL_MODE_NONE); +} + +static inline D3D12_FILL_MODE toD3DFillMode(QRhiGraphicsPipeline::PolygonMode mode) +{ + switch (mode) { + case QRhiGraphicsPipeline::Fill: + return D3D12_FILL_MODE_SOLID; + case QRhiGraphicsPipeline::Line: + return D3D12_FILL_MODE_WIREFRAME; + } + Q_UNREACHABLE_RETURN(D3D12_FILL_MODE_SOLID); +} + +static inline D3D12_COMPARISON_FUNC toD3DCompareOp(QRhiGraphicsPipeline::CompareOp op) +{ + switch (op) { + case QRhiGraphicsPipeline::Never: + return D3D12_COMPARISON_FUNC_NEVER; + case QRhiGraphicsPipeline::Less: + return D3D12_COMPARISON_FUNC_LESS; + case QRhiGraphicsPipeline::Equal: + return D3D12_COMPARISON_FUNC_EQUAL; + case QRhiGraphicsPipeline::LessOrEqual: + return D3D12_COMPARISON_FUNC_LESS_EQUAL; + case QRhiGraphicsPipeline::Greater: + return D3D12_COMPARISON_FUNC_GREATER; + case QRhiGraphicsPipeline::NotEqual: + return D3D12_COMPARISON_FUNC_NOT_EQUAL; + case QRhiGraphicsPipeline::GreaterOrEqual: + return D3D12_COMPARISON_FUNC_GREATER_EQUAL; + case QRhiGraphicsPipeline::Always: + return D3D12_COMPARISON_FUNC_ALWAYS; + } + Q_UNREACHABLE_RETURN(D3D12_COMPARISON_FUNC_ALWAYS); +} + +static inline D3D12_STENCIL_OP toD3DStencilOp(QRhiGraphicsPipeline::StencilOp op) +{ + switch (op) { + case QRhiGraphicsPipeline::StencilZero: + return D3D12_STENCIL_OP_ZERO; + case QRhiGraphicsPipeline::Keep: + return D3D12_STENCIL_OP_KEEP; + case QRhiGraphicsPipeline::Replace: + return D3D12_STENCIL_OP_REPLACE; + case QRhiGraphicsPipeline::IncrementAndClamp: + return D3D12_STENCIL_OP_INCR_SAT; + case QRhiGraphicsPipeline::DecrementAndClamp: + return D3D12_STENCIL_OP_DECR_SAT; + case QRhiGraphicsPipeline::Invert: + return D3D12_STENCIL_OP_INVERT; + case QRhiGraphicsPipeline::IncrementAndWrap: + return D3D12_STENCIL_OP_INCR; + case QRhiGraphicsPipeline::DecrementAndWrap: + return D3D12_STENCIL_OP_DECR; + } + Q_UNREACHABLE_RETURN(D3D12_STENCIL_OP_KEEP); +} + +static inline D3D12_PRIMITIVE_TOPOLOGY toD3DTopology(QRhiGraphicsPipeline::Topology t, int patchControlPointCount) +{ + switch (t) { + case QRhiGraphicsPipeline::Triangles: + return D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; + case QRhiGraphicsPipeline::TriangleStrip: + return D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; + case QRhiGraphicsPipeline::TriangleFan: + qWarning("Triangle fans are not supported with D3D"); + return D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; + case QRhiGraphicsPipeline::Lines: + return D3D_PRIMITIVE_TOPOLOGY_LINELIST; + case QRhiGraphicsPipeline::LineStrip: + return D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; + case QRhiGraphicsPipeline::Points: + return D3D_PRIMITIVE_TOPOLOGY_POINTLIST; + case QRhiGraphicsPipeline::Patches: + Q_ASSERT(patchControlPointCount >= 1 && patchControlPointCount <= 32); + return D3D_PRIMITIVE_TOPOLOGY(D3D_PRIMITIVE_TOPOLOGY_1_CONTROL_POINT_PATCHLIST + (patchControlPointCount - 1)); + } + Q_UNREACHABLE_RETURN(D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST); +} + +static inline D3D12_PRIMITIVE_TOPOLOGY_TYPE toD3DTopologyType(QRhiGraphicsPipeline::Topology t) +{ + switch (t) { + case QRhiGraphicsPipeline::Triangles: + case QRhiGraphicsPipeline::TriangleStrip: + case QRhiGraphicsPipeline::TriangleFan: + return D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE; + case QRhiGraphicsPipeline::Lines: + case QRhiGraphicsPipeline::LineStrip: + return D3D12_PRIMITIVE_TOPOLOGY_TYPE_LINE; + case QRhiGraphicsPipeline::Points: + return D3D12_PRIMITIVE_TOPOLOGY_TYPE_POINT; + case QRhiGraphicsPipeline::Patches: + return D3D12_PRIMITIVE_TOPOLOGY_TYPE_PATCH; + } + Q_UNREACHABLE_RETURN(D3D12_PRIMITIVE_TOPOLOGY_TYPE_TRIANGLE); +} + +static inline DXGI_FORMAT toD3DAttributeFormat(QRhiVertexInputAttribute::Format format) +{ + switch (format) { + case QRhiVertexInputAttribute::Float4: + return DXGI_FORMAT_R32G32B32A32_FLOAT; + case QRhiVertexInputAttribute::Float3: + return DXGI_FORMAT_R32G32B32_FLOAT; + case QRhiVertexInputAttribute::Float2: + return DXGI_FORMAT_R32G32_FLOAT; + case QRhiVertexInputAttribute::Float: + return DXGI_FORMAT_R32_FLOAT; + case QRhiVertexInputAttribute::UNormByte4: + return DXGI_FORMAT_R8G8B8A8_UNORM; + case QRhiVertexInputAttribute::UNormByte2: + return DXGI_FORMAT_R8G8_UNORM; + case QRhiVertexInputAttribute::UNormByte: + return DXGI_FORMAT_R8_UNORM; + case QRhiVertexInputAttribute::UInt4: + return DXGI_FORMAT_R32G32B32A32_UINT; + case QRhiVertexInputAttribute::UInt3: + return DXGI_FORMAT_R32G32B32_UINT; + case QRhiVertexInputAttribute::UInt2: + return DXGI_FORMAT_R32G32_UINT; + case QRhiVertexInputAttribute::UInt: + return DXGI_FORMAT_R32_UINT; + case QRhiVertexInputAttribute::SInt4: + return DXGI_FORMAT_R32G32B32A32_SINT; + case QRhiVertexInputAttribute::SInt3: + return DXGI_FORMAT_R32G32B32_SINT; + case QRhiVertexInputAttribute::SInt2: + return DXGI_FORMAT_R32G32_SINT; + case QRhiVertexInputAttribute::SInt: + return DXGI_FORMAT_R32_SINT; + case QRhiVertexInputAttribute::Half4: + // Note: D3D does not support half3. Pass through half3 as half4. + case QRhiVertexInputAttribute::Half3: + return DXGI_FORMAT_R16G16B16A16_FLOAT; + case QRhiVertexInputAttribute::Half2: + return DXGI_FORMAT_R16G16_FLOAT; + case QRhiVertexInputAttribute::Half: + return DXGI_FORMAT_R16_FLOAT; + case QRhiVertexInputAttribute::UShort4: + // Note: D3D does not support UShort3. Pass through UShort3 as UShort4. + case QRhiVertexInputAttribute::UShort3: + return DXGI_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort2: + return DXGI_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return DXGI_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + // Note: D3D does not support SShort3. Pass through SShort3 as SShort4. + case QRhiVertexInputAttribute::SShort3: + return DXGI_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort2: + return DXGI_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return DXGI_FORMAT_R16_SINT; + } + Q_UNREACHABLE_RETURN(DXGI_FORMAT_R32G32B32A32_FLOAT); +} + +QD3D12GraphicsPipeline::QD3D12GraphicsPipeline(QRhiImplementation *rhi) + : QRhiGraphicsPipeline(rhi) +{ +} + +QD3D12GraphicsPipeline::~QD3D12GraphicsPipeline() +{ + destroy(); +} + +void QD3D12GraphicsPipeline::destroy() +{ + if (handle.isNull()) + return; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) { + rhiD->releaseQueue.deferredReleasePipeline(handle); + rhiD->releaseQueue.deferredReleaseRootSignature(rootSigHandle); + } + + handle = {}; + stageData = {}; + + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12GraphicsPipeline::create() +{ + if (!handle.isNull()) + destroy(); + + QRHI_RES_RHI(QRhiD3D12); + if (!rhiD->sanityCheckGraphicsPipeline(this)) + return false; + + rhiD->pipelineCreationStart(); + + QByteArray shaderBytecode[5]; + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + const QD3D12Stage d3dStage = qd3d12_stage(shaderStage.type()); + stageData[d3dStage].valid = true; + stageData[d3dStage].stage = d3dStage; + auto cacheIt = rhiD->shaderBytecodeCache.data.constFind(shaderStage); + if (cacheIt != rhiD->shaderBytecodeCache.data.constEnd()) { + shaderBytecode[d3dStage] = cacheIt->bytecode; + stageData[d3dStage].nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; + } else { + QString error; + QShaderKey shaderKey; + int compileFlags = 0; + if (m_flags.testFlag(CompileShadersWithDebugInfo)) + compileFlags |= int(HlslCompileFlag::WithDebugInfo); + const QByteArray bytecode = compileHlslShaderSource(shaderStage.shader(), + shaderStage.shaderVariant(), + compileFlags, + &error, + &shaderKey); + if (bytecode.isEmpty()) { + qWarning("HLSL graphics shader compilation failed: %s", qPrintable(error)); + return false; + } + + shaderBytecode[d3dStage] = bytecode; + stageData[d3dStage].nativeResourceBindingMap = shaderStage.shader().nativeResourceBindingMap(shaderKey); + rhiD->shaderBytecodeCache.insertWithCapacityLimit(shaderStage, + { bytecode, stageData[d3dStage].nativeResourceBindingMap }); + } + } + + QD3D12ShaderResourceBindings *srbD = QRHI_RES(QD3D12ShaderResourceBindings, m_shaderResourceBindings); + if (srbD) { + rootSigHandle = srbD->createRootSignature(stageData.data(), 5); + if (rootSigHandle.isNull()) { + qWarning("Failed to create root signature"); + return false; + } + } + ID3D12RootSignature *rootSig = nullptr; + if (QD3D12RootSignature *rs = rhiD->rootSignaturePool.lookupRef(rootSigHandle)) + rootSig = rs->rootSig; + if (!rootSig) { + qWarning("Cannot create graphics pipeline state without root signature"); + return false; + } + + QD3D12RenderPassDescriptor *rpD = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); + const DXGI_SAMPLE_DESC sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, DXGI_FORMAT(rpD->colorFormat[0])); + + struct { + QD3D12PipelineStateSubObject<ID3D12RootSignature *, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_ROOT_SIGNATURE> rootSig; + QD3D12PipelineStateSubObject<D3D12_INPUT_LAYOUT_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_INPUT_LAYOUT> inputLayout; + QD3D12PipelineStateSubObject<D3D12_PRIMITIVE_TOPOLOGY_TYPE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_PRIMITIVE_TOPOLOGY> primitiveTopology; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_VS> VS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_HS> HS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DS> DS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_GS> GS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_PS> PS; + QD3D12PipelineStateSubObject<D3D12_RASTERIZER_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_RASTERIZER> rasterizerState; + QD3D12PipelineStateSubObject<D3D12_DEPTH_STENCIL_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DEPTH_STENCIL> depthStencilState; + QD3D12PipelineStateSubObject<D3D12_BLEND_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_BLEND> blendState; + QD3D12PipelineStateSubObject<D3D12_RT_FORMAT_ARRAY, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_RENDER_TARGET_FORMATS> rtFormats; + QD3D12PipelineStateSubObject<DXGI_FORMAT, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DEPTH_STENCIL_FORMAT> dsFormat; + QD3D12PipelineStateSubObject<DXGI_SAMPLE_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_SAMPLE_DESC> sampleDesc; + QD3D12PipelineStateSubObject<UINT, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_SAMPLE_MASK> sampleMask; + QD3D12PipelineStateSubObject<D3D12_VIEW_INSTANCING_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_VIEW_INSTANCING> viewInstancingDesc; + } stream; + + stream.rootSig.object = rootSig; + + QVarLengthArray<D3D12_INPUT_ELEMENT_DESC, 4> inputDescs; + QByteArrayList matrixSliceSemantics; + if (!shaderBytecode[VS].isEmpty()) { + for (auto it = m_vertexInputLayout.cbeginAttributes(), itEnd = m_vertexInputLayout.cendAttributes(); + it != itEnd; ++it) + { + D3D12_INPUT_ELEMENT_DESC desc = {}; + // The output from SPIRV-Cross uses TEXCOORD<location> as the + // semantic, except for matrices that are unrolled into consecutive + // vec2/3/4s attributes and need TEXCOORD<location>_ as + // SemanticName and row/column index as SemanticIndex. + const int matrixSlice = it->matrixSlice(); + if (matrixSlice < 0) { + desc.SemanticName = "TEXCOORD"; + desc.SemanticIndex = UINT(it->location()); + } else { + QByteArray sem; + sem.resize(16); + qsnprintf(sem.data(), sem.size(), "TEXCOORD%d_", it->location() - matrixSlice); + matrixSliceSemantics.append(sem); + desc.SemanticName = matrixSliceSemantics.last().constData(); + desc.SemanticIndex = UINT(matrixSlice); + } + desc.Format = toD3DAttributeFormat(it->format()); + desc.InputSlot = UINT(it->binding()); + desc.AlignedByteOffset = it->offset(); + const QRhiVertexInputBinding *inputBinding = m_vertexInputLayout.bindingAt(it->binding()); + if (inputBinding->classification() == QRhiVertexInputBinding::PerInstance) { + desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_INSTANCE_DATA; + desc.InstanceDataStepRate = inputBinding->instanceStepRate(); + } else { + desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA; + } + inputDescs.append(desc); + } + } + + stream.inputLayout.object.NumElements = inputDescs.count(); + stream.inputLayout.object.pInputElementDescs = inputDescs.isEmpty() ? nullptr : inputDescs.constData(); + + stream.primitiveTopology.object = toD3DTopologyType(m_topology); + topology = toD3DTopology(m_topology, m_patchControlPointCount); + + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + const int d3dStage = qd3d12_stage(shaderStage.type()); + switch (d3dStage) { + case VS: + stream.VS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.VS.object.BytecodeLength = shaderBytecode[d3dStage].size(); + break; + case HS: + stream.HS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.HS.object.BytecodeLength = shaderBytecode[d3dStage].size(); + break; + case DS: + stream.DS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.DS.object.BytecodeLength = shaderBytecode[d3dStage].size(); + break; + case GS: + stream.GS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.GS.object.BytecodeLength = shaderBytecode[d3dStage].size(); + break; + case PS: + stream.PS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.PS.object.BytecodeLength = shaderBytecode[d3dStage].size(); + break; + default: + Q_UNREACHABLE(); + break; + } + } + + stream.rasterizerState.object.FillMode = toD3DFillMode(m_polygonMode); + stream.rasterizerState.object.CullMode = toD3DCullMode(m_cullMode); + stream.rasterizerState.object.FrontCounterClockwise = m_frontFace == CCW; + stream.rasterizerState.object.DepthBias = m_depthBias; + stream.rasterizerState.object.SlopeScaledDepthBias = m_slopeScaledDepthBias; + stream.rasterizerState.object.DepthClipEnable = TRUE; + stream.rasterizerState.object.MultisampleEnable = sampleDesc.Count > 1; + + stream.depthStencilState.object.DepthEnable = m_depthTest; + stream.depthStencilState.object.DepthWriteMask = m_depthWrite ? D3D12_DEPTH_WRITE_MASK_ALL : D3D12_DEPTH_WRITE_MASK_ZERO; + stream.depthStencilState.object.DepthFunc = toD3DCompareOp(m_depthOp); + stream.depthStencilState.object.StencilEnable = m_stencilTest; + if (m_stencilTest) { + stream.depthStencilState.object.StencilReadMask = UINT8(m_stencilReadMask); + stream.depthStencilState.object.StencilWriteMask = UINT8(m_stencilWriteMask); + stream.depthStencilState.object.FrontFace.StencilFailOp = toD3DStencilOp(m_stencilFront.failOp); + stream.depthStencilState.object.FrontFace.StencilDepthFailOp = toD3DStencilOp(m_stencilFront.depthFailOp); + stream.depthStencilState.object.FrontFace.StencilPassOp = toD3DStencilOp(m_stencilFront.passOp); + stream.depthStencilState.object.FrontFace.StencilFunc = toD3DCompareOp(m_stencilFront.compareOp); + stream.depthStencilState.object.BackFace.StencilFailOp = toD3DStencilOp(m_stencilBack.failOp); + stream.depthStencilState.object.BackFace.StencilDepthFailOp = toD3DStencilOp(m_stencilBack.depthFailOp); + stream.depthStencilState.object.BackFace.StencilPassOp = toD3DStencilOp(m_stencilBack.passOp); + stream.depthStencilState.object.BackFace.StencilFunc = toD3DCompareOp(m_stencilBack.compareOp); + } + + stream.blendState.object.IndependentBlendEnable = m_targetBlends.count() > 1; + for (int i = 0, ie = m_targetBlends.count(); i != ie; ++i) { + const QRhiGraphicsPipeline::TargetBlend &b(m_targetBlends[i]); + D3D12_RENDER_TARGET_BLEND_DESC blend = {}; + blend.BlendEnable = b.enable; + blend.SrcBlend = toD3DBlendFactor(b.srcColor, true); + blend.DestBlend = toD3DBlendFactor(b.dstColor, true); + blend.BlendOp = toD3DBlendOp(b.opColor); + blend.SrcBlendAlpha = toD3DBlendFactor(b.srcAlpha, false); + blend.DestBlendAlpha = toD3DBlendFactor(b.dstAlpha, false); + blend.BlendOpAlpha = toD3DBlendOp(b.opAlpha); + blend.RenderTargetWriteMask = toD3DColorWriteMask(b.colorWrite); + stream.blendState.object.RenderTarget[i] = blend; + } + if (m_targetBlends.isEmpty()) { + D3D12_RENDER_TARGET_BLEND_DESC blend = {}; + blend.RenderTargetWriteMask = D3D12_COLOR_WRITE_ENABLE_ALL; + stream.blendState.object.RenderTarget[0] = blend; + } + + stream.rtFormats.object.NumRenderTargets = rpD->colorAttachmentCount; + for (int i = 0; i < rpD->colorAttachmentCount; ++i) + stream.rtFormats.object.RTFormats[i] = DXGI_FORMAT(rpD->colorFormat[i]); + + stream.dsFormat.object = rpD->hasDepthStencil ? DXGI_FORMAT(rpD->dsFormat) : DXGI_FORMAT_UNKNOWN; + + stream.sampleDesc.object = sampleDesc; + + stream.sampleMask.object = 0xFFFFFFFF; + + viewInstanceMask = 0; + const bool isMultiView = m_multiViewCount >= 2; + stream.viewInstancingDesc.object.ViewInstanceCount = isMultiView ? m_multiViewCount : 0; + QVarLengthArray<D3D12_VIEW_INSTANCE_LOCATION, 4> viewInstanceLocations; + if (isMultiView) { + for (int i = 0; i < m_multiViewCount; ++i) { + viewInstanceMask |= (1 << i); + viewInstanceLocations.append({ 0, UINT(i) }); + } + stream.viewInstancingDesc.object.pViewInstanceLocations = viewInstanceLocations.constData(); + } + + const D3D12_PIPELINE_STATE_STREAM_DESC streamDesc = { sizeof(stream), &stream }; + + ID3D12PipelineState *pso = nullptr; + HRESULT hr = rhiD->dev->CreatePipelineState(&streamDesc, __uuidof(ID3D12PipelineState), reinterpret_cast<void **>(&pso)); + if (FAILED(hr)) { + qWarning("Failed to create graphics pipeline state: %s", + qPrintable(QSystemError::windowsComString(hr))); + rhiD->rootSignaturePool.remove(rootSigHandle); + rootSigHandle = {}; + return false; + } + + handle = QD3D12Pipeline::addToPool(&rhiD->pipelinePool, QD3D12Pipeline::Graphics, pso); + + rhiD->pipelineCreationEnd(); + generation += 1; + rhiD->registerResource(this); + return true; +} + +QD3D12ComputePipeline::QD3D12ComputePipeline(QRhiImplementation *rhi) + : QRhiComputePipeline(rhi) +{ +} + +QD3D12ComputePipeline::~QD3D12ComputePipeline() +{ + destroy(); +} + +void QD3D12ComputePipeline::destroy() +{ + if (handle.isNull()) + return; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) { + rhiD->releaseQueue.deferredReleasePipeline(handle); + rhiD->releaseQueue.deferredReleaseRootSignature(rootSigHandle); + } + + handle = {}; + stageData = {}; + + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12ComputePipeline::create() +{ + if (!handle.isNull()) + destroy(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->pipelineCreationStart(); + + stageData.valid = true; + stageData.stage = CS; + + QByteArray shaderBytecode; + auto cacheIt = rhiD->shaderBytecodeCache.data.constFind(m_shaderStage); + if (cacheIt != rhiD->shaderBytecodeCache.data.constEnd()) { + shaderBytecode = cacheIt->bytecode; + stageData.nativeResourceBindingMap = cacheIt->nativeResourceBindingMap; + } else { + QString error; + QShaderKey shaderKey; + int compileFlags = 0; + if (m_flags.testFlag(CompileShadersWithDebugInfo)) + compileFlags |= int(HlslCompileFlag::WithDebugInfo); + const QByteArray bytecode = compileHlslShaderSource(m_shaderStage.shader(), + m_shaderStage.shaderVariant(), + compileFlags, + &error, + &shaderKey); + if (bytecode.isEmpty()) { + qWarning("HLSL compute shader compilation failed: %s", qPrintable(error)); + return false; + } + + shaderBytecode = bytecode; + stageData.nativeResourceBindingMap = m_shaderStage.shader().nativeResourceBindingMap(shaderKey); + rhiD->shaderBytecodeCache.insertWithCapacityLimit(m_shaderStage, { bytecode, + stageData.nativeResourceBindingMap }); + } + + QD3D12ShaderResourceBindings *srbD = QRHI_RES(QD3D12ShaderResourceBindings, m_shaderResourceBindings); + if (srbD) { + rootSigHandle = srbD->createRootSignature(&stageData, 1); + if (rootSigHandle.isNull()) { + qWarning("Failed to create root signature"); + return false; + } + } + ID3D12RootSignature *rootSig = nullptr; + if (QD3D12RootSignature *rs = rhiD->rootSignaturePool.lookupRef(rootSigHandle)) + rootSig = rs->rootSig; + if (!rootSig) { + qWarning("Cannot create compute pipeline state without root signature"); + return false; + } + + struct { + QD3D12PipelineStateSubObject<ID3D12RootSignature *, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_ROOT_SIGNATURE> rootSig; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_CS> CS; + } stream; + stream.rootSig.object = rootSig; + stream.CS.object.pShaderBytecode = shaderBytecode.constData(); + stream.CS.object.BytecodeLength = shaderBytecode.size(); + const D3D12_PIPELINE_STATE_STREAM_DESC streamDesc = { sizeof(stream), &stream }; + ID3D12PipelineState *pso = nullptr; + HRESULT hr = rhiD->dev->CreatePipelineState(&streamDesc, __uuidof(ID3D12PipelineState), reinterpret_cast<void **>(&pso)); + if (FAILED(hr)) { + qWarning("Failed to create compute pipeline state: %s", + qPrintable(QSystemError::windowsComString(hr))); + rhiD->rootSignaturePool.remove(rootSigHandle); + rootSigHandle = {}; + return false; + } + + handle = QD3D12Pipeline::addToPool(&rhiD->pipelinePool, QD3D12Pipeline::Compute, pso); + + rhiD->pipelineCreationEnd(); + generation += 1; + rhiD->registerResource(this); + return true; +} + +// This is a lot like in the Metal backend: we need to now the rtv and dsv +// formats to create a graphics pipeline, and that's exactly what our +// "renderpass descriptor" is going to hold. +QD3D12RenderPassDescriptor::QD3D12RenderPassDescriptor(QRhiImplementation *rhi) + : QRhiRenderPassDescriptor(rhi) +{ + serializedFormatData.reserve(16); +} + +QD3D12RenderPassDescriptor::~QD3D12RenderPassDescriptor() +{ + destroy(); +} + +void QD3D12RenderPassDescriptor::destroy() +{ + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); +} + +bool QD3D12RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const +{ + if (!other) + return false; + + const QD3D12RenderPassDescriptor *o = QRHI_RES(const QD3D12RenderPassDescriptor, other); + + if (colorAttachmentCount != o->colorAttachmentCount) + return false; + + if (hasDepthStencil != o->hasDepthStencil) + return false; + + for (int i = 0; i < colorAttachmentCount; ++i) { + if (colorFormat[i] != o->colorFormat[i]) + return false; + } + + if (hasDepthStencil) { + if (dsFormat != o->dsFormat) + return false; + } + + return true; +} + +void QD3D12RenderPassDescriptor::updateSerializedFormat() +{ + serializedFormatData.clear(); + auto p = std::back_inserter(serializedFormatData); + + *p++ = colorAttachmentCount; + *p++ = hasDepthStencil; + for (int i = 0; i < colorAttachmentCount; ++i) + *p++ = colorFormat[i]; + *p++ = hasDepthStencil ? dsFormat : 0; +} + +QRhiRenderPassDescriptor *QD3D12RenderPassDescriptor::newCompatibleRenderPassDescriptor() const +{ + QD3D12RenderPassDescriptor *rpD = new QD3D12RenderPassDescriptor(m_rhi); + rpD->colorAttachmentCount = colorAttachmentCount; + rpD->hasDepthStencil = hasDepthStencil; + memcpy(rpD->colorFormat, colorFormat, sizeof(colorFormat)); + rpD->dsFormat = dsFormat; + + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); + return rpD; +} + +QVector<quint32> QD3D12RenderPassDescriptor::serializedFormat() const +{ + return serializedFormatData; +} + +QD3D12CommandBuffer::QD3D12CommandBuffer(QRhiImplementation *rhi) + : QRhiCommandBuffer(rhi) +{ + resetState(); +} + +QD3D12CommandBuffer::~QD3D12CommandBuffer() +{ + destroy(); +} + +void QD3D12CommandBuffer::destroy() +{ + // nothing to do here, the command list is not owned by us +} + +const QRhiNativeHandles *QD3D12CommandBuffer::nativeHandles() +{ + nativeHandlesStruct.commandList = cmdList; + return &nativeHandlesStruct; +} + +QD3D12SwapChainRenderTarget::QD3D12SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain), + d(rhi) +{ +} + +QD3D12SwapChainRenderTarget::~QD3D12SwapChainRenderTarget() +{ + destroy(); +} + +void QD3D12SwapChainRenderTarget::destroy() +{ + // nothing to do here +} + +QSize QD3D12SwapChainRenderTarget::pixelSize() const +{ + return d.pixelSize; +} + +float QD3D12SwapChainRenderTarget::devicePixelRatio() const +{ + return d.dpr; +} + +int QD3D12SwapChainRenderTarget::sampleCount() const +{ + return d.sampleCount; +} + +QD3D12SwapChain::QD3D12SwapChain(QRhiImplementation *rhi) + : QRhiSwapChain(rhi), + rtWrapper(rhi, this), + rtWrapperRight(rhi, this), + cbWrapper(rhi) +{ +} + +QD3D12SwapChain::~QD3D12SwapChain() +{ + destroy(); +} + +void QD3D12SwapChain::destroy() +{ + if (!swapChain) + return; + + releaseBuffers(); + + swapChain->Release(); + swapChain = nullptr; + sourceSwapChain1->Release(); + sourceSwapChain1 = nullptr; + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + FrameResources &fr(frameRes[i]); + if (fr.fence) + fr.fence->Release(); + if (fr.fenceEvent) + CloseHandle(fr.fenceEvent); + if (fr.cmdList) + fr.cmdList->Release(); + fr = {}; + } + + if (dcompVisual) { + dcompVisual->Release(); + dcompVisual = nullptr; + } + + if (dcompTarget) { + dcompTarget->Release(); + dcompTarget = nullptr; + } + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) { + rhiD->swapchains.remove(this); + rhiD->unregisterResource(this); + } +} + +void QD3D12SwapChain::releaseBuffers() +{ + QRHI_RES_RHI(QRhiD3D12); + rhiD->waitGpu(); + for (UINT i = 0; i < BUFFER_COUNT; ++i) { + rhiD->resourcePool.remove(colorBuffers[i]); + rhiD->rtvPool.release(rtvs[i], 1); + if (stereo) + rhiD->rtvPool.release(rtvsRight[i], 1); + if (!msaaBuffers[i].isNull()) + rhiD->resourcePool.remove(msaaBuffers[i]); + if (msaaRtvs[i].isValid()) + rhiD->rtvPool.release(msaaRtvs[i], 1); + } +} + +void QD3D12SwapChain::waitCommandCompletionForFrameSlot(int frameSlot) +{ + FrameResources &fr(frameRes[frameSlot]); + if (fr.fence->GetCompletedValue() < fr.fenceCounter) { + fr.fence->SetEventOnCompletion(fr.fenceCounter, fr.fenceEvent); + WaitForSingleObject(fr.fenceEvent, INFINITE); + } +} + +void QD3D12SwapChain::addCommandCompletionSignalForCurrentFrameSlot() +{ + QRHI_RES_RHI(QRhiD3D12); + FrameResources &fr(frameRes[currentFrameSlot]); + fr.fenceCounter += 1u; + rhiD->cmdQueue->Signal(fr.fence, fr.fenceCounter); +} + +QRhiCommandBuffer *QD3D12SwapChain::currentFrameCommandBuffer() +{ + return &cbWrapper; +} + +QRhiRenderTarget *QD3D12SwapChain::currentFrameRenderTarget() +{ + return &rtWrapper; +} + +QRhiRenderTarget *QD3D12SwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + return !stereo || targetBuffer == StereoTargetBuffer::LeftBuffer ? &rtWrapper : &rtWrapperRight; +} + +QSize QD3D12SwapChain::surfacePixelSize() +{ + Q_ASSERT(m_window); + return m_window->size() * m_window->devicePixelRatio(); +} + +bool QD3D12SwapChain::isFormatSupported(Format f) +{ + if (f == SDR) + return true; + + if (!m_window) { + qWarning("Attempted to call isFormatSupported() without a window set"); + return false; + } + + QRHI_RES_RHI(QRhiD3D12); + DXGI_OUTPUT_DESC1 desc1; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) { + if (desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) + return f == QRhiSwapChain::HDRExtendedSrgbLinear || f == QRhiSwapChain::HDR10; + } + + return false; +} + +QRhiSwapChainHdrInfo QD3D12SwapChain::hdrInfo() +{ + QRhiSwapChainHdrInfo info = QRhiSwapChain::hdrInfo(); + // Must use m_window, not window, given this may be called before createOrResize(). + if (m_window) { + QRHI_RES_RHI(QRhiD3D12); + DXGI_OUTPUT_DESC1 hdrOutputDesc; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { + info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; + info.limits.luminanceInNits.minLuminance = hdrOutputDesc.MinLuminance; + info.limits.luminanceInNits.maxLuminance = hdrOutputDesc.MaxLuminance; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; // 1.0 = 80 nits + info.sdrWhiteLevel = QRhiD3D::sdrWhiteLevelInNits(hdrOutputDesc); + } + } + return info; +} + +QRhiRenderPassDescriptor *QD3D12SwapChain::newCompatibleRenderPassDescriptor() +{ + // not yet built so cannot rely on data computed in createOrResize() + chooseFormats(); + + QD3D12RenderPassDescriptor *rpD = new QD3D12RenderPassDescriptor(m_rhi); + rpD->colorAttachmentCount = 1; + rpD->hasDepthStencil = m_depthStencil != nullptr; + rpD->colorFormat[0] = int(srgbAdjustedColorFormat); + rpD->dsFormat = QD3D12RenderBuffer::DS_FORMAT; + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); + return rpD; +} + +bool QRhiD3D12::ensureDirectCompositionDevice() +{ + if (dcompDevice) + return true; + + qCDebug(QRHI_LOG_INFO, "Creating Direct Composition device (needed for semi-transparent windows)"); + dcompDevice = QRhiD3D::createDirectCompositionDevice(); + return dcompDevice ? true : false; +} + +static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; +static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; + +void QD3D12SwapChain::chooseFormats() +{ + colorFormat = DEFAULT_FORMAT; + srgbAdjustedColorFormat = m_flags.testFlag(sRGB) ? DEFAULT_SRGB_FORMAT : DEFAULT_FORMAT; + hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G22_NONE_P709; // SDR + DXGI_OUTPUT_DESC1 hdrOutputDesc; + QRHI_RES_RHI(QRhiD3D12); + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { + // https://docs.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range + if (hdrOutputDesc.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) { + switch (m_format) { + case HDRExtendedSrgbLinear: + colorFormat = DXGI_FORMAT_R16G16B16A16_FLOAT; + hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G10_NONE_P709; + srgbAdjustedColorFormat = colorFormat; + break; + case HDR10: + colorFormat = DXGI_FORMAT_R10G10B10A2_UNORM; + hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020; + srgbAdjustedColorFormat = colorFormat; + break; + default: + break; + } + } else { + // This happens also when Use HDR is set to Off in the Windows + // Display settings. Show a helpful warning, but continue with the + // default non-HDR format. + qWarning("The output associated with the window is not HDR capable " + "(or Use HDR is Off in the Display Settings), ignoring HDR format request"); + } + } + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, colorFormat); +} + +bool QD3D12SwapChain::createOrResize() +{ + // Can be called multiple times due to window resizes - that is not the + // same as a simple destroy+create (as with other resources). Just need to + // resize the buffers then. + + const bool needsRegistration = !window || window != m_window; + + // except if the window actually changes + if (window && window != m_window) + destroy(); + + window = m_window; + m_currentPixelSize = surfacePixelSize(); + pixelSize = m_currentPixelSize; + + if (pixelSize.isEmpty()) + return false; + + HWND hwnd = reinterpret_cast<HWND>(window->winId()); + HRESULT hr; + QRHI_RES_RHI(QRhiD3D12); + stereo = m_window->format().stereo() && rhiD->dxgiFactory->IsWindowedStereoEnabled(); + + if (m_flags.testFlag(SurfaceHasPreMulAlpha) || m_flags.testFlag(SurfaceHasNonPreMulAlpha)) { + if (rhiD->ensureDirectCompositionDevice()) { + if (!dcompTarget) { + hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, false, &dcompTarget); + if (FAILED(hr)) { + qWarning("Failed to create Direct Composition target for the window: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } + if (dcompTarget && !dcompVisual) { + hr = rhiD->dcompDevice->CreateVisual(&dcompVisual); + if (FAILED(hr)) { + qWarning("Failed to create DirectComposition visual: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } + } + // simple consistency check + if (window->requestedFormat().alphaBufferSize() <= 0) + qWarning("Swapchain says surface has alpha but the window has no alphaBufferSize set. " + "This may lead to problems."); + } + + swapInterval = m_flags.testFlag(QRhiSwapChain::NoVSync) ? 0 : 1; + swapChainFlags = 0; + if (swapInterval == 0 && rhiD->supportsAllowTearing) + swapChainFlags |= DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING; + + if (!swapChain) { + chooseFormats(); + + DXGI_SWAP_CHAIN_DESC1 desc = {}; + desc.Width = UINT(pixelSize.width()); + desc.Height = UINT(pixelSize.height()); + desc.Format = colorFormat; + desc.SampleDesc.Count = 1; + desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; + desc.BufferCount = BUFFER_COUNT; + desc.Flags = swapChainFlags; + desc.Scaling = DXGI_SCALING_NONE; + desc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_DISCARD; + desc.Stereo = stereo; + + if (dcompVisual) { + // With DirectComposition setting AlphaMode to STRAIGHT fails the + // swapchain creation, whereas the result seems to be identical + // with any of the other values, including IGNORE. (?) + desc.AlphaMode = DXGI_ALPHA_MODE_PREMULTIPLIED; + + // DirectComposition has its own limitations, cannot use + // SCALING_NONE. So with semi-transparency requested we are forced + // to SCALING_STRETCH. + desc.Scaling = DXGI_SCALING_STRETCH; + } + + if (dcompVisual) + hr = rhiD->dxgiFactory->CreateSwapChainForComposition(rhiD->cmdQueue, &desc, nullptr, &sourceSwapChain1); + else + hr = rhiD->dxgiFactory->CreateSwapChainForHwnd(rhiD->cmdQueue, hwnd, &desc, nullptr, nullptr, &sourceSwapChain1); + + // If failed and we tried a HDR format, then try with SDR. This + // matches other backends, such as Vulkan where if the format is + // not supported, the default one is used instead. + if (FAILED(hr) && m_format != SDR) { + colorFormat = DEFAULT_FORMAT; + desc.Format = DEFAULT_FORMAT; + if (dcompVisual) + hr = rhiD->dxgiFactory->CreateSwapChainForComposition(rhiD->cmdQueue, &desc, nullptr, &sourceSwapChain1); + else + hr = rhiD->dxgiFactory->CreateSwapChainForHwnd(rhiD->cmdQueue, hwnd, &desc, nullptr, nullptr, &sourceSwapChain1); + } + + if (SUCCEEDED(hr)) { + if (FAILED(sourceSwapChain1->QueryInterface(__uuidof(IDXGISwapChain3), reinterpret_cast<void **>(&swapChain)))) { + qWarning("IDXGISwapChain3 not available"); + return false; + } + if (m_format != SDR) { + hr = swapChain->SetColorSpace1(hdrColorSpace); + if (FAILED(hr)) { + qWarning("Failed to set color space on swapchain: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } + if (dcompVisual) { + hr = dcompVisual->SetContent(swapChain); + if (SUCCEEDED(hr)) { + hr = dcompTarget->SetRoot(dcompVisual); + if (FAILED(hr)) { + qWarning("Failed to associate Direct Composition visual with the target: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } else { + qWarning("Failed to set content for Direct Composition visual: %s", + qPrintable(QSystemError::windowsComString(hr))); + } + } else { + // disable Alt+Enter; not relevant when using DirectComposition + rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); + } + } + if (FAILED(hr)) { + qWarning("Failed to create D3D12 swapchain: %s" + " (Width=%u Height=%u Format=%u SampleCount=%u BufferCount=%u Scaling=%u SwapEffect=%u Stereo=%u)", + qPrintable(QSystemError::windowsComString(hr)), + desc.Width, desc.Height, UINT(desc.Format), desc.SampleDesc.Count, + desc.BufferCount, UINT(desc.Scaling), UINT(desc.SwapEffect), UINT(desc.Stereo)); + return false; + } + + for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { + hr = rhiD->dev->CreateFence(0, + D3D12_FENCE_FLAG_NONE, + __uuidof(ID3D12Fence), + reinterpret_cast<void **>(&frameRes[i].fence)); + if (FAILED(hr)) { + qWarning("Failed to create fence for swapchain: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + frameRes[i].fenceEvent = CreateEvent(nullptr, FALSE, FALSE, nullptr); + + frameRes[i].fenceCounter = 0; + } + } else { + releaseBuffers(); + hr = swapChain->ResizeBuffers(BUFFER_COUNT, + UINT(pixelSize.width()), + UINT(pixelSize.height()), + colorFormat, + swapChainFlags); + if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { + qWarning("Device loss detected in ResizeBuffers()"); + rhiD->deviceLost = true; + return false; + } else if (FAILED(hr)) { + qWarning("Failed to resize D3D12 swapchain: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + + for (UINT i = 0; i < BUFFER_COUNT; ++i) { + ID3D12Resource *colorBuffer; + hr = swapChain->GetBuffer(i, __uuidof(ID3D12Resource), reinterpret_cast<void **>(&colorBuffer)); + if (FAILED(hr)) { + qWarning("Failed to get buffer %u for D3D12 swapchain: %s", + i, qPrintable(QSystemError::windowsComString(hr))); + return false; + } + colorBuffers[i] = QD3D12Resource::addToPool(&rhiD->resourcePool, colorBuffer, D3D12_RESOURCE_STATE_PRESENT); + rtvs[i] = rhiD->rtvPool.allocate(1); + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = srgbAdjustedColorFormat; + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2D; + rhiD->dev->CreateRenderTargetView(colorBuffer, &rtvDesc, rtvs[i].cpuHandle); + + if (stereo) { + rtvsRight[i] = rhiD->rtvPool.allocate(1); + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = srgbAdjustedColorFormat; + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.ArraySize = 1; + rtvDesc.Texture2DArray.FirstArraySlice = 1; + rhiD->dev->CreateRenderTargetView(colorBuffer, &rtvDesc, rtvsRight[i].cpuHandle); + } + } + + if (m_depthStencil && m_depthStencil->sampleCount() != m_sampleCount) { + qWarning("Depth-stencil buffer's sampleCount (%d) does not match color buffers' sample count (%d). Expect problems.", + m_depthStencil->sampleCount(), m_sampleCount); + } + if (m_depthStencil && m_depthStencil->pixelSize() != pixelSize) { + if (m_depthStencil->flags().testFlag(QRhiRenderBuffer::UsedWithSwapChainOnly)) { + m_depthStencil->setPixelSize(pixelSize); + if (!m_depthStencil->create()) + qWarning("Failed to rebuild swapchain's associated depth-stencil buffer for size %dx%d", + pixelSize.width(), pixelSize.height()); + } else { + qWarning("Depth-stencil buffer's size (%dx%d) does not match the surface size (%dx%d). Expect problems.", + m_depthStencil->pixelSize().width(), m_depthStencil->pixelSize().height(), + pixelSize.width(), pixelSize.height()); + } + } + + ds = m_depthStencil ? QRHI_RES(QD3D12RenderBuffer, m_depthStencil) : nullptr; + + if (sampleDesc.Count > 1) { + for (UINT i = 0; i < BUFFER_COUNT; ++i) { + D3D12_RESOURCE_DESC resourceDesc = {}; + resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; + resourceDesc.Width = UINT64(pixelSize.width()); + resourceDesc.Height = UINT(pixelSize.height()); + resourceDesc.DepthOrArraySize = 1; + resourceDesc.MipLevels = 1; + resourceDesc.Format = srgbAdjustedColorFormat; + resourceDesc.SampleDesc = sampleDesc; + resourceDesc.Layout = D3D12_TEXTURE_LAYOUT_UNKNOWN; + resourceDesc.Flags = D3D12_RESOURCE_FLAG_ALLOW_RENDER_TARGET; + D3D12_CLEAR_VALUE clearValue = {}; + clearValue.Format = colorFormat; + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + HRESULT hr = rhiD->vma.createResource(D3D12_HEAP_TYPE_DEFAULT, + &resourceDesc, + D3D12_RESOURCE_STATE_RENDER_TARGET, + &clearValue, + &allocation, + __uuidof(ID3D12Resource), + reinterpret_cast<void **>(&resource)); + if (FAILED(hr)) { + qWarning("Failed to create MSAA color buffer: %s", qPrintable(QSystemError::windowsComString(hr))); + return false; + } + msaaBuffers[i] = QD3D12Resource::addToPool(&rhiD->resourcePool, resource, D3D12_RESOURCE_STATE_RENDER_TARGET, allocation); + msaaRtvs[i] = rhiD->rtvPool.allocate(1); + if (!msaaRtvs[i].isValid()) + return false; + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = srgbAdjustedColorFormat; + rtvDesc.ViewDimension = sampleDesc.Count > 1 ? D3D12_RTV_DIMENSION_TEXTURE2DMS + : D3D12_RTV_DIMENSION_TEXTURE2D; + rhiD->dev->CreateRenderTargetView(resource, &rtvDesc, msaaRtvs[i].cpuHandle); + } + } + + currentBackBufferIndex = swapChain->GetCurrentBackBufferIndex(); + currentFrameSlot = 0; + + rtWrapper.setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget + QD3D12SwapChainRenderTarget *rtD = QRHI_RES(QD3D12SwapChainRenderTarget, &rtWrapper); + rtD->d.rp = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); + rtD->d.pixelSize = pixelSize; + rtD->d.dpr = float(window->devicePixelRatio()); + rtD->d.sampleCount = int(sampleDesc.Count); + rtD->d.colorAttCount = 1; + rtD->d.dsAttCount = m_depthStencil ? 1 : 0; + + rtWrapperRight.setRenderPassDescriptor(m_renderPassDesc); + QD3D12SwapChainRenderTarget *rtDr = QRHI_RES(QD3D12SwapChainRenderTarget, &rtWrapperRight); + rtDr->d.rp = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); + rtDr->d.pixelSize = pixelSize; + rtDr->d.dpr = float(window->devicePixelRatio()); + rtDr->d.sampleCount = int(sampleDesc.Count); + rtDr->d.colorAttCount = 1; + rtDr->d.dsAttCount = m_depthStencil ? 1 : 0; + + if (needsRegistration) { + rhiD->swapchains.insert(this); + rhiD->registerResource(this); + } + + return true; +} + +QT_END_NAMESPACE + +#endif // __ID3D12Device2_INTERFACE_DEFINED__ diff --git a/src/gui/rhi/qrhid3d12_p.h b/src/gui/rhi/qrhid3d12_p.h new file mode 100644 index 0000000000..3f9abbb5ac --- /dev/null +++ b/src/gui/rhi/qrhid3d12_p.h @@ -0,0 +1,1248 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHID3D12_P_H +#define QRHID3D12_P_H + +// +// W A R N I N G +// ------------- +// +// This file is not part of the Qt API. It exists purely as an +// implementation detail. This header file may change from version to +// version without notice, or even be removed. +// +// We mean it. +// + +#include "qrhi_p.h" +#include <QWindow> +#include <QBitArray> + +#include <optional> +#include <array> + +#include <d3d12.h> +#include <d3d12sdklayers.h> +#include <dxgi1_6.h> +#include <dcomp.h> + +#include "D3D12MemAlloc.h" + +// ID3D12Device2 and ID3D12GraphicsCommandList1 and types and enums introduced +// with those are hard requirements now. These should be declared in any +// moderately recent d3d12.h, but if it is an SDK from before Windows 10 +// version 1703 then these types could be missing. In the absence of other +// options, handle this by skipping all the code and making QRhi::create() fail +// in such builds. +#ifdef __ID3D12Device2_INTERFACE_DEFINED__ +#define QRHI_D3D12_AVAILABLE + +QT_BEGIN_NAMESPACE + +static const int QD3D12_FRAMES_IN_FLIGHT = 2; + +class QRhiD3D12; + +struct QD3D12Descriptor +{ + D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = {}; + D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = {}; + + bool isValid() const { return cpuHandle.ptr != 0; } +}; + +struct QD3D12ReleaseQueue; + +struct QD3D12DescriptorHeap +{ + bool isValid() const { return heap && capacity; } + bool create(ID3D12Device *device, + quint32 descriptorCount, + D3D12_DESCRIPTOR_HEAP_TYPE heapType, + D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags); + void createWithExisting(const QD3D12DescriptorHeap &other, + quint32 offsetInDescriptors, + quint32 descriptorCount); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + QD3D12Descriptor get(quint32 count); + QD3D12Descriptor at(quint32 index) const; + quint32 remainingCapacity() const { return capacity - head; } + + QD3D12Descriptor incremented(const QD3D12Descriptor &descriptor, quint32 offsetInDescriptors) const + { + D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = descriptor.cpuHandle; + cpuHandle.ptr += offsetInDescriptors * descriptorByteSize; + D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = descriptor.gpuHandle; + if (gpuHandle.ptr) + gpuHandle.ptr += offsetInDescriptors * descriptorByteSize; + return { cpuHandle, gpuHandle }; + } + + ID3D12DescriptorHeap *heap = nullptr; + quint32 capacity = 0; + QD3D12Descriptor heapStart; + quint32 head = 0; + quint32 descriptorByteSize = 0; + D3D12_DESCRIPTOR_HEAP_TYPE heapType; + D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags; +}; + +struct QD3D12CpuDescriptorPool +{ + bool isValid() const { return !heaps.isEmpty(); } + bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE heapType, const char *debugName = ""); + void destroy(); + + QD3D12Descriptor allocate(quint32 count); + void release(const QD3D12Descriptor &descriptor, quint32 count); + + static const int DESCRIPTORS_PER_HEAP = 256; + + struct HeapWithMap { + QD3D12DescriptorHeap heap; + QBitArray map; + static HeapWithMap init(const QD3D12DescriptorHeap &heap, quint32 descriptorCount) { + HeapWithMap result; + result.heap = heap; + result.map.resize(descriptorCount); + return result; + } + }; + + ID3D12Device *device; + quint32 descriptorByteSize; + QVector<HeapWithMap> heaps; + const char *debugName; +}; + +struct QD3D12QueryHeap +{ + bool isValid() const { return heap && capacity; } + bool create(ID3D12Device *device, + quint32 queryCount, + D3D12_QUERY_HEAP_TYPE heapType); + void destroy(); + + ID3D12QueryHeap *heap = nullptr; + quint32 capacity = 0; +}; + +struct QD3D12StagingArea +{ + static const quint32 ALIGNMENT = D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT; // 512 so good enough both for cb and texdata + + struct Allocation { + quint8 *p = nullptr; + D3D12_GPU_VIRTUAL_ADDRESS gpuAddr = 0; + ID3D12Resource *buffer = nullptr; + quint32 bufferOffset = 0; + bool isValid() const { return p != nullptr; } + }; + + bool isValid() const { return allocation && mem.isValid(); } + bool create(QRhiD3D12 *rhi, quint32 capacity, D3D12_HEAP_TYPE heapType); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + Allocation get(quint32 byteSize); + + quint32 remainingCapacity() const + { + return capacity - head; + } + + static quint32 allocSizeForArray(quint32 size, int count = 1) + { + return count * ((size + ALIGNMENT - 1) & ~(ALIGNMENT - 1)); + } + + Allocation mem; + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + quint32 head; + quint32 capacity; +}; + +struct QD3D12ObjectHandle +{ + quint32 index = 0; + quint32 generation = 0; + + // the default, null handle is guaranteed to give ObjectPool::isValid() == false + bool isNull() const { return index == 0 && generation == 0; } +}; + +inline bool operator==(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept +{ + return a.index == b.index && a.generation == b.generation; +} + +inline bool operator!=(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept +{ + return !(a == b); +} + +template<typename T> +struct QD3D12ObjectPool +{ + void create(const char *debugName = "") + { + this->debugName = debugName; + Q_ASSERT(data.isEmpty()); + data.append(Data()); // index 0 is always invalid + } + + void destroy() { + int leakCount = 0; // will nicely destroy everything here, but warn about it if enabled + for (Data &d : data) { + if (d.object.has_value()) { + leakCount += 1; + d.object->releaseResources(); + } + } + data.clear(); +#ifndef QT_NO_DEBUG + // debug builds: just do it always + static bool leakCheck = true; +#else + // release builds: opt-in + static bool leakCheck = qEnvironmentVariableIntValue("QT_RHI_LEAK_CHECK"); +#endif + if (leakCheck) { + if (leakCount > 0) { + qWarning("QD3D12ObjectPool::destroy(): Pool %p '%s' had %d unreleased objects", + this, debugName, leakCount); + } + } + } + + bool isValid(const QD3D12ObjectHandle &handle) const + { + return handle.index > 0 + && handle.index < quint32(data.count()) + && handle.generation > 0 + && handle.generation == data[handle.index].generation + && data[handle.index].object.has_value(); + } + + T lookup(const QD3D12ObjectHandle &handle) const + { + return isValid(handle) ? *data[handle.index].object : T(); + } + + const T *lookupRef(const QD3D12ObjectHandle &handle) const + { + return isValid(handle) ? &*data[handle.index].object : nullptr; + } + + T *lookupRef(const QD3D12ObjectHandle &handle) + { + return isValid(handle) ? &*data[handle.index].object : nullptr; + } + + QD3D12ObjectHandle add(const T &object) + { + Q_ASSERT(!data.isEmpty()); + const quint32 count = quint32(data.count()); + quint32 index = 1; // index 0 is always invalid + for (; index < count; ++index) { + if (!data[index].object.has_value()) + break; + } + if (index < count) { + data[index].object = object; + quint32 &generation = data[index].generation; + generation += 1u; + return { index, generation }; + } else { + data.append({ object, 1 }); + return { count, 1 }; + } + } + + void remove(const QD3D12ObjectHandle &handle) + { + if (T *object = lookupRef(handle)) { + object->releaseResources(); + data[handle.index].object.reset(); + } + } + + const char *debugName; + struct Data { + std::optional<T> object; + quint32 generation = 0; + }; + QVector<Data> data; +}; + +struct QD3D12Resource +{ + ID3D12Resource *resource; + D3D12_RESOURCE_STATES state; + D3D12_RESOURCE_DESC desc; + D3D12MA::Allocation *allocation; + void *cpuMapPtr; + enum { UavUsageRead = 0x01, UavUsageWrite = 0x02 }; + int uavUsage; + bool owns; + + // note that this assumes the allocation (if there is one) and the resource + // are separately releaseable, see D3D12MemAlloc docs + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Resource> *pool, + ID3D12Resource *resource, + D3D12_RESOURCE_STATES state, + D3D12MA::Allocation *allocation = nullptr, + void *cpuMapPtr = nullptr) + { + Q_ASSERT(resource); + return pool->add({ resource, state, resource->GetDesc(), allocation, cpuMapPtr, 0, true }); + } + + // for QRhiTexture::createFrom() where the ID3D12Resource is not owned by us + static QD3D12ObjectHandle addNonOwningToPool(QD3D12ObjectPool<QD3D12Resource> *pool, + ID3D12Resource *resource, + D3D12_RESOURCE_STATES state) + { + Q_ASSERT(resource); + return pool->add({ resource, state, resource->GetDesc(), nullptr, nullptr, 0, false }); + } + + void releaseResources() + { + if (owns) { + // order matters: resource first, then the allocation + resource->Release(); + if (allocation) + allocation->Release(); + } + } +}; + +struct QD3D12Pipeline +{ + enum Type { + Graphics, + Compute + }; + Type type; + ID3D12PipelineState *pso; + + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Pipeline> *pool, + Type type, + ID3D12PipelineState *pso) + { + return pool->add({ type, pso }); + } + + void releaseResources() + { + pso->Release(); + } +}; + +struct QD3D12RootSignature +{ + ID3D12RootSignature *rootSig; + + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12RootSignature> *pool, + ID3D12RootSignature *rootSig) + { + return pool->add({ rootSig }); + } + + void releaseResources() + { + rootSig->Release(); + } +}; + +struct QD3D12ReleaseQueue +{ + void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool, + QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool, + QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool) + { + this->resourcePool = resourcePool; + this->pipelinePool = pipelinePool; + this->rootSignaturePool = rootSignaturePool; + } + + void deferredReleaseResource(const QD3D12ObjectHandle &handle); + void deferredReleaseResourceWithViews(const QD3D12ObjectHandle &handle, + QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount); + void deferredReleasePipeline(const QD3D12ObjectHandle &handle); + void deferredReleaseRootSignature(const QD3D12ObjectHandle &handle); + void deferredReleaseCallback(std::function<void(void*)> callback, void *userData); + void deferredReleaseResourceAndAllocation(ID3D12Resource *resource, + D3D12MA::Allocation *allocation); + void deferredReleaseDescriptorHeap(ID3D12DescriptorHeap *heap); + void deferredReleaseViews(QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount); + + void activatePendingDeferredReleaseRequests(int frameSlot); + void executeDeferredReleases(int frameSlot, bool forced = false); + void releaseAll(); + + struct DeferredReleaseEntry { + enum Type { + Resource, + Pipeline, + RootSignature, + Callback, + ResourceAndAllocation, + DescriptorHeap, + Views + }; + Type type = Resource; + std::optional<int> frameSlotToBeReleasedIn; + QD3D12ObjectHandle handle; + QD3D12CpuDescriptorPool *poolForViews = nullptr; + QD3D12Descriptor viewsStart; + int viewCount = 0; + std::function<void(void*)> callback = nullptr; + void *callbackUserData = nullptr; + QPair<ID3D12Resource *, D3D12MA::Allocation *> resourceAndAllocation = {}; + ID3D12DescriptorHeap *descriptorHeap = nullptr; + }; + QVector<DeferredReleaseEntry> queue; + QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; + QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool = nullptr; + QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool = nullptr; +}; + +struct QD3D12CommandBuffer; + +struct QD3D12ResourceBarrierGenerator +{ + static const int PREALLOC = 16; + + void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool) + { + this->resourcePool = resourcePool; + } + + void addTransitionBarrier(const QD3D12ObjectHandle &resourceHandle, D3D12_RESOURCE_STATES stateAfter); + void enqueueBufferedTransitionBarriers(QD3D12CommandBuffer *cbD); + void enqueueSubresourceTransitionBarrier(QD3D12CommandBuffer *cbD, + const QD3D12ObjectHandle &resourceHandle, + UINT subresource, + D3D12_RESOURCE_STATES stateBefore, + D3D12_RESOURCE_STATES stateAfter); + void enqueueUavBarrier(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &resourceHandle); + + struct TransitionResourceBarrier { + QD3D12ObjectHandle resourceHandle; + D3D12_RESOURCE_STATES stateBefore; + D3D12_RESOURCE_STATES stateAfter; + }; + QVarLengthArray<TransitionResourceBarrier, PREALLOC> transitionResourceBarriers; + QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; +}; + +struct QD3D12ShaderBytecodeCache +{ + struct Shader { + Shader() = default; + Shader(const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) + : bytecode(bytecode), nativeResourceBindingMap(rbm) + { } + QByteArray bytecode; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + }; + + QHash<QRhiShaderStage, Shader> data; + + void insertWithCapacityLimit(const QRhiShaderStage &key, const Shader &s); +}; + +struct QD3D12ShaderVisibleDescriptorHeap +{ + bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE type, quint32 perFrameDescriptorCount); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + QD3D12DescriptorHeap heap; + QD3D12DescriptorHeap perFrameHeapSlice[QD3D12_FRAMES_IN_FLIGHT]; +}; + +// wrap foreign struct so we can legally supply equality operators and qHash: +struct Q_D3D12_SAMPLER_DESC +{ + D3D12_SAMPLER_DESC desc; + + friend bool operator==(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept + { + return lhs.desc.Filter == rhs.desc.Filter + && lhs.desc.AddressU == rhs.desc.AddressU + && lhs.desc.AddressV == rhs.desc.AddressV + && lhs.desc.AddressW == rhs.desc.AddressW + && lhs.desc.MipLODBias == rhs.desc.MipLODBias + && lhs.desc.MaxAnisotropy == rhs.desc.MaxAnisotropy + && lhs.desc.ComparisonFunc == rhs.desc.ComparisonFunc + // BorderColor is never used, skip it + && lhs.desc.MinLOD == rhs.desc.MinLOD + && lhs.desc.MaxLOD == rhs.desc.MaxLOD; + } + + friend bool operator!=(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept + { + return !(lhs == rhs); + } + + friend size_t qHash(const Q_D3D12_SAMPLER_DESC &key, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, key.desc.Filter); + seed = hash(seed, key.desc.AddressU); + seed = hash(seed, key.desc.AddressV); + seed = hash(seed, key.desc.AddressW); + seed = hash(seed, key.desc.MipLODBias); + seed = hash(seed, key.desc.MaxAnisotropy); + seed = hash(seed, key.desc.ComparisonFunc); + // BorderColor is never used, skip it + seed = hash(seed, key.desc.MinLOD); + seed = hash(seed, key.desc.MaxLOD); + return seed; + } +}; + +struct QD3D12SamplerManager +{ + const quint32 MAX_SAMPLERS = 512; + + bool create(ID3D12Device *device); + void destroy(); + + QD3D12Descriptor getShaderVisibleDescriptor(const D3D12_SAMPLER_DESC &desc); + + ID3D12Device *device = nullptr; + QD3D12ShaderVisibleDescriptorHeap shaderVisibleSamplerHeap; + QHash<Q_D3D12_SAMPLER_DESC, QD3D12Descriptor> gpuMap; +}; + +enum QD3D12Stage { VS = 0, HS, DS, GS, PS, CS }; + +static inline QD3D12Stage qd3d12_stage(QRhiShaderStage::Type type) +{ + switch (type) { + case QRhiShaderStage::Vertex: + return VS; + case QRhiShaderStage::TessellationControl: + return HS; + case QRhiShaderStage::TessellationEvaluation: + return DS; + case QRhiShaderStage::Geometry: + return GS; + case QRhiShaderStage::Fragment: + return PS; + case QRhiShaderStage::Compute: + return CS; + } + Q_UNREACHABLE_RETURN(VS); +} + +static inline D3D12_SHADER_VISIBILITY qd3d12_stageToVisibility(QD3D12Stage s) +{ + switch (s) { + case VS: + return D3D12_SHADER_VISIBILITY_VERTEX; + case HS: + return D3D12_SHADER_VISIBILITY_HULL; + case DS: + return D3D12_SHADER_VISIBILITY_DOMAIN; + case GS: + return D3D12_SHADER_VISIBILITY_GEOMETRY; + case PS: + return D3D12_SHADER_VISIBILITY_PIXEL; + case CS: + return D3D12_SHADER_VISIBILITY_ALL; + } + Q_UNREACHABLE_RETURN(D3D12_SHADER_VISIBILITY_ALL); +} + +static inline QRhiShaderResourceBinding::StageFlag qd3d12_stageToSrb(QD3D12Stage s) +{ + switch (s) { + case VS: + return QRhiShaderResourceBinding::VertexStage; + case HS: + return QRhiShaderResourceBinding::TessellationControlStage; + case DS: + return QRhiShaderResourceBinding::TessellationEvaluationStage; + case GS: + return QRhiShaderResourceBinding::GeometryStage; + case PS: + return QRhiShaderResourceBinding::FragmentStage; + case CS: + return QRhiShaderResourceBinding::ComputeStage; + } + Q_UNREACHABLE_RETURN(QRhiShaderResourceBinding::VertexStage); +} + +struct QD3D12ShaderStageData +{ + bool valid = false; // to allow simple arrays where unused stages are indicated by !valid + QD3D12Stage stage = VS; + QShader::NativeResourceBindingMap nativeResourceBindingMap; +}; + +struct QD3D12ShaderResourceBindings; + +struct QD3D12ShaderResourceVisitor +{ + enum StorageOp { Load = 0, Store, LoadStore }; + + QD3D12ShaderResourceVisitor(const QD3D12ShaderResourceBindings *srb, + const QD3D12ShaderStageData *stageData, + int stageCount) + : srb(srb), + stageData(stageData), + stageCount(stageCount) + { + } + + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::UniformBufferData &, int, int)> uniformBuffer = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> texture = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> sampler = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageImageData &, StorageOp, int)> storageImage = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageBufferData &, StorageOp, int)> storageBuffer = nullptr; + + void visit(); + + const QD3D12ShaderResourceBindings *srb; + const QD3D12ShaderStageData *stageData; + int stageCount; +}; + +struct QD3D12Readback +{ + // common + int frameSlot = -1; + QRhiReadbackResult *result = nullptr; + QD3D12StagingArea staging; + quint32 byteSize = 0; + // textures + quint32 bytesPerLine = 0; + QSize pixelSize; + QRhiTexture::Format format = QRhiTexture::UnknownFormat; + quint32 stagingRowPitch = 0; +}; + +struct QD3D12MipmapGenerator +{ + bool create(QRhiD3D12 *rhiD); + void destroy(); + void generate(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &textureHandle); + + QRhiD3D12 *rhiD; + QD3D12ObjectHandle rootSigHandle; + QD3D12ObjectHandle pipelineHandle; +}; + +struct QD3D12MemoryAllocator +{ + bool create(ID3D12Device *device, IDXGIAdapter1 *adapter); + void destroy(); + + HRESULT createResource(D3D12_HEAP_TYPE heapType, + const D3D12_RESOURCE_DESC *resourceDesc, + D3D12_RESOURCE_STATES initialState, + const D3D12_CLEAR_VALUE *optimizedClearValue, + D3D12MA::Allocation **maybeAllocation, + REFIID riidResource, + void **ppvResource); + + void getBudget(D3D12MA::Budget *localBudget, D3D12MA::Budget *nonLocalBudget); + + bool isUsingD3D12MA() const { return allocator != nullptr; } + + ID3D12Device *device = nullptr; + D3D12MA::Allocator *allocator = nullptr; +}; + +struct QD3D12Buffer : public QRhiBuffer +{ + QD3D12Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QD3D12Buffer(); + + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + void executeHostWritesForFrameSlot(int frameSlot); + + QD3D12ObjectHandle handles[QD3D12_FRAMES_IN_FLIGHT] = {}; + struct HostWrite { + quint32 offset; + QRhiBufferData data; + }; + QVarLengthArray<HostWrite, 16> pendingHostWrites[QD3D12_FRAMES_IN_FLIGHT]; + friend class QRhiD3D12; + friend struct QD3D12CommandBuffer; +}; + +struct QD3D12RenderBuffer : public QRhiRenderBuffer +{ + QD3D12RenderBuffer(QRhiImplementation *rhi, + Type type, + const QSize &pixelSize, + int sampleCount, + Flags flags, + QRhiTexture::Format backingFormatHint); + ~QD3D12RenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + static const DXGI_FORMAT DS_FORMAT = DXGI_FORMAT_D24_UNORM_S8_UINT; + + QD3D12ObjectHandle handle; + QD3D12Descriptor rtv; + QD3D12Descriptor dsv; + DXGI_FORMAT dxgiFormat; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12Texture : public QRhiTexture +{ + QD3D12Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QD3D12Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + void setNativeLayout(int layout) override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + + QD3D12ObjectHandle handle; + QD3D12Descriptor srv; + DXGI_FORMAT dxgiFormat; + DXGI_FORMAT srvFormat; + DXGI_FORMAT rtFormat; // RTV/DSV/UAV + uint mipLevelCount; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D12; + friend struct QD3D12CommandBuffer; +}; + +struct QD3D12Sampler : public QRhiSampler +{ + QD3D12Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QD3D12Sampler(); + void destroy() override; + bool create() override; + + QD3D12Descriptor lookupOrCreateShaderVisibleDescriptor(); + + D3D12_SAMPLER_DESC desc = {}; + QD3D12Descriptor shaderVisibleDescriptor; +}; + +struct QD3D12RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QD3D12RenderPassDescriptor(QRhiImplementation *rhi); + ~QD3D12RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + + void updateSerializedFormat(); + + static const int MAX_COLOR_ATTACHMENTS = 8; + int colorAttachmentCount = 0; + bool hasDepthStencil = false; + int colorFormat[MAX_COLOR_ATTACHMENTS]; + int dsFormat; + QVector<quint32> serializedFormatData; +}; + +struct QD3D12RenderTargetData +{ + QD3D12RenderTargetData(QRhiImplementation *) { } + + QD3D12RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + static const int MAX_COLOR_ATTACHMENTS = QD3D12RenderPassDescriptor::MAX_COLOR_ATTACHMENTS; + D3D12_CPU_DESCRIPTOR_HANDLE rtv[MAX_COLOR_ATTACHMENTS]; + D3D12_CPU_DESCRIPTOR_HANDLE dsv; +}; + +struct QD3D12SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QD3D12SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QD3D12SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QD3D12RenderTargetData d; +}; + +struct QD3D12TextureRenderTarget : public QRhiTextureRenderTarget +{ + QD3D12TextureRenderTarget(QRhiImplementation *rhi, + const QRhiTextureRenderTargetDescription &desc, + Flags flags); + ~QD3D12TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QD3D12RenderTargetData d; + bool ownsRtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; + QD3D12Descriptor rtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; + bool ownsDsv = false; + QD3D12Descriptor dsv; + friend class QRhiD3D12; +}; + +struct QD3D12ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QD3D12ShaderResourceBindings(QRhiImplementation *rhi); + ~QD3D12ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QD3D12ObjectHandle createRootSignature(const QD3D12ShaderStageData *stageData, int stageCount); + + struct VisitorData { + QVarLengthArray<D3D12_ROOT_PARAMETER1, 2> cbParams[6]; + + D3D12_ROOT_PARAMETER1 srvTables[6] = {}; + QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> srvRanges[6]; + quint32 currentSrvRangeOffset[6] = {}; + + QVarLengthArray<D3D12_ROOT_PARAMETER1, 4> samplerTables[6]; + std::array<D3D12_DESCRIPTOR_RANGE1, 16> samplerRanges[6] = {}; + int samplerRangeHeads[6] = {}; + + D3D12_ROOT_PARAMETER1 uavTables[6] = {}; + QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> uavRanges[6]; + quint32 currentUavRangeOffset[6] = {}; + } visitorData; + + + void visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int shaderRegister, + int binding); + void visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + void visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + + bool hasDynamicOffset = false; + uint generation = 0; + + friend class QRhiD3D12; + friend struct QD3D12ShaderResourceVisitor; +}; + +struct QD3D12GraphicsPipeline : public QRhiGraphicsPipeline +{ + QD3D12GraphicsPipeline(QRhiImplementation *rhi); + ~QD3D12GraphicsPipeline(); + void destroy() override; + bool create() override; + + QD3D12ObjectHandle handle; + QD3D12ObjectHandle rootSigHandle; + std::array<QD3D12ShaderStageData, 5> stageData; + D3D12_PRIMITIVE_TOPOLOGY topology; + UINT viewInstanceMask = 0; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12ComputePipeline : public QRhiComputePipeline +{ + QD3D12ComputePipeline(QRhiImplementation *rhi); + ~QD3D12ComputePipeline(); + void destroy() override; + bool create() override; + + QD3D12ObjectHandle handle; + QD3D12ObjectHandle rootSigHandle; + QD3D12ShaderStageData stageData; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12CommandBuffer : public QRhiCommandBuffer +{ + QD3D12CommandBuffer(QRhiImplementation *rhi); + ~QD3D12CommandBuffer(); + void destroy() override; + + const QRhiNativeHandles *nativeHandles(); + + ID3D12GraphicsCommandList1 *cmdList = nullptr; // not owned + QRhiD3D12CommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + void resetState() + { + recordingPass = NoPass; + currentTarget = nullptr; + + resetPerPassState(); + } + + void resetPerPassState() + { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentIndexBuffer = {}; + currentIndexOffset = 0; + currentIndexFormat = DXGI_FORMAT_R16_UINT; + currentVertexBuffers = {}; + currentVertexOffsets = {}; + } + + // per-frame + PassType recordingPass; + QRhiRenderTarget *currentTarget; + + // per-pass + QD3D12GraphicsPipeline *currentGraphicsPipeline; + QD3D12ComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + QD3D12ObjectHandle currentIndexBuffer; + quint32 currentIndexOffset; + DXGI_FORMAT currentIndexFormat; + std::array<QD3D12ObjectHandle, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexBuffers; + std::array<quint32, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexOffsets; + + // global + double lastGpuTime = 0; + + // per-setShaderResources + struct VisitorData { + QVarLengthArray<QPair<QD3D12ObjectHandle, quint32>, 4> cbufs[6]; + QVarLengthArray<QD3D12Descriptor, 8> srvs[6]; + QVarLengthArray<QD3D12Descriptor, 8> samplers[6]; + QVarLengthArray<QPair<QD3D12ObjectHandle, D3D12_UNORDERED_ACCESS_VIEW_DESC>, 4> uavs[6]; + } visitorData; + + void visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int shaderRegister, + int binding, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets); + void visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + void visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); +}; + +struct QD3D12SwapChain : public QRhiSwapChain +{ + QD3D12SwapChain(QRhiImplementation *rhi); + ~QD3D12SwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + QRhiSwapChainHdrInfo hdrInfo() override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void releaseBuffers(); + void waitCommandCompletionForFrameSlot(int frameSlot); + void addCommandCompletionSignalForCurrentFrameSlot(); + void chooseFormats(); + + QWindow *window = nullptr; + IDXGISwapChain1 *sourceSwapChain1 = nullptr; + IDXGISwapChain3 *swapChain = nullptr; + QSize pixelSize; + UINT swapInterval = 1; + UINT swapChainFlags = 0; + BOOL stereo = false; + DXGI_FORMAT colorFormat; + DXGI_FORMAT srgbAdjustedColorFormat; + DXGI_COLOR_SPACE_TYPE hdrColorSpace; + IDCompositionTarget *dcompTarget = nullptr; + IDCompositionVisual *dcompVisual = nullptr; + static const UINT BUFFER_COUNT = 3; + QD3D12ObjectHandle colorBuffers[BUFFER_COUNT]; + QD3D12Descriptor rtvs[BUFFER_COUNT]; + QD3D12Descriptor rtvsRight[BUFFER_COUNT]; + DXGI_SAMPLE_DESC sampleDesc; + QD3D12ObjectHandle msaaBuffers[BUFFER_COUNT]; + QD3D12Descriptor msaaRtvs[BUFFER_COUNT]; + QD3D12RenderBuffer *ds = nullptr; + UINT currentBackBufferIndex = 0; + QD3D12SwapChainRenderTarget rtWrapper; + QD3D12SwapChainRenderTarget rtWrapperRight; + QD3D12CommandBuffer cbWrapper; + + struct FrameResources { + ID3D12Fence *fence = nullptr; + HANDLE fenceEvent = nullptr; + UINT64 fenceCounter = 0; + ID3D12GraphicsCommandList1 *cmdList = nullptr; + } frameRes[QD3D12_FRAMES_IN_FLIGHT]; + + int currentFrameSlot = 0; // index in frameRes +}; + +template<typename T, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE Type> +struct alignas(void*) QD3D12PipelineStateSubObject +{ + D3D12_PIPELINE_STATE_SUBOBJECT_TYPE type = Type; + T object = {}; +}; + +class QRhiD3D12 : public QRhiImplementation +{ +public: + // 16MB * QD3D12_FRAMES_IN_FLIGHT; buffer and texture upload staging data that + // gets no space from this will get their own temporary staging areas. + static const quint32 SMALL_STAGING_AREA_BYTES_PER_FRAME = 16 * 1024 * 1024; + + static const quint32 SHADER_VISIBLE_CBV_SRV_UAV_HEAP_PER_FRAME_START_SIZE = 16384; + + QRhiD3D12(QRhiD3D12InitParams *params, QRhiD3D12NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void waitGpu(); + DXGI_SAMPLE_DESC effectiveSampleDesc(int sampleCount, DXGI_FORMAT format) const; + bool ensureDirectCompositionDevice(); + bool startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList1 **cmdList); + void enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); + void finishActiveReadbacks(bool forced = false); + bool ensureShaderVisibleDescriptorHeapCapacity(QD3D12ShaderVisibleDescriptorHeap *h, + D3D12_DESCRIPTOR_HEAP_TYPE type, + int frameSlot, + quint32 neededDescriptorCount, + bool *gotNew); + void bindShaderVisibleHeaps(QD3D12CommandBuffer *cbD); + + bool debugLayer = false; + ID3D12Device2 *dev = nullptr; + D3D_FEATURE_LEVEL minimumFeatureLevel = D3D_FEATURE_LEVEL(0); + LUID adapterLuid = {}; + bool importedDevice = false; + bool importedCommandQueue = false; + QRhi::Flags rhiFlags; + IDXGIFactory2 *dxgiFactory = nullptr; + bool supportsAllowTearing = false; + IDXGIAdapter1 *activeAdapter = nullptr; + QRhiDriverInfo driverInfoStruct; + QRhiD3D12NativeHandles nativeHandlesStruct; + bool deviceLost = false; + ID3D12CommandQueue *cmdQueue = nullptr; + ID3D12Fence *fullFence = nullptr; + HANDLE fullFenceEvent = nullptr; + UINT64 fullFenceCounter = 0; + ID3D12CommandAllocator *cmdAllocators[QD3D12_FRAMES_IN_FLIGHT] = {}; + QD3D12MemoryAllocator vma; + QD3D12CpuDescriptorPool rtvPool; + QD3D12CpuDescriptorPool dsvPool; + QD3D12CpuDescriptorPool cbvSrvUavPool; + QD3D12ObjectPool<QD3D12Resource> resourcePool; + QD3D12ObjectPool<QD3D12Pipeline> pipelinePool; + QD3D12ObjectPool<QD3D12RootSignature> rootSignaturePool; + QD3D12ReleaseQueue releaseQueue; + QD3D12ResourceBarrierGenerator barrierGen; + QD3D12SamplerManager samplerMgr; + QD3D12MipmapGenerator mipmapGen; + QD3D12StagingArea smallStagingAreas[QD3D12_FRAMES_IN_FLIGHT]; + QD3D12ShaderVisibleDescriptorHeap shaderVisibleCbvSrvUavHeap; + UINT64 timestampTicksPerSecond = 0; + QD3D12QueryHeap timestampQueryHeap; + QD3D12StagingArea timestampReadbackArea; + IDCompositionDevice *dcompDevice = nullptr; + QD3D12SwapChain *currentSwapChain = nullptr; + QSet<QD3D12SwapChain *> swapchains; + QD3D12ShaderBytecodeCache shaderBytecodeCache; + QVarLengthArray<QD3D12Readback, 4> activeReadbacks; + bool offscreenActive = false; + QD3D12CommandBuffer *offscreenCb[QD3D12_FRAMES_IN_FLIGHT] = {}; + + struct { + bool multiView = false; + bool textureViewFormat = false; + } caps; +}; + +QT_END_NAMESPACE + +#endif // __ID3D12Device2_INTERFACE_DEFINED__ + +#endif diff --git a/src/gui/rhi/qrhid3dhelpers.cpp b/src/gui/rhi/qrhid3dhelpers.cpp new file mode 100644 index 0000000000..216c358cbe --- /dev/null +++ b/src/gui/rhi/qrhid3dhelpers.cpp @@ -0,0 +1,172 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhid3dhelpers_p.h" +#include <QtCore/private/qsystemlibrary_p.h> +#include <QtCore/private/qsystemerror_p.h> + +QT_BEGIN_NAMESPACE + +namespace QRhiD3D { + +bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result) +{ + bool ok = false; + QRect wr = w->geometry(); + wr = QRect(wr.topLeft() * w->devicePixelRatio(), wr.size() * w->devicePixelRatio()); + const QPoint center = wr.center(); + IDXGIOutput *currentOutput = nullptr; + IDXGIOutput *output = nullptr; + for (UINT i = 0; adapter->EnumOutputs(i, &output) != DXGI_ERROR_NOT_FOUND; ++i) { + DXGI_OUTPUT_DESC desc; + output->GetDesc(&desc); + const RECT r = desc.DesktopCoordinates; + const QRect dr(QPoint(r.left, r.top), QPoint(r.right - 1, r.bottom - 1)); + if (dr.contains(center)) { + currentOutput = output; + break; + } else { + output->Release(); + } + } + if (currentOutput) { + ok = SUCCEEDED(currentOutput->QueryInterface(__uuidof(IDXGIOutput6), reinterpret_cast<void **>(result))); + currentOutput->Release(); + } + return ok; +} + +bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result) +{ + bool ok = false; + IDXGIOutput6 *out6 = nullptr; + if (output6ForWindow(w, adapter, &out6)) { + ok = SUCCEEDED(out6->GetDesc1(result)); + out6->Release(); + } + return ok; +} + +float sdrWhiteLevelInNits(const DXGI_OUTPUT_DESC1 &outputDesc) +{ + QVector<DISPLAYCONFIG_PATH_INFO> pathInfos; + uint32_t pathInfoCount, modeInfoCount; + LONG result; + do { + if (GetDisplayConfigBufferSizes(QDC_ONLY_ACTIVE_PATHS, &pathInfoCount, &modeInfoCount) == ERROR_SUCCESS) { + pathInfos.resize(pathInfoCount); + QVector<DISPLAYCONFIG_MODE_INFO> modeInfos(modeInfoCount); + result = QueryDisplayConfig(QDC_ONLY_ACTIVE_PATHS, &pathInfoCount, pathInfos.data(), &modeInfoCount, modeInfos.data(), nullptr); + } else { + return 200.0f; + } + } while (result == ERROR_INSUFFICIENT_BUFFER); + + MONITORINFOEX monitorInfo = {}; + monitorInfo.cbSize = sizeof(monitorInfo); + GetMonitorInfo(outputDesc.Monitor, &monitorInfo); + + for (const DISPLAYCONFIG_PATH_INFO &info : pathInfos) { + DISPLAYCONFIG_SOURCE_DEVICE_NAME deviceName = {}; + deviceName.header.type = DISPLAYCONFIG_DEVICE_INFO_GET_SOURCE_NAME; + deviceName.header.size = sizeof(deviceName); + deviceName.header.adapterId = info.sourceInfo.adapterId; + deviceName.header.id = info.sourceInfo.id; + if (DisplayConfigGetDeviceInfo(&deviceName.header) == ERROR_SUCCESS) { + if (!wcscmp(monitorInfo.szDevice, deviceName.viewGdiDeviceName)) { + DISPLAYCONFIG_SDR_WHITE_LEVEL whiteLevel = {}; + whiteLevel.header.type = DISPLAYCONFIG_DEVICE_INFO_GET_SDR_WHITE_LEVEL; + whiteLevel.header.size = sizeof(DISPLAYCONFIG_SDR_WHITE_LEVEL); + whiteLevel.header.adapterId = info.targetInfo.adapterId; + whiteLevel.header.id = info.targetInfo.id; + if (DisplayConfigGetDeviceInfo(&whiteLevel.header) == ERROR_SUCCESS) + return whiteLevel.SDRWhiteLevel * 80 / 1000.0f; + } + } + } + + return 200.0f; +} + +pD3DCompile resolveD3DCompile() +{ + for (const wchar_t *libraryName : {L"D3DCompiler_47", L"D3DCompiler_43"}) { + QSystemLibrary library(libraryName); + if (library.load()) { + if (auto symbol = library.resolve("D3DCompile")) + return reinterpret_cast<pD3DCompile>(symbol); + } else { + qWarning("Failed to load D3DCompiler_47/43.dll"); + } + } + return nullptr; +} + +IDCompositionDevice *createDirectCompositionDevice() +{ + QSystemLibrary dcomplib(QStringLiteral("dcomp")); + typedef HRESULT (__stdcall *DCompositionCreateDeviceFuncPtr)( + _In_opt_ IDXGIDevice *dxgiDevice, + _In_ REFIID iid, + _Outptr_ void **dcompositionDevice); + DCompositionCreateDeviceFuncPtr func = reinterpret_cast<DCompositionCreateDeviceFuncPtr>( + dcomplib.resolve("DCompositionCreateDevice")); + if (!func) { + qWarning("Unable to resolve DCompositionCreateDevice, perhaps dcomp.dll is missing?"); + return nullptr; + } + IDCompositionDevice *device = nullptr; + HRESULT hr = func(nullptr, __uuidof(IDCompositionDevice), reinterpret_cast<void **>(&device)); + if (FAILED(hr)) { + qWarning("Failed to create Direct Composition device: %s", + qPrintable(QSystemError::windowsComString(hr))); + return nullptr; + } + return device; +} + +#ifdef QRHI_D3D12_HAS_DXC +std::pair<IDxcCompiler *, IDxcLibrary *> createDxcCompiler() +{ + QSystemLibrary dxclib(QStringLiteral("dxcompiler")); + // this will not be in the system library location, hence onlySystemDirectory==false + if (!dxclib.load(false)) { + qWarning("Failed to load dxcompiler.dll"); + return {}; + } + DxcCreateInstanceProc func = reinterpret_cast<DxcCreateInstanceProc>(dxclib.resolve("DxcCreateInstance")); + if (!func) { + qWarning("Unable to resolve DxcCreateInstance"); + return {}; + } + IDxcCompiler *compiler = nullptr; + HRESULT hr = func(CLSID_DxcCompiler, __uuidof(IDxcCompiler), reinterpret_cast<void**>(&compiler)); + if (FAILED(hr)) { + qWarning("Failed to create dxc compiler instance: %s", + qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + IDxcLibrary *library = nullptr; + hr = func(CLSID_DxcLibrary, __uuidof(IDxcLibrary), reinterpret_cast<void**>(&library)); + if (FAILED(hr)) { + qWarning("Failed to create dxc library instance: %s", + qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + return { compiler, library }; +} +#endif + +void fillDriverInfo(QRhiDriverInfo *info, const DXGI_ADAPTER_DESC1 &desc) +{ + const QString name = QString::fromUtf16(reinterpret_cast<const char16_t *>(desc.Description)); + info->deviceName = name.toUtf8(); + info->deviceId = desc.DeviceId; + info->vendorId = desc.VendorId; + info->deviceType = (desc.Flags & DXGI_ADAPTER_FLAG_SOFTWARE) ? QRhiDriverInfo::CpuDevice + : QRhiDriverInfo::UnknownDevice; +} + +} // namespace + +QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhid3dhelpers_p.h b/src/gui/rhi/qrhid3dhelpers_p.h new file mode 100644 index 0000000000..f31cdc8d11 --- /dev/null +++ b/src/gui/rhi/qrhid3dhelpers_p.h @@ -0,0 +1,53 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHID3DHELPERS_P_H +#define QRHID3DHELPERS_P_H + +// +// W A R N I N G +// ------------- +// +// This file is not part of the Qt API. It exists purely as an +// implementation detail. This header file may change from version to +// version without notice, or even be removed. +// +// We mean it. +// + +#include <rhi/qrhi.h> + +#include <QtGui/qwindow.h> + +#include <dxgi1_6.h> +#include <dcomp.h> +#include <d3dcompiler.h> + +#if __has_include(<dxcapi.h>) +#include <dxcapi.h> +#define QRHI_D3D12_HAS_DXC +#endif + +QT_BEGIN_NAMESPACE + +namespace QRhiD3D { + +bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result); +bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result); +float sdrWhiteLevelInNits(const DXGI_OUTPUT_DESC1 &outputDesc); + +pD3DCompile resolveD3DCompile(); + +IDCompositionDevice *createDirectCompositionDevice(); + +#ifdef QRHI_D3D12_HAS_DXC +std::pair<IDxcCompiler *, IDxcLibrary *> createDxcCompiler(); +#endif + +void fillDriverInfo(QRhiDriverInfo *info, const DXGI_ADAPTER_DESC1 &desc); + +} // namespace + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qrhigles2.cpp b/src/gui/rhi/qrhigles2.cpp index 05465e40cc..62830c291d 100644 --- a/src/gui/rhi/qrhigles2.cpp +++ b/src/gui/rhi/qrhigles2.cpp @@ -1,48 +1,14 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhigles2_p_p.h" -#include <QWindow> +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhigles2_p.h" #include <QOffscreenSurface> #include <QOpenGLContext> +#include <QtCore/qmap.h> #include <QtGui/private/qopenglextensions_p.h> #include <QtGui/private/qopenglprogrambinarycache_p.h> +#include <QtGui/private/qwindow_p.h> +#include <qpa/qplatformopenglcontext.h> #include <qmath.h> QT_BEGIN_NAMESPACE @@ -62,13 +28,17 @@ QT_BEGIN_NAMESPACE /*! \class QRhiGles2InitParams - \internal \inmodule QtGui + \since 6.6 \brief OpenGL specific initialization parameters. - An OpenGL-based QRhi needs an already created QOffscreenSurface at minimum. - Additionally, while optional, it is recommended that the QWindow the first - QRhiSwapChain will target is passed in as well. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + An OpenGL-based QRhi needs an already created QSurface that can be used in + combination with QOpenGLContext. Most commonly, this is a QOffscreenSurface + in practice. Additionally, while optional, it is recommended that the QWindow + the first QRhiSwapChain will target is passed in as well. \badcode QOffscreenSurface *fallbackSurface = QRhiGles2InitParams::newFallbackSurface(); @@ -85,21 +55,21 @@ QT_BEGIN_NAMESPACE thread) are satisfied. The implicitly created context is destroyed automatically together with the QRhi. - The QSurfaceFormat for the context is specified in \l format. The + The QSurfaceFormat for the context is specified in \c format. The constructor sets this to QSurfaceFormat::defaultFormat() so applications - that use QSurfaceFormat::setDefaultFormat() do not need to set the format - again. + that call QSurfaceFormat::setDefaultFormat() with the appropriate settings + before the constructor runs will not need to change value of \c format. - \note The depth and stencil buffer sizes are set automatically to 24 and 8 - when no size was explicitly set for these buffers in \l format. As there - are possible adjustments to \l format, applications can use - adjustedFormat() to query the effective format that is passed to - QOpenGLContext::setFormat() internally. + \note Remember to set the depth and stencil buffer sizes to 24 and 8 when + the renderer relies on depth or stencil testing, either in the global + default QSurfaceFormat, or, alternatively, separately in all the involved + QSurfaceFormat instances: in \c format, the format argument passed to + newFallbackSurface(), and on any QWindow that is used with the QRhi. - A QOffscreenSurface has to be specified in \l fallbackSurface. In order to - prevent mistakes in threaded situations, this is never created - automatically by the QRhi since, like QWindow, QOffscreenSurface can only - be created on the gui/main thread. + A QSurface has to be specified in \c fallbackSurface. In order to prevent + mistakes in threaded situations, this is never created automatically by the + QRhi because, like QWindow, instances of QSurface subclasses can often be + created on the gui/main thread only. As a convenience, applications can use newFallbackSurface() which creates and returns a QOffscreenSurface that is compatible with the QOpenGLContext @@ -107,23 +77,27 @@ QT_BEGIN_NAMESPACE of the returned QOffscreenSurface is transferred to the caller and the QRhi will not destroy it. - \note QRhiSwapChain can only target QWindow instances that have their - surface type set to QSurface::OpenGLSurface. + \note With the OpenGL backend, QRhiSwapChain can only target QWindow + instances that have their surface type set to QSurface::OpenGLSurface or + QSurface::RasterGLSurface. - \note \l window is optional. It is recommended to specify it whenever + \note \c window is optional. It is recommended to specify it whenever possible, in order to avoid problems on multi-adapter and multi-screen - systems. When \l window is not set, the very first - QOpenGLContext::makeCurrent() happens with \l fallbackSurface which may be + systems. When \c window is not set, the very first + QOpenGLContext::makeCurrent() happens with \c fallbackSurface which may be an invisible window on some platforms (for example, Windows) and that may trigger unexpected problems in some cases. + In case resource sharing with an existing QOpenGLContext is desired, \c + shareContext can be set to an existing QOpenGLContext. Alternatively, + Qt::AA_ShareOpenGLContexts is honored as well, when enabled. + \section2 Working with existing OpenGL contexts When interoperating with another graphics engine, it may be necessary to get a QRhi instance that uses the same OpenGL context. This can be achieved by passing a pointer to a QRhiGles2NativeHandles to QRhi::create(). The - \l{QRhiGles2NativeHandles::context}{context} must be set to a non-null - value. + \c{QRhiGles2NativeHandles::context} must be set to a non-null value then. An alternative approach is to create a QOpenGLContext that \l{QOpenGLContext::setShareContext()}{shares resources} with the other @@ -134,12 +108,50 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiGles2InitParams::format + + The QSurfaceFormat, initialized to QSurfaceFormat::defaultFormat() by default. +*/ + +/*! + \variable QRhiGles2InitParams::fallbackSurface + + A QSurface compatible with \l format. Typically a QOffscreenSurface. + Providing this is mandatory. Be aware of the threading implications: a + QOffscreenSurface, like QWindow, must only ever be created and destroyed on + the main (gui) thread, even if the QRhi is created and operates on another + thread. +*/ + +/*! + \variable QRhiGles2InitParams::window + + Optional, but setting it is recommended when targeting a QWindow with the + QRhi. +*/ + +/*! + \variable QRhiGles2InitParams::shareContext + + Optional, the QOpenGLContext to share resource with. QRhi creates its own + context, and setting this member to a valid QOpenGLContext leads to calling + \l{QOpenGLContext::setShareContext()}{setShareContext()} with it. +*/ + +/*! \class QRhiGles2NativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the OpenGL context used by the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiGles2NativeHandles::context +*/ + #ifndef GL_BGRA #define GL_BGRA 0x80E1 #endif @@ -204,6 +216,10 @@ QT_BEGIN_NAMESPACE #define GL_DEPTH_COMPONENT32F 0x8CAC #endif +#ifndef GL_UNSIGNED_INT_24_8 +#define GL_UNSIGNED_INT_24_8 0x84FA +#endif + #ifndef GL_STENCIL_INDEX #define GL_STENCIL_INDEX 0x1901 #endif @@ -229,7 +245,7 @@ QT_BEGIN_NAMESPACE #endif #ifndef GL_FRAMEBUFFER_SRGB -#define GL_FRAMEBUFFER_SRGB 0x8DB9 +#define GL_FRAMEBUFFER_SRGB 0x8DB9 #endif #ifndef GL_READ_FRAMEBUFFER @@ -280,16 +296,48 @@ QT_BEGIN_NAMESPACE #define GL_COMPUTE_SHADER 0x91B9 #endif -#ifndef GL_ALL_BARRIER_BITS -#define GL_ALL_BARRIER_BITS 0xFFFFFFFF +#ifndef GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT +#define GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT 0x00000001 +#endif + +#ifndef GL_ELEMENT_ARRAY_BARRIER_BIT +#define GL_ELEMENT_ARRAY_BARRIER_BIT 0x00000002 +#endif + +#ifndef GL_UNIFORM_BARRIER_BIT +#define GL_UNIFORM_BARRIER_BIT 0x00000004 +#endif + +#ifndef GL_BUFFER_UPDATE_BARRIER_BIT +#define GL_BUFFER_UPDATE_BARRIER_BIT 0x00000200 +#endif + +#ifndef GL_SHADER_STORAGE_BARRIER_BIT +#define GL_SHADER_STORAGE_BARRIER_BIT 0x00002000 +#endif + +#ifndef GL_TEXTURE_FETCH_BARRIER_BIT +#define GL_TEXTURE_FETCH_BARRIER_BIT 0x00000008 #endif #ifndef GL_SHADER_IMAGE_ACCESS_BARRIER_BIT #define GL_SHADER_IMAGE_ACCESS_BARRIER_BIT 0x00000020 #endif -#ifndef GL_SHADER_STORAGE_BARRIER_BIT -#define GL_SHADER_STORAGE_BARRIER_BIT 0x00002000 +#ifndef GL_PIXEL_BUFFER_BARRIER_BIT +#define GL_PIXEL_BUFFER_BARRIER_BIT 0x00000080 +#endif + +#ifndef GL_TEXTURE_UPDATE_BARRIER_BIT +#define GL_TEXTURE_UPDATE_BARRIER_BIT 0x00000100 +#endif + +#ifndef GL_FRAMEBUFFER_BARRIER_BIT +#define GL_FRAMEBUFFER_BARRIER_BIT 0x00000400 +#endif + +#ifndef GL_ALL_BARRIER_BITS +#define GL_ALL_BARRIER_BITS 0xFFFFFFFF #endif #ifndef GL_VERTEX_PROGRAM_POINT_SIZE @@ -312,6 +360,10 @@ QT_BEGIN_NAMESPACE #define GL_TEXTURE_2D_MULTISAMPLE 0x9100 #endif +#ifndef GL_TEXTURE_2D_MULTISAMPLE_ARRAY +#define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102 +#endif + #ifndef GL_TEXTURE_EXTERNAL_OES #define GL_TEXTURE_EXTERNAL_OES 0x8D65 #endif @@ -356,6 +408,126 @@ QT_BEGIN_NAMESPACE #define GL_TEXTURE_WRAP_R 0x8072 #endif +#ifndef GL_TEXTURE_RECTANGLE +#define GL_TEXTURE_RECTANGLE 0x84F5 +#endif + +#ifndef GL_TEXTURE_2D_ARRAY +#define GL_TEXTURE_2D_ARRAY 0x8C1A +#endif + +#ifndef GL_MAX_ARRAY_TEXTURE_LAYERS +#define GL_MAX_ARRAY_TEXTURE_LAYERS 0x88FF +#endif + +#ifndef GL_MAX_VERTEX_UNIFORM_COMPONENTS +#define GL_MAX_VERTEX_UNIFORM_COMPONENTS 0x8B4A +#endif + +#ifndef GL_MAX_FRAGMENT_UNIFORM_COMPONENTS +#define GL_MAX_FRAGMENT_UNIFORM_COMPONENTS 0x8B49 +#endif + +#ifndef GL_MAX_VERTEX_UNIFORM_VECTORS +#define GL_MAX_VERTEX_UNIFORM_VECTORS 0x8DFB +#endif + +#ifndef GL_MAX_FRAGMENT_UNIFORM_VECTORS +#define GL_MAX_FRAGMENT_UNIFORM_VECTORS 0x8DFD +#endif + +#ifndef GL_RGB10_A2 +#define GL_RGB10_A2 0x8059 +#endif + +#ifndef GL_UNSIGNED_INT_2_10_10_10_REV +#define GL_UNSIGNED_INT_2_10_10_10_REV 0x8368 +#endif + +#ifndef GL_MAX_VARYING_COMPONENTS +#define GL_MAX_VARYING_COMPONENTS 0x8B4B +#endif + +#ifndef GL_MAX_VARYING_FLOATS +#define GL_MAX_VARYING_FLOATS 0x8B4B +#endif + +#ifndef GL_MAX_VARYING_VECTORS +#define GL_MAX_VARYING_VECTORS 0x8DFC +#endif + +#ifndef GL_TESS_CONTROL_SHADER +#define GL_TESS_CONTROL_SHADER 0x8E88 +#endif + +#ifndef GL_TESS_EVALUATION_SHADER +#define GL_TESS_EVALUATION_SHADER 0x8E87 +#endif + +#ifndef GL_PATCH_VERTICES +#define GL_PATCH_VERTICES 0x8E72 +#endif + +#ifndef GL_LINE +#define GL_LINE 0x1B01 +#endif + +#ifndef GL_FILL +#define GL_FILL 0x1B02 +#endif + +#ifndef GL_PATCHES +#define GL_PATCHES 0x000E +#endif + +#ifndef GL_GEOMETRY_SHADER +#define GL_GEOMETRY_SHADER 0x8DD9 +#endif + +#ifndef GL_BACK_LEFT +#define GL_BACK_LEFT 0x0402 +#endif + +#ifndef GL_BACK_RIGHT +#define GL_BACK_RIGHT 0x0403 +#endif + +#ifndef GL_TEXTURE_1D +# define GL_TEXTURE_1D 0x0DE0 +#endif + +#ifndef GL_TEXTURE_1D_ARRAY +# define GL_TEXTURE_1D_ARRAY 0x8C18 +#endif + +#ifndef GL_HALF_FLOAT +#define GL_HALF_FLOAT 0x140B +#endif + +#ifndef GL_MAX_VERTEX_OUTPUT_COMPONENTS +#define GL_MAX_VERTEX_OUTPUT_COMPONENTS 0x9122 +#endif + +#ifndef GL_TIMESTAMP +#define GL_TIMESTAMP 0x8E28 +#endif + +#ifndef GL_QUERY_RESULT +#define GL_QUERY_RESULT 0x8866 +#endif + +#ifndef GL_QUERY_RESULT_AVAILABLE +#define GL_QUERY_RESULT_AVAILABLE 0x8867 +#endif + +#ifndef GL_BUFFER +#define GL_BUFFER 0x82E0 +#endif + +#ifndef GL_PROGRAM +#define GL_PROGRAM 0x82E2 +#endif + /*! Constructs a new QRhiGles2InitParams. @@ -367,28 +539,12 @@ QRhiGles2InitParams::QRhiGles2InitParams() } /*! - \return the QSurfaceFormat that will be set on the QOpenGLContext before - calling QOpenGLContext::create(). This format is based on \a format, but - may be adjusted. Applicable only when QRhi creates the context. - Applications are advised to set this format on their QWindow in order to - avoid potential BAD_MATCH failures. - */ -QSurfaceFormat QRhiGles2InitParams::adjustedFormat(const QSurfaceFormat &format) -{ - QSurfaceFormat fmt = format; - - if (fmt.depthBufferSize() == -1) - fmt.setDepthBufferSize(24); - if (fmt.stencilBufferSize() == -1) - fmt.setStencilBufferSize(8); - - return fmt; -} - -/*! \return a new QOffscreenSurface that can be used with a QRhi by passing it via a QRhiGles2InitParams. + When \a format is not specified, its default value is the global default + format settable via QSurfaceFormat::setDefaultFormat(). + \a format is adjusted as appropriate in order to avoid having problems afterwards due to an incompatible context and surface. @@ -400,7 +556,7 @@ QSurfaceFormat QRhiGles2InitParams::adjustedFormat(const QSurfaceFormat &format) */ QOffscreenSurface *QRhiGles2InitParams::newFallbackSurface(const QSurfaceFormat &format) { - QSurfaceFormat fmt = adjustedFormat(format); + QSurfaceFormat fmt = format; // To resolve all fields in the format as much as possible, create a context. // This may be heavy, but allows avoiding BAD_MATCH on some systems. @@ -421,9 +577,10 @@ QOffscreenSurface *QRhiGles2InitParams::newFallbackSurface(const QSurfaceFormat QRhiGles2::QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *importDevice) : ofr(this) { - requestedFormat = QRhiGles2InitParams::adjustedFormat(params->format); + requestedFormat = params->format; fallbackSurface = params->fallbackSurface; maybeWindow = params->window; // may be null + maybeShareContext = params->shareContext; // may be null importedContext = importDevice != nullptr; if (importedContext) { @@ -435,21 +592,50 @@ QRhiGles2::QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *import } } -bool QRhiGles2::ensureContext(QSurface *surface) const +static inline QSurface *currentSurfaceForCurrentContext(QOpenGLContext *ctx) { - bool nativeWindowGone = false; - if (surface && surface->surfaceClass() == QSurface::Window && !surface->surfaceHandle()) { - surface = fallbackSurface; - nativeWindowGone = true; - } + if (QOpenGLContext::currentContext() != ctx) + return nullptr; + + QSurface *currentSurface = ctx->surface(); + if (!currentSurface) + return nullptr; + + if (currentSurface->surfaceClass() == QSurface::Window && !currentSurface->surfaceHandle()) + return nullptr; - if (!surface) - surface = fallbackSurface; + return currentSurface; +} + +QSurface *QRhiGles2::evaluateFallbackSurface() const +{ + // With Apple's deprecated OpenGL support we need to minimize the usage of + // QOffscreenSurface since delicate problems can pop up with + // NSOpenGLContext and drawables. +#if defined(Q_OS_MACOS) + return maybeWindow && maybeWindow->handle() ? static_cast<QSurface *>(maybeWindow) : fallbackSurface; +#else + return fallbackSurface; +#endif +} - if (needsMakeCurrent) - needsMakeCurrent = false; - else if (!nativeWindowGone && QOpenGLContext::currentContext() == ctx && (surface == fallbackSurface || ctx->surface() == surface)) +bool QRhiGles2::ensureContext(QSurface *surface) const +{ + if (!surface) { + // null means any surface is good because not going to render + if (currentSurfaceForCurrentContext(ctx)) + return true; + // if the context is not already current with a valid surface, use our + // fallback surface, but platform specific quirks may apply + surface = evaluateFallbackSurface(); + } else if (surface->surfaceClass() == QSurface::Window && !surface->surfaceHandle()) { + // the window is not usable anymore (no native window underneath), behave as if offscreen + surface = evaluateFallbackSurface(); + } else if (!needsMakeCurrentDueToSwap && currentSurfaceForCurrentContext(ctx) == surface) { + // bail out if the makeCurrent is not necessary return true; + } + needsMakeCurrentDueToSwap = false; if (!ctx->makeCurrent(surface)) { if (ctx->isValid()) { @@ -464,6 +650,58 @@ bool QRhiGles2::ensureContext(QSurface *surface) const return true; } +static inline GLenum toGlCompressedTextureFormat(QRhiTexture::Format format, QRhiTexture::Flags flags) +{ + const bool srgb = flags.testFlag(QRhiTexture::sRGB); + switch (format) { + case QRhiTexture::BC1: + return srgb ? 0x8C4C : 0x83F0; + case QRhiTexture::BC2: + return srgb ? 0x8C4E : 0x83F2; + case QRhiTexture::BC3: + return srgb ? 0x8C4F : 0x83F3; + + case QRhiTexture::ETC2_RGB8: + return srgb ? 0x9275 : 0x9274; + case QRhiTexture::ETC2_RGB8A1: + return srgb ? 0x9277 : 0x9276; + case QRhiTexture::ETC2_RGBA8: + return srgb ? 0x9279 : 0x9278; + + case QRhiTexture::ASTC_4x4: + return srgb ? 0x93D0 : 0x93B0; + case QRhiTexture::ASTC_5x4: + return srgb ? 0x93D1 : 0x93B1; + case QRhiTexture::ASTC_5x5: + return srgb ? 0x93D2 : 0x93B2; + case QRhiTexture::ASTC_6x5: + return srgb ? 0x93D3 : 0x93B3; + case QRhiTexture::ASTC_6x6: + return srgb ? 0x93D4 : 0x93B4; + case QRhiTexture::ASTC_8x5: + return srgb ? 0x93D5 : 0x93B5; + case QRhiTexture::ASTC_8x6: + return srgb ? 0x93D6 : 0x93B6; + case QRhiTexture::ASTC_8x8: + return srgb ? 0x93D7 : 0x93B7; + case QRhiTexture::ASTC_10x5: + return srgb ? 0x93D8 : 0x93B8; + case QRhiTexture::ASTC_10x6: + return srgb ? 0x93D9 : 0x93B9; + case QRhiTexture::ASTC_10x8: + return srgb ? 0x93DA : 0x93BA; + case QRhiTexture::ASTC_10x10: + return srgb ? 0x93DB : 0x93BB; + case QRhiTexture::ASTC_12x10: + return srgb ? 0x93DC : 0x93BC; + case QRhiTexture::ASTC_12x12: + return srgb ? 0x93DD : 0x93BD; + + default: + return 0; // this is reachable, just return an invalid format + } +} + bool QRhiGles2::create(QRhi::Flags flags) { Q_ASSERT(fallbackSurface); @@ -472,9 +710,14 @@ bool QRhiGles2::create(QRhi::Flags flags) if (!importedContext) { ctx = new QOpenGLContext; ctx->setFormat(requestedFormat); - if (QOpenGLContext *shareContext = qt_gl_global_share_context()) { + if (maybeShareContext) { + ctx->setShareContext(maybeShareContext); + ctx->setScreen(maybeShareContext->screen()); + } else if (QOpenGLContext *shareContext = qt_gl_global_share_context()) { ctx->setShareContext(shareContext); ctx->setScreen(shareContext->screen()); + } else if (maybeWindow) { + ctx->setScreen(maybeWindow->screen()); } if (!ctx->create()) { qWarning("QRhiGles2: Failed to create context"); @@ -489,6 +732,40 @@ bool QRhiGles2::create(QRhi::Flags flags) return false; f = static_cast<QOpenGLExtensions *>(ctx->extraFunctions()); + const QSurfaceFormat actualFormat = ctx->format(); + caps.gles = actualFormat.renderableType() == QSurfaceFormat::OpenGLES; + + if (!caps.gles) { + glPolygonMode = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum)>( + ctx->getProcAddress(QByteArrayLiteral("glPolygonMode"))); + + glTexImage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)( + GLenum, GLint, GLint, GLsizei, GLint, GLenum, GLenum, const void *)>( + ctx->getProcAddress(QByteArrayLiteral("glTexImage1D"))); + + glTexStorage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLint, GLenum, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glTexStorage1D"))); + + glTexSubImage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)( + GLenum, GLint, GLint, GLsizei, GLenum, GLenum, const GLvoid *)>( + ctx->getProcAddress(QByteArrayLiteral("glTexSubImage1D"))); + + glCopyTexSubImage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLint, GLint, GLint, + GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glCopyTexSubImage1D"))); + + glCompressedTexImage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)( + GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, const GLvoid *)>( + ctx->getProcAddress(QByteArrayLiteral("glCompressedTexImage1D"))); + + glCompressedTexSubImage1D = reinterpret_cast<void(QOPENGLF_APIENTRYP)( + GLenum, GLint, GLint, GLsizei, GLenum, GLsizei, const GLvoid *)>( + ctx->getProcAddress(QByteArrayLiteral("glCompressedTexSubImage1D"))); + + glFramebufferTexture1D = + reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLenum, GLuint, GLint)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTexture1D"))); + } const char *vendor = reinterpret_cast<const char *>(f->glGetString(GL_VENDOR)); const char *renderer = reinterpret_cast<const char *>(f->glGetString(GL_RENDERER)); @@ -507,26 +784,78 @@ bool QRhiGles2::create(QRhi::Flags flags) if (version) driverInfoStruct.deviceName += QByteArray(version); - const QSurfaceFormat actualFormat = ctx->format(); - caps.ctxMajor = actualFormat.majorVersion(); caps.ctxMinor = actualFormat.minorVersion(); GLint n = 0; f->glGetIntegerv(GL_NUM_COMPRESSED_TEXTURE_FORMATS, &n); - supportedCompressedFormats.resize(n); - if (n > 0) - f->glGetIntegerv(GL_COMPRESSED_TEXTURE_FORMATS, supportedCompressedFormats.data()); + if (n > 0) { + QVarLengthArray<GLint, 16> compressedTextureFormats(n); + f->glGetIntegerv(GL_COMPRESSED_TEXTURE_FORMATS, compressedTextureFormats.data()); + for (GLint format : compressedTextureFormats) + supportedCompressedFormats.insert(format); + + } + // The above looks nice, if only it worked always. With GLES the list we + // query is likely the full list of compressed formats (mostly anything + // that can be decoded). With OpenGL however the list is not required to + // include all formats due to the way the spec is worded. For instance, we + // cannot rely on ASTC formats being present in the list on non-ES. Some + // drivers do include them (Intel, NVIDIA), some don't (Mesa). On the other + // hand, relying on extension strings only is not ok: for example, Intel + // reports GL_KHR_texture_compression_astc_ldr whereas NVIDIA doesn't. So + // the only reasonable thing to do is to query the list always and then see + // if there is something we can add - if not already in there. + std::array<QRhiTexture::Flags, 2> textureVariantFlags; + textureVariantFlags[0] = {}; + textureVariantFlags[1] = QRhiTexture::sRGB; + if (f->hasOpenGLExtension(QOpenGLExtensions::DDSTextureCompression)) { + for (QRhiTexture::Flags f : textureVariantFlags) { + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::BC1, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::BC2, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::BC3, f)); + } + } + if (f->hasOpenGLExtension(QOpenGLExtensions::ETC2TextureCompression)) { + for (QRhiTexture::Flags f : textureVariantFlags) { + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ETC2_RGB8, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ETC2_RGB8A1, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ETC2_RGBA8, f)); + } + } + if (f->hasOpenGLExtension(QOpenGLExtensions::ASTCTextureCompression)) { + for (QRhiTexture::Flags f : textureVariantFlags) { + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_4x4, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_5x4, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_5x5, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_6x5, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_6x6, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_8x5, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_8x6, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_8x8, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_10x5, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_10x8, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_10x10, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_12x10, f)); + supportedCompressedFormats.insert(toGlCompressedTextureFormat(QRhiTexture::ASTC_12x12, f)); + } + } f->glGetIntegerv(GL_MAX_TEXTURE_SIZE, &caps.maxTextureSize); - if (caps.ctxMajor >= 3 || actualFormat.renderableType() == QSurfaceFormat::OpenGL) { + if (!caps.gles || caps.ctxMajor >= 3) { + // non-ES or ES 3.0+ f->glGetIntegerv(GL_MAX_DRAW_BUFFERS, &caps.maxDrawBuffers); + caps.hasDrawBuffersFunc = true; f->glGetIntegerv(GL_MAX_SAMPLES, &caps.maxSamples); caps.maxSamples = qMax(1, caps.maxSamples); } else { + // ES 2.0 / WebGL 1 caps.maxDrawBuffers = 1; - caps.maxSamples = 1; + caps.hasDrawBuffersFunc = false; + // This does not mean MSAA is not supported, just that we cannot query + // the supported sample counts. Assume that 4x is always supported. + caps.maxSamples = 4; } caps.msaaRenderBuffer = f->hasOpenGLExtension(QOpenGLExtensions::FramebufferMultisample) @@ -535,20 +864,27 @@ bool QRhiGles2::create(QRhi::Flags flags) caps.npotTextureFull = f->hasOpenGLFeature(QOpenGLFunctions::NPOTTextures) && f->hasOpenGLFeature(QOpenGLFunctions::NPOTTextureRepeat); - caps.gles = actualFormat.renderableType() == QSurfaceFormat::OpenGLES; if (caps.gles) caps.fixedIndexPrimitiveRestart = caps.ctxMajor >= 3; // ES 3.0 else caps.fixedIndexPrimitiveRestart = caps.ctxMajor > 4 || (caps.ctxMajor == 4 && caps.ctxMinor >= 3); // 4.3 - if (caps.fixedIndexPrimitiveRestart) + if (caps.fixedIndexPrimitiveRestart) { +#ifdef Q_OS_WASM + // WebGL 2 behaves as if GL_PRIMITIVE_RESTART_FIXED_INDEX was always + // enabled (i.e. matching D3D/Metal), and the value cannot be passed to + // glEnable, so skip the call. +#else f->glEnable(GL_PRIMITIVE_RESTART_FIXED_INDEX); +#endif + } caps.bgraExternalFormat = f->hasOpenGLExtension(QOpenGLExtensions::BGRATextureFormat); caps.bgraInternalFormat = caps.bgraExternalFormat && caps.gles; caps.r8Format = f->hasOpenGLFeature(QOpenGLFunctions::TextureRGFormats); caps.r16Format = f->hasOpenGLExtension(QOpenGLExtensions::Sized16Formats); caps.floatFormats = caps.ctxMajor >= 3; // 3.0 or ES 3.0 + caps.rgb10Formats = caps.ctxMajor >= 3; // 3.0 or ES 3.0 caps.depthTexture = caps.ctxMajor >= 3; // 3.0 or ES 3.0 caps.packedDepthStencil = f->hasOpenGLExtension(QOpenGLExtensions::PackedDepthStencil); #ifdef Q_OS_WASM @@ -556,7 +892,13 @@ bool QRhiGles2::create(QRhi::Flags flags) #else caps.needsDepthStencilCombinedAttach = false; #endif - caps.srgbCapableDefaultFramebuffer = f->hasOpenGLExtension(QOpenGLExtensions::SRGBFrameBuffer); + + // QOpenGLExtensions::SRGBFrameBuffer is not useful here. We need to know if + // controlling the sRGB-on-shader-write state is supported, not that if the + // default framebuffer is sRGB-capable. And there are two different + // extensions for desktop and ES. + caps.srgbWriteControl = ctx->hasExtension("GL_EXT_framebuffer_sRGB") || ctx->hasExtension("GL_EXT_sRGB_write_control"); + caps.coreProfile = actualFormat.profile() == QSurfaceFormat::CoreProfile; if (caps.gles) @@ -631,9 +973,67 @@ bool QRhiGles2::create(QRhi::Flags flags) caps.texture3D = caps.ctxMajor >= 3; // 3.0 + if (caps.gles) + caps.texture1D = false; // ES + else + caps.texture1D = glTexImage1D && (caps.ctxMajor >= 2); // 2.0 + + if (caps.gles) + caps.tessellation = caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2); // ES 3.2 + else + caps.tessellation = caps.ctxMajor >= 4; // 4.0 + + if (caps.gles) + caps.geometryShader = caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2); // ES 3.2 + else + caps.geometryShader = caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2); // 3.2 + + if (caps.ctxMajor >= 3) { // 3.0 or ES 3.0 + GLint maxArraySize = 0; + f->glGetIntegerv(GL_MAX_ARRAY_TEXTURE_LAYERS, &maxArraySize); + caps.maxTextureArraySize = maxArraySize; + } else { + caps.maxTextureArraySize = 0; + } + + // The ES 2.0 spec only has MAX_xxxx_VECTORS. ES 3.0 and up has both + // *VECTORS and *COMPONENTS. OpenGL 2.0-4.0 only has MAX_xxxx_COMPONENTS. + // 4.1 and above has both. What a mess. + if (caps.gles) { + GLint maxVertexUniformVectors = 0; + f->glGetIntegerv(GL_MAX_VERTEX_UNIFORM_VECTORS, &maxVertexUniformVectors); + GLint maxFragmentUniformVectors = 0; + f->glGetIntegerv(GL_MAX_FRAGMENT_UNIFORM_VECTORS, &maxFragmentUniformVectors); + caps.maxUniformVectors = qMin(maxVertexUniformVectors, maxFragmentUniformVectors); + } else { + GLint maxVertexUniformComponents = 0; + f->glGetIntegerv(GL_MAX_VERTEX_UNIFORM_COMPONENTS, &maxVertexUniformComponents); + GLint maxFragmentUniformComponents = 0; + f->glGetIntegerv(GL_MAX_FRAGMENT_UNIFORM_COMPONENTS, &maxFragmentUniformComponents); + caps.maxUniformVectors = qMin(maxVertexUniformComponents, maxFragmentUniformComponents) / 4; + } + + f->glGetIntegerv(GL_MAX_VERTEX_ATTRIBS, &caps.maxVertexInputs); + + if (caps.gles) { + f->glGetIntegerv(GL_MAX_VARYING_VECTORS, &caps.maxVertexOutputs); + } else if (caps.ctxMajor >= 3) { + GLint components = 0; + f->glGetIntegerv(caps.coreProfile ? GL_MAX_VERTEX_OUTPUT_COMPONENTS : GL_MAX_VARYING_COMPONENTS, &components); + caps.maxVertexOutputs = components / 4; + } else { + // OpenGL before 3.0 only has this, and not the same as + // MAX_VARYING_COMPONENTS strictly speaking, but will do. + GLint components = 0; + f->glGetIntegerv(GL_MAX_VARYING_FLOATS, &components); + if (components > 0) + caps.maxVertexOutputs = components / 4; + } + if (!caps.gles) { f->glEnable(GL_VERTEX_PROGRAM_POINT_SIZE); - f->glEnable(GL_POINT_SPRITE); + if (!caps.coreProfile) + f->glEnable(GL_POINT_SPRITE); } // else (with gles) these are always on // Match D3D and others when it comes to seamless cubemap filtering. @@ -642,6 +1042,60 @@ bool QRhiGles2::create(QRhi::Flags flags) if (!caps.gles && (caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2))) f->glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS); + caps.halfAttributes = f->hasOpenGLExtension(QOpenGLExtensions::HalfFloatVertex); + + // We always require GL_OVR_multiview2 for symmetry with other backends. + caps.multiView = f->hasOpenGLExtension(QOpenGLExtensions::MultiView) + && f->hasOpenGLExtension(QOpenGLExtensions::MultiViewExtended); + if (caps.multiView) { + glFramebufferTextureMultiviewOVR = + reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLuint, GLint, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTextureMultiviewOVR"))); + } + + // Only do timestamp queries on OpenGL 3.3+. + caps.timestamps = !caps.gles && (caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 3)); + if (caps.timestamps) { + glQueryCounter = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLuint, GLenum)>( + ctx->getProcAddress(QByteArrayLiteral("glQueryCounter"))); + glGetQueryObjectui64v = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLuint, GLenum, quint64 *)>( + ctx->getProcAddress(QByteArrayLiteral("glGetQueryObjectui64v"))); + if (!glQueryCounter || !glGetQueryObjectui64v) + caps.timestamps = false; + } + + // glObjectLabel is available on OpenGL ES 3.2+ and OpenGL 4.3+ + if (caps.gles) + caps.objectLabel = caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2); + else + caps.objectLabel = caps.ctxMajor > 4 || (caps.ctxMajor == 4 && caps.ctxMinor >= 3); + if (caps.objectLabel) { + glObjectLabel = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLuint, GLsizei, const GLchar *)>( + ctx->getProcAddress(QByteArrayLiteral("glObjectLabel"))); + } + + if (caps.gles) { + // This is the third way to get multisample rendering with GLES. (1. is + // multisample render buffer -> resolve to texture; 2. is multisample + // texture with GLES 3.1; 3. is this, avoiding the explicit multisample + // buffer and should be more efficient with tiled architectures. + // Interesting also because 2. does not seem to work in practice on + // devices such as the Quest 3) + caps.glesMultisampleRenderToTexture = ctx->hasExtension("GL_EXT_multisampled_render_to_texture"); + if (caps.glesMultisampleRenderToTexture) { + glFramebufferTexture2DMultisampleEXT = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLenum, GLuint, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTexture2DMultisampleEXT"))); + } + caps.glesMultiviewMultisampleRenderToTexture = ctx->hasExtension("GL_OVR_multiview_multisampled_render_to_texture"); + if (caps.glesMultiviewMultisampleRenderToTexture) { + glFramebufferTextureMultisampleMultiviewOVR = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLuint, GLint, GLsizei, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTextureMultisampleMultiviewOVR"))); + } + } else { + caps.glesMultisampleRenderToTexture = false; + caps.glesMultiviewMultisampleRenderToTexture = false; + } + nativeHandlesStruct.context = ctx; contextLost = false; @@ -657,6 +1111,11 @@ void QRhiGles2::destroy() ensureContext(); executeDeferredReleases(); + if (ofr.tsQueries[0]) { + f->glDeleteQueries(2, ofr.tsQueries); + ofr.tsQueries[0] = ofr.tsQueries[1] = 0; + } + if (vao) { f->glDeleteVertexArrays(1, &vao); vao = 0; @@ -676,7 +1135,7 @@ void QRhiGles2::destroy() void QRhiGles2::executeDeferredReleases() { - for (int i = releaseQueue.count() - 1; i >= 0; --i) { + for (int i = releaseQueue.size() - 1; i >= 0; --i) { const QRhiGles2::DeferredReleaseEntry &e(releaseQueue[i]); switch (e.type) { case QRhiGles2::DeferredReleaseEntry::Buffer: @@ -694,6 +1153,7 @@ void QRhiGles2::executeDeferredReleases() break; case QRhiGles2::DeferredReleaseEntry::TextureRenderTarget: f->glDeleteFramebuffers(1, &e.textureRenderTarget.framebuffer); + f->glDeleteTextures(1, &e.textureRenderTarget.nonMsaaThrowawayDepthTexture); break; default: Q_UNREACHABLE(); @@ -713,23 +1173,12 @@ QList<int> QRhiGles2::supportedSampleCounts() const return supportedSampleCountList; } -int QRhiGles2::effectiveSampleCount(int sampleCount) const -{ - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - const int s = qBound(1, sampleCount, 64); - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return 1; - } - return s; -} - QRhiSwapChain *QRhiGles2::createSwapChain() { return new QGles2SwapChain(this); } -QRhiBuffer *QRhiGles2::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, int size) +QRhiBuffer *QRhiGles2::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) { return new QGles2Buffer(this, type, usage, size); } @@ -761,58 +1210,6 @@ QMatrix4x4 QRhiGles2::clipSpaceCorrMatrix() const return QMatrix4x4(); // identity } -static inline GLenum toGlCompressedTextureFormat(QRhiTexture::Format format, QRhiTexture::Flags flags) -{ - const bool srgb = flags.testFlag(QRhiTexture::sRGB); - switch (format) { - case QRhiTexture::BC1: - return srgb ? 0x8C4C : 0x83F0; - case QRhiTexture::BC2: - return srgb ? 0x8C4E : 0x83F2; - case QRhiTexture::BC3: - return srgb ? 0x8C4F : 0x83F3; - - case QRhiTexture::ETC2_RGB8: - return srgb ? 0x9275 : 0x9274; - case QRhiTexture::ETC2_RGB8A1: - return srgb ? 0x9277 : 0x9276; - case QRhiTexture::ETC2_RGBA8: - return srgb ? 0x9279 : 0x9278; - - case QRhiTexture::ASTC_4x4: - return srgb ? 0x93D0 : 0x93B0; - case QRhiTexture::ASTC_5x4: - return srgb ? 0x93D1 : 0x93B1; - case QRhiTexture::ASTC_5x5: - return srgb ? 0x93D2 : 0x93B2; - case QRhiTexture::ASTC_6x5: - return srgb ? 0x93D3 : 0x93B3; - case QRhiTexture::ASTC_6x6: - return srgb ? 0x93D4 : 0x93B4; - case QRhiTexture::ASTC_8x5: - return srgb ? 0x93D5 : 0x93B5; - case QRhiTexture::ASTC_8x6: - return srgb ? 0x93D6 : 0x93B6; - case QRhiTexture::ASTC_8x8: - return srgb ? 0x93D7 : 0x93B7; - case QRhiTexture::ASTC_10x5: - return srgb ? 0x93D8 : 0x93B8; - case QRhiTexture::ASTC_10x6: - return srgb ? 0x93D9 : 0x93B9; - case QRhiTexture::ASTC_10x8: - return srgb ? 0x93DA : 0x93BA; - case QRhiTexture::ASTC_10x10: - return srgb ? 0x93DB : 0x93BB; - case QRhiTexture::ASTC_12x10: - return srgb ? 0x93DC : 0x93BC; - case QRhiTexture::ASTC_12x12: - return srgb ? 0x93DD : 0x93BD; - - default: - return 0; // this is reachable, just return an invalid format - } -} - static inline void toGlTextureFormat(QRhiTexture::Format format, const QRhiGles2::Caps &caps, GLenum *glintformat, GLenum *glsizedintformat, GLenum *glformat, GLenum *gltype) @@ -884,6 +1281,12 @@ static inline void toGlTextureFormat(QRhiTexture::Format format, const QRhiGles2 *glformat = GL_RED; *gltype = GL_FLOAT; break; + case QRhiTexture::RGB10A2: + *glintformat = GL_RGB10_A2; + *glsizedintformat = *glintformat; + *glformat = GL_RGBA; + *gltype = GL_UNSIGNED_INT_2_10_10_10_REV; + break; case QRhiTexture::D16: *glintformat = GL_DEPTH_COMPONENT16; *glsizedintformat = *glintformat; @@ -894,13 +1297,13 @@ static inline void toGlTextureFormat(QRhiTexture::Format format, const QRhiGles2 *glintformat = GL_DEPTH_COMPONENT24; *glsizedintformat = *glintformat; *glformat = GL_DEPTH_COMPONENT; - *gltype = GL_UNSIGNED_SHORT; + *gltype = GL_UNSIGNED_INT; break; case QRhiTexture::D24S8: *glintformat = GL_DEPTH24_STENCIL8; *glsizedintformat = *glintformat; *glformat = GL_DEPTH_STENCIL; - *gltype = GL_UNSIGNED_SHORT; + *gltype = GL_UNSIGNED_INT_24_8; break; case QRhiTexture::D32F: *glintformat = GL_DEPTH_COMPONENT32F; @@ -957,6 +1360,9 @@ bool QRhiGles2::isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture case QRhiTexture::R32F: return caps.floatFormats; + case QRhiTexture::RGB10A2: + return caps.rgb10Formats; + default: break; } @@ -974,7 +1380,7 @@ bool QRhiGles2::isFeatureSupported(QRhi::Feature feature) const case QRhi::DebugMarkers: return false; case QRhi::Timestamps: - return false; + return caps.timestamps; case QRhi::Instancing: return caps.instancing; case QRhi::CustomInstanceStepRate: @@ -1027,9 +1433,34 @@ bool QRhiGles2::isFeatureSupported(QRhi::Feature feature) const return caps.texture3D; case QRhi::RenderTo3DTextureSlice: return caps.texture3D; - default: - Q_UNREACHABLE(); + case QRhi::TextureArrays: + return caps.maxTextureArraySize > 0; + case QRhi::Tessellation: + return caps.tessellation; + case QRhi::GeometryShader: + return caps.geometryShader; + case QRhi::TextureArrayRange: + return false; + case QRhi::NonFillPolygonMode: + return !caps.gles; + case QRhi::OneDimensionalTextures: + return caps.texture1D; + case QRhi::OneDimensionalTextureMipmaps: + return caps.texture1D; + case QRhi::HalfAttributes: + return caps.halfAttributes; + case QRhi::RenderToOneDimensionalTexture: + return caps.texture1D; + case QRhi::ThreeDimensionalTextureMipmaps: + return caps.texture3D; + case QRhi::MultiView: + return caps.multiView && caps.maxTextureArraySize > 0; + case QRhi::TextureViewFormat: return false; + case QRhi::ResolveDepthStencil: + return true; + default: + Q_UNREACHABLE_RETURN(false); } } @@ -1058,9 +1489,16 @@ int QRhiGles2::resourceLimit(QRhi::ResourceLimit limit) const return caps.maxThreadGroupsY; case QRhi::MaxThreadGroupZ: return caps.maxThreadGroupsZ; + case QRhi::TextureArraySizeMax: + return 2048; + case QRhi::MaxUniformBufferRange: + return int(qMin<qint64>(INT_MAX, caps.maxUniformVectors * qint64(16))); + case QRhi::MaxVertexInputs: + return caps.maxVertexInputs; + case QRhi::MaxVertexOutputs: + return caps.maxVertexOutputs; default: - Q_UNREACHABLE(); - return 0; + Q_UNREACHABLE_RETURN(0); } } @@ -1074,9 +1512,11 @@ QRhiDriverInfo QRhiGles2::driverInfo() const return driverInfoStruct; } -void QRhiGles2::sendVMemStatsToProfiler() +QRhiStats QRhiGles2::statistics() { - // nothing to do here + QRhiStats result; + result.totalPipelineCreationTime = totalPipelineCreationTime(); + return result; } bool QRhiGles2::makeThreadLocalNativeContextCurrent() @@ -1125,8 +1565,8 @@ QByteArray QRhiGles2::pipelineCacheData() memset(&header, 0, sizeof(header)); header.rhiId = pipelineCacheRhiId(); header.arch = quint32(sizeof(void*)); - header.programBinaryCount = m_pipelineCache.count(); - const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.count())); + header.programBinaryCount = m_pipelineCache.size(); + const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.size())); if (driverStrLen) memcpy(header.driver, driverInfoStruct.deviceName.constData(), driverStrLen); header.driver[driverStrLen] = '\0'; @@ -1176,7 +1616,7 @@ void QRhiGles2::setPipelineCacheData(const QByteArray &data) const size_t headerSize = sizeof(QGles2PipelineCacheDataHeader); if (data.size() < qsizetype(headerSize)) { - qWarning("setPipelineCacheData: Invalid blob size (header incomplete)"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (header incomplete)"); return; } const size_t dataOffset = headerSize; @@ -1185,27 +1625,27 @@ void QRhiGles2::setPipelineCacheData(const QByteArray &data) const quint32 rhiId = pipelineCacheRhiId(); if (header.rhiId != rhiId) { - qWarning("setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", - rhiId, header.rhiId); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", + rhiId, header.rhiId); return; } const quint32 arch = quint32(sizeof(void*)); if (header.arch != arch) { - qWarning("setPipelineCacheData: Architecture does not match (%u, %u)", - arch, header.arch); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Architecture does not match (%u, %u)", + arch, header.arch); return; } if (header.programBinaryCount == 0) return; - const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.count())); + const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.size())); if (strncmp(header.driver, driverInfoStruct.deviceName.constData(), driverStrLen)) { - qWarning("setPipelineCacheData: OpenGL vendor/renderer/version does not match"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: OpenGL vendor/renderer/version does not match"); return; } if (data.size() < qsizetype(dataOffset + header.dataSize)) { - qWarning("setPipelineCacheData: Invalid blob size (data incomplete)"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (data incomplete)"); return; } @@ -1233,7 +1673,7 @@ void QRhiGles2::setPipelineCacheData(const QByteArray &data) m_pipelineCache.insert(key, { format, data }); } - qCDebug(QRHI_LOG_INFO, "Seeded pipeline cache with %d program binaries", int(m_pipelineCache.count())); + qCDebug(QRHI_LOG_INFO, "Seeded pipeline cache with %d program binaries", int(m_pipelineCache.size())); } QRhiRenderBuffer *QRhiGles2::createRenderBuffer(QRhiRenderBuffer::Type type, const QSize &pixelSize, @@ -1244,10 +1684,10 @@ QRhiRenderBuffer *QRhiGles2::createRenderBuffer(QRhiRenderBuffer::Type type, con } QRhiTexture *QRhiGles2::createTexture(QRhiTexture::Format format, - const QSize &pixelSize, int depth, + const QSize &pixelSize, int depth, int arraySize, int sampleCount, QRhiTexture::Flags flags) { - return new QGles2Texture(this, format, pixelSize, depth, sampleCount, flags); + return new QGles2Texture(this, format, pixelSize, depth, arraySize, sampleCount, flags); } QRhiSampler *QRhiGles2::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, @@ -1315,13 +1755,14 @@ void QRhiGles2::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind QGles2ShaderResourceBindings *srbD = QRHI_RES(QGles2ShaderResourceBindings, srb); if (cbD->passNeedsResourceTracking) { QRhiPassResourceTracker &passResTracker(cbD->passResTrackers[cbD->currentPassResTrackerIndex]); - for (int i = 0, ie = srbD->m_bindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->m_bindings.at(i).data(); + for (int i = 0, ie = srbD->m_bindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings.at(i)); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: // no BufUniformRead / AccessUniform because no real uniform buffers are used break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: for (int elem = 0; elem < b->u.stex.count; ++elem) { trackedRegisterTexture(&passResTracker, QRHI_RES(QGles2Texture, b->u.stex.texSamplers[elem].tex), @@ -1592,10 +2033,14 @@ const QRhiNativeHandles *QRhiGles2::nativeHandles(QRhiCommandBuffer *cb) return nullptr; } -static inline void addBoundaryCommand(QGles2CommandBuffer *cbD, QGles2CommandBuffer::Command::Cmd type) +static inline void addBoundaryCommand(QGles2CommandBuffer *cbD, QGles2CommandBuffer::Command::Cmd type, GLuint tsQuery = 0) { QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = type; + if (type == QGles2CommandBuffer::Command::BeginFrame) + cmd.args.beginFrame.timestampQuery = tsQuery; + else if (type == QGles2CommandBuffer::Command::EndFrame) + cmd.args.endFrame.timestampQuery = tsQuery; } void QRhiGles2::beginExternal(QRhiCommandBuffer *cb) @@ -1624,8 +2069,14 @@ void QRhiGles2::beginExternal(QRhiCommandBuffer *cb) cbD->resetCommands(); - if (vao) + if (vao) { f->glBindVertexArray(0); + } else { + f->glBindBuffer(GL_ARRAY_BUFFER, 0); + f->glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, 0); + if (caps.compute) + f->glBindBuffer(GL_SHADER_STORAGE_BUFFER, 0); + } } void QRhiGles2::endExternal(QRhiCommandBuffer *cb) @@ -1649,21 +2100,36 @@ void QRhiGles2::endExternal(QRhiCommandBuffer *cb) enqueueBindFramebuffer(cbD->currentTarget, cbD); } +double QRhiGles2::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QGles2CommandBuffer *cbD = QRHI_RES(QGles2CommandBuffer, cb); + return cbD->lastGpuTime; +} + QRhi::FrameOpResult QRhiGles2::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags) { QGles2SwapChain *swapChainD = QRHI_RES(QGles2SwapChain, swapChain); if (!ensureContext(swapChainD->surface)) return contextLost ? QRhi::FrameOpDeviceLost : QRhi::FrameOpError; - currentSwapChain = swapChainD; + ctx->handle()->beginFrame(); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - QRHI_PROF_F(beginSwapChainFrame(swapChain)); + currentSwapChain = swapChainD; executeDeferredReleases(); swapChainD->cb.resetState(); - addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::BeginFrame); + if (swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]) { + double elapsedSec = 0; + if (swapChainD->timestamps.tryQueryTimestamps(swapChainD->currentTimestampPairIndex, this, &elapsedSec)) + swapChainD->cb.lastGpuTime = elapsedSec; + } + + GLuint tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + GLuint tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + const bool recordTimestamps = tsStart && tsEnd && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + + addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::BeginFrame, recordTimestamps ? tsStart : 0); return QRhi::FrameOpSuccess; } @@ -1673,26 +2139,33 @@ QRhi::FrameOpResult QRhiGles2::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame QGles2SwapChain *swapChainD = QRHI_RES(QGles2SwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); - addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::EndFrame); + GLuint tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + GLuint tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + const bool recordTimestamps = tsStart && tsEnd && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + if (recordTimestamps) { + swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex] = true; + swapChainD->currentTimestampPairIndex = (swapChainD->currentTimestampPairIndex + 1) % QGles2SwapChainTimestamps::TIMESTAMP_PAIRS; + } + + addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::EndFrame, recordTimestamps ? tsEnd : 0); if (!ensureContext(swapChainD->surface)) return contextLost ? QRhi::FrameOpDeviceLost : QRhi::FrameOpError; executeCommandBuffer(&swapChainD->cb); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - // this must be done before the swap - QRHI_PROF_F(endSwapChainFrame(swapChain, swapChainD->frameCount + 1)); - if (swapChainD->surface && !flags.testFlag(QRhi::SkipPresent)) { ctx->swapBuffers(swapChainD->surface); - needsMakeCurrent = true; + needsMakeCurrentDueToSwap = true; } else { f->glFlush(); } swapChainD->frameCount += 1; currentSwapChain = nullptr; + + ctx->handle()->endFrame(); + return QRhi::FrameOpSuccess; } @@ -1706,7 +2179,12 @@ QRhi::FrameOpResult QRhiGles2::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: executeDeferredReleases(); ofr.cbWrapper.resetState(); - addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::BeginFrame); + if (rhiFlags.testFlag(QRhi::EnableTimestamps) && caps.timestamps) { + if (!ofr.tsQueries[0]) + f->glGenQueries(2, ofr.tsQueries); + } + + addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::BeginFrame, ofr.tsQueries[0]); *cb = &ofr.cbWrapper; return QRhi::FrameOpSuccess; @@ -1718,13 +2196,29 @@ QRhi::FrameOpResult QRhiGles2::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_ASSERT(ofr.active); ofr.active = false; - addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::EndFrame); + addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::EndFrame, ofr.tsQueries[1]); if (!ensureContext()) return contextLost ? QRhi::FrameOpDeviceLost : QRhi::FrameOpError; executeCommandBuffer(&ofr.cbWrapper); + // Just as endFrame() does a flush when skipping the swapBuffers(), do it + // here as well. This has the added benefit of playing nice when rendering + // to a texture from a context and then consuming that texture from + // another, sharing context. + f->glFlush(); + + if (ofr.tsQueries[0]) { + quint64 timestamps[2]; + glGetQueryObjectui64v(ofr.tsQueries[1], GL_QUERY_RESULT, ×tamps[1]); + glGetQueryObjectui64v(ofr.tsQueries[0], GL_QUERY_RESULT, ×tamps[0]); + if (timestamps[1] >= timestamps[0]) { + const quint64 nanoseconds = timestamps[1] - timestamps[0]; + ofr.cbWrapper.lastGpuTime = nanoseconds / 1000000000.0; // seconds + } + } + return QRhi::FrameOpSuccess; } @@ -1746,6 +2240,12 @@ QRhi::FrameOpResult QRhiGles2::finish() executeCommandBuffer(¤tSwapChain->cb); currentSwapChain->cb.resetCommands(); } + // Do an actual glFinish(). May seem superfluous, but this is what + // matches most other backends e.g. Vulkan/Metal that do a heavyweight + // wait-for-idle blocking in their finish(). More importantly, this + // allows clients simply call finish() in threaded or shared context + // situations where one explicitly needs to do a glFlush or Finish. + f->glFinish(); } return QRhi::FrameOpSuccess; } @@ -1765,9 +2265,30 @@ static bool textureAccessIsWrite(QGles2Texture::Access access) || access == QGles2Texture::AccessFramebuffer; } +static inline GLbitfield barriersForBuffer() +{ + return GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT + | GL_ELEMENT_ARRAY_BARRIER_BIT + | GL_UNIFORM_BARRIER_BIT + | GL_BUFFER_UPDATE_BARRIER_BIT + | GL_SHADER_STORAGE_BARRIER_BIT; +} + +static inline GLbitfield barriersForTexture() +{ + return GL_TEXTURE_FETCH_BARRIER_BIT + | GL_SHADER_IMAGE_ACCESS_BARRIER_BIT + | GL_PIXEL_BUFFER_BARRIER_BIT + | GL_TEXTURE_UPDATE_BARRIER_BIT + | GL_FRAMEBUFFER_BARRIER_BIT; +} + void QRhiGles2::trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *bufD, QGles2Buffer::Access access) { Q_ASSERT(cbD->recordingPass == QGles2CommandBuffer::NoPass); // this is for resource updates only + if (!bufD->m_usage.testFlag(QRhiBuffer::StorageBuffer)) + return; + const QGles2Buffer::Access prevAccess = bufD->usageState.access; if (access == prevAccess) return; @@ -1779,7 +2300,7 @@ void QRhiGles2::trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *buf // for now. QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::Barrier; - cmd.args.barrier.barriers = GL_ALL_BARRIER_BITS; + cmd.args.barrier.barriers = barriersForBuffer(); } bufD->usageState.access = access; @@ -1788,6 +2309,9 @@ void QRhiGles2::trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *buf void QRhiGles2::trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Texture::Access access) { Q_ASSERT(cbD->recordingPass == QGles2CommandBuffer::NoPass); // this is for resource updates only + if (!texD->m_flags.testFlag(QRhiTexture::UsedWithLoadStore)) + return; + const QGles2Texture::Access prevAccess = texD->usageState.access; if (access == prevAccess) return; @@ -1795,7 +2319,7 @@ void QRhiGles2::trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *tex if (textureAccessIsWrite(prevAccess)) { QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::Barrier; - cmd.args.barrier.barriers = GL_ALL_BARRIER_BITS; + cmd.args.barrier.barriers = barriersForTexture(); } texD->usageState.access = access; @@ -1807,37 +2331,58 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb trackedImageBarrier(cbD, texD, QGles2Texture::AccessUpdate); const bool isCompressed = isCompressedFormat(texD->m_format); const bool isCubeMap = texD->m_flags.testFlag(QRhiTexture::CubeMap); + const bool is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + const bool is1D = texD->m_flags.testFlag(QRhiTexture::OneDimensional); + const bool isArray = texD->m_flags.testFlag(QRhiTexture::TextureArray); const GLenum faceTargetBase = isCubeMap ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : texD->target; + const GLenum effectiveTarget = faceTargetBase + (isCubeMap ? uint(layer) : 0u); const QPoint dp = subresDesc.destinationTopLeft(); const QByteArray rawData = subresDesc.data(); - if (!subresDesc.image().isNull()) { - QImage img = subresDesc.image(); - QSize size = img.size(); + + auto setCmdByNotCompressedData = [&](const void* data, QSize size, quint32 dataStride) + { + quint32 bytesPerLine = 0; + quint32 bytesPerPixel = 0; + textureFormatInfo(texD->m_format, size, &bytesPerLine, nullptr, &bytesPerPixel); + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::SubImage; - if (!subresDesc.sourceSize().isEmpty() || !subresDesc.sourceTopLeft().isNull()) { - const QPoint sp = subresDesc.sourceTopLeft(); - if (!subresDesc.sourceSize().isEmpty()) - size = subresDesc.sourceSize(); - img = img.copy(sp.x(), sp.y(), size.width(), size.height()); - } cmd.args.subImage.target = texD->target; cmd.args.subImage.texture = texD->texture; - cmd.args.subImage.faceTarget = faceTargetBase + uint(layer); + cmd.args.subImage.faceTarget = effectiveTarget; cmd.args.subImage.level = level; cmd.args.subImage.dx = dp.x(); - cmd.args.subImage.dy = dp.y(); - cmd.args.subImage.dz = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) ? layer : 0; + cmd.args.subImage.dy = is1D && isArray ? layer : dp.y(); + cmd.args.subImage.dz = is3D || isArray ? layer : 0; cmd.args.subImage.w = size.width(); cmd.args.subImage.h = size.height(); cmd.args.subImage.glformat = texD->glformat; cmd.args.subImage.gltype = texD->gltype; - cmd.args.subImage.rowStartAlign = 4; - cmd.args.subImage.rowLength = 0; - cmd.args.subImage.data = cbD->retainImage(img); + + if (dataStride == 0) + dataStride = bytesPerLine; + + cmd.args.subImage.rowStartAlign = (dataStride & 3) ? 1 : 4; + cmd.args.subImage.rowLength = bytesPerPixel ? dataStride / bytesPerPixel : 0; + + cmd.args.subImage.data = data; + }; + + if (!subresDesc.image().isNull()) { + QImage img = subresDesc.image(); + QSize size = img.size(); + if (!subresDesc.sourceSize().isEmpty() || !subresDesc.sourceTopLeft().isNull()) { + const QPoint sp = subresDesc.sourceTopLeft(); + if (!subresDesc.sourceSize().isEmpty()) + size = subresDesc.sourceSize(); + img = img.copy(sp.x(), sp.y(), size.width(), size.height()); + } + + setCmdByNotCompressedData(cbD->retainImage(img), size, img.bytesPerLine()); } else if (!rawData.isEmpty() && isCompressed) { - const bool is3D = texD->flags().testFlag(QRhiTexture::ThreeDimensional); - if ((texD->flags().testFlag(QRhiTexture::UsedAsCompressedAtlas) || is3D) + const int depth = qMax(1, texD->m_depth); + const int arraySize = qMax(0, texD->m_arraySize); + if ((texD->flags().testFlag(QRhiTexture::UsedAsCompressedAtlas) || is3D || isArray) && !texD->zeroInitialized) { // Create on first upload since glCompressedTexImage2D cannot take @@ -1848,18 +2393,20 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb quint32 byteSize = 0; compressedFormatInfo(texD->m_format, texD->m_pixelSize, nullptr, &byteSize, nullptr); if (is3D) - byteSize *= texD->m_depth; + byteSize *= depth; + if (isArray) + byteSize *= arraySize; QByteArray zeroBuf(byteSize, 0); QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::CompressedImage; cmd.args.compressedImage.target = texD->target; cmd.args.compressedImage.texture = texD->texture; - cmd.args.compressedImage.faceTarget = faceTargetBase + uint(layer); + cmd.args.compressedImage.faceTarget = effectiveTarget; cmd.args.compressedImage.level = level; cmd.args.compressedImage.glintformat = texD->glintformat; cmd.args.compressedImage.w = texD->m_pixelSize.width(); - cmd.args.compressedImage.h = texD->m_pixelSize.height(); - cmd.args.compressedImage.depth = is3D ? texD->m_depth : 0; + cmd.args.compressedImage.h = is1D && isArray ? arraySize : texD->m_pixelSize.height(); + cmd.args.compressedImage.depth = is3D ? depth : (isArray ? arraySize : 0); cmd.args.compressedImage.size = byteSize; cmd.args.compressedImage.data = cbD->retainData(zeroBuf); texD->zeroInitialized = true; @@ -1872,11 +2419,11 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb cmd.cmd = QGles2CommandBuffer::Command::CompressedSubImage; cmd.args.compressedSubImage.target = texD->target; cmd.args.compressedSubImage.texture = texD->texture; - cmd.args.compressedSubImage.faceTarget = faceTargetBase + uint(layer); + cmd.args.compressedSubImage.faceTarget = effectiveTarget; cmd.args.compressedSubImage.level = level; cmd.args.compressedSubImage.dx = dp.x(); - cmd.args.compressedSubImage.dy = dp.y(); - cmd.args.compressedSubImage.dz = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) ? layer : 0; + cmd.args.compressedSubImage.dy = is1D && isArray ? layer : dp.y(); + cmd.args.compressedSubImage.dz = is3D || isArray ? layer : 0; cmd.args.compressedSubImage.w = size.width(); cmd.args.compressedSubImage.h = size.height(); cmd.args.compressedSubImage.glintformat = texD->glintformat; @@ -1887,43 +2434,20 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb cmd.cmd = QGles2CommandBuffer::Command::CompressedImage; cmd.args.compressedImage.target = texD->target; cmd.args.compressedImage.texture = texD->texture; - cmd.args.compressedImage.faceTarget = faceTargetBase + uint(layer); + cmd.args.compressedImage.faceTarget = effectiveTarget; cmd.args.compressedImage.level = level; cmd.args.compressedImage.glintformat = texD->glintformat; cmd.args.compressedImage.w = size.width(); - cmd.args.compressedImage.h = size.height(); - cmd.args.compressedImage.depth = is3D ? texD->m_depth : 0; + cmd.args.compressedImage.h = is1D && isArray ? arraySize : size.height(); + cmd.args.compressedImage.depth = is3D ? depth : (isArray ? arraySize : 0); cmd.args.compressedImage.size = rawData.size(); cmd.args.compressedImage.data = cbD->retainData(rawData); } } else if (!rawData.isEmpty()) { const QSize size = subresDesc.sourceSize().isEmpty() ? q->sizeForMipLevel(level, texD->m_pixelSize) : subresDesc.sourceSize(); - quint32 bytesPerLine = 0; - quint32 bytesPerPixel = 0; - textureFormatInfo(texD->m_format, size, &bytesPerLine, nullptr, &bytesPerPixel); - QGles2CommandBuffer::Command &cmd(cbD->commands.get()); - cmd.cmd = QGles2CommandBuffer::Command::SubImage; - cmd.args.subImage.target = texD->target; - cmd.args.subImage.texture = texD->texture; - cmd.args.subImage.faceTarget = faceTargetBase + uint(layer); - cmd.args.subImage.level = level; - cmd.args.subImage.dx = dp.x(); - cmd.args.subImage.dy = dp.y(); - cmd.args.subImage.dz = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) ? layer : 0; - cmd.args.subImage.w = size.width(); - cmd.args.subImage.h = size.height(); - cmd.args.subImage.glformat = texD->glformat; - cmd.args.subImage.gltype = texD->gltype; - // Default unpack alignment (row start aligment - // requirement) is 4. QImage guarantees 4 byte aligned - // row starts, but our raw data here does not. - cmd.args.subImage.rowStartAlign = (bytesPerLine & 3) ? 1 : 4; - if (subresDesc.dataStride() && bytesPerPixel) - cmd.args.subImage.rowLength = subresDesc.dataStride() / bytesPerPixel; - else - cmd.args.subImage.rowLength = 0; - cmd.args.subImage.data = cbD->retainData(rawData); + + setCmdByNotCompressedData(cbD->retainData(rawData), size, subresDesc.dataStride()); } else { qWarning("Invalid texture upload for %p layer=%d mip=%d", texD, layer, level); } @@ -1990,9 +2514,9 @@ void QRhiGles2::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate const QRhiResourceUpdateBatchPrivate::TextureOp &u(ud->textureOps[opIdx]); if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Upload) { QGles2Texture *texD = QRHI_RES(QGles2Texture, u.dst); - for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { + for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) enqueueSubresUpload(texD, cbD, layer, level, subresDesc); } } @@ -2019,21 +2543,26 @@ void QRhiGles2::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::CopyTex; + const bool srcHasZ = srcD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || srcD->m_flags.testFlag(QRhiTexture::TextureArray); + const bool dstHasZ = dstD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || dstD->m_flags.testFlag(QRhiTexture::TextureArray); + const bool dstIs1dArray = dstD->m_flags.testFlag(QRhiTexture::OneDimensional) + && dstD->m_flags.testFlag(QRhiTexture::TextureArray); + cmd.args.copyTex.srcTarget = srcD->target; - cmd.args.copyTex.srcFaceTarget = srcFaceTargetBase + uint(u.desc.sourceLayer()); + cmd.args.copyTex.srcFaceTarget = srcFaceTargetBase + (srcHasZ ? 0u : uint(u.desc.sourceLayer())); cmd.args.copyTex.srcTexture = srcD->texture; cmd.args.copyTex.srcLevel = u.desc.sourceLevel(); cmd.args.copyTex.srcX = sp.x(); cmd.args.copyTex.srcY = sp.y(); - cmd.args.copyTex.srcZ = srcD->m_flags.testFlag(QRhiTexture::ThreeDimensional) ? u.desc.sourceLayer() : 0; + cmd.args.copyTex.srcZ = srcHasZ ? u.desc.sourceLayer() : 0; cmd.args.copyTex.dstTarget = dstD->target; - cmd.args.copyTex.dstFaceTarget = dstFaceTargetBase + uint(u.desc.destinationLayer()); + cmd.args.copyTex.dstFaceTarget = dstFaceTargetBase + (dstHasZ ? 0u : uint(u.desc.destinationLayer())); cmd.args.copyTex.dstTexture = dstD->texture; cmd.args.copyTex.dstLevel = u.desc.destinationLevel(); cmd.args.copyTex.dstX = dp.x(); - cmd.args.copyTex.dstY = dp.y(); - cmd.args.copyTex.dstZ = dstD->m_flags.testFlag(QRhiTexture::ThreeDimensional) ? u.desc.destinationLayer() : 0; + cmd.args.copyTex.dstY = dstIs1dArray ? u.desc.destinationLayer() : dp.y(); + cmd.args.copyTex.dstZ = dstHasZ ? u.desc.destinationLayer() : 0; cmd.args.copyTex.w = copySize.width(); cmd.args.copyTex.h = copySize.height(); @@ -2051,7 +2580,9 @@ void QRhiGles2::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate cmd.args.readPixels.w = readImageSize.width(); cmd.args.readPixels.h = readImageSize.height(); cmd.args.readPixels.format = texD->m_format; - if (texD->m_flags.testFlag(QRhiTexture::ThreeDimensional)) { + if (texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) + || texD->m_flags.testFlag(QRhiTexture::TextureArray)) + { cmd.args.readPixels.readTarget = texD->target; cmd.args.readPixels.slice3D = u.rb.layer(); } else { @@ -2089,9 +2620,10 @@ static inline GLenum toGlTopology(QRhiGraphicsPipeline::Topology t) return GL_LINE_STRIP; case QRhiGraphicsPipeline::Points: return GL_POINTS; + case QRhiGraphicsPipeline::Patches: + return GL_PATCHES; default: - Q_UNREACHABLE(); - return GL_TRIANGLES; + Q_UNREACHABLE_RETURN(GL_TRIANGLES); } } @@ -2103,8 +2635,7 @@ static inline GLenum toGlCullMode(QRhiGraphicsPipeline::CullMode c) case QRhiGraphicsPipeline::Back: return GL_BACK; default: - Q_UNREACHABLE(); - return GL_BACK; + Q_UNREACHABLE_RETURN(GL_BACK); } } @@ -2116,8 +2647,7 @@ static inline GLenum toGlFrontFace(QRhiGraphicsPipeline::FrontFace f) case QRhiGraphicsPipeline::CW: return GL_CW; default: - Q_UNREACHABLE(); - return GL_CCW; + Q_UNREACHABLE_RETURN(GL_CCW); } } @@ -2161,8 +2691,7 @@ static inline GLenum toGlBlendFactor(QRhiGraphicsPipeline::BlendFactor f) qWarning("Unsupported blend factor %d", f); return GL_ZERO; default: - Q_UNREACHABLE(); - return GL_ZERO; + Q_UNREACHABLE_RETURN(GL_ZERO); } } @@ -2180,8 +2709,7 @@ static inline GLenum toGlBlendOp(QRhiGraphicsPipeline::BlendOp op) case QRhiGraphicsPipeline::Max: return GL_MAX; default: - Q_UNREACHABLE(); - return GL_FUNC_ADD; + Q_UNREACHABLE_RETURN(GL_FUNC_ADD); } } @@ -2205,8 +2733,7 @@ static inline GLenum toGlCompareOp(QRhiGraphicsPipeline::CompareOp op) case QRhiGraphicsPipeline::Always: return GL_ALWAYS; default: - Q_UNREACHABLE(); - return GL_ALWAYS; + Q_UNREACHABLE_RETURN(GL_ALWAYS); } } @@ -2230,8 +2757,19 @@ static inline GLenum toGlStencilOp(QRhiGraphicsPipeline::StencilOp op) case QRhiGraphicsPipeline::DecrementAndWrap: return GL_DECR_WRAP; default: - Q_UNREACHABLE(); - return GL_KEEP; + Q_UNREACHABLE_RETURN(GL_KEEP); + } +} + +static inline GLenum toGlPolygonMode(QRhiGraphicsPipeline::PolygonMode mode) +{ + switch (mode) { + case QRhiGraphicsPipeline::PolygonMode::Fill: + return GL_FILL; + case QRhiGraphicsPipeline::PolygonMode::Line: + return GL_LINE; + default: + Q_UNREACHABLE_RETURN(GL_FILL); } } @@ -2249,8 +2787,7 @@ static inline GLenum toGlMinFilter(QRhiSampler::Filter f, QRhiSampler::Filter m) else return m == QRhiSampler::Nearest ? GL_LINEAR_MIPMAP_NEAREST : GL_LINEAR_MIPMAP_LINEAR; default: - Q_UNREACHABLE(); - return GL_LINEAR; + Q_UNREACHABLE_RETURN(GL_LINEAR); } } @@ -2262,8 +2799,7 @@ static inline GLenum toGlMagFilter(QRhiSampler::Filter f) case QRhiSampler::Linear: return GL_LINEAR; default: - Q_UNREACHABLE(); - return GL_LINEAR; + Q_UNREACHABLE_RETURN(GL_LINEAR); } } @@ -2277,8 +2813,7 @@ static inline GLenum toGlWrapMode(QRhiSampler::AddressMode m) case QRhiSampler::Mirror: return GL_MIRRORED_REPEAT; default: - Q_UNREACHABLE(); - return GL_CLAMP_TO_EDGE; + Q_UNREACHABLE_RETURN(GL_CLAMP_TO_EDGE); } } @@ -2302,8 +2837,7 @@ static inline GLenum toGlTextureCompareFunc(QRhiSampler::CompareOp op) case QRhiSampler::Always: return GL_ALWAYS; default: - Q_UNREACHABLE(); - return GL_NEVER; + Q_UNREACHABLE_RETURN(GL_NEVER); } } @@ -2432,6 +2966,8 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) const QGles2CommandBuffer::Command &cmd(*it); switch (cmd.cmd) { case QGles2CommandBuffer::Command::BeginFrame: + if (cmd.args.beginFrame.timestampQuery) + glQueryCounter(cmd.args.beginFrame.timestampQuery, GL_TIMESTAMP); if (caps.coreProfile) { if (!vao) f->glGenVertexArrays(1, &vao); @@ -2448,8 +2984,18 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) f->glVertexAttribDivisor(GLuint(i), 0); state.instancedAttributesUsed = false; } +#ifdef Q_OS_WASM + for (int i = 0; i < CommandBufferExecTrackedState::TRACKED_ATTRIB_COUNT; ++i) { + if (state.enabledAttribArrays[i]) { + f->glDisableVertexAttribArray(GLuint(i)); + state.enabledAttribArrays[i] = false; + } + } +#endif if (vao) f->glBindVertexArray(0); + if (cmd.args.endFrame.timestampQuery) + glQueryCounter(cmd.args.endFrame.timestampQuery, GL_TIMESTAMP); break; case QGles2CommandBuffer::Command::ResetFrame: if (vao) @@ -2578,6 +3124,54 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) type = GL_INT; size = 1; break; + case QRhiVertexInputAttribute::Half4: + type = GL_HALF_FLOAT; + size = 4; + break; + case QRhiVertexInputAttribute::Half3: + type = GL_HALF_FLOAT; + size = 3; + break; + case QRhiVertexInputAttribute::Half2: + type = GL_HALF_FLOAT; + size = 2; + break; + case QRhiVertexInputAttribute::Half: + type = GL_HALF_FLOAT; + size = 1; + break; + case QRhiVertexInputAttribute::UShort4: + type = GL_UNSIGNED_SHORT; + size = 4; + break; + case QRhiVertexInputAttribute::UShort3: + type = GL_UNSIGNED_SHORT; + size = 3; + break; + case QRhiVertexInputAttribute::UShort2: + type = GL_UNSIGNED_SHORT; + size = 2; + break; + case QRhiVertexInputAttribute::UShort: + type = GL_UNSIGNED_SHORT; + size = 1; + break; + case QRhiVertexInputAttribute::SShort4: + type = GL_SHORT; + size = 4; + break; + case QRhiVertexInputAttribute::SShort3: + type = GL_SHORT; + size = 3; + break; + case QRhiVertexInputAttribute::SShort2: + type = GL_SHORT; + size = 2; + break; + case QRhiVertexInputAttribute::SShort: + type = GL_SHORT; + size = 1; + break; default: break; } @@ -2604,7 +3198,7 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) f->glEnableVertexAttribArray(GLuint(locationIdx)); } if (inputBinding->classification() == QRhiVertexInputBinding::PerInstance && caps.instancing) { - f->glVertexAttribDivisor(GLuint(locationIdx), GLuint(inputBinding->instanceStepRate())); + f->glVertexAttribDivisor(GLuint(locationIdx), inputBinding->instanceStepRate()); if (Q_LIKELY(locationIdx < CommandBufferExecTrackedState::TRACKED_ATTRIB_COUNT)) state.nonzeroAttribDivisor[locationIdx] = true; else @@ -2701,24 +3295,32 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) cmd.args.bindShaderResources.dynamicOffsetCount); break; case QGles2CommandBuffer::Command::BindFramebuffer: + { + QVarLengthArray<GLenum, 8> bufs; if (cmd.args.bindFramebuffer.fbo) { f->glBindFramebuffer(GL_FRAMEBUFFER, cmd.args.bindFramebuffer.fbo); + const int colorAttCount = cmd.args.bindFramebuffer.colorAttCount; + bufs.append(colorAttCount > 0 ? GL_COLOR_ATTACHMENT0 : GL_NONE); if (caps.maxDrawBuffers > 1) { - const int colorAttCount = cmd.args.bindFramebuffer.colorAttCount; - QVarLengthArray<GLenum, 8> bufs; - for (int i = 0; i < colorAttCount; ++i) + for (int i = 1; i < colorAttCount; ++i) bufs.append(GL_COLOR_ATTACHMENT0 + uint(i)); - f->glDrawBuffers(colorAttCount, bufs.constData()); } } else { f->glBindFramebuffer(GL_FRAMEBUFFER, ctx->defaultFramebufferObject()); + if (cmd.args.bindFramebuffer.stereo && cmd.args.bindFramebuffer.stereoTarget == QRhiSwapChain::RightBuffer) + bufs.append(GL_BACK_RIGHT); + else + bufs.append(caps.gles ? GL_BACK : GL_BACK_LEFT); } - if (caps.srgbCapableDefaultFramebuffer) { + if (caps.hasDrawBuffersFunc) + f->glDrawBuffers(bufs.count(), bufs.constData()); + if (caps.srgbWriteControl) { if (cmd.args.bindFramebuffer.srgb) f->glEnable(GL_FRAMEBUFFER_SRGB); else f->glDisable(GL_FRAMEBUFFER_SRGB); } + } break; case QGles2CommandBuffer::Command::Clear: f->glDisable(GL_SCISSOR_TEST); @@ -2730,8 +3332,10 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) f->glDepthMask(GL_TRUE); f->glClearDepthf(cmd.args.clear.d); } - if (cmd.args.clear.mask & GL_STENCIL_BUFFER_BIT) + if (cmd.args.clear.mask & GL_STENCIL_BUFFER_BIT) { + f->glStencilMask(0xFF); f->glClearStencil(GLint(cmd.args.clear.s)); + } f->glClear(cmd.args.clear.mask); cbD->graphicsPassState.reset(); // altered depth/color write, invalidate in order to avoid confusing the state tracking break; @@ -2742,7 +3346,7 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::GetBufferSubData: { - QRhiBufferReadbackResult *result = cmd.args.getBufferSubData.result; + QRhiReadbackResult *result = cmd.args.getBufferSubData.result; bindVertexIndexBufferWithStateReset(&state, f, cmd.args.getBufferSubData.target, cmd.args.getBufferSubData.buffer); if (caps.gles) { if (caps.properMapBuffer) { @@ -2772,19 +3376,29 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) GLuint fbo; f->glGenFramebuffers(1, &fbo); f->glBindFramebuffer(GL_FRAMEBUFFER, fbo); - if (cmd.args.copyTex.srcTarget == GL_TEXTURE_3D) { + if (cmd.args.copyTex.srcTarget == GL_TEXTURE_3D + || cmd.args.copyTex.srcTarget == GL_TEXTURE_2D_ARRAY + || cmd.args.copyTex.srcTarget == GL_TEXTURE_1D_ARRAY) { f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.copyTex.srcTexture, cmd.args.copyTex.srcLevel, cmd.args.copyTex.srcZ); + } else if (cmd.args.copyTex.srcTarget == GL_TEXTURE_1D) { + glFramebufferTexture1D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.copyTex.srcTarget, cmd.args.copyTex.srcTexture, + cmd.args.copyTex.srcLevel); } else { f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.copyTex.srcFaceTarget, cmd.args.copyTex.srcTexture, cmd.args.copyTex.srcLevel); } f->glBindTexture(cmd.args.copyTex.dstTarget, cmd.args.copyTex.dstTexture); - if (cmd.args.copyTex.dstTarget == GL_TEXTURE_3D) { + if (cmd.args.copyTex.dstTarget == GL_TEXTURE_3D || cmd.args.copyTex.dstTarget == GL_TEXTURE_2D_ARRAY) { f->glCopyTexSubImage3D(cmd.args.copyTex.dstTarget, cmd.args.copyTex.dstLevel, cmd.args.copyTex.dstX, cmd.args.copyTex.dstY, cmd.args.copyTex.dstZ, cmd.args.copyTex.srcX, cmd.args.copyTex.srcY, cmd.args.copyTex.w, cmd.args.copyTex.h); + } else if (cmd.args.copyTex.dstTarget == GL_TEXTURE_1D) { + glCopyTexSubImage1D(cmd.args.copyTex.dstTarget, cmd.args.copyTex.dstLevel, + cmd.args.copyTex.dstX, cmd.args.copyTex.srcX, + cmd.args.copyTex.srcY, cmd.args.copyTex.w); } else { f->glCopyTexSubImage2D(cmd.args.copyTex.dstFaceTarget, cmd.args.copyTex.dstLevel, cmd.args.copyTex.dstX, cmd.args.copyTex.dstY, @@ -2811,6 +3425,9 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) if (cmd.args.readPixels.slice3D >= 0) { f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, tex, mipLevel, cmd.args.readPixels.slice3D); + } else if (cmd.args.readPixels.readTarget == GL_TEXTURE_1D) { + glFramebufferTexture1D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.readPixels.readTarget, tex, mipLevel); } else { f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.readPixels.readTarget, tex, mipLevel); @@ -2852,10 +3469,22 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) result->data.resize(w * h * 8); f->glReadPixels(0, 0, w, h, GL_RGBA, GL_HALF_FLOAT, result->data.data()); break; + case QRhiTexture::R16F: + result->data.resize(w * h * 2); + f->glReadPixels(0, 0, w, h, GL_RED, GL_HALF_FLOAT, result->data.data()); + break; + case QRhiTexture::R32F: + result->data.resize(w * h * 4); + f->glReadPixels(0, 0, w, h, GL_RED, GL_FLOAT, result->data.data()); + break; case QRhiTexture::RGBA32F: result->data.resize(w * h * 16); f->glReadPixels(0, 0, w, h, GL_RGBA, GL_FLOAT, result->data.data()); break; + case QRhiTexture::RGB10A2: + result->data.resize(w * h * 4); + f->glReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, result->data.data()); + break; default: result->data.resize(w * h * 4); f->glReadPixels(0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, result->data.data()); @@ -2880,12 +3509,17 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) f->glPixelStorei(GL_UNPACK_ALIGNMENT, cmd.args.subImage.rowStartAlign); if (cmd.args.subImage.rowLength != 0) f->glPixelStorei(GL_UNPACK_ROW_LENGTH, cmd.args.subImage.rowLength); - if (cmd.args.subImage.target == GL_TEXTURE_3D) { + if (cmd.args.subImage.target == GL_TEXTURE_3D || cmd.args.subImage.target == GL_TEXTURE_2D_ARRAY) { f->glTexSubImage3D(cmd.args.subImage.target, cmd.args.subImage.level, cmd.args.subImage.dx, cmd.args.subImage.dy, cmd.args.subImage.dz, cmd.args.subImage.w, cmd.args.subImage.h, 1, cmd.args.subImage.glformat, cmd.args.subImage.gltype, cmd.args.subImage.data); + } else if (cmd.args.subImage.target == GL_TEXTURE_1D) { + glTexSubImage1D(cmd.args.subImage.target, cmd.args.subImage.level, + cmd.args.subImage.dx, cmd.args.subImage.w, + cmd.args.subImage.glformat, cmd.args.subImage.gltype, + cmd.args.subImage.data); } else { f->glTexSubImage2D(cmd.args.subImage.faceTarget, cmd.args.subImage.level, cmd.args.subImage.dx, cmd.args.subImage.dy, @@ -2900,11 +3534,16 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::CompressedImage: f->glBindTexture(cmd.args.compressedImage.target, cmd.args.compressedImage.texture); - if (cmd.args.compressedImage.target == GL_TEXTURE_3D) { + if (cmd.args.compressedImage.target == GL_TEXTURE_3D || cmd.args.compressedImage.target == GL_TEXTURE_2D_ARRAY) { f->glCompressedTexImage3D(cmd.args.compressedImage.target, cmd.args.compressedImage.level, cmd.args.compressedImage.glintformat, cmd.args.compressedImage.w, cmd.args.compressedImage.h, cmd.args.compressedImage.depth, 0, cmd.args.compressedImage.size, cmd.args.compressedImage.data); + } else if (cmd.args.compressedImage.target == GL_TEXTURE_1D) { + glCompressedTexImage1D( + cmd.args.compressedImage.target, cmd.args.compressedImage.level, + cmd.args.compressedImage.glintformat, cmd.args.compressedImage.w, 0, + cmd.args.compressedImage.size, cmd.args.compressedImage.data); } else { f->glCompressedTexImage2D(cmd.args.compressedImage.faceTarget, cmd.args.compressedImage.level, cmd.args.compressedImage.glintformat, @@ -2914,12 +3553,18 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::CompressedSubImage: f->glBindTexture(cmd.args.compressedSubImage.target, cmd.args.compressedSubImage.texture); - if (cmd.args.compressedSubImage.target == GL_TEXTURE_3D) { + if (cmd.args.compressedSubImage.target == GL_TEXTURE_3D || cmd.args.compressedSubImage.target == GL_TEXTURE_2D_ARRAY) { f->glCompressedTexSubImage3D(cmd.args.compressedSubImage.target, cmd.args.compressedSubImage.level, cmd.args.compressedSubImage.dx, cmd.args.compressedSubImage.dy, cmd.args.compressedSubImage.dz, cmd.args.compressedSubImage.w, cmd.args.compressedSubImage.h, 1, cmd.args.compressedSubImage.glintformat, cmd.args.compressedSubImage.size, cmd.args.compressedSubImage.data); + } else if (cmd.args.compressedImage.target == GL_TEXTURE_1D) { + glCompressedTexSubImage1D( + cmd.args.compressedSubImage.target, cmd.args.compressedSubImage.level, + cmd.args.compressedSubImage.dx, cmd.args.compressedSubImage.w, + cmd.args.compressedSubImage.glintformat, cmd.args.compressedSubImage.size, + cmd.args.compressedSubImage.data); } else { f->glCompressedTexSubImage2D(cmd.args.compressedSubImage.faceTarget, cmd.args.compressedSubImage.level, cmd.args.compressedSubImage.dx, cmd.args.compressedSubImage.dy, @@ -2930,19 +3575,128 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::BlitFromRenderbuffer: { + // Altering the scissor state, so reset the stored state, although + // not strictly required as long as blit is done in endPass() only. + cbD->graphicsPassState.reset(); + f->glDisable(GL_SCISSOR_TEST); + GLuint fbo[2]; + f->glGenFramebuffers(2, fbo); + f->glBindFramebuffer(GL_READ_FRAMEBUFFER, fbo[0]); + const bool ds = cmd.args.blitFromRenderbuffer.isDepthStencil; + if (ds) { + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + } else { + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + } + f->glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo[1]); + if (cmd.args.blitFromRenderbuffer.target == GL_TEXTURE_3D || cmd.args.blitFromRenderbuffer.target == GL_TEXTURE_2D_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + } else { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + } + } else { + if (ds) { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + } else { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + } + } + f->glBlitFramebuffer(0, 0, cmd.args.blitFromRenderbuffer.w, cmd.args.blitFromRenderbuffer.h, + 0, 0, cmd.args.blitFromRenderbuffer.w, cmd.args.blitFromRenderbuffer.h, + ds ? GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT : GL_COLOR_BUFFER_BIT, + GL_NEAREST); // Qt 5 used Nearest when resolving samples, stick to that + f->glBindFramebuffer(GL_FRAMEBUFFER, ctx->defaultFramebufferObject()); + f->glDeleteFramebuffers(2, fbo); + } + break; + case QGles2CommandBuffer::Command::BlitFromTexture: + { + // Altering the scissor state, so reset the stored state, although + // not strictly required as long as blit is done in endPass() only. + cbD->graphicsPassState.reset(); + f->glDisable(GL_SCISSOR_TEST); GLuint fbo[2]; f->glGenFramebuffers(2, fbo); f->glBindFramebuffer(GL_READ_FRAMEBUFFER, fbo[0]); - f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, - GL_RENDERBUFFER, cmd.args.blitFromRb.renderbuffer); + const bool ds = cmd.args.blitFromTexture.isDepthStencil; + if (cmd.args.blitFromTexture.srcTarget == GL_TEXTURE_2D_MULTISAMPLE_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + } else { + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + } + } else { + if (ds) { + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + } else { + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + } + } f->glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo[1]); - - f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromRb.target, - cmd.args.blitFromRb.texture, cmd.args.blitFromRb.dstLevel); - f->glBlitFramebuffer(0, 0, cmd.args.blitFromRb.w, cmd.args.blitFromRb.h, - 0, 0, cmd.args.blitFromRb.w, cmd.args.blitFromRb.h, - GL_COLOR_BUFFER_BIT, - GL_LINEAR); + if (cmd.args.blitFromTexture.dstTarget == GL_TEXTURE_3D || cmd.args.blitFromTexture.dstTarget == GL_TEXTURE_2D_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + } else { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + } + } else { + if (ds) { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + } else { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + } + } + f->glBlitFramebuffer(0, 0, cmd.args.blitFromTexture.w, cmd.args.blitFromTexture.h, + 0, 0, cmd.args.blitFromTexture.w, cmd.args.blitFromTexture.h, + ds ? GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT : GL_COLOR_BUFFER_BIT, + GL_NEAREST); // Qt 5 used Nearest when resolving samples, stick to that f->glBindFramebuffer(GL_FRAMEBUFFER, ctx->defaultFramebufferObject()); f->glDeleteFramebuffers(2, fbo); } @@ -2976,12 +3730,12 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) for (auto it = tracker.cbeginBuffers(), itEnd = tracker.cendBuffers(); it != itEnd; ++it) { QGles2Buffer::Access accessBeforePass = QGles2Buffer::Access(it->stateAtPassBegin.access); if (bufferAccessIsWrite(accessBeforePass)) - barriers |= GL_ALL_BARRIER_BITS; + barriers |= barriersForBuffer(); } for (auto it = tracker.cbeginTextures(), itEnd = tracker.cendTextures(); it != itEnd; ++it) { QGles2Texture::Access accessBeforePass = QGles2Texture::Access(it->stateAtPassBegin.access); if (textureAccessIsWrite(accessBeforePass)) - barriers |= GL_ALL_BARRIER_BITS; + barriers |= barriersForTexture(); } if (barriers) f->glMemoryBarrier(barriers); @@ -2991,6 +3745,13 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) if (caps.compute) f->glMemoryBarrier(cmd.args.barrier.barriers); break; + case QGles2CommandBuffer::Command::InvalidateFramebuffer: + if (caps.gles && caps.ctxMajor >= 3) { + f->glInvalidateFramebuffer(GL_DRAW_FRAMEBUFFER, + cmd.args.invalidateFramebuffer.attCount, + cmd.args.invalidateFramebuffer.att); + } + break; default: break; } @@ -3039,6 +3800,12 @@ void QRhiGles2::executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2Grap f->glFrontFace(frontFace); } + const GLenum polygonMode = toGlPolygonMode(psD->m_polygonMode); + if (glPolygonMode && (forceUpdate || polygonMode != state.polygonMode)) { + state.polygonMode = polygonMode; + glPolygonMode(GL_FRONT_AND_BACK, polygonMode); + } + if (!psD->m_targetBlends.isEmpty()) { // We do not have MRT support here, meaning all targets use the blend // params from the first one. This is technically incorrect, even if @@ -3179,12 +3946,21 @@ void QRhiGles2::executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2Grap } } + if (psD->m_topology == QRhiGraphicsPipeline::Patches) { + const int cpCount = psD->m_patchControlPointCount; + if (forceUpdate || cpCount != state.cpCount) { + state.cpCount = cpCount; + f->glPatchParameteri(GL_PATCH_VERTICES, qMax(1, cpCount)); + } + } + f->glUseProgram(psD->program); } -static inline void qrhi_std140_to_packed(float *dst, int vecSize, int elemCount, const void *src) +template <typename T> +static inline void qrhi_std140_to_packed(T *dst, int vecSize, int elemCount, const void *src) { - const float *p = reinterpret_cast<const float *>(src); + const T *p = reinterpret_cast<const T *>(src); for (int i = 0; i < elemCount; ++i) { for (int j = 0; j < vecSize; ++j) dst[vecSize * i + j] = *p++; @@ -3192,6 +3968,59 @@ static inline void qrhi_std140_to_packed(float *dst, int vecSize, int elemCount, } } +void QRhiGles2::bindCombinedSampler(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Sampler *samplerD, + void *ps, uint psGeneration, int glslLocation, + int *texUnit, bool *activeTexUnitAltered) +{ + const bool samplerStateValid = texD->samplerState == samplerD->d; + const bool cachedStateInRange = *texUnit < 16; + bool updateTextureBinding = true; + if (samplerStateValid && cachedStateInRange) { + // If we already encountered the same texture with + // the same pipeline for this texture unit in the + // current pass, then the shader program already + // has the uniform set. As in a 3D scene one model + // often has more than one associated texture map, + // the savings here can become significant, + // depending on the scene. + if (cbD->textureUnitState[*texUnit].ps == ps + && cbD->textureUnitState[*texUnit].psGeneration == psGeneration + && cbD->textureUnitState[*texUnit].texture == texD->texture) + { + updateTextureBinding = false; + } + } + if (updateTextureBinding) { + f->glActiveTexture(GL_TEXTURE0 + uint(*texUnit)); + *activeTexUnitAltered = true; + f->glBindTexture(texD->target, texD->texture); + f->glUniform1i(glslLocation, *texUnit); + if (cachedStateInRange) { + cbD->textureUnitState[*texUnit].ps = ps; + cbD->textureUnitState[*texUnit].psGeneration = psGeneration; + cbD->textureUnitState[*texUnit].texture = texD->texture; + } + } + ++(*texUnit); + if (!samplerStateValid) { + f->glTexParameteri(texD->target, GL_TEXTURE_MIN_FILTER, GLint(samplerD->d.glminfilter)); + f->glTexParameteri(texD->target, GL_TEXTURE_MAG_FILTER, GLint(samplerD->d.glmagfilter)); + f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_S, GLint(samplerD->d.glwraps)); + f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_T, GLint(samplerD->d.glwrapt)); + if (caps.texture3D) + f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_R, GLint(samplerD->d.glwrapr)); + if (caps.textureCompareMode) { + if (samplerD->d.gltexcomparefunc != GL_NEVER) { + f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_MODE, GL_COMPARE_REF_TO_TEXTURE); + f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_FUNC, GLint(samplerD->d.gltexcomparefunc)); + } else { + f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_MODE, GL_NONE); + } + } + texD->samplerState = samplerD->d; + } +} + void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, QRhiGraphicsPipeline *maybeGraphicsPs, QRhiComputePipeline *maybeComputePs, QRhiShaderResourceBindings *srb, @@ -3200,14 +4029,29 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, QGles2ShaderResourceBindings *srbD = QRHI_RES(QGles2ShaderResourceBindings, srb); int texUnit = 1; // start from unit 1, keep 0 for resource mgmt stuff to avoid clashes bool activeTexUnitAltered = false; - QVarLengthArray<float, 256> packedFloatArray; + union data32_t { + float f; + qint32 i; + }; + QVarLengthArray<data32_t, 256> packedArray; QGles2UniformDescriptionVector &uniforms(maybeGraphicsPs ? QRHI_RES(QGles2GraphicsPipeline, maybeGraphicsPs)->uniforms : QRHI_RES(QGles2ComputePipeline, maybeComputePs)->uniforms); QGles2UniformState *uniformState = maybeGraphicsPs ? QRHI_RES(QGles2GraphicsPipeline, maybeGraphicsPs)->uniformState : QRHI_RES(QGles2ComputePipeline, maybeComputePs)->uniformState; + struct SeparateTexture { + QGles2Texture *texture; + int binding; + int elem; + }; + QVarLengthArray<SeparateTexture, 8> separateTextureBindings; + struct SeparateSampler { + QGles2Sampler *sampler; + int binding; + }; + QVarLengthArray<SeparateSampler, 4> separateSamplerBindings; - for (int i = 0, ie = srbD->m_bindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->m_bindings.at(i).data(); + for (int i = 0, ie = srbD->m_bindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings.at(i)); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: @@ -3221,7 +4065,7 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, } QGles2Buffer *bufD = QRHI_RES(QGles2Buffer, b->u.ubuf.buf); const char *bufView = bufD->data.constData() + viewOffset; - for (const QGles2UniformDescription &uniform : qAsConst(uniforms)) { + for (const QGles2UniformDescription &uniform : std::as_const(uniforms)) { if (uniform.binding == b->binding) { // in a uniform buffer everything is at least 4 byte aligned // so this should not cause unaligned reads @@ -3233,11 +4077,16 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, && uniform.type != QShaderDescription::Vec2 && uniform.type != QShaderDescription::Vec3 && uniform.type != QShaderDescription::Vec4 + && uniform.type != QShaderDescription::Int + && uniform.type != QShaderDescription::Int2 + && uniform.type != QShaderDescription::Int3 + && uniform.type != QShaderDescription::Int4 && uniform.type != QShaderDescription::Mat3 && uniform.type != QShaderDescription::Mat4) { qWarning("Uniform with buffer binding %d, buffer offset %d, type %d is an array, " - "but arrays are only supported for float, vec2, vec3, vec4, mat3 and mat4. " + "but arrays are only supported for float, vec2, vec3, vec4, int, " + "ivec2, ivec3, ivec4, mat3 and mat4. " "Only the first element will be set.", uniform.binding, uniform.offset, uniform.type); } @@ -3267,9 +4116,9 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, } } else { // input is 16 bytes per element as per std140, have to convert to packed - packedFloatArray.resize(elemCount); - qrhi_std140_to_packed(packedFloatArray.data(), 1, elemCount, src); - f->glUniform1fv(uniform.glslLocation, elemCount, packedFloatArray.constData()); + packedArray.resize(elemCount); + qrhi_std140_to_packed(&packedArray.data()->f, 1, elemCount, src); + f->glUniform1fv(uniform.glslLocation, elemCount, &packedArray.constData()->f); } } break; @@ -3293,9 +4142,9 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, f->glUniform2fv(uniform.glslLocation, 1, v); } } else { - packedFloatArray.resize(elemCount * 2); - qrhi_std140_to_packed(packedFloatArray.data(), 2, elemCount, src); - f->glUniform2fv(uniform.glslLocation, elemCount, packedFloatArray.constData()); + packedArray.resize(elemCount * 2); + qrhi_std140_to_packed(&packedArray.data()->f, 2, elemCount, src); + f->glUniform2fv(uniform.glslLocation, elemCount, &packedArray.constData()->f); } } break; @@ -3321,9 +4170,9 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, f->glUniform3fv(uniform.glslLocation, 1, v); } } else { - packedFloatArray.resize(elemCount * 3); - qrhi_std140_to_packed(packedFloatArray.data(), 3, elemCount, src); - f->glUniform3fv(uniform.glslLocation, elemCount, packedFloatArray.constData()); + packedArray.resize(elemCount * 3); + qrhi_std140_to_packed(&packedArray.data()->f, 3, elemCount, src); + f->glUniform3fv(uniform.glslLocation, elemCount, &packedArray.constData()->f); } } break; @@ -3351,7 +4200,7 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, f->glUniform4fv(uniform.glslLocation, 1, v); } } else { - f->glUniform4fv(uniform.glslLocation, qMax(1, uniform.arrayDim), reinterpret_cast<const float *>(src)); + f->glUniform4fv(uniform.glslLocation, elemCount, reinterpret_cast<const float *>(src)); } } break; @@ -3370,9 +4219,9 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, memcpy(mat + 6, srcMat + 8, 3 * sizeof(float)); f->glUniformMatrix3fv(uniform.glslLocation, 1, GL_FALSE, mat); } else { - packedFloatArray.resize(elemCount * 9); - qrhi_std140_to_packed(packedFloatArray.data(), 3, elemCount * 3, src); - f->glUniformMatrix3fv(uniform.glslLocation, elemCount, GL_FALSE, packedFloatArray.constData()); + packedArray.resize(elemCount * 9); + qrhi_std140_to_packed(&packedArray.data()->f, 3, elemCount * 3, src); + f->glUniformMatrix3fv(uniform.glslLocation, elemCount, GL_FALSE, &packedArray.constData()->f); } } break; @@ -3380,16 +4229,43 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, f->glUniformMatrix4fv(uniform.glslLocation, qMax(1, uniform.arrayDim), GL_FALSE, reinterpret_cast<const float *>(src)); break; case QShaderDescription::Int: - f->glUniform1i(uniform.glslLocation, *reinterpret_cast<const qint32 *>(src)); + { + const int elemCount = uniform.arrayDim; + if (elemCount < 1) { + f->glUniform1i(uniform.glslLocation, *reinterpret_cast<const qint32 *>(src)); + } else { + packedArray.resize(elemCount); + qrhi_std140_to_packed(&packedArray.data()->i, 1, elemCount, src); + f->glUniform1iv(uniform.glslLocation, elemCount, &packedArray.constData()->i); + } + } break; case QShaderDescription::Int2: - f->glUniform2iv(uniform.glslLocation, 1, reinterpret_cast<const qint32 *>(src)); + { + const int elemCount = uniform.arrayDim; + if (elemCount < 1) { + f->glUniform2iv(uniform.glslLocation, 1, reinterpret_cast<const qint32 *>(src)); + } else { + packedArray.resize(elemCount * 2); + qrhi_std140_to_packed(&packedArray.data()->i, 2, elemCount, src); + f->glUniform2iv(uniform.glslLocation, elemCount, &packedArray.constData()->i); + } + } break; case QShaderDescription::Int3: - f->glUniform3iv(uniform.glslLocation, 1, reinterpret_cast<const qint32 *>(src)); + { + const int elemCount = uniform.arrayDim; + if (elemCount < 1) { + f->glUniform3iv(uniform.glslLocation, 1, reinterpret_cast<const qint32 *>(src)); + } else { + packedArray.resize(elemCount * 3); + qrhi_std140_to_packed(&packedArray.data()->i, 3, elemCount, src); + f->glUniform3iv(uniform.glslLocation, elemCount, &packedArray.constData()->i); + } + } break; case QShaderDescription::Int4: - f->glUniform4iv(uniform.glslLocation, 1, reinterpret_cast<const qint32 *>(src)); + f->glUniform4iv(uniform.glslLocation, qMax(1, uniform.arrayDim), reinterpret_cast<const qint32 *>(src)); break; case QShaderDescription::Uint: f->glUniform1ui(uniform.glslLocation, *reinterpret_cast<const quint32 *>(src)); @@ -3441,64 +4317,33 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, QGles2Texture *texD = QRHI_RES(QGles2Texture, b->u.stex.texSamplers[elem].tex); QGles2Sampler *samplerD = QRHI_RES(QGles2Sampler, b->u.stex.texSamplers[elem].sampler); for (const QGles2SamplerDescription &shaderSampler : samplers) { - if (shaderSampler.binding == b->binding) { - const bool samplerStateValid = texD->samplerState == samplerD->d; - const bool cachedStateInRange = texUnit < 16; - bool updateTextureBinding = true; - if (samplerStateValid && cachedStateInRange) { - // If we already encountered the same texture with - // the same pipeline for this texture unit in the - // current pass, then the shader program already - // has the uniform set. As in a 3D scene one model - // often has more than one associated texture map, - // the savings here can become significant, - // depending on the scene. - if (cbD->textureUnitState[texUnit].ps == ps - && cbD->textureUnitState[texUnit].psGeneration == psGeneration - && cbD->textureUnitState[texUnit].texture == texD->texture) - { - updateTextureBinding = false; - } - } - if (updateTextureBinding) { - f->glActiveTexture(GL_TEXTURE0 + uint(texUnit)); - activeTexUnitAltered = true; - f->glBindTexture(texD->target, texD->texture); - f->glUniform1i(shaderSampler.glslLocation + elem, texUnit); - if (cachedStateInRange) { - cbD->textureUnitState[texUnit].ps = ps; - cbD->textureUnitState[texUnit].psGeneration = psGeneration; - cbD->textureUnitState[texUnit].texture = texD->texture; - } - } - ++texUnit; - if (!samplerStateValid) { - f->glTexParameteri(texD->target, GL_TEXTURE_MIN_FILTER, GLint(samplerD->d.glminfilter)); - f->glTexParameteri(texD->target, GL_TEXTURE_MAG_FILTER, GLint(samplerD->d.glmagfilter)); - f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_S, GLint(samplerD->d.glwraps)); - f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_T, GLint(samplerD->d.glwrapt)); - if (caps.texture3D) - f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_R, GLint(samplerD->d.glwrapr)); - if (caps.textureCompareMode) { - if (samplerD->d.gltexcomparefunc != GL_NEVER) { - f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_MODE, GL_COMPARE_REF_TO_TEXTURE); - f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_FUNC, GLint(samplerD->d.gltexcomparefunc)); - } else { - f->glTexParameteri(texD->target, GL_TEXTURE_COMPARE_MODE, GL_NONE); - } - } - texD->samplerState = samplerD->d; - } + if (shaderSampler.combinedBinding == b->binding) { + const int loc = shaderSampler.glslLocation + elem; + bindCombinedSampler(cbD, texD, samplerD, ps, psGeneration, loc, &texUnit, &activeTexUnitAltered); + break; } } } } break; + case QRhiShaderResourceBinding::Texture: + for (int elem = 0; elem < b->u.stex.count; ++elem) { + QGles2Texture *texD = QRHI_RES(QGles2Texture, b->u.stex.texSamplers[elem].tex); + separateTextureBindings.append({ texD, b->binding, elem }); + } + break; + case QRhiShaderResourceBinding::Sampler: + { + QGles2Sampler *samplerD = QRHI_RES(QGles2Sampler, b->u.stex.texSamplers[0].sampler); + separateSamplerBindings.append({ samplerD, b->binding }); + } + break; case QRhiShaderResourceBinding::ImageLoad: case QRhiShaderResourceBinding::ImageStore: case QRhiShaderResourceBinding::ImageLoadStore: { QGles2Texture *texD = QRHI_RES(QGles2Texture, b->u.simage.tex); + Q_ASSERT(texD->m_flags.testFlag(QRhiTexture::UsedWithLoadStore)); const bool layered = texD->m_flags.testFlag(QRhiTexture::CubeMap); GLenum access = GL_READ_WRITE; if (b->type == QRhiShaderResourceBinding::ImageLoad) @@ -3515,6 +4360,7 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, case QRhiShaderResourceBinding::BufferLoadStore: { QGles2Buffer *bufD = QRHI_RES(QGles2Buffer, b->u.sbuf.buf); + Q_ASSERT(bufD->m_usage.testFlag(QRhiBuffer::StorageBuffer)); if (b->u.sbuf.offset == 0 && b->u.sbuf.maybeSize == 0) f->glBindBufferBase(GL_SHADER_STORAGE_BUFFER, GLuint(b->binding), bufD->buffer); else @@ -3528,6 +4374,35 @@ void QRhiGles2::bindShaderResources(QGles2CommandBuffer *cbD, } } + if (!separateTextureBindings.isEmpty() || !separateSamplerBindings.isEmpty()) { + const QGles2SamplerDescriptionVector &samplers(maybeGraphicsPs ? QRHI_RES(QGles2GraphicsPipeline, maybeGraphicsPs)->samplers + : QRHI_RES(QGles2ComputePipeline, maybeComputePs)->samplers); + void *ps; + uint psGeneration; + if (maybeGraphicsPs) { + ps = maybeGraphicsPs; + psGeneration = QRHI_RES(QGles2GraphicsPipeline, maybeGraphicsPs)->generation; + } else { + ps = maybeComputePs; + psGeneration = QRHI_RES(QGles2ComputePipeline, maybeComputePs)->generation; + } + for (const QGles2SamplerDescription &shaderSampler : samplers) { + if (shaderSampler.combinedBinding >= 0) + continue; + for (const SeparateSampler &sepSampler : separateSamplerBindings) { + if (sepSampler.binding != shaderSampler.sbinding) + continue; + for (const SeparateTexture &sepTex : separateTextureBindings) { + if (sepTex.binding != shaderSampler.tbinding) + continue; + const int loc = shaderSampler.glslLocation + sepTex.elem; + bindCombinedSampler(cbD, sepTex.texture, sepSampler.sampler, ps, psGeneration, + loc, &texUnit, &activeTexUnitAltered); + } + } + } + } + if (activeTexUnitAltered) f->glActiveTexture(GL_TEXTURE0); } @@ -3548,15 +4423,21 @@ QGles2RenderTargetData *QRhiGles2::enqueueBindFramebuffer(QRhiRenderTarget *rt, QGles2CommandBuffer::Command &fbCmd(cbD->commands.get()); fbCmd.cmd = QGles2CommandBuffer::Command::BindFramebuffer; + static const bool doClearBuffers = qEnvironmentVariableIntValue("QT_GL_NO_CLEAR_BUFFERS") == 0; + static const bool doClearColorBuffer = qEnvironmentVariableIntValue("QT_GL_NO_CLEAR_COLOR_BUFFER") == 0; + switch (rt->resourceType()) { - case QRhiResource::RenderTarget: - rtD = &QRHI_RES(QGles2ReferenceRenderTarget, rt)->d; + case QRhiResource::SwapChainRenderTarget: + rtD = &QRHI_RES(QGles2SwapChainRenderTarget, rt)->d; if (wantsColorClear) - *wantsColorClear = true; + *wantsColorClear = doClearBuffers && doClearColorBuffer; if (wantsDsClear) - *wantsDsClear = true; + *wantsDsClear = doClearBuffers; fbCmd.args.bindFramebuffer.fbo = 0; fbCmd.args.bindFramebuffer.colorAttCount = 1; + fbCmd.args.bindFramebuffer.stereo = rtD->stereoTarget.has_value(); + if (fbCmd.args.bindFramebuffer.stereo) + fbCmd.args.bindFramebuffer.stereoTarget = rtD->stereoTarget.value(); break; case QRhiResource::TextureRenderTarget: { @@ -3568,6 +4449,7 @@ QGles2RenderTargetData *QRhiGles2::enqueueBindFramebuffer(QRhiRenderTarget *rt, *wantsDsClear = !rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents); fbCmd.args.bindFramebuffer.fbo = rtTex->framebuffer; fbCmd.args.bindFramebuffer.colorAttCount = rtD->colorAttCount; + fbCmd.args.bindFramebuffer.stereo = false; for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); it != itEnd; ++it) @@ -3608,7 +4490,7 @@ QGles2RenderTargetData *QRhiGles2::enqueueBindFramebuffer(QRhiRenderTarget *rt, void QRhiGles2::enqueueBarriersForPass(QGles2CommandBuffer *cbD) { cbD->passResTrackers.append(QRhiPassResourceTracker()); - cbD->currentPassResTrackerIndex = cbD->passResTrackers.count() - 1; + cbD->currentPassResTrackerIndex = cbD->passResTrackers.size() - 1; QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::BarriersForPass; cmd.args.barriersForPass.trackerIndex = cbD->currentPassResTrackerIndex; @@ -3631,6 +4513,12 @@ void QRhiGles2::beginPass(QRhiCommandBuffer *cb, // glMemoryBarrier() calls based on that tracker when submitted. enqueueBarriersForPass(cbD); + if (rt->resourceType() == QRhiRenderTarget::TextureRenderTarget) { + QGles2TextureRenderTarget *rtTex = QRHI_RES(QGles2TextureRenderTarget, rt); + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QGles2Texture, QGles2RenderBuffer>(rtTex->description(), rtTex->d.currentResIdList)) + rtTex->create(); + } + bool wantsColorClear, wantsDsClear; QGles2RenderTargetData *rtD = enqueueBindFramebuffer(rt, cbD, &wantsColorClear, &wantsDsClear); @@ -3662,28 +4550,129 @@ void QRhiGles2::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resource if (cbD->currentTarget->resourceType() == QRhiResource::TextureRenderTarget) { QGles2TextureRenderTarget *rtTex = QRHI_RES(QGles2TextureRenderTarget, cbD->currentTarget); - if (rtTex->m_desc.cbeginColorAttachments() != rtTex->m_desc.cendColorAttachments()) { - // handle only 1 color attachment and only (msaa) renderbuffer - const QRhiColorAttachment &colorAtt(*rtTex->m_desc.cbeginColorAttachments()); - if (colorAtt.resolveTexture()) { - Q_ASSERT(colorAtt.renderBuffer()); + for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); + it != itEnd; ++it) + { + const QRhiColorAttachment &colorAtt(*it); + if (!colorAtt.resolveTexture()) + continue; + + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + const QSize size = resolveTexD->pixelSize(); + if (colorAtt.renderBuffer()) { QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, colorAtt.renderBuffer()); - const QSize size = colorAtt.resolveTexture()->pixelSize(); if (rbD->pixelSize() != size) { qWarning("Resolve source (%dx%d) and target (%dx%d) size does not match", rbD->pixelSize().width(), rbD->pixelSize().height(), size.width(), size.height()); } QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::BlitFromRenderbuffer; - cmd.args.blitFromRb.renderbuffer = rbD->renderbuffer; - cmd.args.blitFromRb.w = size.width(); - cmd.args.blitFromRb.h = size.height(); - QGles2Texture *colorTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); - const GLenum faceTargetBase = colorTexD->m_flags.testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X - : colorTexD->target; - cmd.args.blitFromRb.target = faceTargetBase + uint(colorAtt.resolveLayer()); - cmd.args.blitFromRb.texture = colorTexD->texture; - cmd.args.blitFromRb.dstLevel = colorAtt.resolveLevel(); + cmd.args.blitFromRenderbuffer.renderbuffer = rbD->renderbuffer; + cmd.args.blitFromRenderbuffer.w = size.width(); + cmd.args.blitFromRenderbuffer.h = size.height(); + if (resolveTexD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromRenderbuffer.target = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(colorAtt.resolveLayer()); + else + cmd.args.blitFromRenderbuffer.target = resolveTexD->target; + cmd.args.blitFromRenderbuffer.dstTexture = resolveTexD->texture; + cmd.args.blitFromRenderbuffer.dstLevel = colorAtt.resolveLevel(); + const bool hasZ = resolveTexD->m_flags.testFlag(QRhiTexture::ThreeDimensional) + || resolveTexD->m_flags.testFlag(QRhiTexture::TextureArray); + cmd.args.blitFromRenderbuffer.dstLayer = hasZ ? colorAtt.resolveLayer() : 0; + cmd.args.blitFromRenderbuffer.isDepthStencil = false; + } else if (caps.glesMultisampleRenderToTexture) { + // Nothing to do, resolving into colorAtt.resolveTexture() is automatic, + // colorAtt.texture() is in fact not used for anything. + } else { + Q_ASSERT(colorAtt.texture()); + QGles2Texture *texD = QRHI_RES(QGles2Texture, colorAtt.texture()); + if (texD->pixelSize() != size) { + qWarning("Resolve source (%dx%d) and target (%dx%d) size does not match", + texD->pixelSize().width(), texD->pixelSize().height(), size.width(), size.height()); + } + const int resolveCount = colorAtt.multiViewCount() >= 2 ? colorAtt.multiViewCount() : 1; + for (int resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + const int srcLayer = colorAtt.layer() + resolveIdx; + const int dstLayer = colorAtt.resolveLayer() + resolveIdx; + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromTexture; + if (texD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromTexture.srcTarget = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(srcLayer); + else + cmd.args.blitFromTexture.srcTarget = texD->target; + cmd.args.blitFromTexture.srcTexture = texD->texture; + cmd.args.blitFromTexture.srcLevel = colorAtt.level(); + cmd.args.blitFromTexture.srcLayer = 0; + if (texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || texD->m_flags.testFlag(QRhiTexture::TextureArray)) + cmd.args.blitFromTexture.srcLayer = srcLayer; + cmd.args.blitFromTexture.w = size.width(); + cmd.args.blitFromTexture.h = size.height(); + if (resolveTexD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromTexture.dstTarget = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(dstLayer); + else + cmd.args.blitFromTexture.dstTarget = resolveTexD->target; + cmd.args.blitFromTexture.dstTexture = resolveTexD->texture; + cmd.args.blitFromTexture.dstLevel = colorAtt.resolveLevel(); + cmd.args.blitFromTexture.dstLayer = 0; + if (resolveTexD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || resolveTexD->m_flags.testFlag(QRhiTexture::TextureArray)) + cmd.args.blitFromTexture.dstLayer = dstLayer; + cmd.args.blitFromTexture.isDepthStencil = false; + } + } + } + + if (rtTex->m_desc.depthResolveTexture()) { + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, rtTex->m_desc.depthResolveTexture()); + const QSize size = depthResolveTexD->pixelSize(); + if (rtTex->m_desc.depthStencilBuffer()) { + QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, rtTex->m_desc.depthStencilBuffer()); + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromRenderbuffer; + cmd.args.blitFromRenderbuffer.renderbuffer = rbD->renderbuffer; + cmd.args.blitFromRenderbuffer.w = size.width(); + cmd.args.blitFromRenderbuffer.h = size.height(); + cmd.args.blitFromRenderbuffer.target = depthResolveTexD->target; + cmd.args.blitFromRenderbuffer.dstTexture = depthResolveTexD->texture; + cmd.args.blitFromRenderbuffer.dstLevel = 0; + cmd.args.blitFromRenderbuffer.dstLayer = 0; + cmd.args.blitFromRenderbuffer.isDepthStencil = true; + } else if (caps.glesMultisampleRenderToTexture) { + // Nothing to do, resolving into depthResolveTexture() is automatic. + } else { + QGles2Texture *depthTexD = QRHI_RES(QGles2Texture, rtTex->m_desc.depthTexture()); + const int resolveCount = depthTexD->arraySize() >= 2 ? depthTexD->arraySize() : 1; + for (int resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromTexture; + cmd.args.blitFromTexture.srcTarget = depthTexD->target; + cmd.args.blitFromTexture.srcTexture = depthTexD->texture; + cmd.args.blitFromTexture.srcLevel = 0; + cmd.args.blitFromTexture.srcLayer = resolveIdx; + cmd.args.blitFromTexture.w = size.width(); + cmd.args.blitFromTexture.h = size.height(); + cmd.args.blitFromTexture.dstTarget = depthResolveTexD->target; + cmd.args.blitFromTexture.dstTexture = depthResolveTexD->texture; + cmd.args.blitFromTexture.dstLevel = 0; + cmd.args.blitFromTexture.dstLayer = resolveIdx; + cmd.args.blitFromTexture.isDepthStencil = true; + } + } + } + + const bool mayDiscardDepthStencil = + (rtTex->m_desc.depthStencilBuffer() + || (rtTex->m_desc.depthTexture() && rtTex->m_flags.testFlag(QRhiTextureRenderTarget::DoNotStoreDepthStencilContents))) + && !rtTex->m_desc.depthResolveTexture(); + if (mayDiscardDepthStencil) { + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::InvalidateFramebuffer; + if (caps.needsDepthStencilCombinedAttach) { + cmd.args.invalidateFramebuffer.attCount = 1; + cmd.args.invalidateFramebuffer.att[0] = GL_DEPTH_STENCIL_ATTACHMENT; + } else { + cmd.args.invalidateFramebuffer.attCount = 2; + cmd.args.invalidateFramebuffer.att[0] = GL_DEPTH_ATTACHMENT; + cmd.args.invalidateFramebuffer.att[1] = GL_STENCIL_ATTACHMENT; } } } @@ -3776,9 +4765,9 @@ void QRhiGles2::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) accessAndIsNewFlag = { 0, false }; QGles2ShaderResourceBindings *srbD = QRHI_RES(QGles2ShaderResourceBindings, cbD->currentComputeSrb); - const int bindingCount = srbD->m_bindings.count(); + const int bindingCount = srbD->m_bindings.size(); for (int i = 0; i < bindingCount; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->m_bindings.at(i).data(); + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings.at(i)); switch (b->type) { case QRhiShaderResourceBinding::ImageLoad: case QRhiShaderResourceBinding::ImageStore: @@ -3842,17 +4831,22 @@ static inline GLenum toGlShaderType(QRhiShaderStage::Type type) switch (type) { case QRhiShaderStage::Vertex: return GL_VERTEX_SHADER; + case QRhiShaderStage::TessellationControl: + return GL_TESS_CONTROL_SHADER; + case QRhiShaderStage::TessellationEvaluation: + return GL_TESS_EVALUATION_SHADER; + case QRhiShaderStage::Geometry: + return GL_GEOMETRY_SHADER; case QRhiShaderStage::Fragment: return GL_FRAGMENT_SHADER; case QRhiShaderStage::Compute: return GL_COMPUTE_SHADER; default: - Q_UNREACHABLE(); - return GL_VERTEX_SHADER; + Q_UNREACHABLE_RETURN(GL_VERTEX_SHADER); } } -QByteArray QRhiGles2::shaderSource(const QRhiShaderStage &shaderStage, int *glslVersion) +QByteArray QRhiGles2::shaderSource(const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion) { const QShader bakedShader = shaderStage.shader(); QList<int> versionsToTry; @@ -3871,8 +4865,8 @@ QByteArray QRhiGles2::shaderSource(const QRhiShaderStage &shaderStage, int *glsl QShaderVersion ver(v, QShaderVersion::GlslEs); source = bakedShader.shader({ QShader::GlslShader, ver, shaderStage.shaderVariant() }).shader(); if (!source.isEmpty()) { - if (glslVersion) - *glslVersion = v; + if (shaderVersion) + *shaderVersion = ver; break; } } @@ -3905,8 +4899,8 @@ QByteArray QRhiGles2::shaderSource(const QRhiShaderStage &shaderStage, int *glsl for (int v : versionsToTry) { source = bakedShader.shader({ QShader::GlslShader, v, shaderStage.shaderVariant() }).shader(); if (!source.isEmpty()) { - if (glslVersion) - *glslVersion = v; + if (shaderVersion) + *shaderVersion = v; break; } } @@ -3918,9 +4912,9 @@ QByteArray QRhiGles2::shaderSource(const QRhiShaderStage &shaderStage, int *glsl return source; } -bool QRhiGles2::compileShader(GLuint program, const QRhiShaderStage &shaderStage, int *glslVersion) +bool QRhiGles2::compileShader(GLuint program, const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion) { - const QByteArray source = shaderSource(shaderStage, glslVersion); + const QByteArray source = shaderSource(shaderStage, shaderVersion); if (source.isEmpty()) return false; @@ -3931,7 +4925,7 @@ bool QRhiGles2::compileShader(GLuint program, const QRhiShaderStage &shaderStage } else { shader = f->glCreateShader(toGlShaderType(shaderStage.type())); const char *srcStr = source.constData(); - const GLint srcLength = source.count(); + const GLint srcLength = source.size(); f->glShaderSource(shader, 1, &srcStr, &srcLength); f->glCompileShader(shader); GLint compiled = 0; @@ -3948,7 +4942,7 @@ bool QRhiGles2::compileShader(GLuint program, const QRhiShaderStage &shaderStage qWarning("Failed to compile shader: %s\nSource was:\n%s", log.constData(), source.constData()); return false; } - if (m_shaderCache.count() >= MAX_SHADER_CACHE_ENTRIES) { + if (m_shaderCache.size() >= MAX_SHADER_CACHE_ENTRIES) { // Use the simplest strategy: too many cached shaders -> drop them all. for (uint shader : m_shaderCache) f->glDeleteShader(shader); // does not actually get released yet when attached to a not-yet-released program @@ -3987,7 +4981,7 @@ void QRhiGles2::registerUniformIfActive(const QShaderDescription::BlockVariable int binding, int baseOffset, GLuint program, - QSet<int> *activeUniformLocations, + QDuplicateTracker<int, 256> *activeUniformLocations, QGles2UniformDescriptionVector *dst) { if (var.type == QShaderDescription::Struct) { @@ -4004,9 +4998,8 @@ void QRhiGles2::registerUniformIfActive(const QShaderDescription::BlockVariable // that is not the case, it won't break anything, but we'll generate // unnecessary glUniform* calls then. uniform.glslLocation = f->glGetUniformLocation(program, name.constData()); - if (uniform.glslLocation >= 0 && !activeUniformLocations->contains(uniform.glslLocation)) { - activeUniformLocations->insert(uniform.glslLocation); - if (var.arrayDims.count() > 1) { + if (uniform.glslLocation >= 0 && !activeUniformLocations->hasSeen(uniform.glslLocation)) { + if (var.arrayDims.size() > 1) { qWarning("Array '%s' has more than one dimension. This is not supported.", var.name.constData()); return; @@ -4021,7 +5014,7 @@ void QRhiGles2::registerUniformIfActive(const QShaderDescription::BlockVariable void QRhiGles2::gatherUniforms(GLuint program, const QShaderDescription::UniformBlock &ub, - QSet<int> *activeUniformLocations, + QDuplicateTracker<int, 256> *activeUniformLocations, QGles2UniformDescriptionVector *dst) { QByteArray prefix = ub.structName + '.'; @@ -4035,7 +5028,7 @@ void QRhiGles2::gatherUniforms(GLuint program, registerUniformIfActive(structMember, structPrefix + ".", ub.binding, baseOffset, program, activeUniformLocations, dst); } else { - if (blockMember.arrayDims.count() > 1) { + if (blockMember.arrayDims.size() > 1) { qWarning("Array of struct '%s' has more than one dimension. Only the first " "dimension is used.", blockMember.name.constData()); @@ -4063,7 +5056,23 @@ void QRhiGles2::gatherSamplers(GLuint program, QGles2SamplerDescription sampler; sampler.glslLocation = f->glGetUniformLocation(program, v.name.constData()); if (sampler.glslLocation >= 0) { - sampler.binding = v.binding; + sampler.combinedBinding = v.binding; + sampler.tbinding = -1; + sampler.sbinding = -1; + dst->append(sampler); + } +} + +void QRhiGles2::gatherGeneratedSamplers(GLuint program, + const QShader::SeparateToCombinedImageSamplerMapping &mapping, + QGles2SamplerDescriptionVector *dst) +{ + QGles2SamplerDescription sampler; + sampler.glslLocation = f->glGetUniformLocation(program, mapping.combinedSamplerName.constData()); + if (sampler.glslLocation >= 0) { + sampler.combinedBinding = -1; + sampler.tbinding = mapping.textureBinding; + sampler.sbinding = mapping.samplerBinding; dst->append(sampler); } } @@ -4104,13 +5113,18 @@ static inline QShader::Stage toShaderStage(QRhiShaderStage::Type type) switch (type) { case QRhiShaderStage::Vertex: return QShader::VertexStage; + case QRhiShaderStage::TessellationControl: + return QShader::TessellationControlStage; + case QRhiShaderStage::TessellationEvaluation: + return QShader::TessellationEvaluationStage; + case QRhiShaderStage::Geometry: + return QShader::GeometryStage; case QRhiShaderStage::Fragment: return QShader::FragmentStage; case QRhiShaderStage::Compute: return QShader::ComputeStage; default: - Q_UNREACHABLE(); - return QShader::VertexStage; + Q_UNREACHABLE_RETURN(QShader::VertexStage); } } @@ -4233,7 +5247,7 @@ void QRhiGles2::trySaveToPipelineCache(GLuint program, const QByteArray &cacheKe } } -QGles2Buffer::QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size) +QGles2Buffer::QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) : QRhiBuffer(rhi, type, usage, size) { } @@ -4256,10 +5270,10 @@ void QGles2Buffer::destroy() buffer = 0; QRHI_RES_RHI(QRhiGles2); - rhiD->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QGles2Buffer::create() @@ -4268,7 +5282,6 @@ bool QGles2Buffer::create() destroy(); QRHI_RES_RHI(QRhiGles2); - QRHI_PROF; nonZeroSize = m_size <= 0 ? 256 : m_size; @@ -4278,7 +5291,6 @@ bool QGles2Buffer::create() return false; } data.resize(nonZeroSize); - QRHI_PROF_F(newBuffer(this, uint(nonZeroSize), 0, 1)); return true; } @@ -4295,9 +5307,11 @@ bool QGles2Buffer::create() rhiD->f->glBindBuffer(targetForDataOps, buffer); rhiD->f->glBufferData(targetForDataOps, nonZeroSize, nullptr, m_type == Dynamic ? GL_DYNAMIC_DRAW : GL_STATIC_DRAW); + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_BUFFER, buffer, -1, m_objectName.constData()); + usageState.access = AccessNone; - QRHI_PROF_F(newBuffer(this, uint(nonZeroSize), 1, 0)); rhiD->registerResource(this); return true; } @@ -4366,11 +5380,11 @@ void QGles2RenderBuffer::destroy() stencilRenderbuffer = 0; QRHI_RES_RHI(QRhiGles2); - if (owns) - rhiD->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseRenderBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) { + if (owns) + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QGles2RenderBuffer::create() @@ -4379,14 +5393,11 @@ bool QGles2RenderBuffer::create() destroy(); QRHI_RES_RHI(QRhiGles2); - QRHI_PROF; samples = rhiD->effectiveSampleCount(m_sampleCount); if (m_flags.testFlag(UsedWithSwapChainOnly)) { - if (m_type == DepthStencil) { - QRHI_PROF_F(newRenderBuffer(this, false, true, samples)); + if (m_type == DepthStencil) return true; - } qWarning("RenderBuffer: UsedWithSwapChainOnly is meaningless in combination with Color"); } @@ -4426,7 +5437,6 @@ bool QGles2RenderBuffer::create() rhiD->f->glRenderbufferStorage(GL_RENDERBUFFER, stencilStorage, size.width(), size.height()); } - QRHI_PROF_F(newRenderBuffer(this, false, false, samples)); break; case QRhiRenderBuffer::Color: { @@ -4447,7 +5457,6 @@ bool QGles2RenderBuffer::create() rhiD->f->glRenderbufferStorage(GL_RENDERBUFFER, internalFormat, size.width(), size.height()); } - QRHI_PROF_F(newRenderBuffer(this, false, false, samples)); } break; default: @@ -4455,7 +5464,11 @@ bool QGles2RenderBuffer::create() break; } + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_RENDERBUFFER, renderbuffer, -1, m_objectName.constData()); + owns = true; + generation += 1; rhiD->registerResource(this); return true; } @@ -4479,10 +5492,8 @@ bool QGles2RenderBuffer::createFrom(NativeRenderBuffer src) renderbuffer = src.object; - QRHI_PROF; - QRHI_PROF_F(newRenderBuffer(this, false, false, samples)); - owns = false; + generation += 1; rhiD->registerResource(this); return true; } @@ -4496,8 +5507,8 @@ QRhiTexture::Format QGles2RenderBuffer::backingFormat() const } QGles2Texture::QGles2Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags) - : QRhiTexture(rhi, format, pixelSize, depth, sampleCount, flags) + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags) { } @@ -4521,11 +5532,11 @@ void QGles2Texture::destroy() zeroInitialized = false; QRHI_RES_RHI(QRhiGles2); - if (owns) - rhiD->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseTexture(this)); - rhiD->unregisterResource(this); + if (rhiD) { + if (owns) + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QGles2Texture::prepareCreate(QSize *adjustedSize) @@ -4537,12 +5548,15 @@ bool QGles2Texture::prepareCreate(QSize *adjustedSize) if (!rhiD->ensureContext()) return false; - const QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; - const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(QRhiTexture::TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); const bool hasMipMaps = m_flags.testFlag(MipMapped); const bool isCompressed = rhiD->isCompressedFormat(m_format); + const bool is1D = m_flags.testFlag(OneDimensional); + + const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); if (is3D && !rhiD->caps.texture3D) { qWarning("3D textures are not supported"); @@ -4552,16 +5566,46 @@ bool QGles2Texture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both cube and 3D"); return false; } - m_depth = qMax(1, m_depth); + if (isArray && is3D) { + qWarning("Texture cannot be both array and 3D"); + return false; + } + if (is1D && !rhiD->caps.texture1D) { + qWarning("1D textures are not supported"); + return false; + } + if (is1D && is3D) { + qWarning("Texture cannot be both 1D and 3D"); + return false; + } + if (is1D && isCube) { + qWarning("Texture cannot be both 1D and cube"); + return false; + } + if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } + if (m_arraySize > 0 && !isArray) { + qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); + return false; + } + if (m_arraySize < 1 && isArray) { + qWarning("Texture is an array but array size is %d", m_arraySize); + return false; + } + + target = isCube ? GL_TEXTURE_CUBE_MAP + : m_sampleCount > 1 ? (isArray ? GL_TEXTURE_2D_MULTISAMPLE_ARRAY : GL_TEXTURE_2D_MULTISAMPLE) + : (is3D ? GL_TEXTURE_3D + : (is1D ? (isArray ? GL_TEXTURE_1D_ARRAY : GL_TEXTURE_1D) + : (isArray ? GL_TEXTURE_2D_ARRAY : GL_TEXTURE_2D))); - target = isCube ? GL_TEXTURE_CUBE_MAP - : m_sampleCount > 1 ? GL_TEXTURE_2D_MULTISAMPLE : (is3D ? GL_TEXTURE_3D : GL_TEXTURE_2D); if (m_flags.testFlag(ExternalOES)) target = GL_TEXTURE_EXTERNAL_OES; + else if (m_flags.testFlag(TextureRectangleGL)) + target = GL_TEXTURE_RECTANGLE; mipLevelCount = hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1; gltype = GL_UNSIGNED_BYTE; @@ -4603,21 +5647,35 @@ bool QGles2Texture::create() rhiD->f->glGenTextures(1, &texture); const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(QRhiTexture::TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); const bool hasMipMaps = m_flags.testFlag(MipMapped); const bool isCompressed = rhiD->isCompressedFormat(m_format); + const bool is1D = m_flags.testFlag(OneDimensional); + if (!isCompressed) { rhiD->f->glBindTexture(target, texture); if (!m_flags.testFlag(UsedWithLoadStore)) { - if (is3D) { + if (is1D) { + for (int level = 0; level < mipLevelCount; ++level) { + const QSize mipSize = rhiD->q->sizeForMipLevel(level, size); + if (isArray) + rhiD->f->glTexImage2D(target, level, GLint(glintformat), mipSize.width(), + qMax(0, m_arraySize), 0, glformat, gltype, nullptr); + else + rhiD->glTexImage1D(target, level, GLint(glintformat), mipSize.width(), 0, + glformat, gltype, nullptr); + } + } else if (is3D || isArray) { + const int layerCount = is3D ? qMax(1, m_depth) : qMax(0, m_arraySize); if (hasMipMaps) { for (int level = 0; level != mipLevelCount; ++level) { const QSize mipSize = rhiD->q->sizeForMipLevel(level, size); - rhiD->f->glTexImage3D(target, level, GLint(glintformat), mipSize.width(), mipSize.height(), m_depth, + rhiD->f->glTexImage3D(target, level, GLint(glintformat), mipSize.width(), mipSize.height(), layerCount, 0, glformat, gltype, nullptr); } } else { - rhiD->f->glTexImage3D(target, 0, GLint(glintformat), size.width(), size.height(), m_depth, + rhiD->f->glTexImage3D(target, 0, GLint(glintformat), size.width(), size.height(), layerCount, 0, glformat, gltype, nullptr); } } else if (hasMipMaps || isCube) { @@ -4631,17 +5689,32 @@ bool QGles2Texture::create() } } } else { - rhiD->f->glTexImage2D(target, 0, GLint(glintformat), size.width(), size.height(), - 0, glformat, gltype, nullptr); + // 2D texture. For multisample textures the GLES 3.1 + // glStorage2DMultisample must be used for portability. + if (m_sampleCount > 1 && rhiD->caps.multisampledTexture) { + // internal format must be sized + rhiD->f->glTexStorage2DMultisample(target, m_sampleCount, glsizedintformat, + size.width(), size.height(), GL_TRUE); + } else { + rhiD->f->glTexImage2D(target, 0, GLint(glintformat), size.width(), size.height(), + 0, glformat, gltype, nullptr); + } } } else { // Must be specified with immutable storage functions otherwise // bindImageTexture may fail. Also, the internal format must be a // sized format here. - if (is3D) - rhiD->f->glTexStorage3D(target, mipLevelCount, glsizedintformat, size.width(), size.height(), m_depth); + if (is1D && !isArray) + rhiD->glTexStorage1D(target, mipLevelCount, glsizedintformat, size.width()); + else if (!is1D && (is3D || isArray)) + rhiD->f->glTexStorage3D(target, mipLevelCount, glsizedintformat, size.width(), size.height(), + is3D ? qMax(1, m_depth) : qMax(0, m_arraySize)); + else if (m_sampleCount > 1) + rhiD->f->glTexStorage2DMultisample(target, m_sampleCount, glsizedintformat, + size.width(), size.height(), GL_TRUE); else - rhiD->f->glTexStorage2D(target, mipLevelCount, glsizedintformat, size.width(), size.height()); + rhiD->f->glTexStorage2D(target, mipLevelCount, glsizedintformat, size.width(), + is1D ? qMax(0, m_arraySize) : size.height()); } specified = true; } else { @@ -4651,8 +5724,8 @@ bool QGles2Texture::create() specified = false; } - QRHI_PROF; - QRHI_PROF_F(newTexture(this, true, mipLevelCount, isCube ? 6 : 1, 1)); + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_TEXTURE, texture, -1, m_objectName.constData()); owns = true; @@ -4674,13 +5747,10 @@ bool QGles2Texture::createFrom(QRhiTexture::NativeTexture src) specified = true; zeroInitialized = true; - QRHI_RES_RHI(QRhiGles2); - QRHI_PROF; - QRHI_PROF_F(newTexture(this, false, mipLevelCount, m_flags.testFlag(CubeMap) ? 6 : 1, 1)); - owns = false; generation += 1; + QRHI_RES_RHI(QRhiGles2); rhiD->registerResource(this); return true; } @@ -4703,7 +5773,9 @@ QGles2Sampler::~QGles2Sampler() void QGles2Sampler::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2Sampler::create() @@ -4716,6 +5788,8 @@ bool QGles2Sampler::create() d.gltexcomparefunc = toGlTextureCompareFunc(m_compareOp); generation += 1; + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(this, false); return true; } @@ -4732,7 +5806,9 @@ QGles2RenderPassDescriptor::~QGles2RenderPassDescriptor() void QGles2RenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -4743,36 +5819,44 @@ bool QGles2RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *ot QRhiRenderPassDescriptor *QGles2RenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; } -QGles2ReferenceRenderTarget::QGles2ReferenceRenderTarget(QRhiImplementation *rhi) - : QRhiRenderTarget(rhi), +QVector<quint32> QGles2RenderPassDescriptor::serializedFormat() const +{ + return {}; +} + +QGles2SwapChainRenderTarget::QGles2SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain), d(rhi) { } -QGles2ReferenceRenderTarget::~QGles2ReferenceRenderTarget() +QGles2SwapChainRenderTarget::~QGles2SwapChainRenderTarget() { destroy(); } -void QGles2ReferenceRenderTarget::destroy() +void QGles2SwapChainRenderTarget::destroy() { // nothing to do here } -QSize QGles2ReferenceRenderTarget::pixelSize() const +QSize QGles2SwapChainRenderTarget::pixelSize() const { return d.pixelSize; } -float QGles2ReferenceRenderTarget::devicePixelRatio() const +float QGles2SwapChainRenderTarget::devicePixelRatio() const { return d.dpr; } -int QGles2ReferenceRenderTarget::sampleCount() const +int QGles2SwapChainRenderTarget::sampleCount() const { return d.sampleCount; } @@ -4799,18 +5883,24 @@ void QGles2TextureRenderTarget::destroy() e.type = QRhiGles2::DeferredReleaseEntry::TextureRenderTarget; e.textureRenderTarget.framebuffer = framebuffer; + e.textureRenderTarget.nonMsaaThrowawayDepthTexture = nonMsaaThrowawayDepthTexture; framebuffer = 0; + nonMsaaThrowawayDepthTexture = 0; QRHI_RES_RHI(QRhiGles2); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } QRhiRenderPassDescriptor *QGles2TextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; } bool QGles2TextureRenderTarget::create() @@ -4820,13 +5910,13 @@ bool QGles2TextureRenderTarget::create() if (framebuffer) destroy(); - const bool hasColorAttachments = m_desc.cbeginColorAttachments() != m_desc.cendColorAttachments(); + const bool hasColorAttachments = m_desc.colorAttachmentCount() > 0; Q_ASSERT(hasColorAttachments || m_desc.depthTexture()); Q_ASSERT(!m_desc.depthStencilBuffer() || !m_desc.depthTexture()); const bool hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); if (hasColorAttachments) { - const int count = m_desc.cendColorAttachments() - m_desc.cbeginColorAttachments(); + const int count = int(m_desc.colorAttachmentCount()); if (count > rhiD->caps.maxDrawBuffers) { qWarning("QGles2TextureRenderTarget: Too many color attachments (%d, max is %d)", count, rhiD->caps.maxDrawBuffers); @@ -4843,6 +5933,7 @@ bool QGles2TextureRenderTarget::create() d.colorAttCount = 0; int attIndex = 0; + int multiViewCount = 0; for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it, ++attIndex) { d.colorAttCount += 1; const QRhiColorAttachment &colorAtt(*it); @@ -4852,17 +5943,57 @@ bool QGles2TextureRenderTarget::create() if (texture) { QGles2Texture *texD = QRHI_RES(QGles2Texture, texture); Q_ASSERT(texD->texture && texD->specified); - if (texD->flags().testFlag(QRhiTexture::ThreeDimensional)) { - rhiD->f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), texD->texture, - colorAtt.level(), colorAtt.layer()); + if (texD->flags().testFlag(QRhiTexture::ThreeDimensional) || texD->flags().testFlag(QRhiTexture::TextureArray)) { + if (colorAtt.multiViewCount() < 2) { + rhiD->f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), texD->texture, + colorAtt.level(), colorAtt.layer()); + } else { + multiViewCount = colorAtt.multiViewCount(); + if (texD->sampleCount() > 1 && rhiD->caps.glesMultiviewMultisampleRenderToTexture && colorAtt.resolveTexture()) { + // Special path for GLES and GL_OVR_multiview_multisampled_render_to_texture: + // ignore the color attachment's (multisample) texture + // array and give the resolve texture array to GL. (no + // explicit resolving is needed by us later on) + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_COLOR_ATTACHMENT0 + uint(attIndex), + resolveTexD->texture, + colorAtt.resolveLevel(), + texD->sampleCount(), + colorAtt.resolveLayer(), + multiViewCount); + } else { + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, + GL_COLOR_ATTACHMENT0 + uint(attIndex), + texD->texture, + colorAtt.level(), + colorAtt.layer(), + multiViewCount); + } + } + } else if (texD->flags().testFlag(QRhiTexture::OneDimensional)) { + rhiD->glFramebufferTexture1D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), + texD->target + uint(colorAtt.layer()), texD->texture, + colorAtt.level()); } else { - const GLenum faceTargetBase = texD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : texD->target; - rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.layer()), - texD->texture, colorAtt.level()); + if (texD->sampleCount() > 1 && rhiD->caps.glesMultisampleRenderToTexture && colorAtt.resolveTexture()) { + // Special path for GLES and GL_EXT_multisampled_render_to_texture: + // ignore the color attachment's (multisample) texture and + // give the resolve texture to GL. (no explicit resolving is + // needed by us later on) + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + const GLenum faceTargetBase = resolveTexD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : resolveTexD->target; + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.resolveLayer()), + resolveTexD->texture, colorAtt.level(), texD->sampleCount()); + } else { + const GLenum faceTargetBase = texD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : texD->target; + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.layer()), + texD->texture, colorAtt.level()); + } } if (attIndex == 0) { d.pixelSize = rhiD->q->sizeForMipLevel(colorAtt.level(), texD->pixelSize()); - d.sampleCount = 1; + d.sampleCount = texD->sampleCount(); } } else if (renderBuffer) { QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, renderBuffer); @@ -4883,12 +6014,14 @@ bool QGles2TextureRenderTarget::create() } else { rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, depthRbD->renderbuffer); - if (depthRbD->stencilRenderbuffer) + if (depthRbD->stencilRenderbuffer) { rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, depthRbD->stencilRenderbuffer); - else // packed + } else { + // packed depth-stencil rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, depthRbD->renderbuffer); + } } if (d.colorAttCount == 0) { d.pixelSize = depthRbD->pixelSize(); @@ -4896,11 +6029,105 @@ bool QGles2TextureRenderTarget::create() } } else { QGles2Texture *depthTexD = QRHI_RES(QGles2Texture, m_desc.depthTexture()); - rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->target, - depthTexD->texture, 0); + if (multiViewCount < 2) { + if (depthTexD->sampleCount() > 1 && rhiD->caps.glesMultisampleRenderToTexture && m_desc.depthResolveTexture()) { + // Special path for GLES and + // GL_EXT_multisampled_render_to_texture, for depth-stencil. + // Relevant only when depthResolveTexture is set. + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, m_desc.depthResolveTexture()); + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthResolveTexD->target, + depthResolveTexD->texture, 0, depthTexD->sampleCount()); + if (rhiD->isStencilSupportingFormat(depthResolveTexD->format())) { + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthResolveTexD->target, + depthResolveTexD->texture, 0, depthTexD->sampleCount()); + } + } else { + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->target, + depthTexD->texture, 0); + if (rhiD->isStencilSupportingFormat(depthTexD->format())) { + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthTexD->target, + depthTexD->texture, 0); + } + } + } else { + if (depthTexD->sampleCount() > 1 && rhiD->caps.glesMultiviewMultisampleRenderToTexture) { + // And so it turns out + // https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt + // does not work with multisample 2D texture arrays. (at least + // that's what Issue 30 in the extension spec seems to imply) + // + // There is https://registry.khronos.org/OpenGL/extensions/EXT/EXT_multiview_texture_multisample.txt + // that seems to resolve that, but that does not seem to + // work (or not available) on GLES devices such as the Quest 3. + // + // So instead, on GLES we can use the + // multisample-multiview-auto-resolving version (which in + // turn is not supported on desktop GL e.g. by NVIDIA), too + // bad we have a multisample depth texture array here as + // every other API out there requires that. So, in absence + // of a depthResolveTexture, create a temporary one ignoring + // what the user has already created. + // + if (!m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture()) { + qWarning("Attempted to create a multiview+multisample QRhiTextureRenderTarget, but DoNotStoreDepthStencilContents was not set." + " This path has no choice but to behave as if DoNotStoreDepthStencilContents was set, because QRhi is forced to create" + " a throwaway non-multisample depth texture here. Set the flag to silence this warning, or set a depthResolveTexture."); + } + if (m_desc.depthResolveTexture()) { + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, m_desc.depthResolveTexture()); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_DEPTH_ATTACHMENT, + depthResolveTexD->texture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + if (rhiD->isStencilSupportingFormat(depthResolveTexD->format())) { + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_STENCIL_ATTACHMENT, + depthResolveTexD->texture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + } + } else { + if (!nonMsaaThrowawayDepthTexture) { + rhiD->f->glGenTextures(1, &nonMsaaThrowawayDepthTexture); + rhiD->f->glBindTexture(GL_TEXTURE_2D_ARRAY, nonMsaaThrowawayDepthTexture); + rhiD->f->glTexStorage3D(GL_TEXTURE_2D_ARRAY, 1, GL_DEPTH24_STENCIL8, + depthTexD->pixelSize().width(), depthTexD->pixelSize().height(), multiViewCount); + } + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_DEPTH_ATTACHMENT, + nonMsaaThrowawayDepthTexture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_STENCIL_ATTACHMENT, + nonMsaaThrowawayDepthTexture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + } + } else { + // The depth texture here must be an array with at least + // multiViewCount elements, and the format should be D24 or D32F + // for depth only, or D24S8 for depth and stencil. + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->texture, + 0, 0, multiViewCount); + if (rhiD->isStencilSupportingFormat(depthTexD->format())) { + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthTexD->texture, + 0, 0, multiViewCount); + } + } + } if (d.colorAttCount == 0) { d.pixelSize = depthTexD->pixelSize(); - d.sampleCount = 1; + d.sampleCount = depthTexD->sampleCount(); } } d.dsAttCount = 1; @@ -4917,12 +6144,20 @@ bool QGles2TextureRenderTarget::create() return false; } + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_FRAMEBUFFER, framebuffer, -1, m_objectName.constData()); + + QRhiRenderTargetAttachmentTracker::updateResIdList<QGles2Texture, QGles2RenderBuffer>(m_desc, &d.currentResIdList); + rhiD->registerResource(this); return true; } QSize QGles2TextureRenderTarget::pixelSize() const { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QGles2Texture, QGles2RenderBuffer>(m_desc, d.currentResIdList)) + const_cast<QGles2TextureRenderTarget *>(this)->create(); + return d.pixelSize; } @@ -4948,7 +6183,9 @@ QGles2ShaderResourceBindings::~QGles2ShaderResourceBindings() void QGles2ShaderResourceBindings::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2ShaderResourceBindings::create() @@ -4958,8 +6195,8 @@ bool QGles2ShaderResourceBindings::create() return false; hasDynamicOffset = false; - for (int i = 0, ie = m_bindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = m_bindings.at(i).data(); + for (int i = 0, ie = m_bindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = QRhiImplementation::shaderResourceBindingData(m_bindings.at(i)); if (b->type == QRhiShaderResourceBinding::UniformBuffer) { if (b->u.ubuf.hasDynamicOffset) { hasDynamicOffset = true; @@ -4971,9 +6208,16 @@ bool QGles2ShaderResourceBindings::create() rhiD->updateLayoutDesc(this); generation += 1; + rhiD->registerResource(this, false); return true; } +void QGles2ShaderResourceBindings::updateResources(UpdateFlags flags) +{ + Q_UNUSED(flags); + generation += 1; +} + QGles2GraphicsPipeline::QGles2GraphicsPipeline(QRhiImplementation *rhi) : QRhiGraphicsPipeline(rhi) { @@ -4999,9 +6243,20 @@ void QGles2GraphicsPipeline::destroy() samplers.clear(); QRHI_RES_RHI(QRhiGles2); - rhiD->releaseQueue.append(e); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } +} - rhiD->unregisterResource(this); +static inline bool isGraphicsStage(const QRhiShaderStage &shaderStage) +{ + const QRhiShaderStage::Type t = shaderStage.type(); + return t == QRhiShaderStage::Vertex + || t == QRhiShaderStage::TessellationControl + || t == QRhiShaderStage::TessellationEvaluation + || t == QRhiShaderStage::Geometry + || t == QRhiShaderStage::Fragment; } bool QGles2GraphicsPipeline::create() @@ -5014,6 +6269,7 @@ bool QGles2GraphicsPipeline::create() if (!rhiD->ensureContext()) return false; + rhiD->pipelineCreationStart(); if (!rhiD->sanityCheckGraphicsPipeline(this)) return false; @@ -5021,40 +6277,72 @@ bool QGles2GraphicsPipeline::create() program = rhiD->f->glCreateProgram(); - QShaderDescription vsDesc; - QShaderDescription fsDesc; - for (const QRhiShaderStage &shaderStage : qAsConst(m_shaderStages)) { - if (shaderStage.type() == QRhiShaderStage::Vertex) - vsDesc = shaderStage.shader().description(); - else if (shaderStage.type() == QRhiShaderStage::Fragment) - fsDesc = shaderStage.shader().description(); + enum { + VtxIdx = 0, + TCIdx, + TEIdx, + GeomIdx, + FragIdx, + LastIdx + }; + const auto descIdxForStage = [](const QRhiShaderStage &shaderStage) { + switch (shaderStage.type()) { + case QRhiShaderStage::Vertex: + return VtxIdx; + case QRhiShaderStage::TessellationControl: + return TCIdx; + case QRhiShaderStage::TessellationEvaluation: + return TEIdx; + case QRhiShaderStage::Geometry: + return GeomIdx; + case QRhiShaderStage::Fragment: + return FragIdx; + default: + break; + } + Q_UNREACHABLE_RETURN(VtxIdx); + }; + QShaderDescription desc[LastIdx]; + QShader::SeparateToCombinedImageSamplerMappingList samplerMappingList[LastIdx]; + bool vertexFragmentOnly = true; + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + if (isGraphicsStage(shaderStage)) { + const int idx = descIdxForStage(shaderStage); + if (idx != VtxIdx && idx != FragIdx) + vertexFragmentOnly = false; + QShader shader = shaderStage.shader(); + QShaderVersion shaderVersion; + desc[idx] = shader.description(); + if (!rhiD->shaderSource(shaderStage, &shaderVersion).isEmpty()) { + samplerMappingList[idx] = shader.separateToCombinedImageSamplerMappingList( + { QShader::GlslShader, shaderVersion, shaderStage.shaderVariant() }); + } + } } QByteArray cacheKey; QRhiGles2::ProgramCacheResult cacheResult = rhiD->tryLoadFromDiskOrPipelineCache(m_shaderStages.constData(), - m_shaderStages.count(), + m_shaderStages.size(), program, - vsDesc.inputVariables(), + desc[VtxIdx].inputVariables(), &cacheKey); if (cacheResult == QRhiGles2::ProgramCacheError) return false; if (cacheResult == QRhiGles2::ProgramCacheMiss) { - for (const QRhiShaderStage &shaderStage : qAsConst(m_shaderStages)) { - if (shaderStage.type() == QRhiShaderStage::Vertex) { - if (!rhiD->compileShader(program, shaderStage, nullptr)) - return false; - } else if (shaderStage.type() == QRhiShaderStage::Fragment) { + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + if (isGraphicsStage(shaderStage)) { if (!rhiD->compileShader(program, shaderStage, nullptr)) return false; } } // important when GLSL <= 150 is used that does not have location qualifiers - for (const QShaderDescription::InOutVariable &inVar : vsDesc.inputVariables()) + for (const QShaderDescription::InOutVariable &inVar : desc[VtxIdx].inputVariables()) rhiD->f->glBindAttribLocation(program, GLuint(inVar.location), inVar.name); - rhiD->sanityCheckVertexFragmentInterface(vsDesc, fsDesc); + if (vertexFragmentOnly) + rhiD->sanityCheckVertexFragmentInterface(desc[VtxIdx], desc[FragIdx]); if (!rhiD->linkProgram(program)) return false; @@ -5080,13 +6368,19 @@ bool QGles2GraphicsPipeline::create() // Use the same work area for the vertex & fragment stages, thus ensuring // that we will not do superfluous glUniform calls for uniforms that are // present in both shaders. - QSet<int> activeUniformLocations; - - for (const QShaderDescription::UniformBlock &ub : vsDesc.uniformBlocks()) - rhiD->gatherUniforms(program, ub, &activeUniformLocations, &uniforms); - - for (const QShaderDescription::UniformBlock &ub : fsDesc.uniformBlocks()) - rhiD->gatherUniforms(program, ub, &activeUniformLocations, &uniforms); + QDuplicateTracker<int, 256> activeUniformLocations; + + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + if (isGraphicsStage(shaderStage)) { + const int idx = descIdxForStage(shaderStage); + for (const QShaderDescription::UniformBlock &ub : desc[idx].uniformBlocks()) + rhiD->gatherUniforms(program, ub, &activeUniformLocations, &uniforms); + for (const QShaderDescription::InOutVariable &v : desc[idx].combinedImageSamplers()) + rhiD->gatherSamplers(program, v, &samplers); + for (const QShader::SeparateToCombinedImageSamplerMapping &mapping : samplerMappingList[idx]) + rhiD->gatherGeneratedSamplers(program, mapping, &samplers); + } + } std::sort(uniforms.begin(), uniforms.end(), [](const QGles2UniformDescription &a, const QGles2UniformDescription &b) @@ -5094,17 +6388,15 @@ bool QGles2GraphicsPipeline::create() return a.offset < b.offset; }); - for (const QShaderDescription::InOutVariable &v : vsDesc.combinedImageSamplers()) - rhiD->gatherSamplers(program, v, &samplers); - - for (const QShaderDescription::InOutVariable &v : fsDesc.combinedImageSamplers()) - rhiD->gatherSamplers(program, v, &samplers); - memset(uniformState, 0, sizeof(uniformState)); currentSrb = nullptr; currentSrbGeneration = 0; + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_PROGRAM, program, -1, m_objectName.constData()); + + rhiD->pipelineCreationEnd(); generation += 1; rhiD->registerResource(this); return true; @@ -5135,9 +6427,10 @@ void QGles2ComputePipeline::destroy() samplers.clear(); QRHI_RES_RHI(QRhiGles2); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QGles2ComputePipeline::create() @@ -5150,7 +6443,16 @@ bool QGles2ComputePipeline::create() if (!rhiD->ensureContext()) return false; + rhiD->pipelineCreationStart(); + const QShaderDescription csDesc = m_shaderStage.shader().description(); + QShader::SeparateToCombinedImageSamplerMappingList csSamplerMappingList; + QShaderVersion shaderVersion; + if (!rhiD->shaderSource(m_shaderStage, &shaderVersion).isEmpty()) { + csSamplerMappingList = m_shaderStage.shader().separateToCombinedImageSamplerMappingList( + { QShader::GlslShader, shaderVersion, m_shaderStage.shaderVariant() }); + } + program = rhiD->f->glCreateProgram(); QByteArray cacheKey; @@ -5183,11 +6485,13 @@ bool QGles2ComputePipeline::create() } } - QSet<int> activeUniformLocations; + QDuplicateTracker<int, 256> activeUniformLocations; for (const QShaderDescription::UniformBlock &ub : csDesc.uniformBlocks()) rhiD->gatherUniforms(program, ub, &activeUniformLocations, &uniforms); for (const QShaderDescription::InOutVariable &v : csDesc.combinedImageSamplers()) rhiD->gatherSamplers(program, v, &samplers); + for (const QShader::SeparateToCombinedImageSamplerMapping &mapping : csSamplerMappingList) + rhiD->gatherGeneratedSamplers(program, mapping, &samplers); // storage images and buffers need no special steps here @@ -5196,6 +6500,7 @@ bool QGles2ComputePipeline::create() currentSrb = nullptr; currentSrbGeneration = 0; + rhiD->pipelineCreationEnd(); generation += 1; rhiD->registerResource(this); return true; @@ -5219,7 +6524,9 @@ void QGles2CommandBuffer::destroy() QGles2SwapChain::QGles2SwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), - rt(rhi), + rt(rhi, this), + rtLeft(rhi, this), + rtRight(rhi, this), cb(rhi) { } @@ -5231,8 +6538,9 @@ QGles2SwapChain::~QGles2SwapChain() void QGles2SwapChain::destroy() { - QRHI_PROF; - QRHI_PROF_F(releaseSwapChain(this)); + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } QRhiCommandBuffer *QGles2SwapChain::currentFrameCommandBuffer() @@ -5245,19 +6553,59 @@ QRhiRenderTarget *QGles2SwapChain::currentFrameRenderTarget() return &rt; } +QRhiRenderTarget *QGles2SwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + if (targetBuffer == LeftBuffer) + return rtLeft.d.isValid() ? &rtLeft : &rt; + else if (targetBuffer == RightBuffer) + return rtRight.d.isValid() ? &rtRight : &rt; + else + Q_UNREACHABLE_RETURN(nullptr); +} + QSize QGles2SwapChain::surfacePixelSize() { Q_ASSERT(m_window); - return m_window->size() * m_window->devicePixelRatio(); + if (QPlatformWindow *platformWindow = m_window->handle()) + // Prefer using QPlatformWindow geometry and DPR in order to avoid + // errors due to rounded QWindow geometry. + return platformWindow->geometry().size() * platformWindow->devicePixelRatio(); + else + return m_window->size() * m_window->devicePixelRatio(); +} + +bool QGles2SwapChain::isFormatSupported(Format f) +{ + return f == SDR; } QRhiRenderPassDescriptor *QGles2SwapChain::newCompatibleRenderPassDescriptor() { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; +} + +void QGles2SwapChain::initSwapChainRenderTarget(QGles2SwapChainRenderTarget *rt) +{ + rt->setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget + rt->d.rp = QRHI_RES(QGles2RenderPassDescriptor, m_renderPassDesc); + rt->d.pixelSize = pixelSize; + rt->d.dpr = float(m_window->devicePixelRatio()); + rt->d.sampleCount = qBound(1, m_sampleCount, 64); + rt->d.colorAttCount = 1; + rt->d.dsAttCount = m_depthStencil ? 1 : 0; + rt->d.srgbUpdateAndBlend = m_flags.testFlag(QRhiSwapChain::sRGB); } bool QGles2SwapChain::createOrResize() { + // can be called multiple times due to window resizes + const bool needsRegistration = !surface || surface != m_window; + if (surface && surface != m_window) + destroy(); + surface = m_window; m_currentPixelSize = surfacePixelSize(); pixelSize = m_currentPixelSize; @@ -5269,21 +6617,70 @@ bool QGles2SwapChain::createOrResize() m_depthStencil->create(); } - rt.d.rp = QRHI_RES(QGles2RenderPassDescriptor, m_renderPassDesc); - rt.d.pixelSize = pixelSize; - rt.d.dpr = float(m_window->devicePixelRatio()); - rt.d.sampleCount = qBound(1, m_sampleCount, 64); - rt.d.colorAttCount = 1; - rt.d.dsAttCount = m_depthStencil ? 1 : 0; - rt.d.srgbUpdateAndBlend = m_flags.testFlag(QRhiSwapChain::sRGB); + initSwapChainRenderTarget(&rt); + + if (m_window->format().stereo()) { + initSwapChainRenderTarget(&rtLeft); + rtLeft.d.stereoTarget = QRhiSwapChain::LeftBuffer; + initSwapChainRenderTarget(&rtRight); + rtRight.d.stereoTarget = QRhiSwapChain::RightBuffer; + } frameCount = 0; - QRHI_PROF; - // make something up - QRHI_PROF_F(resizeSwapChain(this, 2, m_sampleCount > 1 ? 2 : 0, m_sampleCount)); + QRHI_RES_RHI(QRhiGles2); + if (rhiD->rhiFlags.testFlag(QRhi::EnableTimestamps) && rhiD->caps.timestamps) + timestamps.prepare(rhiD); + + // The only reason to register this fairly fake gl swapchain + // object with no native resources underneath is to be able to + // implement a safe destroy(). + if (needsRegistration) + rhiD->registerResource(this, false); return true; } +void QGles2SwapChainTimestamps::prepare(QRhiGles2 *rhiD) +{ + if (!query[0]) + rhiD->f->glGenQueries(TIMESTAMP_PAIRS * 2, query); +} + +void QGles2SwapChainTimestamps::destroy(QRhiGles2 *rhiD) +{ + rhiD->f->glDeleteQueries(TIMESTAMP_PAIRS * 2, query); + memset(active, 0, sizeof(active)); + memset(query, 0, sizeof(query)); +} + +bool QGles2SwapChainTimestamps::tryQueryTimestamps(int pairIndex, QRhiGles2 *rhiD, double *elapsedSec) +{ + if (!active[pairIndex]) + return false; + + GLuint tsStart = query[pairIndex * 2]; + GLuint tsEnd = query[pairIndex * 2 + 1]; + + GLuint ready = GL_FALSE; + rhiD->f->glGetQueryObjectuiv(tsEnd, GL_QUERY_RESULT_AVAILABLE, &ready); + + if (!ready) + return false; + + bool result = false; + quint64 timestamps[2]; + rhiD->glGetQueryObjectui64v(tsStart, GL_QUERY_RESULT, ×tamps[0]); + rhiD->glGetQueryObjectui64v(tsEnd, GL_QUERY_RESULT, ×tamps[1]); + + if (timestamps[1] >= timestamps[0]) { + const quint64 nanoseconds = timestamps[1] - timestamps[0]; + *elapsedSec = nanoseconds / 1000000000.0; + result = true; + } + + active[pairIndex] = false; + return result; +} + QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhigles2_p.h b/src/gui/rhi/qrhigles2_p.h index 7273b8ee9e..4305186c02 100644 --- a/src/gui/rhi/qrhigles2_p.h +++ b/src/gui/rhi/qrhigles2_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIGLES2_H -#define QRHIGLES2_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHIGLES2_P_H +#define QRHIGLES2_P_H // // W A R N I N G @@ -51,32 +15,1121 @@ // We mean it. // -#include <private/qrhi_p.h> -#include <QtGui/qsurfaceformat.h> +#include "qrhi_p.h" +#include <rhi/qshaderdescription.h> +#include <qopengl.h> +#include <QByteArray> +#include <QWindow> +#include <QPointer> +#include <QtCore/private/qduplicatetracker_p.h> +#include <optional> QT_BEGIN_NAMESPACE -class QOpenGLContext; -class QOffscreenSurface; -class QWindow; +class QOpenGLExtensions; +class QRhiGles2; + +struct QGles2Buffer : public QRhiBuffer +{ + QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QGles2Buffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + quint32 nonZeroSize = 0; + GLuint buffer = 0; + GLenum targetForDataOps; + QByteArray data; + enum Access { + AccessNone, + AccessVertex, + AccessIndex, + AccessUniform, + AccessStorageRead, + AccessStorageWrite, + AccessStorageReadWrite, + AccessUpdate + }; + struct UsageState { + Access access; + }; + UsageState usageState; + friend class QRhiGles2; +}; + +struct QGles2RenderBuffer : public QRhiRenderBuffer +{ + QGles2RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QGles2RenderBuffer(); + void destroy() override; + bool create() override; + bool createFrom(NativeRenderBuffer src) override; + QRhiTexture::Format backingFormat() const override; + + GLuint renderbuffer = 0; + GLuint stencilRenderbuffer = 0; // when packed depth-stencil not supported + int samples; + bool owns = true; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2SamplerData +{ + GLenum glminfilter = 0; + GLenum glmagfilter = 0; + GLenum glwraps = 0; + GLenum glwrapt = 0; + GLenum glwrapr = 0; + GLenum gltexcomparefunc = 0; +}; + +inline bool operator==(const QGles2SamplerData &a, const QGles2SamplerData &b) +{ + return a.glminfilter == b.glminfilter + && a.glmagfilter == b.glmagfilter + && a.glwraps == b.glwraps + && a.glwrapt == b.glwrapt + && a.glwrapr == b.glwrapr + && a.gltexcomparefunc == b.gltexcomparefunc; +} + +inline bool operator!=(const QGles2SamplerData &a, const QGles2SamplerData &b) +{ + return !(a == b); +} + +struct QGles2Texture : public QRhiTexture +{ + QGles2Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QGles2Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + + GLuint texture = 0; + bool owns = true; + GLenum target; + GLenum glintformat; + GLenum glsizedintformat; + GLenum glformat; + GLenum gltype; + QGles2SamplerData samplerState; + bool specified = false; + bool zeroInitialized = false; + int mipLevelCount = 0; + + enum Access { + AccessNone, + AccessSample, + AccessFramebuffer, + AccessStorageRead, + AccessStorageWrite, + AccessStorageReadWrite, + AccessUpdate, + AccessRead + }; + struct UsageState { + Access access; + }; + UsageState usageState; + + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2Sampler : public QRhiSampler +{ + QGles2Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QGles2Sampler(); + void destroy() override; + bool create() override; + + QGles2SamplerData d; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QGles2RenderPassDescriptor(QRhiImplementation *rhi); + ~QGles2RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QGles2RenderTargetData +{ + QGles2RenderTargetData(QRhiImplementation *) { } + + bool isValid() const { return rp != nullptr; } + + QGles2RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + bool srgbUpdateAndBlend = false; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + std::optional<QRhiSwapChain::StereoTargetBuffer> stereoTarget; +}; + +struct QGles2SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QGles2SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QGles2SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QGles2RenderTargetData d; +}; + +struct QGles2TextureRenderTarget : public QRhiTextureRenderTarget +{ + QGles2TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QGles2TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QGles2RenderTargetData d; + GLuint framebuffer = 0; + GLuint nonMsaaThrowawayDepthTexture = 0; + friend class QRhiGles2; +}; + +struct QGles2ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QGles2ShaderResourceBindings(QRhiImplementation *rhi); + ~QGles2ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + bool hasDynamicOffset = false; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2UniformDescription +{ + QShaderDescription::VariableType type; + int glslLocation; + int binding; + quint32 offset; + quint32 size; + int arrayDim; +}; + +Q_DECLARE_TYPEINFO(QGles2UniformDescription, Q_RELOCATABLE_TYPE); + +struct QGles2SamplerDescription +{ + int glslLocation; + int combinedBinding; + int tbinding; + int sbinding; +}; + +Q_DECLARE_TYPEINFO(QGles2SamplerDescription, Q_RELOCATABLE_TYPE); + +using QGles2UniformDescriptionVector = QVarLengthArray<QGles2UniformDescription, 8>; +using QGles2SamplerDescriptionVector = QVarLengthArray<QGles2SamplerDescription, 4>; + +struct QGles2UniformState +{ + static constexpr int MAX_TRACKED_LOCATION = 1023; + int componentCount; + float v[4]; +}; + +struct QGles2GraphicsPipeline : public QRhiGraphicsPipeline +{ + QGles2GraphicsPipeline(QRhiImplementation *rhi); + ~QGles2GraphicsPipeline(); + void destroy() override; + bool create() override; + + GLuint program = 0; + GLenum drawMode = GL_TRIANGLES; + QGles2UniformDescriptionVector uniforms; + QGles2SamplerDescriptionVector samplers; + QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; + QRhiShaderResourceBindings *currentSrb = nullptr; + uint currentSrbGeneration = 0; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2ComputePipeline : public QRhiComputePipeline +{ + QGles2ComputePipeline(QRhiImplementation *rhi); + ~QGles2ComputePipeline(); + void destroy() override; + bool create() override; + + GLuint program = 0; + QGles2UniformDescriptionVector uniforms; + QGles2SamplerDescriptionVector samplers; + QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; + QRhiShaderResourceBindings *currentSrb = nullptr; + uint currentSrbGeneration = 0; + uint generation = 0; + friend class QRhiGles2; +}; -struct Q_GUI_EXPORT QRhiGles2InitParams : public QRhiInitParams +struct QGles2CommandBuffer : public QRhiCommandBuffer { - QRhiGles2InitParams(); + QGles2CommandBuffer(QRhiImplementation *rhi); + ~QGles2CommandBuffer(); + void destroy() override; + + // keep at a reasonably low value otherwise sizeof Command explodes + static const int MAX_DYNAMIC_OFFSET_COUNT = 8; + + struct Command { + enum Cmd { + BeginFrame, + EndFrame, + ResetFrame, + Viewport, + Scissor, + BlendConstants, + StencilRef, + BindVertexBuffer, + BindIndexBuffer, + Draw, + DrawIndexed, + BindGraphicsPipeline, + BindShaderResources, + BindFramebuffer, + Clear, + BufferSubData, + GetBufferSubData, + CopyTex, + ReadPixels, + SubImage, + CompressedImage, + CompressedSubImage, + BlitFromRenderbuffer, + BlitFromTexture, + GenMip, + BindComputePipeline, + Dispatch, + BarriersForPass, + Barrier, + InvalidateFramebuffer + }; + Cmd cmd; + + // QRhi*/QGles2* references should be kept at minimum (so no + // QRhiTexture/Buffer/etc. pointers). + union Args { + struct { + GLuint timestampQuery; + } beginFrame; + struct { + GLuint timestampQuery; + } endFrame; + struct { + float x, y, w, h; + float d0, d1; + } viewport; + struct { + int x, y, w, h; + } scissor; + struct { + float r, g, b, a; + } blendConstants; + struct { + quint32 ref; + QRhiGraphicsPipeline *ps; + } stencilRef; + struct { + QRhiGraphicsPipeline *ps; + GLuint buffer; + quint32 offset; + int binding; + } bindVertexBuffer; + struct { + GLuint buffer; + quint32 offset; + GLenum type; + } bindIndexBuffer; + struct { + QRhiGraphicsPipeline *ps; + quint32 vertexCount; + quint32 firstVertex; + quint32 instanceCount; + quint32 baseInstance; + } draw; + struct { + QRhiGraphicsPipeline *ps; + quint32 indexCount; + quint32 firstIndex; + quint32 instanceCount; + quint32 baseInstance; + qint32 baseVertex; + } drawIndexed; + struct { + QRhiGraphicsPipeline *ps; + } bindGraphicsPipeline; + struct { + QRhiGraphicsPipeline *maybeGraphicsPs; + QRhiComputePipeline *maybeComputePs; + QRhiShaderResourceBindings *srb; + int dynamicOffsetCount; + uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offset + } bindShaderResources; + struct { + GLbitfield mask; + float c[4]; + float d; + quint32 s; + } clear; + struct { + GLuint fbo; + bool srgb; + int colorAttCount; + bool stereo; + QRhiSwapChain::StereoTargetBuffer stereoTarget; + } bindFramebuffer; + struct { + GLenum target; + GLuint buffer; + int offset; + int size; + const void *data; // must come from retainData() + } bufferSubData; + struct { + QRhiReadbackResult *result; + GLenum target; + GLuint buffer; + int offset; + int size; + } getBufferSubData; + struct { + GLenum srcTarget; + GLenum srcFaceTarget; + GLuint srcTexture; + int srcLevel; + int srcX; + int srcY; + int srcZ; + GLenum dstTarget; + GLuint dstTexture; + GLenum dstFaceTarget; + int dstLevel; + int dstX; + int dstY; + int dstZ; + int w; + int h; + } copyTex; + struct { + QRhiReadbackResult *result; + GLuint texture; + int w; + int h; + QRhiTexture::Format format; + GLenum readTarget; + int level; + int slice3D; + } readPixels; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + int dx; + int dy; + int dz; + int w; + int h; + GLenum glformat; + GLenum gltype; + int rowStartAlign; + int rowLength; + const void *data; // must come from retainImage() + } subImage; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + GLenum glintformat; + int w; + int h; + int depth; + int size; + const void *data; // must come from retainData() + } compressedImage; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + int dx; + int dy; + int dz; + int w; + int h; + GLenum glintformat; + int size; + const void *data; // must come from retainData() + } compressedSubImage; + struct { + GLuint renderbuffer; + int w; + int h; + GLenum target; + GLuint dstTexture; + int dstLevel; + int dstLayer; + bool isDepthStencil; + } blitFromRenderbuffer; + struct { + GLenum srcTarget; + GLuint srcTexture; + int srcLevel; + int srcLayer; + int w; + int h; + GLenum dstTarget; + GLuint dstTexture; + int dstLevel; + int dstLayer; + bool isDepthStencil; + } blitFromTexture; + struct { + GLenum target; + GLuint texture; + } genMip; + struct { + QRhiComputePipeline *ps; + } bindComputePipeline; + struct { + GLuint x; + GLuint y; + GLuint z; + } dispatch; + struct { + int trackerIndex; + } barriersForPass; + struct { + GLbitfield barriers; + } barrier; + struct { + int attCount; + GLenum att[3]; + } invalidateFramebuffer; + } args; + }; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; - QSurfaceFormat format; - QOffscreenSurface *fallbackSurface = nullptr; - QWindow *window = nullptr; + QRhiBackendCommandList<Command> commands; + QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; + int currentPassResTrackerIndex; - static QOffscreenSurface *newFallbackSurface(const QSurfaceFormat &format = QSurfaceFormat::defaultFormat()); - static QSurfaceFormat adjustedFormat(const QSurfaceFormat &format = QSurfaceFormat::defaultFormat()); + PassType recordingPass; + bool passNeedsResourceTracking; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + + struct GraphicsPassState { + bool valid = false; + bool scissor; + bool cullFace; + GLenum cullMode; + GLenum frontFace; + bool blendEnabled; + struct ColorMask { bool r, g, b, a; } colorMask; + struct Blend { + GLenum srcColor; + GLenum dstColor; + GLenum srcAlpha; + GLenum dstAlpha; + GLenum opColor; + GLenum opAlpha; + } blend; + bool depthTest; + bool depthWrite; + GLenum depthFunc; + bool stencilTest; + GLuint stencilReadMask; + GLuint stencilWriteMask; + struct StencilFace { + GLenum func; + GLenum failOp; + GLenum zfailOp; + GLenum zpassOp; + } stencil[2]; // front, back + bool polyOffsetFill; + float polyOffsetFactor; + float polyOffsetUnits; + float lineWidth; + int cpCount; + GLenum polygonMode; + void reset() { valid = false; } + struct { + // not part of QRhiGraphicsPipeline but used by setGraphicsPipeline() + GLint stencilRef = 0; + } dynamic; + } graphicsPassState; + + struct ComputePassState { + enum Access { + Read = 0x01, + Write = 0x02 + }; + QHash<QRhiResource *, QPair<int, bool> > writtenResources; + void reset() { + writtenResources.clear(); + } + } computePassState; + + struct TextureUnitState { + void *ps; + uint psGeneration; + uint texture; + } textureUnitState[16]; + + QVarLengthArray<QByteArray, 4> dataRetainPool; + QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; + QVarLengthArray<QImage, 4> imageRetainPool; + + // relies heavily on implicit sharing (no copies of the actual data will be made) + const void *retainData(const QByteArray &data) { + dataRetainPool.append(data); + return dataRetainPool.last().constData(); + } + const uchar *retainBufferData(const QRhiBufferData &data) { + bufferDataRetainPool.append(data); + return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); + } + const void *retainImage(const QImage &image) { + imageRetainPool.append(image); + return imageRetainPool.last().constBits(); + } + void resetCommands() { + commands.reset(); + dataRetainPool.clear(); + bufferDataRetainPool.clear(); + imageRetainPool.clear(); + + passResTrackers.clear(); + currentPassResTrackerIndex = -1; + } + void resetState() { + recordingPass = NoPass; + passNeedsResourceTracking = true; + // do not zero lastGpuTime + currentTarget = nullptr; + resetCommands(); + resetCachedState(); + } + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + graphicsPassState.reset(); + computePassState.reset(); + memset(textureUnitState, 0, sizeof(textureUnitState)); + } }; -struct Q_GUI_EXPORT QRhiGles2NativeHandles : public QRhiNativeHandles +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, + const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) +{ + return a.func == b.func + && a.failOp == b.failOp + && a.zfailOp == b.zfailOp + && a.zpassOp == b.zpassOp; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, + const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) { - QOpenGLContext *context = nullptr; + return !(a == b); +} + +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, + const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) +{ + return a.r == b.r && a.g == b.g && a.b == b.b && a.a == b.a; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, + const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) +{ + return !(a == b); +} + +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::Blend &a, + const QGles2CommandBuffer::GraphicsPassState::Blend &b) +{ + return a.srcColor == b.srcColor + && a.dstColor == b.dstColor + && a.srcAlpha == b.srcAlpha + && a.dstAlpha == b.dstAlpha + && a.opColor == b.opColor + && a.opAlpha == b.opAlpha; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::Blend &a, + const QGles2CommandBuffer::GraphicsPassState::Blend &b) +{ + return !(a == b); +} + +struct QGles2SwapChainTimestamps +{ + static const int TIMESTAMP_PAIRS = 2; + + bool active[TIMESTAMP_PAIRS] = {}; + GLuint query[TIMESTAMP_PAIRS * 2] = {}; + + void prepare(QRhiGles2 *rhiD); + void destroy(QRhiGles2 *rhiD); + bool tryQueryTimestamps(int pairIndex, QRhiGles2 *rhiD, double *elapsedSec); }; +struct QGles2SwapChain : public QRhiSwapChain +{ + QGles2SwapChain(QRhiImplementation *rhi); + ~QGles2SwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void initSwapChainRenderTarget(QGles2SwapChainRenderTarget *rt); + + QSurface *surface = nullptr; + QSize pixelSize; + QGles2SwapChainRenderTarget rt; + QGles2SwapChainRenderTarget rtLeft; + QGles2SwapChainRenderTarget rtRight; + QGles2CommandBuffer cb; + int frameCount = 0; + QGles2SwapChainTimestamps timestamps; + int currentTimestampPairIndex = 0; +}; + +class QRhiGles2 : public QRhiImplementation +{ +public: + QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + bool ensureContext(QSurface *surface = nullptr) const; + QSurface *evaluateFallbackSurface() const; + void executeDeferredReleases(); + void trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *bufD, QGles2Buffer::Access access); + void trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Texture::Access access); + void enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cbD, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, + QGles2Buffer *bufD, + QRhiPassResourceTracker::BufferAccess access, + QRhiPassResourceTracker::BufferStage stage); + void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, + QGles2Texture *texD, + QRhiPassResourceTracker::TextureAccess access, + QRhiPassResourceTracker::TextureStage stage); + void executeCommandBuffer(QRhiCommandBuffer *cb); + void executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2GraphicsPipeline *psD); + void bindCombinedSampler(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Sampler *samplerD, + void *ps, uint psGeneration, int glslLocation, + int *texUnit, bool *activeTexUnitAltered); + void bindShaderResources(QGles2CommandBuffer *cbD, + QRhiGraphicsPipeline *maybeGraphicsPs, QRhiComputePipeline *maybeComputePs, + QRhiShaderResourceBindings *srb, + const uint *dynOfsPairs, int dynOfsCount); + QGles2RenderTargetData *enqueueBindFramebuffer(QRhiRenderTarget *rt, QGles2CommandBuffer *cbD, + bool *wantsColorClear = nullptr, bool *wantsDsClear = nullptr); + void enqueueBarriersForPass(QGles2CommandBuffer *cbD); + QByteArray shaderSource(const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); + bool compileShader(GLuint program, const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); + bool linkProgram(GLuint program); + void registerUniformIfActive(const QShaderDescription::BlockVariable &var, + const QByteArray &namePrefix, int binding, int baseOffset, + GLuint program, + QDuplicateTracker<int, 256> *activeUniformLocations, + QGles2UniformDescriptionVector *dst); + void gatherUniforms(GLuint program, const QShaderDescription::UniformBlock &ub, + QDuplicateTracker<int, 256> *activeUniformLocations, QGles2UniformDescriptionVector *dst); + void gatherSamplers(GLuint program, const QShaderDescription::InOutVariable &v, + QGles2SamplerDescriptionVector *dst); + void gatherGeneratedSamplers(GLuint program, + const QShader::SeparateToCombinedImageSamplerMapping &mapping, + QGles2SamplerDescriptionVector *dst); + void sanityCheckVertexFragmentInterface(const QShaderDescription &vsDesc, const QShaderDescription &fsDesc); + bool isProgramBinaryDiskCacheEnabled() const; + + enum ProgramCacheResult { + ProgramCacheHit, + ProgramCacheMiss, + ProgramCacheError + }; + ProgramCacheResult tryLoadFromDiskOrPipelineCache(const QRhiShaderStage *stages, + int stageCount, + GLuint program, + const QVector<QShaderDescription::InOutVariable> &inputVars, + QByteArray *cacheKey); + void trySaveToDiskCache(GLuint program, const QByteArray &cacheKey); + void trySaveToPipelineCache(GLuint program, const QByteArray &cacheKey, bool force = false); + + QRhi::Flags rhiFlags; + QOpenGLContext *ctx = nullptr; + bool importedContext = false; + QSurfaceFormat requestedFormat; + QSurface *fallbackSurface = nullptr; + QPointer<QWindow> maybeWindow = nullptr; + QOpenGLContext *maybeShareContext = nullptr; + mutable bool needsMakeCurrentDueToSwap = false; + QOpenGLExtensions *f = nullptr; + void (QOPENGLF_APIENTRYP glPolygonMode) (GLenum, GLenum) = nullptr; + void(QOPENGLF_APIENTRYP glTexImage1D)(GLenum, GLint, GLint, GLsizei, GLint, GLenum, GLenum, + const void *) = nullptr; + void(QOPENGLF_APIENTRYP glTexStorage1D)(GLenum, GLint, GLenum, GLsizei) = nullptr; + void(QOPENGLF_APIENTRYP glTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, GLenum, + const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glCopyTexSubImage1D)(GLenum, GLint, GLint, GLint, GLint, + GLsizei) = nullptr; + void(QOPENGLF_APIENTRYP glCompressedTexImage1D)(GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, + const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glCompressedTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, + GLsizei, const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glFramebufferTexture1D)(GLenum, GLenum, GLenum, GLuint, + GLint) = nullptr; + void(QOPENGLF_APIENTRYP glFramebufferTextureMultiviewOVR)(GLenum, GLenum, GLuint, GLint, + GLint, GLsizei) = nullptr; + void (QOPENGLF_APIENTRYP glQueryCounter)(GLuint, GLenum) = nullptr; + void (QOPENGLF_APIENTRYP glGetQueryObjectui64v)(GLuint, GLenum, quint64 *) = nullptr; + void (QOPENGLF_APIENTRYP glObjectLabel)(GLenum, GLuint, GLsizei, const GLchar *) = nullptr; + void (QOPENGLF_APIENTRYP glFramebufferTexture2DMultisampleEXT)(GLenum, GLenum, GLenum, GLuint, GLint, GLsizei) = nullptr; + void (QOPENGLF_APIENTRYP glFramebufferTextureMultisampleMultiviewOVR)(GLenum, GLenum, GLuint, GLint, GLsizei, GLint, GLsizei) = nullptr; + uint vao = 0; + struct Caps { + Caps() + : ctxMajor(2), + ctxMinor(0), + maxTextureSize(2048), + maxDrawBuffers(4), + maxSamples(16), + maxTextureArraySize(0), + maxThreadGroupsPerDimension(0), + maxThreadsPerThreadGroup(0), + maxThreadGroupsX(0), + maxThreadGroupsY(0), + maxThreadGroupsZ(0), + maxUniformVectors(4096), + maxVertexInputs(8), + maxVertexOutputs(8), + msaaRenderBuffer(false), + multisampledTexture(false), + npotTextureFull(true), + gles(false), + fixedIndexPrimitiveRestart(false), + bgraExternalFormat(false), + bgraInternalFormat(false), + r8Format(false), + r16Format(false), + floatFormats(false), + rgb10Formats(false), + depthTexture(false), + packedDepthStencil(false), + needsDepthStencilCombinedAttach(false), + srgbWriteControl(false), + coreProfile(false), + uniformBuffers(false), + elementIndexUint(false), + depth24(false), + rgba8Format(false), + instancing(false), + baseVertex(false), + compute(false), + textureCompareMode(false), + properMapBuffer(false), + nonBaseLevelFramebufferTexture(false), + texelFetch(false), + intAttributes(true), + screenSpaceDerivatives(false), + programBinary(false), + texture3D(false), + tessellation(false), + geometryShader(false), + texture1D(false), + hasDrawBuffersFunc(false), + halfAttributes(false), + multiView(false), + timestamps(false), + objectLabel(false) + { } + int ctxMajor; + int ctxMinor; + int maxTextureSize; + int maxDrawBuffers; + int maxSamples; + int maxTextureArraySize; + int maxThreadGroupsPerDimension; + int maxThreadsPerThreadGroup; + int maxThreadGroupsX; + int maxThreadGroupsY; + int maxThreadGroupsZ; + int maxUniformVectors; + int maxVertexInputs; + int maxVertexOutputs; + // Multisample fb and blit are supported (GLES 3.0 or OpenGL 3.x). Not + // the same as multisample textures! + uint msaaRenderBuffer : 1; + uint multisampledTexture : 1; + uint npotTextureFull : 1; + uint gles : 1; + uint fixedIndexPrimitiveRestart : 1; + uint bgraExternalFormat : 1; + uint bgraInternalFormat : 1; + uint r8Format : 1; + uint r16Format : 1; + uint floatFormats : 1; + uint rgb10Formats : 1; + uint depthTexture : 1; + uint packedDepthStencil : 1; + uint needsDepthStencilCombinedAttach : 1; + uint srgbWriteControl : 1; + uint coreProfile : 1; + uint uniformBuffers : 1; + uint elementIndexUint : 1; + uint depth24 : 1; + uint rgba8Format : 1; + uint instancing : 1; + uint baseVertex : 1; + uint compute : 1; + uint textureCompareMode : 1; + uint properMapBuffer : 1; + uint nonBaseLevelFramebufferTexture : 1; + uint texelFetch : 1; + uint intAttributes : 1; + uint screenSpaceDerivatives : 1; + uint programBinary : 1; + uint texture3D : 1; + uint tessellation : 1; + uint geometryShader : 1; + uint texture1D : 1; + uint hasDrawBuffersFunc : 1; + uint halfAttributes : 1; + uint multiView : 1; + uint timestamps : 1; + uint objectLabel : 1; + uint glesMultisampleRenderToTexture : 1; + uint glesMultiviewMultisampleRenderToTexture : 1; + } caps; + QGles2SwapChain *currentSwapChain = nullptr; + QSet<GLint> supportedCompressedFormats; + mutable QList<int> supportedSampleCountList; + QRhiGles2NativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + mutable bool contextLost = false; + + struct DeferredReleaseEntry { + enum Type { + Buffer, + Pipeline, + Texture, + RenderBuffer, + TextureRenderTarget + }; + Type type; + union { + struct { + GLuint buffer; + } buffer; + struct { + GLuint program; + } pipeline; + struct { + GLuint texture; + } texture; + struct { + GLuint renderbuffer; + GLuint renderbuffer2; + } renderbuffer; + struct { + GLuint framebuffer; + GLuint nonMsaaThrowawayDepthTexture; + } textureRenderTarget; + }; + }; + QList<DeferredReleaseEntry> releaseQueue; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } + bool active = false; + QGles2CommandBuffer cbWrapper; + GLuint tsQueries[2] = {}; + } ofr; + + QHash<QRhiShaderStage, uint> m_shaderCache; + + struct PipelineCacheData { + quint32 format; + QByteArray data; + }; + QHash<QByteArray, PipelineCacheData> m_pipelineCache; +}; + +Q_DECLARE_TYPEINFO(QRhiGles2::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhigles2_p_p.h b/src/gui/rhi/qrhigles2_p_p.h deleted file mode 100644 index 8a0e0ec2a1..0000000000 --- a/src/gui/rhi/qrhigles2_p_p.h +++ /dev/null @@ -1,1045 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIGLES2_P_H -#define QRHIGLES2_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhigles2_p.h" -#include "qrhi_p_p.h" -#include "qshaderdescription_p.h" -#include <qopengl.h> -#include <QByteArray> -#include <QSurface> - -QT_BEGIN_NAMESPACE - -class QOpenGLExtensions; - -struct QGles2Buffer : public QRhiBuffer -{ - QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size); - ~QGles2Buffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - int nonZeroSize = 0; - GLuint buffer = 0; - GLenum targetForDataOps; - QByteArray data; - enum Access { - AccessNone, - AccessVertex, - AccessIndex, - AccessUniform, - AccessStorageRead, - AccessStorageWrite, - AccessStorageReadWrite, - AccessUpdate - }; - struct UsageState { - Access access; - }; - UsageState usageState; - friend class QRhiGles2; -}; - -struct QGles2RenderBuffer : public QRhiRenderBuffer -{ - QGles2RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QGles2RenderBuffer(); - void destroy() override; - bool create() override; - bool createFrom(NativeRenderBuffer src) override; - QRhiTexture::Format backingFormat() const override; - - GLuint renderbuffer = 0; - GLuint stencilRenderbuffer = 0; // when packed depth-stencil not supported - int samples; - bool owns = true; - friend class QRhiGles2; -}; - -struct QGles2SamplerData -{ - GLenum glminfilter = 0; - GLenum glmagfilter = 0; - GLenum glwraps = 0; - GLenum glwrapt = 0; - GLenum glwrapr = 0; - GLenum gltexcomparefunc = 0; -}; - -inline bool operator==(const QGles2SamplerData &a, const QGles2SamplerData &b) -{ - return a.glminfilter == b.glminfilter - && a.glmagfilter == b.glmagfilter - && a.glwraps == b.glwraps - && a.glwrapt == b.glwrapt - && a.glwrapr == b.glwrapr - && a.gltexcomparefunc == b.gltexcomparefunc; -} - -inline bool operator!=(const QGles2SamplerData &a, const QGles2SamplerData &b) -{ - return !(a == b); -} - -struct QGles2Texture : public QRhiTexture -{ - QGles2Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags); - ~QGles2Texture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - - GLuint texture = 0; - bool owns = true; - GLenum target; - GLenum glintformat; - GLenum glsizedintformat; - GLenum glformat; - GLenum gltype; - QGles2SamplerData samplerState; - bool specified = false; - bool zeroInitialized = false; - int mipLevelCount = 0; - - enum Access { - AccessNone, - AccessSample, - AccessFramebuffer, - AccessStorageRead, - AccessStorageWrite, - AccessStorageReadWrite, - AccessUpdate, - AccessRead - }; - struct UsageState { - Access access; - }; - UsageState usageState; - - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2Sampler : public QRhiSampler -{ - QGles2Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QGles2Sampler(); - void destroy() override; - bool create() override; - - QGles2SamplerData d; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2RenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QGles2RenderPassDescriptor(QRhiImplementation *rhi); - ~QGles2RenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; -}; - -struct QGles2RenderTargetData -{ - QGles2RenderTargetData(QRhiImplementation *) { } - - QGles2RenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - bool srgbUpdateAndBlend = false; -}; - -struct QGles2ReferenceRenderTarget : public QRhiRenderTarget -{ - QGles2ReferenceRenderTarget(QRhiImplementation *rhi); - ~QGles2ReferenceRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QGles2RenderTargetData d; -}; - -struct QGles2TextureRenderTarget : public QRhiTextureRenderTarget -{ - QGles2TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QGles2TextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QGles2RenderTargetData d; - GLuint framebuffer = 0; - friend class QRhiGles2; -}; - -struct QGles2ShaderResourceBindings : public QRhiShaderResourceBindings -{ - QGles2ShaderResourceBindings(QRhiImplementation *rhi); - ~QGles2ShaderResourceBindings(); - void destroy() override; - bool create() override; - - bool hasDynamicOffset = false; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2UniformDescription -{ - QShaderDescription::VariableType type; - int glslLocation; - int binding; - uint offset; - int size; - int arrayDim; -}; - -Q_DECLARE_TYPEINFO(QGles2UniformDescription, Q_RELOCATABLE_TYPE); - -struct QGles2SamplerDescription -{ - int glslLocation; - int binding; -}; - -Q_DECLARE_TYPEINFO(QGles2SamplerDescription, Q_RELOCATABLE_TYPE); - -using QGles2UniformDescriptionVector = QVarLengthArray<QGles2UniformDescription, 8>; -using QGles2SamplerDescriptionVector = QVarLengthArray<QGles2SamplerDescription, 4>; - -struct QGles2UniformState -{ - static constexpr int MAX_TRACKED_LOCATION = 1023; - int componentCount; - float v[4]; -}; - -struct QGles2GraphicsPipeline : public QRhiGraphicsPipeline -{ - QGles2GraphicsPipeline(QRhiImplementation *rhi); - ~QGles2GraphicsPipeline(); - void destroy() override; - bool create() override; - - GLuint program = 0; - GLenum drawMode = GL_TRIANGLES; - QGles2UniformDescriptionVector uniforms; - QGles2SamplerDescriptionVector samplers; - QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; - QRhiShaderResourceBindings *currentSrb = nullptr; - uint currentSrbGeneration = 0; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2ComputePipeline : public QRhiComputePipeline -{ - QGles2ComputePipeline(QRhiImplementation *rhi); - ~QGles2ComputePipeline(); - void destroy() override; - bool create() override; - - GLuint program = 0; - QGles2UniformDescriptionVector uniforms; - QGles2SamplerDescriptionVector samplers; - QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; - QRhiShaderResourceBindings *currentSrb = nullptr; - uint currentSrbGeneration = 0; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2CommandBuffer : public QRhiCommandBuffer -{ - QGles2CommandBuffer(QRhiImplementation *rhi); - ~QGles2CommandBuffer(); - void destroy() override; - - // keep at a reasonably low value otherwise sizeof Command explodes - static const int MAX_DYNAMIC_OFFSET_COUNT = 8; - - struct Command { - enum Cmd { - BeginFrame, - EndFrame, - ResetFrame, - Viewport, - Scissor, - BlendConstants, - StencilRef, - BindVertexBuffer, - BindIndexBuffer, - Draw, - DrawIndexed, - BindGraphicsPipeline, - BindShaderResources, - BindFramebuffer, - Clear, - BufferSubData, - GetBufferSubData, - CopyTex, - ReadPixels, - SubImage, - CompressedImage, - CompressedSubImage, - BlitFromRenderbuffer, - GenMip, - BindComputePipeline, - Dispatch, - BarriersForPass, - Barrier - }; - Cmd cmd; - - // QRhi*/QGles2* references should be kept at minimum (so no - // QRhiTexture/Buffer/etc. pointers). - union Args { - struct { - float x, y, w, h; - float d0, d1; - } viewport; - struct { - int x, y, w, h; - } scissor; - struct { - float r, g, b, a; - } blendConstants; - struct { - quint32 ref; - QRhiGraphicsPipeline *ps; - } stencilRef; - struct { - QRhiGraphicsPipeline *ps; - GLuint buffer; - quint32 offset; - int binding; - } bindVertexBuffer; - struct { - GLuint buffer; - quint32 offset; - GLenum type; - } bindIndexBuffer; - struct { - QRhiGraphicsPipeline *ps; - quint32 vertexCount; - quint32 firstVertex; - quint32 instanceCount; - quint32 baseInstance; - } draw; - struct { - QRhiGraphicsPipeline *ps; - quint32 indexCount; - quint32 firstIndex; - quint32 instanceCount; - quint32 baseInstance; - qint32 baseVertex; - } drawIndexed; - struct { - QRhiGraphicsPipeline *ps; - } bindGraphicsPipeline; - struct { - QRhiGraphicsPipeline *maybeGraphicsPs; - QRhiComputePipeline *maybeComputePs; - QRhiShaderResourceBindings *srb; - int dynamicOffsetCount; - uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offset - } bindShaderResources; - struct { - GLbitfield mask; - float c[4]; - float d; - quint32 s; - } clear; - struct { - GLuint fbo; - bool srgb; - int colorAttCount; - } bindFramebuffer; - struct { - GLenum target; - GLuint buffer; - int offset; - int size; - const void *data; // must come from retainData() - } bufferSubData; - struct { - QRhiBufferReadbackResult *result; - GLenum target; - GLuint buffer; - int offset; - int size; - } getBufferSubData; - struct { - GLenum srcTarget; - GLenum srcFaceTarget; - GLuint srcTexture; - int srcLevel; - int srcX; - int srcY; - int srcZ; - GLenum dstTarget; - GLuint dstTexture; - GLenum dstFaceTarget; - int dstLevel; - int dstX; - int dstY; - int dstZ; - int w; - int h; - } copyTex; - struct { - QRhiReadbackResult *result; - GLuint texture; - int w; - int h; - QRhiTexture::Format format; - GLenum readTarget; - int level; - int slice3D; - } readPixels; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - int dx; - int dy; - int dz; - int w; - int h; - GLenum glformat; - GLenum gltype; - int rowStartAlign; - int rowLength; - const void *data; // must come from retainImage() - } subImage; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - GLenum glintformat; - int w; - int h; - int depth; - int size; - const void *data; // must come from retainData() - } compressedImage; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - int dx; - int dy; - int dz; - int w; - int h; - GLenum glintformat; - int size; - const void *data; // must come from retainData() - } compressedSubImage; - struct { - GLuint renderbuffer; - int w; - int h; - GLenum target; - GLuint texture; - int dstLevel; - } blitFromRb; - struct { - GLenum target; - GLuint texture; - } genMip; - struct { - QRhiComputePipeline *ps; - } bindComputePipeline; - struct { - GLuint x; - GLuint y; - GLuint z; - } dispatch; - struct { - int trackerIndex; - } barriersForPass; - struct { - GLbitfield barriers; - } barrier; - } args; - }; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - QRhiBackendCommandList<Command> commands; - QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; - int currentPassResTrackerIndex; - - PassType recordingPass; - bool passNeedsResourceTracking; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - - struct GraphicsPassState { - bool valid = false; - bool scissor; - bool cullFace; - GLenum cullMode; - GLenum frontFace; - bool blendEnabled; - struct ColorMask { bool r, g, b, a; } colorMask; - struct Blend { - GLenum srcColor; - GLenum dstColor; - GLenum srcAlpha; - GLenum dstAlpha; - GLenum opColor; - GLenum opAlpha; - } blend; - bool depthTest; - bool depthWrite; - GLenum depthFunc; - bool stencilTest; - GLuint stencilReadMask; - GLuint stencilWriteMask; - struct StencilFace { - GLenum func; - GLenum failOp; - GLenum zfailOp; - GLenum zpassOp; - } stencil[2]; // front, back - bool polyOffsetFill; - float polyOffsetFactor; - float polyOffsetUnits; - float lineWidth; - void reset() { valid = false; } - struct { - // not part of QRhiGraphicsPipeline but used by setGraphicsPipeline() - GLint stencilRef = 0; - } dynamic; - } graphicsPassState; - - struct ComputePassState { - enum Access { - Read = 0x01, - Write = 0x02 - }; - QHash<QRhiResource *, QPair<int, bool> > writtenResources; - void reset() { - writtenResources.clear(); - } - } computePassState; - - struct TextureUnitState { - void *ps; - uint psGeneration; - uint texture; - } textureUnitState[16]; - - QVarLengthArray<QByteArray, 4> dataRetainPool; - QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; - QVarLengthArray<QImage, 4> imageRetainPool; - - // relies heavily on implicit sharing (no copies of the actual data will be made) - const void *retainData(const QByteArray &data) { - dataRetainPool.append(data); - return dataRetainPool.last().constData(); - } - const uchar *retainBufferData(const QRhiBufferData &data) { - bufferDataRetainPool.append(data); - return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); - } - const void *retainImage(const QImage &image) { - imageRetainPool.append(image); - return imageRetainPool.last().constBits(); - } - void resetCommands() { - commands.reset(); - dataRetainPool.clear(); - bufferDataRetainPool.clear(); - imageRetainPool.clear(); - - passResTrackers.clear(); - currentPassResTrackerIndex = -1; - } - void resetState() { - recordingPass = NoPass; - passNeedsResourceTracking = true; - currentTarget = nullptr; - resetCommands(); - resetCachedState(); - } - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - graphicsPassState.reset(); - computePassState.reset(); - memset(textureUnitState, 0, sizeof(textureUnitState)); - } -}; - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, - const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) -{ - return a.func == b.func - && a.failOp == b.failOp - && a.zfailOp == b.zfailOp - && a.zpassOp == b.zpassOp; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, - const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) -{ - return !(a == b); -} - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, - const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) -{ - return a.r == b.r && a.g == b.g && a.b == b.b && a.a == b.a; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, - const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) -{ - return !(a == b); -} - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::Blend &a, - const QGles2CommandBuffer::GraphicsPassState::Blend &b) -{ - return a.srcColor == b.srcColor - && a.dstColor == b.dstColor - && a.srcAlpha == b.srcAlpha - && a.dstAlpha == b.dstAlpha - && a.opColor == b.opColor - && a.opAlpha == b.opAlpha; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::Blend &a, - const QGles2CommandBuffer::GraphicsPassState::Blend &b) -{ - return !(a == b); -} - -struct QGles2SwapChain : public QRhiSwapChain -{ - QGles2SwapChain(QRhiImplementation *rhi); - ~QGles2SwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - QSurface *surface = nullptr; - QSize pixelSize; - QGles2ReferenceRenderTarget rt; - QGles2CommandBuffer cb; - int frameCount = 0; -}; - -class QRhiGles2 : public QRhiImplementation -{ -public: - QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *importDevice = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - void sendVMemStatsToProfiler() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - bool ensureContext(QSurface *surface = nullptr) const; - void executeDeferredReleases(); - void trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *bufD, QGles2Buffer::Access access); - void trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Texture::Access access); - void enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cbD, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, - QGles2Buffer *bufD, - QRhiPassResourceTracker::BufferAccess access, - QRhiPassResourceTracker::BufferStage stage); - void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, - QGles2Texture *texD, - QRhiPassResourceTracker::TextureAccess access, - QRhiPassResourceTracker::TextureStage stage); - void executeCommandBuffer(QRhiCommandBuffer *cb); - void executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2GraphicsPipeline *psD); - void bindShaderResources(QGles2CommandBuffer *cbD, - QRhiGraphicsPipeline *maybeGraphicsPs, QRhiComputePipeline *maybeComputePs, - QRhiShaderResourceBindings *srb, - const uint *dynOfsPairs, int dynOfsCount); - QGles2RenderTargetData *enqueueBindFramebuffer(QRhiRenderTarget *rt, QGles2CommandBuffer *cbD, - bool *wantsColorClear = nullptr, bool *wantsDsClear = nullptr); - void enqueueBarriersForPass(QGles2CommandBuffer *cbD); - int effectiveSampleCount(int sampleCount) const; - QByteArray shaderSource(const QRhiShaderStage &shaderStage, int *glslVersion); - bool compileShader(GLuint program, const QRhiShaderStage &shaderStage, int *glslVersion); - bool linkProgram(GLuint program); - void registerUniformIfActive(const QShaderDescription::BlockVariable &var, - const QByteArray &namePrefix, int binding, int baseOffset, - GLuint program, - QSet<int> *activeUniformLocations, - QGles2UniformDescriptionVector *dst); - void gatherUniforms(GLuint program, const QShaderDescription::UniformBlock &ub, - QSet<int> *activeUniformLocations, QGles2UniformDescriptionVector *dst); - void gatherSamplers(GLuint program, const QShaderDescription::InOutVariable &v, - QGles2SamplerDescriptionVector *dst); - void sanityCheckVertexFragmentInterface(const QShaderDescription &vsDesc, const QShaderDescription &fsDesc); - bool isProgramBinaryDiskCacheEnabled() const; - - enum ProgramCacheResult { - ProgramCacheHit, - ProgramCacheMiss, - ProgramCacheError - }; - ProgramCacheResult tryLoadFromDiskOrPipelineCache(const QRhiShaderStage *stages, - int stageCount, - GLuint program, - const QVector<QShaderDescription::InOutVariable> &inputVars, - QByteArray *cacheKey); - void trySaveToDiskCache(GLuint program, const QByteArray &cacheKey); - void trySaveToPipelineCache(GLuint program, const QByteArray &cacheKey, bool force = false); - - QRhi::Flags rhiFlags; - QOpenGLContext *ctx = nullptr; - bool importedContext = false; - QSurfaceFormat requestedFormat; - QSurface *fallbackSurface = nullptr; - QWindow *maybeWindow = nullptr; - mutable bool needsMakeCurrent = false; - QOpenGLExtensions *f = nullptr; - uint vao = 0; - struct Caps { - Caps() - : ctxMajor(2), - ctxMinor(0), - maxTextureSize(2048), - maxDrawBuffers(4), - maxSamples(16), - maxThreadGroupsPerDimension(0), - maxThreadsPerThreadGroup(0), - maxThreadGroupsX(0), - maxThreadGroupsY(0), - maxThreadGroupsZ(0), - msaaRenderBuffer(false), - multisampledTexture(false), - npotTextureFull(true), - gles(false), - fixedIndexPrimitiveRestart(false), - bgraExternalFormat(false), - bgraInternalFormat(false), - r8Format(false), - r16Format(false), - floatFormats(false), - depthTexture(false), - packedDepthStencil(false), - needsDepthStencilCombinedAttach(false), - srgbCapableDefaultFramebuffer(false), - coreProfile(false), - uniformBuffers(false), - elementIndexUint(false), - depth24(false), - rgba8Format(false), - instancing(false), - baseVertex(false), - compute(false), - textureCompareMode(false), - properMapBuffer(false), - nonBaseLevelFramebufferTexture(false), - texelFetch(false), - intAttributes(true), - screenSpaceDerivatives(false), - programBinary(false), - texture3D(false) - { } - int ctxMajor; - int ctxMinor; - int maxTextureSize; - int maxDrawBuffers; - int maxSamples; - int maxThreadGroupsPerDimension; - int maxThreadsPerThreadGroup; - int maxThreadGroupsX; - int maxThreadGroupsY; - int maxThreadGroupsZ; - // Multisample fb and blit are supported (GLES 3.0 or OpenGL 3.x). Not - // the same as multisample textures! - uint msaaRenderBuffer : 1; - uint multisampledTexture : 1; - uint npotTextureFull : 1; - uint gles : 1; - uint fixedIndexPrimitiveRestart : 1; - uint bgraExternalFormat : 1; - uint bgraInternalFormat : 1; - uint r8Format : 1; - uint r16Format : 1; - uint floatFormats : 1; - uint depthTexture : 1; - uint packedDepthStencil : 1; - uint needsDepthStencilCombinedAttach : 1; - uint srgbCapableDefaultFramebuffer : 1; - uint coreProfile : 1; - uint uniformBuffers : 1; - uint elementIndexUint : 1; - uint depth24 : 1; - uint rgba8Format : 1; - uint instancing : 1; - uint baseVertex : 1; - uint compute : 1; - uint textureCompareMode : 1; - uint properMapBuffer : 1; - uint nonBaseLevelFramebufferTexture : 1; - uint texelFetch : 1; - uint intAttributes : 1; - uint screenSpaceDerivatives : 1; - uint programBinary : 1; - uint texture3D : 1; - } caps; - QGles2SwapChain *currentSwapChain = nullptr; - QList<GLint> supportedCompressedFormats; - mutable QList<int> supportedSampleCountList; - QRhiGles2NativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - mutable bool contextLost = false; - - struct DeferredReleaseEntry { - enum Type { - Buffer, - Pipeline, - Texture, - RenderBuffer, - TextureRenderTarget - }; - Type type; - union { - struct { - GLuint buffer; - } buffer; - struct { - GLuint program; - } pipeline; - struct { - GLuint texture; - } texture; - struct { - GLuint renderbuffer; - GLuint renderbuffer2; - } renderbuffer; - struct { - GLuint framebuffer; - } textureRenderTarget; - }; - }; - QList<DeferredReleaseEntry> releaseQueue; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } - bool active = false; - QGles2CommandBuffer cbWrapper; - } ofr; - - QHash<QRhiShaderStage, uint> m_shaderCache; - - struct PipelineCacheData { - quint32 format; - QByteArray data; - }; - QHash<QByteArray, PipelineCacheData> m_pipelineCache; -}; - -Q_DECLARE_TYPEINFO(QRhiGles2::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhimetal.mm b/src/gui/rhi/qrhimetal.mm index cf11f12ead..b99afc596c 100644 --- a/src/gui/rhi/qrhimetal.mm +++ b/src/gui/rhi/qrhimetal.mm @@ -1,46 +1,18 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhimetal_p_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhimetal_p.h" +#include "qshader_p.h" #include <QGuiApplication> #include <QWindow> +#include <QUrl> +#include <QFile> +#include <QTemporaryFile> +#include <QFileInfo> #include <qmath.h> +#include <QOperatingSystemVersion> + +#include <QtCore/private/qcore_mac_p.h> #ifdef Q_OS_MACOS #include <AppKit/AppKit.h> @@ -68,19 +40,28 @@ QT_BEGIN_NAMESPACE #error ARC not supported #endif -// Note: we expect everything here pass the Metal API validation when running -// in Debug mode in XCode. Some of the issues that break validation are not -// obvious and not visible when running outside XCode. -// -// An exception is the nextDrawable Called Early blah blah warning, which is -// plain and simply false. +// Even though the macOS 13 MTLBinaryArchive problem (QTBUG-106703) seems +// to be solved in later 13.x releases, we have reports from old Intel hardware +// and older macOS versions where this causes problems (QTBUG-114338). +// Thus we no longer do OS version based differentiation, but rather have a +// single toggle that is currently on, and so QRhi::(set)pipelineCache() +// does nothing with Metal. +#define QRHI_METAL_DISABLE_BINARY_ARCHIVE + +// We should be able to operate with command buffers that do not automatically +// retain/release the resources used by them. (since we have logic that mirrors +// other backends such as the Vulkan one anyway) +#define QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES /*! \class QRhiMetalInitParams \inmodule QtRhi - \internal + \since 6.6 \brief Metal specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Metal-based QRhi needs no special parameters for initialization. \badcode @@ -88,10 +69,13 @@ QT_BEGIN_NAMESPACE rhi = QRhi::create(QRhi::Metal, ¶ms); \endcode - \note Metal API validation cannot be enabled by the application. Instead, - run the debug build of the application in XCode. Generating a - \c{.xcodeproj} file via \c{qmake -spec macx-xcode} provides a convenient - way to enable this. + \note Metal API validation cannot be enabled programmatically by the QRhi. + Instead, either run the debug build of the application in XCode, by + generating a \c{.xcodeproj} file via \c{cmake -G Xcode}, or set the + environment variable \c{METAL_DEVICE_WRAPPER_TYPE=1}. The variable needs to + be set early on in the environment, perferably before starting the process; + attempting to set it at QRhi creation time is not functional in practice. + (too late probably) \note QRhiSwapChain can only target QWindow instances that have their surface type set to QSurface::MetalSurface. @@ -110,14 +94,30 @@ QT_BEGIN_NAMESPACE /*! \class QRhiMetalNativeHandles \inmodule QtRhi - \internal + \since 6.6 \brief Holds the Metal device used by the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiMetalNativeHandles::dev + + Set to a valid MTLDevice to import an existing device. +*/ + +/*! + \variable QRhiMetalNativeHandles::cmdQueue + + Set to a valid MTLCommandQueue when importing an existing command queue. + When \nullptr, QRhi will create a new command queue. +*/ + +/*! \class QRhiMetalCommandBufferNativeHandles \inmodule QtRhi - \internal + \since 6.6 \brief Holds the MTLCommandBuffer and MTLRenderCommandEncoder objects that are backing a QRhiCommandBuffer. \note The command buffer object is only guaranteed to be valid while @@ -129,14 +129,28 @@ QT_BEGIN_NAMESPACE \note The command encoder is only valid while recording a pass, that is, between \l{QRhiCommandBuffer::beginPass()} - \l{QRhiCommandBuffer::endPass()}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiMetalCommandBufferNativeHandles::commandBuffer +*/ + +/*! + \variable QRhiMetalCommandBufferNativeHandles::encoder +*/ + struct QMetalShader { id<MTLLibrary> lib = nil; id<MTLFunction> func = nil; - std::array<uint, 3> localSize; + std::array<uint, 3> localSize = {}; + uint outputVertexCount = 0; + QShaderDescription desc; QShader::NativeResourceBindingMap nativeResourceBindingMap; + QShader::NativeShaderInfo nativeShaderInfo; void destroy() { nativeResourceBindingMap.clear(); @@ -149,11 +163,14 @@ struct QMetalShader struct QRhiMetalData { - QRhiMetalData(QRhiImplementation *rhi) : ofr(rhi) { } + QRhiMetalData(QRhiMetal *rhi) : q(rhi), ofr(rhi) { } + QRhiMetal *q; id<MTLDevice> dev = nil; id<MTLCommandQueue> cmdQueue = nil; + API_AVAILABLE(macosx(11.0), ios(14.0)) id<MTLBinaryArchive> binArch = nil; + id<MTLCommandBuffer> newCommandBuffer(); MTLRenderPassDescriptor *createDefaultRenderPass(bool hasDepthStencil, const QColor &colorClearValue, const QRhiDepthStencilClearValue &depthStencilClearValue, @@ -161,6 +178,11 @@ struct QRhiMetalData id<MTLLibrary> createMetalLib(const QShader &shader, QShader::Variant shaderVariant, QString *error, QByteArray *entryPoint, QShaderKey *activeKey); id<MTLFunction> createMSLShaderFunction(id<MTLLibrary> lib, const QByteArray &entryPoint); + bool setupBinaryArchive(NSURL *sourceFileUrl = nil); + void addRenderPipelineToBinaryArchive(MTLRenderPipelineDescriptor *rpDesc); + void trySeedingRenderPipelineFromBinaryArchive(MTLRenderPipelineDescriptor *rpDesc); + void addComputePipelineToBinaryArchive(MTLComputePipelineDescriptor *cpDesc); + void trySeedingComputePipelineFromBinaryArchive(MTLComputePipelineDescriptor *cpDesc); struct DeferredReleaseEntry { enum Type { @@ -168,7 +190,9 @@ struct QRhiMetalData RenderBuffer, Texture, Sampler, - StagingBuffer + StagingBuffer, + GraphicsPipeline, + ComputePipeline }; Type type; int lastActiveFrameSlot; // -1 if not used otherwise 0..FRAMES_IN_FLIGHT-1 @@ -190,6 +214,15 @@ struct QRhiMetalData struct { id<MTLBuffer> buffer; } stagingBuffer; + struct { + id<MTLRenderPipelineState> pipelineState; + id<MTLDepthStencilState> depthStencilState; + std::array<id<MTLComputePipelineState>, 3> tessVertexComputeState; + id<MTLComputePipelineState> tessTessControlComputeState; + } graphicsPipeline; + struct { + id<MTLComputePipelineState> pipelineState; + } computePipeline; }; }; QVector<DeferredReleaseEntry> releaseQueue; @@ -197,6 +230,7 @@ struct QRhiMetalData struct OffscreenFrame { OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } bool active = false; + double lastGpuTime = 0; QMetalCommandBuffer cbWrapper; } ofr; @@ -211,6 +245,17 @@ struct QRhiMetalData }; QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; + struct BufferReadback + { + int activeFrameSlot = -1; + QRhiReadbackResult *result; + quint32 offset; + quint32 readSize; + id<MTLBuffer> buf; + }; + + QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; + MTLCaptureManager *captureMgr; id<MTLCaptureScope> captureScope = nil; @@ -228,7 +273,7 @@ struct QMetalBufferData bool slotted; id<MTLBuffer> buf[QMTL_FRAMES_IN_FLIGHT]; struct BufferUpdate { - int offset; + quint32 offset; QRhiBufferData data; }; QVarLengthArray<BufferUpdate, 16> pendingUpdates[QMTL_FRAMES_IN_FLIGHT]; @@ -259,15 +304,45 @@ struct QMetalSamplerData id<MTLSamplerState> samplerState = nil; }; +struct QMetalShaderResourceBindingsData { + struct Stage { + struct Buffer { + int nativeBinding; + id<MTLBuffer> mtlbuf; + quint32 offset; + }; + struct Texture { + int nativeBinding; + id<MTLTexture> mtltex; + }; + struct Sampler { + int nativeBinding; + id<MTLSamplerState> mtlsampler; + }; + QVarLengthArray<Buffer, 8> buffers; + QVarLengthArray<Texture, 8> textures; + QVarLengthArray<Sampler, 8> samplers; + QRhiBatchedBindings<id<MTLBuffer> > bufferBatches; + QRhiBatchedBindings<NSUInteger> bufferOffsetBatches; + QRhiBatchedBindings<id<MTLTexture> > textureBatches; + QRhiBatchedBindings<id<MTLSamplerState> > samplerBatches; + } res[QRhiMetal::SUPPORTED_STAGES]; + enum { VERTEX = 0, FRAGMENT = 1, COMPUTE = 2, TESSCTRL = 3, TESSEVAL = 4 }; +}; + struct QMetalCommandBufferData { id<MTLCommandBuffer> cb; + double lastGpuTime = 0; id<MTLRenderCommandEncoder> currentRenderPassEncoder; id<MTLComputeCommandEncoder> currentComputePassEncoder; + id<MTLComputeCommandEncoder> tessellationComputeEncoder; MTLRenderPassDescriptor *currentPassRpDesc; int currentFirstVertexBinding; QRhiBatchedBindings<id<MTLBuffer> > currentVertexInputsBuffers; QRhiBatchedBindings<NSUInteger> currentVertexInputOffsets; + id<MTLDepthStencilState> currentDepthStencilState; + QMetalShaderResourceBindingsData currentShaderResourceBindingState; }; struct QMetalRenderTargetData @@ -293,22 +368,77 @@ struct QMetalRenderTargetData struct { ColorAtt colorAtt[QMetalRenderPassDescriptor::MAX_COLOR_ATTACHMENTS]; id<MTLTexture> dsTex = nil; + id<MTLTexture> dsResolveTex = nil; bool hasStencil = false; bool depthNeedsStore = false; + bool preserveColor = false; + bool preserveDs = false; } fb; + + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; }; struct QMetalGraphicsPipelineData { + QMetalGraphicsPipeline *q = nullptr; id<MTLRenderPipelineState> ps = nil; id<MTLDepthStencilState> ds = nil; MTLPrimitiveType primitiveType; MTLWinding winding; MTLCullMode cullMode; + MTLTriangleFillMode triangleFillMode; float depthBias; float slopeScaledDepthBias; QMetalShader vs; QMetalShader fs; + struct ExtraBufferManager { + enum class WorkBufType { + DeviceLocal, + HostVisible + }; + QMetalBuffer *acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type = WorkBufType::DeviceLocal); + QVector<QMetalBuffer *> deviceLocalWorkBuffers; + QVector<QMetalBuffer *> hostVisibleWorkBuffers; + } extraBufMgr; + struct Tessellation { + QMetalGraphicsPipelineData *q = nullptr; + bool enabled = false; + bool failed = false; + uint inControlPointCount; + uint outControlPointCount; + QMetalShader compVs[3]; + std::array<id<MTLComputePipelineState>, 3> vertexComputeState = {}; + id<MTLComputePipelineState> tessControlComputeState = nil; + QMetalShader compTesc; + QMetalShader vertTese; + quint32 vsCompOutputBufferSize(quint32 vertexOrIndexCount, quint32 instanceCount) const + { + // max vertex output components = resourceLimit(MaxVertexOutputs) * 4 = 60 + return vertexOrIndexCount * instanceCount * sizeof(float) * 60; + } + quint32 tescCompOutputBufferSize(quint32 patchCount) const + { + return outControlPointCount * patchCount * sizeof(float) * 60; + } + quint32 tescCompPatchOutputBufferSize(quint32 patchCount) const + { + // assume maxTessellationControlPerPatchOutputComponents is 128 + return patchCount * sizeof(float) * 128; + } + quint32 patchCountForDrawCall(quint32 vertexOrIndexCount, quint32 instanceCount) const + { + return ((vertexOrIndexCount + inControlPointCount - 1) / inControlPointCount) * instanceCount; + } + static int vsCompVariantToIndex(QShader::Variant vertexCompVariant); + id<MTLComputePipelineState> vsCompPipeline(QRhiMetal *rhiD, QShader::Variant vertexCompVariant); + id<MTLComputePipelineState> tescCompPipeline(QRhiMetal *rhiD); + id<MTLRenderPipelineState> teseFragRenderPipeline(QRhiMetal *rhiD, QMetalGraphicsPipeline *pipeline); + } tess; + void setupVertexInputDescriptor(MTLVertexDescriptor *desc); + void setupStageInputDescriptor(MTLStageInputOutputDescriptor *desc); + + // SPIRV-Cross buffer size buffers + QMetalBuffer *bufferSizeBuffer = nullptr; }; struct QMetalComputePipelineData @@ -316,6 +446,9 @@ struct QMetalComputePipelineData id<MTLComputePipelineState> ps = nil; QMetalShader cs; MTLSize localSize; + + // SPIRV-Cross buffer size buffers + QMetalBuffer *bufferSizeBuffer = nullptr; }; struct QMetalSwapChainData @@ -323,10 +456,16 @@ struct QMetalSwapChainData CAMetalLayer *layer = nullptr; id<CAMetalDrawable> curDrawable = nil; dispatch_semaphore_t sem[QMTL_FRAMES_IN_FLIGHT]; + double lastGpuTime[QMTL_FRAMES_IN_FLIGHT]; MTLRenderPassDescriptor *rp = nullptr; id<MTLTexture> msaaTex[QMTL_FRAMES_IN_FLIGHT]; QRhiTexture::Format rhiColorFormat; MTLPixelFormat colorFormat; +#ifdef Q_OS_MACOS + bool liveResizeObserverSet = false; + QMacNotificationObserver liveResizeStartObserver; + QMacNotificationObserver liveResizeEndObserver; +#endif }; QRhiMetal::QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice) @@ -337,7 +476,7 @@ QRhiMetal::QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *import importedDevice = importDevice != nullptr; if (importedDevice) { - if (d->dev) { + if (importDevice->dev) { d->dev = (id<MTLDevice>) importDevice->dev; importedCmdQueue = importDevice->cmdQueue != nullptr; if (importedCmdQueue) @@ -360,9 +499,55 @@ inline Int aligned(Int v, Int byteAlign) return (v + byteAlign - 1) & ~(byteAlign - 1); } +bool QRhiMetal::probe(QRhiMetalInitParams *params) +{ + Q_UNUSED(params); + id<MTLDevice> dev = MTLCreateSystemDefaultDevice(); + if (dev) { + [dev release]; + return true; + } + return false; +} + +id<MTLCommandBuffer> QRhiMetalData::newCommandBuffer() +{ +#ifdef QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES + // Do not let the command buffer mess with the refcount of objects. We do + // have a proper render loop and will manage lifetimes similarly to other + // backends (Vulkan). + return [cmdQueue commandBufferWithUnretainedReferences]; +#else + return [cmdQueue commandBuffer]; +#endif +} + +bool QRhiMetalData::setupBinaryArchive(NSURL *sourceFileUrl) +{ +#ifdef QRHI_METAL_DISABLE_BINARY_ARCHIVE + return false; +#endif + + if (@available(macOS 11.0, iOS 14.0, *)) { + [binArch release]; + MTLBinaryArchiveDescriptor *binArchDesc = [MTLBinaryArchiveDescriptor new]; + binArchDesc.url = sourceFileUrl; + NSError *err = nil; + binArch = [dev newBinaryArchiveWithDescriptor: binArchDesc error: &err]; + [binArchDesc release]; + if (!binArch) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("newBinaryArchiveWithDescriptor failed: %s", qPrintable(msg)); + return false; + } + return true; + } + return false; +} + bool QRhiMetal::create(QRhi::Flags flags) { - Q_UNUSED(flags); + rhiFlags = flags; if (importedDevice) [d->dev retain]; @@ -377,10 +562,12 @@ bool QRhiMetal::create(QRhi::Flags flags) const QString deviceName = QString::fromNSString([d->dev name]); qCDebug(QRHI_LOG_INFO, "Metal device: %s", qPrintable(deviceName)); driverInfoStruct.deviceName = deviceName.toUtf8(); - driverInfoStruct.deviceId = [d->dev registryID]; -#ifdef Q_OS_IOS - driverInfoStruct.deviceType = QRhiDriverInfo::IntegratedDevice; -#else + + // deviceId and vendorId stay unset for now. Note that registryID is not + // suitable as deviceId because it does not seem stable on macOS and can + // apparently change when the system is rebooted. + +#ifdef Q_OS_MACOS if (@available(macOS 10.15, *)) { const MTLDeviceLocation deviceLocation = [d->dev location]; switch (deviceLocation) { @@ -397,8 +584,14 @@ bool QRhiMetal::create(QRhi::Flags flags) break; } } +#else + driverInfoStruct.deviceType = QRhiDriverInfo::IntegratedDevice; #endif + const QOperatingSystemVersion ver = QOperatingSystemVersion::current(); + osMajor = ver.majorVersion(); + osMinor = ver.minorVersion(); + if (importedCmdQueue) [d->cmdQueue retain]; else @@ -414,11 +607,18 @@ bool QRhiMetal::create(QRhi::Flags flags) #if defined(Q_OS_MACOS) caps.maxTextureSize = 16384; + caps.baseVertexAndInstance = true; + if (@available(macOS 10.15, *)) + caps.isAppleGPU = [d->dev supportsFamily:MTLGPUFamilyApple7]; + caps.maxThreadGroupSize = 1024; + caps.multiView = true; #elif defined(Q_OS_TVOS) if ([d->dev supportsFeatureSet: MTLFeatureSet(30003)]) // MTLFeatureSet_tvOS_GPUFamily2_v1 caps.maxTextureSize = 16384; else caps.maxTextureSize = 8192; + caps.baseVertexAndInstance = false; + caps.isAppleGPU = true; #elif defined(Q_OS_IOS) // welcome to feature set hell if ([d->dev supportsFeatureSet: MTLFeatureSet(16)] // MTLFeatureSet_iOS_GPUFamily5_v1 @@ -426,15 +626,34 @@ bool QRhiMetal::create(QRhi::Flags flags) || [d->dev supportsFeatureSet: MTLFeatureSet(4)]) // MTLFeatureSet_iOS_GPUFamily3_v1 { caps.maxTextureSize = 16384; + caps.baseVertexAndInstance = true; } else if ([d->dev supportsFeatureSet: MTLFeatureSet(3)] // MTLFeatureSet_iOS_GPUFamily2_v2 || [d->dev supportsFeatureSet: MTLFeatureSet(2)]) // MTLFeatureSet_iOS_GPUFamily1_v2 { caps.maxTextureSize = 8192; + caps.baseVertexAndInstance = false; } else { caps.maxTextureSize = 4096; + caps.baseVertexAndInstance = false; + } + caps.isAppleGPU = true; + if (@available(iOS 13, *)) { + if ([d->dev supportsFamily: MTLGPUFamilyApple4]) + caps.maxThreadGroupSize = 1024; + if ([d->dev supportsFamily: MTLGPUFamilyApple5]) + caps.multiView = true; } #endif + caps.supportedSampleCounts = { 1 }; + for (int sampleCount : { 2, 4, 8 }) { + if ([d->dev supportsTextureSampleCount: sampleCount]) + caps.supportedSampleCounts.append(sampleCount); + } + + if (rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + d->setupBinaryArchive(); + nativeHandlesStruct.dev = (MTLDevice *) d->dev; nativeHandlesStruct.cmdQueue = (MTLCommandQueue *) d->cmdQueue; @@ -453,6 +672,11 @@ void QRhiMetal::destroy() [d->captureScope release]; d->captureScope = nil; + if (@available(macOS 11.0, iOS 14.0, *)) { + [d->binArch release]; + d->binArch = nil; + } + [d->cmdQueue release]; if (!importedCmdQueue) d->cmdQueue = nil; @@ -464,18 +688,7 @@ void QRhiMetal::destroy() QVector<int> QRhiMetal::supportedSampleCounts() const { - return { 1, 2, 4, 8 }; -} - -int QRhiMetal::effectiveSampleCount(int sampleCount) const -{ - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - const int s = qBound(1, sampleCount, 64); - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return 1; - } - return s; + return caps.supportedSampleCounts; } QRhiSwapChain *QRhiMetal::createSwapChain() @@ -483,7 +696,7 @@ QRhiSwapChain *QRhiMetal::createSwapChain() return new QMetalSwapChain(this); } -QRhiBuffer *QRhiMetal::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, int size) +QRhiBuffer *QRhiMetal::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) { return new QMetalBuffer(this, type, usage, size); } @@ -526,16 +739,32 @@ bool QRhiMetal::isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture { Q_UNUSED(flags); + bool supportsFamilyMac2 = false; // needed for BC* formats + bool supportsFamilyApple3 = false; + #ifdef Q_OS_MACOS - if (format >= QRhiTexture::ETC2_RGB8 && format <= QRhiTexture::ETC2_RGBA8) - return false; - if (format >= QRhiTexture::ASTC_4x4 && format <= QRhiTexture::ASTC_12x12) - return false; + supportsFamilyMac2 = true; + if (caps.isAppleGPU) + supportsFamilyApple3 = true; #else - if (format >= QRhiTexture::BC1 && format <= QRhiTexture::BC7) - return false; + supportsFamilyApple3 = true; #endif + // BC5 is not available for any Apple hardare + if (format == QRhiTexture::BC5) + return false; + + if (!supportsFamilyApple3) { + if (format >= QRhiTexture::ETC2_RGB8 && format <= QRhiTexture::ETC2_RGBA8) + return false; + if (format >= QRhiTexture::ASTC_4x4 && format <= QRhiTexture::ASTC_12x12) + return false; + } + + if (!supportsFamilyMac2) + if (format >= QRhiTexture::BC1 && format <= QRhiTexture::BC7) + return false; + return true; } @@ -549,7 +778,7 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const case QRhi::DebugMarkers: return true; case QRhi::Timestamps: - return false; + return true; case QRhi::Instancing: return true; case QRhi::CustomInstanceStepRate: @@ -573,9 +802,9 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const case QRhi::VertexShaderPointSize: return true; case QRhi::BaseVertex: - return true; + return caps.baseVertexAndInstance; case QRhi::BaseInstance: - return true; + return caps.baseVertexAndInstance; case QRhi::TriangleFanTopology: return false; case QRhi::ReadBackNonUniformBuffer: @@ -593,7 +822,12 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const case QRhi::ReadBackAnyTextureFormat: return true; case QRhi::PipelineCacheDataLoadSave: - return false; + { + if (@available(macOS 11.0, iOS 14.0, *)) + return true; + else + return false; + } case QRhi::ImageDataStride: return true; case QRhi::RenderBufferImport: @@ -602,6 +836,32 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::RenderTo3DTextureSlice: return true; + case QRhi::TextureArrays: + return true; + case QRhi::Tessellation: + return true; + case QRhi::GeometryShader: + return false; + case QRhi::TextureArrayRange: + return false; + case QRhi::NonFillPolygonMode: + return true; + case QRhi::OneDimensionalTextures: + return true; + case QRhi::OneDimensionalTextureMipmaps: + return false; + case QRhi::HalfAttributes: + return true; + case QRhi::RenderToOneDimensionalTexture: + return false; + case QRhi::ThreeDimensionalTextureMipmaps: + return true; + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return false; + case QRhi::ResolveDepthStencil: + return true; default: Q_UNREACHABLE(); return false; @@ -630,11 +890,15 @@ int QRhiMetal::resourceLimit(QRhi::ResourceLimit limit) const case QRhi::MaxThreadGroupY: Q_FALLTHROUGH(); case QRhi::MaxThreadGroupZ: -#if defined(Q_OS_MACOS) - return 1024; -#else - return 512; -#endif + return caps.maxThreadGroupSize; + case QRhi::TextureArraySizeMax: + return 2048; + case QRhi::MaxUniformBufferRange: + return 65536; + case QRhi::MaxVertexInputs: + return 31; + case QRhi::MaxVertexOutputs: + return 15; // use the minimum from MTLGPUFamily1/2/3 default: Q_UNREACHABLE(); return 0; @@ -651,9 +915,11 @@ QRhiDriverInfo QRhiMetal::driverInfo() const return driverInfoStruct; } -void QRhiMetal::sendVMemStatsToProfiler() +QRhiStats QRhiMetal::statistics() { - // nothing to do here + QRhiStats result; + result.totalPipelineCreationTime = totalPipelineCreationTime(); + return result; } bool QRhiMetal::makeThreadLocalNativeContextCurrent() @@ -675,14 +941,133 @@ bool QRhiMetal::isDeviceLost() const return false; } +struct QMetalPipelineCacheDataHeader +{ + quint32 rhiId; + quint32 arch; + quint32 dataSize; + quint32 osMajor; + quint32 osMinor; + char driver[236]; +}; + QByteArray QRhiMetal::pipelineCacheData() { - return QByteArray(); + Q_STATIC_ASSERT(sizeof(QMetalPipelineCacheDataHeader) == 256); + QByteArray data; + if (@available(macOS 11.0, iOS 14.0, *)) { + if (!d->binArch || !rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + return data; + + QTemporaryFile tmp; + if (!tmp.open()) { + qCDebug(QRHI_LOG_INFO, "pipelineCacheData: Failed to create temporary file for Metal"); + return data; + } + tmp.close(); // the file exists until the tmp dtor runs + + const QString fn = QFileInfo(tmp.fileName()).absoluteFilePath(); + NSURL *url = QUrl::fromLocalFile(fn).toNSURL(); + NSError *err = nil; + if (![d->binArch serializeToURL: url error: &err]) { + const QString msg = QString::fromNSString(err.localizedDescription); + // Some of these "errors" are not actual errors. (think of "Nothing to serialize") + qCDebug(QRHI_LOG_INFO, "Failed to serialize MTLBinaryArchive: %s", qPrintable(msg)); + return data; + } + + QFile f(fn); + if (!f.open(QIODevice::ReadOnly)) { + qCDebug(QRHI_LOG_INFO, "pipelineCacheData: Failed to reopen temporary file"); + return data; + } + const QByteArray blob = f.readAll(); + f.close(); + + const size_t headerSize = sizeof(QMetalPipelineCacheDataHeader); + const quint32 dataSize = quint32(blob.size()); + + data.resize(headerSize + dataSize); + + QMetalPipelineCacheDataHeader header = {}; + header.rhiId = pipelineCacheRhiId(); + header.arch = quint32(sizeof(void*)); + header.dataSize = quint32(dataSize); + header.osMajor = osMajor; + header.osMinor = osMinor; + const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.length())); + if (driverStrLen) + memcpy(header.driver, driverInfoStruct.deviceName.constData(), driverStrLen); + header.driver[driverStrLen] = '\0'; + + memcpy(data.data(), &header, headerSize); + memcpy(data.data() + headerSize, blob.constData(), dataSize); + } + return data; } void QRhiMetal::setPipelineCacheData(const QByteArray &data) { - Q_UNUSED(data); + if (data.isEmpty()) + return; + + const size_t headerSize = sizeof(QMetalPipelineCacheDataHeader); + if (data.size() < qsizetype(headerSize)) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (header incomplete)"); + return; + } + + const size_t dataOffset = headerSize; + QMetalPipelineCacheDataHeader header; + memcpy(&header, data.constData(), headerSize); + + const quint32 rhiId = pipelineCacheRhiId(); + if (header.rhiId != rhiId) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", + rhiId, header.rhiId); + return; + } + + const quint32 arch = quint32(sizeof(void*)); + if (header.arch != arch) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Architecture does not match (%u, %u)", + arch, header.arch); + return; + } + + if (header.osMajor != osMajor || header.osMinor != osMinor) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: OS version does not match (%u.%u, %u.%u)", + osMajor, osMinor, header.osMajor, header.osMinor); + return; + } + + const size_t driverStrLen = qMin(sizeof(header.driver) - 1, size_t(driverInfoStruct.deviceName.length())); + if (strncmp(header.driver, driverInfoStruct.deviceName.constData(), driverStrLen)) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Metal device name does not match"); + return; + } + + if (data.size() < qsizetype(dataOffset + header.dataSize)) { + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size (data incomplete)"); + return; + } + + if (@available(macOS 11.0, iOS 14.0, *)) { + const char *p = data.constData() + dataOffset; + + QTemporaryFile tmp; + if (!tmp.open()) { + qCDebug(QRHI_LOG_INFO, "pipelineCacheData: Failed to create temporary file for Metal"); + return; + } + tmp.write(p, header.dataSize); + tmp.close(); // the file exists until the tmp dtor runs + + const QString fn = QFileInfo(tmp.fileName()).absoluteFilePath(); + NSURL *url = QUrl::fromLocalFile(fn).toNSURL(); + if (d->setupBinaryArchive(url)) + qCDebug(QRHI_LOG_INFO, "Created MTLBinaryArchive with initial data of %u bytes", header.dataSize); + } } QRhiRenderBuffer *QRhiMetal::createRenderBuffer(QRhiRenderBuffer::Type type, const QSize &pixelSize, @@ -693,10 +1078,10 @@ QRhiRenderBuffer *QRhiMetal::createRenderBuffer(QRhiRenderBuffer::Type type, con } QRhiTexture *QRhiMetal::createTexture(QRhiTexture::Format format, - const QSize &pixelSize, int depth, + const QSize &pixelSize, int depth, int arraySize, int sampleCount, QRhiTexture::Flags flags) { - return new QMetalTexture(this, format, pixelSize, depth, sampleCount, flags); + return new QMetalTexture(this, format, pixelSize, depth, arraySize, sampleCount, flags); } QRhiSampler *QRhiMetal::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, @@ -752,6 +1137,136 @@ static inline int mapBinding(int binding, return -1; } +static inline void bindStageBuffers(QMetalCommandBuffer *cbD, + int stage, + const QRhiBatchedBindings<id<MTLBuffer>>::Batch &bufferBatch, + const QRhiBatchedBindings<NSUInteger>::Batch &offsetBatch) +{ + switch (stage) { + case QMetalShaderResourceBindingsData::VERTEX: + [cbD->d->currentRenderPassEncoder setVertexBuffers: bufferBatch.resources.constData() + offsets: offsetBatch.resources.constData() + withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::FRAGMENT: + [cbD->d->currentRenderPassEncoder setFragmentBuffers: bufferBatch.resources.constData() + offsets: offsetBatch.resources.constData() + withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::COMPUTE: + [cbD->d->currentComputePassEncoder setBuffers: bufferBatch.resources.constData() + offsets: offsetBatch.resources.constData() + withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::TESSCTRL: + case QMetalShaderResourceBindingsData::TESSEVAL: + // do nothing. These are used later for tessellation + break; + default: + Q_UNREACHABLE(); + break; + } +} + +static inline void bindStageTextures(QMetalCommandBuffer *cbD, + int stage, + const QRhiBatchedBindings<id<MTLTexture>>::Batch &textureBatch) +{ + switch (stage) { + case QMetalShaderResourceBindingsData::VERTEX: + [cbD->d->currentRenderPassEncoder setVertexTextures: textureBatch.resources.constData() + withRange: NSMakeRange(textureBatch.startBinding, NSUInteger(textureBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::FRAGMENT: + [cbD->d->currentRenderPassEncoder setFragmentTextures: textureBatch.resources.constData() + withRange: NSMakeRange(textureBatch.startBinding, NSUInteger(textureBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::COMPUTE: + [cbD->d->currentComputePassEncoder setTextures: textureBatch.resources.constData() + withRange: NSMakeRange(textureBatch.startBinding, NSUInteger(textureBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::TESSCTRL: + case QMetalShaderResourceBindingsData::TESSEVAL: + // do nothing. These are used later for tessellation + break; + default: + Q_UNREACHABLE(); + break; + } +} + +static inline void bindStageSamplers(QMetalCommandBuffer *cbD, + int encoderStage, + const QRhiBatchedBindings<id<MTLSamplerState>>::Batch &samplerBatch) +{ + switch (encoderStage) { + case QMetalShaderResourceBindingsData::VERTEX: + [cbD->d->currentRenderPassEncoder setVertexSamplerStates: samplerBatch.resources.constData() + withRange: NSMakeRange(samplerBatch.startBinding, NSUInteger(samplerBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::FRAGMENT: + [cbD->d->currentRenderPassEncoder setFragmentSamplerStates: samplerBatch.resources.constData() + withRange: NSMakeRange(samplerBatch.startBinding, NSUInteger(samplerBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::COMPUTE: + [cbD->d->currentComputePassEncoder setSamplerStates: samplerBatch.resources.constData() + withRange: NSMakeRange(samplerBatch.startBinding, NSUInteger(samplerBatch.resources.count()))]; + break; + case QMetalShaderResourceBindingsData::TESSCTRL: + case QMetalShaderResourceBindingsData::TESSEVAL: + // do nothing. These are used later for tessellation + break; + default: + Q_UNREACHABLE(); + break; + } +} + +// Helper that is not used during the common vertex+fragment and compute +// pipelines, but is necessary when tessellation is involved and so the +// graphics pipeline is under the hood a combination of multiple compute and +// render pipelines. We need to be able to set the buffers, textures, samplers +// when a switching between render and compute encoders. +static inline void rebindShaderResources(QMetalCommandBuffer *cbD, int resourceStage, int encoderStage, + const QMetalShaderResourceBindingsData *customBindingState = nullptr) +{ + const QMetalShaderResourceBindingsData *bindingData = customBindingState ? customBindingState : &cbD->d->currentShaderResourceBindingState; + + for (int i = 0, ie = bindingData->res[resourceStage].bufferBatches.batches.count(); i != ie; ++i) { + const auto &bufferBatch(bindingData->res[resourceStage].bufferBatches.batches[i]); + const auto &offsetBatch(bindingData->res[resourceStage].bufferOffsetBatches.batches[i]); + bindStageBuffers(cbD, encoderStage, bufferBatch, offsetBatch); + } + + for (int i = 0, ie = bindingData->res[resourceStage].textureBatches.batches.count(); i != ie; ++i) { + const auto &batch(bindingData->res[resourceStage].textureBatches.batches[i]); + bindStageTextures(cbD, encoderStage, batch); + } + + for (int i = 0, ie = bindingData->res[resourceStage].samplerBatches.batches.count(); i != ie; ++i) { + const auto &batch(bindingData->res[resourceStage].samplerBatches.batches[i]); + bindStageSamplers(cbD, encoderStage, batch); + } +} + +static inline QRhiShaderResourceBinding::StageFlag toRhiSrbStage(int stage) +{ + switch (stage) { + case QMetalShaderResourceBindingsData::VERTEX: + return QRhiShaderResourceBinding::StageFlag::VertexStage; + case QMetalShaderResourceBindingsData::TESSCTRL: + return QRhiShaderResourceBinding::StageFlag::TessellationControlStage; + case QMetalShaderResourceBindingsData::TESSEVAL: + return QRhiShaderResourceBinding::StageFlag::TessellationEvaluationStage; + case QMetalShaderResourceBindingsData::FRAGMENT: + return QRhiShaderResourceBinding::StageFlag::FragmentStage; + case QMetalShaderResourceBindingsData::COMPUTE: + return QRhiShaderResourceBinding::StageFlag::ComputeStage; + } + + Q_UNREACHABLE_RETURN(QRhiShaderResourceBinding::StageFlag::VertexStage); +} + void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD, QMetalCommandBuffer *cbD, int dynamicOffsetCount, @@ -759,38 +1274,16 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD bool offsetOnlyChange, const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[SUPPORTED_STAGES]) { - struct Stage { - struct Buffer { - int nativeBinding; - id<MTLBuffer> mtlbuf; - uint offset; - }; - struct Texture { - int nativeBinding; - id<MTLTexture> mtltex; - }; - struct Sampler { - int nativeBinding; - id<MTLSamplerState> mtlsampler; - }; - QVarLengthArray<Buffer, 8> buffers; - QVarLengthArray<Texture, 8> textures; - QVarLengthArray<Sampler, 8> samplers; - QRhiBatchedBindings<id<MTLBuffer> > bufferBatches; - QRhiBatchedBindings<NSUInteger> bufferOffsetBatches; - QRhiBatchedBindings<id<MTLTexture> > textureBatches; - QRhiBatchedBindings<id<MTLSamplerState> > samplerBatches; - } res[SUPPORTED_STAGES]; - enum { VERTEX = 0, FRAGMENT = 1, COMPUTE = 2 }; + QMetalShaderResourceBindingsData bindingData; - for (const QRhiShaderResourceBinding &binding : qAsConst(srbD->sortedBindings)) { - const QRhiShaderResourceBinding::Data *b = binding.data(); + for (const QRhiShaderResourceBinding &binding : std::as_const(srbD->sortedBindings)) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(binding); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: { QMetalBuffer *bufD = QRHI_RES(QMetalBuffer, b->u.ubuf.buf); id<MTLBuffer> mtlbuf = bufD->d->buf[bufD->d->slotted ? currentFrameSlot : 0]; - uint offset = uint(b->u.ubuf.offset); + quint32 offset = b->u.ubuf.offset; for (int i = 0; i < dynamicOffsetCount; ++i) { const QRhiCommandBuffer::DynamicOffset &dynOfs(dynamicOffsets[i]); if (dynOfs.first == b->binding) { @@ -798,51 +1291,38 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD break; } } - if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { - const int nativeBinding = mapBinding(b->binding, VERTEX, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[VERTEX].buffers.append({ nativeBinding, mtlbuf, offset }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { - const int nativeBinding = mapBinding(b->binding, FRAGMENT, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[FRAGMENT].buffers.append({ nativeBinding, mtlbuf, offset }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { - const int nativeBinding = mapBinding(b->binding, COMPUTE, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[COMPUTE].buffers.append({ nativeBinding, mtlbuf, offset }); + + for (int stage = 0; stage < SUPPORTED_STAGES; ++stage) { + if (b->stage.testFlag(toRhiSrbStage(stage))) { + const int nativeBinding = mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Buffer); + if (nativeBinding >= 0) + bindingData.res[stage].buffers.append({ nativeBinding, mtlbuf, offset }); + } } } break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; for (int elem = 0; elem < data->count; ++elem) { QMetalTexture *texD = QRHI_RES(QMetalTexture, b->u.stex.texSamplers[elem].tex); QMetalSampler *samplerD = QRHI_RES(QMetalSampler, b->u.stex.texSamplers[elem].sampler); - if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { - const int nativeBindingTexture = mapBinding(b->binding, VERTEX, nativeResourceBindingMaps, BindingType::Texture); - const int nativeBindingSampler = mapBinding(b->binding, VERTEX, nativeResourceBindingMaps, BindingType::Sampler); - if (nativeBindingTexture >= 0 && nativeBindingSampler >= 0) { - res[VERTEX].textures.append({ nativeBindingTexture + elem, texD->d->tex }); - res[VERTEX].samplers.append({ nativeBindingSampler + elem, samplerD->d->samplerState }); - } - } - if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { - const int nativeBindingTexture = mapBinding(b->binding, FRAGMENT, nativeResourceBindingMaps, BindingType::Texture); - const int nativeBindingSampler = mapBinding(b->binding, FRAGMENT, nativeResourceBindingMaps, BindingType::Sampler); - if (nativeBindingTexture >= 0 && nativeBindingSampler >= 0) { - res[FRAGMENT].textures.append({ nativeBindingTexture + elem, texD->d->tex }); - res[FRAGMENT].samplers.append({ nativeBindingSampler + elem, samplerD->d->samplerState }); - } - } - if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { - const int nativeBindingTexture = mapBinding(b->binding, COMPUTE, nativeResourceBindingMaps, BindingType::Texture); - const int nativeBindingSampler = mapBinding(b->binding, COMPUTE, nativeResourceBindingMaps, BindingType::Sampler); - if (nativeBindingTexture >= 0 && nativeBindingSampler >= 0) { - res[COMPUTE].textures.append({ nativeBindingTexture + elem, texD->d->tex }); - res[COMPUTE].samplers.append({ nativeBindingSampler + elem, samplerD->d->samplerState }); + + for (int stage = 0; stage < SUPPORTED_STAGES; ++stage) { + if (b->stage.testFlag(toRhiSrbStage(stage))) { + // Must handle all three cases (combined, separate, separate): + // first = texture binding, second = sampler binding + // first = texture binding + // first = sampler binding (i.e. BindingType::Texture...) + const int textureBinding = mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Texture); + const int samplerBinding = texD && samplerD ? mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Sampler) + : (samplerD ? mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Texture) : -1); + if (textureBinding >= 0 && texD) + bindingData.res[stage].textures.append({ textureBinding + elem, texD->d->tex }); + if (samplerBinding >= 0) + bindingData.res[stage].samplers.append({ samplerBinding + elem, samplerD->d->samplerState }); } } } @@ -854,20 +1334,13 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD { QMetalTexture *texD = QRHI_RES(QMetalTexture, b->u.simage.tex); id<MTLTexture> t = texD->d->viewForLevel(b->u.simage.level); - if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { - const int nativeBinding = mapBinding(b->binding, VERTEX, nativeResourceBindingMaps, BindingType::Texture); - if (nativeBinding >= 0) - res[VERTEX].textures.append({ nativeBinding, t }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { - const int nativeBinding = mapBinding(b->binding, FRAGMENT, nativeResourceBindingMaps, BindingType::Texture); - if (nativeBinding >= 0) - res[FRAGMENT].textures.append({ nativeBinding, t }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { - const int nativeBinding = mapBinding(b->binding, COMPUTE, nativeResourceBindingMaps, BindingType::Texture); - if (nativeBinding >= 0) - res[COMPUTE].textures.append({ nativeBinding, t }); + + for (int stage = 0; stage < SUPPORTED_STAGES; ++stage) { + if (b->stage.testFlag(toRhiSrbStage(stage))) { + const int nativeBinding = mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Texture); + if (nativeBinding >= 0) + bindingData.res[stage].textures.append({ nativeBinding, t }); + } } } break; @@ -877,21 +1350,13 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD { QMetalBuffer *bufD = QRHI_RES(QMetalBuffer, b->u.sbuf.buf); id<MTLBuffer> mtlbuf = bufD->d->buf[0]; - uint offset = uint(b->u.sbuf.offset); - if (b->stage.testFlag(QRhiShaderResourceBinding::VertexStage)) { - const int nativeBinding = mapBinding(b->binding, VERTEX, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[VERTEX].buffers.append({ nativeBinding, mtlbuf, offset }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::FragmentStage)) { - const int nativeBinding = mapBinding(b->binding, FRAGMENT, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[FRAGMENT].buffers.append({ nativeBinding, mtlbuf, offset }); - } - if (b->stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) { - const int nativeBinding = mapBinding(b->binding, COMPUTE, nativeResourceBindingMaps, BindingType::Buffer); - if (nativeBinding >= 0) - res[COMPUTE].buffers.append({ nativeBinding, mtlbuf, offset }); + quint32 offset = b->u.sbuf.offset; + for (int stage = 0; stage < SUPPORTED_STAGES; ++stage) { + if (b->stage.testFlag(toRhiSrbStage(stage))) { + const int nativeBinding = mapBinding(b->binding, stage, nativeResourceBindingMaps, BindingType::Buffer); + if (nativeBinding >= 0) + bindingData.res[stage].buffers.append({ nativeBinding, mtlbuf, offset }); + } } } break; @@ -902,9 +1367,10 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD } for (int stage = 0; stage < SUPPORTED_STAGES; ++stage) { - if (cbD->recordingPass != QMetalCommandBuffer::RenderPass && (stage == VERTEX || stage == FRAGMENT)) + if (cbD->recordingPass != QMetalCommandBuffer::RenderPass && (stage == QMetalShaderResourceBindingsData::VERTEX || stage == QMetalShaderResourceBindingsData::FRAGMENT + || stage == QMetalShaderResourceBindingsData::TESSCTRL || stage == QMetalShaderResourceBindingsData::TESSEVAL)) continue; - if (cbD->recordingPass != QMetalCommandBuffer::ComputePass && stage == COMPUTE) + if (cbD->recordingPass != QMetalCommandBuffer::ComputePass && (stage == QMetalShaderResourceBindingsData::COMPUTE)) continue; // QRhiBatchedBindings works with the native bindings and expects @@ -912,104 +1378,107 @@ void QRhiMetal::enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD // on the QRhi (SPIR-V) binding) is not helpful in this regard, so we // have to sort here every time. - std::sort(res[stage].buffers.begin(), res[stage].buffers.end(), [](const Stage::Buffer &a, const Stage::Buffer &b) { + std::sort(bindingData.res[stage].buffers.begin(), bindingData.res[stage].buffers.end(), [](const QMetalShaderResourceBindingsData::Stage::Buffer &a, const QMetalShaderResourceBindingsData::Stage::Buffer &b) { return a.nativeBinding < b.nativeBinding; }); - for (const Stage::Buffer &buf : qAsConst(res[stage].buffers)) { - res[stage].bufferBatches.feed(buf.nativeBinding, buf.mtlbuf); - res[stage].bufferOffsetBatches.feed(buf.nativeBinding, buf.offset); + for (const QMetalShaderResourceBindingsData::Stage::Buffer &buf : std::as_const(bindingData.res[stage].buffers)) { + bindingData.res[stage].bufferBatches.feed(buf.nativeBinding, buf.mtlbuf); + bindingData.res[stage].bufferOffsetBatches.feed(buf.nativeBinding, buf.offset); } - res[stage].bufferBatches.finish(); - res[stage].bufferOffsetBatches.finish(); - - for (int i = 0, ie = res[stage].bufferBatches.batches.count(); i != ie; ++i) { - const auto &bufferBatch(res[stage].bufferBatches.batches[i]); - const auto &offsetBatch(res[stage].bufferOffsetBatches.batches[i]); - switch (stage) { - case VERTEX: - [cbD->d->currentRenderPassEncoder setVertexBuffers: bufferBatch.resources.constData() - offsets: offsetBatch.resources.constData() - withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; - break; - case FRAGMENT: - [cbD->d->currentRenderPassEncoder setFragmentBuffers: bufferBatch.resources.constData() - offsets: offsetBatch.resources.constData() - withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; - break; - case COMPUTE: - [cbD->d->currentComputePassEncoder setBuffers: bufferBatch.resources.constData() - offsets: offsetBatch.resources.constData() - withRange: NSMakeRange(bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; - break; - default: - Q_UNREACHABLE(); - break; + bindingData.res[stage].bufferBatches.finish(); + bindingData.res[stage].bufferOffsetBatches.finish(); + + for (int i = 0, ie = bindingData.res[stage].bufferBatches.batches.count(); i != ie; ++i) { + const auto &bufferBatch(bindingData.res[stage].bufferBatches.batches[i]); + const auto &offsetBatch(bindingData.res[stage].bufferOffsetBatches.batches[i]); + // skip setting Buffer binding if the current state is already correct + if (cbD->d->currentShaderResourceBindingState.res[stage].bufferBatches.batches.count() > i + && cbD->d->currentShaderResourceBindingState.res[stage].bufferOffsetBatches.batches.count() > i + && bufferBatch == cbD->d->currentShaderResourceBindingState.res[stage].bufferBatches.batches[i] + && offsetBatch == cbD->d->currentShaderResourceBindingState.res[stage].bufferOffsetBatches.batches[i]) + { + continue; } + bindStageBuffers(cbD, stage, bufferBatch, offsetBatch); } if (offsetOnlyChange) continue; - std::sort(res[stage].textures.begin(), res[stage].textures.end(), [](const Stage::Texture &a, const Stage::Texture &b) { + std::sort(bindingData.res[stage].textures.begin(), bindingData.res[stage].textures.end(), [](const QMetalShaderResourceBindingsData::Stage::Texture &a, const QMetalShaderResourceBindingsData::Stage::Texture &b) { return a.nativeBinding < b.nativeBinding; }); - std::sort(res[stage].samplers.begin(), res[stage].samplers.end(), [](const Stage::Sampler &a, const Stage::Sampler &b) { + std::sort(bindingData.res[stage].samplers.begin(), bindingData.res[stage].samplers.end(), [](const QMetalShaderResourceBindingsData::Stage::Sampler &a, const QMetalShaderResourceBindingsData::Stage::Sampler &b) { return a.nativeBinding < b.nativeBinding; }); - for (const Stage::Texture &t : qAsConst(res[stage].textures)) - res[stage].textureBatches.feed(t.nativeBinding, t.mtltex); + for (const QMetalShaderResourceBindingsData::Stage::Texture &t : std::as_const(bindingData.res[stage].textures)) + bindingData.res[stage].textureBatches.feed(t.nativeBinding, t.mtltex); - for (const Stage::Sampler &s : qAsConst(res[stage].samplers)) - res[stage].samplerBatches.feed(s.nativeBinding, s.mtlsampler); + for (const QMetalShaderResourceBindingsData::Stage::Sampler &s : std::as_const(bindingData.res[stage].samplers)) + bindingData.res[stage].samplerBatches.feed(s.nativeBinding, s.mtlsampler); - res[stage].textureBatches.finish(); - res[stage].samplerBatches.finish(); + bindingData.res[stage].textureBatches.finish(); + bindingData.res[stage].samplerBatches.finish(); - for (int i = 0, ie = res[stage].textureBatches.batches.count(); i != ie; ++i) { - const auto &batch(res[stage].textureBatches.batches[i]); - switch (stage) { - case VERTEX: - [cbD->d->currentRenderPassEncoder setVertexTextures: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - case FRAGMENT: - [cbD->d->currentRenderPassEncoder setFragmentTextures: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - case COMPUTE: - [cbD->d->currentComputePassEncoder setTextures: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - default: - Q_UNREACHABLE(); - break; + for (int i = 0, ie = bindingData.res[stage].textureBatches.batches.count(); i != ie; ++i) { + const auto &batch(bindingData.res[stage].textureBatches.batches[i]); + // skip setting Texture binding if the current state is already correct + if (cbD->d->currentShaderResourceBindingState.res[stage].textureBatches.batches.count() > i + && batch == cbD->d->currentShaderResourceBindingState.res[stage].textureBatches.batches[i]) + { + continue; } + bindStageTextures(cbD, stage, batch); } - for (int i = 0, ie = res[stage].samplerBatches.batches.count(); i != ie; ++i) { - const auto &batch(res[stage].samplerBatches.batches[i]); - switch (stage) { - case VERTEX: - [cbD->d->currentRenderPassEncoder setVertexSamplerStates: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - case FRAGMENT: - [cbD->d->currentRenderPassEncoder setFragmentSamplerStates: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - case COMPUTE: - [cbD->d->currentComputePassEncoder setSamplerStates: batch.resources.constData() - withRange: NSMakeRange(batch.startBinding, NSUInteger(batch.resources.count()))]; - break; - default: - Q_UNREACHABLE(); - break; + + for (int i = 0, ie = bindingData.res[stage].samplerBatches.batches.count(); i != ie; ++i) { + const auto &batch(bindingData.res[stage].samplerBatches.batches[i]); + // skip setting Sampler State if the current state is already correct + if (cbD->d->currentShaderResourceBindingState.res[stage].samplerBatches.batches.count() > i + && batch == cbD->d->currentShaderResourceBindingState.res[stage].samplerBatches.batches[i]) + { + continue; } + bindStageSamplers(cbD, stage, batch); } } + + cbD->d->currentShaderResourceBindingState = bindingData; +} + +void QMetalGraphicsPipeline::makeActiveForCurrentRenderPassEncoder(QMetalCommandBuffer *cbD) +{ + [cbD->d->currentRenderPassEncoder setRenderPipelineState: d->ps]; + + if (cbD->d->currentDepthStencilState != d->ds) { + [cbD->d->currentRenderPassEncoder setDepthStencilState: d->ds]; + cbD->d->currentDepthStencilState = d->ds; + } + + if (cbD->currentCullMode == -1 || d->cullMode != uint(cbD->currentCullMode)) { + [cbD->d->currentRenderPassEncoder setCullMode: d->cullMode]; + cbD->currentCullMode = int(d->cullMode); + } + if (cbD->currentTriangleFillMode == -1 || d->triangleFillMode != uint(cbD->currentTriangleFillMode)) { + [cbD->d->currentRenderPassEncoder setTriangleFillMode: d->triangleFillMode]; + cbD->currentTriangleFillMode = int(d->triangleFillMode); + } + if (cbD->currentFrontFaceWinding == -1 || d->winding != uint(cbD->currentFrontFaceWinding)) { + [cbD->d->currentRenderPassEncoder setFrontFacingWinding: d->winding]; + cbD->currentFrontFaceWinding = int(d->winding); + } + if (!qFuzzyCompare(d->depthBias, cbD->currentDepthBiasValues.first) + || !qFuzzyCompare(d->slopeScaledDepthBias, cbD->currentDepthBiasValues.second)) + { + [cbD->d->currentRenderPassEncoder setDepthBias: d->depthBias + slopeScale: d->slopeScaledDepthBias + clamp: 0.0f]; + cbD->currentDepthBiasValues = { d->depthBias, d->slopeScaledDepthBias }; + } } void QRhiMetal::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline *ps) @@ -1018,29 +1487,24 @@ void QRhiMetal::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline Q_ASSERT(cbD->recordingPass == QMetalCommandBuffer::RenderPass); QMetalGraphicsPipeline *psD = QRHI_RES(QMetalGraphicsPipeline, ps); - if (cbD->currentGraphicsPipeline != ps || cbD->currentPipelineGeneration != psD->generation) { - cbD->currentGraphicsPipeline = ps; - cbD->currentComputePipeline = nullptr; - cbD->currentPipelineGeneration = psD->generation; + if (cbD->currentGraphicsPipeline == psD && cbD->currentPipelineGeneration == psD->generation) + return; - [cbD->d->currentRenderPassEncoder setRenderPipelineState: psD->d->ps]; - [cbD->d->currentRenderPassEncoder setDepthStencilState: psD->d->ds]; + cbD->currentGraphicsPipeline = psD; + cbD->currentComputePipeline = nullptr; + cbD->currentPipelineGeneration = psD->generation; - if (cbD->currentCullMode == -1 || psD->d->cullMode != uint(cbD->currentCullMode)) { - [cbD->d->currentRenderPassEncoder setCullMode: psD->d->cullMode]; - cbD->currentCullMode = int(psD->d->cullMode); - } - if (cbD->currentFrontFaceWinding == -1 || psD->d->winding != uint(cbD->currentFrontFaceWinding)) { - [cbD->d->currentRenderPassEncoder setFrontFacingWinding: psD->d->winding]; - cbD->currentFrontFaceWinding = int(psD->d->winding); + if (!psD->d->tess.enabled && !psD->d->tess.failed) { + psD->makeActiveForCurrentRenderPassEncoder(cbD); + } else { + // mark work buffers that can now be safely reused as reusable + for (QMetalBuffer *workBuf : psD->d->extraBufMgr.deviceLocalWorkBuffers) { + if (workBuf && workBuf->lastActiveFrameSlot == currentFrameSlot) + workBuf->lastActiveFrameSlot = -1; } - if (!qFuzzyCompare(psD->d->depthBias, cbD->currentDepthBiasValues.first) - || !qFuzzyCompare(psD->d->slopeScaledDepthBias, cbD->currentDepthBiasValues.second)) - { - [cbD->d->currentRenderPassEncoder setDepthBias: psD->d->depthBias - slopeScale: psD->d->slopeScaledDepthBias - clamp: 0.0f]; - cbD->currentDepthBiasValues = { psD->d->depthBias, psD->d->slopeScaledDepthBias }; + for (QMetalBuffer *workBuf : psD->d->extraBufMgr.hostVisibleWorkBuffers) { + if (workBuf && workBuf->lastActiveFrameSlot == currentFrameSlot) + workBuf->lastActiveFrameSlot = -1; } } @@ -1053,8 +1517,8 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind { QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); Q_ASSERT(cbD->recordingPass != QMetalCommandBuffer::NoPass); - QMetalGraphicsPipeline *gfxPsD = QRHI_RES(QMetalGraphicsPipeline, cbD->currentGraphicsPipeline); - QMetalComputePipeline *compPsD = QRHI_RES(QMetalComputePipeline, cbD->currentComputePipeline); + QMetalGraphicsPipeline *gfxPsD = cbD->currentGraphicsPipeline; + QMetalComputePipeline *compPsD = cbD->currentComputePipeline; if (!srb) { if (gfxPsD) @@ -1068,9 +1532,15 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind bool hasDynamicOffsetInSrb = false; bool resNeedsRebind = false; + // SPIRV-Cross buffer size buffers + // Need to determine storage buffer sizes here as this is the last opportunity for storage + // buffer bindings (offset, size) to be specified before draw / dispatch call + const bool needsBufferSizeBuffer = (compPsD && compPsD->d->bufferSizeBuffer) || (gfxPsD && gfxPsD->d->bufferSizeBuffer); + QMap<QRhiShaderResourceBinding::StageFlag, QMap<int, quint32>> storageBufferSizes; + // do buffer writes, figure out if we need to rebind, and mark as in-use for (int i = 0, ie = srbD->sortedBindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->sortedBindings.at(i).data(); + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings.at(i)); QMetalShaderResourceBindings::BoundResourceData &bd(srbD->boundResourceData[i]); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: @@ -1091,8 +1561,10 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind } break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; if (bd.stex.count != data->count) { bd.stex.count = data->count; resNeedsRebind = true; @@ -1100,19 +1572,26 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind for (int elem = 0; elem < data->count; ++elem) { QMetalTexture *texD = QRHI_RES(QMetalTexture, data->texSamplers[elem].tex); QMetalSampler *samplerD = QRHI_RES(QMetalSampler, data->texSamplers[elem].sampler); - if (texD->generation != bd.stex.d[elem].texGeneration - || texD->m_id != bd.stex.d[elem].texId - || samplerD->generation != bd.stex.d[elem].samplerGeneration - || samplerD->m_id != bd.stex.d[elem].samplerId) + Q_ASSERT(texD || samplerD); + const quint64 texId = texD ? texD->m_id : 0; + const uint texGen = texD ? texD->generation : 0; + const quint64 samplerId = samplerD ? samplerD->m_id : 0; + const uint samplerGen = samplerD ? samplerD->generation : 0; + if (texGen != bd.stex.d[elem].texGeneration + || texId != bd.stex.d[elem].texId + || samplerGen != bd.stex.d[elem].samplerGeneration + || samplerId != bd.stex.d[elem].samplerId) { resNeedsRebind = true; - bd.stex.d[elem].texId = texD->m_id; - bd.stex.d[elem].texGeneration = texD->generation; - bd.stex.d[elem].samplerId = samplerD->m_id; - bd.stex.d[elem].samplerGeneration = samplerD->generation; + bd.stex.d[elem].texId = texId; + bd.stex.d[elem].texGeneration = texGen; + bd.stex.d[elem].samplerId = samplerId; + bd.stex.d[elem].samplerGeneration = samplerGen; } - texD->lastActiveFrameSlot = currentFrameSlot; - samplerD->lastActiveFrameSlot = currentFrameSlot; + if (texD) + texD->lastActiveFrameSlot = currentFrameSlot; + if (samplerD) + samplerD->lastActiveFrameSlot = currentFrameSlot; } } break; @@ -1135,6 +1614,17 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind { QMetalBuffer *bufD = QRHI_RES(QMetalBuffer, b->u.sbuf.buf); Q_ASSERT(bufD->m_usage.testFlag(QRhiBuffer::StorageBuffer)); + + if (needsBufferSizeBuffer) { + for (int i = 0; i < 6; ++i) { + const QRhiShaderResourceBinding::StageFlag stage = + QRhiShaderResourceBinding::StageFlag(1 << i); + if (b->stage.testFlag(stage)) { + storageBufferSizes[stage][b->binding] = b->u.sbuf.maybeSize ? b->u.sbuf.maybeSize : bufD->size(); + } + } + } + executeBufferHostWritesForCurrentFrame(bufD); if (bufD->generation != bd.sbuf.generation || bufD->m_id != bd.sbuf.id) { resNeedsRebind = true; @@ -1150,26 +1640,141 @@ void QRhiMetal::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind } } + if (needsBufferSizeBuffer) { + QMetalBuffer *bufD = nullptr; + QVarLengthArray<QPair<QMetalShader *, QRhiShaderResourceBinding::StageFlag>, 4> shaders; + + if (compPsD) { + bufD = compPsD->d->bufferSizeBuffer; + Q_ASSERT(compPsD->d->cs.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)); + shaders.append(qMakePair(&compPsD->d->cs, QRhiShaderResourceBinding::StageFlag::ComputeStage)); + } else { + bufD = gfxPsD->d->bufferSizeBuffer; + if (gfxPsD->d->tess.enabled) { + + // Assumptions + // * We only use one of the compute vertex shader variants in a pipeline at any one time + // * The vertex shader variants all have the same storage block bindings + // * The vertex shader variants all have the same native resource binding map + // * The vertex shader variants all have the same MslBufferSizeBufferBinding requirement + // * The vertex shader variants all have the same MslBufferSizeBufferBinding binding + // => We only need to use one vertex shader variant to generate the identical shader + // resource bindings + Q_ASSERT(gfxPsD->d->tess.compVs[0].desc.storageBlocks() == gfxPsD->d->tess.compVs[1].desc.storageBlocks()); + Q_ASSERT(gfxPsD->d->tess.compVs[0].desc.storageBlocks() == gfxPsD->d->tess.compVs[2].desc.storageBlocks()); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeResourceBindingMap == gfxPsD->d->tess.compVs[1].nativeResourceBindingMap); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeResourceBindingMap == gfxPsD->d->tess.compVs[2].nativeResourceBindingMap); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding) + == gfxPsD->d->tess.compVs[1].nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding) + == gfxPsD->d->tess.compVs[2].nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding] + == gfxPsD->d->tess.compVs[1].nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding]); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding] + == gfxPsD->d->tess.compVs[2].nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding]); + + if (gfxPsD->d->tess.compVs[0].nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) + shaders.append(qMakePair(&gfxPsD->d->tess.compVs[0], QRhiShaderResourceBinding::StageFlag::VertexStage)); + + if (gfxPsD->d->tess.compTesc.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) + shaders.append(qMakePair(&gfxPsD->d->tess.compTesc, QRhiShaderResourceBinding::StageFlag::TessellationControlStage)); + + if (gfxPsD->d->tess.vertTese.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) + shaders.append(qMakePair(&gfxPsD->d->tess.vertTese, QRhiShaderResourceBinding::StageFlag::TessellationEvaluationStage)); + + } else { + if (gfxPsD->d->vs.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) + shaders.append(qMakePair(&gfxPsD->d->vs, QRhiShaderResourceBinding::StageFlag::VertexStage)); + } + if (gfxPsD->d->fs.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) + shaders.append(qMakePair(&gfxPsD->d->fs, QRhiShaderResourceBinding::StageFlag::FragmentStage)); + } + + quint32 offset = 0; + for (const QPair<QMetalShader *, QRhiShaderResourceBinding::StageFlag> &shader : shaders) { + + const int binding = shader.first->nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding]; + + // if we don't have a srb entry for the buffer size buffer + if (!(storageBufferSizes.contains(shader.second) && storageBufferSizes[shader.second].contains(binding))) { + + int maxNativeBinding = 0; + for (const QShaderDescription::StorageBlock &block : shader.first->desc.storageBlocks()) + maxNativeBinding = qMax(maxNativeBinding, shader.first->nativeResourceBindingMap[block.binding].first); + + const int size = (maxNativeBinding + 1) * sizeof(int); + + Q_ASSERT(offset + size <= bufD->size()); + srbD->sortedBindings.append(QRhiShaderResourceBinding::bufferLoad(binding, shader.second, bufD, offset, size)); + + QMetalShaderResourceBindings::BoundResourceData bd; + bd.sbuf.id = bufD->m_id; + bd.sbuf.generation = bufD->generation; + srbD->boundResourceData.append(bd); + } + + // create the buffer size buffer data + QVarLengthArray<int, 8> bufferSizeBufferData; + Q_ASSERT(storageBufferSizes.contains(shader.second)); + const QMap<int, quint32> &sizes(storageBufferSizes[shader.second]); + for (const QShaderDescription::StorageBlock &block : shader.first->desc.storageBlocks()) { + const int index = shader.first->nativeResourceBindingMap[block.binding].first; + + // if the native binding is -1, the buffer is present but not accessed in the shader + if (index < 0) + continue; + + if (bufferSizeBufferData.size() <= index) + bufferSizeBufferData.resize(index + 1); + + Q_ASSERT(sizes.contains(block.binding)); + bufferSizeBufferData[index] = sizes[block.binding]; + } + + QRhiBufferData data; + const quint32 size = bufferSizeBufferData.size() * sizeof(int); + data.assign(reinterpret_cast<const char *>(bufferSizeBufferData.constData()), size); + Q_ASSERT(offset + size <= bufD->size()); + bufD->d->pendingUpdates[bufD->d->slotted ? currentFrameSlot : 0].append({ offset, data }); + + // buffer offsets must be 32byte aligned + offset += ((size + 31) / 32) * 32; + } + + executeBufferHostWritesForCurrentFrame(bufD); + bufD->lastActiveFrameSlot = currentFrameSlot; + } + // make sure the resources for the correct slot get bound const int resSlot = hasSlottedResourceInSrb ? currentFrameSlot : 0; if (hasSlottedResourceInSrb && cbD->currentResSlot != resSlot) resNeedsRebind = true; - const bool srbChanged = gfxPsD ? (cbD->currentGraphicsSrb != srb) : (cbD->currentComputeSrb != srb); + const bool srbChanged = gfxPsD ? (cbD->currentGraphicsSrb != srbD) : (cbD->currentComputeSrb != srbD); const bool srbRebuilt = cbD->currentSrbGeneration != srbD->generation; // dynamic uniform buffer offsets always trigger a rebind if (hasDynamicOffsetInSrb || resNeedsRebind || srbChanged || srbRebuilt) { - const QShader::NativeResourceBindingMap *resBindMaps[SUPPORTED_STAGES] = { nullptr, nullptr, nullptr }; + const QShader::NativeResourceBindingMap *resBindMaps[SUPPORTED_STAGES] = { nullptr, nullptr, nullptr, nullptr, nullptr }; if (gfxPsD) { - cbD->currentGraphicsSrb = srb; + cbD->currentGraphicsSrb = srbD; cbD->currentComputeSrb = nullptr; - resBindMaps[0] = &gfxPsD->d->vs.nativeResourceBindingMap; - resBindMaps[1] = &gfxPsD->d->fs.nativeResourceBindingMap; + if (gfxPsD->d->tess.enabled) { + // If tessellating, we don't know which compVs shader to use until the draw call is + // made. They should all have the same native resource binding map, so pick one. + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeResourceBindingMap == gfxPsD->d->tess.compVs[1].nativeResourceBindingMap); + Q_ASSERT(gfxPsD->d->tess.compVs[0].nativeResourceBindingMap == gfxPsD->d->tess.compVs[2].nativeResourceBindingMap); + resBindMaps[QMetalShaderResourceBindingsData::VERTEX] = &gfxPsD->d->tess.compVs[0].nativeResourceBindingMap; + resBindMaps[QMetalShaderResourceBindingsData::TESSCTRL] = &gfxPsD->d->tess.compTesc.nativeResourceBindingMap; + resBindMaps[QMetalShaderResourceBindingsData::TESSEVAL] = &gfxPsD->d->tess.vertTese.nativeResourceBindingMap; + } else { + resBindMaps[QMetalShaderResourceBindingsData::VERTEX] = &gfxPsD->d->vs.nativeResourceBindingMap; + } + resBindMaps[QMetalShaderResourceBindingsData::FRAGMENT] = &gfxPsD->d->fs.nativeResourceBindingMap; } else { cbD->currentGraphicsSrb = nullptr; - cbD->currentComputeSrb = srb; - resBindMaps[2] = &compPsD->d->cs.nativeResourceBindingMap; + cbD->currentComputeSrb = srbD; + resBindMaps[QMetalShaderResourceBindingsData::COMPUTE] = &compPsD->d->cs.nativeResourceBindingMap; } cbD->currentSrbGeneration = srbD->generation; cbD->currentResSlot = resSlot; @@ -1200,13 +1805,13 @@ void QRhiMetal::setVertexInput(QRhiCommandBuffer *cb, offsets.finish(); // same binding space for vertex and constant buffers - work it around - QRhiShaderResourceBindings *srb = cbD->currentGraphicsSrb; + QMetalShaderResourceBindings *srbD = cbD->currentGraphicsSrb; // There's nothing guaranteeing setShaderResources() was called before // setVertexInput()... but whatever srb will get bound will have to be // layout-compatible anyways so maxBinding is the same. - if (!srb) - srb = cbD->currentGraphicsPipeline->shaderResourceBindings(); - const int firstVertexBinding = QRHI_RES(QMetalShaderResourceBindings, srb)->maxBinding + 1; + if (!srbD) + srbD = QRHI_RES(QMetalShaderResourceBindings, cbD->currentGraphicsPipeline->shaderResourceBindings()); + const int firstVertexBinding = srbD->maxBinding + 1; if (firstVertexBinding != cbD->d->currentFirstVertexBinding || buffers != cbD->d->currentVertexInputsBuffers @@ -1230,7 +1835,7 @@ void QRhiMetal::setVertexInput(QRhiCommandBuffer *cb, QMetalBuffer *ibufD = QRHI_RES(QMetalBuffer, indexBuf); executeBufferHostWritesForCurrentFrame(ibufD); ibufD->lastActiveFrameSlot = currentFrameSlot; - cbD->currentIndexBuffer = indexBuf; + cbD->currentIndexBuffer = ibufD; cbD->currentIndexOffset = indexOffset; cbD->currentIndexFormat = indexFormat; } else { @@ -1246,7 +1851,7 @@ void QRhiMetal::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) // x,y is top-left in MTLViewportRect but bottom-left in QRhiViewport float x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, viewport.viewport(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<UnBounded>(outputSize, viewport.viewport(), &x, &y, &w, &h)) return; MTLViewport vp; @@ -1259,8 +1864,10 @@ void QRhiMetal::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) [cbD->d->currentRenderPassEncoder setViewport: vp]; - if (!QRHI_RES(QMetalGraphicsPipeline, cbD->currentGraphicsPipeline)->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)) { + if (cbD->currentGraphicsPipeline + && !cbD->currentGraphicsPipeline->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)) { MTLScissorRect s; + qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, viewport.viewport(), &x, &y, &w, &h); s.x = NSUInteger(x); s.y = NSUInteger(y); s.width = NSUInteger(w); @@ -1273,12 +1880,12 @@ void QRhiMetal::setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) { QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); Q_ASSERT(cbD->recordingPass == QMetalCommandBuffer::RenderPass); - Q_ASSERT(QRHI_RES(QMetalGraphicsPipeline, cbD->currentGraphicsPipeline)->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)); + Q_ASSERT(cbD->currentGraphicsPipeline->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)); const QSize outputSize = cbD->currentTarget->pixelSize(); // x,y is top-left in MTLScissorRect but bottom-left in QRhiScissor int x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, scissor.scissor(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, scissor.scissor(), &x, &y, &w, &h)) return; MTLScissorRect s; @@ -1307,15 +1914,321 @@ void QRhiMetal::setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) [cbD->d->currentRenderPassEncoder setStencilReferenceValue: refValue]; } +static id<MTLComputeCommandEncoder> tessellationComputeEncoder(QMetalCommandBuffer *cbD) +{ + if (cbD->d->currentRenderPassEncoder) { + [cbD->d->currentRenderPassEncoder endEncoding]; + cbD->d->currentRenderPassEncoder = nil; + } + + if (!cbD->d->tessellationComputeEncoder) + cbD->d->tessellationComputeEncoder = [cbD->d->cb computeCommandEncoder]; + + return cbD->d->tessellationComputeEncoder; +} + +static void endTessellationComputeEncoding(QMetalCommandBuffer *cbD) +{ + if (cbD->d->tessellationComputeEncoder) { + [cbD->d->tessellationComputeEncoder endEncoding]; + cbD->d->tessellationComputeEncoder = nil; + } + + QMetalRenderTargetData * rtD = nullptr; + + switch (cbD->currentTarget->resourceType()) { + case QRhiResource::SwapChainRenderTarget: + rtD = QRHI_RES(QMetalSwapChainRenderTarget, cbD->currentTarget)->d; + break; + case QRhiResource::TextureRenderTarget: + rtD = QRHI_RES(QMetalTextureRenderTarget, cbD->currentTarget)->d; + break; + default: + break; + } + + Q_ASSERT(rtD); + + QVarLengthArray<MTLLoadAction, 4> oldColorLoad; + for (uint i = 0; i < uint(rtD->colorAttCount); ++i) { + oldColorLoad.append(cbD->d->currentPassRpDesc.colorAttachments[i].loadAction); + if (cbD->d->currentPassRpDesc.colorAttachments[i].storeAction != MTLStoreActionDontCare) + cbD->d->currentPassRpDesc.colorAttachments[i].loadAction = MTLLoadActionLoad; + } + + MTLLoadAction oldDepthLoad; + MTLLoadAction oldStencilLoad; + if (rtD->dsAttCount) { + oldDepthLoad = cbD->d->currentPassRpDesc.depthAttachment.loadAction; + if (cbD->d->currentPassRpDesc.depthAttachment.storeAction != MTLStoreActionDontCare) + cbD->d->currentPassRpDesc.depthAttachment.loadAction = MTLLoadActionLoad; + + oldStencilLoad = cbD->d->currentPassRpDesc.stencilAttachment.loadAction; + if (cbD->d->currentPassRpDesc.stencilAttachment.storeAction != MTLStoreActionDontCare) + cbD->d->currentPassRpDesc.stencilAttachment.loadAction = MTLLoadActionLoad; + } + + cbD->d->currentRenderPassEncoder = [cbD->d->cb renderCommandEncoderWithDescriptor: cbD->d->currentPassRpDesc]; + cbD->resetPerPassCachedState(); + + for (uint i = 0; i < uint(rtD->colorAttCount); ++i) { + cbD->d->currentPassRpDesc.colorAttachments[i].loadAction = oldColorLoad[i]; + } + + if (rtD->dsAttCount) { + cbD->d->currentPassRpDesc.depthAttachment.loadAction = oldDepthLoad; + cbD->d->currentPassRpDesc.stencilAttachment.loadAction = oldStencilLoad; + } + +} + +void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) +{ + QMetalCommandBuffer *cbD = args.cbD; + QMetalGraphicsPipeline *graphicsPipeline = cbD->currentGraphicsPipeline; + if (graphicsPipeline->d->tess.failed) + return; + + const bool indexed = args.type != TessDrawArgs::NonIndexed; + const quint32 instanceCount = indexed ? args.drawIndexed.instanceCount : args.draw.instanceCount; + const quint32 vertexOrIndexCount = indexed ? args.drawIndexed.indexCount : args.draw.vertexCount; + + QMetalGraphicsPipelineData::Tessellation &tess(graphicsPipeline->d->tess); + QMetalGraphicsPipelineData::ExtraBufferManager &extraBufMgr(graphicsPipeline->d->extraBufMgr); + const quint32 patchCount = tess.patchCountForDrawCall(vertexOrIndexCount, instanceCount); + QMetalBuffer *vertOutBuf = nullptr; + QMetalBuffer *tescOutBuf = nullptr; + QMetalBuffer *tescPatchOutBuf = nullptr; + QMetalBuffer *tescFactorBuf = nullptr; + QMetalBuffer *tescParamsBuf = nullptr; + id<MTLComputeCommandEncoder> vertTescComputeEncoder = tessellationComputeEncoder(cbD); + + // Step 1: vertex shader (as compute) + { + id<MTLComputeCommandEncoder> computeEncoder = vertTescComputeEncoder; + QShader::Variant shaderVariant = QShader::NonIndexedVertexAsComputeShader; + if (args.type == TessDrawArgs::U16Indexed) + shaderVariant = QShader::UInt16IndexedVertexAsComputeShader; + else if (args.type == TessDrawArgs::U32Indexed) + shaderVariant = QShader::UInt32IndexedVertexAsComputeShader; + const int varIndex = QMetalGraphicsPipelineData::Tessellation::vsCompVariantToIndex(shaderVariant); + id<MTLComputePipelineState> computePipelineState = tess.vsCompPipeline(this, shaderVariant); + [computeEncoder setComputePipelineState: computePipelineState]; + + // Make uniform buffers, textures, and samplers (meant for the + // vertex stage from the client's point of view) visible in the + // "vertex as compute" shader + cbD->d->currentComputePassEncoder = computeEncoder; + rebindShaderResources(cbD, QMetalShaderResourceBindingsData::VERTEX, QMetalShaderResourceBindingsData::COMPUTE); + cbD->d->currentComputePassEncoder = nil; + + const QMap<int, int> &ebb(tess.compVs[varIndex].nativeShaderInfo.extraBufferBindings); + const int outputBufferBinding = ebb.value(QShaderPrivate::MslTessVertTescOutputBufferBinding, -1); + const int indexBufferBinding = ebb.value(QShaderPrivate::MslTessVertIndicesBufferBinding, -1); + + if (outputBufferBinding >= 0) { + const quint32 workBufSize = tess.vsCompOutputBufferSize(vertexOrIndexCount, instanceCount); + vertOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); + if (!vertOutBuf) + return; + [computeEncoder setBuffer: vertOutBuf->d->buf[0] offset: 0 atIndex: outputBufferBinding]; + } + + if (indexBufferBinding >= 0) + [computeEncoder setBuffer: (id<MTLBuffer>) args.drawIndexed.indexBuffer offset: 0 atIndex: indexBufferBinding]; + + for (int i = 0, ie = cbD->d->currentVertexInputsBuffers.batches.count(); i != ie; ++i) { + const auto &bufferBatch(cbD->d->currentVertexInputsBuffers.batches[i]); + const auto &offsetBatch(cbD->d->currentVertexInputOffsets.batches[i]); + [computeEncoder setBuffers: bufferBatch.resources.constData() + offsets: offsetBatch.resources.constData() + withRange: NSMakeRange(uint(cbD->d->currentFirstVertexBinding) + bufferBatch.startBinding, NSUInteger(bufferBatch.resources.count()))]; + } + + if (indexed) { + [computeEncoder setStageInRegion: MTLRegionMake2D(args.drawIndexed.vertexOffset, args.drawIndexed.firstInstance, + args.drawIndexed.indexCount, args.drawIndexed.instanceCount)]; + } else { + [computeEncoder setStageInRegion: MTLRegionMake2D(args.draw.firstVertex, args.draw.firstInstance, + args.draw.vertexCount, args.draw.instanceCount)]; + } + + [computeEncoder dispatchThreads: MTLSizeMake(vertexOrIndexCount, instanceCount, 1) + threadsPerThreadgroup: MTLSizeMake(computePipelineState.threadExecutionWidth, 1, 1)]; + } + + // Step 2: tessellation control shader (as compute) + { + id<MTLComputeCommandEncoder> computeEncoder = vertTescComputeEncoder; + id<MTLComputePipelineState> computePipelineState = tess.tescCompPipeline(this); + [computeEncoder setComputePipelineState: computePipelineState]; + + cbD->d->currentComputePassEncoder = computeEncoder; + rebindShaderResources(cbD, QMetalShaderResourceBindingsData::TESSCTRL, QMetalShaderResourceBindingsData::COMPUTE); + cbD->d->currentComputePassEncoder = nil; + + const QMap<int, int> &ebb(tess.compTesc.nativeShaderInfo.extraBufferBindings); + const int outputBufferBinding = ebb.value(QShaderPrivate::MslTessVertTescOutputBufferBinding, -1); + const int patchOutputBufferBinding = ebb.value(QShaderPrivate::MslTessTescPatchOutputBufferBinding, -1); + const int tessFactorBufferBinding = ebb.value(QShaderPrivate::MslTessTescTessLevelBufferBinding, -1); + const int paramsBufferBinding = ebb.value(QShaderPrivate::MslTessTescParamsBufferBinding, -1); + const int inputBufferBinding = ebb.value(QShaderPrivate::MslTessTescInputBufferBinding, -1); + + if (outputBufferBinding >= 0) { + const quint32 workBufSize = tess.tescCompOutputBufferSize(patchCount); + tescOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); + if (!tescOutBuf) + return; + [computeEncoder setBuffer: tescOutBuf->d->buf[0] offset: 0 atIndex: outputBufferBinding]; + } + + if (patchOutputBufferBinding >= 0) { + const quint32 workBufSize = tess.tescCompPatchOutputBufferSize(patchCount); + tescPatchOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); + if (!tescPatchOutBuf) + return; + [computeEncoder setBuffer: tescPatchOutBuf->d->buf[0] offset: 0 atIndex: patchOutputBufferBinding]; + } + + if (tessFactorBufferBinding >= 0) { + tescFactorBuf = extraBufMgr.acquireWorkBuffer(this, patchCount * sizeof(MTLQuadTessellationFactorsHalf)); + [computeEncoder setBuffer: tescFactorBuf->d->buf[0] offset: 0 atIndex: tessFactorBufferBinding]; + } + + if (paramsBufferBinding >= 0) { + struct { + quint32 inControlPointCount; + quint32 patchCount; + } params; + tescParamsBuf = extraBufMgr.acquireWorkBuffer(this, sizeof(params), QMetalGraphicsPipelineData::ExtraBufferManager::WorkBufType::HostVisible); + if (!tescParamsBuf) + return; + params.inControlPointCount = tess.inControlPointCount; + params.patchCount = patchCount; + id<MTLBuffer> paramsBuf = tescParamsBuf->d->buf[0]; + char *p = reinterpret_cast<char *>([paramsBuf contents]); + memcpy(p, ¶ms, sizeof(params)); + [computeEncoder setBuffer: paramsBuf offset: 0 atIndex: paramsBufferBinding]; + } + + if (vertOutBuf && inputBufferBinding >= 0) + [computeEncoder setBuffer: vertOutBuf->d->buf[0] offset: 0 atIndex: inputBufferBinding]; + + int sgSize = int(computePipelineState.threadExecutionWidth); + int wgSize = std::lcm(tess.outControlPointCount, sgSize); + while (wgSize > caps.maxThreadGroupSize) { + sgSize /= 2; + wgSize = std::lcm(tess.outControlPointCount, sgSize); + } + [computeEncoder dispatchThreads: MTLSizeMake(patchCount * tess.outControlPointCount, 1, 1) + threadsPerThreadgroup: MTLSizeMake(wgSize, 1, 1)]; + } + + // Much of the state in the QMetalCommandBuffer is going to be reset + // when we get a new render encoder. Save what we need. (cheaper than + // starting to walk over the srb again) + const QMetalShaderResourceBindingsData resourceBindings = cbD->d->currentShaderResourceBindingState; + + endTessellationComputeEncoding(cbD); + + // Step 3: tessellation evaluation (as vertex) + fragment shader + { + // No need to call tess.teseFragRenderPipeline because it was done + // once and we know the result is stored in the standard place + // (graphicsPipeline->d->ps). + + graphicsPipeline->makeActiveForCurrentRenderPassEncoder(cbD); + id<MTLRenderCommandEncoder> renderEncoder = cbD->d->currentRenderPassEncoder; + + rebindShaderResources(cbD, QMetalShaderResourceBindingsData::TESSEVAL, QMetalShaderResourceBindingsData::VERTEX, &resourceBindings); + rebindShaderResources(cbD, QMetalShaderResourceBindingsData::FRAGMENT, QMetalShaderResourceBindingsData::FRAGMENT, &resourceBindings); + + const QMap<int, int> &ebb(tess.compTesc.nativeShaderInfo.extraBufferBindings); + const int outputBufferBinding = ebb.value(QShaderPrivate::MslTessVertTescOutputBufferBinding, -1); + const int patchOutputBufferBinding = ebb.value(QShaderPrivate::MslTessTescPatchOutputBufferBinding, -1); + const int tessFactorBufferBinding = ebb.value(QShaderPrivate::MslTessTescTessLevelBufferBinding, -1); + + if (outputBufferBinding >= 0 && tescOutBuf) + [renderEncoder setVertexBuffer: tescOutBuf->d->buf[0] offset: 0 atIndex: outputBufferBinding]; + + if (patchOutputBufferBinding >= 0 && tescPatchOutBuf) + [renderEncoder setVertexBuffer: tescPatchOutBuf->d->buf[0] offset: 0 atIndex: patchOutputBufferBinding]; + + if (tessFactorBufferBinding >= 0 && tescFactorBuf) { + [renderEncoder setTessellationFactorBuffer: tescFactorBuf->d->buf[0] offset: 0 instanceStride: 0]; + [renderEncoder setVertexBuffer: tescFactorBuf->d->buf[0] offset: 0 atIndex: tessFactorBufferBinding]; + } + + [cbD->d->currentRenderPassEncoder drawPatches: tess.outControlPointCount + patchStart: 0 + patchCount: patchCount + patchIndexBuffer: nil + patchIndexBufferOffset: 0 + instanceCount: 1 + baseInstance: 0]; + } +} + +void QRhiMetal::adjustForMultiViewDraw(quint32 *instanceCount, QRhiCommandBuffer *cb) +{ + QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); + const int multiViewCount = cbD->currentGraphicsPipeline->m_multiViewCount; + if (multiViewCount <= 1) + return; + + const QMap<int, int> &ebb(cbD->currentGraphicsPipeline->d->vs.nativeShaderInfo.extraBufferBindings); + const int viewMaskBufBinding = ebb.value(QShaderPrivate::MslMultiViewMaskBufferBinding, -1); + if (viewMaskBufBinding == -1) { + qWarning("No extra buffer for multiview in the vertex shader; was it built with --view-count specified?"); + return; + } + struct { + quint32 viewOffset; + quint32 viewCount; + } multiViewInfo; + multiViewInfo.viewOffset = 0; + multiViewInfo.viewCount = quint32(multiViewCount); + QMetalBuffer *buf = cbD->currentGraphicsPipeline->d->extraBufMgr.acquireWorkBuffer(this, sizeof(multiViewInfo), + QMetalGraphicsPipelineData::ExtraBufferManager::WorkBufType::HostVisible); + if (buf) { + id<MTLBuffer> mtlbuf = buf->d->buf[0]; + char *p = reinterpret_cast<char *>([mtlbuf contents]); + memcpy(p, &multiViewInfo, sizeof(multiViewInfo)); + [cbD->d->currentRenderPassEncoder setVertexBuffer: mtlbuf offset: 0 atIndex: viewMaskBufBinding]; + // The instance count is adjusted for layered rendering. The vertex shader is expected to contain something like: + // uint gl_ViewIndex = spvViewMask[0] + (gl_InstanceIndex - gl_BaseInstance) % spvViewMask[1]; + // where spvViewMask is the buffer with multiViewInfo passed in above. + *instanceCount *= multiViewCount; + } +} + void QRhiMetal::draw(QRhiCommandBuffer *cb, quint32 vertexCount, quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) { QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); Q_ASSERT(cbD->recordingPass == QMetalCommandBuffer::RenderPass); - [cbD->d->currentRenderPassEncoder drawPrimitives: - QRHI_RES(QMetalGraphicsPipeline, cbD->currentGraphicsPipeline)->d->primitiveType - vertexStart: firstVertex vertexCount: vertexCount instanceCount: instanceCount baseInstance: firstInstance]; + if (cbD->currentGraphicsPipeline->d->tess.enabled) { + TessDrawArgs a; + a.cbD = cbD; + a.type = TessDrawArgs::NonIndexed; + a.draw.vertexCount = vertexCount; + a.draw.instanceCount = instanceCount; + a.draw.firstVertex = firstVertex; + a.draw.firstInstance = firstInstance; + tessellatedDraw(a); + return; + } + + adjustForMultiViewDraw(&instanceCount, cb); + + if (caps.baseVertexAndInstance) { + [cbD->d->currentRenderPassEncoder drawPrimitives: cbD->currentGraphicsPipeline->d->primitiveType + vertexStart: firstVertex vertexCount: vertexCount instanceCount: instanceCount baseInstance: firstInstance]; + } else { + [cbD->d->currentRenderPassEncoder drawPrimitives: cbD->currentGraphicsPipeline->d->primitiveType + vertexStart: firstVertex vertexCount: vertexCount instanceCount: instanceCount]; + } } void QRhiMetal::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, @@ -1328,19 +2241,44 @@ void QRhiMetal::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, return; const quint32 indexOffset = cbD->currentIndexOffset + firstIndex * (cbD->currentIndexFormat == QRhiCommandBuffer::IndexUInt16 ? 2 : 4); - Q_ASSERT(indexOffset == aligned<quint32>(indexOffset, 4)); - - QMetalBuffer *ibufD = QRHI_RES(QMetalBuffer, cbD->currentIndexBuffer); - id<MTLBuffer> mtlbuf = ibufD->d->buf[ibufD->d->slotted ? currentFrameSlot : 0]; + Q_ASSERT(indexOffset == aligned(indexOffset, 4u)); + + QMetalBuffer *ibufD = cbD->currentIndexBuffer; + id<MTLBuffer> mtlibuf = ibufD->d->buf[ibufD->d->slotted ? currentFrameSlot : 0]; + + if (cbD->currentGraphicsPipeline->d->tess.enabled) { + TessDrawArgs a; + a.cbD = cbD; + a.type = cbD->currentIndexFormat == QRhiCommandBuffer::IndexUInt16 ? TessDrawArgs::U16Indexed : TessDrawArgs::U32Indexed; + a.drawIndexed.indexCount = indexCount; + a.drawIndexed.instanceCount = instanceCount; + a.drawIndexed.firstIndex = firstIndex; + a.drawIndexed.vertexOffset = vertexOffset; + a.drawIndexed.firstInstance = firstInstance; + a.drawIndexed.indexBuffer = mtlibuf; + tessellatedDraw(a); + return; + } - [cbD->d->currentRenderPassEncoder drawIndexedPrimitives: QRHI_RES(QMetalGraphicsPipeline, cbD->currentGraphicsPipeline)->d->primitiveType - indexCount: indexCount - indexType: cbD->currentIndexFormat == QRhiCommandBuffer::IndexUInt16 ? MTLIndexTypeUInt16 : MTLIndexTypeUInt32 - indexBuffer: mtlbuf - indexBufferOffset: indexOffset - instanceCount: instanceCount - baseVertex: vertexOffset - baseInstance: firstInstance]; + adjustForMultiViewDraw(&instanceCount, cb); + + if (caps.baseVertexAndInstance) { + [cbD->d->currentRenderPassEncoder drawIndexedPrimitives: cbD->currentGraphicsPipeline->d->primitiveType + indexCount: indexCount + indexType: cbD->currentIndexFormat == QRhiCommandBuffer::IndexUInt16 ? MTLIndexTypeUInt16 : MTLIndexTypeUInt32 + indexBuffer: mtlibuf + indexBufferOffset: indexOffset + instanceCount: instanceCount + baseVertex: vertexOffset + baseInstance: firstInstance]; + } else { + [cbD->d->currentRenderPassEncoder drawIndexedPrimitives: cbD->currentGraphicsPipeline->d->primitiveType + indexCount: indexCount + indexType: cbD->currentIndexFormat == QRhiCommandBuffer::IndexUInt16 ? MTLIndexTypeUInt16 : MTLIndexTypeUInt32 + indexBuffer: mtlibuf + indexBufferOffset: indexOffset + instanceCount: instanceCount]; + } } void QRhiMetal::debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) @@ -1394,34 +2332,39 @@ void QRhiMetal::endExternal(QRhiCommandBuffer *cb) cbD->resetPerPassCachedState(); } +double QRhiMetal::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); + return cbD->d->lastGpuTime; +} + QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); QMetalSwapChain *swapChainD = QRHI_RES(QMetalSwapChain, swapChain); + currentSwapChain = swapChainD; + currentFrameSlot = swapChainD->currentFrameSlot; - // This is a bit messed up since for this swapchain we want to wait for the - // commands+present to complete, while for others just for the commands - // (for this same frame slot) but not sure how to do that in a sane way so - // wait for full cb completion for now. - for (QMetalSwapChain *sc : qAsConst(swapchains)) { - dispatch_semaphore_t sem = sc->d->sem[swapChainD->currentFrameSlot]; - dispatch_semaphore_wait(sem, DISPATCH_TIME_FOREVER); + // If we are too far ahead, block. This is also what ensures that any + // resource used in the previous frame for this slot is now not in use + // anymore by the GPU. + dispatch_semaphore_wait(swapChainD->d->sem[currentFrameSlot], DISPATCH_TIME_FOREVER); + + // Do this also for any other swapchain's commands with the same frame slot + // While this reduces concurrency, it keeps resource usage safe: swapchain + // A starting its frame 0, followed by swapchain B starting its own frame 0 + // will make B wait for A's frame 0 commands, so if a resource is written + // in B's frame or when B checks for pending resource releases, that won't + // mess up A's in-flight commands (as they are not in flight anymore). + for (QMetalSwapChain *sc : std::as_const(swapchains)) { if (sc != swapChainD) - dispatch_semaphore_signal(sem); + sc->waitUntilCompleted(currentFrameSlot); // wait+signal } - currentSwapChain = swapChainD; - currentFrameSlot = swapChainD->currentFrameSlot; - if (swapChainD->ds) - swapChainD->ds->lastActiveFrameSlot = currentFrameSlot; - [d->captureScope beginScope]; - // Do not let the command buffer mess with the refcount of objects. We do - // have a proper render loop and will manage lifetimes similarly to other - // backends (Vulkan). - swapChainD->cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + swapChainD->cbWrapper.d->cb = d->newCommandBuffer(); QMetalRenderTargetData::ColorAtt colorAtt; if (swapChainD->samples > 1) { @@ -1433,14 +2376,16 @@ QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF swapChainD->rtWrapper.d->fb.colorAtt[0] = colorAtt; swapChainD->rtWrapper.d->fb.dsTex = swapChainD->ds ? swapChainD->ds->d->tex : nil; + swapChainD->rtWrapper.d->fb.dsResolveTex = nil; swapChainD->rtWrapper.d->fb.hasStencil = swapChainD->ds ? true : false; swapChainD->rtWrapper.d->fb.depthNeedsStore = false; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - QRHI_PROF_F(beginSwapChainFrame(swapChain)); + if (swapChainD->ds) + swapChainD->ds->lastActiveFrameSlot = currentFrameSlot; executeDeferredReleases(); - swapChainD->cbWrapper.resetState(); + swapChainD->cbWrapper.resetState(swapChainD->d->lastGpuTime[currentFrameSlot]); + swapChainD->d->lastGpuTime[currentFrameSlot] = 0; finishActiveReadbacks(); return QRhi::FrameOpSuccess; @@ -1451,28 +2396,47 @@ QRhi::FrameOpResult QRhiMetal::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame QMetalSwapChain *swapChainD = QRHI_RES(QMetalSwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); + __block int thisFrameSlot = currentFrameSlot; + [swapChainD->cbWrapper.d->cb addCompletedHandler: ^(id<MTLCommandBuffer> cb) { + swapChainD->d->lastGpuTime[thisFrameSlot] += cb.GPUEndTime - cb.GPUStartTime; + dispatch_semaphore_signal(swapChainD->d->sem[thisFrameSlot]); + }]; + +#ifdef QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES + // When Metal API validation diagnostics is enabled in Xcode the texture is + // released before the command buffer is done with it. Manually keep it alive + // to work around this. + id<MTLTexture> drawableTexture = [swapChainD->d->curDrawable.texture retain]; + [swapChainD->cbWrapper.d->cb addCompletedHandler:^(id<MTLCommandBuffer>) { + [drawableTexture release]; + }]; +#endif + const bool needsPresent = !flags.testFlag(QRhi::SkipPresent); - if (needsPresent) { - auto drawable = swapChainD->d->curDrawable; - [swapChainD->cbWrapper.d->cb addScheduledHandler:^(id<MTLCommandBuffer>) { + const bool presentsWithTransaction = swapChainD->d->layer.presentsWithTransaction; + if (!presentsWithTransaction && needsPresent) { + // beginFrame-endFrame without a render pass inbetween means there is no drawable. + if (id<CAMetalDrawable> drawable = swapChainD->d->curDrawable) + [swapChainD->cbWrapper.d->cb presentDrawable: drawable]; + } + + [swapChainD->cbWrapper.d->cb commit]; + + if (presentsWithTransaction && needsPresent) { + // beginFrame-endFrame without a render pass inbetween means there is no drawable. + if (id<CAMetalDrawable> drawable = swapChainD->d->curDrawable) { + // The layer has presentsWithTransaction set to true to avoid flicker on resizing, + // so here it is important to follow what the Metal docs say when it comes to the + // issuing the present. + [swapChainD->cbWrapper.d->cb waitUntilScheduled]; [drawable present]; - }]; + } } // Must not hold on to the drawable, regardless of needsPresent [swapChainD->d->curDrawable release]; swapChainD->d->curDrawable = nil; - __block int thisFrameSlot = currentFrameSlot; - [swapChainD->cbWrapper.d->cb addCompletedHandler: ^(id<MTLCommandBuffer>) { - dispatch_semaphore_signal(swapChainD->d->sem[thisFrameSlot]); - }]; - - [swapChainD->cbWrapper.d->cb commit]; - - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - QRHI_PROF_F(endSwapChainFrame(swapChain, swapChainD->frameCount + 1)); - [d->captureScope endScope]; if (needsPresent) @@ -1488,23 +2452,17 @@ QRhi::FrameOpResult QRhiMetal::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: Q_UNUSED(flags); currentFrameSlot = (currentFrameSlot + 1) % QMTL_FRAMES_IN_FLIGHT; - if (swapchains.count() > 1) { - for (QMetalSwapChain *sc : qAsConst(swapchains)) { - // wait+signal is the general pattern to ensure the commands for a - // given frame slot have completed (if sem is 1, we go 0 then 1; if - // sem is 0 we go -1, block, completion increments to 0, then us to 1) - dispatch_semaphore_t sem = sc->d->sem[currentFrameSlot]; - dispatch_semaphore_wait(sem, DISPATCH_TIME_FOREVER); - dispatch_semaphore_signal(sem); - } - } + + for (QMetalSwapChain *sc : std::as_const(swapchains)) + sc->waitUntilCompleted(currentFrameSlot); d->ofr.active = true; *cb = &d->ofr.cbWrapper; - d->ofr.cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + d->ofr.cbWrapper.d->cb = d->newCommandBuffer(); executeDeferredReleases(); - d->ofr.cbWrapper.resetState(); + d->ofr.cbWrapper.resetState(d->ofr.lastGpuTime); + d->ofr.lastGpuTime = 0; finishActiveReadbacks(); return QRhi::FrameOpSuccess; @@ -1516,10 +2474,13 @@ QRhi::FrameOpResult QRhiMetal::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_ASSERT(d->ofr.active); d->ofr.active = false; - [d->ofr.cbWrapper.d->cb commit]; + id<MTLCommandBuffer> cb = d->ofr.cbWrapper.d->cb; + [cb commit]; // offscreen frames wait for completion, unlike swapchain ones - [d->ofr.cbWrapper.d->cb waitUntilCompleted]; + [cb waitUntilCompleted]; + + d->ofr.lastGpuTime += cb.GPUEndTime - cb.GPUStartTime; finishActiveReadbacks(true); @@ -1543,16 +2504,14 @@ QRhi::FrameOpResult QRhiMetal::finish() } } - for (QMetalSwapChain *sc : qAsConst(swapchains)) { + for (QMetalSwapChain *sc : std::as_const(swapchains)) { for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) { if (currentSwapChain && sc == currentSwapChain && i == currentFrameSlot) { // no wait as this is the thing we're going to be commit below and // beginFrame decremented sem already and going to be signaled by endFrame continue; } - dispatch_semaphore_t sem = sc->d->sem[i]; - dispatch_semaphore_wait(sem, DISPATCH_TIME_FOREVER); - dispatch_semaphore_signal(sem); + sc->waitUntilCompleted(i); } } @@ -1562,10 +2521,13 @@ QRhi::FrameOpResult QRhiMetal::finish() } if (inFrame) { - if (d->ofr.active) - d->ofr.cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; - else - swapChainD->cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + if (d->ofr.active) { + d->ofr.lastGpuTime += cb.GPUEndTime - cb.GPUStartTime; + d->ofr.cbWrapper.d->cb = d->newCommandBuffer(); + } else { + swapChainD->d->lastGpuTime[currentFrameSlot] += cb.GPUEndTime - cb.GPUStartTime; + swapChainD->cbWrapper.d->cb = d->newCommandBuffer(); + } } executeDeferredReleases(true); @@ -1742,7 +2704,15 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate { QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); + + id<MTLBlitCommandEncoder> blitEnc = nil; + auto ensureBlit = [&blitEnc, cbD, this]() { + if (!blitEnc) { + blitEnc = [cbD->d->cb blitCommandEncoder]; + if (debugMarkers) + [blitEnc pushDebugGroup: @"Texture upload/copy"]; + } + }; for (int opIdx = 0; opIdx < ud->activeBufferOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::BufferOp &u(ud->bufferOps[opIdx]); @@ -1766,25 +2736,33 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate QMetalBuffer *bufD = QRHI_RES(QMetalBuffer, u.buf); executeBufferHostWritesForCurrentFrame(bufD); const int idx = bufD->d->slotted ? currentFrameSlot : 0; - char *p = reinterpret_cast<char *>([bufD->d->buf[idx] contents]); - if (p) { - u.result->data.resize(u.readSize); - memcpy(u.result->data.data(), p + u.offset, size_t(u.readSize)); + if (bufD->m_type == QRhiBuffer::Dynamic) { + char *p = reinterpret_cast<char *>([bufD->d->buf[idx] contents]); + if (p) { + u.result->data.resize(u.readSize); + memcpy(u.result->data.data(), p + u.offset, size_t(u.readSize)); + } + if (u.result->completed) + u.result->completed(); + } else { + QRhiMetalData::BufferReadback readback; + readback.activeFrameSlot = idx; + readback.buf = bufD->d->buf[idx]; + readback.offset = u.offset; + readback.readSize = u.readSize; + readback.result = u.result; + d->activeBufferReadbacks.append(readback); +#ifdef Q_OS_MACOS + if (bufD->d->managed) { + // On non-Apple Silicon, manually synchronize memory from GPU to CPU + ensureBlit(); + [blitEnc synchronizeResource:readback.buf]; + } +#endif } - if (u.result->completed) - u.result->completed(); } } - id<MTLBlitCommandEncoder> blitEnc = nil; - auto ensureBlit = [&blitEnc, cbD, this] { - if (!blitEnc) { - blitEnc = [cbD->d->cb blitCommandEncoder]; - if (debugMarkers) - [blitEnc pushDebugGroup: @"Texture upload/copy"]; - } - }; - for (int opIdx = 0; opIdx < ud->activeTextureOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::TextureOp &u(ud->textureOps[opIdx]); if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Upload) { @@ -1792,7 +2770,7 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate qsizetype stagingSize = 0; for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) stagingSize += subresUploadByteSize(subresDesc); } } @@ -1801,13 +2779,12 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate Q_ASSERT(!utexD->d->stagingBuf[currentFrameSlot]); utexD->d->stagingBuf[currentFrameSlot] = [d->dev newBufferWithLength: NSUInteger(stagingSize) options: MTLResourceStorageModeShared]; - QRHI_PROF_F(newTextureStagingArea(utexD, currentFrameSlot, quint32(stagingSize))); void *mp = [utexD->d->stagingBuf[currentFrameSlot] contents]; qsizetype curOfs = 0; for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) enqueueSubresUpload(utexD, mp, blitEnc, layer, level, subresDesc, &curOfs); } } @@ -1820,7 +2797,6 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate e.stagingBuffer.buffer = utexD->d->stagingBuf[currentFrameSlot]; utexD->d->stagingBuf[currentFrameSlot] = nil; d->releaseQueue.append(e); - QRHI_PROF_F(releaseTextureStagingArea(utexD, currentFrameSlot)); } else if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Copy) { Q_ASSERT(u.src && u.dst); QMetalTexture *srcD = QRHI_RES(QMetalTexture, u.src); @@ -1882,10 +2858,6 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate textureFormatInfo(readback.format, readback.pixelSize, &bpl, &readback.bufSize, nullptr); readback.buf = [d->dev newBufferWithLength: readback.bufSize options: MTLResourceStorageModeShared]; - QRHI_PROF_F(newReadbackBuffer(qint64(qintptr(readback.buf)), - texD ? static_cast<QRhiResource *>(texD) : static_cast<QRhiResource *>(swapChainD), - readback.bufSize)); - ensureBlit(); [blitEnc copyFromTexture: src sourceSlice: NSUInteger(is3D ? 0 : u.rb.layer()) @@ -1923,17 +2895,17 @@ void QRhiMetal::executeBufferHostWritesForSlot(QMetalBuffer *bufD, int slot) return; void *p = [bufD->d->buf[slot] contents]; - int changeBegin = -1; - int changeEnd = -1; - for (const QMetalBufferData::BufferUpdate &u : qAsConst(bufD->d->pendingUpdates[slot])) { + quint32 changeBegin = UINT32_MAX; + quint32 changeEnd = 0; + for (const QMetalBufferData::BufferUpdate &u : std::as_const(bufD->d->pendingUpdates[slot])) { memcpy(static_cast<char *>(p) + u.offset, u.data.constData(), size_t(u.data.size())); - if (changeBegin == -1 || u.offset < changeBegin) + if (u.offset < changeBegin) changeBegin = u.offset; - if (changeEnd == -1 || u.offset + u.data.size() > changeEnd) + if (u.offset + u.data.size() > changeEnd) changeEnd = u.offset + u.data.size(); } #ifdef Q_OS_MACOS - if (changeBegin >= 0 && bufD->d->managed) + if (changeBegin < UINT32_MAX && changeBegin < changeEnd && bufD->d->managed) [bufD->d->buf[slot] didModifyRange: NSMakeRange(NSUInteger(changeBegin), NSUInteger(changeEnd - changeBegin))]; #endif @@ -1967,8 +2939,8 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, QMetalRenderTargetData *rtD = nullptr; switch (rt->resourceType()) { - case QRhiResource::RenderTarget: - rtD = QRHI_RES(QMetalReferenceRenderTarget, rt)->d; + case QRhiResource::SwapChainRenderTarget: + rtD = QRHI_RES(QMetalSwapChainRenderTarget, rt)->d; cbD->d->currentPassRpDesc = d->createDefaultRenderPass(rtD->dsAttCount, colorClearValue, depthStencilClearValue, rtD->colorAttCount); if (rtD->colorAttCount) { QMetalRenderTargetData::ColorAtt &color0(rtD->fb.colorAtt[0]); @@ -1998,22 +2970,27 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, { QMetalTextureRenderTarget *rtTex = QRHI_RES(QMetalTextureRenderTarget, rt); rtD = rtTex->d; + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QMetalTexture, QMetalRenderBuffer>(rtTex->description(), rtD->currentResIdList)) + rtTex->create(); cbD->d->currentPassRpDesc = d->createDefaultRenderPass(rtD->dsAttCount, colorClearValue, depthStencilClearValue, rtD->colorAttCount); - if (rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents)) { + if (rtD->fb.preserveColor) { for (uint i = 0; i < uint(rtD->colorAttCount); ++i) cbD->d->currentPassRpDesc.colorAttachments[i].loadAction = MTLLoadActionLoad; } - if (rtD->dsAttCount && rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents)) { + if (rtD->dsAttCount && rtD->fb.preserveDs) { cbD->d->currentPassRpDesc.depthAttachment.loadAction = MTLLoadActionLoad; cbD->d->currentPassRpDesc.stencilAttachment.loadAction = MTLLoadActionLoad; } for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); it != itEnd; ++it) { - if (it->texture()) + if (it->texture()) { QRHI_RES(QMetalTexture, it->texture())->lastActiveFrameSlot = currentFrameSlot; - else if (it->renderBuffer()) + if (it->multiViewCount() >= 2) + cbD->d->currentPassRpDesc.renderTargetArrayLength = NSUInteger(it->multiViewCount()); + } else if (it->renderBuffer()) { QRHI_RES(QMetalRenderBuffer, it->renderBuffer())->lastActiveFrameSlot = currentFrameSlot; + } if (it->resolveTexture()) QRHI_RES(QMetalTexture, it->resolveTexture())->lastActiveFrameSlot = currentFrameSlot; } @@ -2021,6 +2998,8 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, QRHI_RES(QMetalRenderBuffer, rtTex->m_desc.depthStencilBuffer())->lastActiveFrameSlot = currentFrameSlot; if (rtTex->m_desc.depthTexture()) QRHI_RES(QMetalTexture, rtTex->m_desc.depthTexture())->lastActiveFrameSlot = currentFrameSlot; + if (rtTex->m_desc.depthResolveTexture()) + QRHI_RES(QMetalTexture, rtTex->m_desc.depthResolveTexture())->lastActiveFrameSlot = currentFrameSlot; } break; default: @@ -2034,7 +3013,8 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, cbD->d->currentPassRpDesc.colorAttachments[i].depthPlane = NSUInteger(rtD->fb.colorAtt[i].slice); cbD->d->currentPassRpDesc.colorAttachments[i].level = NSUInteger(rtD->fb.colorAtt[i].level); if (rtD->fb.colorAtt[i].resolveTex) { - cbD->d->currentPassRpDesc.colorAttachments[i].storeAction = MTLStoreActionMultisampleResolve; + cbD->d->currentPassRpDesc.colorAttachments[i].storeAction = rtD->fb.preserveColor ? MTLStoreActionStoreAndMultisampleResolve + : MTLStoreActionMultisampleResolve; cbD->d->currentPassRpDesc.colorAttachments[i].resolveTexture = rtD->fb.colorAtt[i].resolveTex; cbD->d->currentPassRpDesc.colorAttachments[i].resolveSlice = NSUInteger(rtD->fb.colorAtt[i].resolveLayer); cbD->d->currentPassRpDesc.colorAttachments[i].resolveLevel = NSUInteger(rtD->fb.colorAtt[i].resolveLevel); @@ -2047,6 +3027,15 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, cbD->d->currentPassRpDesc.stencilAttachment.texture = rtD->fb.hasStencil ? rtD->fb.dsTex : nil; if (rtD->fb.depthNeedsStore) // Depth/Stencil is set to DontCare by default, override if needed cbD->d->currentPassRpDesc.depthAttachment.storeAction = MTLStoreActionStore; + if (rtD->fb.dsResolveTex) { + cbD->d->currentPassRpDesc.depthAttachment.storeAction = rtD->fb.depthNeedsStore ? MTLStoreActionStoreAndMultisampleResolve + : MTLStoreActionMultisampleResolve; + cbD->d->currentPassRpDesc.depthAttachment.resolveTexture = rtD->fb.dsResolveTex; + if (rtD->fb.hasStencil) { + cbD->d->currentPassRpDesc.stencilAttachment.resolveTexture = rtD->fb.dsResolveTex; + cbD->d->currentPassRpDesc.stencilAttachment.storeAction = cbD->d->currentPassRpDesc.depthAttachment.storeAction; + } + } } cbD->d->currentRenderPassEncoder = [cbD->d->cb renderCommandEncoderWithDescriptor: cbD->d->currentPassRpDesc]; @@ -2104,9 +3093,9 @@ void QRhiMetal::setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *p Q_ASSERT(cbD->recordingPass == QMetalCommandBuffer::ComputePass); QMetalComputePipeline *psD = QRHI_RES(QMetalComputePipeline, ps); - if (cbD->currentComputePipeline != ps || cbD->currentPipelineGeneration != psD->generation) { + if (cbD->currentComputePipeline != psD || cbD->currentPipelineGeneration != psD->generation) { cbD->currentGraphicsPipeline = nullptr; - cbD->currentComputePipeline = ps; + cbD->currentComputePipeline = psD; cbD->currentPipelineGeneration = psD->generation; [cbD->d->currentComputePassEncoder setComputePipelineState: psD->d->ps]; @@ -2171,6 +3160,17 @@ void QRhiMetal::executeDeferredReleases(bool forced) case QRhiMetalData::DeferredReleaseEntry::StagingBuffer: [e.stagingBuffer.buffer release]; break; + case QRhiMetalData::DeferredReleaseEntry::GraphicsPipeline: + [e.graphicsPipeline.pipelineState release]; + [e.graphicsPipeline.depthStencilState release]; + [e.graphicsPipeline.tessVertexComputeState[0] release]; + [e.graphicsPipeline.tessVertexComputeState[1] release]; + [e.graphicsPipeline.tessVertexComputeState[2] release]; + [e.graphicsPipeline.tessTessControlComputeState release]; + break; + case QRhiMetalData::DeferredReleaseEntry::ComputePipeline: + [e.computePipeline.pipelineState release]; + break; default: break; } @@ -2182,7 +3182,6 @@ void QRhiMetal::executeDeferredReleases(bool forced) void QRhiMetal::finishActiveReadbacks(bool forced) { QVarLengthArray<std::function<void()>, 4> completedCallbacks; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); for (int i = d->activeTextureReadbacks.count() - 1; i >= 0; --i) { const QRhiMetalData::TextureReadback &readback(d->activeTextureReadbacks[i]); @@ -2194,12 +3193,26 @@ void QRhiMetal::finishActiveReadbacks(bool forced) memcpy(readback.result->data.data(), p, readback.bufSize); [readback.buf release]; - QRHI_PROF_F(releaseReadbackBuffer(qint64(qintptr(readback.buf)))); + if (readback.result->completed) + completedCallbacks.append(readback.result->completed); + + d->activeTextureReadbacks.remove(i); + } + } + + for (int i = d->activeBufferReadbacks.count() - 1; i >= 0; --i) { + const QRhiMetalData::BufferReadback &readback(d->activeBufferReadbacks[i]); + if (forced || currentFrameSlot == readback.activeFrameSlot + || readback.activeFrameSlot < 0) { + readback.result->data.resize(readback.readSize); + char *p = reinterpret_cast<char *>([readback.buf contents]); + Q_ASSERT(p); + memcpy(readback.result->data.data(), p + readback.offset, size_t(readback.readSize)); if (readback.result->completed) completedCallbacks.append(readback.result->completed); - d->activeTextureReadbacks.removeLast(); + d->activeBufferReadbacks.remove(i); } } @@ -2207,7 +3220,7 @@ void QRhiMetal::finishActiveReadbacks(bool forced) f(); } -QMetalBuffer::QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size) +QMetalBuffer::QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) : QRhiBuffer(rhi, type, usage, size), d(new QMetalBufferData) { @@ -2237,10 +3250,10 @@ void QMetalBuffer::destroy() } QRHI_RES_RHI(QRhiMetal); - rhiD->d->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QMetalBuffer::create() @@ -2253,13 +3266,15 @@ bool QMetalBuffer::create() return false; } - const uint nonZeroSize = m_size <= 0 ? 256 : uint(m_size); - const uint roundedSize = m_usage.testFlag(QRhiBuffer::UniformBuffer) ? aligned<uint>(nonZeroSize, 256) : nonZeroSize; + const quint32 nonZeroSize = m_size <= 0 ? 256 : m_size; + const quint32 roundedSize = m_usage.testFlag(QRhiBuffer::UniformBuffer) ? aligned(nonZeroSize, 256u) : nonZeroSize; d->managed = false; MTLResourceOptions opts = MTLResourceStorageModeShared; + + QRHI_RES_RHI(QRhiMetal); #ifdef Q_OS_MACOS - if (m_type != Dynamic) { + if (!rhiD->caps.isAppleGPU && m_type != Dynamic) { opts = MTLResourceStorageModeManaged; d->managed = true; } @@ -2270,8 +3285,10 @@ bool QMetalBuffer::create() // Static maps to on macOS) is not safe when another frame reading from the // same buffer is still in flight. d->slotted = !m_usage.testFlag(QRhiBuffer::StorageBuffer); // except for SSBOs written in the shader + // and a special case for internal work buffers + if (int(m_usage) == WorkBufPoolUsage) + d->slotted = false; - QRHI_RES_RHI(QRhiMetal); for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) { if (i == 0 || d->slotted) { d->buf[i] = [rhiD->d->dev newBufferWithLength: roundedSize options: opts]; @@ -2286,9 +3303,6 @@ bool QMetalBuffer::create() } } - QRHI_PROF; - QRHI_PROF_F(newBuffer(this, roundedSize, d->slotted ? QMTL_FRAMES_IN_FLIGHT : 1, 0)); - lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -2337,11 +3351,12 @@ void QMetalBuffer::endFullDynamicBufferUpdateForCurrentFrame() #endif } -static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QRhiTexture::Flags flags, const QRhiMetalData *d) +static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QRhiTexture::Flags flags, const QRhiMetal *d) { #ifndef Q_OS_MACOS Q_UNUSED(d); #endif + const bool srgb = flags.testFlag(QRhiTexture::sRGB); switch (format) { case QRhiTexture::RGBA8: @@ -2376,13 +3391,16 @@ static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QR case QRhiTexture::R32F: return MTLPixelFormatR32Float; + case QRhiTexture::RGB10A2: + return MTLPixelFormatRGB10A2Unorm; + #ifdef Q_OS_MACOS case QRhiTexture::D16: return MTLPixelFormatDepth16Unorm; case QRhiTexture::D24: - return [d->dev isDepth24Stencil8PixelFormatSupported] ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float; + return [d->d->dev isDepth24Stencil8PixelFormatSupported] ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float; case QRhiTexture::D24S8: - return [d->dev isDepth24Stencil8PixelFormatSupported] ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; + return [d->d->dev isDepth24Stencil8PixelFormatSupported] ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; #else case QRhiTexture::D16: return MTLPixelFormatDepth32Float; @@ -2405,7 +3423,7 @@ static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QR return MTLPixelFormatBC4_RUnorm; case QRhiTexture::BC5: qWarning("QRhiMetal does not support BC5"); - return MTLPixelFormatRGBA8Unorm; + return MTLPixelFormatInvalid; case QRhiTexture::BC6H: return MTLPixelFormatBC6H_RGBUfloat; case QRhiTexture::BC7: @@ -2419,7 +3437,7 @@ static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QR case QRhiTexture::BC6H: case QRhiTexture::BC7: qWarning("QRhiMetal: BCx compression not supported on this platform"); - return MTLPixelFormatRGBA8Unorm; + return MTLPixelFormatInvalid; #endif #ifndef Q_OS_MACOS @@ -2460,32 +3478,129 @@ static inline MTLPixelFormat toMetalTextureFormat(QRhiTexture::Format format, QR return srgb ? MTLPixelFormatASTC_12x12_sRGB : MTLPixelFormatASTC_12x12_LDR; #else case QRhiTexture::ETC2_RGB8: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatETC2_RGB8_sRGB : MTLPixelFormatETC2_RGB8; + } + qWarning("QRhiMetal: ETC2 compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ETC2_RGB8A1: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatETC2_RGB8A1_sRGB : MTLPixelFormatETC2_RGB8A1; + } + qWarning("QRhiMetal: ETC2 compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ETC2_RGBA8: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatEAC_RGBA8_sRGB : MTLPixelFormatEAC_RGBA8; + } qWarning("QRhiMetal: ETC2 compression not supported on this platform"); - return MTLPixelFormatRGBA8Unorm; - + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_4x4: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_4x4_sRGB : MTLPixelFormatASTC_4x4_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_5x4: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_5x4_sRGB : MTLPixelFormatASTC_5x4_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_5x5: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_5x5_sRGB : MTLPixelFormatASTC_5x5_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_6x5: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_6x5_sRGB : MTLPixelFormatASTC_6x5_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_6x6: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_6x6_sRGB : MTLPixelFormatASTC_6x6_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_8x5: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_8x5_sRGB : MTLPixelFormatASTC_8x5_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_8x6: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_8x6_sRGB : MTLPixelFormatASTC_8x6_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_8x8: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_8x8_sRGB : MTLPixelFormatASTC_8x8_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_10x5: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_10x5_sRGB : MTLPixelFormatASTC_10x5_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_10x6: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_10x6_sRGB : MTLPixelFormatASTC_10x6_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_10x8: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_10x8_sRGB : MTLPixelFormatASTC_10x8_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_10x10: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_10x10_sRGB : MTLPixelFormatASTC_10x10_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_12x10: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_12x10_sRGB : MTLPixelFormatASTC_12x10_LDR; + } + qWarning("QRhiMetal: ASTC compression not supported on this platform"); + return MTLPixelFormatInvalid; case QRhiTexture::ASTC_12x12: + if (d->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) + return srgb ? MTLPixelFormatASTC_12x12_sRGB : MTLPixelFormatASTC_12x12_LDR; + } qWarning("QRhiMetal: ASTC compression not supported on this platform"); - return MTLPixelFormatRGBA8Unorm; + return MTLPixelFormatInvalid; #endif default: Q_UNREACHABLE(); - return MTLPixelFormatRGBA8Unorm; + return MTLPixelFormatInvalid; } } @@ -2516,10 +3631,10 @@ void QMetalRenderBuffer::destroy() d->tex = nil; QRHI_RES_RHI(QRhiMetal); - rhiD->d->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseRenderBuffer(this)); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QMetalRenderBuffer::create() @@ -2542,16 +3657,23 @@ bool QMetalRenderBuffer::create() desc.resourceOptions = MTLResourceStorageModePrivate; desc.usage = MTLTextureUsageRenderTarget; - bool transientBacking = false; switch (m_type) { case DepthStencil: #ifdef Q_OS_MACOS - desc.storageMode = MTLStorageModePrivate; - d->format = rhiD->d->dev.depth24Stencil8PixelFormatSupported - ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; + if (rhiD->caps.isAppleGPU) { + if (@available(macOS 11.0, *)) { + desc.storageMode = MTLStorageModeMemoryless; + d->format = MTLPixelFormatDepth32Float_Stencil8; + } else { + Q_UNREACHABLE(); + } + } else { + desc.storageMode = MTLStorageModePrivate; + d->format = rhiD->d->dev.depth24Stencil8PixelFormatSupported + ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; + } #else desc.storageMode = MTLStorageModeMemoryless; - transientBacking = true; d->format = MTLPixelFormatDepth32Float_Stencil8; #endif desc.pixelFormat = d->format; @@ -2559,7 +3681,7 @@ bool QMetalRenderBuffer::create() case Color: desc.storageMode = MTLStorageModePrivate; if (m_backingFormatHint != QRhiTexture::UnknownFormat) - d->format = toMetalTextureFormat(m_backingFormatHint, {}, rhiD->d); + d->format = toMetalTextureFormat(m_backingFormatHint, {}, rhiD); else d->format = MTLPixelFormatRGBA8Unorm; desc.pixelFormat = d->format; @@ -2575,9 +3697,6 @@ bool QMetalRenderBuffer::create() if (!m_objectName.isEmpty()) d->tex.label = [NSString stringWithUTF8String: m_objectName.constData()]; - QRHI_PROF; - QRHI_PROF_F(newRenderBuffer(this, transientBacking, false, samples)); - lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -2593,8 +3712,8 @@ QRhiTexture::Format QMetalRenderBuffer::backingFormat() const } QMetalTexture::QMetalTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags) - : QRhiTexture(rhi, format, pixelSize, depth, sampleCount, flags), + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags), d(new QMetalTextureData(this)) { for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) @@ -2633,10 +3752,10 @@ void QMetalTexture::destroy() } QRHI_RES_RHI(QRhiMetal); - rhiD->d->releaseQueue.append(e); - QRHI_PROF; - QRHI_PROF_F(releaseTexture(this)); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QMetalTexture::prepareCreate(QSize *adjustedSize) @@ -2644,13 +3763,17 @@ bool QMetalTexture::prepareCreate(QSize *adjustedSize) if (d->tex) destroy(); - const QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); const bool hasMipMaps = m_flags.testFlag(MipMapped); + const bool is1D = m_flags.testFlag(OneDimensional); + + const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); QRHI_RES_RHI(QRhiMetal); - d->format = toMetalTextureFormat(m_format, m_flags, rhiD->d); + d->format = toMetalTextureFormat(m_format, m_flags, rhiD); mipLevelCount = hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1; samples = rhiD->effectiveSampleCount(m_sampleCount); if (samples > 1) { @@ -2671,11 +3794,30 @@ bool QMetalTexture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both cube and 3D"); return false; } - m_depth = qMax(1, m_depth); + if (isArray && is3D) { + qWarning("Texture cannot be both array and 3D"); + return false; + } + if (is1D && is3D) { + qWarning("Texture cannot be both 1D and 3D"); + return false; + } + if (is1D && isCube) { + qWarning("Texture cannot be both 1D and cube"); + return false; + } if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } + if (m_arraySize > 0 && !isArray) { + qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); + return false; + } + if (m_arraySize < 1 && isArray) { + qWarning("Texture is an array but array size is %d", m_arraySize); + return false; + } if (adjustedSize) *adjustedSize = size; @@ -2693,19 +3835,36 @@ bool QMetalTexture::create() const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); - if (isCube) + const bool isArray = m_flags.testFlag(TextureArray); + const bool is1D = m_flags.testFlag(OneDimensional); + if (isCube) { desc.textureType = MTLTextureTypeCube; - else if (is3D) + } else if (is3D) { desc.textureType = MTLTextureType3D; - else + } else if (is1D) { + desc.textureType = isArray ? MTLTextureType1DArray : MTLTextureType1D; + } else if (isArray) { +#ifdef Q_OS_IOS + if (@available(iOS 14, *)) { + desc.textureType = samples > 1 ? MTLTextureType2DMultisampleArray : MTLTextureType2DArray; + } else { + desc.textureType = MTLTextureType2DArray; + } +#else + desc.textureType = samples > 1 ? MTLTextureType2DMultisampleArray : MTLTextureType2DArray; +#endif + } else { desc.textureType = samples > 1 ? MTLTextureType2DMultisample : MTLTextureType2D; + } desc.pixelFormat = d->format; desc.width = NSUInteger(size.width()); desc.height = NSUInteger(size.height()); - desc.depth = is3D ? m_depth : 1; + desc.depth = is3D ? qMax(1, m_depth) : 1; desc.mipmapLevelCount = NSUInteger(mipLevelCount); if (samples > 1) desc.sampleCount = NSUInteger(samples); + if (isArray) + desc.arrayLength = NSUInteger(qMax(0, m_arraySize)); desc.resourceOptions = MTLResourceStorageModePrivate; desc.storageMode = MTLStorageModePrivate; desc.usage = MTLTextureUsageShaderRead; @@ -2723,9 +3882,6 @@ bool QMetalTexture::create() d->owns = true; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, true, mipLevelCount, isCube ? 6 : 1, samples)); - lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -2745,9 +3901,6 @@ bool QMetalTexture::createFrom(QRhiTexture::NativeTexture src) d->owns = false; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, false, mipLevelCount, m_flags.testFlag(CubeMap) ? 6 : 1, samples)); - lastActiveFrameSlot = -1; generation += 1; QRHI_RES_RHI(QRhiMetal); @@ -2768,8 +3921,10 @@ id<MTLTexture> QMetalTextureData::viewForLevel(int level) const MTLTextureType type = [tex textureType]; const bool isCube = q->m_flags.testFlag(QRhiTexture::CubeMap); + const bool isArray = q->m_flags.testFlag(QRhiTexture::TextureArray); id<MTLTexture> view = [tex newTextureViewWithPixelFormat: format textureType: type - levels: NSMakeRange(NSUInteger(level), 1) slices: NSMakeRange(0, isCube ? 6 : 1)]; + levels: NSMakeRange(NSUInteger(level), 1) + slices: NSMakeRange(0, isCube ? 6 : (isArray ? qMax(0, q->m_arraySize) : 1))]; perLevelViews[level] = view; return view; @@ -2801,8 +3956,10 @@ void QMetalSampler::destroy() d->samplerState = nil; QRHI_RES_RHI(QRhiMetal); - rhiD->d->releaseQueue.append(e); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } static inline MTLSamplerMinMagFilter toMetalFilter(QRhiSampler::Filter f) @@ -2902,6 +4059,7 @@ bool QMetalSampler::create() QMetalRenderPassDescriptor::QMetalRenderPassDescriptor(QRhiImplementation *rhi) : QRhiRenderPassDescriptor(rhi) { + serializedFormatData.reserve(16); } QMetalRenderPassDescriptor::~QMetalRenderPassDescriptor() @@ -2911,7 +4069,9 @@ QMetalRenderPassDescriptor::~QMetalRenderPassDescriptor() void QMetalRenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } bool QMetalRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -2940,44 +4100,66 @@ bool QMetalRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *ot return true; } +void QMetalRenderPassDescriptor::updateSerializedFormat() +{ + serializedFormatData.clear(); + auto p = std::back_inserter(serializedFormatData); + + *p++ = colorAttachmentCount; + *p++ = hasDepthStencil; + for (int i = 0; i < colorAttachmentCount; ++i) + *p++ = colorFormat[i]; + *p++ = hasDepthStencil ? dsFormat : 0; +} + QRhiRenderPassDescriptor *QMetalRenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - QMetalRenderPassDescriptor *rp = new QMetalRenderPassDescriptor(m_rhi); - rp->colorAttachmentCount = colorAttachmentCount; - rp->hasDepthStencil = hasDepthStencil; - memcpy(rp->colorFormat, colorFormat, sizeof(colorFormat)); - rp->dsFormat = dsFormat; - return rp; + QMetalRenderPassDescriptor *rpD = new QMetalRenderPassDescriptor(m_rhi); + rpD->colorAttachmentCount = colorAttachmentCount; + rpD->hasDepthStencil = hasDepthStencil; + memcpy(rpD->colorFormat, colorFormat, sizeof(colorFormat)); + rpD->dsFormat = dsFormat; + + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiMetal); + rhiD->registerResource(rpD, false); + return rpD; } -QMetalReferenceRenderTarget::QMetalReferenceRenderTarget(QRhiImplementation *rhi) - : QRhiRenderTarget(rhi), +QVector<quint32> QMetalRenderPassDescriptor::serializedFormat() const +{ + return serializedFormatData; +} + +QMetalSwapChainRenderTarget::QMetalSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain), d(new QMetalRenderTargetData) { } -QMetalReferenceRenderTarget::~QMetalReferenceRenderTarget() +QMetalSwapChainRenderTarget::~QMetalSwapChainRenderTarget() { destroy(); delete d; } -void QMetalReferenceRenderTarget::destroy() +void QMetalSwapChainRenderTarget::destroy() { // nothing to do here } -QSize QMetalReferenceRenderTarget::pixelSize() const +QSize QMetalSwapChainRenderTarget::pixelSize() const { return d->pixelSize; } -float QMetalReferenceRenderTarget::devicePixelRatio() const +float QMetalSwapChainRenderTarget::devicePixelRatio() const { return d->dpr; } -int QMetalReferenceRenderTarget::sampleCount() const +int QMetalSwapChainRenderTarget::sampleCount() const { return d->sampleCount; } @@ -2998,12 +4180,14 @@ QMetalTextureRenderTarget::~QMetalTextureRenderTarget() void QMetalTextureRenderTarget::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } QRhiRenderPassDescriptor *QMetalTextureRenderTarget::newCompatibleRenderPassDescriptor() { - const int colorAttachmentCount = m_desc.cendColorAttachments() - m_desc.cbeginColorAttachments(); + const int colorAttachmentCount = int(m_desc.colorAttachmentCount()); QMetalRenderPassDescriptor *rpD = new QMetalRenderPassDescriptor(m_rhi); rpD->colorAttachmentCount = colorAttachmentCount; rpD->hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); @@ -3020,14 +4204,17 @@ QRhiRenderPassDescriptor *QMetalTextureRenderTarget::newCompatibleRenderPassDesc else if (m_desc.depthStencilBuffer()) rpD->dsFormat = int(QRHI_RES(QMetalRenderBuffer, m_desc.depthStencilBuffer())->d->format); + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiMetal); + rhiD->registerResource(rpD, false); return rpD; } bool QMetalTextureRenderTarget::create() { QRHI_RES_RHI(QRhiMetal); - const bool hasColorAttachments = m_desc.cbeginColorAttachments() != m_desc.cendColorAttachments(); - Q_ASSERT(hasColorAttachments || m_desc.depthTexture()); + Q_ASSERT(m_desc.colorAttachmentCount() > 0 || m_desc.depthTexture()); Q_ASSERT(!m_desc.depthStencilBuffer() || !m_desc.depthTexture()); const bool hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); @@ -3071,8 +4258,9 @@ bool QMetalTextureRenderTarget::create() if (m_desc.depthTexture()) { QMetalTexture *depthTexD = QRHI_RES(QMetalTexture, m_desc.depthTexture()); d->fb.dsTex = depthTexD->d->tex; - d->fb.hasStencil = false; - d->fb.depthNeedsStore = true; + d->fb.hasStencil = rhiD->isStencilSupportingFormat(depthTexD->format()); + d->fb.depthNeedsStore = !m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture(); + d->fb.preserveDs = m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents); if (d->colorAttCount == 0) { d->pixelSize = depthTexD->pixelSize(); d->sampleCount = depthTexD->samples; @@ -3082,21 +4270,35 @@ bool QMetalTextureRenderTarget::create() d->fb.dsTex = depthRbD->d->tex; d->fb.hasStencil = true; d->fb.depthNeedsStore = false; + d->fb.preserveDs = false; if (d->colorAttCount == 0) { d->pixelSize = depthRbD->pixelSize(); d->sampleCount = depthRbD->samples; } } + if (m_desc.depthResolveTexture()) { + QMetalTexture *depthResolveTexD = QRHI_RES(QMetalTexture, m_desc.depthResolveTexture()); + d->fb.dsResolveTex = depthResolveTexD->d->tex; + } d->dsAttCount = 1; } else { d->dsAttCount = 0; } + if (d->colorAttCount > 0) + d->fb.preserveColor = m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents); + + QRhiRenderTargetAttachmentTracker::updateResIdList<QMetalTexture, QMetalRenderBuffer>(m_desc, &d->currentResIdList); + + rhiD->registerResource(this, false); return true; } QSize QMetalTextureRenderTarget::pixelSize() const { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QMetalTexture, QMetalRenderBuffer>(m_desc, d->currentResIdList)) + const_cast<QMetalTextureRenderTarget *>(this)->create(); + return d->pixelSize; } @@ -3124,6 +4326,10 @@ void QMetalShaderResourceBindings::destroy() { sortedBindings.clear(); maxBinding = -1; + + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } bool QMetalShaderResourceBindings::create() @@ -3138,13 +4344,9 @@ bool QMetalShaderResourceBindings::create() rhiD->updateLayoutDesc(this); std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); - std::sort(sortedBindings.begin(), sortedBindings.end(), - [](const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) - { - return a.data()->binding < b.data()->binding; - }); + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); if (!sortedBindings.isEmpty()) - maxBinding = sortedBindings.last().data()->binding; + maxBinding = QRhiImplementation::shaderResourceBindingData(sortedBindings.last())->binding; else maxBinding = -1; @@ -3154,13 +4356,29 @@ bool QMetalShaderResourceBindings::create() memset(&bd, 0, sizeof(BoundResourceData)); generation += 1; + rhiD->registerResource(this, false); return true; } +void QMetalShaderResourceBindings::updateResources(UpdateFlags flags) +{ + sortedBindings.clear(); + std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); + if (!flags.testFlag(BindingsAreSorted)) + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); + + for (BoundResourceData &bd : boundResourceData) + memset(&bd, 0, sizeof(BoundResourceData)); + + generation += 1; +} + QMetalGraphicsPipeline::QMetalGraphicsPipeline(QRhiImplementation *rhi) : QRhiGraphicsPipeline(rhi), d(new QMetalGraphicsPipelineData) { + d->q = this; + d->tess.q = d; } QMetalGraphicsPipeline::~QMetalGraphicsPipeline() @@ -3171,21 +4389,48 @@ QMetalGraphicsPipeline::~QMetalGraphicsPipeline() void QMetalGraphicsPipeline::destroy() { - QRHI_RES_RHI(QRhiMetal); - d->vs.destroy(); d->fs.destroy(); - [d->ds release]; - d->ds = nil; + d->tess.compVs[0].destroy(); + d->tess.compVs[1].destroy(); + d->tess.compVs[2].destroy(); - if (!d->ps) + d->tess.compTesc.destroy(); + d->tess.vertTese.destroy(); + + qDeleteAll(d->extraBufMgr.deviceLocalWorkBuffers); + d->extraBufMgr.deviceLocalWorkBuffers.clear(); + qDeleteAll(d->extraBufMgr.hostVisibleWorkBuffers); + d->extraBufMgr.hostVisibleWorkBuffers.clear(); + + delete d->bufferSizeBuffer; + d->bufferSizeBuffer = nullptr; + + if (!d->ps && !d->ds + && !d->tess.vertexComputeState[0] && !d->tess.vertexComputeState[1] && !d->tess.vertexComputeState[2] + && !d->tess.tessControlComputeState) + { return; + } - [d->ps release]; + QRhiMetalData::DeferredReleaseEntry e; + e.type = QRhiMetalData::DeferredReleaseEntry::GraphicsPipeline; + e.lastActiveFrameSlot = lastActiveFrameSlot; + e.graphicsPipeline.pipelineState = d->ps; + e.graphicsPipeline.depthStencilState = d->ds; + e.graphicsPipeline.tessVertexComputeState = d->tess.vertexComputeState; + e.graphicsPipeline.tessTessControlComputeState = d->tess.tessControlComputeState; d->ps = nil; + d->ds = nil; + d->tess.vertexComputeState = {}; + d->tess.tessControlComputeState = nil; - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiMetal); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } static inline MTLVertexFormat toMetalAttributeFormat(QRhiVertexInputAttribute::Format format) @@ -3221,6 +4466,30 @@ static inline MTLVertexFormat toMetalAttributeFormat(QRhiVertexInputAttribute::F return MTLVertexFormatInt2; case QRhiVertexInputAttribute::SInt: return MTLVertexFormatInt; + case QRhiVertexInputAttribute::Half4: + return MTLVertexFormatHalf4; + case QRhiVertexInputAttribute::Half3: + return MTLVertexFormatHalf3; + case QRhiVertexInputAttribute::Half2: + return MTLVertexFormatHalf2; + case QRhiVertexInputAttribute::Half: + return MTLVertexFormatHalf; + case QRhiVertexInputAttribute::UShort4: + return MTLVertexFormatUShort4; + case QRhiVertexInputAttribute::UShort3: + return MTLVertexFormatUShort3; + case QRhiVertexInputAttribute::UShort2: + return MTLVertexFormatUShort2; + case QRhiVertexInputAttribute::UShort: + return MTLVertexFormatUShort; + case QRhiVertexInputAttribute::SShort4: + return MTLVertexFormatShort4; + case QRhiVertexInputAttribute::SShort3: + return MTLVertexFormatShort3; + case QRhiVertexInputAttribute::SShort2: + return MTLVertexFormatShort2; + case QRhiVertexInputAttribute::SShort: + return MTLVertexFormatShort; default: Q_UNREACHABLE(); return MTLVertexFormatFloat4; @@ -3376,6 +4645,24 @@ static inline MTLPrimitiveType toMetalPrimitiveType(QRhiGraphicsPipeline::Topolo } } +static inline MTLPrimitiveTopologyClass toMetalPrimitiveTopologyClass(QRhiGraphicsPipeline::Topology t) +{ + switch (t) { + case QRhiGraphicsPipeline::Triangles: + case QRhiGraphicsPipeline::TriangleStrip: + case QRhiGraphicsPipeline::TriangleFan: + return MTLPrimitiveTopologyClassTriangle; + case QRhiGraphicsPipeline::Lines: + case QRhiGraphicsPipeline::LineStrip: + return MTLPrimitiveTopologyClassLine; + case QRhiGraphicsPipeline::Points: + return MTLPrimitiveTopologyClassPoint; + default: + Q_UNREACHABLE(); + return MTLPrimitiveTopologyClassTriangle; + } +} + static inline MTLCullMode toMetalCullMode(QRhiGraphicsPipeline::CullMode c) { switch (c) { @@ -3391,15 +4678,80 @@ static inline MTLCullMode toMetalCullMode(QRhiGraphicsPipeline::CullMode c) } } +static inline MTLTriangleFillMode toMetalTriangleFillMode(QRhiGraphicsPipeline::PolygonMode mode) +{ + switch (mode) { + case QRhiGraphicsPipeline::Fill: + return MTLTriangleFillModeFill; + case QRhiGraphicsPipeline::Line: + return MTLTriangleFillModeLines; + default: + Q_UNREACHABLE(); + return MTLTriangleFillModeFill; + } +} + +static inline MTLWinding toMetalTessellationWindingOrder(QShaderDescription::TessellationWindingOrder w) +{ + switch (w) { + case QShaderDescription::CwTessellationWindingOrder: + return MTLWindingClockwise; + case QShaderDescription::CcwTessellationWindingOrder: + return MTLWindingCounterClockwise; + default: + // this is reachable, consider a tess.eval. shader not declaring it, the value is then Unknown + return MTLWindingCounterClockwise; + } +} + +static inline MTLTessellationPartitionMode toMetalTessellationPartitionMode(QShaderDescription::TessellationPartitioning p) +{ + switch (p) { + case QShaderDescription::EqualTessellationPartitioning: + return MTLTessellationPartitionModePow2; + case QShaderDescription::FractionalEvenTessellationPartitioning: + return MTLTessellationPartitionModeFractionalEven; + case QShaderDescription::FractionalOddTessellationPartitioning: + return MTLTessellationPartitionModeFractionalOdd; + default: + // this is reachable, consider a tess.eval. shader not declaring it, the value is then Unknown + return MTLTessellationPartitionModePow2; + } +} + +static inline MTLLanguageVersion toMetalLanguageVersion(const QShaderVersion &version) +{ + int v = version.version(); + return MTLLanguageVersion(((v / 10) << 16) + (v % 10)); +} + id<MTLLibrary> QRhiMetalData::createMetalLib(const QShader &shader, QShader::Variant shaderVariant, QString *error, QByteArray *entryPoint, QShaderKey *activeKey) { - QShaderKey key = { QShader::MetalLibShader, 20, shaderVariant }; - QShaderCode mtllib = shader.shader(key); - if (mtllib.shader().isEmpty()) { - key.setSourceVersion(12); - mtllib = shader.shader(key); + QVarLengthArray<int, 8> versions; + if (@available(macOS 13, iOS 16, *)) + versions << 30; + if (@available(macOS 12, iOS 15, *)) + versions << 24; + if (@available(macOS 11, iOS 14, *)) + versions << 23; + if (@available(macOS 10.15, iOS 13, *)) + versions << 22; + if (@available(macOS 10.14, iOS 12, *)) + versions << 21; + versions << 20 << 12; + + const QList<QShaderKey> shaders = shader.availableShaders(); + + QShaderKey key; + + for (const int &version : versions) { + key = { QShader::Source::MetalLibShader, version, shaderVariant }; + if (shaders.contains(key)) + break; } + + QShaderCode mtllib = shader.shader(key); if (!mtllib.shader().isEmpty()) { dispatch_data_t data = dispatch_data_create(mtllib.shader().constData(), size_t(mtllib.shader().size()), @@ -3418,12 +4770,13 @@ id<MTLLibrary> QRhiMetalData::createMetalLib(const QShader &shader, QShader::Var } } - key = { QShader::MslShader, 20, shaderVariant }; - QShaderCode mslSource = shader.shader(key); - if (mslSource.shader().isEmpty()) { - key.setSourceVersion(12); - mslSource = shader.shader(key); + for (const int &version : versions) { + key = { QShader::Source::MslShader, version, shaderVariant }; + if (shaders.contains(key)) + break; } + + QShaderCode mslSource = shader.shader(key); if (mslSource.shader().isEmpty()) { qWarning() << "No MSL 2.0 or 1.2 code found in baked shader" << shader; return nil; @@ -3431,7 +4784,7 @@ id<MTLLibrary> QRhiMetalData::createMetalLib(const QShader &shader, QShader::Var NSString *src = [NSString stringWithUTF8String: mslSource.shader().constData()]; MTLCompileOptions *opts = [[MTLCompileOptions alloc] init]; - opts.languageVersion = key.sourceVersion() == 20 ? MTLLanguageVersion2_0 : MTLLanguageVersion1_2; + opts.languageVersion = toMetalLanguageVersion(key.sourceVersion()); NSError *err = nil; id<MTLLibrary> lib = [dev newLibraryWithSource: src options: opts error: &err]; [opts release]; @@ -3453,55 +4806,192 @@ id<MTLLibrary> QRhiMetalData::createMetalLib(const QShader &shader, QShader::Var id<MTLFunction> QRhiMetalData::createMSLShaderFunction(id<MTLLibrary> lib, const QByteArray &entryPoint) { - NSString *name = [NSString stringWithUTF8String: entryPoint.constData()]; - id<MTLFunction> f = [lib newFunctionWithName: name]; - [name release]; - return f; + return [lib newFunctionWithName:[NSString stringWithUTF8String:entryPoint.constData()]]; } -bool QMetalGraphicsPipeline::create() +void QMetalGraphicsPipeline::setupAttachmentsInMetalRenderPassDescriptor(void *metalRpDesc, QMetalRenderPassDescriptor *rpD) { - if (d->ps) - destroy(); + MTLRenderPipelineDescriptor *rpDesc = reinterpret_cast<MTLRenderPipelineDescriptor *>(metalRpDesc); + + if (rpD->colorAttachmentCount) { + // defaults when no targetBlends are provided + rpDesc.colorAttachments[0].pixelFormat = MTLPixelFormat(rpD->colorFormat[0]); + rpDesc.colorAttachments[0].writeMask = MTLColorWriteMaskAll; + rpDesc.colorAttachments[0].blendingEnabled = false; + + Q_ASSERT(m_targetBlends.count() == rpD->colorAttachmentCount + || (m_targetBlends.isEmpty() && rpD->colorAttachmentCount == 1)); + + for (uint i = 0, ie = uint(m_targetBlends.count()); i != ie; ++i) { + const QRhiGraphicsPipeline::TargetBlend &b(m_targetBlends[int(i)]); + rpDesc.colorAttachments[i].pixelFormat = MTLPixelFormat(rpD->colorFormat[i]); + rpDesc.colorAttachments[i].blendingEnabled = b.enable; + rpDesc.colorAttachments[i].sourceRGBBlendFactor = toMetalBlendFactor(b.srcColor); + rpDesc.colorAttachments[i].destinationRGBBlendFactor = toMetalBlendFactor(b.dstColor); + rpDesc.colorAttachments[i].rgbBlendOperation = toMetalBlendOp(b.opColor); + rpDesc.colorAttachments[i].sourceAlphaBlendFactor = toMetalBlendFactor(b.srcAlpha); + rpDesc.colorAttachments[i].destinationAlphaBlendFactor = toMetalBlendFactor(b.dstAlpha); + rpDesc.colorAttachments[i].alphaBlendOperation = toMetalBlendOp(b.opAlpha); + rpDesc.colorAttachments[i].writeMask = toMetalColorWriteMask(b.colorWrite); + } + } + + if (rpD->hasDepthStencil) { + // Must only be set when a depth-stencil buffer will actually be bound, + // validation blows up otherwise. + MTLPixelFormat fmt = MTLPixelFormat(rpD->dsFormat); + rpDesc.depthAttachmentPixelFormat = fmt; +#if defined(Q_OS_MACOS) + if (fmt != MTLPixelFormatDepth16Unorm && fmt != MTLPixelFormatDepth32Float) +#else + if (fmt != MTLPixelFormatDepth32Float) +#endif + rpDesc.stencilAttachmentPixelFormat = fmt; + } QRHI_RES_RHI(QRhiMetal); - if (!rhiD->sanityCheckGraphicsPipeline(this)) - return false; + rpDesc.sampleCount = NSUInteger(rhiD->effectiveSampleCount(m_sampleCount)); +} +void QMetalGraphicsPipeline::setupMetalDepthStencilDescriptor(void *metalDsDesc) +{ + MTLDepthStencilDescriptor *dsDesc = reinterpret_cast<MTLDepthStencilDescriptor *>(metalDsDesc); + + dsDesc.depthCompareFunction = m_depthTest ? toMetalCompareOp(m_depthOp) : MTLCompareFunctionAlways; + dsDesc.depthWriteEnabled = m_depthWrite; + if (m_stencilTest) { + dsDesc.frontFaceStencil = [[MTLStencilDescriptor alloc] init]; + dsDesc.frontFaceStencil.stencilFailureOperation = toMetalStencilOp(m_stencilFront.failOp); + dsDesc.frontFaceStencil.depthFailureOperation = toMetalStencilOp(m_stencilFront.depthFailOp); + dsDesc.frontFaceStencil.depthStencilPassOperation = toMetalStencilOp(m_stencilFront.passOp); + dsDesc.frontFaceStencil.stencilCompareFunction = toMetalCompareOp(m_stencilFront.compareOp); + dsDesc.frontFaceStencil.readMask = m_stencilReadMask; + dsDesc.frontFaceStencil.writeMask = m_stencilWriteMask; + + dsDesc.backFaceStencil = [[MTLStencilDescriptor alloc] init]; + dsDesc.backFaceStencil.stencilFailureOperation = toMetalStencilOp(m_stencilBack.failOp); + dsDesc.backFaceStencil.depthFailureOperation = toMetalStencilOp(m_stencilBack.depthFailOp); + dsDesc.backFaceStencil.depthStencilPassOperation = toMetalStencilOp(m_stencilBack.passOp); + dsDesc.backFaceStencil.stencilCompareFunction = toMetalCompareOp(m_stencilBack.compareOp); + dsDesc.backFaceStencil.readMask = m_stencilReadMask; + dsDesc.backFaceStencil.writeMask = m_stencilWriteMask; + } +} + +void QMetalGraphicsPipeline::mapStates() +{ + d->winding = m_frontFace == CCW ? MTLWindingCounterClockwise : MTLWindingClockwise; + d->cullMode = toMetalCullMode(m_cullMode); + d->triangleFillMode = toMetalTriangleFillMode(m_polygonMode); + d->depthBias = float(m_depthBias); + d->slopeScaledDepthBias = m_slopeScaledDepthBias; +} + +void QMetalGraphicsPipelineData::setupVertexInputDescriptor(MTLVertexDescriptor *desc) +{ // same binding space for vertex and constant buffers - work it around - const int firstVertexBinding = QRHI_RES(QMetalShaderResourceBindings, m_shaderResourceBindings)->maxBinding + 1; + // should be in native resource binding not SPIR-V, but this will work anyway + const int firstVertexBinding = QRHI_RES(QMetalShaderResourceBindings, q->shaderResourceBindings())->maxBinding + 1; - MTLVertexDescriptor *inputLayout = [MTLVertexDescriptor vertexDescriptor]; - for (auto it = m_vertexInputLayout.cbeginAttributes(), itEnd = m_vertexInputLayout.cendAttributes(); + QRhiVertexInputLayout vertexInputLayout = q->vertexInputLayout(); + for (auto it = vertexInputLayout.cbeginAttributes(), itEnd = vertexInputLayout.cendAttributes(); it != itEnd; ++it) { const uint loc = uint(it->location()); - inputLayout.attributes[loc].format = toMetalAttributeFormat(it->format()); - inputLayout.attributes[loc].offset = NSUInteger(it->offset()); - inputLayout.attributes[loc].bufferIndex = NSUInteger(firstVertexBinding + it->binding()); + desc.attributes[loc].format = decltype(desc.attributes[loc].format)(toMetalAttributeFormat(it->format())); + desc.attributes[loc].offset = NSUInteger(it->offset()); + desc.attributes[loc].bufferIndex = NSUInteger(firstVertexBinding + it->binding()); } int bindingIndex = 0; - for (auto it = m_vertexInputLayout.cbeginBindings(), itEnd = m_vertexInputLayout.cendBindings(); + const NSUInteger viewCount = qMax<NSUInteger>(1, q->multiViewCount()); + for (auto it = vertexInputLayout.cbeginBindings(), itEnd = vertexInputLayout.cendBindings(); it != itEnd; ++it, ++bindingIndex) { const uint layoutIdx = uint(firstVertexBinding + bindingIndex); - inputLayout.layouts[layoutIdx].stepFunction = - it->classification() == QRhiVertexInputBinding::PerInstance - ? MTLVertexStepFunctionPerInstance : MTLVertexStepFunctionPerVertex; - inputLayout.layouts[layoutIdx].stepRate = NSUInteger(it->instanceStepRate()); - inputLayout.layouts[layoutIdx].stride = it->stride(); + desc.layouts[layoutIdx].stepFunction = + it->classification() == QRhiVertexInputBinding::PerInstance + ? MTLVertexStepFunctionPerInstance : MTLVertexStepFunctionPerVertex; + desc.layouts[layoutIdx].stepRate = NSUInteger(it->instanceStepRate()); + if (desc.layouts[layoutIdx].stepFunction == MTLVertexStepFunctionPerInstance) + desc.layouts[layoutIdx].stepRate *= viewCount; + desc.layouts[layoutIdx].stride = it->stride(); } +} - MTLRenderPipelineDescriptor *rpDesc = [[MTLRenderPipelineDescriptor alloc] init]; +void QMetalGraphicsPipelineData::setupStageInputDescriptor(MTLStageInputOutputDescriptor *desc) +{ + // same binding space for vertex and constant buffers - work it around + // should be in native resource binding not SPIR-V, but this will work anyway + const int firstVertexBinding = QRHI_RES(QMetalShaderResourceBindings, q->shaderResourceBindings())->maxBinding + 1; + + QRhiVertexInputLayout vertexInputLayout = q->vertexInputLayout(); + for (auto it = vertexInputLayout.cbeginAttributes(), itEnd = vertexInputLayout.cendAttributes(); + it != itEnd; ++it) + { + const uint loc = uint(it->location()); + desc.attributes[loc].format = decltype(desc.attributes[loc].format)(toMetalAttributeFormat(it->format())); + desc.attributes[loc].offset = NSUInteger(it->offset()); + desc.attributes[loc].bufferIndex = NSUInteger(firstVertexBinding + it->binding()); + } + int bindingIndex = 0; + for (auto it = vertexInputLayout.cbeginBindings(), itEnd = vertexInputLayout.cendBindings(); + it != itEnd; ++it, ++bindingIndex) + { + const uint layoutIdx = uint(firstVertexBinding + bindingIndex); + if (desc.indexBufferIndex) { + desc.layouts[layoutIdx].stepFunction = + it->classification() == QRhiVertexInputBinding::PerInstance + ? MTLStepFunctionThreadPositionInGridY : MTLStepFunctionThreadPositionInGridXIndexed; + } else { + desc.layouts[layoutIdx].stepFunction = + it->classification() == QRhiVertexInputBinding::PerInstance + ? MTLStepFunctionThreadPositionInGridY : MTLStepFunctionThreadPositionInGridX; + } + desc.layouts[layoutIdx].stepRate = NSUInteger(it->instanceStepRate()); + desc.layouts[layoutIdx].stride = it->stride(); + } +} - rpDesc.vertexDescriptor = inputLayout; +void QRhiMetalData::trySeedingRenderPipelineFromBinaryArchive(MTLRenderPipelineDescriptor *rpDesc) +{ + if (@available(macOS 11.0, iOS 14.0, *)) { + if (binArch) { + NSArray *binArchArray = [NSArray arrayWithObjects: binArch, nil]; + rpDesc.binaryArchives = binArchArray; + } + } +} - // mutability cannot be determined (slotted buffers could be set as +void QRhiMetalData::addRenderPipelineToBinaryArchive(MTLRenderPipelineDescriptor *rpDesc) +{ + if (@available(macOS 11.0, iOS 14.0, *)) { + if (binArch) { + NSError *err = nil; + if (![binArch addRenderPipelineFunctionsWithDescriptor: rpDesc error: &err]) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("Failed to collect render pipeline functions to binary archive: %s", qPrintable(msg)); + } + } + } +} + +bool QMetalGraphicsPipeline::createVertexFragmentPipeline() +{ + QRHI_RES_RHI(QRhiMetal); + + MTLVertexDescriptor *vertexDesc = [MTLVertexDescriptor vertexDescriptor]; + d->setupVertexInputDescriptor(vertexDesc); + + MTLRenderPipelineDescriptor *rpDesc = [[MTLRenderPipelineDescriptor alloc] init]; + rpDesc.vertexDescriptor = vertexDesc; + + // Mutability cannot be determined (slotted buffers could be set as // MTLMutabilityImmutable, but then we potentially need a different // descriptor for each buffer combination as this depends on the actual - // buffers not just the resource binding layout) so leave it at the default + // buffers not just the resource binding layout), so leave + // rpDesc.vertex/fragmentBuffers at the defaults. - for (const QRhiShaderStage &shaderStage : qAsConst(m_shaderStages)) { + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { auto cacheIt = rhiD->d->shaderCache.constFind(shaderStage); if (cacheIt != rhiD->d->shaderCache.constEnd()) { switch (shaderStage.type()) { @@ -3547,8 +5037,9 @@ bool QMetalGraphicsPipeline::create() case QRhiShaderStage::Vertex: d->vs.lib = lib; d->vs.func = func; - if (const QShader::NativeResourceBindingMap *map = shader.nativeResourceBindingMap(activeKey)) - d->vs.nativeResourceBindingMap = *map; + d->vs.nativeResourceBindingMap = shader.nativeResourceBindingMap(activeKey); + d->vs.desc = shader.description(); + d->vs.nativeShaderInfo = shader.nativeShaderInfo(activeKey); rhiD->d->shaderCache.insert(shaderStage, d->vs); [d->vs.lib retain]; [d->vs.func retain]; @@ -3557,8 +5048,9 @@ bool QMetalGraphicsPipeline::create() case QRhiShaderStage::Fragment: d->fs.lib = lib; d->fs.func = func; - if (const QShader::NativeResourceBindingMap *map = shader.nativeResourceBindingMap(activeKey)) - d->fs.nativeResourceBindingMap = *map; + d->fs.nativeResourceBindingMap = shader.nativeResourceBindingMap(activeKey); + d->fs.desc = shader.description(); + d->fs.nativeShaderInfo = shader.nativeShaderInfo(activeKey); rhiD->d->shaderCache.insert(shaderStage, d->fs); [d->fs.lib retain]; [d->fs.func retain]; @@ -3573,85 +5065,838 @@ bool QMetalGraphicsPipeline::create() } QMetalRenderPassDescriptor *rpD = QRHI_RES(QMetalRenderPassDescriptor, m_renderPassDesc); + setupAttachmentsInMetalRenderPassDescriptor(rpDesc, rpD); - if (rpD->colorAttachmentCount) { - // defaults when no targetBlends are provided - rpDesc.colorAttachments[0].pixelFormat = MTLPixelFormat(rpD->colorFormat[0]); - rpDesc.colorAttachments[0].writeMask = MTLColorWriteMaskAll; - rpDesc.colorAttachments[0].blendingEnabled = false; + if (m_multiViewCount >= 2) + rpDesc.inputPrimitiveTopology = toMetalPrimitiveTopologyClass(m_topology); - Q_ASSERT(m_targetBlends.count() == rpD->colorAttachmentCount - || (m_targetBlends.isEmpty() && rpD->colorAttachmentCount == 1)); + rhiD->d->trySeedingRenderPipelineFromBinaryArchive(rpDesc); - for (uint i = 0, ie = uint(m_targetBlends.count()); i != ie; ++i) { - const QRhiGraphicsPipeline::TargetBlend &b(m_targetBlends[int(i)]); - rpDesc.colorAttachments[i].pixelFormat = MTLPixelFormat(rpD->colorFormat[i]); - rpDesc.colorAttachments[i].blendingEnabled = b.enable; - rpDesc.colorAttachments[i].sourceRGBBlendFactor = toMetalBlendFactor(b.srcColor); - rpDesc.colorAttachments[i].destinationRGBBlendFactor = toMetalBlendFactor(b.dstColor); - rpDesc.colorAttachments[i].rgbBlendOperation = toMetalBlendOp(b.opColor); - rpDesc.colorAttachments[i].sourceAlphaBlendFactor = toMetalBlendFactor(b.srcAlpha); - rpDesc.colorAttachments[i].destinationAlphaBlendFactor = toMetalBlendFactor(b.dstAlpha); - rpDesc.colorAttachments[i].alphaBlendOperation = toMetalBlendOp(b.opAlpha); - rpDesc.colorAttachments[i].writeMask = toMetalColorWriteMask(b.colorWrite); + if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + rhiD->d->addRenderPipelineToBinaryArchive(rpDesc); + + NSError *err = nil; + d->ps = [rhiD->d->dev newRenderPipelineStateWithDescriptor: rpDesc error: &err]; + [rpDesc release]; + if (!d->ps) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("Failed to create render pipeline state: %s", qPrintable(msg)); + return false; + } + + MTLDepthStencilDescriptor *dsDesc = [[MTLDepthStencilDescriptor alloc] init]; + setupMetalDepthStencilDescriptor(dsDesc); + d->ds = [rhiD->d->dev newDepthStencilStateWithDescriptor: dsDesc]; + [dsDesc release]; + + d->primitiveType = toMetalPrimitiveType(m_topology); + mapStates(); + + return true; +} + +int QMetalGraphicsPipelineData::Tessellation::vsCompVariantToIndex(QShader::Variant vertexCompVariant) +{ + switch (vertexCompVariant) { + case QShader::NonIndexedVertexAsComputeShader: + return 0; + case QShader::UInt32IndexedVertexAsComputeShader: + return 1; + case QShader::UInt16IndexedVertexAsComputeShader: + return 2; + default: + break; + } + return -1; +} + +id<MTLComputePipelineState> QMetalGraphicsPipelineData::Tessellation::vsCompPipeline(QRhiMetal *rhiD, QShader::Variant vertexCompVariant) +{ + const int varIndex = vsCompVariantToIndex(vertexCompVariant); + if (varIndex >= 0 && vertexComputeState[varIndex]) + return vertexComputeState[varIndex]; + + id<MTLFunction> func = nil; + if (varIndex >= 0) + func = compVs[varIndex].func; + + if (!func) { + qWarning("No compute function found for vertex shader translated for tessellation, this should not happen"); + return nil; + } + + const QMap<int, int> &ebb(compVs[varIndex].nativeShaderInfo.extraBufferBindings); + const int indexBufferBinding = ebb.value(QShaderPrivate::MslTessVertIndicesBufferBinding, -1); + + MTLComputePipelineDescriptor *cpDesc = [MTLComputePipelineDescriptor new]; + cpDesc.computeFunction = func; + cpDesc.threadGroupSizeIsMultipleOfThreadExecutionWidth = YES; + cpDesc.stageInputDescriptor = [MTLStageInputOutputDescriptor stageInputOutputDescriptor]; + if (indexBufferBinding >= 0) { + if (vertexCompVariant == QShader::UInt32IndexedVertexAsComputeShader) { + cpDesc.stageInputDescriptor.indexType = MTLIndexTypeUInt32; + cpDesc.stageInputDescriptor.indexBufferIndex = indexBufferBinding; + } else if (vertexCompVariant == QShader::UInt16IndexedVertexAsComputeShader) { + cpDesc.stageInputDescriptor.indexType = MTLIndexTypeUInt16; + cpDesc.stageInputDescriptor.indexBufferIndex = indexBufferBinding; } } + q->setupStageInputDescriptor(cpDesc.stageInputDescriptor); - if (rpD->hasDepthStencil) { - // Must only be set when a depth-stencil buffer will actually be bound, - // validation blows up otherwise. - MTLPixelFormat fmt = MTLPixelFormat(rpD->dsFormat); - rpDesc.depthAttachmentPixelFormat = fmt; -#ifdef Q_OS_MACOS - if (fmt != MTLPixelFormatDepth16Unorm && fmt != MTLPixelFormatDepth32Float) -#else - if (fmt != MTLPixelFormatDepth32Float) -#endif - rpDesc.stencilAttachmentPixelFormat = fmt; + rhiD->d->trySeedingComputePipelineFromBinaryArchive(cpDesc); + + if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + rhiD->d->addComputePipelineToBinaryArchive(cpDesc); + + NSError *err = nil; + id<MTLComputePipelineState> ps = [rhiD->d->dev newComputePipelineStateWithDescriptor: cpDesc + options: MTLPipelineOptionNone + reflection: nil + error: &err]; + [cpDesc release]; + if (!ps) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("Failed to create compute pipeline state: %s", qPrintable(msg)); + } else { + vertexComputeState[varIndex] = ps; } + // not retained, the only owner is vertexComputeState and so the QRhiGraphicsPipeline + return ps; +} - rpDesc.sampleCount = NSUInteger(rhiD->effectiveSampleCount(m_sampleCount)); +id<MTLComputePipelineState> QMetalGraphicsPipelineData::Tessellation::tescCompPipeline(QRhiMetal *rhiD) +{ + if (tessControlComputeState) + return tessControlComputeState; + + MTLComputePipelineDescriptor *cpDesc = [MTLComputePipelineDescriptor new]; + cpDesc.computeFunction = compTesc.func; + + rhiD->d->trySeedingComputePipelineFromBinaryArchive(cpDesc); + + if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + rhiD->d->addComputePipelineToBinaryArchive(cpDesc); NSError *err = nil; - d->ps = [rhiD->d->dev newRenderPipelineStateWithDescriptor: rpDesc error: &err]; - if (!d->ps) { + id<MTLComputePipelineState> ps = [rhiD->d->dev newComputePipelineStateWithDescriptor: cpDesc + options: MTLPipelineOptionNone + reflection: nil + error: &err]; + [cpDesc release]; + if (!ps) { const QString msg = QString::fromNSString(err.localizedDescription); - qWarning("Failed to create render pipeline state: %s", qPrintable(msg)); - [rpDesc release]; - return false; + qWarning("Failed to create compute pipeline state: %s", qPrintable(msg)); + } else { + tessControlComputeState = ps; } + // not retained, the only owner is tessControlComputeState and so the QRhiGraphicsPipeline + return ps; +} + +static inline bool indexTaken(quint32 index, quint64 indices) +{ + return (indices >> index) & 0x1; +} + +static inline void takeIndex(quint32 index, quint64 &indices) +{ + indices |= 1 << index; +} + +static inline int nextAttributeIndex(quint64 indices) +{ + // Maximum number of vertex attributes per vertex descriptor. There does + // not appear to be a way to query this from the implementation. + // https://developer.apple.com/metal/Metal-Feature-Set-Tables.pdf indicates + // that all GPU families have a value of 31. + static const int maxVertexAttributes = 31; + + for (int index = 0; index < maxVertexAttributes; ++index) { + if (!indexTaken(index, indices)) + return index; + } + + Q_UNREACHABLE_RETURN(-1); +} + +static inline int aligned(quint32 offset, quint32 alignment) +{ + return ((offset + alignment - 1) / alignment) * alignment; +} + +template<typename T> +static void addUnusedVertexAttribute(const T &variable, QRhiMetal *rhiD, quint32 &offset, quint32 &vertexAlignment) +{ + + int elements = 1; + for (const int dim : variable.arrayDims) + elements *= dim; + + if (variable.type == QShaderDescription::VariableType::Struct) { + for (int element = 0; element < elements; ++element) { + for (const auto &member : variable.structMembers) { + addUnusedVertexAttribute(member, rhiD, offset, vertexAlignment); + } + } + } else { + const QRhiVertexInputAttribute::Format format = rhiD->shaderDescVariableFormatToVertexInputFormat(variable.type); + const quint32 size = rhiD->byteSizePerVertexForVertexInputFormat(format); + + // MSL specification 3.0 says alignment = size for non packed scalars and vectors + const quint32 alignment = size; + vertexAlignment = std::max(vertexAlignment, alignment); + + for (int element = 0; element < elements; ++element) { + // adjust alignment + offset = aligned(offset, alignment); + offset += size; + } + } +} + +template<typename T> +static void addVertexAttribute(const T &variable, int binding, QRhiMetal *rhiD, int &index, quint32 &offset, MTLVertexAttributeDescriptorArray *attributes, quint64 &indices, quint32 &vertexAlignment) +{ + + int elements = 1; + for (const int dim : variable.arrayDims) + elements *= dim; + + if (variable.type == QShaderDescription::VariableType::Struct) { + for (int element = 0; element < elements; ++element) { + for (const auto &member : variable.structMembers) { + addVertexAttribute(member, binding, rhiD, index, offset, attributes, indices, vertexAlignment); + } + } + } else { + const QRhiVertexInputAttribute::Format format = rhiD->shaderDescVariableFormatToVertexInputFormat(variable.type); + const quint32 size = rhiD->byteSizePerVertexForVertexInputFormat(format); + + // MSL specification 3.0 says alignment = size for non packed scalars and vectors + const quint32 alignment = size; + vertexAlignment = std::max(vertexAlignment, alignment); + + for (int element = 0; element < elements; ++element) { + Q_ASSERT(!indexTaken(index, indices)); + + // adjust alignment + offset = aligned(offset, alignment); + + attributes[index].bufferIndex = binding; + attributes[index].format = toMetalAttributeFormat(format); + attributes[index].offset = offset; + + takeIndex(index, indices); + index++; + if (indexTaken(index, indices)) + index = nextAttributeIndex(indices); + + offset += size; + } + } +} + +static inline bool matches(const QList<QShaderDescription::BlockVariable> &a, const QList<QShaderDescription::BlockVariable> &b) +{ + if (a.size() == b.size()) { + bool match = true; + for (int i = 0; i < a.size() && match; ++i) { + match &= a[i].type == b[i].type + && a[i].arrayDims == b[i].arrayDims + && matches(a[i].structMembers, b[i].structMembers); + } + return match; + } + + return false; +} + +static inline bool matches(const QShaderDescription::InOutVariable &a, const QShaderDescription::InOutVariable &b) +{ + return a.location == b.location + && a.type == b.type + && a.perPatch == b.perPatch + && matches(a.structMembers, b.structMembers); +} + +// +// Create the tessellation evaluation render pipeline state +// +// The tesc runs as a compute shader in a compute pipeline and writes per patch and per patch +// control point data into separate storage buffers. The tese runs as a vertex shader in a render +// pipeline. Our task is to generate a render pipeline descriptor for the tese that pulls vertices +// from these buffers. +// +// As the buffers we are pulling vertices from are written by a compute pipeline, they follow the +// MSL alignment conventions which we must take into account when generating our +// MTLVertexDescriptor. We must include the user defined tese input attributes, and any builtins +// that were used. +// +// SPIRV-Cross generates the MSL tese shader code with input attribute indices that reflect the +// specified GLSL locations. Interface blocks are flattened with each member having an incremented +// attribute index. SPIRV-Cross reports an error on compilation if there are clashes in the index +// address space. +// +// After the user specified attributes are processed, SPIRV-Cross places the in-use builtins at the +// next available (lowest value) attribute index. Tese builtins are processed in the following +// order: +// +// in gl_PerVertex +// { +// vec4 gl_Position; +// float gl_PointSize; +// float gl_ClipDistance[]; +// }; +// +// patch in float gl_TessLevelOuter[4]; +// patch in float gl_TessLevelInner[2]; +// +// Enumerations in QShaderDescription::BuiltinType are defined in this order. +// +// For quads, SPIRV-Cross places MTLQuadTessellationFactorsHalf per patch in the tessellation +// factor buffer. For triangles it uses MTLTriangleTessellationFactorsHalf. +// +// It should be noted that SPIRV-Cross handles the following builtin inputs internally, with no +// host side support required. +// +// in vec3 gl_TessCoord; +// in int gl_PatchVerticesIn; +// in int gl_PrimitiveID; +// +id<MTLRenderPipelineState> QMetalGraphicsPipelineData::Tessellation::teseFragRenderPipeline(QRhiMetal *rhiD, QMetalGraphicsPipeline *pipeline) +{ + if (pipeline->d->ps) + return pipeline->d->ps; + + MTLRenderPipelineDescriptor *rpDesc = [[MTLRenderPipelineDescriptor alloc] init]; + MTLVertexDescriptor *vertexDesc = [MTLVertexDescriptor vertexDescriptor]; + + // tesc output buffers + const QMap<int, int> &ebb(compTesc.nativeShaderInfo.extraBufferBindings); + const int tescOutputBufferBinding = ebb.value(QShaderPrivate::MslTessVertTescOutputBufferBinding, -1); + const int tescPatchOutputBufferBinding = ebb.value(QShaderPrivate::MslTessTescPatchOutputBufferBinding, -1); + const int tessFactorBufferBinding = ebb.value(QShaderPrivate::MslTessTescTessLevelBufferBinding, -1); + quint32 offsetInTescOutput = 0; + quint32 offsetInTescPatchOutput = 0; + quint32 offsetInTessFactorBuffer = 0; + quint32 tescOutputAlignment = 0; + quint32 tescPatchOutputAlignment = 0; + quint32 tessFactorAlignment = 0; + QSet<int> usedBuffers; + + // tesc output variables in ascending location order + QMap<int, QShaderDescription::InOutVariable> tescOutVars; + for (const auto &tescOutVar : compTesc.desc.outputVariables()) + tescOutVars[tescOutVar.location] = tescOutVar; + + // tese input variables in ascending location order + QMap<int, QShaderDescription::InOutVariable> teseInVars; + for (const auto &teseInVar : vertTese.desc.inputVariables()) + teseInVars[teseInVar.location] = teseInVar; + + // bit mask tracking usage of vertex attribute indices + quint64 indices = 0; + + for (QShaderDescription::InOutVariable &tescOutVar : tescOutVars) { + + int index = tescOutVar.location; + int binding = -1; + quint32 *offset = nullptr; + quint32 *alignment = nullptr; + + if (tescOutVar.perPatch) { + binding = tescPatchOutputBufferBinding; + offset = &offsetInTescPatchOutput; + alignment = &tescPatchOutputAlignment; + } else { + tescOutVar.arrayDims.removeLast(); + binding = tescOutputBufferBinding; + offset = &offsetInTescOutput; + alignment = &tescOutputAlignment; + } + + if (teseInVars.contains(index)) { + + if (!matches(teseInVars[index], tescOutVar)) { + qWarning() << "mismatched tessellation control output -> tesssellation evaluation input at location" << index; + qWarning() << " tesc out:" << tescOutVar; + qWarning() << " tese in:" << teseInVars[index]; + } + + if (binding != -1) { + addVertexAttribute(tescOutVar, binding, rhiD, index, *offset, vertexDesc.attributes, indices, *alignment); + usedBuffers << binding; + } else { + qWarning() << "baked tessellation control shader missing output buffer binding information"; + addUnusedVertexAttribute(tescOutVar, rhiD, *offset, *alignment); + } + + } else { + qWarning() << "missing tessellation evaluation input for tessellation control output:" << tescOutVar; + addUnusedVertexAttribute(tescOutVar, rhiD, *offset, *alignment); + } + + teseInVars.remove(tescOutVar.location); + } + + for (const QShaderDescription::InOutVariable &teseInVar : teseInVars) + qWarning() << "missing tessellation control output for tessellation evaluation input:" << teseInVar; + + // tesc output builtins in ascending location order + QMap<QShaderDescription::BuiltinType, QShaderDescription::BuiltinVariable> tescOutBuiltins; + for (const auto &tescOutBuiltin : compTesc.desc.outputBuiltinVariables()) + tescOutBuiltins[tescOutBuiltin.type] = tescOutBuiltin; + + // tese input builtins in ascending location order + QMap<QShaderDescription::BuiltinType, QShaderDescription::BuiltinVariable> teseInBuiltins; + for (const auto &teseInBuiltin : vertTese.desc.inputBuiltinVariables()) + teseInBuiltins[teseInBuiltin.type] = teseInBuiltin; + + const bool trianglesMode = vertTese.desc.tessellationMode() == QShaderDescription::TrianglesTessellationMode; + bool tessLevelAdded = false; + + for (const QShaderDescription::BuiltinVariable &builtin : tescOutBuiltins) { + + QShaderDescription::InOutVariable variable; + int binding = -1; + quint32 *offset = nullptr; + quint32 *alignment = nullptr; + + switch (builtin.type) { + case QShaderDescription::BuiltinType::PositionBuiltin: + variable.type = QShaderDescription::VariableType::Vec4; + binding = tescOutputBufferBinding; + offset = &offsetInTescOutput; + alignment = &tescOutputAlignment; + break; + case QShaderDescription::BuiltinType::PointSizeBuiltin: + variable.type = QShaderDescription::VariableType::Float; + binding = tescOutputBufferBinding; + offset = &offsetInTescOutput; + alignment = &tescOutputAlignment; + break; + case QShaderDescription::BuiltinType::ClipDistanceBuiltin: + variable.type = QShaderDescription::VariableType::Float; + variable.arrayDims = builtin.arrayDims; + binding = tescOutputBufferBinding; + offset = &offsetInTescOutput; + alignment = &tescOutputAlignment; + break; + case QShaderDescription::BuiltinType::TessLevelOuterBuiltin: + variable.type = QShaderDescription::VariableType::Half4; + binding = tessFactorBufferBinding; + offset = &offsetInTessFactorBuffer; + tessLevelAdded = trianglesMode; + alignment = &tessFactorAlignment; + break; + case QShaderDescription::BuiltinType::TessLevelInnerBuiltin: + if (trianglesMode) { + if (!tessLevelAdded) { + variable.type = QShaderDescription::VariableType::Half4; + binding = tessFactorBufferBinding; + offsetInTessFactorBuffer = 0; + offset = &offsetInTessFactorBuffer; + alignment = &tessFactorAlignment; + tessLevelAdded = true; + } else { + teseInBuiltins.remove(builtin.type); + continue; + } + } else { + variable.type = QShaderDescription::VariableType::Half2; + binding = tessFactorBufferBinding; + offsetInTessFactorBuffer = 8; + offset = &offsetInTessFactorBuffer; + alignment = &tessFactorAlignment; + } + break; + default: + Q_UNREACHABLE(); + break; + } + + if (teseInBuiltins.contains(builtin.type)) { + if (binding != -1) { + int index = nextAttributeIndex(indices); + addVertexAttribute(variable, binding, rhiD, index, *offset, vertexDesc.attributes, indices, *alignment); + usedBuffers << binding; + } else { + qWarning() << "baked tessellation control shader missing output buffer binding information"; + addUnusedVertexAttribute(variable, rhiD, *offset, *alignment); + } + } else { + addUnusedVertexAttribute(variable, rhiD, *offset, *alignment); + } + + teseInBuiltins.remove(builtin.type); + } + + for (const QShaderDescription::BuiltinVariable &builtin : teseInBuiltins) { + switch (builtin.type) { + case QShaderDescription::BuiltinType::PositionBuiltin: + case QShaderDescription::BuiltinType::PointSizeBuiltin: + case QShaderDescription::BuiltinType::ClipDistanceBuiltin: + qWarning() << "missing tessellation control output for tessellation evaluation builtin input:" << builtin; + break; + default: + break; + } + } + + if (usedBuffers.contains(tescOutputBufferBinding)) { + vertexDesc.layouts[tescOutputBufferBinding].stepFunction = MTLVertexStepFunctionPerPatchControlPoint; + vertexDesc.layouts[tescOutputBufferBinding].stride = aligned(offsetInTescOutput, tescOutputAlignment); + } + + if (usedBuffers.contains(tescPatchOutputBufferBinding)) { + vertexDesc.layouts[tescPatchOutputBufferBinding].stepFunction = MTLVertexStepFunctionPerPatch; + vertexDesc.layouts[tescPatchOutputBufferBinding].stride = aligned(offsetInTescPatchOutput, tescPatchOutputAlignment); + } + + if (usedBuffers.contains(tessFactorBufferBinding)) { + vertexDesc.layouts[tessFactorBufferBinding].stepFunction = MTLVertexStepFunctionPerPatch; + vertexDesc.layouts[tessFactorBufferBinding].stride = trianglesMode ? sizeof(MTLTriangleTessellationFactorsHalf) : sizeof(MTLQuadTessellationFactorsHalf); + } + + rpDesc.vertexDescriptor = vertexDesc; + rpDesc.vertexFunction = vertTese.func; + rpDesc.fragmentFunction = pipeline->d->fs.func; + + // The portable, cross-API approach is to use CCW, the results are then + // identical (assuming the applied clipSpaceCorrMatrix) for all the 3D + // APIs. The tess.eval. GLSL shader is thus expected to specify ccw. If it + // doesn't, things may not work as expected. + rpDesc.tessellationOutputWindingOrder = toMetalTessellationWindingOrder(vertTese.desc.tessellationWindingOrder()); + + rpDesc.tessellationPartitionMode = toMetalTessellationPartitionMode(vertTese.desc.tessellationPartitioning()); + + QMetalRenderPassDescriptor *rpD = QRHI_RES(QMetalRenderPassDescriptor, pipeline->renderPassDescriptor()); + pipeline->setupAttachmentsInMetalRenderPassDescriptor(rpDesc, rpD); + + rhiD->d->trySeedingRenderPipelineFromBinaryArchive(rpDesc); + + if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + rhiD->d->addRenderPipelineToBinaryArchive(rpDesc); + + NSError *err = nil; + id<MTLRenderPipelineState> ps = [rhiD->d->dev newRenderPipelineStateWithDescriptor: rpDesc error: &err]; [rpDesc release]; + if (!ps) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("Failed to create render pipeline state for tessellation: %s", qPrintable(msg)); + } else { + // ps is stored in the QMetalGraphicsPipelineData so the end result in this + // regard is no different from what createVertexFragmentPipeline does + pipeline->d->ps = ps; + } + return ps; +} - MTLDepthStencilDescriptor *dsDesc = [[MTLDepthStencilDescriptor alloc] init]; - dsDesc.depthCompareFunction = m_depthTest ? toMetalCompareOp(m_depthOp) : MTLCompareFunctionAlways; - dsDesc.depthWriteEnabled = m_depthWrite; - if (m_stencilTest) { - dsDesc.frontFaceStencil = [[MTLStencilDescriptor alloc] init]; - dsDesc.frontFaceStencil.stencilFailureOperation = toMetalStencilOp(m_stencilFront.failOp); - dsDesc.frontFaceStencil.depthFailureOperation = toMetalStencilOp(m_stencilFront.depthFailOp); - dsDesc.frontFaceStencil.depthStencilPassOperation = toMetalStencilOp(m_stencilFront.passOp); - dsDesc.frontFaceStencil.stencilCompareFunction = toMetalCompareOp(m_stencilFront.compareOp); - dsDesc.frontFaceStencil.readMask = m_stencilReadMask; - dsDesc.frontFaceStencil.writeMask = m_stencilWriteMask; +QMetalBuffer *QMetalGraphicsPipelineData::ExtraBufferManager::acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type) +{ + QVector<QMetalBuffer *> *workBuffers = type == WorkBufType::DeviceLocal ? &deviceLocalWorkBuffers : &hostVisibleWorkBuffers; - dsDesc.backFaceStencil = [[MTLStencilDescriptor alloc] init]; - dsDesc.backFaceStencil.stencilFailureOperation = toMetalStencilOp(m_stencilBack.failOp); - dsDesc.backFaceStencil.depthFailureOperation = toMetalStencilOp(m_stencilBack.depthFailOp); - dsDesc.backFaceStencil.depthStencilPassOperation = toMetalStencilOp(m_stencilBack.passOp); - dsDesc.backFaceStencil.stencilCompareFunction = toMetalCompareOp(m_stencilBack.compareOp); - dsDesc.backFaceStencil.readMask = m_stencilReadMask; - dsDesc.backFaceStencil.writeMask = m_stencilWriteMask; + // Check if something is reusable as-is. + for (QMetalBuffer *workBuf : *workBuffers) { + if (workBuf && workBuf->lastActiveFrameSlot == -1 && workBuf->size() >= size) { + workBuf->lastActiveFrameSlot = rhiD->currentFrameSlot; + return workBuf; + } + } + + // Once the pool is above a certain threshold, see if there is something + // unused (but too small) and recreate that our size. + if (workBuffers->count() > QMTL_FRAMES_IN_FLIGHT * 8) { + for (QMetalBuffer *workBuf : *workBuffers) { + if (workBuf && workBuf->lastActiveFrameSlot == -1) { + workBuf->setSize(size); + if (workBuf->create()) { + workBuf->lastActiveFrameSlot = rhiD->currentFrameSlot; + return workBuf; + } + } + } + } + + // Add a new buffer to the pool. + QMetalBuffer *buf; + if (type == WorkBufType::DeviceLocal) { + // for GPU->GPU data (non-slotted, not necessarily host writable) + buf = new QMetalBuffer(rhiD, QRhiBuffer::Static, QRhiBuffer::UsageFlags(QMetalBuffer::WorkBufPoolUsage), size); + } else { + // for CPU->GPU (non-slotted, host writable/coherent) + buf = new QMetalBuffer(rhiD, QRhiBuffer::Dynamic, QRhiBuffer::UsageFlags(QMetalBuffer::WorkBufPoolUsage), size); + } + if (buf->create()) { + buf->lastActiveFrameSlot = rhiD->currentFrameSlot; + workBuffers->append(buf); + return buf; + } + + qWarning("Failed to acquire work buffer of size %u", size); + return nullptr; +} + +bool QMetalGraphicsPipeline::createTessellationPipelines(const QShader &tessVert, const QShader &tesc, const QShader &tese, const QShader &tessFrag) +{ + QRHI_RES_RHI(QRhiMetal); + QString error; + QByteArray entryPoint; + QShaderKey activeKey; + + const QShaderDescription tescDesc = tesc.description(); + const QShaderDescription teseDesc = tese.description(); + d->tess.inControlPointCount = uint(m_patchControlPointCount); + d->tess.outControlPointCount = tescDesc.tessellationOutputVertexCount(); + if (!d->tess.outControlPointCount) + d->tess.outControlPointCount = teseDesc.tessellationOutputVertexCount(); + + if (!d->tess.outControlPointCount) { + qWarning("Failed to determine output vertex count from the tessellation control or evaluation shader, cannot tessellate"); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + + if (m_multiViewCount >= 2) + qWarning("Multiview is not supported with tessellation"); + + // Now the vertex shader is a compute shader. + // It should have three dedicated *VertexAsComputeShader variants. + // What the requested variant was (Standard or Batchable) plays no role here. + // (the Qt Quick scenegraph does not use tessellation with its materials) + // Create all three versions. + + bool variantsPresent[3] = {}; + const QVector<QShaderKey> tessVertKeys = tessVert.availableShaders(); + for (const QShaderKey &k : tessVertKeys) { + switch (k.sourceVariant()) { + case QShader::NonIndexedVertexAsComputeShader: + variantsPresent[0] = true; + break; + case QShader::UInt32IndexedVertexAsComputeShader: + variantsPresent[1] = true; + break; + case QShader::UInt16IndexedVertexAsComputeShader: + variantsPresent[2] = true; + break; + default: + break; + } + } + if (!(variantsPresent[0] && variantsPresent[1] && variantsPresent[2])) { + qWarning("Vertex shader is not prepared for Metal tessellation. Cannot tessellate. " + "Perhaps the relevant variants (UInt32IndexedVertexAsComputeShader et al) were not generated? " + "Try passing --msltess to qsb."); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + + int varIndex = 0; // Will map NonIndexed as 0, UInt32 as 1, UInt16 as 2. Do not change this ordering. + for (QShader::Variant variant : { + QShader::NonIndexedVertexAsComputeShader, + QShader::UInt32IndexedVertexAsComputeShader, + QShader::UInt16IndexedVertexAsComputeShader }) + { + id<MTLLibrary> lib = rhiD->d->createMetalLib(tessVert, variant, &error, &entryPoint, &activeKey); + if (!lib) { + qWarning("MSL shader compilation failed for vertex-as-compute shader %d: %s", int(variant), qPrintable(error)); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + id<MTLFunction> func = rhiD->d->createMSLShaderFunction(lib, entryPoint); + if (!func) { + qWarning("MSL function for entry point %s not found", entryPoint.constData()); + [lib release]; + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + QMetalShader &compVs(d->tess.compVs[varIndex]); + compVs.lib = lib; + compVs.func = func; + compVs.desc = tessVert.description(); + compVs.nativeResourceBindingMap = tessVert.nativeResourceBindingMap(activeKey); + compVs.nativeShaderInfo = tessVert.nativeShaderInfo(activeKey); + + // pre-create all three MTLComputePipelineStates + if (!d->tess.vsCompPipeline(rhiD, variant)) { + qWarning("Failed to pre-generate compute pipeline for vertex compute shader (tessellation variant %d)", int(variant)); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + + ++varIndex; + } + + // Pipeline #2 is a compute that runs the tessellation control (compute) shader + id<MTLLibrary> tessControlLib = rhiD->d->createMetalLib(tesc, QShader::StandardShader, &error, &entryPoint, &activeKey); + if (!tessControlLib) { + qWarning("MSL shader compilation failed for tessellation control compute shader: %s", qPrintable(error)); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + id<MTLFunction> tessControlFunc = rhiD->d->createMSLShaderFunction(tessControlLib, entryPoint); + if (!tessControlFunc) { + qWarning("MSL function for entry point %s not found", entryPoint.constData()); + [tessControlLib release]; + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + d->tess.compTesc.lib = tessControlLib; + d->tess.compTesc.func = tessControlFunc; + d->tess.compTesc.desc = tesc.description(); + d->tess.compTesc.nativeResourceBindingMap = tesc.nativeResourceBindingMap(activeKey); + d->tess.compTesc.nativeShaderInfo = tesc.nativeShaderInfo(activeKey); + if (!d->tess.tescCompPipeline(rhiD)) { + qWarning("Failed to pre-generate compute pipeline for tessellation control shader"); + d->tess.enabled = false; + d->tess.failed = true; + return false; } + // Pipeline #3 is a render pipeline with the tessellation evaluation (vertex) + the fragment shader + id<MTLLibrary> tessEvalLib = rhiD->d->createMetalLib(tese, QShader::StandardShader, &error, &entryPoint, &activeKey); + if (!tessEvalLib) { + qWarning("MSL shader compilation failed for tessellation evaluation vertex shader: %s", qPrintable(error)); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + id<MTLFunction> tessEvalFunc = rhiD->d->createMSLShaderFunction(tessEvalLib, entryPoint); + if (!tessEvalFunc) { + qWarning("MSL function for entry point %s not found", entryPoint.constData()); + [tessEvalLib release]; + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + d->tess.vertTese.lib = tessEvalLib; + d->tess.vertTese.func = tessEvalFunc; + d->tess.vertTese.desc = tese.description(); + d->tess.vertTese.nativeResourceBindingMap = tese.nativeResourceBindingMap(activeKey); + d->tess.vertTese.nativeShaderInfo = tese.nativeShaderInfo(activeKey); + + id<MTLLibrary> fragLib = rhiD->d->createMetalLib(tessFrag, QShader::StandardShader, &error, &entryPoint, &activeKey); + if (!fragLib) { + qWarning("MSL shader compilation failed for fragment shader: %s", qPrintable(error)); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + id<MTLFunction> fragFunc = rhiD->d->createMSLShaderFunction(fragLib, entryPoint); + if (!fragFunc) { + qWarning("MSL function for entry point %s not found", entryPoint.constData()); + [fragLib release]; + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + d->fs.lib = fragLib; + d->fs.func = fragFunc; + d->fs.desc = tessFrag.description(); + d->fs.nativeShaderInfo = tessFrag.nativeShaderInfo(activeKey); + d->fs.nativeResourceBindingMap = tessFrag.nativeResourceBindingMap(activeKey); + + if (!d->tess.teseFragRenderPipeline(rhiD, this)) { + qWarning("Failed to pre-generate render pipeline for tessellation evaluation + fragment shader"); + d->tess.enabled = false; + d->tess.failed = true; + return false; + } + + MTLDepthStencilDescriptor *dsDesc = [[MTLDepthStencilDescriptor alloc] init]; + setupMetalDepthStencilDescriptor(dsDesc); d->ds = [rhiD->d->dev newDepthStencilStateWithDescriptor: dsDesc]; [dsDesc release]; - d->primitiveType = toMetalPrimitiveType(m_topology); - d->winding = m_frontFace == CCW ? MTLWindingCounterClockwise : MTLWindingClockwise; - d->cullMode = toMetalCullMode(m_cullMode); - d->depthBias = float(m_depthBias); - d->slopeScaledDepthBias = m_slopeScaledDepthBias; + // no primitiveType + mapStates(); + + return true; +} + +bool QMetalGraphicsPipeline::create() +{ + destroy(); // no early test, always invoke and leave it to destroy to decide what to clean up + + QRHI_RES_RHI(QRhiMetal); + rhiD->pipelineCreationStart(); + if (!rhiD->sanityCheckGraphicsPipeline(this)) + return false; + + // See if tessellation is involved. Things will be very different, if so. + QShader tessVert; + QShader tesc; + QShader tese; + QShader tessFrag; + for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { + switch (shaderStage.type()) { + case QRhiShaderStage::Vertex: + tessVert = shaderStage.shader(); + break; + case QRhiShaderStage::TessellationControl: + tesc = shaderStage.shader(); + break; + case QRhiShaderStage::TessellationEvaluation: + tese = shaderStage.shader(); + break; + case QRhiShaderStage::Fragment: + tessFrag = shaderStage.shader(); + break; + default: + break; + } + } + d->tess.enabled = tesc.isValid() && tese.isValid() && m_topology == Patches && m_patchControlPointCount > 0; + d->tess.failed = false; + + bool ok = d->tess.enabled ? createTessellationPipelines(tessVert, tesc, tese, tessFrag) : createVertexFragmentPipeline(); + if (!ok) + return false; + + // SPIRV-Cross buffer size buffers + int buffers = 0; + QVarLengthArray<QMetalShader *, 6> shaders; + if (d->tess.enabled) { + shaders.append(&d->tess.compVs[0]); + shaders.append(&d->tess.compVs[1]); + shaders.append(&d->tess.compVs[2]); + shaders.append(&d->tess.compTesc); + shaders.append(&d->tess.vertTese); + } else { + shaders.append(&d->vs); + } + shaders.append(&d->fs); + + for (QMetalShader *shader : shaders) { + if (shader->nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) { + const int binding = shader->nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding]; + shader->nativeResourceBindingMap[binding] = qMakePair(binding, -1); + int maxNativeBinding = 0; + for (const QShaderDescription::StorageBlock &block : shader->desc.storageBlocks()) + maxNativeBinding = qMax(maxNativeBinding, shader->nativeResourceBindingMap[block.binding].first); + + // we use one buffer to hold data for all graphics shader stages, each with a different offset. + // buffer offsets must be 32byte aligned - adjust buffer count accordingly + buffers += ((maxNativeBinding + 1 + 7) / 8) * 8; + } + } + + if (buffers) { + if (!d->bufferSizeBuffer) + d->bufferSizeBuffer = new QMetalBuffer(rhiD, QRhiBuffer::Static, QRhiBuffer::StorageBuffer, buffers * sizeof(int)); + + d->bufferSizeBuffer->setSize(buffers * sizeof(int)); + d->bufferSizeBuffer->create(); + } + rhiD->pipelineCreationEnd(); lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -3672,17 +5917,48 @@ QMetalComputePipeline::~QMetalComputePipeline() void QMetalComputePipeline::destroy() { - QRHI_RES_RHI(QRhiMetal); - d->cs.destroy(); if (!d->ps) return; - [d->ps release]; + delete d->bufferSizeBuffer; + d->bufferSizeBuffer = nullptr; + + QRhiMetalData::DeferredReleaseEntry e; + e.type = QRhiMetalData::DeferredReleaseEntry::ComputePipeline; + e.lastActiveFrameSlot = lastActiveFrameSlot; + e.computePipeline.pipelineState = d->ps; d->ps = nil; - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiMetal); + if (rhiD) { + rhiD->d->releaseQueue.append(e); + rhiD->unregisterResource(this); + } +} + +void QRhiMetalData::trySeedingComputePipelineFromBinaryArchive(MTLComputePipelineDescriptor *cpDesc) +{ + if (@available(macOS 11.0, iOS 14.0, *)) { + if (binArch) { + NSArray *binArchArray = [NSArray arrayWithObjects: binArch, nil]; + cpDesc.binaryArchives = binArchArray; + } + } +} + +void QRhiMetalData::addComputePipelineToBinaryArchive(MTLComputePipelineDescriptor *cpDesc) +{ + if (@available(macOS 11.0, iOS 14.0, *)) { + if (binArch) { + NSError *err = nil; + if (![binArch addComputePipelineFunctionsWithDescriptor: cpDesc error: &err]) { + const QString msg = QString::fromNSString(err.localizedDescription); + qWarning("Failed to collect compute pipeline functions to binary archive: %s", qPrintable(msg)); + } + } + } } bool QMetalComputePipeline::create() @@ -3691,6 +5967,7 @@ bool QMetalComputePipeline::create() destroy(); QRHI_RES_RHI(QRhiMetal); + rhiD->pipelineCreationStart(); auto cacheIt = rhiD->d->shaderCache.constFind(m_shaderStage); if (cacheIt != rhiD->d->shaderCache.constEnd()) { @@ -3715,8 +5992,15 @@ bool QMetalComputePipeline::create() d->cs.lib = lib; d->cs.func = func; d->cs.localSize = shader.description().computeShaderLocalSize(); - if (const QShader::NativeResourceBindingMap *map = shader.nativeResourceBindingMap(activeKey)) - d->cs.nativeResourceBindingMap = *map; + d->cs.nativeResourceBindingMap = shader.nativeResourceBindingMap(activeKey); + d->cs.desc = shader.description(); + d->cs.nativeShaderInfo = shader.nativeShaderInfo(activeKey); + + // SPIRV-Cross buffer size buffers + if (d->cs.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) { + const int binding = d->cs.nativeShaderInfo.extraBufferBindings[QShaderPrivate::MslBufferSizeBufferBinding]; + d->cs.nativeResourceBindingMap[binding] = qMakePair(binding, -1); + } if (rhiD->d->shaderCache.count() >= QRhiMetal::MAX_SHADER_CACHE_ENTRIES) { for (QMetalShader &s : rhiD->d->shaderCache) @@ -3731,14 +6015,42 @@ bool QMetalComputePipeline::create() d->localSize = MTLSizeMake(d->cs.localSize[0], d->cs.localSize[1], d->cs.localSize[2]); + MTLComputePipelineDescriptor *cpDesc = [MTLComputePipelineDescriptor new]; + cpDesc.computeFunction = d->cs.func; + + rhiD->d->trySeedingComputePipelineFromBinaryArchive(cpDesc); + + if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) + rhiD->d->addComputePipelineToBinaryArchive(cpDesc); + NSError *err = nil; - d->ps = [rhiD->d->dev newComputePipelineStateWithFunction: d->cs.func error: &err]; + d->ps = [rhiD->d->dev newComputePipelineStateWithDescriptor: cpDesc + options: MTLPipelineOptionNone + reflection: nil + error: &err]; + [cpDesc release]; if (!d->ps) { const QString msg = QString::fromNSString(err.localizedDescription); - qWarning("Failed to create render pipeline state: %s", qPrintable(msg)); + qWarning("Failed to create compute pipeline state: %s", qPrintable(msg)); return false; } + // SPIRV-Cross buffer size buffers + if (d->cs.nativeShaderInfo.extraBufferBindings.contains(QShaderPrivate::MslBufferSizeBufferBinding)) { + int buffers = 0; + for (const QShaderDescription::StorageBlock &block : d->cs.desc.storageBlocks()) + buffers = qMax(buffers, d->cs.nativeResourceBindingMap[block.binding].first); + + buffers += 1; + + if (!d->bufferSizeBuffer) + d->bufferSizeBuffer = new QMetalBuffer(rhiD, QRhiBuffer::Static, QRhiBuffer::StorageBuffer, buffers * sizeof(int)); + + d->bufferSizeBuffer->setSize(buffers * sizeof(int)); + d->bufferSizeBuffer->create(); + } + + rhiD->pipelineCreationEnd(); lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -3770,10 +6082,12 @@ const QRhiNativeHandles *QMetalCommandBuffer::nativeHandles() return &nativeHandlesStruct; } -void QMetalCommandBuffer::resetState() +void QMetalCommandBuffer::resetState(double lastGpuTime) { + d->lastGpuTime = lastGpuTime; d->currentRenderPassEncoder = nil; d->currentComputePassEncoder = nil; + d->tessellationComputeEncoder = nil; d->currentPassRpDesc = nil; resetPerPassState(); } @@ -3798,9 +6112,12 @@ void QMetalCommandBuffer::resetPerPassCachedState() currentIndexOffset = 0; currentIndexFormat = QRhiCommandBuffer::IndexUInt16; currentCullMode = -1; + currentTriangleFillMode = -1; currentFrontFaceWinding = -1; currentDepthBiasValues = { 0.0f, 0.0f }; + d->currentShaderResourceBindingState = {}; + d->currentDepthStencilState = nil; d->currentFirstVertexBinding = -1; d->currentVertexInputsBuffers.clear(); d->currentVertexInputOffsets.clear(); @@ -3808,7 +6125,7 @@ void QMetalCommandBuffer::resetPerPassCachedState() QMetalSwapChain::QMetalSwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), - rtWrapper(rhi), + rtWrapper(rhi, this), cbWrapper(rhi), d(new QMetalSwapChainData) { @@ -3832,8 +6149,7 @@ void QMetalSwapChain::destroy() for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) { if (d->sem[i]) { // the semaphores cannot be released if they do not have the initial value - dispatch_semaphore_wait(d->sem[i], DISPATCH_TIME_FOREVER); - dispatch_semaphore_signal(d->sem[i]); + waitUntilCompleted(i); dispatch_release(d->sem[i]); d->sem[i] = nullptr; @@ -3845,18 +6161,23 @@ void QMetalSwapChain::destroy() d->msaaTex[i] = nil; } +#ifdef Q_OS_MACOS + d->liveResizeStartObserver.remove(); + d->liveResizeEndObserver.remove(); + d->liveResizeObserverSet = false; +#endif + d->layer = nullptr; + m_proxyData = {}; [d->curDrawable release]; d->curDrawable = nil; QRHI_RES_RHI(QRhiMetal); - rhiD->swapchains.remove(this); - - QRHI_PROF; - QRHI_PROF_F(releaseSwapChain(this)); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->swapchains.remove(this); + rhiD->unregisterResource(this); + } } QRhiCommandBuffer *QMetalSwapChain::currentFrameCommandBuffer() @@ -3869,6 +6190,9 @@ QRhiRenderTarget *QMetalSwapChain::currentFrameRenderTarget() return &rtWrapper; } +// view.layer should ideally be called on the main thread, otherwise the UI +// Thread Checker in Xcode drops a warning. Hence trying to proxy it through +// QRhiSwapChainProxyData instead of just calling this function directly. static inline CAMetalLayer *layerForWindow(QWindow *window) { Q_ASSERT(window); @@ -3881,21 +6205,51 @@ static inline CAMetalLayer *layerForWindow(QWindow *window) return static_cast<CAMetalLayer *>(view.layer); } +// If someone calls this, it is hopefully from the main thread, and they will +// then set the returned data on the QRhiSwapChain, so it won't need to query +// the layer on its own later on. +QRhiSwapChainProxyData QRhiMetal::updateSwapChainProxyData(QWindow *window) +{ + QRhiSwapChainProxyData d; + d.reserved[0] = layerForWindow(window); + return d; +} + QSize QMetalSwapChain::surfacePixelSize() { Q_ASSERT(m_window); CAMetalLayer *layer = d->layer; if (!layer) - layer = layerForWindow(m_window); + layer = qrhi_objectFromProxyData<CAMetalLayer>(&m_proxyData, m_window, QRhi::Metal, 0); - CGSize layerSize = layer.bounds.size; - layerSize.width *= layer.contentsScale; - layerSize.height *= layer.contentsScale; - return QSizeF::fromCGSize(layerSize).toSize(); + Q_ASSERT(layer); + int height = (int)layer.bounds.size.height; + int width = (int)layer.bounds.size.width; + width *= layer.contentsScale; + height *= layer.contentsScale; + return QSize(width, height); +} + +bool QMetalSwapChain::isFormatSupported(Format f) +{ + if (f == HDRExtendedSrgbLinear) { + if (@available(macOS 10.11, iOS 16.0, *)) + return hdrInfo().limits.colorComponentValue.maxPotentialColorComponentValue > 1.0f; + else + return false; + } else if (f == HDRExtendedDisplayP3Linear) { + if (@available(macOS 11.0, iOS 14.0, *)) + return hdrInfo().limits.colorComponentValue.maxPotentialColorComponentValue > 1.0f; + else + return false; + } + return f == SDR; } QRhiRenderPassDescriptor *QMetalSwapChain::newCompatibleRenderPassDescriptor() { + QRHI_RES_RHI(QRhiMetal); + chooseFormats(); // ensure colorFormat and similar are filled out QMetalRenderPassDescriptor *rpD = new QMetalRenderPassDescriptor(m_rhi); @@ -3906,13 +6260,15 @@ QRhiRenderPassDescriptor *QMetalSwapChain::newCompatibleRenderPassDescriptor() #ifdef Q_OS_MACOS // m_depthStencil may not be built yet so cannot rely on computed fields in it - QRHI_RES_RHI(QRhiMetal); rpD->dsFormat = rhiD->d->dev.depth24Stencil8PixelFormatSupported ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; #else rpD->dsFormat = MTLPixelFormatDepth32Float_Stencil8; #endif + rpD->updateSerializedFormat(); + + rhiD->registerResource(rpD, false); return rpD; } @@ -3921,10 +6277,26 @@ void QMetalSwapChain::chooseFormats() QRHI_RES_RHI(QRhiMetal); samples = rhiD->effectiveSampleCount(m_sampleCount); // pick a format that is allowed for CAMetalLayer.pixelFormat + if (m_format == HDRExtendedSrgbLinear || m_format == HDRExtendedDisplayP3Linear) { + d->colorFormat = MTLPixelFormatRGBA16Float; + d->rhiColorFormat = QRhiTexture::RGBA16F; + return; + } d->colorFormat = m_flags.testFlag(sRGB) ? MTLPixelFormatBGRA8Unorm_sRGB : MTLPixelFormatBGRA8Unorm; d->rhiColorFormat = QRhiTexture::BGRA8; } +void QMetalSwapChain::waitUntilCompleted(int slot) +{ + // wait+signal is the general pattern to ensure the commands for a + // given frame slot have completed (if sem is 1, we go 0 then 1; if + // sem is 0 we go -1, block, completion increments to 0, then us to 1) + + dispatch_semaphore_t sem = d->sem[slot]; + dispatch_semaphore_wait(sem, DISPATCH_TIME_FOREVER); + dispatch_semaphore_signal(sem); +} + bool QMetalSwapChain::createOrResize() { Q_ASSERT(m_window); @@ -3946,13 +6318,25 @@ bool QMetalSwapChain::createOrResize() return false; } - d->layer = layerForWindow(window); + d->layer = qrhi_objectFromProxyData<CAMetalLayer>(&m_proxyData, window, QRhi::Metal, 0); Q_ASSERT(d->layer); chooseFormats(); if (d->colorFormat != d->layer.pixelFormat) d->layer.pixelFormat = d->colorFormat; + if (m_format == HDRExtendedSrgbLinear) { + if (@available(macOS 10.11, iOS 16.0, *)) { + d->layer.colorspace = CGColorSpaceCreateWithName(kCGColorSpaceExtendedLinearSRGB); + d->layer.wantsExtendedDynamicRangeContent = YES; + } + } else if (m_format == HDRExtendedDisplayP3Linear) { + if (@available(macOS 11.0, iOS 16.0, *)) { + d->layer.colorspace = CGColorSpaceCreateWithName(kCGColorSpaceExtendedLinearDisplayP3); + d->layer.wantsExtendedDynamicRangeContent = YES; + } + } + if (m_flags.testFlag(UsedAsTransferSource)) d->layer.framebufferOnly = NO; @@ -3976,9 +6360,12 @@ bool QMetalSwapChain::createOrResize() // Now set the layer's drawableSize which will stay set to the same value // until the next createOrResize(), thus ensuring atomicity with regards to // the drawable size in frames. - CGSize layerSize = d->layer.bounds.size; - layerSize.width *= d->layer.contentsScale; - layerSize.height *= d->layer.contentsScale; + int width = (int)d->layer.bounds.size.width; + int height = (int)d->layer.bounds.size.height; + CGSize layerSize = CGSizeMake(width, height); + const float scaleFactor = d->layer.contentsScale; + layerSize.width *= scaleFactor; + layerSize.height *= scaleFactor; d->layer.drawableSize = layerSize; m_currentPixelSize = QSizeF::fromCGSize(layerSize).toSize(); @@ -3986,10 +6373,39 @@ bool QMetalSwapChain::createOrResize() [d->layer setDevice: rhiD->d->dev]; +#ifdef Q_OS_MACOS + // Can only use presentsWithTransaction (to get smooth resizing) when + // presenting from the main (gui) thread. We predict that based on the + // thread this function is called on since if the QRhiSwapChain is + // initialied on a given thread then that's almost certainly the thread on + // which the QRhi renders and presents. + const bool canUsePresentsWithTransaction = NSThread.isMainThread; + + // Have an env.var. just in case it turns out presentsWithTransaction is + // not desired in some specific case. + static bool allowPresentsWithTransaction = !qEnvironmentVariableIntValue("QT_MTL_NO_TRANSACTION"); + + if (allowPresentsWithTransaction && canUsePresentsWithTransaction && !d->liveResizeObserverSet) { + d->liveResizeObserverSet = true; + NSView *view = reinterpret_cast<NSView *>(window->winId()); + NSWindow *window = view.window; + if (window) { + qCDebug(QRHI_LOG_INFO, "will set presentsWithTransaction during live resize"); + d->liveResizeStartObserver = QMacNotificationObserver(window, NSWindowWillStartLiveResizeNotification, [this] { + d->layer.presentsWithTransaction = true; + }); + d->liveResizeEndObserver = QMacNotificationObserver(window, NSWindowDidEndLiveResizeNotification, [this] { + d->layer.presentsWithTransaction = false; + }); + } + } +#endif + [d->curDrawable release]; d->curDrawable = nil; for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) { + d->lastGpuTime[i] = 0; if (!d->sem[i]) d->sem[i] = dispatch_semaphore_create(QMTL_FRAMES_IN_FLIGHT - 1); } @@ -4015,13 +6431,15 @@ bool QMetalSwapChain::createOrResize() } } + rtWrapper.setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget rtWrapper.d->pixelSize = pixelSize; - rtWrapper.d->dpr = float(window->devicePixelRatio()); + rtWrapper.d->dpr = scaleFactor; rtWrapper.d->sampleCount = samples; rtWrapper.d->colorAttCount = 1; rtWrapper.d->dsAttCount = ds ? 1 : 0; - qCDebug(QRHI_LOG_INFO, "got CAMetalLayer, size %dx%d", pixelSize.width(), pixelSize.height()); + qCDebug(QRHI_LOG_INFO, "got CAMetalLayer, pixel size %dx%d (scale %.2f)", + pixelSize.width(), pixelSize.height(), scaleFactor); if (samples > 1) { MTLTextureDescriptor *desc = [[MTLTextureDescriptor alloc] init]; @@ -4040,13 +6458,39 @@ bool QMetalSwapChain::createOrResize() [desc release]; } - QRHI_PROF; - QRHI_PROF_F(resizeSwapChain(this, QMTL_FRAMES_IN_FLIGHT, samples > 1 ? QMTL_FRAMES_IN_FLIGHT : 0, samples)); - if (needsRegistration) rhiD->registerResource(this); return true; } +QRhiSwapChainHdrInfo QMetalSwapChain::hdrInfo() +{ + QRhiSwapChainHdrInfo info; + info.limitsType = QRhiSwapChainHdrInfo::ColorComponentValue; + info.limits.colorComponentValue.maxColorComponentValue = 1; + info.limits.colorComponentValue.maxPotentialColorComponentValue = 1; + info.luminanceBehavior = QRhiSwapChainHdrInfo::DisplayReferred; // 1.0 = SDR white + info.sdrWhiteLevel = 200; // typical value, but dummy (don't know the real one); won't matter due to being display-referred + + if (m_window) { + // Must use m_window, not window, given this may be called before createOrResize(). +#if defined(Q_OS_MACOS) + NSView *view = reinterpret_cast<NSView *>(m_window->winId()); + NSScreen *screen = view.window.screen; + info.limits.colorComponentValue.maxColorComponentValue = screen.maximumExtendedDynamicRangeColorComponentValue; + info.limits.colorComponentValue.maxPotentialColorComponentValue = screen.maximumPotentialExtendedDynamicRangeColorComponentValue; +#elif defined(Q_OS_IOS) + if (@available(iOS 16.0, *)) { + UIView *view = reinterpret_cast<UIView *>(m_window->winId()); + UIScreen *screen = view.window.windowScene.screen; + info.limits.colorComponentValue.maxColorComponentValue = view.window.windowScene.screen.currentEDRHeadroom; + info.limits.colorComponentValue.maxPotentialColorComponentValue = screen.potentialEDRHeadroom; + } +#endif + } + + return info; +} + QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhimetal_p.h b/src/gui/rhi/qrhimetal_p.h index 58025c02f2..f539148b2c 100644 --- a/src/gui/rhi/qrhimetal_p.h +++ b/src/gui/rhi/qrhimetal_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIMETAL_H -#define QRHIMETAL_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHIMETAL_P_H +#define QRHIMETAL_P_H // // W A R N I N G @@ -51,29 +15,494 @@ // We mean it. // -#include <private/qrhi_p.h> - -Q_FORWARD_DECLARE_OBJC_CLASS(MTLDevice); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandQueue); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandBuffer); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLRenderCommandEncoder); +#include "qrhi_p.h" +#include <QWindow> QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiMetalInitParams : public QRhiInitParams +static const int QMTL_FRAMES_IN_FLIGHT = 2; + +// have to hide the ObjC stuff, this header cannot contain MTL* at all +struct QMetalBufferData; + +struct QMetalBuffer : public QRhiBuffer +{ + QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QMetalBuffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + QMetalBufferData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; + + static constexpr int WorkBufPoolUsage = 1 << 8; + static_assert(WorkBufPoolUsage > QRhiBuffer::StorageBuffer); +}; + +struct QMetalRenderBufferData; + +struct QMetalRenderBuffer : public QRhiRenderBuffer +{ + QMetalRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QMetalRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + QMetalRenderBufferData *d; + int samples = 1; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalTextureData; + +struct QMetalTexture : public QRhiTexture +{ + QMetalTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QMetalTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + + QMetalTextureData *d; + int mipLevelCount = 0; + int samples = 1; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; + friend struct QMetalTextureData; +}; + +struct QMetalSamplerData; + +struct QMetalSampler : public QRhiSampler +{ + QMetalSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QMetalSampler(); + void destroy() override; + bool create() override; + + QMetalSamplerData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; +}; + +struct QMetalRenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QMetalRenderPassDescriptor(QRhiImplementation *rhi); + ~QMetalRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + + void updateSerializedFormat(); + + // there is no MTLRenderPassDescriptor here as one will be created for each pass in beginPass() + + // but the things needed for the render pipeline descriptor have to be provided + static const int MAX_COLOR_ATTACHMENTS = 8; + int colorAttachmentCount = 0; + bool hasDepthStencil = false; + int colorFormat[MAX_COLOR_ATTACHMENTS]; + int dsFormat; + QVector<quint32> serializedFormatData; +}; + +struct QMetalRenderTargetData; + +struct QMetalSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QMetalSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QMetalSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QMetalRenderTargetData *d; +}; + +struct QMetalTextureRenderTarget : public QRhiTextureRenderTarget +{ + QMetalTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QMetalTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QMetalRenderTargetData *d; + friend class QRhiMetal; +}; + +struct QMetalShaderResourceBindings : public QRhiShaderResourceBindings +{ + QMetalShaderResourceBindings(QRhiImplementation *rhi); + ~QMetalShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + int maxBinding = -1; + + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData; + + uint generation = 0; + friend class QRhiMetal; +}; + +struct QMetalGraphicsPipelineData; +struct QMetalCommandBuffer; + +struct QMetalGraphicsPipeline : public QRhiGraphicsPipeline +{ + QMetalGraphicsPipeline(QRhiImplementation *rhi); + ~QMetalGraphicsPipeline(); + void destroy() override; + bool create() override; + + void makeActiveForCurrentRenderPassEncoder(QMetalCommandBuffer *cbD); + void setupAttachmentsInMetalRenderPassDescriptor(void *metalRpDesc, QMetalRenderPassDescriptor *rpD); + void setupMetalDepthStencilDescriptor(void *metalDsDesc); + void mapStates(); + bool createVertexFragmentPipeline(); + bool createTessellationPipelines(const QShader &tessVert, const QShader &tesc, const QShader &tese, const QShader &tessFrag); + + QMetalGraphicsPipelineData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalComputePipelineData; + +struct QMetalComputePipeline : public QRhiComputePipeline +{ + QMetalComputePipeline(QRhiImplementation *rhi); + ~QMetalComputePipeline(); + void destroy() override; + bool create() override; + + QMetalComputePipelineData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalCommandBufferData; +struct QMetalSwapChain; + +struct QMetalCommandBuffer : public QRhiCommandBuffer { + QMetalCommandBuffer(QRhiImplementation *rhi); + ~QMetalCommandBuffer(); + void destroy() override; + + QMetalCommandBufferData *d = nullptr; + QRhiMetalCommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + // per-pass (render or compute command encoder) persistent state + PassType recordingPass; + QRhiRenderTarget *currentTarget; + + // per-pass (render or compute command encoder) volatile (cached) state + QMetalGraphicsPipeline *currentGraphicsPipeline; + QMetalComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QMetalShaderResourceBindings *currentGraphicsSrb; + QMetalShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + int currentResSlot; + QMetalBuffer *currentIndexBuffer; + quint32 currentIndexOffset; + QRhiCommandBuffer::IndexFormat currentIndexFormat; + int currentCullMode; + int currentTriangleFillMode; + int currentFrontFaceWinding; + QPair<float, float> currentDepthBiasValues; + + const QRhiNativeHandles *nativeHandles(); + void resetState(double lastGpuTime = 0); + void resetPerPassState(); + void resetPerPassCachedState(); }; -struct Q_GUI_EXPORT QRhiMetalNativeHandles : public QRhiNativeHandles +struct QMetalSwapChainData; + +struct QMetalSwapChain : public QRhiSwapChain { - MTLDevice *dev = nullptr; - MTLCommandQueue *cmdQueue = nullptr; + QMetalSwapChain(QRhiImplementation *rhi); + ~QMetalSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + + bool createOrResize() override; + + virtual QRhiSwapChainHdrInfo hdrInfo() override; + + void chooseFormats(); + void waitUntilCompleted(int slot); + + QWindow *window = nullptr; + QSize pixelSize; + int currentFrameSlot = 0; // 0..QMTL_FRAMES_IN_FLIGHT-1 + int frameCount = 0; + int samples = 1; + QMetalSwapChainRenderTarget rtWrapper; + QMetalCommandBuffer cbWrapper; + QMetalRenderBuffer *ds = nullptr; + QMetalSwapChainData *d = nullptr; }; -struct Q_GUI_EXPORT QRhiMetalCommandBufferNativeHandles : public QRhiNativeHandles +struct QRhiMetalData; + +class QRhiMetal : public QRhiImplementation { - MTLCommandBuffer *commandBuffer = nullptr; - MTLRenderCommandEncoder *encoder = nullptr; +public: + QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice = nullptr); + ~QRhiMetal(); + + static bool probe(QRhiMetalInitParams *params); + static QRhiSwapChainProxyData updateSwapChainProxyData(QWindow *window); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void executeDeferredReleases(bool forced = false); + void finishActiveReadbacks(bool forced = false); + qsizetype subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; + void enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEncPtr, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc, + qsizetype *curOfs); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void executeBufferHostWritesForSlot(QMetalBuffer *bufD, int slot); + void executeBufferHostWritesForCurrentFrame(QMetalBuffer *bufD); + static const int SUPPORTED_STAGES = 5; + void enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD, + QMetalCommandBuffer *cbD, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets, + bool offsetOnlyChange, + const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[SUPPORTED_STAGES]); + struct TessDrawArgs { + QMetalCommandBuffer *cbD; + enum { + NonIndexed, + U16Indexed, + U32Indexed + } type; + struct NonIndexedArgs { + quint32 vertexCount; + quint32 instanceCount; + quint32 firstVertex; + quint32 firstInstance; + }; + struct IndexedArgs { + quint32 indexCount; + quint32 instanceCount; + quint32 firstIndex; + qint32 vertexOffset; + quint32 firstInstance; + void *indexBuffer; + }; + union { + NonIndexedArgs draw; + IndexedArgs drawIndexed; + }; + }; + void tessellatedDraw(const TessDrawArgs &args); + void adjustForMultiViewDraw(quint32 *instanceCount, QRhiCommandBuffer *cb); + + QRhi::Flags rhiFlags; + bool importedDevice = false; + bool importedCmdQueue = false; + QMetalSwapChain *currentSwapChain = nullptr; + QSet<QMetalSwapChain *> swapchains; + QRhiMetalNativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + quint32 osMajor = 0; + quint32 osMinor = 0; + + struct { + int maxTextureSize = 4096; + bool baseVertexAndInstance = true; + QVector<int> supportedSampleCounts; + bool isAppleGPU = false; + int maxThreadGroupSize = 512; + bool multiView = false; + } caps; + + QRhiMetalData *d = nullptr; }; QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhimetal_p_p.h b/src/gui/rhi/qrhimetal_p_p.h deleted file mode 100644 index 3a22b76b22..0000000000 --- a/src/gui/rhi/qrhimetal_p_p.h +++ /dev/null @@ -1,486 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIMETAL_P_H -#define QRHIMETAL_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhimetal_p.h" -#include "qrhi_p_p.h" -#include <QWindow> - -QT_BEGIN_NAMESPACE - -static const int QMTL_FRAMES_IN_FLIGHT = 2; - -// have to hide the ObjC stuff, this header cannot contain MTL* at all -struct QMetalBufferData; - -struct QMetalBuffer : public QRhiBuffer -{ - QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size); - ~QMetalBuffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - QMetalBufferData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; -}; - -struct QMetalRenderBufferData; - -struct QMetalRenderBuffer : public QRhiRenderBuffer -{ - QMetalRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QMetalRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - QMetalRenderBufferData *d; - int samples = 1; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalTextureData; - -struct QMetalTexture : public QRhiTexture -{ - QMetalTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags); - ~QMetalTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - - QMetalTextureData *d; - int mipLevelCount = 0; - int samples = 1; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; - friend struct QMetalTextureData; -}; - -struct QMetalSamplerData; - -struct QMetalSampler : public QRhiSampler -{ - QMetalSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QMetalSampler(); - void destroy() override; - bool create() override; - - QMetalSamplerData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; -}; - -struct QMetalRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QMetalRenderPassDescriptor(QRhiImplementation *rhi); - ~QMetalRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - - // there is no MTLRenderPassDescriptor here as one will be created for each pass in beginPass() - - // but the things needed for the render pipeline descriptor have to be provided - static const int MAX_COLOR_ATTACHMENTS = 8; - int colorAttachmentCount = 0; - bool hasDepthStencil = false; - int colorFormat[MAX_COLOR_ATTACHMENTS]; - int dsFormat; -}; - -struct QMetalRenderTargetData; - -struct QMetalReferenceRenderTarget : public QRhiRenderTarget -{ - QMetalReferenceRenderTarget(QRhiImplementation *rhi); - ~QMetalReferenceRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QMetalRenderTargetData *d; -}; - -struct QMetalTextureRenderTarget : public QRhiTextureRenderTarget -{ - QMetalTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QMetalTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QMetalRenderTargetData *d; - friend class QRhiMetal; -}; - -struct QMetalShaderResourceBindings : public QRhiShaderResourceBindings -{ - QMetalShaderResourceBindings(QRhiImplementation *rhi); - ~QMetalShaderResourceBindings(); - void destroy() override; - bool create() override; - - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - int maxBinding = -1; - - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData; - - uint generation = 0; - friend class QRhiMetal; -}; - -struct QMetalGraphicsPipelineData; - -struct QMetalGraphicsPipeline : public QRhiGraphicsPipeline -{ - QMetalGraphicsPipeline(QRhiImplementation *rhi); - ~QMetalGraphicsPipeline(); - void destroy() override; - bool create() override; - - QMetalGraphicsPipelineData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalComputePipelineData; - -struct QMetalComputePipeline : public QRhiComputePipeline -{ - QMetalComputePipeline(QRhiImplementation *rhi); - ~QMetalComputePipeline(); - void destroy() override; - bool create() override; - - QMetalComputePipelineData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalCommandBufferData; -struct QMetalSwapChain; - -struct QMetalCommandBuffer : public QRhiCommandBuffer -{ - QMetalCommandBuffer(QRhiImplementation *rhi); - ~QMetalCommandBuffer(); - void destroy() override; - - QMetalCommandBufferData *d = nullptr; - QRhiMetalCommandBufferNativeHandles nativeHandlesStruct; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - // per-pass (render or compute command encoder) persistent state - PassType recordingPass; - QRhiRenderTarget *currentTarget; - - // per-pass (render or compute command encoder) volatile (cached) state - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - int currentResSlot; - QRhiBuffer *currentIndexBuffer; - quint32 currentIndexOffset; - QRhiCommandBuffer::IndexFormat currentIndexFormat; - int currentCullMode; - int currentFrontFaceWinding; - QPair<float, float> currentDepthBiasValues; - - const QRhiNativeHandles *nativeHandles(); - void resetState(); - void resetPerPassState(); - void resetPerPassCachedState(); -}; - -struct QMetalSwapChainData; - -struct QMetalSwapChain : public QRhiSwapChain -{ - QMetalSwapChain(QRhiImplementation *rhi); - ~QMetalSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - QSize surfacePixelSize() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - - bool createOrResize() override; - - void chooseFormats(); - - QWindow *window = nullptr; - QSize pixelSize; - int currentFrameSlot = 0; // 0..QMTL_FRAMES_IN_FLIGHT-1 - int frameCount = 0; - int samples = 1; - QMetalReferenceRenderTarget rtWrapper; - QMetalCommandBuffer cbWrapper; - QMetalRenderBuffer *ds = nullptr; - QMetalSwapChainData *d = nullptr; -}; - -struct QRhiMetalData; - -class QRhiMetal : public QRhiImplementation -{ -public: - QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice = nullptr); - ~QRhiMetal(); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - void sendVMemStatsToProfiler() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void executeDeferredReleases(bool forced = false); - void finishActiveReadbacks(bool forced = false); - qsizetype subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; - void enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEncPtr, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc, - qsizetype *curOfs); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void executeBufferHostWritesForSlot(QMetalBuffer *bufD, int slot); - void executeBufferHostWritesForCurrentFrame(QMetalBuffer *bufD); - static const int SUPPORTED_STAGES = 3; - void enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD, - QMetalCommandBuffer *cbD, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets, - bool offsetOnlyChange, - const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[SUPPORTED_STAGES]); - int effectiveSampleCount(int sampleCount) const; - - bool importedDevice = false; - bool importedCmdQueue = false; - QMetalSwapChain *currentSwapChain = nullptr; - QSet<QMetalSwapChain *> swapchains; - QRhiMetalNativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - - struct { - int maxTextureSize = 4096; - } caps; - - QRhiMetalData *d = nullptr; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhinull.cpp b/src/gui/rhi/qrhinull.cpp index d8dd77a790..566b922c1b 100644 --- a/src/gui/rhi/qrhinull.cpp +++ b/src/gui/rhi/qrhinull.cpp @@ -1,43 +1,7 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhinull_p_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhinull_p.h" #include <qmath.h> #include <QPainter> @@ -45,10 +9,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiNullInitParams - \internal \inmodule QtGui + \since 6.6 \brief Null backend specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Null QRhi needs no special parameters for initialization. \badcode @@ -59,15 +26,17 @@ QT_BEGIN_NAMESPACE The Null backend does not issue any graphics calls and creates no resources. All QRhi operations will succeed as normal so applications can still be run, albeit potentially at an unthrottled speed, depending on - their frame rendering strategy. The backend reports resources to - QRhiProfiler as usual. + their frame rendering strategy. */ /*! \class QRhiNullNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Empty. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ QRhiNull::QRhiNull(QRhiNullInitParams *params) @@ -96,7 +65,7 @@ QRhiSwapChain *QRhiNull::createSwapChain() return new QNullSwapChain(this); } -QRhiBuffer *QRhiNull::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, int size) +QRhiBuffer *QRhiNull::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) { return new QNullBuffer(this, type, usage, size); } @@ -162,10 +131,17 @@ int QRhiNull::resourceLimit(QRhi::ResourceLimit limit) const return 0; case QRhi::MaxThreadGroupZ: return 0; - default: - Q_UNREACHABLE(); - return 0; + case QRhi::TextureArraySizeMax: + return 2048; + case QRhi::MaxUniformBufferRange: + return 65536; + case QRhi::MaxVertexInputs: + return 32; + case QRhi::MaxVertexOutputs: + return 32; } + + Q_UNREACHABLE_RETURN(0); } const QRhiNativeHandles *QRhiNull::nativeHandles() @@ -180,9 +156,9 @@ QRhiDriverInfo QRhiNull::driverInfo() const return info; } -void QRhiNull::sendVMemStatsToProfiler() +QRhiStats QRhiNull::statistics() { - // nothing to do here + return {}; } bool QRhiNull::makeThreadLocalNativeContextCurrent() @@ -219,10 +195,10 @@ QRhiRenderBuffer *QRhiNull::createRenderBuffer(QRhiRenderBuffer::Type type, cons } QRhiTexture *QRhiNull::createTexture(QRhiTexture::Format format, - const QSize &pixelSize, int depth, + const QSize &pixelSize, int depth, int arraySize, int sampleCount, QRhiTexture::Flags flags) { - return new QNullTexture(this, format, pixelSize, depth, sampleCount, flags); + return new QNullTexture(this, format, pixelSize, depth, arraySize, sampleCount, flags); } QRhiSampler *QRhiNull::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, @@ -374,12 +350,16 @@ void QRhiNull::endExternal(QRhiCommandBuffer *cb) Q_UNUSED(cb); } +double QRhiNull::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + Q_UNUSED(cb); + return 0; +} + QRhi::FrameOpResult QRhiNull::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); currentSwapChain = swapChain; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - QRHI_PROF_F(beginSwapChainFrame(swapChain)); return QRhi::FrameOpSuccess; } @@ -387,9 +367,6 @@ QRhi::FrameOpResult QRhiNull::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameF { Q_UNUSED(flags); QNullSwapChain *swapChainD = QRHI_RES(QNullSwapChain, swapChain); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - QRHI_PROF_F(endSwapChainFrame(swapChain, swapChainD->frameCount + 1)); - QRHI_PROF_F(swapChainFrameGpuTime(swapChain, 0.000666f)); swapChainD->frameCount += 1; currentSwapChain = nullptr; return QRhi::FrameOpSuccess; @@ -416,9 +393,9 @@ QRhi::FrameOpResult QRhiNull::finish() void QRhiNull::simulateTextureUpload(const QRhiResourceUpdateBatchPrivate::TextureOp &u) { QNullTexture *texD = QRHI_RES(QNullTexture, u.dst); - for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { + for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) { + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) { if (!subresDesc.image().isNull()) { const QImage src = subresDesc.image(); QPainter painter(&texD->image[layer][level]); @@ -491,7 +468,7 @@ void QRhiNull::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *re QNullBuffer *bufD = QRHI_RES(QNullBuffer, u.buf); memcpy(bufD->data + u.offset, u.data.constData(), size_t(u.data.size())); } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::Read) { - QRhiBufferReadbackResult *result = u.result; + QRhiReadbackResult *result = u.result; result->data.resize(u.readSize); QNullBuffer *bufD = QRHI_RES(QNullBuffer, u.buf); memcpy(result->data.data(), bufD->data + u.offset, size_t(u.readSize)); @@ -549,12 +526,18 @@ void QRhiNull::beginPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates, QRhiCommandBuffer::BeginPassFlags flags) { - Q_UNUSED(rt); Q_UNUSED(colorClearValue); Q_UNUSED(depthStencilClearValue); Q_UNUSED(flags); + if (resourceUpdates) resourceUpdate(cb, resourceUpdates); + + if (rt->resourceType() == QRhiRenderTarget::TextureRenderTarget) { + QNullTextureRenderTarget *rtTex = QRHI_RES(QNullTextureRenderTarget, rt); + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QNullTexture, QNullRenderBuffer>(rtTex->description(), rtTex->d.currentResIdList)) + rtTex->create(); + } } void QRhiNull::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -578,7 +561,7 @@ void QRhiNull::endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *re resourceUpdate(cb, resourceUpdates); } -QNullBuffer::QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size) +QNullBuffer::QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) : QRhiBuffer(rhi, type, usage, size) { } @@ -593,17 +576,22 @@ void QNullBuffer::destroy() delete[] data; data = nullptr; - QRHI_PROF; - QRHI_PROF_F(releaseBuffer(this)); + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullBuffer::create() { + if (data) + destroy(); + data = new char[m_size]; memset(data, 0, m_size); - QRHI_PROF; - QRHI_PROF_F(newBuffer(this, uint(m_size), 1, 0)); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); + return true; } @@ -627,14 +615,24 @@ QNullRenderBuffer::~QNullRenderBuffer() void QNullRenderBuffer::destroy() { - QRHI_PROF; - QRHI_PROF_F(releaseRenderBuffer(this)); + valid = false; + + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullRenderBuffer::create() { - QRHI_PROF; - QRHI_PROF_F(newRenderBuffer(this, false, false, 1)); + if (valid) + destroy(); + + valid = true; + generation += 1; + + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); + return true; } @@ -644,8 +642,8 @@ QRhiTexture::Format QNullRenderBuffer::backingFormat() const } QNullTexture::QNullTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags) - : QRhiTexture(rhi, format, pixelSize, depth, sampleCount, flags) + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags) { } @@ -656,20 +654,33 @@ QNullTexture::~QNullTexture() void QNullTexture::destroy() { - QRHI_PROF; - QRHI_PROF_F(releaseTexture(this)); + valid = false; + + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullTexture::create() { + if (valid) + destroy(); + + valid = true; + QRHI_RES_RHI(QRhiNull); const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool isArray = m_flags.testFlag(TextureArray); const bool hasMipMaps = m_flags.testFlag(MipMapped); - QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; - m_depth = qMax(1, m_depth); + const bool is1D = m_flags.testFlags(OneDimensional); + QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); const int mipLevelCount = hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1; - const int layerCount = is3D ? m_depth : (isCube ? 6 : 1); + const int layerCount = is3D ? qMax(1, m_depth) + : (isCube ? 6 + : (isArray ? qMax(0, m_arraySize) + : 1)); if (m_format == RGBA8) { image.resize(layerCount); @@ -682,21 +693,26 @@ bool QNullTexture::create() } } - QRHI_PROF; - QRHI_PROF_F(newTexture(this, true, mipLevelCount, layerCount, 1)); + generation += 1; + + rhiD->registerResource(this); + return true; } bool QNullTexture::createFrom(QRhiTexture::NativeTexture src) { Q_UNUSED(src); + if (valid) + destroy(); + + valid = true; + + generation += 1; + QRHI_RES_RHI(QRhiNull); - const bool isCube = m_flags.testFlag(CubeMap); - const bool hasMipMaps = m_flags.testFlag(MipMapped); - QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; - const int mipLevelCount = hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, false, mipLevelCount, isCube ? 6 : 1, 1)); + rhiD->registerResource(this); + return true; } @@ -713,10 +729,15 @@ QNullSampler::~QNullSampler() void QNullSampler::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullSampler::create() { + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); return true; } @@ -732,6 +753,9 @@ QNullRenderPassDescriptor::~QNullRenderPassDescriptor() void QNullRenderPassDescriptor::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -742,35 +766,43 @@ bool QNullRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *oth QRhiRenderPassDescriptor *QNullRenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; +} + +QVector<quint32> QNullRenderPassDescriptor::serializedFormat() const +{ + return {}; } -QNullReferenceRenderTarget::QNullReferenceRenderTarget(QRhiImplementation *rhi) - : QRhiRenderTarget(rhi), +QNullSwapChainRenderTarget::QNullSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain), d(rhi) { } -QNullReferenceRenderTarget::~QNullReferenceRenderTarget() +QNullSwapChainRenderTarget::~QNullSwapChainRenderTarget() { destroy(); } -void QNullReferenceRenderTarget::destroy() +void QNullSwapChainRenderTarget::destroy() { } -QSize QNullReferenceRenderTarget::pixelSize() const +QSize QNullSwapChainRenderTarget::pixelSize() const { return d.pixelSize; } -float QNullReferenceRenderTarget::devicePixelRatio() const +float QNullSwapChainRenderTarget::devicePixelRatio() const { return d.dpr; } -int QNullReferenceRenderTarget::sampleCount() const +int QNullSwapChainRenderTarget::sampleCount() const { return 1; } @@ -790,11 +822,17 @@ QNullTextureRenderTarget::~QNullTextureRenderTarget() void QNullTextureRenderTarget::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } QRhiRenderPassDescriptor *QNullTextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; } bool QNullTextureRenderTarget::create() @@ -811,11 +849,16 @@ bool QNullTextureRenderTarget::create() } else if (m_desc.depthTexture()) { d.pixelSize = m_desc.depthTexture()->pixelSize(); } + QRhiRenderTargetAttachmentTracker::updateResIdList<QNullTexture, QNullRenderBuffer>(m_desc, &d.currentResIdList); + rhiD->registerResource(this); return true; } QSize QNullTextureRenderTarget::pixelSize() const { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QNullTexture, QNullRenderBuffer>(m_desc, d.currentResIdList)) + const_cast<QNullTextureRenderTarget *>(this)->create(); + return d.pixelSize; } @@ -841,6 +884,9 @@ QNullShaderResourceBindings::~QNullShaderResourceBindings() void QNullShaderResourceBindings::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullShaderResourceBindings::create() @@ -851,9 +897,15 @@ bool QNullShaderResourceBindings::create() rhiD->updateLayoutDesc(this); + rhiD->registerResource(this, false); return true; } +void QNullShaderResourceBindings::updateResources(UpdateFlags flags) +{ + Q_UNUSED(flags); +} + QNullGraphicsPipeline::QNullGraphicsPipeline(QRhiImplementation *rhi) : QRhiGraphicsPipeline(rhi) { @@ -866,6 +918,9 @@ QNullGraphicsPipeline::~QNullGraphicsPipeline() void QNullGraphicsPipeline::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullGraphicsPipeline::create() @@ -874,6 +929,7 @@ bool QNullGraphicsPipeline::create() if (!rhiD->sanityCheckGraphicsPipeline(this)) return false; + rhiD->registerResource(this); return true; } @@ -889,10 +945,15 @@ QNullComputePipeline::~QNullComputePipeline() void QNullComputePipeline::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullComputePipeline::create() { + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); return true; } @@ -913,7 +974,7 @@ void QNullCommandBuffer::destroy() QNullSwapChain::QNullSwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), - rt(rhi), + rt(rhi, this), cb(rhi) { } @@ -925,8 +986,9 @@ QNullSwapChain::~QNullSwapChain() void QNullSwapChain::destroy() { - QRHI_PROF; - QRHI_PROF_F(releaseSwapChain(this)); + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } QRhiCommandBuffer *QNullSwapChain::currentFrameCommandBuffer() @@ -944,19 +1006,37 @@ QSize QNullSwapChain::surfacePixelSize() return QSize(1280, 720); } +bool QNullSwapChain::isFormatSupported(Format f) +{ + return f == SDR; +} + QRhiRenderPassDescriptor *QNullSwapChain::newCompatibleRenderPassDescriptor() { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; } bool QNullSwapChain::createOrResize() { + const bool needsRegistration = !window || window != m_window; + if (window && window != m_window) + destroy(); + + window = m_window; m_currentPixelSize = surfacePixelSize(); + rt.setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget rt.d.rp = QRHI_RES(QNullRenderPassDescriptor, m_renderPassDesc); rt.d.pixelSize = m_currentPixelSize; frameCount = 0; - QRHI_PROF; - QRHI_PROF_F(resizeSwapChain(this, 1, 0, 1)); + + if (needsRegistration) { + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); + } + return true; } diff --git a/src/gui/rhi/qrhinull_p.h b/src/gui/rhi/qrhinull_p.h index 8b45d3b254..fc266b4f38 100644 --- a/src/gui/rhi/qrhinull_p.h +++ b/src/gui/rhi/qrhinull_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHINULL_H -#define QRHINULL_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHINULL_P_H +#define QRHINULL_P_H // // W A R N I N G @@ -51,16 +15,279 @@ // We mean it. // -#include <private/qrhi_p.h> +#include "qrhi_p.h" QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiNullInitParams : public QRhiInitParams +struct QNullBuffer : public QRhiBuffer +{ + QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QNullBuffer(); + void destroy() override; + bool create() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + + char *data = nullptr; +}; + +struct QNullRenderBuffer : public QRhiRenderBuffer +{ + QNullRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QNullRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + bool valid = false; + uint generation = 0; +}; + +struct QNullTexture : public QRhiTexture { + QNullTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QNullTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + + bool valid = false; + QVarLengthArray<std::array<QImage, QRhi::MAX_MIP_LEVELS>, 6> image; + uint generation = 0; }; -struct Q_GUI_EXPORT QRhiNullNativeHandles : public QRhiNativeHandles +struct QNullSampler : public QRhiSampler { + QNullSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QNullSampler(); + void destroy() override; + bool create() override; +}; + +struct QNullRenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QNullRenderPassDescriptor(QRhiImplementation *rhi); + ~QNullRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QNullRenderTargetData +{ + QNullRenderTargetData(QRhiImplementation *) { } + + QNullRenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; +}; + +struct QNullSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QNullSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QNullSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QNullRenderTargetData d; +}; + +struct QNullTextureRenderTarget : public QRhiTextureRenderTarget +{ + QNullTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QNullTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QNullRenderTargetData d; +}; + +struct QNullShaderResourceBindings : public QRhiShaderResourceBindings +{ + QNullShaderResourceBindings(QRhiImplementation *rhi); + ~QNullShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; +}; + +struct QNullGraphicsPipeline : public QRhiGraphicsPipeline +{ + QNullGraphicsPipeline(QRhiImplementation *rhi); + ~QNullGraphicsPipeline(); + void destroy() override; + bool create() override; +}; + +struct QNullComputePipeline : public QRhiComputePipeline +{ + QNullComputePipeline(QRhiImplementation *rhi); + ~QNullComputePipeline(); + void destroy() override; + bool create() override; +}; + +struct QNullCommandBuffer : public QRhiCommandBuffer +{ + QNullCommandBuffer(QRhiImplementation *rhi); + ~QNullCommandBuffer(); + void destroy() override; +}; + +struct QNullSwapChain : public QRhiSwapChain +{ + QNullSwapChain(QRhiImplementation *rhi); + ~QNullSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + QWindow *window = nullptr; + QNullSwapChainRenderTarget rt; + QNullCommandBuffer cb; + int frameCount = 0; +}; + +class QRhiNull : public QRhiImplementation +{ +public: + QRhiNull(QRhiNullInitParams *params); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void simulateTextureUpload(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + void simulateTextureCopy(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + void simulateTextureGenMips(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + + QRhiNullNativeHandles nativeHandlesStruct; + QRhiSwapChain *currentSwapChain = nullptr; + QNullCommandBuffer offscreenCommandBuffer; }; QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhinull_p_p.h b/src/gui/rhi/qrhinull_p_p.h deleted file mode 100644 index b064e7ad37..0000000000 --- a/src/gui/rhi/qrhinull_p_p.h +++ /dev/null @@ -1,320 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHINULL_P_H -#define QRHINULL_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhinull_p.h" -#include "qrhi_p_p.h" - -QT_BEGIN_NAMESPACE - -struct QNullBuffer : public QRhiBuffer -{ - QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size); - ~QNullBuffer(); - void destroy() override; - bool create() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - - char *data = nullptr; -}; - -struct QNullRenderBuffer : public QRhiRenderBuffer -{ - QNullRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QNullRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; -}; - -struct QNullTexture : public QRhiTexture -{ - QNullTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags); - ~QNullTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - - QVarLengthArray<std::array<QImage, QRhi::MAX_MIP_LEVELS>, 6> image; -}; - -struct QNullSampler : public QRhiSampler -{ - QNullSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QNullSampler(); - void destroy() override; - bool create() override; -}; - -struct QNullRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QNullRenderPassDescriptor(QRhiImplementation *rhi); - ~QNullRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; -}; - -struct QNullRenderTargetData -{ - QNullRenderTargetData(QRhiImplementation *) { } - - QNullRenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; -}; - -struct QNullReferenceRenderTarget : public QRhiRenderTarget -{ - QNullReferenceRenderTarget(QRhiImplementation *rhi); - ~QNullReferenceRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QNullRenderTargetData d; -}; - -struct QNullTextureRenderTarget : public QRhiTextureRenderTarget -{ - QNullTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QNullTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QNullRenderTargetData d; -}; - -struct QNullShaderResourceBindings : public QRhiShaderResourceBindings -{ - QNullShaderResourceBindings(QRhiImplementation *rhi); - ~QNullShaderResourceBindings(); - void destroy() override; - bool create() override; -}; - -struct QNullGraphicsPipeline : public QRhiGraphicsPipeline -{ - QNullGraphicsPipeline(QRhiImplementation *rhi); - ~QNullGraphicsPipeline(); - void destroy() override; - bool create() override; -}; - -struct QNullComputePipeline : public QRhiComputePipeline -{ - QNullComputePipeline(QRhiImplementation *rhi); - ~QNullComputePipeline(); - void destroy() override; - bool create() override; -}; - -struct QNullCommandBuffer : public QRhiCommandBuffer -{ - QNullCommandBuffer(QRhiImplementation *rhi); - ~QNullCommandBuffer(); - void destroy() override; -}; - -struct QNullSwapChain : public QRhiSwapChain -{ - QNullSwapChain(QRhiImplementation *rhi); - ~QNullSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - QNullReferenceRenderTarget rt; - QNullCommandBuffer cb; - int frameCount = 0; -}; - -class QRhiNull : public QRhiImplementation -{ -public: - QRhiNull(QRhiNullInitParams *params); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - void sendVMemStatsToProfiler() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void simulateTextureUpload(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - void simulateTextureCopy(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - void simulateTextureGenMips(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - - QRhiNullNativeHandles nativeHandlesStruct; - QRhiSwapChain *currentSwapChain = nullptr; - QNullCommandBuffer offscreenCommandBuffer; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhiprofiler.cpp b/src/gui/rhi/qrhiprofiler.cpp deleted file mode 100644 index 76798dc396..0000000000 --- a/src/gui/rhi/qrhiprofiler.cpp +++ /dev/null @@ -1,611 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhiprofiler_p_p.h" -#include "qrhi_p_p.h" -#include <QtCore/qiodevice.h> - -QT_BEGIN_NAMESPACE - -/*! - \class QRhiProfiler - \internal - \inmodule QtGui - - \brief Collects resource and timing information from an active QRhi. - - A QRhiProfiler is present for each QRhi. Query it via QRhi::profiler(). The - profiler is active only when the QRhi was created with - QRhi::EnableProfiling. No data is collected otherwise. - - \note GPU timings are only available when QRhi indicates that - QRhi::Timestamps is supported. - - Besides collecting data from the QRhi implementations, some additional - values are calculated. For example, for textures and similar resources the - profiler gives an estimate of the complete amount of memory the resource - needs. - - \section2 Output Format - - The output is comma-separated text. Each line has a number of - comma-separated entries and each line ends with a comma. - - For example: - - \badcode - 1,0,140446057946208,Triangle vbuf,type,0,usage,1,logical_size,84,effective_size,84,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,0,140446057947376,Triangle ubuf,type,2,usage,4,logical_size,68,effective_size,256,backing_gpu_buf_count,2,backing_cpu_buf_count,0, - 1,1,140446057950416,,type,0,usage,1,logical_size,112,effective_size,112,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,1,140446057950544,,type,0,usage,2,logical_size,12,effective_size,12,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,1,140446057947440,,type,2,usage,4,logical_size,68,effective_size,256,backing_gpu_buf_count,2,backing_cpu_buf_count,0, - 1,1,140446057984784,Cube vbuf (textured),type,0,usage,1,logical_size,720,effective_size,720,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,1,140446057982528,Cube ubuf (textured),type,2,usage,4,logical_size,68,effective_size,256,backing_gpu_buf_count,2,backing_cpu_buf_count,0, - 7,8,140446058913648,Qt texture,width,256,height,256,format,1,owns_native_resource,1,mip_count,9,layer_count,1,effective_sample_count,1,approx_byte_size,349524, - 1,8,140446058795856,Cube vbuf (textured with offscreen),type,0,usage,1,logical_size,720,effective_size,720,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,8,140446058947920,Cube ubuf (textured with offscreen),type,2,usage,4,logical_size,68,effective_size,256,backing_gpu_buf_count,2,backing_cpu_buf_count,0, - 7,8,140446058794928,Texture for offscreen content,width,512,height,512,format,1,owns_native_resource,1,mip_count,1,layer_count,1,effective_sample_count,1,approx_byte_size,1048576, - 1,8,140446058963904,Triangle vbuf,type,0,usage,1,logical_size,84,effective_size,84,backing_gpu_buf_count,1,backing_cpu_buf_count,0, - 1,8,140446058964560,Triangle ubuf,type,2,usage,4,logical_size,68,effective_size,256,backing_gpu_buf_count,2,backing_cpu_buf_count,0, - 5,9,140446057945392,,type,0,width,1280,height,720,effective_sample_count,1,transient_backing,0,winsys_backing,0,approx_byte_size,3686400, - 11,9,140446057944592,,width,1280,height,720,buffer_count,2,msaa_buffer_count,0,effective_sample_count,1,approx_total_byte_size,7372800, - 9,9,140446058913648,Qt texture,slot,0,size,262144, - 10,9,140446058913648,Qt texture,slot,0, - 17,2019,140446057944592,,frames_since_resize,121,min_ms_frame_delta,9,max_ms_frame_delta,33,Favg_ms_frame_delta,16.1167, - 18,2019,140446057944592,,frames_since_resize,121,min_ms_frame_build,0,max_ms_frame_build,1,Favg_ms_frame_build,0.00833333, - 17,4019,140446057944592,,frames_since_resize,241,min_ms_frame_delta,15,max_ms_frame_delta,17,Favg_ms_frame_delta,16.0583, - 18,4019,140446057944592,,frames_since_resize,241,min_ms_frame_build,0,max_ms_frame_build,0,Favg_ms_frame_build,0, - 12,5070,140446057944592,, - 2,5079,140446057947376,Triangle ubuf, - 2,5079,140446057946208,Triangle vbuf, - 2,5079,140446057947440,, - 2,5079,140446057950544,, - 2,5079,140446057950416,, - 8,5079,140446058913648,Qt texture, - 2,5079,140446057982528,Cube ubuf (textured), - 2,5079,140446057984784,Cube vbuf (textured), - 2,5079,140446058964560,Triangle ubuf, - 2,5079,140446058963904,Triangle vbuf, - 8,5079,140446058794928,Texture for offscreen content, - 2,5079,140446058947920,Cube ubuf (textured with offscreen), - 2,5079,140446058795856,Cube vbuf (textured with offscreen), - 6,5079,140446057945392,, - \endcode - - Each line starts with \c op, \c timestamp, \c res, \c name where op is a - value from StreamOp, timestamp is a recording timestamp in milliseconds - (qint64), res is a number (quint64) referring to the QRhiResource the entry - refers to, or 0 if not applicable. \c name is the value of - QRhiResource::name() and may be empty as well. The \c name will never - contain a comma. - - This is followed by any number of \c{key, value} pairs where \c key is an - unspecified string and \c value is a number. If \c key starts with \c F, it - indicates the value is a float. Otherwise assume that the value is a - qint64. - */ - -/*! - \enum QRhiProfiler::StreamOp - Describes an entry in the profiler's output stream. - - \value NewBuffer A buffer is created - \value ReleaseBuffer A buffer is destroyed - \value NewBufferStagingArea A staging buffer for buffer upload is created - \value ReleaseBufferStagingArea A staging buffer for buffer upload is destroyed - \value NewRenderBuffer A renderbuffer is created - \value ReleaseRenderBuffer A renderbuffer is destroyed - \value NewTexture A texture is created - \value ReleaseTexture A texture is destroyed - \value NewTextureStagingArea A staging buffer for texture upload is created - \value ReleaseTextureStagingArea A staging buffer for texture upload is destroyed - \value ResizeSwapChain A swapchain is created or resized - \value ReleaseSwapChain A swapchain is destroyed - \value NewReadbackBuffer A staging buffer for readback is created - \value ReleaseReadbackBuffer A staging buffer for readback is destroyed - \value GpuMemAllocStats GPU memory allocator statistics - \value GpuFrameTime GPU frame times - \value FrameToFrameTime CPU frame-to-frame times - \value FrameBuildTime CPU beginFrame-endFrame times - */ - -/*! - \class QRhiProfiler::CpuTime - \internal - \inmodule QtGui - \brief Contains CPU-side frame timings. - - Once sufficient number of frames have been rendered, the minimum, maximum, - and average values (in milliseconds) from various measurements are made - available in this struct queriable from QRhiProfiler::frameToFrameTimes() - and QRhiProfiler::frameBuildTimes(). - - \sa QRhiProfiler::setFrameTimingWriteInterval() - */ - -/*! - \class QRhiProfiler::GpuTime - \internal - \inmodule QtGui - \brief Contains GPU-side frame timings. - - Once sufficient number of frames have been rendered, the minimum, maximum, - and average values (in milliseconds) calculated from GPU command buffer - timestamps are made available in this struct queriable from - QRhiProfiler::gpuFrameTimes(). - - \sa QRhiProfiler::setFrameTimingWriteInterval() - */ - -/*! - \internal - */ -QRhiProfiler::QRhiProfiler() - : d(new QRhiProfilerPrivate) -{ - d->ts.start(); -} - -/*! - Destructor. - */ -QRhiProfiler::~QRhiProfiler() -{ - // Flush because there is a high chance we have writes that were made since - // the event loop last ran. (esp. relevant for network devices like QTcpSocket) - if (d->outputDevice) - d->outputDevice->waitForBytesWritten(1000); - - delete d; -} - -/*! - Sets the output \a device. - - \note No output will be generated when QRhi::EnableProfiling was not set. - */ -void QRhiProfiler::setDevice(QIODevice *device) -{ - d->outputDevice = device; -} - -/*! - Requests writing a GpuMemAllocStats entry into the output, when applicable. - Backends that do not support this will ignore the request. This is an - explicit request since getting the allocator status and statistics may be - an expensive operation. - */ -void QRhiProfiler::addVMemAllocatorStats() -{ - if (d->rhiDWhenEnabled) - d->rhiDWhenEnabled->sendVMemStatsToProfiler(); -} - -/*! - \return the currently set frame timing writeout interval. - */ -int QRhiProfiler::frameTimingWriteInterval() const -{ - return d->frameTimingWriteInterval; -} - -/*! - Sets the number of frames that need to be rendered before the collected CPU - and GPU timings are processed (min, max, average are calculated) to \a - frameCount. - - The default value is 120. - */ -void QRhiProfiler::setFrameTimingWriteInterval(int frameCount) -{ - if (frameCount > 0) - d->frameTimingWriteInterval = frameCount; -} - -/*! - \return min, max, and avg in milliseconds for the time that elapsed between two - QRhi::endFrame() calls. - - \note The values are all 0 until at least frameTimingWriteInterval() frames - have been rendered. - */ -QRhiProfiler::CpuTime QRhiProfiler::frameToFrameTimes(QRhiSwapChain *sc) const -{ - auto it = d->swapchains.constFind(sc); - if (it != d->swapchains.constEnd()) - return it->frameToFrameTime; - - return QRhiProfiler::CpuTime(); -} - -/*! - \return min, max, and avg in milliseconds for the time that elapsed between - a QRhi::beginFrame() and QRhi::endFrame(). - - \note The values are all 0 until at least frameTimingWriteInterval() frames - have been rendered. - */ -QRhiProfiler::CpuTime QRhiProfiler::frameBuildTimes(QRhiSwapChain *sc) const -{ - auto it = d->swapchains.constFind(sc); - if (it != d->swapchains.constEnd()) - return it->beginToEndFrameTime; - - return QRhiProfiler::CpuTime(); -} - -/*! - \return min, max, and avg in milliseconds for the GPU time that is spent on - one frame. - - \note The values are all 0 until at least frameTimingWriteInterval() frames - have been rendered. - - The GPU times should only be compared between runs on the same GPU of the - same system with the same backend. Comparing times for different graphics - cards or for different backends can give misleading results. The numbers are - not meant to be comparable that way. - - \note Some backends have no support for this, and even for those that have, - it is not guaranteed that the driver will support it at run time. Support - can be checked via QRhi::Timestamps. - */ -QRhiProfiler::GpuTime QRhiProfiler::gpuFrameTimes(QRhiSwapChain *sc) const -{ - auto it = d->swapchains.constFind(sc); - if (it != d->swapchains.constEnd()) - return it->gpuFrameTime; - - return QRhiProfiler::GpuTime(); -} - -void QRhiProfilerPrivate::startEntry(QRhiProfiler::StreamOp op, qint64 timestamp, QRhiResource *res) -{ - buf.clear(); - buf.append(QByteArray::number(op)); - buf.append(','); - buf.append(QByteArray::number(timestamp)); - buf.append(','); - buf.append(QByteArray::number(quint64(quintptr(res)))); - buf.append(','); - if (res) - buf.append(res->name()); - buf.append(','); -} - -void QRhiProfilerPrivate::writeInt(const char *key, qint64 v) -{ - Q_ASSERT(key[0] != 'F'); - buf.append(key); - buf.append(','); - buf.append(QByteArray::number(v)); - buf.append(','); -} - -void QRhiProfilerPrivate::writeFloat(const char *key, float f) -{ - Q_ASSERT(key[0] == 'F'); - buf.append(key); - buf.append(','); - buf.append(QByteArray::number(double(f))); - buf.append(','); -} - -void QRhiProfilerPrivate::endEntry() -{ - buf.append('\n'); - outputDevice->write(buf); -} - -void QRhiProfilerPrivate::newBuffer(QRhiBuffer *buf, quint32 realSize, int backingGpuBufCount, int backingCpuBufCount) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::NewBuffer, ts.elapsed(), buf); - writeInt("type", buf->type()); - writeInt("usage", buf->usage()); - writeInt("logical_size", buf->size()); - writeInt("effective_size", realSize); - writeInt("backing_gpu_buf_count", backingGpuBufCount); - writeInt("backing_cpu_buf_count", backingCpuBufCount); - endEntry(); -} - -void QRhiProfilerPrivate::releaseBuffer(QRhiBuffer *buf) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseBuffer, ts.elapsed(), buf); - endEntry(); -} - -void QRhiProfilerPrivate::newBufferStagingArea(QRhiBuffer *buf, int slot, quint32 size) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::NewBufferStagingArea, ts.elapsed(), buf); - writeInt("slot", slot); - writeInt("size", size); - endEntry(); -} - -void QRhiProfilerPrivate::releaseBufferStagingArea(QRhiBuffer *buf, int slot) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseBufferStagingArea, ts.elapsed(), buf); - writeInt("slot", slot); - endEntry(); -} - -void QRhiProfilerPrivate::newRenderBuffer(QRhiRenderBuffer *rb, bool transientBacking, bool winSysBacking, int sampleCount) -{ - if (!outputDevice) - return; - - const QRhiRenderBuffer::Type type = rb->type(); - const QSize sz = rb->pixelSize(); - // just make up something, ds is likely D24S8 while color is RGBA8 or similar - const QRhiTexture::Format assumedFormat = type == QRhiRenderBuffer::DepthStencil ? QRhiTexture::D32F : QRhiTexture::RGBA8; - quint32 byteSize = rhiDWhenEnabled->approxByteSizeForTexture(assumedFormat, sz, 1, 1, 1); - if (sampleCount > 1) - byteSize *= uint(sampleCount); - - startEntry(QRhiProfiler::NewRenderBuffer, ts.elapsed(), rb); - writeInt("type", type); - writeInt("width", sz.width()); - writeInt("height", sz.height()); - writeInt("effective_sample_count", sampleCount); - writeInt("transient_backing", transientBacking); - writeInt("winsys_backing", winSysBacking); - writeInt("approx_byte_size", byteSize); - endEntry(); -} - -void QRhiProfilerPrivate::releaseRenderBuffer(QRhiRenderBuffer *rb) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseRenderBuffer, ts.elapsed(), rb); - endEntry(); -} - -void QRhiProfilerPrivate::newTexture(QRhiTexture *tex, bool owns, int mipCount, int layerCount, int sampleCount) -{ - if (!outputDevice) - return; - - const QRhiTexture::Format format = tex->format(); - const QSize sz = tex->pixelSize(); - const int depth = tex->depth(); - quint32 byteSize = rhiDWhenEnabled->approxByteSizeForTexture(format, sz, depth, mipCount, layerCount); - if (sampleCount > 1) - byteSize *= uint(sampleCount); - - startEntry(QRhiProfiler::NewTexture, ts.elapsed(), tex); - writeInt("width", sz.width()); - writeInt("height", sz.height()); - writeInt("format", format); - writeInt("owns_native_resource", owns); - writeInt("mip_count", mipCount); - writeInt("layer_count", layerCount); - writeInt("effective_sample_count", sampleCount); - writeInt("approx_byte_size", byteSize); - endEntry(); -} - -void QRhiProfilerPrivate::releaseTexture(QRhiTexture *tex) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseTexture, ts.elapsed(), tex); - endEntry(); -} - -void QRhiProfilerPrivate::newTextureStagingArea(QRhiTexture *tex, int slot, quint32 size) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::NewTextureStagingArea, ts.elapsed(), tex); - writeInt("slot", slot); - writeInt("size", size); - endEntry(); -} - -void QRhiProfilerPrivate::releaseTextureStagingArea(QRhiTexture *tex, int slot) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseTextureStagingArea, ts.elapsed(), tex); - writeInt("slot", slot); - endEntry(); -} - -void QRhiProfilerPrivate::resizeSwapChain(QRhiSwapChain *sc, int bufferCount, int msaaBufferCount, int sampleCount) -{ - if (!outputDevice) - return; - - const QSize sz = sc->currentPixelSize(); - quint32 byteSize = rhiDWhenEnabled->approxByteSizeForTexture(QRhiTexture::BGRA8, sz, 1, 1, 1); - byteSize = byteSize * uint(bufferCount) + byteSize * uint(msaaBufferCount) * uint(sampleCount); - - startEntry(QRhiProfiler::ResizeSwapChain, ts.elapsed(), sc); - writeInt("width", sz.width()); - writeInt("height", sz.height()); - writeInt("buffer_count", bufferCount); - writeInt("msaa_buffer_count", msaaBufferCount); - writeInt("effective_sample_count", sampleCount); - writeInt("approx_total_byte_size", byteSize); - endEntry(); -} - -void QRhiProfilerPrivate::releaseSwapChain(QRhiSwapChain *sc) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseSwapChain, ts.elapsed(), sc); - endEntry(); -} - -template<typename T> -void calcTiming(QList<T> *vec, T *minDelta, T *maxDelta, float *avgDelta) -{ - if (vec->isEmpty()) - return; - - *minDelta = *maxDelta = 0; - float totalDelta = 0; - for (T delta : qAsConst(*vec)) { - totalDelta += float(delta); - if (*minDelta == 0 || delta < *minDelta) - *minDelta = delta; - if (*maxDelta == 0 || delta > *maxDelta) - *maxDelta = delta; - } - *avgDelta = totalDelta / vec->count(); - - vec->clear(); -} - -void QRhiProfilerPrivate::beginSwapChainFrame(QRhiSwapChain *sc) -{ - Sc &scd(swapchains[sc]); - scd.beginToEndTimer.start(); -} - -void QRhiProfilerPrivate::endSwapChainFrame(QRhiSwapChain *sc, int frameCount) -{ - Sc &scd(swapchains[sc]); - if (!scd.frameToFrameRunning) { - scd.frameToFrameTimer.start(); - scd.frameToFrameRunning = true; - return; - } - - scd.frameToFrameSamples.append(scd.frameToFrameTimer.restart()); - if (scd.frameToFrameSamples.count() >= frameTimingWriteInterval) { - calcTiming(&scd.frameToFrameSamples, - &scd.frameToFrameTime.minTime, &scd.frameToFrameTime.maxTime, &scd.frameToFrameTime.avgTime); - if (outputDevice) { - startEntry(QRhiProfiler::FrameToFrameTime, ts.elapsed(), sc); - writeInt("frames_since_resize", frameCount); - writeInt("min_ms_frame_delta", scd.frameToFrameTime.minTime); - writeInt("max_ms_frame_delta", scd.frameToFrameTime.maxTime); - writeFloat("Favg_ms_frame_delta", scd.frameToFrameTime.avgTime); - endEntry(); - } - } - - scd.beginToEndSamples.append(scd.beginToEndTimer.elapsed()); - if (scd.beginToEndSamples.count() >= frameTimingWriteInterval) { - calcTiming(&scd.beginToEndSamples, - &scd.beginToEndFrameTime.minTime, &scd.beginToEndFrameTime.maxTime, &scd.beginToEndFrameTime.avgTime); - if (outputDevice) { - startEntry(QRhiProfiler::FrameBuildTime, ts.elapsed(), sc); - writeInt("frames_since_resize", frameCount); - writeInt("min_ms_frame_build", scd.beginToEndFrameTime.minTime); - writeInt("max_ms_frame_build", scd.beginToEndFrameTime.maxTime); - writeFloat("Favg_ms_frame_build", scd.beginToEndFrameTime.avgTime); - endEntry(); - } - } -} - -void QRhiProfilerPrivate::swapChainFrameGpuTime(QRhiSwapChain *sc, float gpuTime) -{ - Sc &scd(swapchains[sc]); - scd.gpuFrameSamples.append(gpuTime); - if (scd.gpuFrameSamples.count() >= frameTimingWriteInterval) { - calcTiming(&scd.gpuFrameSamples, - &scd.gpuFrameTime.minTime, &scd.gpuFrameTime.maxTime, &scd.gpuFrameTime.avgTime); - if (outputDevice) { - startEntry(QRhiProfiler::GpuFrameTime, ts.elapsed(), sc); - writeFloat("Fmin_ms_gpu_frame_time", scd.gpuFrameTime.minTime); - writeFloat("Fmax_ms_gpu_frame_time", scd.gpuFrameTime.maxTime); - writeFloat("Favg_ms_gpu_frame_time", scd.gpuFrameTime.avgTime); - endEntry(); - } - } -} - -void QRhiProfilerPrivate::newReadbackBuffer(qint64 id, QRhiResource *src, quint32 size) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::NewReadbackBuffer, ts.elapsed(), src); - writeInt("id", id); - writeInt("size", size); - endEntry(); -} - -void QRhiProfilerPrivate::releaseReadbackBuffer(qint64 id) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::ReleaseReadbackBuffer, ts.elapsed(), nullptr); - writeInt("id", id); - endEntry(); -} - -void QRhiProfilerPrivate::vmemStat(uint realAllocCount, uint subAllocCount, quint32 totalSize, quint32 unusedSize) -{ - if (!outputDevice) - return; - - startEntry(QRhiProfiler::GpuMemAllocStats, ts.elapsed(), nullptr); - writeInt("real_alloc_count", realAllocCount); - writeInt("sub_alloc_count", subAllocCount); - writeInt("total_size", totalSize); - writeInt("unused_size", unusedSize); - endEntry(); -} - -QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhiprofiler_p.h b/src/gui/rhi/qrhiprofiler_p.h deleted file mode 100644 index 9707fb5972..0000000000 --- a/src/gui/rhi/qrhiprofiler_p.h +++ /dev/null @@ -1,123 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIPROFILER_H -#define QRHIPROFILER_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include <private/qrhi_p.h> - -QT_BEGIN_NAMESPACE - -class QRhiProfilerPrivate; -class QIODevice; - -class Q_GUI_EXPORT QRhiProfiler -{ -public: - enum StreamOp { - NewBuffer = 1, - ReleaseBuffer, - NewBufferStagingArea, - ReleaseBufferStagingArea, - NewRenderBuffer, - ReleaseRenderBuffer, - NewTexture, - ReleaseTexture, - NewTextureStagingArea, - ReleaseTextureStagingArea, - ResizeSwapChain, - ReleaseSwapChain, - NewReadbackBuffer, - ReleaseReadbackBuffer, - GpuMemAllocStats, - GpuFrameTime, - FrameToFrameTime, - FrameBuildTime - }; - - ~QRhiProfiler(); - - void setDevice(QIODevice *device); - - void addVMemAllocatorStats(); - - int frameTimingWriteInterval() const; - void setFrameTimingWriteInterval(int frameCount); - - struct CpuTime { - qint64 minTime = 0; - qint64 maxTime = 0; - float avgTime = 0; - }; - - struct GpuTime { - float minTime = 0; - float maxTime = 0; - float avgTime = 0; - }; - - CpuTime frameToFrameTimes(QRhiSwapChain *sc) const; - CpuTime frameBuildTimes(QRhiSwapChain *sc) const; // beginFrame - endFrame - GpuTime gpuFrameTimes(QRhiSwapChain *sc) const; - -private: - Q_DISABLE_COPY(QRhiProfiler) - QRhiProfiler(); - QRhiProfilerPrivate *d; - friend class QRhiImplementation; - friend class QRhiProfilerPrivate; -}; - -Q_DECLARE_TYPEINFO(QRhiProfiler::CpuTime, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiProfiler::GpuTime, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhiprofiler_p_p.h b/src/gui/rhi/qrhiprofiler_p_p.h deleted file mode 100644 index 7d0eec8614..0000000000 --- a/src/gui/rhi/qrhiprofiler_p_p.h +++ /dev/null @@ -1,124 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIPROFILER_P_H -#define QRHIPROFILER_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhiprofiler_p.h" -#include <QElapsedTimer> -#include <QHash> - -QT_BEGIN_NAMESPACE - -class QRhiProfilerPrivate -{ -public: - static QRhiProfilerPrivate *get(QRhiProfiler *p) { return p->d; } - - void newBuffer(QRhiBuffer *buf, quint32 realSize, int backingGpuBufCount, int backingCpuBufCount); - void releaseBuffer(QRhiBuffer *buf); - void newBufferStagingArea(QRhiBuffer *buf, int slot, quint32 size); - void releaseBufferStagingArea(QRhiBuffer *buf, int slot); - - void newRenderBuffer(QRhiRenderBuffer *rb, bool transientBacking, bool winSysBacking, int sampleCount); - void releaseRenderBuffer(QRhiRenderBuffer *rb); - - void newTexture(QRhiTexture *tex, bool owns, int mipCount, int layerCount, int sampleCount); - void releaseTexture(QRhiTexture *tex); - void newTextureStagingArea(QRhiTexture *tex, int slot, quint32 size); - void releaseTextureStagingArea(QRhiTexture *tex, int slot); - - void resizeSwapChain(QRhiSwapChain *sc, int bufferCount, int msaaBufferCount, int sampleCount); - void releaseSwapChain(QRhiSwapChain *sc); - - void beginSwapChainFrame(QRhiSwapChain *sc); - void endSwapChainFrame(QRhiSwapChain *sc, int frameCount); - void swapChainFrameGpuTime(QRhiSwapChain *sc, float gpuTimeMs); - - void newReadbackBuffer(qint64 id, QRhiResource *src, quint32 size); - void releaseReadbackBuffer(qint64 id); - - void vmemStat(uint realAllocCount, uint subAllocCount, quint32 totalSize, quint32 unusedSize); - - void startEntry(QRhiProfiler::StreamOp op, qint64 timestamp, QRhiResource *res); - void writeInt(const char *key, qint64 v); - void writeFloat(const char *key, float f); - void endEntry(); - - QRhiImplementation *rhiDWhenEnabled = nullptr; - QIODevice *outputDevice = nullptr; - QElapsedTimer ts; - QByteArray buf; - static const int DEFAULT_FRAME_TIMING_WRITE_INTERVAL = 120; // frames - int frameTimingWriteInterval = DEFAULT_FRAME_TIMING_WRITE_INTERVAL; - struct Sc { - Sc() { - frameToFrameSamples.reserve(DEFAULT_FRAME_TIMING_WRITE_INTERVAL); - beginToEndSamples.reserve(DEFAULT_FRAME_TIMING_WRITE_INTERVAL); - gpuFrameSamples.reserve(DEFAULT_FRAME_TIMING_WRITE_INTERVAL); - } - QElapsedTimer frameToFrameTimer; - bool frameToFrameRunning = false; - QElapsedTimer beginToEndTimer; - QList<qint64> frameToFrameSamples; - QList<qint64> beginToEndSamples; - QList<float> gpuFrameSamples; - QRhiProfiler::CpuTime frameToFrameTime; - QRhiProfiler::CpuTime beginToEndFrameTime; - QRhiProfiler::GpuTime gpuFrameTime; - }; - QHash<QRhiSwapChain *, Sc> swapchains; -}; - -Q_DECLARE_TYPEINFO(QRhiProfilerPrivate::Sc, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhivulkan.cpp b/src/gui/rhi/qrhivulkan.cpp index 64b979b8c7..f0b51146cc 100644 --- a/src/gui/rhi/qrhivulkan.cpp +++ b/src/gui/rhi/qrhivulkan.cpp @@ -1,46 +1,11 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qrhivulkan_p_p.h" -#include "qrhivulkanext_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhivulkan_p.h" +#include <qpa/qplatformvulkaninstance.h> #define VMA_IMPLEMENTATION +#define VMA_DYNAMIC_VULKAN_FUNCTIONS 1 #define VMA_STATIC_VULKAN_FUNCTIONS 0 #define VMA_RECORDING_ENABLED 0 #define VMA_DEDICATED_ALLOCATION 0 @@ -49,12 +14,16 @@ #endif QT_WARNING_PUSH QT_WARNING_DISABLE_GCC("-Wsuggest-override") +#if defined(Q_CC_CLANG) && Q_CC_CLANG >= 1100 +QT_WARNING_DISABLE_CLANG("-Wdeprecated-copy") +#endif #include "vk_mem_alloc.h" QT_WARNING_POP #include <qmath.h> #include <QVulkanFunctions> #include <QtGui/qwindow.h> +#include <optional> QT_BEGIN_NAMESPACE @@ -90,10 +59,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiVulkanInitParams - \internal \inmodule QtGui + \since 6.6 \brief Vulkan specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Vulkan-based QRhi needs at minimum a valid QVulkanInstance. It is up to the user to ensure this is available and initialized. This is typically done in main() similarly to the following: @@ -196,34 +168,121 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiVulkanInitParams::inst + + The QVulkanInstance that has already been successfully + \l{QVulkanInstance::create()}{created}, required. +*/ + +/*! + \variable QRhiVulkanInitParams::window + + Optional, but recommended when targeting a QWindow. +*/ + +/*! + \variable QRhiVulkanInitParams::deviceExtensions + + Optional, empty by default. The list of Vulkan device extensions to enable. + Unsupported extensions are ignored gracefully. +*/ + +/*! \class QRhiVulkanNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Collects device, queue, and other Vulkan objects that are used by the QRhi. \note Ownership of the Vulkan objects is never transferred. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiVulkanNativeHandles::physDev + + When different from \nullptr, specifies the Vulkan physical device to use. +*/ + +/*! + \variable QRhiVulkanNativeHandles::dev + + When wanting to import not just a physical device, but also use an already + existing VkDevice, set this and the graphics queue index and family index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueueFamilyIdx + + Graphics queue family index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueueIdx + + Graphics queue index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::vmemAllocator + + Relevant only when importing an existing memory allocator object, + leave it set to \nullptr otherwise. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueue + + Output only, not used by QRhi::create(), only set by the + QRhi::nativeHandles() accessor. The graphics VkQueue used by the QRhi. +*/ + +/*! + \variable QRhiVulkanNativeHandles::inst + + Output only, not used by QRhi::create(), only set by the + QRhi::nativeHandles() accessor. The QVulkanInstance used by the QRhi. +*/ + +/*! \class QRhiVulkanCommandBufferNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the Vulkan command buffer object that is backing a QRhiCommandBuffer. \note The Vulkan command buffer object is only guaranteed to be valid, and in recording state, while recording a frame. That is, between a \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} or \l{QRhi::beginOffscreenFrame()}{beginOffscreenFrame()} - - \l{QRhi::endOffsrceenFrame()}{endOffscreenFrame()} pair. + \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} pair. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiVulkanCommandBufferNativeHandles::commandBuffer + + The VkCommandBuffer object. +*/ + +/*! \class QRhiVulkanRenderPassNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the Vulkan render pass object backing a QRhiRenderPassDescriptor. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiVulkanRenderPassNativeHandles::renderPass + + The VkRenderPass object. +*/ + template <class Int> inline Int aligned(Int v, Int byteAlign) { @@ -232,84 +291,14 @@ inline Int aligned(Int v, Int byteAlign) static QVulkanInstance *globalVulkanInstance; -static void VKAPI_PTR wrap_vkGetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, VkPhysicalDeviceProperties* pProperties) -{ - globalVulkanInstance->functions()->vkGetPhysicalDeviceProperties(physicalDevice, pProperties); -} - -static void VKAPI_PTR wrap_vkGetPhysicalDeviceMemoryProperties(VkPhysicalDevice physicalDevice, VkPhysicalDeviceMemoryProperties* pMemoryProperties) -{ - globalVulkanInstance->functions()->vkGetPhysicalDeviceMemoryProperties(physicalDevice, pMemoryProperties); -} - -static VkResult VKAPI_PTR wrap_vkAllocateMemory(VkDevice device, const VkMemoryAllocateInfo* pAllocateInfo, const VkAllocationCallbacks* pAllocator, VkDeviceMemory* pMemory) +static VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL wrap_vkGetInstanceProcAddr(VkInstance, const char *pName) { - return globalVulkanInstance->deviceFunctions(device)->vkAllocateMemory(device, pAllocateInfo, pAllocator, pMemory); + return globalVulkanInstance->getInstanceProcAddr(pName); } -void VKAPI_PTR wrap_vkFreeMemory(VkDevice device, VkDeviceMemory memory, const VkAllocationCallbacks* pAllocator) +static VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL wrap_vkGetDeviceProcAddr(VkDevice device, const char *pName) { - globalVulkanInstance->deviceFunctions(device)->vkFreeMemory(device, memory, pAllocator); -} - -VkResult VKAPI_PTR wrap_vkMapMemory(VkDevice device, VkDeviceMemory memory, VkDeviceSize offset, VkDeviceSize size, VkMemoryMapFlags flags, void** ppData) -{ - return globalVulkanInstance->deviceFunctions(device)->vkMapMemory(device, memory, offset, size, flags, ppData); -} - -void VKAPI_PTR wrap_vkUnmapMemory(VkDevice device, VkDeviceMemory memory) -{ - globalVulkanInstance->deviceFunctions(device)->vkUnmapMemory(device, memory); -} - -VkResult VKAPI_PTR wrap_vkFlushMappedMemoryRanges(VkDevice device, uint32_t memoryRangeCount, const VkMappedMemoryRange* pMemoryRanges) -{ - return globalVulkanInstance->deviceFunctions(device)->vkFlushMappedMemoryRanges(device, memoryRangeCount, pMemoryRanges); -} - -VkResult VKAPI_PTR wrap_vkInvalidateMappedMemoryRanges(VkDevice device, uint32_t memoryRangeCount, const VkMappedMemoryRange* pMemoryRanges) -{ - return globalVulkanInstance->deviceFunctions(device)->vkInvalidateMappedMemoryRanges(device, memoryRangeCount, pMemoryRanges); -} - -VkResult VKAPI_PTR wrap_vkBindBufferMemory(VkDevice device, VkBuffer buffer, VkDeviceMemory memory, VkDeviceSize memoryOffset) -{ - return globalVulkanInstance->deviceFunctions(device)->vkBindBufferMemory(device, buffer, memory, memoryOffset); -} - -VkResult VKAPI_PTR wrap_vkBindImageMemory(VkDevice device, VkImage image, VkDeviceMemory memory, VkDeviceSize memoryOffset) -{ - return globalVulkanInstance->deviceFunctions(device)->vkBindImageMemory(device, image, memory, memoryOffset); -} - -void VKAPI_PTR wrap_vkGetBufferMemoryRequirements(VkDevice device, VkBuffer buffer, VkMemoryRequirements* pMemoryRequirements) -{ - globalVulkanInstance->deviceFunctions(device)->vkGetBufferMemoryRequirements(device, buffer, pMemoryRequirements); -} - -void VKAPI_PTR wrap_vkGetImageMemoryRequirements(VkDevice device, VkImage image, VkMemoryRequirements* pMemoryRequirements) -{ - globalVulkanInstance->deviceFunctions(device)->vkGetImageMemoryRequirements(device, image, pMemoryRequirements); -} - -VkResult VKAPI_PTR wrap_vkCreateBuffer(VkDevice device, const VkBufferCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkBuffer* pBuffer) -{ - return globalVulkanInstance->deviceFunctions(device)->vkCreateBuffer(device, pCreateInfo, pAllocator, pBuffer); -} - -void VKAPI_PTR wrap_vkDestroyBuffer(VkDevice device, VkBuffer buffer, const VkAllocationCallbacks* pAllocator) -{ - globalVulkanInstance->deviceFunctions(device)->vkDestroyBuffer(device, buffer, pAllocator); -} - -VkResult VKAPI_PTR wrap_vkCreateImage(VkDevice device, const VkImageCreateInfo* pCreateInfo, const VkAllocationCallbacks* pAllocator, VkImage* pImage) -{ - return globalVulkanInstance->deviceFunctions(device)->vkCreateImage(device, pCreateInfo, pAllocator, pImage); -} - -void VKAPI_PTR wrap_vkDestroyImage(VkDevice device, VkImage image, const VkAllocationCallbacks* pAllocator) -{ - globalVulkanInstance->deviceFunctions(device)->vkDestroyImage(device, image, pAllocator); + return globalVulkanInstance->functions()->vkGetDeviceProcAddr(device, pName); } static inline VmaAllocation toVmaAllocation(QVkAlloc a) @@ -322,6 +311,13 @@ static inline VmaAllocator toVmaAllocator(QVkAllocator a) return reinterpret_cast<VmaAllocator>(a); } +/*! + \return the list of instance extensions that are expected to be enabled on + the QVulkanInstance that is used for the Vulkan-based QRhi. + + The returned list can be safely passed to QVulkanInstance::setExtensions() + as-is, because unsupported extensions are filtered out automatically. + */ QByteArrayList QRhiVulkanInitParams::preferredInstanceExtensions() { return { @@ -329,12 +325,18 @@ QByteArrayList QRhiVulkanInitParams::preferredInstanceExtensions() }; } +/*! + \return the list of device extensions that are expected to be enabled on the + \c VkDevice when creating a Vulkan-based QRhi with an externally created + \c VkDevice object. + */ QByteArrayList QRhiVulkanInitParams::preferredExtensionsForImportedDevice() { return { QByteArrayLiteral("VK_KHR_swapchain"), - QByteArrayLiteral("VK_EXT_debug_marker"), - QByteArrayLiteral("VK_EXT_vertex_attribute_divisor") + QByteArrayLiteral("VK_EXT_vertex_attribute_divisor"), + QByteArrayLiteral("VK_KHR_create_renderpass2"), + QByteArrayLiteral("VK_KHR_depth_stencil_resolve") }; } @@ -361,20 +363,19 @@ QRhiVulkan::QRhiVulkan(QRhiVulkanInitParams *params, QRhiVulkanNativeHandles *im } } -static bool qvk_debug_filter(VkDebugReportFlagsEXT flags, VkDebugReportObjectTypeEXT objectType, uint64_t object, - size_t location, int32_t messageCode, const char *pLayerPrefix, const char *pMessage) +static bool qvk_debug_filter(QVulkanInstance::DebugMessageSeverityFlags severity, + QVulkanInstance::DebugMessageTypeFlags type, + const void *callbackData) { - Q_UNUSED(flags); - Q_UNUSED(objectType); - Q_UNUSED(object); - Q_UNUSED(location); - Q_UNUSED(messageCode); - Q_UNUSED(pLayerPrefix); + Q_UNUSED(severity); + Q_UNUSED(type); +#ifdef VK_EXT_debug_utils + const VkDebugUtilsMessengerCallbackDataEXT *d = static_cast<const VkDebugUtilsMessengerCallbackDataEXT *>(callbackData); // Filter out certain misleading validation layer messages, as per // VulkanMemoryAllocator documentation. - if (strstr(pMessage, "Mapping an image with layout") - && strstr(pMessage, "can result in undefined behavior if this memory is used by the device")) + if (strstr(d->pMessage, "Mapping an image with layout") + && strstr(d->pMessage, "can result in undefined behavior if this memory is used by the device")) { return true; } @@ -385,9 +386,11 @@ static bool qvk_debug_filter(VkDebugReportFlagsEXT flags, VkDebugReportObjectTyp // then move on to another pool. If there is a real error, a qWarning // message is shown by allocateDescriptorSet(), so the validation warning // does not have any value and is just noise. - if (strstr(pMessage, "VUID-VkDescriptorSetAllocateInfo-descriptorPool-00307")) + if (strstr(d->pMessage, "VUID-VkDescriptorSetAllocateInfo-descriptorPool-00307")) return true; - +#else + Q_UNUSED(callbackData); +#endif return false; } @@ -417,11 +420,19 @@ bool QRhiVulkan::create(QRhi::Flags flags) return false; } - globalVulkanInstance = inst; // assume this will not change during the lifetime of the entire application + rhiFlags = flags; + qCDebug(QRHI_LOG_INFO, "Initializing QRhi Vulkan backend %p with flags %d", this, int(rhiFlags)); + globalVulkanInstance = inst; // used for function resolving in vkmemalloc callbacks f = inst->functions(); + if (QRHI_LOG_INFO().isEnabled(QtDebugMsg)) { + qCDebug(QRHI_LOG_INFO, "Enabled instance extensions:"); + for (const char *ext : inst->extensions()) + qCDebug(QRHI_LOG_INFO, " %s", ext); + } - rhiFlags = flags; + caps = {}; + caps.debugUtils = inst->extensions().contains(QByteArrayLiteral("VK_EXT_debug_utils")); QList<VkQueueFamilyProperties> queueFamilyProps; auto queryQueueFamilyProps = [this, &queueFamilyProps] { @@ -502,50 +513,125 @@ bool QRhiVulkan::create(QRhi::Flags flags) physDevProperties.deviceType); } + caps.apiVersion = inst->apiVersion(); + + // Check the physical device API version against the instance API version, + // they do not have to match, which means whatever version was set in the + // QVulkanInstance may not be legally used with a given device if the + // physical device has a lower version. + const QVersionNumber physDevApiVersion(VK_VERSION_MAJOR(physDevProperties.apiVersion), + VK_VERSION_MINOR(physDevProperties.apiVersion)); // patch version left out intentionally + if (physDevApiVersion < caps.apiVersion) { + qCDebug(QRHI_LOG_INFO) << "Instance has api version" << caps.apiVersion + << "whereas the chosen physical device has" << physDevApiVersion + << "- restricting to the latter"; + caps.apiVersion = physDevApiVersion; + } + driverInfoStruct.deviceName = QByteArray(physDevProperties.deviceName); driverInfoStruct.deviceId = physDevProperties.deviceID; driverInfoStruct.vendorId = physDevProperties.vendorID; driverInfoStruct.deviceType = toRhiDeviceType(physDevProperties.deviceType); - f->vkGetPhysicalDeviceFeatures(physDev, &physDevFeatures); + bool featuresQueried = false; +#ifdef VK_VERSION_1_1 + VkPhysicalDeviceFeatures2 physDevFeaturesChainable = {}; + physDevFeaturesChainable.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; +#endif + + // Vulkan >=1.2 headers at build time, >=1.2 implementation at run time +#ifdef VK_VERSION_1_2 + if (!featuresQueried) { + // Vulkan11Features, Vulkan12Features, etc. are only in Vulkan 1.2 and newer. + if (caps.apiVersion >= QVersionNumber(1, 2)) { + physDevFeatures11IfApi12OrNewer = {}; + physDevFeatures11IfApi12OrNewer.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES; + physDevFeatures12 = {}; + physDevFeatures12.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES; +#ifdef VK_VERSION_1_3 + physDevFeatures13 = {}; + physDevFeatures13.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_3_FEATURES; +#endif + physDevFeaturesChainable.pNext = &physDevFeatures11IfApi12OrNewer; + physDevFeatures11IfApi12OrNewer.pNext = &physDevFeatures12; +#ifdef VK_VERSION_1_3 + if (caps.apiVersion >= QVersionNumber(1, 3)) + physDevFeatures12.pNext = &physDevFeatures13; +#endif + f->vkGetPhysicalDeviceFeatures2(physDev, &physDevFeaturesChainable); + memcpy(&physDevFeatures, &physDevFeaturesChainable.features, sizeof(VkPhysicalDeviceFeatures)); + featuresQueried = true; + } + } +#endif // VK_VERSION_1_2 + + // Vulkan >=1.1 headers at build time, 1.1 implementation at run time +#ifdef VK_VERSION_1_1 + if (!featuresQueried) { + // Vulkan versioning nightmares: if the runtime API version is 1.1, + // there is no Vulkan11Features (introduced in 1.2+, the headers might + // have the types and structs, but the Vulkan implementation version at + // run time is what matters). But there are individual feature structs. + // For multiview, it is important to get this right since at the time of + // writing Quest 3 Android is a Vulkan 1.1 implementation at run time on + // the headset. + if (caps.apiVersion == QVersionNumber(1, 1)) { + multiviewFeaturesIfApi11 = {}; + multiviewFeaturesIfApi11.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES; + physDevFeaturesChainable.pNext = &multiviewFeaturesIfApi11; + f->vkGetPhysicalDeviceFeatures2(physDev, &physDevFeaturesChainable); + memcpy(&physDevFeatures, &physDevFeaturesChainable.features, sizeof(VkPhysicalDeviceFeatures)); + featuresQueried = true; + } + } +#endif + + if (!featuresQueried) { + // If the API version at run time is 1.0 (or we are building with + // ancient 1.0 headers), then do the Vulkan 1.0 query. + f->vkGetPhysicalDeviceFeatures(physDev, &physDevFeatures); + featuresQueried = true; + } // Choose queue and create device, unless the device was specified in importParams. if (!importedDevice) { // We only support combined graphics+present queues. When it comes to // compute, only combined graphics+compute queue is used, compute gets // disabled otherwise. - gfxQueueFamilyIdx = -1; - int computelessGfxQueueCandidateIdx = -1; + std::optional<uint32_t> gfxQueueFamilyIdxOpt; + std::optional<uint32_t> computelessGfxQueueCandidateIdxOpt; queryQueueFamilyProps(); - for (int i = 0; i < queueFamilyProps.count(); ++i) { - qCDebug(QRHI_LOG_INFO, "queue family %d: flags=0x%x count=%d", + const uint32_t queueFamilyCount = uint32_t(queueFamilyProps.size()); + for (uint32_t i = 0; i < queueFamilyCount; ++i) { + qCDebug(QRHI_LOG_INFO, "queue family %u: flags=0x%x count=%u", i, queueFamilyProps[i].queueFlags, queueFamilyProps[i].queueCount); - if (gfxQueueFamilyIdx == -1 + if (!gfxQueueFamilyIdxOpt.has_value() && (queueFamilyProps[i].queueFlags & VK_QUEUE_GRAPHICS_BIT) - && (!maybeWindow || inst->supportsPresent(physDev, uint32_t(i), maybeWindow))) + && (!maybeWindow || inst->supportsPresent(physDev, i, maybeWindow))) { if (queueFamilyProps[i].queueFlags & VK_QUEUE_COMPUTE_BIT) - gfxQueueFamilyIdx = i; - else if (computelessGfxQueueCandidateIdx == -1) - computelessGfxQueueCandidateIdx = i; + gfxQueueFamilyIdxOpt = i; + else if (!computelessGfxQueueCandidateIdxOpt.has_value()) + computelessGfxQueueCandidateIdxOpt = i; } } - if (gfxQueueFamilyIdx == -1) { - if (computelessGfxQueueCandidateIdx != -1) { - gfxQueueFamilyIdx = computelessGfxQueueCandidateIdx; + if (gfxQueueFamilyIdxOpt.has_value()) { + gfxQueueFamilyIdx = gfxQueueFamilyIdxOpt.value(); + } else { + if (computelessGfxQueueCandidateIdxOpt.has_value()) { + gfxQueueFamilyIdx = computelessGfxQueueCandidateIdxOpt.value(); } else { qWarning("No graphics (or no graphics+present) queue family found"); return false; } } - VkDeviceQueueCreateInfo queueInfo[2]; + VkDeviceQueueCreateInfo queueInfo = {}; const float prio[] = { 0 }; - memset(queueInfo, 0, sizeof(queueInfo)); - queueInfo[0].sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; - queueInfo[0].queueFamilyIndex = uint32_t(gfxQueueFamilyIdx); - queueInfo[0].queueCount = 1; - queueInfo[0].pQueuePriorities = prio; + queueInfo.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_CREATE_INFO; + queueInfo.queueFamilyIndex = gfxQueueFamilyIdx; + queueInfo.queueCount = 1; + queueInfo.pQueuePriorities = prio; QList<const char *> devLayers; if (inst->layers().contains("VK_LAYER_KHRONOS_validation")) @@ -557,44 +643,69 @@ bool QRhiVulkan::create(QRhi::Flags flags) if (devExtCount) { QList<VkExtensionProperties> extProps(devExtCount); f->vkEnumerateDeviceExtensionProperties(physDev, nullptr, &devExtCount, extProps.data()); - for (const VkExtensionProperties &p : qAsConst(extProps)) + for (const VkExtensionProperties &p : std::as_const(extProps)) devExts.append({ p.extensionName, p.specVersion }); } - qCDebug(QRHI_LOG_INFO, "%d device extensions available", int(devExts.count())); + qCDebug(QRHI_LOG_INFO, "%d device extensions available", int(devExts.size())); QList<const char *> requestedDevExts; requestedDevExts.append("VK_KHR_swapchain"); - caps.debugMarkers = false; - if (devExts.contains(VK_EXT_DEBUG_MARKER_EXTENSION_NAME)) { - requestedDevExts.append(VK_EXT_DEBUG_MARKER_EXTENSION_NAME); - caps.debugMarkers = true; + const bool hasPhysDevProp2 = inst->extensions().contains(QByteArrayLiteral("VK_KHR_get_physical_device_properties2")); + + if (devExts.contains(QByteArrayLiteral("VK_KHR_portability_subset"))) { + if (hasPhysDevProp2) { + requestedDevExts.append("VK_KHR_portability_subset"); + } else { + qWarning("VK_KHR_portability_subset should be enabled on the device " + "but the instance does not have VK_KHR_get_physical_device_properties2 enabled. " + "Expect problems."); + } } - caps.vertexAttribDivisor = false; +#ifdef VK_EXT_vertex_attribute_divisor if (devExts.contains(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME)) { - if (inst->extensions().contains(QByteArrayLiteral("VK_KHR_get_physical_device_properties2"))) { + if (hasPhysDevProp2) { requestedDevExts.append(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME); caps.vertexAttribDivisor = true; } } +#endif + +#ifdef VK_KHR_create_renderpass2 + if (devExts.contains(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME)) { + requestedDevExts.append(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME); + caps.renderPass2KHR = true; + } +#endif + +#ifdef VK_KHR_depth_stencil_resolve + if (devExts.contains(VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME)) { + requestedDevExts.append(VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME); + caps.depthStencilResolveKHR = true; + } +#endif for (const QByteArray &ext : requestedDeviceExtensions) { - if (!ext.isEmpty()) { - if (devExts.contains(ext)) + if (!ext.isEmpty() && !requestedDevExts.contains(ext)) { + if (devExts.contains(ext)) { requestedDevExts.append(ext.constData()); - else - qWarning("Device extension %s is not supported", ext.constData()); + } else { + qWarning("Device extension %s requested in QRhiVulkanInitParams is not supported", + ext.constData()); + } } } QByteArrayList envExtList = qgetenv("QT_VULKAN_DEVICE_EXTENSIONS").split(';'); for (const QByteArray &ext : envExtList) { if (!ext.isEmpty() && !requestedDevExts.contains(ext)) { - if (devExts.contains(ext)) + if (devExts.contains(ext)) { requestedDevExts.append(ext.constData()); - else - qWarning("Device extension %s is not supported", ext.constData()); + } else { + qWarning("Device extension %s requested in QT_VULKAN_DEVICE_EXTENSIONS is not supported", + ext.constData()); + } } } @@ -604,29 +715,50 @@ bool QRhiVulkan::create(QRhi::Flags flags) qCDebug(QRHI_LOG_INFO, " %s", ext); } - VkDeviceCreateInfo devInfo; - memset(&devInfo, 0, sizeof(devInfo)); + VkDeviceCreateInfo devInfo = {}; devInfo.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO; devInfo.queueCreateInfoCount = 1; - devInfo.pQueueCreateInfos = queueInfo; - devInfo.enabledLayerCount = uint32_t(devLayers.count()); + devInfo.pQueueCreateInfos = &queueInfo; + devInfo.enabledLayerCount = uint32_t(devLayers.size()); devInfo.ppEnabledLayerNames = devLayers.constData(); - devInfo.enabledExtensionCount = uint32_t(requestedDevExts.count()); + devInfo.enabledExtensionCount = uint32_t(requestedDevExts.size()); devInfo.ppEnabledExtensionNames = requestedDevExts.constData(); - VkPhysicalDeviceFeatures features; - memset(&features, 0, sizeof(features)); - if (physDevFeatures.wideLines) - features.wideLines = VK_TRUE; - if (physDevFeatures.largePoints) - features.largePoints = VK_TRUE; - if (physDevFeatures.textureCompressionETC2) - features.textureCompressionETC2 = VK_TRUE; - if (physDevFeatures.textureCompressionASTC_LDR) - features.textureCompressionASTC_LDR = VK_TRUE; - if (physDevFeatures.textureCompressionBC) - features.textureCompressionBC = VK_TRUE; - devInfo.pEnabledFeatures = &features; + // Enable all features that are reported as supported, except + // robustness because that potentially affects performance. + // + // Enabling all features mainly serves third-party renderers that may + // use the VkDevice created here. For the record, the backend here + // optionally relies on the following features, meaning just for our + // (QRhi/Quick/Quick 3D) purposes it would be sufficient to + // enable-if-supported only the following: + // + // wideLines, largePoints, fillModeNonSolid, + // tessellationShader, geometryShader + // textureCompressionETC2, textureCompressionASTC_LDR, textureCompressionBC + +#ifdef VK_VERSION_1_1 + physDevFeaturesChainable.features.robustBufferAccess = VK_FALSE; +#endif +#ifdef VK_VERSION_1_3 + physDevFeatures13.robustImageAccess = VK_FALSE; +#endif + +#ifdef VK_VERSION_1_1 + if (caps.apiVersion >= QVersionNumber(1, 1)) { + // For a >=1.2 implementation at run time, this will enable all + // (1.0-1.3) features reported as supported, except the ones we turn + // off explicitly above. For a 1.1 implementation at run time, this + // only enables the 1.0 and multiview features reported as + // supported. We will not be bothering with the Vulkan 1.1 + // individual feature struct nonsense. + devInfo.pNext = &physDevFeaturesChainable; + } else +#endif + { + physDevFeatures.robustBufferAccess = VK_FALSE; + devInfo.pEnabledFeatures = &physDevFeatures; + } VkResult err = f->vkCreateDevice(physDev, &devInfo, nullptr, &dev); if (err != VK_SUCCESS) { @@ -635,14 +767,27 @@ bool QRhiVulkan::create(QRhi::Flags flags) } } else { qCDebug(QRHI_LOG_INFO, "Using imported device %p", dev); + + // Here we have no way to tell if the extensions got enabled or not. + // Pretend it's all there and supported. If getProcAddress fails, we'll + // handle that gracefully. + caps.vertexAttribDivisor = true; + caps.renderPass2KHR = true; + caps.depthStencilResolveKHR = true; } + vkGetPhysicalDeviceSurfaceCapabilitiesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR>( + inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfaceCapabilitiesKHR")); + vkGetPhysicalDeviceSurfaceFormatsKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfaceFormatsKHR>( + inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfaceFormatsKHR")); + vkGetPhysicalDeviceSurfacePresentModesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfacePresentModesKHR>( + inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfacePresentModesKHR")); + df = inst->deviceFunctions(dev); - VkCommandPoolCreateInfo poolInfo; - memset(&poolInfo, 0, sizeof(poolInfo)); + VkCommandPoolCreateInfo poolInfo = {}; poolInfo.sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO; - poolInfo.queueFamilyIndex = uint32_t(gfxQueueFamilyIdx); + poolInfo.queueFamilyIndex = gfxQueueFamilyIdx; for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { VkResult err = df->vkCreateCommandPool(dev, &poolInfo, nullptr, &cmdPool[i]); if (err != VK_SUCCESS) { @@ -651,13 +796,10 @@ bool QRhiVulkan::create(QRhi::Flags flags) } } - if (gfxQueueFamilyIdx < 0) { - // this is when importParams is faulty and did not specify the queue family index - qWarning("No queue family index provided"); - return false; - } + qCDebug(QRHI_LOG_INFO, "Using queue family index %u and queue index %u", + gfxQueueFamilyIdx, gfxQueueIdx); - df->vkGetDeviceQueue(dev, uint32_t(gfxQueueFamilyIdx), gfxQueueIdx, &gfxQueue); + df->vkGetDeviceQueue(dev, gfxQueueFamilyIdx, gfxQueueIdx, &gfxQueue); if (queueFamilyProps.isEmpty()) queryQueueFamilyProps(); @@ -672,35 +814,51 @@ bool QRhiVulkan::create(QRhi::Flags flags) caps.wideLines = physDevFeatures.wideLines; - caps.texture3DSliceAs2D = inst->apiVersion() >= QVersionNumber(1, 1); + caps.texture3DSliceAs2D = caps.apiVersion >= QVersionNumber(1, 1); + + caps.tessellation = physDevFeatures.tessellationShader; + caps.geometryShader = physDevFeatures.geometryShader; + + caps.nonFillPolygonMode = physDevFeatures.fillModeNonSolid; + +#ifdef VK_VERSION_1_2 + if (caps.apiVersion >= QVersionNumber(1, 2)) + caps.multiView = physDevFeatures11IfApi12OrNewer.multiview; +#endif + +#ifdef VK_VERSION_1_1 + if (caps.apiVersion == QVersionNumber(1, 1)) + caps.multiView = multiviewFeaturesIfApi11.multiview; +#endif + + // With Vulkan 1.2 renderpass2 and depth_stencil_resolve are core, but we + // have to support the case of 1.1 + extensions, in particular for the Quest + // 3 (Android, Vulkan 1.1 at the time of writing). Therefore, always rely on + // the KHR extension for now. +#ifdef VK_KHR_create_renderpass2 + if (caps.renderPass2KHR) { + vkCreateRenderPass2KHR = reinterpret_cast<PFN_vkCreateRenderPass2KHR>(f->vkGetDeviceProcAddr(dev, "vkCreateRenderPass2KHR")); + if (!vkCreateRenderPass2KHR) // handle it gracefully, the caps flag may be incorrect when using an imported VkDevice + caps.renderPass2KHR = false; + } +#endif if (!importedAllocator) { - VmaVulkanFunctions afuncs; - afuncs.vkGetPhysicalDeviceProperties = wrap_vkGetPhysicalDeviceProperties; - afuncs.vkGetPhysicalDeviceMemoryProperties = wrap_vkGetPhysicalDeviceMemoryProperties; - afuncs.vkAllocateMemory = wrap_vkAllocateMemory; - afuncs.vkFreeMemory = wrap_vkFreeMemory; - afuncs.vkMapMemory = wrap_vkMapMemory; - afuncs.vkUnmapMemory = wrap_vkUnmapMemory; - afuncs.vkFlushMappedMemoryRanges = wrap_vkFlushMappedMemoryRanges; - afuncs.vkInvalidateMappedMemoryRanges = wrap_vkInvalidateMappedMemoryRanges; - afuncs.vkBindBufferMemory = wrap_vkBindBufferMemory; - afuncs.vkBindImageMemory = wrap_vkBindImageMemory; - afuncs.vkGetBufferMemoryRequirements = wrap_vkGetBufferMemoryRequirements; - afuncs.vkGetImageMemoryRequirements = wrap_vkGetImageMemoryRequirements; - afuncs.vkCreateBuffer = wrap_vkCreateBuffer; - afuncs.vkDestroyBuffer = wrap_vkDestroyBuffer; - afuncs.vkCreateImage = wrap_vkCreateImage; - afuncs.vkDestroyImage = wrap_vkDestroyImage; - - VmaAllocatorCreateInfo allocatorInfo; - memset(&allocatorInfo, 0, sizeof(allocatorInfo)); + VmaVulkanFunctions funcs = {}; + funcs.vkGetInstanceProcAddr = wrap_vkGetInstanceProcAddr; + funcs.vkGetDeviceProcAddr = wrap_vkGetDeviceProcAddr; + + VmaAllocatorCreateInfo allocatorInfo = {}; // A QRhi is supposed to be used from one single thread only. Disable // the allocator's own mutexes. This gives a performance boost. allocatorInfo.flags = VMA_ALLOCATOR_CREATE_EXTERNALLY_SYNCHRONIZED_BIT; allocatorInfo.physicalDevice = physDev; allocatorInfo.device = dev; - allocatorInfo.pVulkanFunctions = &afuncs; + allocatorInfo.pVulkanFunctions = &funcs; + allocatorInfo.instance = inst->vkInstance(); + allocatorInfo.vulkanApiVersion = VK_MAKE_VERSION(caps.apiVersion.majorVersion(), + caps.apiVersion.minorVersion(), + caps.apiVersion.microVersion()); VmaAllocator vmaallocator; VkResult err = vmaCreateAllocator(&allocatorInfo, &vmaallocator); if (err != VK_SUCCESS) { @@ -719,8 +877,7 @@ bool QRhiVulkan::create(QRhi::Flags flags) else qWarning("Failed to create initial descriptor pool: %d", err); - VkQueryPoolCreateInfo timestampQueryPoolInfo; - memset(×tampQueryPoolInfo, 0, sizeof(timestampQueryPoolInfo)); + VkQueryPoolCreateInfo timestampQueryPoolInfo = {}; timestampQueryPoolInfo.sType = VK_STRUCTURE_TYPE_QUERY_POOL_CREATE_INFO; timestampQueryPoolInfo.queryType = VK_QUERY_TYPE_TIMESTAMP; timestampQueryPoolInfo.queryCount = QVK_MAX_ACTIVE_TIMESTAMP_PAIRS * 2; @@ -732,12 +889,14 @@ bool QRhiVulkan::create(QRhi::Flags flags) timestampQueryPoolMap.resize(QVK_MAX_ACTIVE_TIMESTAMP_PAIRS); // 1 bit per pair timestampQueryPoolMap.fill(false); - if (caps.debugMarkers) { - vkCmdDebugMarkerBegin = reinterpret_cast<PFN_vkCmdDebugMarkerBeginEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdDebugMarkerBeginEXT")); - vkCmdDebugMarkerEnd = reinterpret_cast<PFN_vkCmdDebugMarkerEndEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdDebugMarkerEndEXT")); - vkCmdDebugMarkerInsert = reinterpret_cast<PFN_vkCmdDebugMarkerInsertEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdDebugMarkerInsertEXT")); - vkDebugMarkerSetObjectName = reinterpret_cast<PFN_vkDebugMarkerSetObjectNameEXT>(f->vkGetDeviceProcAddr(dev, "vkDebugMarkerSetObjectNameEXT")); +#ifdef VK_EXT_debug_utils + if (caps.debugUtils) { + vkSetDebugUtilsObjectNameEXT = reinterpret_cast<PFN_vkSetDebugUtilsObjectNameEXT>(f->vkGetDeviceProcAddr(dev, "vkSetDebugUtilsObjectNameEXT")); + vkCmdBeginDebugUtilsLabelEXT = reinterpret_cast<PFN_vkCmdBeginDebugUtilsLabelEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdBeginDebugUtilsLabelEXT")); + vkCmdEndDebugUtilsLabelEXT = reinterpret_cast<PFN_vkCmdEndDebugUtilsLabelEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdEndDebugUtilsLabelEXT")); + vkCmdInsertDebugUtilsLabelEXT = reinterpret_cast<PFN_vkCmdInsertDebugUtilsLabelEXT>(f->vkGetDeviceProcAddr(dev, "vkCmdInsertDebugUtilsLabelEXT")); } +#endif deviceLost = false; @@ -747,6 +906,7 @@ bool QRhiVulkan::create(QRhi::Flags flags) nativeHandlesStruct.gfxQueueIdx = gfxQueueIdx; nativeHandlesStruct.gfxQueue = gfxQueue; nativeHandlesStruct.vmemAllocator = allocator; + nativeHandlesStruct.inst = inst; return true; } @@ -815,8 +975,7 @@ VkResult QRhiVulkan::createDescriptorPool(VkDescriptorPool *pool) { VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, QVK_STORAGE_BUFFERS_PER_POOL }, { VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, QVK_STORAGE_IMAGES_PER_POOL } }; - VkDescriptorPoolCreateInfo descPoolInfo; - memset(&descPoolInfo, 0, sizeof(descPoolInfo)); + VkDescriptorPoolCreateInfo descPoolInfo = {}; descPoolInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO; // Do not enable vkFreeDescriptorSets - sets are never freed on their own // (good so no trouble with fragmentation), they just deref their pool @@ -838,7 +997,7 @@ bool QRhiVulkan::allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, V return r; }; - int lastPoolIdx = descriptorPools.count() - 1; + int lastPoolIdx = descriptorPools.size() - 1; for (int i = lastPoolIdx; i >= 0; --i) { if (descriptorPools[i].refCount == 0) { df->vkResetDescriptorPool(dev, descriptorPools[i].pool, 0); @@ -858,7 +1017,7 @@ bool QRhiVulkan::allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, V VkResult poolErr = createDescriptorPool(&newPool); if (poolErr == VK_SUCCESS) { descriptorPools.append(newPool); - lastPoolIdx = descriptorPools.count() - 1; + lastPoolIdx = descriptorPools.size() - 1; VkResult err = tryAllocate(lastPoolIdx); if (err != VK_SUCCESS) { qWarning("Failed to allocate descriptor set from new pool too, giving up: %d", err); @@ -901,6 +1060,10 @@ static inline VkFormat toVkTextureFormat(QRhiTexture::Format format, QRhiTexture case QRhiTexture::R32F: return VK_FORMAT_R32_SFLOAT; + case QRhiTexture::RGB10A2: + // intentionally A2B10G10R10, not A2R10G10B10 + return VK_FORMAT_A2B10G10R10_UNORM_PACK32; + case QRhiTexture::D16: return VK_FORMAT_D16_UNORM; case QRhiTexture::D24: @@ -962,12 +1125,11 @@ static inline VkFormat toVkTextureFormat(QRhiTexture::Format format, QRhiTexture return srgb ? VK_FORMAT_ASTC_12x12_SRGB_BLOCK : VK_FORMAT_ASTC_12x12_UNORM_BLOCK; default: - Q_UNREACHABLE(); - return VK_FORMAT_R8G8B8A8_UNORM; + Q_UNREACHABLE_RETURN(VK_FORMAT_R8G8B8A8_UNORM); } } -static inline QRhiTexture::Format colorTextureFormatFromVkFormat(VkFormat format, QRhiTexture::Flags *flags) +static inline QRhiTexture::Format swapchainReadbackTextureFormat(VkFormat format, QRhiTexture::Flags *flags) { switch (format) { case VK_FORMAT_R8G8B8A8_UNORM: @@ -982,24 +1144,14 @@ static inline QRhiTexture::Format colorTextureFormatFromVkFormat(VkFormat format if (flags) (*flags) |= QRhiTexture::sRGB; return QRhiTexture::BGRA8; - case VK_FORMAT_R8_UNORM: - return QRhiTexture::R8; - case VK_FORMAT_R8G8_UNORM: - return QRhiTexture::RG8; - case VK_FORMAT_R8_SRGB: - if (flags) - (*flags) |= QRhiTexture::sRGB; - return QRhiTexture::R8; - case VK_FORMAT_R8G8_SRGB: - if (flags) - (*flags) |= QRhiTexture::sRGB; - return QRhiTexture::RG8; - case VK_FORMAT_R16_UNORM: - return QRhiTexture::R16; - case VK_FORMAT_R16G16_UNORM: - return QRhiTexture::RG16; - default: // this cannot assert, must warn and return unknown - qWarning("VkFormat %d is not a recognized uncompressed color format", format); + case VK_FORMAT_R16G16B16A16_SFLOAT: + return QRhiTexture::RGBA16F; + case VK_FORMAT_R32G32B32A32_SFLOAT: + return QRhiTexture::RGBA32F; + case VK_FORMAT_A2B10G10R10_UNORM_PACK32: + return QRhiTexture::RGB10A2; + default: + qWarning("VkFormat %d cannot be read back", format); break; } return QRhiTexture::UnknownFormat; @@ -1075,8 +1227,7 @@ bool QRhiVulkan::createTransientImage(VkFormat format, VkResult err; for (int i = 0; i < count; ++i) { - VkImageCreateInfo imgInfo; - memset(&imgInfo, 0, sizeof(imgInfo)); + VkImageCreateInfo imgInfo = {}; imgInfo.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; imgInfo.imageType = VK_IMAGE_TYPE_2D; imgInfo.format = format; @@ -1101,8 +1252,7 @@ bool QRhiVulkan::createTransientImage(VkFormat format, df->vkGetImageMemoryRequirements(dev, images[i], &memReq); } - VkMemoryAllocateInfo memInfo; - memset(&memInfo, 0, sizeof(memInfo)); + VkMemoryAllocateInfo memInfo = {}; memInfo.sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO; memInfo.allocationSize = aligned(memReq.size, memReq.alignment) * VkDeviceSize(count); @@ -1130,8 +1280,7 @@ bool QRhiVulkan::createTransientImage(VkFormat format, } ofs += aligned(memReq.size, memReq.alignment); - VkImageViewCreateInfo imgViewInfo; - memset(&imgViewInfo, 0, sizeof(imgViewInfo)); + VkImageViewCreateInfo imgViewInfo = {}; imgViewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; imgViewInfo.image = images[i]; imgViewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; @@ -1185,18 +1334,18 @@ static void fillRenderPassCreateInfo(VkRenderPassCreateInfo *rpInfo, { memset(subpassDesc, 0, sizeof(VkSubpassDescription)); subpassDesc->pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; - subpassDesc->colorAttachmentCount = uint32_t(rpD->colorRefs.count()); + subpassDesc->colorAttachmentCount = uint32_t(rpD->colorRefs.size()); subpassDesc->pColorAttachments = !rpD->colorRefs.isEmpty() ? rpD->colorRefs.constData() : nullptr; subpassDesc->pDepthStencilAttachment = rpD->hasDepthStencil ? &rpD->dsRef : nullptr; subpassDesc->pResolveAttachments = !rpD->resolveRefs.isEmpty() ? rpD->resolveRefs.constData() : nullptr; memset(rpInfo, 0, sizeof(VkRenderPassCreateInfo)); rpInfo->sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; - rpInfo->attachmentCount = uint32_t(rpD->attDescs.count()); + rpInfo->attachmentCount = uint32_t(rpD->attDescs.size()); rpInfo->pAttachments = rpD->attDescs.constData(); rpInfo->subpassCount = 1; rpInfo->pSubpasses = subpassDesc; - rpInfo->dependencyCount = uint32_t(rpD->subpassDeps.count()); + rpInfo->dependencyCount = uint32_t(rpD->subpassDeps.size()); rpInfo->pDependencies = !rpD->subpassDeps.isEmpty() ? rpD->subpassDeps.constData() : nullptr; } @@ -1204,8 +1353,7 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD { // attachment list layout is color (1), ds (0-1), resolve (0-1) - VkAttachmentDescription attDesc; - memset(&attDesc, 0, sizeof(attDesc)); + VkAttachmentDescription attDesc = {}; attDesc.format = colorFormat; attDesc.samples = samples; attDesc.loadOp = VK_ATTACHMENT_LOAD_OP_CLEAR; @@ -1219,6 +1367,8 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD rpD->colorRefs.append({ 0, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }); rpD->hasDepthStencil = hasDepthStencil; + rpD->hasDepthStencilResolve = false; + rpD->multiViewCount = 0; if (hasDepthStencil) { // clear on load + no store + lazy alloc + transient image should play @@ -1253,8 +1403,7 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD } // Replace the first implicit dep (TOP_OF_PIPE / ALL_COMMANDS) with our own. - VkSubpassDependency subpassDep; - memset(&subpassDep, 0, sizeof(subpassDep)); + VkSubpassDependency subpassDep = {}; subpassDep.srcSubpass = VK_SUBPASS_EXTERNAL; subpassDep.dstSubpass = 0; subpassDep.srcStageMask = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT; @@ -1289,29 +1438,190 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD return true; } +struct MultiViewRenderPassSetupHelper +{ + bool prepare(VkRenderPassCreateInfo *rpInfo, int multiViewCount, bool multiViewCap) + { + if (multiViewCount < 2) + return true; + if (!multiViewCap) { + qWarning("Cannot create multiview render pass without support for the Vulkan 1.1 multiview feature"); + return false; + } +#ifdef VK_VERSION_1_1 + uint32_t allViewsMask = 0; + for (uint32_t i = 0; i < uint32_t(multiViewCount); ++i) + allViewsMask |= (1 << i); + multiViewMask = allViewsMask; + multiViewCorrelationMask = allViewsMask; + multiViewInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_MULTIVIEW_CREATE_INFO; + multiViewInfo.subpassCount = 1; + multiViewInfo.pViewMasks = &multiViewMask; + multiViewInfo.correlationMaskCount = 1; + multiViewInfo.pCorrelationMasks = &multiViewCorrelationMask; + rpInfo->pNext = &multiViewInfo; +#endif + return true; + } + +#ifdef VK_VERSION_1_1 + VkRenderPassMultiviewCreateInfo multiViewInfo = {}; + uint32_t multiViewMask = 0; + uint32_t multiViewCorrelationMask = 0; +#endif +}; + +#ifdef VK_KHR_create_renderpass2 +// Effectively converts a VkRenderPassCreateInfo into a VkRenderPassCreateInfo2, +// adding depth-stencil resolve support. Assumes a single subpass and no subpass +// dependencies. +struct RenderPass2SetupHelper +{ + bool prepare(VkRenderPassCreateInfo2 *rpInfo2, const VkRenderPassCreateInfo *rpInfo, const QVkRenderPassDescriptor *rpD, int multiViewCount) { + *rpInfo2 = {}; + + viewMask = 0; + if (multiViewCount >= 2) { + for (uint32_t i = 0; i < uint32_t(multiViewCount); ++i) + viewMask |= (1 << i); + } + + attDescs2.resize(rpInfo->attachmentCount); + for (qsizetype i = 0; i < attDescs2.count(); ++i) { + VkAttachmentDescription2KHR &att2(attDescs2[i]); + const VkAttachmentDescription &att(rpInfo->pAttachments[i]); + att2 = {}; + att2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2; + att2.flags = att.flags; + att2.format = att.format; + att2.samples = att.samples; + att2.loadOp = att.loadOp; + att2.storeOp = att.storeOp; + att2.stencilLoadOp = att.stencilLoadOp; + att2.stencilStoreOp = att.stencilStoreOp; + att2.initialLayout = att.initialLayout; + att2.finalLayout = att.finalLayout; + } + + attRefs2.clear(); + subpass2 = {}; + subpass2.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR; + const VkSubpassDescription &subpassDesc(rpInfo->pSubpasses[0]); + subpass2.flags = subpassDesc.flags; + subpass2.pipelineBindPoint = subpassDesc.pipelineBindPoint; + if (multiViewCount >= 2) + subpass2.viewMask = viewMask; + + // color attachment refs + qsizetype startIndex = attRefs2.count(); + for (uint32_t j = 0; j < subpassDesc.colorAttachmentCount; ++j) { + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(subpassDesc.pColorAttachments[j]); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + } + subpass2.colorAttachmentCount = subpassDesc.colorAttachmentCount; + subpass2.pColorAttachments = attRefs2.constData() + startIndex; + + // color resolve refs + if (subpassDesc.pResolveAttachments) { + startIndex = attRefs2.count(); + for (uint32_t j = 0; j < subpassDesc.colorAttachmentCount; ++j) { + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(subpassDesc.pResolveAttachments[j]); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + } + subpass2.pResolveAttachments = attRefs2.constData() + startIndex; + } + + // depth-stencil ref + if (subpassDesc.pDepthStencilAttachment) { + startIndex = attRefs2.count(); + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(*subpassDesc.pDepthStencilAttachment); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + subpass2.pDepthStencilAttachment = attRefs2.constData() + startIndex; + } + + // depth-stencil resolve ref +#ifdef VK_KHR_depth_stencil_resolve + dsResolveDesc = {}; + if (rpD->hasDepthStencilResolve) { + startIndex = attRefs2.count(); + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = rpD->dsResolveRef.attachment; + attref2.layout = rpD->dsResolveRef.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + dsResolveDesc.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE_KHR; + dsResolveDesc.depthResolveMode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT; + dsResolveDesc.stencilResolveMode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT; + dsResolveDesc.pDepthStencilResolveAttachment = attRefs2.constData() + startIndex; + subpass2.pNext = &dsResolveDesc; + } +#endif + + rpInfo2->sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR; + rpInfo2->pNext = nullptr; // the 1.1 VkRenderPassMultiviewCreateInfo is part of the '2' structs + rpInfo2->flags = rpInfo->flags; + rpInfo2->attachmentCount = rpInfo->attachmentCount; + rpInfo2->pAttachments = attDescs2.constData(); + rpInfo2->subpassCount = 1; + rpInfo2->pSubpasses = &subpass2; + if (multiViewCount >= 2) { + rpInfo2->correlatedViewMaskCount = 1; + rpInfo2->pCorrelatedViewMasks = &viewMask; + } + return true; + } + + QVarLengthArray<VkAttachmentDescription2KHR, 8> attDescs2; + QVarLengthArray<VkAttachmentReference2KHR, 8> attRefs2; + VkSubpassDescription2KHR subpass2; +#ifdef VK_KHR_depth_stencil_resolve + VkSubpassDescriptionDepthStencilResolveKHR dsResolveDesc; +#endif + uint32_t viewMask; +}; +#endif // VK_KHR_create_renderpass2 + bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, const QRhiColorAttachment *firstColorAttachment, const QRhiColorAttachment *lastColorAttachment, bool preserveColor, bool preserveDs, + bool storeDs, QRhiRenderBuffer *depthStencilBuffer, - QRhiTexture *depthTexture) + QRhiTexture *depthTexture, + QRhiTexture *depthResolveTexture) { - // attachment list layout is color (0-8), ds (0-1), resolve (0-8) + // attachment list layout is color (0-8), ds (0-1), resolve (0-8), ds resolve (0-1) + int multiViewCount = 0; for (auto it = firstColorAttachment; it != lastColorAttachment; ++it) { QVkTexture *texD = QRHI_RES(QVkTexture, it->texture()); QVkRenderBuffer *rbD = QRHI_RES(QVkRenderBuffer, it->renderBuffer()); Q_ASSERT(texD || rbD); - const VkFormat vkformat = texD ? texD->vkformat : rbD->vkformat; + const VkFormat vkformat = texD ? texD->viewFormat : rbD->vkformat; const VkSampleCountFlagBits samples = texD ? texD->samples : rbD->samples; - VkAttachmentDescription attDesc; - memset(&attDesc, 0, sizeof(attDesc)); + VkAttachmentDescription attDesc = {}; attDesc.format = vkformat; attDesc.samples = samples; attDesc.loadOp = preserveColor ? VK_ATTACHMENT_LOAD_OP_LOAD : VK_ATTACHMENT_LOAD_OP_CLEAR; - attDesc.storeOp = it->resolveTexture() ? VK_ATTACHMENT_STORE_OP_DONT_CARE : VK_ATTACHMENT_STORE_OP_STORE; + attDesc.storeOp = (it->resolveTexture() && !preserveColor) ? VK_ATTACHMENT_STORE_OP_DONT_CARE : VK_ATTACHMENT_STORE_OP_STORE; attDesc.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; attDesc.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; // this has to interact correctly with activateTextureRenderTarget(), hence leaving in COLOR_ATT @@ -1319,31 +1629,41 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, attDesc.finalLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; rpD->attDescs.append(attDesc); - const VkAttachmentReference ref = { uint32_t(rpD->attDescs.count() - 1), VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; + const VkAttachmentReference ref = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; rpD->colorRefs.append(ref); + + if (it->multiViewCount() >= 2) { + if (multiViewCount > 0 && multiViewCount != it->multiViewCount()) + qWarning("Inconsistent multiViewCount in color attachment set"); + else + multiViewCount = it->multiViewCount(); + } else if (multiViewCount > 0) { + qWarning("Mixing non-multiview color attachments within a multiview render pass"); + } } + Q_ASSERT(multiViewCount == 0 || multiViewCount >= 2); + rpD->multiViewCount = uint32_t(multiViewCount); rpD->hasDepthStencil = depthStencilBuffer || depthTexture; if (rpD->hasDepthStencil) { - const VkFormat dsFormat = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->vkformat + const VkFormat dsFormat = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->viewFormat : QRHI_RES(QVkRenderBuffer, depthStencilBuffer)->vkformat; const VkSampleCountFlagBits samples = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->samples : QRHI_RES(QVkRenderBuffer, depthStencilBuffer)->samples; const VkAttachmentLoadOp loadOp = preserveDs ? VK_ATTACHMENT_LOAD_OP_LOAD : VK_ATTACHMENT_LOAD_OP_CLEAR; - const VkAttachmentStoreOp storeOp = depthTexture ? VK_ATTACHMENT_STORE_OP_STORE : VK_ATTACHMENT_STORE_OP_DONT_CARE; - VkAttachmentDescription attDesc; - memset(&attDesc, 0, sizeof(attDesc)); + const VkAttachmentStoreOp storeOp = storeDs ? VK_ATTACHMENT_STORE_OP_STORE : VK_ATTACHMENT_STORE_OP_DONT_CARE; + VkAttachmentDescription attDesc = {}; attDesc.format = dsFormat; attDesc.samples = samples; attDesc.loadOp = loadOp; attDesc.storeOp = storeOp; attDesc.stencilLoadOp = loadOp; attDesc.stencilStoreOp = storeOp; - attDesc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + attDesc.initialLayout = preserveDs ? VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL : VK_IMAGE_LAYOUT_UNDEFINED; attDesc.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; rpD->attDescs.append(attDesc); } - rpD->dsRef = { uint32_t(rpD->attDescs.count() - 1), VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL }; + rpD->dsRef = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL }; for (auto it = firstColorAttachment; it != lastColorAttachment; ++it) { if (it->resolveTexture()) { @@ -1363,9 +1683,8 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, int(srcFormat), int(dstFormat)); } - VkAttachmentDescription attDesc; - memset(&attDesc, 0, sizeof(attDesc)); - attDesc.format = dstFormat; + VkAttachmentDescription attDesc = {}; + attDesc.format = rtexD->viewFormat; attDesc.samples = VK_SAMPLE_COUNT_1_BIT; attDesc.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; // ignored attDesc.storeOp = VK_ATTACHMENT_STORE_OP_STORE; @@ -1375,14 +1694,39 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, attDesc.finalLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; rpD->attDescs.append(attDesc); - const VkAttachmentReference ref = { uint32_t(rpD->attDescs.count() - 1), VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; + const VkAttachmentReference ref = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; rpD->resolveRefs.append(ref); } else { const VkAttachmentReference ref = { VK_ATTACHMENT_UNUSED, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; rpD->resolveRefs.append(ref); } } - Q_ASSERT(rpD->colorRefs.count() == rpD->resolveRefs.count()); + Q_ASSERT(rpD->colorRefs.size() == rpD->resolveRefs.size()); + + rpD->hasDepthStencilResolve = rpD->hasDepthStencil && depthResolveTexture; + if (rpD->hasDepthStencilResolve) { + QVkTexture *rtexD = QRHI_RES(QVkTexture, depthResolveTexture); + if (rtexD->samples > VK_SAMPLE_COUNT_1_BIT) + qWarning("Resolving into a multisample depth texture is not supported"); + + QVkTexture *texD = QRHI_RES(QVkTexture, depthResolveTexture); + if (texD->vkformat != rtexD->vkformat) { + qWarning("Multisample resolve between different depth-stencil formats (%d and %d) is not supported.", + int(texD->vkformat), int(rtexD->vkformat)); + } + + VkAttachmentDescription attDesc = {}; + attDesc.format = rtexD->viewFormat; + attDesc.samples = VK_SAMPLE_COUNT_1_BIT; + attDesc.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; // ignored + attDesc.storeOp = VK_ATTACHMENT_STORE_OP_STORE; + attDesc.stencilLoadOp = attDesc.loadOp; + attDesc.stencilStoreOp = attDesc.storeOp; + attDesc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + attDesc.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + rpD->attDescs.append(attDesc); + } + rpD->dsResolveRef = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL }; // rpD->subpassDeps stays empty: don't yet know the correct initial/final // access and stage stuff for the implicit deps at this point, so leave it @@ -1393,10 +1737,35 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, VkSubpassDescription subpassDesc; fillRenderPassCreateInfo(&rpInfo, &subpassDesc, rpD); - VkResult err = df->vkCreateRenderPass(dev, &rpInfo, nullptr, &rpD->rp); - if (err != VK_SUCCESS) { - qWarning("Failed to create renderpass: %d", err); + MultiViewRenderPassSetupHelper multiViewHelper; + if (!multiViewHelper.prepare(&rpInfo, multiViewCount, caps.multiView)) return false; + +#ifdef VK_KHR_create_renderpass2 + if (rpD->hasDepthStencilResolve && caps.renderPass2KHR) { + // Use the KHR extension, not the 1.2 core API, in order to support Vulkan 1.1. + VkRenderPassCreateInfo2KHR rpInfo2; + RenderPass2SetupHelper rp2Helper; + if (!rp2Helper.prepare(&rpInfo2, &rpInfo, rpD, multiViewCount)) + return false; + + VkResult err = vkCreateRenderPass2KHR(dev, &rpInfo2, nullptr, &rpD->rp); + if (err != VK_SUCCESS) { + qWarning("Failed to create renderpass (using VkRenderPassCreateInfo2KHR): %d", err); + return false; + } + } else +#endif + { + if (rpD->hasDepthStencilResolve) { + qWarning("Resolving multisample depth-stencil buffers is not supported without " + "VK_KHR_depth_stencil_resolve and VK_KHR_create_renderpass2"); + } + VkResult err = df->vkCreateRenderPass(dev, &rpInfo, nullptr, &rpD->rp); + if (err != VK_SUCCESS) { + qWarning("Failed to create renderpass: %d", err); + return false; + } } return true; @@ -1437,21 +1806,45 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) ? VK_SURFACE_TRANSFORM_IDENTITY_BIT_KHR : surfaceCaps.currentTransform; + // This looks odd but matches how platforms work in practice. + // + // On Windows with NVIDIA for example, the only supportedCompositeAlpha + // value reported is OPAQUE, nothing else. Yet transparency works + // regardless, as long as the native window is set up correctly, so that's + // not something we need to handle here. + // + // On Linux with Intel and Mesa and running on xcb reports, on one + // particular system, INHERIT+PRE_MULTIPLIED. Tranparency works, regardless, + // presumably due to setting INHERIT. + // + // On the same setup with Wayland instead of xcb we see + // OPAQUE+PRE_MULTIPLIED reported. Here transparency won't work unless + // PRE_MULTIPLIED is set. + // + // Therefore our rules are: + // - Prefer INHERIT over OPAQUE. + // - Then based on the request, try the requested alpha mode, but if + // that's not reported as supported, try also the other (PRE/POST, + // POST/PRE) as that is better than nothing. This is not different from + // some other backends, e.g. D3D11 with DirectComposition there is also + // no control over being straight or pre-multiplied. Whereas with + // WGL/GLX/EGL we never had that sort of control. + VkCompositeAlphaFlagBitsKHR compositeAlpha = (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR) ? VK_COMPOSITE_ALPHA_INHERIT_BIT_KHR : VK_COMPOSITE_ALPHA_OPAQUE_BIT_KHR; - if (swapChainD->m_flags.testFlag(QRhiSwapChain::SurfaceHasPreMulAlpha) - && (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR)) - { - compositeAlpha = VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR; - } - - if (swapChainD->m_flags.testFlag(QRhiSwapChain::SurfaceHasNonPreMulAlpha) - && (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR)) - { - compositeAlpha = VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR; + if (swapChainD->m_flags.testFlag(QRhiSwapChain::SurfaceHasPreMulAlpha)) { + if (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR) + compositeAlpha = VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR; + else if (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR) + compositeAlpha = VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR; + } else if (swapChainD->m_flags.testFlag(QRhiSwapChain::SurfaceHasNonPreMulAlpha)) { + if (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR) + compositeAlpha = VK_COMPOSITE_ALPHA_POST_MULTIPLIED_BIT_KHR; + else if (surfaceCaps.supportedCompositeAlpha & VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR) + compositeAlpha = VK_COMPOSITE_ALPHA_PRE_MULTIPLIED_BIT_KHR; } VkImageUsageFlags usage = VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT; @@ -1459,9 +1852,16 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) if (swapChainD->supportsReadback && swapChainD->m_flags.testFlag(QRhiSwapChain::UsedAsTransferSource)) usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT; + const bool stereo = bool(swapChainD->m_window) && (swapChainD->m_window->format().stereo()) + && surfaceCaps.maxImageArrayLayers > 1; + swapChainD->stereo = stereo; + VkPresentModeKHR presentMode = VK_PRESENT_MODE_FIFO_KHR; if (swapChainD->m_flags.testFlag(QRhiSwapChain::NoVSync)) { - if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_MAILBOX_KHR)) + // Stereo has a weird bug, when using VK_PRESENT_MODE_MAILBOX_KHR, + // black screen is shown, but there is no validation error. + // Detected on Windows, with NVidia RTX A series (at least 4000 and 6000) driver 535.98 + if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_MAILBOX_KHR) && !stereo) presentMode = VK_PRESENT_MODE_MAILBOX_KHR; else if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_IMMEDIATE_KHR)) presentMode = VK_PRESENT_MODE_IMMEDIATE_KHR; @@ -1478,15 +1878,14 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) reuseExisting ? "recycled" : "new", reqBufferCount, swapChainD->pixelSize.width(), swapChainD->pixelSize.height(), presentMode); - VkSwapchainCreateInfoKHR swapChainInfo; - memset(&swapChainInfo, 0, sizeof(swapChainInfo)); + VkSwapchainCreateInfoKHR swapChainInfo = {}; swapChainInfo.sType = VK_STRUCTURE_TYPE_SWAPCHAIN_CREATE_INFO_KHR; swapChainInfo.surface = swapChainD->surface; swapChainInfo.minImageCount = reqBufferCount; swapChainInfo.imageFormat = swapChainD->colorFormat; swapChainInfo.imageColorSpace = swapChainD->colorSpace; swapChainInfo.imageExtent = VkExtent2D { uint32_t(swapChainD->pixelSize.width()), uint32_t(swapChainD->pixelSize.height()) }; - swapChainInfo.imageArrayLayers = 1; + swapChainInfo.imageArrayLayers = stereo ? 2u : 1u; swapChainInfo.imageUsage = usage; swapChainInfo.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; swapChainInfo.preTransform = preTransform; @@ -1544,12 +1943,13 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) } } - VkFenceCreateInfo fenceInfo; - memset(&fenceInfo, 0, sizeof(fenceInfo)); + VkFenceCreateInfo fenceInfo = {}; fenceInfo.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; fenceInfo.flags = VK_FENCE_CREATE_SIGNALED_BIT; - swapChainD->imageRes.resize(swapChainD->bufferCount); + // Double up for stereo + swapChainD->imageRes.resize(swapChainD->bufferCount * (stereo ? 2u : 1u)); + for (int i = 0; i < swapChainD->bufferCount; ++i) { QVkSwapChain::ImageResources &image(swapChainD->imageRes[i]); image.image = swapChainImages[i]; @@ -1558,8 +1958,7 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) image.msaaImageView = msaaViews[i]; } - VkImageViewCreateInfo imgViewInfo; - memset(&imgViewInfo, 0, sizeof(imgViewInfo)); + VkImageViewCreateInfo imgViewInfo = {}; imgViewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; imgViewInfo.image = swapChainImages[i]; imgViewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; @@ -1578,11 +1977,40 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) image.lastUse = QVkSwapChain::ImageResources::ScImageUseNone; } + if (stereo) { + for (int i = 0; i < swapChainD->bufferCount; ++i) { + QVkSwapChain::ImageResources &image(swapChainD->imageRes[i + swapChainD->bufferCount]); + image.image = swapChainImages[i]; + if (swapChainD->samples > VK_SAMPLE_COUNT_1_BIT) { + image.msaaImage = msaaImages[i]; + image.msaaImageView = msaaViews[i]; + } + + VkImageViewCreateInfo imgViewInfo = {}; + imgViewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + imgViewInfo.image = swapChainImages[i]; + imgViewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; + imgViewInfo.format = swapChainD->colorFormat; + imgViewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + imgViewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + imgViewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + imgViewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + imgViewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + imgViewInfo.subresourceRange.baseArrayLayer = 1; + imgViewInfo.subresourceRange.levelCount = imgViewInfo.subresourceRange.layerCount = 1; + err = df->vkCreateImageView(dev, &imgViewInfo, nullptr, &image.imageView); + if (err != VK_SUCCESS) { + qWarning("Failed to create swapchain image view %d: %d", i, err); + return false; + } + + image.lastUse = QVkSwapChain::ImageResources::ScImageUseNone; + } + } swapChainD->currentImageIndex = 0; - VkSemaphoreCreateInfo semInfo; - memset(&semInfo, 0, sizeof(semInfo)); + VkSemaphoreCreateInfo semInfo = {}; semInfo.sType = VK_STRUCTURE_TYPE_SEMAPHORE_CREATE_INFO; for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { @@ -1646,7 +2074,7 @@ void QRhiVulkan::releaseSwapChainResources(QRhiSwapChain *swapChain) } } - for (int i = 0; i < swapChainD->bufferCount; ++i) { + for (int i = 0; i < swapChainD->bufferCount * (swapChainD->stereo ? 2 : 1); ++i) { QVkSwapChain::ImageResources &image(swapChainD->imageRes[i]); if (image.fb) { df->vkDestroyFramebuffer(dev, image.fb, nullptr); @@ -1688,7 +2116,25 @@ void QRhiVulkan::ensureCommandPoolForNewFrame() flags |= VK_COMMAND_POOL_RESET_RELEASE_RESOURCES_BIT; // put all command buffers allocated from this slot's pool to initial state - df->vkResetCommandPool(dev, cmdPool[currentFrameSlot], 0); + df->vkResetCommandPool(dev, cmdPool[currentFrameSlot], flags); +} + +double QRhiVulkan::elapsedSecondsFromTimestamp(quint64 timestamp[2], bool *ok) +{ + quint64 mask = 0; + for (quint64 i = 0; i < timestampValidBits; i += 8) + mask |= 0xFFULL << i; + const quint64 ts0 = timestamp[0] & mask; + const quint64 ts1 = timestamp[1] & mask; + const float nsecsPerTick = physDevProperties.limits.timestampPeriod; + if (!qFuzzyIsNull(nsecsPerTick)) { + const float elapsedMs = float(ts1 - ts0) * nsecsPerTick / 1000000.0f; + const double elapsedSec = elapsedMs / 1000.0; + *ok = true; + return elapsedSec; + } + *ok = false; + return 0; } QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags) @@ -1696,7 +2142,8 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin QVkSwapChain *swapChainD = QRHI_RES(QVkSwapChain, swapChain); const int frameResIndex = swapChainD->bufferCount > 1 ? swapChainD->currentFrameSlot : 0; QVkSwapChain::FrameResources &frame(swapChainD->frameRes[frameResIndex]); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); + + inst->handle()->beginFrame(swapChainD->window); if (!frame.imageAcquired) { // Wait if we are too far ahead, i.e. the thread gets throttled based on the presentation rate @@ -1741,32 +2188,6 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin // mess up A's in-flight commands (as they are not in flight anymore). waitCommandCompletion(frameResIndex); - // Now is the time to read the timestamps for the previous frame for this slot. - if (frame.timestampQueryIndex >= 0) { - quint64 timestamp[2] = { 0, 0 }; - VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(frame.timestampQueryIndex), 2, - 2 * sizeof(quint64), timestamp, sizeof(quint64), - VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); - timestampQueryPoolMap.clearBit(frame.timestampQueryIndex / 2); - frame.timestampQueryIndex = -1; - if (err == VK_SUCCESS) { - quint64 mask = 0; - for (quint64 i = 0; i < timestampValidBits; i += 8) - mask |= 0xFFULL << i; - const quint64 ts0 = timestamp[0] & mask; - const quint64 ts1 = timestamp[1] & mask; - const float nsecsPerTick = physDevProperties.limits.timestampPeriod; - if (!qFuzzyIsNull(nsecsPerTick)) { - const float elapsedMs = float(ts1 - ts0) * nsecsPerTick / 1000000.0f; - // now we have the gpu time for the previous frame for this slot, report it - // (does not matter that it is not for this frame) - QRHI_PROF_F(swapChainFrameGpuTime(swapChain, elapsedMs)); - } - } else { - qWarning("Failed to query timestamp: %d", err); - } - } - currentFrameSlot = int(swapChainD->currentFrameSlot); currentSwapChain = swapChainD; if (swapChainD->ds) @@ -1780,34 +2201,56 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin if (cbres != QRhi::FrameOpSuccess) return cbres; - // when profiling is enabled, pick a free query (pair) from the pool - int timestampQueryIdx = -1; - if (profilerPrivateOrNull() && swapChainD->bufferCount > 1) { // no timestamps if not having at least 2 frames in flight - for (int i = 0; i < timestampQueryPoolMap.count(); ++i) { - if (!timestampQueryPoolMap.testBit(i)) { - timestampQueryPoolMap.setBit(i); - timestampQueryIdx = i * 2; - break; - } - } - } - if (timestampQueryIdx >= 0) { - df->vkCmdResetQueryPool(frame.cmdBuf, timestampQueryPool, uint32_t(timestampQueryIdx), 2); - // record timestamp at the start of the command buffer - df->vkCmdWriteTimestamp(frame.cmdBuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, - timestampQueryPool, uint32_t(timestampQueryIdx)); - frame.timestampQueryIndex = timestampQueryIdx; - } - swapChainD->cbWrapper.cb = frame.cmdBuf; QVkSwapChain::ImageResources &image(swapChainD->imageRes[swapChainD->currentImageIndex]); swapChainD->rtWrapper.d.fb = image.fb; - QRHI_PROF_F(beginSwapChainFrame(swapChain)); + if (swapChainD->stereo) { + QVkSwapChain::ImageResources &image( + swapChainD->imageRes[swapChainD->currentImageIndex + swapChainD->bufferCount]); + swapChainD->rtWrapperRight.d.fb = image.fb; + } prepareNewFrame(&swapChainD->cbWrapper); + // Read the timestamps for the previous frame for this slot. + if (frame.timestampQueryIndex >= 0) { + quint64 timestamp[2] = { 0, 0 }; + VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(frame.timestampQueryIndex), 2, + 2 * sizeof(quint64), timestamp, sizeof(quint64), + VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); + timestampQueryPoolMap.clearBit(frame.timestampQueryIndex / 2); + frame.timestampQueryIndex = -1; + if (err == VK_SUCCESS) { + bool ok = false; + const double elapsedSec = elapsedSecondsFromTimestamp(timestamp, &ok); + if (ok) + swapChainD->cbWrapper.lastGpuTime = elapsedSec; + } else { + qWarning("Failed to query timestamp: %d", err); + } + } + + // No timestamps if the client did not opt in, or when not having at least 2 frames in flight. + if (rhiFlags.testFlag(QRhi::EnableTimestamps) && swapChainD->bufferCount > 1) { + int timestampQueryIdx = -1; + for (int i = 0; i < timestampQueryPoolMap.size(); ++i) { + if (!timestampQueryPoolMap.testBit(i)) { + timestampQueryPoolMap.setBit(i); + timestampQueryIdx = i * 2; + break; + } + } + if (timestampQueryIdx >= 0) { + df->vkCmdResetQueryPool(frame.cmdBuf, timestampQueryPool, uint32_t(timestampQueryIdx), 2); + // record timestamp at the start of the command buffer + df->vkCmdWriteTimestamp(frame.cmdBuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, + timestampQueryPool, uint32_t(timestampQueryIdx)); + frame.timestampQueryIndex = timestampQueryIdx; + } + } + return QRhi::FrameOpSuccess; } @@ -1816,6 +2259,10 @@ QRhi::FrameOpResult QRhiVulkan::endFrame(QRhiSwapChain *swapChain, QRhi::EndFram QVkSwapChain *swapChainD = QRHI_RES(QVkSwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); + auto cleanup = qScopeGuard([this, swapChainD] { + inst->handle()->endFrame(swapChainD->window); + }); + recordPrimaryCommandBuffer(&swapChainD->cbWrapper); int frameResIndex = swapChainD->bufferCount > 1 ? swapChainD->currentFrameSlot : 0; @@ -1823,8 +2270,7 @@ QRhi::FrameOpResult QRhiVulkan::endFrame(QRhiSwapChain *swapChain, QRhi::EndFram QVkSwapChain::ImageResources &image(swapChainD->imageRes[swapChainD->currentImageIndex]); if (image.lastUse != QVkSwapChain::ImageResources::ScImageUseRender) { - VkImageMemoryBarrier presTrans; - memset(&presTrans, 0, sizeof(presTrans)); + VkImageMemoryBarrier presTrans = {}; presTrans.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; presTrans.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT; presTrans.newLayout = VK_IMAGE_LAYOUT_PRESENT_SRC_KHR; @@ -1871,14 +2317,9 @@ QRhi::FrameOpResult QRhiVulkan::endFrame(QRhiSwapChain *swapChain, QRhi::EndFram frame.imageSemWaitable = false; frame.cmdFenceWaitable = true; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - // this must be done before the Present - QRHI_PROF_F(endSwapChainFrame(swapChain, swapChainD->frameCount + 1)); - if (needsPresent) { // add the Present to the queue - VkPresentInfoKHR presInfo; - memset(&presInfo, 0, sizeof(presInfo)); + VkPresentInfoKHR presInfo = {}; presInfo.sType = VK_STRUCTURE_TYPE_PRESENT_INFO_KHR; presInfo.swapchainCount = 1; presInfo.pSwapchains = &swapChainD->sc; @@ -1931,8 +2372,8 @@ void QRhiVulkan::prepareNewFrame(QRhiCommandBuffer *cb) // // With multiple swapchains on the same QRhi things get more convoluted // (and currentFrameSlot strictly alternating is not true anymore) but - // beginNonWrapperFrame() solves that by blocking as necessary so the rest - // here is safe regardless. + // begin(Offscreen)Frame() blocks anyway waiting for its current frame + // slot's previous commands to complete so this here is safe regardless. executeDeferredReleases(); @@ -1946,8 +2387,7 @@ void QRhiVulkan::prepareNewFrame(QRhiCommandBuffer *cb) QRhi::FrameOpResult QRhiVulkan::startPrimaryCommandBuffer(VkCommandBuffer *cb) { if (!*cb) { - VkCommandBufferAllocateInfo cmdBufInfo; - memset(&cmdBufInfo, 0, sizeof(cmdBufInfo)); + VkCommandBufferAllocateInfo cmdBufInfo = {}; cmdBufInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO; cmdBufInfo.commandPool = cmdPool[currentFrameSlot]; cmdBufInfo.level = VK_COMMAND_BUFFER_LEVEL_PRIMARY; @@ -1965,8 +2405,7 @@ QRhi::FrameOpResult QRhiVulkan::startPrimaryCommandBuffer(VkCommandBuffer *cb) } } - VkCommandBufferBeginInfo cmdBufBeginInfo; - memset(&cmdBufBeginInfo, 0, sizeof(cmdBufBeginInfo)); + VkCommandBufferBeginInfo cmdBufBeginInfo = {}; cmdBufBeginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; VkResult err = df->vkBeginCommandBuffer(*cb, &cmdBufBeginInfo); @@ -1997,8 +2436,7 @@ QRhi::FrameOpResult QRhiVulkan::endAndSubmitPrimaryCommandBuffer(VkCommandBuffer return QRhi::FrameOpError; } - VkSubmitInfo submitInfo; - memset(&submitInfo, 0, sizeof(submitInfo)); + VkSubmitInfo submitInfo = {}; submitInfo.sType = VK_STRUCTURE_TYPE_SUBMIT_INFO; submitInfo.commandBufferCount = 1; submitInfo.pCommandBuffers = &cb; @@ -2029,7 +2467,7 @@ QRhi::FrameOpResult QRhiVulkan::endAndSubmitPrimaryCommandBuffer(VkCommandBuffer void QRhiVulkan::waitCommandCompletion(int frameSlot) { - for (QVkSwapChain *sc : qAsConst(swapchains)) { + for (QVkSwapChain *sc : std::as_const(swapchains)) { const int frameResIndex = sc->bufferCount > 1 ? frameSlot : 0; QVkSwapChain::FrameResources &frame(sc->frameRes[frameResIndex]); if (frame.cmdFenceWaitable) { @@ -2043,17 +2481,17 @@ void QRhiVulkan::waitCommandCompletion(int frameSlot) QRhi::FrameOpResult QRhiVulkan::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags) { // Switch to the next slot manually. Swapchains do not know about this - // which is good. So for example a - unusual but possible - onscreen, - // onscreen, offscreen, onscreen, onscreen, onscreen sequence of - // begin/endFrame leads to 0, 1, 0, 0, 1, 0. This works because the - // offscreen frame is synchronous in the sense that we wait for execution - // to complete in endFrame, and so no resources used in that frame are busy + // which is good. So for example an onscreen, onscreen, offscreen, + // onscreen, onscreen, onscreen sequence of frames leads to 0, 1, 0, 0, 1, + // 0. (no strict alternation anymore) But this is not different from what + // happens when multiple swapchains are involved. Offscreen frames are + // synchronous anyway in the sense that they wait for execution to complete + // in endOffscreenFrame, so no resources used in that frame are busy // anymore in the next frame. + currentFrameSlot = (currentFrameSlot + 1) % QVK_FRAMES_IN_FLIGHT; - // except that this gets complicated with multiple swapchains so make sure - // any pending commands have finished for the frame slot we are going to use - if (swapchains.count() > 1) - waitCommandCompletion(currentFrameSlot); + + waitCommandCompletion(currentFrameSlot); ensureCommandPoolForNewFrame(); @@ -2065,6 +2503,24 @@ QRhi::FrameOpResult QRhiVulkan::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi prepareNewFrame(cbWrapper); ofr.active = true; + if (rhiFlags.testFlag(QRhi::EnableTimestamps)) { + int timestampQueryIdx = -1; + for (int i = 0; i < timestampQueryPoolMap.size(); ++i) { + if (!timestampQueryPoolMap.testBit(i)) { + timestampQueryPoolMap.setBit(i); + timestampQueryIdx = i * 2; + break; + } + } + if (timestampQueryIdx >= 0) { + df->vkCmdResetQueryPool(cbWrapper->cb, timestampQueryPool, uint32_t(timestampQueryIdx), 2); + // record timestamp at the start of the command buffer + df->vkCmdWriteTimestamp(cbWrapper->cb, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, + timestampQueryPool, uint32_t(timestampQueryIdx)); + ofr.timestampQueryIndex = timestampQueryIdx; + } + } + *cb = cbWrapper; return QRhi::FrameOpSuccess; } @@ -2078,9 +2534,14 @@ QRhi::FrameOpResult QRhiVulkan::endOffscreenFrame(QRhi::EndFrameFlags flags) QVkCommandBuffer *cbWrapper(ofr.cbWrapper[currentFrameSlot]); recordPrimaryCommandBuffer(cbWrapper); + // record another timestamp, when enabled + if (ofr.timestampQueryIndex >= 0) { + df->vkCmdWriteTimestamp(cbWrapper->cb, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, + timestampQueryPool, uint32_t(ofr.timestampQueryIndex + 1)); + } + if (!ofr.cmdFence) { - VkFenceCreateInfo fenceInfo; - memset(&fenceInfo, 0, sizeof(fenceInfo)); + VkFenceCreateInfo fenceInfo = {}; fenceInfo.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; VkResult err = df->vkCreateFence(dev, &fenceInfo, nullptr, &ofr.cmdFence); if (err != VK_SUCCESS) { @@ -2101,6 +2562,24 @@ QRhi::FrameOpResult QRhiVulkan::endOffscreenFrame(QRhi::EndFrameFlags flags) // previous) frame is safe since we waited for completion above. finishActiveReadbacks(true); + // Read the timestamps, if we wrote them. + if (ofr.timestampQueryIndex >= 0) { + quint64 timestamp[2] = { 0, 0 }; + VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(ofr.timestampQueryIndex), 2, + 2 * sizeof(quint64), timestamp, sizeof(quint64), + VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); + timestampQueryPoolMap.clearBit(ofr.timestampQueryIndex / 2); + ofr.timestampQueryIndex = -1; + if (err == VK_SUCCESS) { + bool ok = false; + const double elapsedSec = elapsedSecondsFromTimestamp(timestamp, &ok); + if (ok) + cbWrapper->lastGpuTime = elapsedSec; + } else { + qWarning("Failed to query timestamp: %d", err); + } + } + return QRhi::FrameOpSuccess; } @@ -2172,6 +2651,9 @@ static inline QRhiPassResourceTracker::UsageState toPassTrackerUsageState(const void QRhiVulkan::activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRenderTarget *rtD) { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QVkTexture, QVkRenderBuffer>(rtD->description(), rtD->d.currentResIdList)) + rtD->create(); + rtD->lastActiveFrameSlot = currentFrameSlot; rtD->d.rp->lastActiveFrameSlot = currentFrameSlot; QRhiPassResourceTracker &passResTracker(cbD->passResTrackers[cbD->currentPassResTrackerIndex]); @@ -2215,6 +2697,13 @@ void QRhiVulkan::activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRe QRhiPassResourceTracker::TexDepthOutputStage); depthTexD->lastActiveFrameSlot = currentFrameSlot; } + if (rtD->m_desc.depthResolveTexture()) { + QVkTexture *depthResolveTexD = QRHI_RES(QVkTexture, rtD->m_desc.depthResolveTexture()); + trackedRegisterTexture(&passResTracker, depthResolveTexD, + QRhiPassResourceTracker::TexDepthOutput, + QRhiPassResourceTracker::TexDepthOutputStage); + depthResolveTexD->lastActiveFrameSlot = currentFrameSlot; + } } void QRhiVulkan::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -2233,8 +2722,7 @@ VkCommandBuffer QRhiVulkan::startSecondaryCommandBuffer(QVkRenderTargetData *rtD secondaryCb = freeSecondaryCbs[currentFrameSlot].last(); freeSecondaryCbs[currentFrameSlot].removeLast(); } else { - VkCommandBufferAllocateInfo cmdBufInfo; - memset(&cmdBufInfo, 0, sizeof(cmdBufInfo)); + VkCommandBufferAllocateInfo cmdBufInfo = {}; cmdBufInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_ALLOCATE_INFO; cmdBufInfo.commandPool = cmdPool[currentFrameSlot]; cmdBufInfo.level = VK_COMMAND_BUFFER_LEVEL_SECONDARY; @@ -2247,12 +2735,10 @@ VkCommandBuffer QRhiVulkan::startSecondaryCommandBuffer(QVkRenderTargetData *rtD } } - VkCommandBufferBeginInfo cmdBufBeginInfo; - memset(&cmdBufBeginInfo, 0, sizeof(cmdBufBeginInfo)); + VkCommandBufferBeginInfo cmdBufBeginInfo = {}; cmdBufBeginInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO; cmdBufBeginInfo.flags = rtD ? VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT : 0; - VkCommandBufferInheritanceInfo cmdBufInheritInfo; - memset(&cmdBufInheritInfo, 0, sizeof(cmdBufInheritInfo)); + VkCommandBufferInheritanceInfo cmdBufInheritInfo = {}; cmdBufInheritInfo.sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_INHERITANCE_INFO; cmdBufInheritInfo.subpass = 0; if (rtD) { @@ -2308,8 +2794,8 @@ void QRhiVulkan::beginPass(QRhiCommandBuffer *cb, QVkRenderTargetData *rtD = nullptr; switch (rt->resourceType()) { - case QRhiResource::RenderTarget: - rtD = &QRHI_RES(QVkReferenceRenderTarget, rt)->d; + case QRhiResource::SwapChainRenderTarget: + rtD = &QRHI_RES(QVkSwapChainRenderTarget, rt)->d; rtD->rp->lastActiveFrameSlot = currentFrameSlot; Q_ASSERT(currentSwapChain); currentSwapChain->imageRes[currentSwapChain->currentImageIndex].lastUse = @@ -2334,8 +2820,7 @@ void QRhiVulkan::beginPass(QRhiCommandBuffer *cb, // No copy operations or image layout transitions allowed after this point // (up until endPass) as we are going to begin the renderpass. - VkRenderPassBeginInfo rpBeginInfo; - memset(&rpBeginInfo, 0, sizeof(rpBeginInfo)); + VkRenderPassBeginInfo rpBeginInfo = {}; rpBeginInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO; rpBeginInfo.renderPass = rtD->rp->rp; rpBeginInfo.framebuffer = rtD->fb; @@ -2360,17 +2845,24 @@ void QRhiVulkan::beginPass(QRhiCommandBuffer *cb, float(colorClearValue.alphaF()) } }; cvs.append(cv); } - rpBeginInfo.clearValueCount = uint32_t(cvs.count()); + for (int i = 0; i < rtD->dsResolveAttCount; ++i) { + VkClearValue cv; + cv.depthStencil = { depthStencilClearValue.depthClearValue(), depthStencilClearValue.stencilClearValue() }; + cvs.append(cv); + } + rpBeginInfo.clearValueCount = uint32_t(cvs.size()); QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::BeginRenderPass; cmd.args.beginRenderPass.desc = rpBeginInfo; - cmd.args.beginRenderPass.clearValueIndex = cbD->pools.clearValue.count(); + cmd.args.beginRenderPass.clearValueIndex = cbD->pools.clearValue.size(); cmd.args.beginRenderPass.useSecondaryCb = cbD->passUsesSecondaryCb; - cbD->pools.clearValue.append(cvs.constData(), cvs.count()); + cbD->pools.clearValue.append(cvs.constData(), cvs.size()); if (cbD->passUsesSecondaryCb) cbD->activeSecondaryCbStack.append(startSecondaryCommandBuffer(rtD)); + + cbD->resetCachedState(); } void QRhiVulkan::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -2382,7 +2874,6 @@ void QRhiVulkan::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourc VkCommandBuffer secondaryCb = cbD->activeSecondaryCbStack.last(); cbD->activeSecondaryCbStack.removeLast(); endAndEnqueueSecondaryCommandBuffer(secondaryCb, cbD); - cbD->resetCachedState(); } QVkCommandBuffer::Command &cmd(cbD->commands.get()); @@ -2414,6 +2905,8 @@ void QRhiVulkan::beginComputePass(QRhiCommandBuffer *cb, if (cbD->passUsesSecondaryCb) cbD->activeSecondaryCbStack.append(startSecondaryCommandBuffer()); + + cbD->resetCachedState(); } void QRhiVulkan::endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -2425,7 +2918,6 @@ void QRhiVulkan::endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch * VkCommandBuffer secondaryCb = cbD->activeSecondaryCbStack.last(); cbD->activeSecondaryCbStack.removeLast(); endAndEnqueueSecondaryCommandBuffer(secondaryCb, cbD); - cbD->resetCachedState(); } cbD->recordingPass = QVkCommandBuffer::NoPass; @@ -2496,9 +2988,9 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) accessAndIsNewFlag = { 0, false }; QVkShaderResourceBindings *srbD = QRHI_RES(QVkShaderResourceBindings, cbD->currentComputeSrb); - const int bindingCount = srbD->m_bindings.count(); + const int bindingCount = srbD->m_bindings.size(); for (int i = 0; i < bindingCount; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->m_bindings.at(i).data(); + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings.at(i)); switch (b->type) { case QRhiShaderResourceBinding::ImageLoad: case QRhiShaderResourceBinding::ImageStore: @@ -2531,8 +3023,7 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) if (accessInThisDispatch && !isNewInThisDispatch) { if (it.key()->resourceType() == QRhiResource::Texture) { QVkTexture *texD = QRHI_RES(QVkTexture, it.key()); - VkImageMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkImageMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; // won't care about subresources, pretend the whole resource was written @@ -2548,8 +3039,7 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) imageBarriers.append(barrier); } else { QVkBuffer *bufD = QRHI_RES(QVkBuffer, it.key()); - VkBufferMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkBufferMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; barrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; barrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; @@ -2576,12 +3066,12 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) df->vkCmdPipelineBarrier(secondaryCb, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, 0, nullptr, - imageBarriers.count(), imageBarriers.constData()); + imageBarriers.size(), imageBarriers.constData()); } if (!bufferBarriers.isEmpty()) { df->vkCmdPipelineBarrier(secondaryCb, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT, 0, 0, nullptr, - bufferBarriers.count(), bufferBarriers.constData(), + bufferBarriers.size(), bufferBarriers.constData(), 0, nullptr); } df->vkCmdDispatch(secondaryCb, uint32_t(x), uint32_t(y), uint32_t(z)); @@ -2591,18 +3081,18 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) cmd.cmd = QVkCommandBuffer::Command::ImageBarrier; cmd.args.imageBarrier.srcStageMask = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; cmd.args.imageBarrier.dstStageMask = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - cmd.args.imageBarrier.count = imageBarriers.count(); - cmd.args.imageBarrier.index = cbD->pools.imageBarrier.count(); - cbD->pools.imageBarrier.append(imageBarriers.constData(), imageBarriers.count()); + cmd.args.imageBarrier.count = imageBarriers.size(); + cmd.args.imageBarrier.index = cbD->pools.imageBarrier.size(); + cbD->pools.imageBarrier.append(imageBarriers.constData(), imageBarriers.size()); } if (!bufferBarriers.isEmpty()) { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::BufferBarrier; cmd.args.bufferBarrier.srcStageMask = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; cmd.args.bufferBarrier.dstStageMask = VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; - cmd.args.bufferBarrier.count = bufferBarriers.count(); - cmd.args.bufferBarrier.index = cbD->pools.bufferBarrier.count(); - cbD->pools.bufferBarrier.append(bufferBarriers.constData(), bufferBarriers.count()); + cmd.args.bufferBarrier.count = bufferBarriers.size(); + cmd.args.bufferBarrier.index = cbD->pools.bufferBarrier.size(); + cbD->pools.bufferBarrier.append(bufferBarriers.constData(), bufferBarriers.size()); } QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::Dispatch; @@ -2614,8 +3104,7 @@ void QRhiVulkan::dispatch(QRhiCommandBuffer *cb, int x, int y, int z) VkShaderModule QRhiVulkan::createShader(const QByteArray &spirv) { - VkShaderModuleCreateInfo shaderInfo; - memset(&shaderInfo, 0, sizeof(shaderInfo)); + VkShaderModuleCreateInfo shaderInfo = {}; shaderInfo.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO; shaderInfo.codeSize = size_t(spirv.size()); shaderInfo.pCode = reinterpret_cast<const quint32 *>(spirv.constData()); @@ -2633,8 +3122,7 @@ bool QRhiVulkan::ensurePipelineCache(const void *initialData, size_t initialData if (pipelineCache) return true; - VkPipelineCacheCreateInfo pipelineCacheInfo; - memset(&pipelineCacheInfo, 0, sizeof(pipelineCacheInfo)); + VkPipelineCacheCreateInfo pipelineCacheInfo = {}; pipelineCacheInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO; pipelineCacheInfo.initialDataSize = initialDataSize; pipelineCacheInfo.pInitialData = initialData; @@ -2659,12 +3147,11 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i const bool updateAll = descSetIdx < 0; int frameSlot = updateAll ? 0 : descSetIdx; while (frameSlot < (updateAll ? QVK_FRAMES_IN_FLIGHT : descSetIdx + 1)) { - for (int i = 0, ie = srbD->sortedBindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->sortedBindings.at(i).data(); + for (int i = 0, ie = srbD->sortedBindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings.at(i)); QVkShaderResourceBindings::BoundResourceData &bd(srbD->boundResourceData[frameSlot][i]); - VkWriteDescriptorSet writeInfo; - memset(&writeInfo, 0, sizeof(writeInfo)); + VkWriteDescriptorSet writeInfo = {}; writeInfo.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET; writeInfo.dstSet = srbD->descSets[frameSlot]; writeInfo.dstBinding = uint32_t(b->binding); @@ -2684,17 +3171,17 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i bd.ubuf.generation = bufD->generation; VkDescriptorBufferInfo bufInfo; bufInfo.buffer = bufD->m_type == QRhiBuffer::Dynamic ? bufD->buffers[frameSlot] : bufD->buffers[0]; - bufInfo.offset = VkDeviceSize(b->u.ubuf.offset); - bufInfo.range = VkDeviceSize(b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size); + bufInfo.offset = b->u.ubuf.offset; + bufInfo.range = b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size; // be nice and assert when we know the vulkan device would die a horrible death due to non-aligned reads Q_ASSERT(aligned(bufInfo.offset, ubufAlign) == bufInfo.offset); - bufferInfoIndex = bufferInfos.count(); + bufferInfoIndex = bufferInfos.size(); bufferInfos.append(bufInfo); } break; case QRhiShaderResourceBinding::SampledTexture: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; writeInfo.descriptorCount = data->count; // arrays of combined image samplers are supported writeInfo.descriptorType = VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; ArrayOfImageDesc imageInfo(data->count); @@ -2710,7 +3197,44 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i imageInfo[elem].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; } bd.stex.count = data->count; - imageInfoIndex = imageInfos.count(); + imageInfoIndex = imageInfos.size(); + imageInfos.append(imageInfo); + } + break; + case QRhiShaderResourceBinding::Texture: + { + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; + writeInfo.descriptorCount = data->count; // arrays of (separate) images are supported + writeInfo.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE; + ArrayOfImageDesc imageInfo(data->count); + for (int elem = 0; elem < data->count; ++elem) { + QVkTexture *texD = QRHI_RES(QVkTexture, data->texSamplers[elem].tex); + bd.stex.d[elem].texId = texD->m_id; + bd.stex.d[elem].texGeneration = texD->generation; + bd.stex.d[elem].samplerId = 0; + bd.stex.d[elem].samplerGeneration = 0; + imageInfo[elem].sampler = VK_NULL_HANDLE; + imageInfo[elem].imageView = texD->imageView; + imageInfo[elem].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; + } + bd.stex.count = data->count; + imageInfoIndex = imageInfos.size(); + imageInfos.append(imageInfo); + } + break; + case QRhiShaderResourceBinding::Sampler: + { + QVkSampler *samplerD = QRHI_RES(QVkSampler, b->u.stex.texSamplers[0].sampler); + writeInfo.descriptorType = VK_DESCRIPTOR_TYPE_SAMPLER; + bd.stex.d[0].texId = 0; + bd.stex.d[0].texGeneration = 0; + bd.stex.d[0].samplerId = samplerD->m_id; + bd.stex.d[0].samplerGeneration = samplerD->generation; + ArrayOfImageDesc imageInfo(1); + imageInfo[0].sampler = samplerD->sampler; + imageInfo[0].imageView = VK_NULL_HANDLE; + imageInfo[0].imageLayout = VK_IMAGE_LAYOUT_GENERAL; + imageInfoIndex = imageInfos.size(); imageInfos.append(imageInfo); } break; @@ -2719,7 +3243,7 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i case QRhiShaderResourceBinding::ImageLoadStore: { QVkTexture *texD = QRHI_RES(QVkTexture, b->u.simage.tex); - VkImageView view = texD->imageViewForLevel(b->u.simage.level); + VkImageView view = texD->perLevelImageViewForLoadStore(b->u.simage.level); if (view) { writeInfo.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE; bd.simage.id = texD->m_id; @@ -2728,7 +3252,7 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i imageInfo[0].sampler = VK_NULL_HANDLE; imageInfo[0].imageView = view; imageInfo[0].imageLayout = VK_IMAGE_LAYOUT_GENERAL; - imageInfoIndex = imageInfos.count(); + imageInfoIndex = imageInfos.size(); imageInfos.append(imageInfo); } } @@ -2743,9 +3267,9 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i bd.sbuf.generation = bufD->generation; VkDescriptorBufferInfo bufInfo; bufInfo.buffer = bufD->m_type == QRhiBuffer::Dynamic ? bufD->buffers[frameSlot] : bufD->buffers[0]; - bufInfo.offset = VkDeviceSize(b->u.ubuf.offset); - bufInfo.range = VkDeviceSize(b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size); - bufferInfoIndex = bufferInfos.count(); + bufInfo.offset = b->u.ubuf.offset; + bufInfo.range = b->u.ubuf.maybeSize ? b->u.ubuf.maybeSize : bufD->m_size; + bufferInfoIndex = bufferInfos.size(); bufferInfos.append(bufInfo); } break; @@ -2759,7 +3283,7 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i ++frameSlot; } - for (int i = 0, writeInfoCount = writeInfos.count(); i < writeInfoCount; ++i) { + for (int i = 0, writeInfoCount = writeInfos.size(); i < writeInfoCount; ++i) { const int bufferInfoIndex = infoIndices[i].first; const int imageInfoIndex = infoIndices[i].second; if (bufferInfoIndex >= 0) @@ -2768,7 +3292,7 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i writeInfos[i].pImageInfo = imageInfos[imageInfoIndex].constData(); } - df->vkUpdateDescriptorSets(dev, uint32_t(writeInfos.count()), writeInfos.constData(), 0, nullptr); + df->vkUpdateDescriptorSets(dev, uint32_t(writeInfos.size()), writeInfos.constData(), 0, nullptr); } static inline bool accessIsWrite(VkAccessFlags access) @@ -2800,8 +3324,7 @@ void QRhiVulkan::trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, in return; } - VkBufferMemoryBarrier bufMemBarrier; - memset(&bufMemBarrier, 0, sizeof(bufMemBarrier)); + VkBufferMemoryBarrier bufMemBarrier = {}; bufMemBarrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; bufMemBarrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; bufMemBarrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; @@ -2815,7 +3338,7 @@ void QRhiVulkan::trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, in cmd.args.bufferBarrier.srcStageMask = s.stage; cmd.args.bufferBarrier.dstStageMask = stage; cmd.args.bufferBarrier.count = 1; - cmd.args.bufferBarrier.index = cbD->pools.bufferBarrier.count(); + cmd.args.bufferBarrier.index = cbD->pools.bufferBarrier.size(); cbD->pools.bufferBarrier.append(bufMemBarrier); s.access = access; @@ -2833,8 +3356,7 @@ void QRhiVulkan::trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, return; } - VkImageMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkImageMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; barrier.subresourceRange.aspectMask = aspectMaskForTextureFormat(texD->m_format); barrier.subresourceRange.baseMipLevel = 0; @@ -2857,7 +3379,7 @@ void QRhiVulkan::trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, cmd.args.imageBarrier.srcStageMask = srcStage; cmd.args.imageBarrier.dstStageMask = stage; cmd.args.imageBarrier.count = 1; - cmd.args.imageBarrier.index = cbD->pools.imageBarrier.count(); + cmd.args.imageBarrier.index = cbD->pools.imageBarrier.size(); cbD->pools.imageBarrier.append(barrier); s.layout = layout; @@ -2869,8 +3391,7 @@ void QRhiVulkan::depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuf { Q_ASSERT(cbD->recordingPass == QVkCommandBuffer::NoPass); - VkImageMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkImageMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; barrier.subresourceRange.baseMipLevel = 0; @@ -2892,7 +3413,7 @@ void QRhiVulkan::depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuf cmd.args.imageBarrier.srcStageMask = stages; cmd.args.imageBarrier.dstStageMask = stages; cmd.args.imageBarrier.count = 1; - cmd.args.imageBarrier.index = cbD->pools.imageBarrier.count(); + cmd.args.imageBarrier.index = cbD->pools.imageBarrier.size(); cbD->pools.imageBarrier.append(barrier); } @@ -2904,8 +3425,7 @@ void QRhiVulkan::subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, int startLevel, int levelCount) { Q_ASSERT(cbD->recordingPass == QVkCommandBuffer::NoPass); - VkImageMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkImageMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; barrier.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; barrier.subresourceRange.baseMipLevel = uint32_t(startLevel); @@ -2923,7 +3443,7 @@ void QRhiVulkan::subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, cmd.args.imageBarrier.srcStageMask = srcStage; cmd.args.imageBarrier.dstStageMask = dstStage; cmd.args.imageBarrier.count = 1; - cmd.args.imageBarrier.index = cbD->pools.imageBarrier.count(); + cmd.args.imageBarrier.index = cbD->pools.imageBarrier.size(); cbD->pools.imageBarrier.append(barrier); } @@ -2946,9 +3466,9 @@ void QRhiVulkan::prepareUploadSubres(QVkTexture *texD, int layer, int level, qsizetype imageSizeBytes = 0; const void *src = nullptr; const bool is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); + const bool is1D = texD->m_flags.testFlag(QRhiTexture::OneDimensional); - VkBufferImageCopy copyInfo; - memset(©Info, 0, sizeof(copyInfo)); + VkBufferImageCopy copyInfo = {}; copyInfo.bufferOffset = *curOfs; copyInfo.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; copyInfo.imageSubresource.mipLevel = uint32_t(level); @@ -2957,6 +3477,8 @@ void QRhiVulkan::prepareUploadSubres(QVkTexture *texD, int layer, int level, copyInfo.imageExtent.depth = 1; if (is3D) copyInfo.imageOffset.z = uint32_t(layer); + if (is1D) + copyInfo.imageOffset.y = uint32_t(layer); const QByteArray rawData = subresDesc.data(); const QPoint dp = subresDesc.destinationTopLeft(); @@ -3043,10 +3565,15 @@ void QRhiVulkan::prepareUploadSubres(QVkTexture *texD, int layer, int level, } } +void QRhiVulkan::printExtraErrorInfo(VkResult err) +{ + if (err == VK_ERROR_OUT_OF_DEVICE_MEMORY) + qWarning() << "Out of device memory, current allocator statistics are" << statistics(); +} + void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates) { QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); for (int opIdx = 0; opIdx < ud->activeBufferOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::BufferOp &u(ud->bufferOps[opIdx]); @@ -3064,16 +3591,14 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat Q_ASSERT(u.offset + u.data.size() <= bufD->m_size); if (!bufD->stagingBuffers[currentFrameSlot]) { - VkBufferCreateInfo bufferInfo; - memset(&bufferInfo, 0, sizeof(bufferInfo)); + VkBufferCreateInfo bufferInfo = {}; bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; // must cover the entire buffer - this way multiple, partial updates per frame // are supported even when the staging buffer is reused (Static) - bufferInfo.size = VkDeviceSize(bufD->m_size); + bufferInfo.size = bufD->m_size; bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT; - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_CPU_ONLY; VmaAllocation allocation; @@ -3081,9 +3606,9 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat &bufD->stagingBuffers[currentFrameSlot], &allocation, nullptr); if (err == VK_SUCCESS) { bufD->stagingAllocations[currentFrameSlot] = allocation; - QRHI_PROF_F(newBufferStagingArea(bufD, currentFrameSlot, quint32(bufD->m_size))); } else { - qWarning("Failed to create staging buffer of size %d: %d", bufD->m_size, err); + qWarning("Failed to create staging buffer of size %u: %d", bufD->m_size, err); + printExtraErrorInfo(err); continue; } } @@ -3095,18 +3620,17 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat qWarning("Failed to map buffer: %d", err); continue; } - memcpy(static_cast<uchar *>(p) + u.offset, u.data.constData(), size_t(u.data.size())); + memcpy(static_cast<uchar *>(p) + u.offset, u.data.constData(), u.data.size()); + vmaFlushAllocation(toVmaAllocator(allocator), a, u.offset, u.data.size()); vmaUnmapMemory(toVmaAllocator(allocator), a); - vmaFlushAllocation(toVmaAllocator(allocator), a, VkDeviceSize(u.offset), VkDeviceSize(u.data.size())); trackedBufferBarrier(cbD, bufD, 0, VK_ACCESS_TRANSFER_WRITE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT); - VkBufferCopy copyInfo; - memset(©Info, 0, sizeof(copyInfo)); - copyInfo.srcOffset = VkDeviceSize(u.offset); - copyInfo.dstOffset = VkDeviceSize(u.offset); - copyInfo.size = VkDeviceSize(u.data.size()); + VkBufferCopy copyInfo = {}; + copyInfo.srcOffset = u.offset; + copyInfo.dstOffset = u.offset; + copyInfo.size = u.data.size(); QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::CopyBuffer; @@ -3132,7 +3656,6 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat bufD->stagingBuffers[currentFrameSlot] = VK_NULL_HANDLE; bufD->stagingAllocations[currentFrameSlot] = nullptr; releaseQueue.append(e); - QRHI_PROF_F(releaseBufferStagingArea(bufD, currentFrameSlot)); } } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::Read) { QVkBuffer *bufD = QRHI_RES(QVkBuffer, u.buf); @@ -3143,7 +3666,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat VkResult err = vmaMapMemory(toVmaAllocator(allocator), a, &p); if (err == VK_SUCCESS) { u.result->data.resize(u.readSize); - memcpy(u.result->data.data(), reinterpret_cast<char *>(p) + u.offset, size_t(u.readSize)); + memcpy(u.result->data.data(), reinterpret_cast<char *>(p) + u.offset, u.readSize); vmaUnmapMemory(toVmaAllocator(allocator), a); } if (u.result->completed) @@ -3160,32 +3683,29 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat readback.result = u.result; readback.byteSize = u.readSize; - VkBufferCreateInfo bufferInfo; - memset(&bufferInfo, 0, sizeof(bufferInfo)); + VkBufferCreateInfo bufferInfo = {}; bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; - bufferInfo.size = VkDeviceSize(readback.byteSize); + bufferInfo.size = readback.byteSize; bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT; - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_GPU_TO_CPU; VmaAllocation allocation; VkResult err = vmaCreateBuffer(toVmaAllocator(allocator), &bufferInfo, &allocInfo, &readback.stagingBuf, &allocation, nullptr); if (err == VK_SUCCESS) { readback.stagingAlloc = allocation; - QRHI_PROF_F(newReadbackBuffer(qint64(readback.stagingBuf), bufD, uint(readback.byteSize))); } else { qWarning("Failed to create readback buffer of size %u: %d", readback.byteSize, err); + printExtraErrorInfo(err); continue; } trackedBufferBarrier(cbD, bufD, 0, VK_ACCESS_TRANSFER_READ_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT); - VkBufferCopy copyInfo; - memset(©Info, 0, sizeof(copyInfo)); - copyInfo.srcOffset = VkDeviceSize(u.offset); - copyInfo.size = VkDeviceSize(u.readSize); + VkBufferCopy copyInfo = {}; + copyInfo.srcOffset = u.offset; + copyInfo.size = u.readSize; QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::CopyBuffer; @@ -3206,22 +3726,20 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat QVkTexture *utexD = QRHI_RES(QVkTexture, u.dst); // batch into a single staging buffer and a single CopyBufferToImage with multiple copyInfos VkDeviceSize stagingSize = 0; - for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { + for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(u.subresDesc[layer][level])) + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) stagingSize += subresUploadByteSize(subresDesc); } } Q_ASSERT(!utexD->stagingBuffers[currentFrameSlot]); - VkBufferCreateInfo bufferInfo; - memset(&bufferInfo, 0, sizeof(bufferInfo)); + VkBufferCreateInfo bufferInfo = {}; bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; bufferInfo.size = stagingSize; bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT; - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_CPU_TO_GPU; VmaAllocation allocation; @@ -3229,10 +3747,10 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat &utexD->stagingBuffers[currentFrameSlot], &allocation, nullptr); if (err != VK_SUCCESS) { qWarning("Failed to create image staging buffer of size %d: %d", int(stagingSize), err); + printExtraErrorInfo(err); continue; } utexD->stagingAllocations[currentFrameSlot] = allocation; - QRHI_PROF_F(newTextureStagingArea(utexD, currentFrameSlot, quint32(stagingSize))); BufferImageCopyList copyInfos; size_t curOfs = 0; @@ -3244,19 +3762,19 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat continue; } - for (int layer = 0, maxLayer = u.subresDesc.count(); layer < maxLayer; ++layer) { + for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { const QList<QRhiTextureSubresourceUploadDescription> &srd(u.subresDesc[layer][level]); if (srd.isEmpty()) continue; - for (const QRhiTextureSubresourceUploadDescription &subresDesc : qAsConst(srd)) { + for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(srd)) { prepareUploadSubres(utexD, layer, level, subresDesc, &curOfs, mp, ©Infos); } } } - vmaUnmapMemory(toVmaAllocator(allocator), a); vmaFlushAllocation(toVmaAllocator(allocator), a, 0, stagingSize); + vmaUnmapMemory(toVmaAllocator(allocator), a); trackedImageBarrier(cbD, utexD, VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL, VK_ACCESS_TRANSFER_WRITE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT); @@ -3266,9 +3784,9 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat cmd.args.copyBufferToImage.src = utexD->stagingBuffers[currentFrameSlot]; cmd.args.copyBufferToImage.dst = utexD->image; cmd.args.copyBufferToImage.dstLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL; - cmd.args.copyBufferToImage.count = copyInfos.count(); - cmd.args.copyBufferToImage.bufferImageCopyIndex = cbD->pools.bufferImageCopy.count(); - cbD->pools.bufferImageCopy.append(copyInfos.constData(), copyInfos.count()); + cmd.args.copyBufferToImage.count = copyInfos.size(); + cmd.args.copyBufferToImage.bufferImageCopyIndex = cbD->pools.bufferImageCopy.size(); + cbD->pools.bufferImageCopy.append(copyInfos.constData(), copyInfos.size()); // no reuse of staging, this is intentional QRhiVulkan::DeferredReleaseEntry e; @@ -3279,7 +3797,6 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat utexD->stagingBuffers[currentFrameSlot] = VK_NULL_HANDLE; utexD->stagingAllocations[currentFrameSlot] = nullptr; releaseQueue.append(e); - QRHI_PROF_F(releaseTextureStagingArea(utexD, currentFrameSlot)); // Similarly to buffers, transitioning away from DST is done later, // when a renderpass using the texture is encountered. @@ -3296,9 +3813,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat const bool srcIs3D = srcD->m_flags.testFlag(QRhiTexture::ThreeDimensional); const bool dstIs3D = dstD->m_flags.testFlag(QRhiTexture::ThreeDimensional); - VkImageCopy region; - memset(®ion, 0, sizeof(region)); - + VkImageCopy region = {}; region.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; region.srcSubresource.mipLevel = uint32_t(u.desc.sourceLevel()); region.srcSubresource.baseArrayLayer = srcIs3D ? 0 : uint32_t(u.desc.sourceLayer()); @@ -3365,7 +3880,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat continue; } readback.pixelSize = swapChainD->pixelSize; - readback.format = colorTextureFormatFromVkFormat(swapChainD->colorFormat, nullptr); + readback.format = swapchainReadbackTextureFormat(swapChainD->colorFormat, nullptr); if (readback.format == QRhiTexture::UnknownFormat) continue; @@ -3375,31 +3890,26 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat textureFormatInfo(readback.format, readback.pixelSize, nullptr, &readback.byteSize, nullptr); // Create a host visible readback buffer. - VkBufferCreateInfo bufferInfo; - memset(&bufferInfo, 0, sizeof(bufferInfo)); + VkBufferCreateInfo bufferInfo = {}; bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; bufferInfo.size = readback.byteSize; bufferInfo.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT; - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_GPU_TO_CPU; VmaAllocation allocation; VkResult err = vmaCreateBuffer(toVmaAllocator(allocator), &bufferInfo, &allocInfo, &readback.stagingBuf, &allocation, nullptr); if (err == VK_SUCCESS) { readback.stagingAlloc = allocation; - QRHI_PROF_F(newReadbackBuffer(qint64(readback.stagingBuf), - texD ? static_cast<QRhiResource *>(texD) : static_cast<QRhiResource *>(swapChainD), - readback.byteSize)); } else { qWarning("Failed to create readback buffer of size %u: %d", readback.byteSize, err); + printExtraErrorInfo(err); continue; } // Copy from the (optimal and not host visible) image into the buffer. - VkBufferImageCopy copyDesc; - memset(©Desc, 0, sizeof(copyDesc)); + VkBufferImageCopy copyDesc = {}; copyDesc.bufferOffset = 0; copyDesc.imageSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; copyDesc.imageSubresource.mipLevel = uint32_t(u.rb.level()); @@ -3451,6 +3961,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat QVkTexture *utexD = QRHI_RES(QVkTexture, u.dst); Q_ASSERT(utexD->m_flags.testFlag(QRhiTexture::UsedWithGenerateMips)); const bool isCube = utexD->m_flags.testFlag(QRhiTexture::CubeMap); + const bool isArray = utexD->m_flags.testFlag(QRhiTexture::TextureArray); const bool is3D = utexD->m_flags.testFlag(QRhiTexture::ThreeDimensional); VkImageLayout origLayout = utexD->usageState.layout; @@ -3459,10 +3970,10 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat if (!origStage) origStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT; - for (int layer = 0; layer < (isCube ? 6 : 1); ++layer) { + for (int layer = 0; layer < (isCube ? 6 : (isArray ? qMax(0, utexD->m_arraySize) : 1)); ++layer) { int w = utexD->m_pixelSize.width(); int h = utexD->m_pixelSize.height(); - int depth = is3D ? utexD->m_depth : 1; + int depth = is3D ? qMax(1, utexD->m_depth) : 1; for (int level = 1; level < int(utexD->mipLevelCount); ++level) { if (level == 1) { subresourceBarrier(cbD, utexD->image, @@ -3487,9 +3998,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat layer, 1, level, 1); - VkImageBlit region; - memset(®ion, 0, sizeof(region)); - + VkImageBlit region = {}; region.srcSubresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; region.srcSubresource.mipLevel = uint32_t(level) - 1; region.srcSubresource.baseArrayLayer = uint32_t(layer); @@ -3560,18 +4069,18 @@ void QRhiVulkan::executeBufferHostWritesForSlot(QVkBuffer *bufD, int slot) qWarning("Failed to map buffer: %d", err); return; } - int changeBegin = -1; - int changeEnd = -1; - for (const QVkBuffer::DynamicUpdate &u : qAsConst(bufD->pendingDynamicUpdates[slot])) { - memcpy(static_cast<char *>(p) + u.offset, u.data.constData(), size_t(u.data.size())); - if (changeBegin == -1 || u.offset < changeBegin) + quint32 changeBegin = UINT32_MAX; + quint32 changeEnd = 0; + for (const QVkBuffer::DynamicUpdate &u : std::as_const(bufD->pendingDynamicUpdates[slot])) { + memcpy(static_cast<char *>(p) + u.offset, u.data.constData(), u.data.size()); + if (u.offset < changeBegin) changeBegin = u.offset; - if (changeEnd == -1 || u.offset + u.data.size() > changeEnd) + if (u.offset + u.data.size() > changeEnd) changeEnd = u.offset + u.data.size(); } + if (changeBegin < UINT32_MAX && changeBegin < changeEnd) + vmaFlushAllocation(toVmaAllocator(allocator), a, changeBegin, changeEnd - changeBegin); vmaUnmapMemory(toVmaAllocator(allocator), a); - if (changeBegin >= 0) - vmaFlushAllocation(toVmaAllocator(allocator), a, VkDeviceSize(changeBegin), VkDeviceSize(changeEnd - changeBegin)); bufD->pendingDynamicUpdates[slot].clear(); } @@ -3610,7 +4119,7 @@ static void qrhivk_releaseSampler(const QRhiVulkan::DeferredReleaseEntry &e, VkD void QRhiVulkan::executeDeferredReleases(bool forced) { - for (int i = releaseQueue.count() - 1; i >= 0; --i) { + for (int i = releaseQueue.size() - 1; i >= 0; --i) { const QRhiVulkan::DeferredReleaseEntry &e(releaseQueue[i]); if (forced || currentFrameSlot == e.lastActiveFrameSlot || e.lastActiveFrameSlot < 0) { switch (e.type) { @@ -3643,6 +4152,8 @@ void QRhiVulkan::executeDeferredReleases(bool forced) df->vkDestroyImageView(dev, e.textureRenderTarget.rtv[att], nullptr); df->vkDestroyImageView(dev, e.textureRenderTarget.resrtv[att], nullptr); } + df->vkDestroyImageView(dev, e.textureRenderTarget.dsv, nullptr); + df->vkDestroyImageView(dev, e.textureRenderTarget.resdsv, nullptr); break; case QRhiVulkan::DeferredReleaseEntry::RenderPass: df->vkDestroyRenderPass(dev, e.renderPass.rp, nullptr); @@ -3665,9 +4176,8 @@ void QRhiVulkan::executeDeferredReleases(bool forced) void QRhiVulkan::finishActiveReadbacks(bool forced) { QVarLengthArray<std::function<void()>, 4> completedCallbacks; - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - for (int i = activeTextureReadbacks.count() - 1; i >= 0; --i) { + for (int i = activeTextureReadbacks.size() - 1; i >= 0; --i) { const QRhiVulkan::TextureReadback &readback(activeTextureReadbacks[i]); if (forced || currentFrameSlot == readback.activeFrameSlot || readback.activeFrameSlot < 0) { readback.result->format = readback.format; @@ -3684,7 +4194,6 @@ void QRhiVulkan::finishActiveReadbacks(bool forced) } vmaDestroyBuffer(toVmaAllocator(allocator), readback.stagingBuf, a); - QRHI_PROF_F(releaseReadbackBuffer(qint64(readback.stagingBuf))); if (readback.result->completed) completedCallbacks.append(readback.result->completed); @@ -3693,7 +4202,7 @@ void QRhiVulkan::finishActiveReadbacks(bool forced) } } - for (int i = activeBufferReadbacks.count() - 1; i >= 0; --i) { + for (int i = activeBufferReadbacks.size() - 1; i >= 0; --i) { const QRhiVulkan::BufferReadback &readback(activeBufferReadbacks[i]); if (forced || currentFrameSlot == readback.activeFrameSlot || readback.activeFrameSlot < 0) { VmaAllocation a = toVmaAllocation(readback.stagingAlloc); @@ -3701,14 +4210,13 @@ void QRhiVulkan::finishActiveReadbacks(bool forced) VkResult err = vmaMapMemory(toVmaAllocator(allocator), a, &p); if (err == VK_SUCCESS && p) { readback.result->data.resize(readback.byteSize); - memcpy(readback.result->data.data(), p, size_t(readback.byteSize)); + memcpy(readback.result->data.data(), p, readback.byteSize); vmaUnmapMemory(toVmaAllocator(allocator), a); } else { qWarning("Failed to map buffer readback buffer of size %d: %d", readback.byteSize, err); } vmaDestroyBuffer(toVmaAllocator(allocator), readback.stagingBuf, a); - QRHI_PROF_F(releaseReadbackBuffer(qint64(readback.stagingBuf))); if (readback.result->completed) completedCallbacks.append(readback.result->completed); @@ -3755,33 +4263,26 @@ QList<int> QRhiVulkan::supportedSampleCounts() const return result; } -VkSampleCountFlagBits QRhiVulkan::effectiveSampleCount(int sampleCount) +VkSampleCountFlagBits QRhiVulkan::effectiveSampleCountBits(int sampleCount) { - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - sampleCount = qBound(1, sampleCount, 64); - - if (!supportedSampleCounts().contains(sampleCount)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return VK_SAMPLE_COUNT_1_BIT; - } + const int s = effectiveSampleCount(sampleCount); for (const auto &qvk_sampleCount : qvk_sampleCounts) { - if (qvk_sampleCount.count == sampleCount) + if (qvk_sampleCount.count == s) return qvk_sampleCount.mask; } - Q_UNREACHABLE(); - return VK_SAMPLE_COUNT_1_BIT; + Q_UNREACHABLE_RETURN(VK_SAMPLE_COUNT_1_BIT); } void QRhiVulkan::enqueueTransitionPassResources(QVkCommandBuffer *cbD) { cbD->passResTrackers.append(QRhiPassResourceTracker()); - cbD->currentPassResTrackerIndex = cbD->passResTrackers.count() - 1; + cbD->currentPassResTrackerIndex = cbD->passResTrackers.size() - 1; QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::TransitionPassResources; - cmd.args.transitionResources.trackerIndex = cbD->passResTrackers.count() - 1; + cmd.args.transitionResources.trackerIndex = cbD->passResTrackers.size() - 1; } void QRhiVulkan::recordPrimaryCommandBuffer(QVkCommandBuffer *cbD) @@ -3884,17 +4385,23 @@ void QRhiVulkan::recordPrimaryCommandBuffer(QVkCommandBuffer *cbD) cmd.args.drawIndexed.firstInstance); break; case QVkCommandBuffer::Command::DebugMarkerBegin: - cmd.args.debugMarkerBegin.marker.pMarkerName = - cbD->pools.debugMarkerData[cmd.args.debugMarkerBegin.markerNameIndex].constData(); - vkCmdDebugMarkerBegin(cbD->cb, &cmd.args.debugMarkerBegin.marker); +#ifdef VK_EXT_debug_utils + cmd.args.debugMarkerBegin.label.pLabelName = + cbD->pools.debugMarkerData[cmd.args.debugMarkerBegin.labelNameIndex].constData(); + vkCmdBeginDebugUtilsLabelEXT(cbD->cb, &cmd.args.debugMarkerBegin.label); +#endif break; case QVkCommandBuffer::Command::DebugMarkerEnd: - vkCmdDebugMarkerEnd(cbD->cb); +#ifdef VK_EXT_debug_utils + vkCmdEndDebugUtilsLabelEXT(cbD->cb); +#endif break; case QVkCommandBuffer::Command::DebugMarkerInsert: - cmd.args.debugMarkerInsert.marker.pMarkerName = - cbD->pools.debugMarkerData[cmd.args.debugMarkerInsert.markerNameIndex].constData(); - vkCmdDebugMarkerInsert(cbD->cb, &cmd.args.debugMarkerInsert.marker); +#ifdef VK_EXT_debug_utils + cmd.args.debugMarkerInsert.label.pLabelName = + cbD->pools.debugMarkerData[cmd.args.debugMarkerInsert.labelNameIndex].constData(); + vkCmdInsertDebugUtilsLabelEXT(cbD->cb, &cmd.args.debugMarkerInsert.label); +#endif break; case QVkCommandBuffer::Command::TransitionPassResources: recordTransitionPassResources(cbD, cbD->passResTrackers[cmd.args.transitionResources.trackerIndex]); @@ -3940,10 +4447,16 @@ static inline VkPipelineStageFlags toVkPipelineStage(QRhiPassResourceTracker::Bu return VK_PIPELINE_STAGE_VERTEX_INPUT_BIT; case QRhiPassResourceTracker::BufVertexStage: return VK_PIPELINE_STAGE_VERTEX_SHADER_BIT; + case QRhiPassResourceTracker::BufTCStage: + return VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT; + case QRhiPassResourceTracker::BufTEStage: + return VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT; case QRhiPassResourceTracker::BufFragmentStage: return VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; case QRhiPassResourceTracker::BufComputeStage: return VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + case QRhiPassResourceTracker::BufGeometryStage: + return VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT; default: Q_UNREACHABLE(); break; @@ -4006,6 +4519,10 @@ static inline VkPipelineStageFlags toVkPipelineStage(QRhiPassResourceTracker::Te switch (stage) { case QRhiPassResourceTracker::TexVertexStage: return VK_PIPELINE_STAGE_VERTEX_SHADER_BIT; + case QRhiPassResourceTracker::TexTCStage: + return VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT; + case QRhiPassResourceTracker::TexTEStage: + return VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT; case QRhiPassResourceTracker::TexFragmentStage: return VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT; case QRhiPassResourceTracker::TexColorOutputStage: @@ -4014,6 +4531,8 @@ static inline VkPipelineStageFlags toVkPipelineStage(QRhiPassResourceTracker::Te return VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT; case QRhiPassResourceTracker::TexComputeStage: return VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT; + case QRhiPassResourceTracker::TexGeometryStage: + return VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT; default: Q_UNREACHABLE(); break; @@ -4083,8 +4602,7 @@ void QRhiVulkan::recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhi if (!accessIsWrite(access)) continue; } - VkBufferMemoryBarrier bufMemBarrier; - memset(&bufMemBarrier, 0, sizeof(bufMemBarrier)); + VkBufferMemoryBarrier bufMemBarrier = {}; bufMemBarrier.sType = VK_STRUCTURE_TYPE_BUFFER_MEMORY_BARRIER; bufMemBarrier.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; bufMemBarrier.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED; @@ -4108,8 +4626,7 @@ void QRhiVulkan::recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhi if (!accessIsWrite(access)) continue; } - VkImageMemoryBarrier barrier; - memset(&barrier, 0, sizeof(barrier)); + VkImageMemoryBarrier barrier = {}; barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER; barrier.subresourceRange.aspectMask = aspectMaskForTextureFormat(texD->m_format); barrier.subresourceRange.baseMipLevel = 0; @@ -4134,10 +4651,18 @@ void QRhiVulkan::recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhi QRhiSwapChain *QRhiVulkan::createSwapChain() { + if (!vkGetPhysicalDeviceSurfaceCapabilitiesKHR + || !vkGetPhysicalDeviceSurfaceFormatsKHR + || !vkGetPhysicalDeviceSurfacePresentModesKHR) + { + qWarning("Physical device surface queries not available"); + return nullptr; + } + return new QVkSwapChain(this); } -QRhiBuffer *QRhiVulkan::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, int size) +QRhiBuffer *QRhiVulkan::createBuffer(QRhiBuffer::Type type, QRhiBuffer::UsageFlags usage, quint32 size) { return new QVkBuffer(this, type, usage, size); } @@ -4211,7 +4736,7 @@ bool QRhiVulkan::isFeatureSupported(QRhi::Feature feature) const case QRhi::MultisampleRenderBuffer: return true; case QRhi::DebugMarkers: - return caps.debugMarkers; + return caps.debugUtils; case QRhi::Timestamps: return timestampValidBits != 0; case QRhi::Instancing: @@ -4266,9 +4791,34 @@ bool QRhiVulkan::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::RenderTo3DTextureSlice: return caps.texture3DSliceAs2D; + case QRhi::TextureArrays: + return true; + case QRhi::Tessellation: + return caps.tessellation; + case QRhi::GeometryShader: + return caps.geometryShader; + case QRhi::TextureArrayRange: + return true; + case QRhi::NonFillPolygonMode: + return caps.nonFillPolygonMode; + case QRhi::OneDimensionalTextures: + return true; + case QRhi::OneDimensionalTextureMipmaps: + return true; + case QRhi::HalfAttributes: + return true; + case QRhi::RenderToOneDimensionalTexture: + return true; + case QRhi::ThreeDimensionalTextureMipmaps: + return true; + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return true; + case QRhi::ResolveDepthStencil: + return caps.renderPass2KHR && caps.depthStencilResolveKHR; default: - Q_UNREACHABLE(); - return false; + Q_UNREACHABLE_RETURN(false); } } @@ -4297,9 +4847,16 @@ int QRhiVulkan::resourceLimit(QRhi::ResourceLimit limit) const return int(physDevProperties.limits.maxComputeWorkGroupSize[1]); case QRhi::MaxThreadGroupZ: return int(physDevProperties.limits.maxComputeWorkGroupSize[2]); + case QRhi::TextureArraySizeMax: + return int(physDevProperties.limits.maxImageArrayLayers); + case QRhi::MaxUniformBufferRange: + return int(qMin<uint32_t>(INT_MAX, physDevProperties.limits.maxUniformBufferRange)); + case QRhi::MaxVertexInputs: + return physDevProperties.limits.maxVertexInputAttributes; + case QRhi::MaxVertexOutputs: + return physDevProperties.limits.maxVertexOutputComponents / 4; default: - Q_UNREACHABLE(); - return 0; + Q_UNREACHABLE_RETURN(0); } } @@ -4313,16 +4870,24 @@ QRhiDriverInfo QRhiVulkan::driverInfo() const return driverInfoStruct; } -void QRhiVulkan::sendVMemStatsToProfiler() +QRhiStats QRhiVulkan::statistics() { - QRhiProfilerPrivate *rhiP = profilerPrivateOrNull(); - if (!rhiP) - return; + QRhiStats result; + result.totalPipelineCreationTime = totalPipelineCreationTime(); + + VmaBudget budgets[VK_MAX_MEMORY_HEAPS]; + vmaGetHeapBudgets(toVmaAllocator(allocator), budgets); - VmaStats stats; - vmaCalculateStats(toVmaAllocator(allocator), &stats); - QRHI_PROF_F(vmemStat(stats.total.blockCount, stats.total.allocationCount, - quint32(stats.total.usedBytes), quint32(stats.total.unusedBytes))); + uint32_t count = toVmaAllocator(allocator)->GetMemoryHeapCount(); + for (uint32_t i = 0; i < count; ++i) { + const VmaStatistics &stats(budgets[i].statistics); + result.blockCount += stats.blockCount; + result.allocCount += stats.allocationCount; + result.usedBytes += stats.allocationBytes; + result.unusedBytes += stats.blockBytes - stats.allocationBytes; + } + + return result; } bool QRhiVulkan::makeThreadLocalNativeContextCurrent() @@ -4364,7 +4929,7 @@ QByteArray QRhiVulkan::pipelineCacheData() size_t dataSize = 0; VkResult err = df->vkGetPipelineCacheData(dev, pipelineCache, &dataSize, nullptr); if (err != VK_SUCCESS) { - qWarning("Failed to get pipeline cache data size: %d", err); + qCDebug(QRHI_LOG_INFO, "Failed to get pipeline cache data size: %d", err); return QByteArray(); } const size_t headerSize = sizeof(QVkPipelineCacheDataHeader); @@ -4372,7 +4937,7 @@ QByteArray QRhiVulkan::pipelineCacheData() data.resize(dataOffset + dataSize); err = df->vkGetPipelineCacheData(dev, pipelineCache, &dataSize, data.data() + dataOffset); if (err != VK_SUCCESS) { - qWarning("Failed to get pipeline cache data of %d bytes: %d", int(dataSize), err); + qCDebug(QRHI_LOG_INFO, "Failed to get pipeline cache data of %d bytes: %d", int(dataSize), err); return QByteArray(); } @@ -4384,6 +4949,7 @@ QByteArray QRhiVulkan::pipelineCacheData() header.deviceId = physDevProperties.deviceID; header.dataSize = quint32(dataSize); header.uuidSize = VK_UUID_SIZE; + header.reserved = 0; memcpy(data.data(), &header, headerSize); memcpy(data.data() + headerSize, physDevProperties.pipelineCacheUUID, VK_UUID_SIZE); @@ -4397,7 +4963,7 @@ void QRhiVulkan::setPipelineCacheData(const QByteArray &data) const size_t headerSize = sizeof(QVkPipelineCacheDataHeader); if (data.size() < qsizetype(headerSize)) { - qWarning("setPipelineCacheData: Invalid blob size"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob size"); return; } QVkPipelineCacheDataHeader header; @@ -4405,49 +4971,49 @@ void QRhiVulkan::setPipelineCacheData(const QByteArray &data) const quint32 rhiId = pipelineCacheRhiId(); if (header.rhiId != rhiId) { - qWarning("setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", - rhiId, header.rhiId); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: The data is for a different QRhi version or backend (%u, %u)", + rhiId, header.rhiId); return; } const quint32 arch = quint32(sizeof(void*)); if (header.arch != arch) { - qWarning("setPipelineCacheData: Architecture does not match (%u, %u)", + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Architecture does not match (%u, %u)", arch, header.arch); return; } if (header.driverVersion != physDevProperties.driverVersion) { - qWarning("setPipelineCacheData: driverVersion does not match (%u, %u)", - physDevProperties.driverVersion, header.driverVersion); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: driverVersion does not match (%u, %u)", + physDevProperties.driverVersion, header.driverVersion); return; } if (header.vendorId != physDevProperties.vendorID) { - qWarning("setPipelineCacheData: vendorID does not match (%u, %u)", - physDevProperties.vendorID, header.vendorId); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: vendorID does not match (%u, %u)", + physDevProperties.vendorID, header.vendorId); return; } if (header.deviceId != physDevProperties.deviceID) { - qWarning("setPipelineCacheData: deviceID does not match (%u, %u)", - physDevProperties.deviceID, header.deviceId); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: deviceID does not match (%u, %u)", + physDevProperties.deviceID, header.deviceId); return; } if (header.uuidSize != VK_UUID_SIZE) { - qWarning("setPipelineCacheData: VK_UUID_SIZE does not match (%u, %u)", - quint32(VK_UUID_SIZE), header.uuidSize); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: VK_UUID_SIZE does not match (%u, %u)", + quint32(VK_UUID_SIZE), header.uuidSize); return; } if (data.size() < qsizetype(headerSize + VK_UUID_SIZE)) { - qWarning("setPipelineCacheData: Invalid blob, no uuid"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob, no uuid"); return; } if (memcmp(data.constData() + headerSize, physDevProperties.pipelineCacheUUID, VK_UUID_SIZE)) { - qWarning("setPipelineCacheData: pipelineCacheUUID does not match"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: pipelineCacheUUID does not match"); return; } const size_t dataOffset = headerSize + VK_UUID_SIZE; if (data.size() < qsizetype(dataOffset + header.dataSize)) { - qWarning("setPipelineCacheData: Invalid blob, data missing"); + qCDebug(QRHI_LOG_INFO, "setPipelineCacheData: Invalid blob, data missing"); return; } @@ -4460,7 +5026,7 @@ void QRhiVulkan::setPipelineCacheData(const QByteArray &data) qCDebug(QRHI_LOG_INFO, "Created pipeline cache with initial data of %d bytes", int(header.dataSize)); } else { - qWarning("Failed to create pipeline cache with initial data specified"); + qCDebug(QRHI_LOG_INFO, "Failed to create pipeline cache with initial data specified"); } } @@ -4472,10 +5038,10 @@ QRhiRenderBuffer *QRhiVulkan::createRenderBuffer(QRhiRenderBuffer::Type type, co } QRhiTexture *QRhiVulkan::createTexture(QRhiTexture::Format format, - const QSize &pixelSize, int depth, + const QSize &pixelSize, int depth, int arraySize, int sampleCount, QRhiTexture::Flags flags) { - return new QVkTexture(this, format, pixelSize, depth, sampleCount, flags); + return new QVkTexture(this, format, pixelSize, depth, arraySize, sampleCount, flags); } QRhiSampler *QRhiVulkan::createSampler(QRhiSampler::Filter magFilter, QRhiSampler::Filter minFilter, @@ -4555,8 +5121,8 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin // Do host writes and mark referenced shader resources as in-use. // Also prepare to ensure the descriptor set we are going to bind refers to up-to-date Vk objects. - for (int i = 0, ie = srbD->sortedBindings.count(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = srbD->sortedBindings[i].data(); + for (int i = 0, ie = srbD->sortedBindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings[i]); QVkShaderResourceBindings::BoundResourceData &bd(descSetBd[i]); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: @@ -4584,8 +5150,10 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin } break; case QRhiShaderResourceBinding::SampledTexture: + case QRhiShaderResourceBinding::Texture: + case QRhiShaderResourceBinding::Sampler: { - const QRhiShaderResourceBinding::Data::SampledTextureData *data = &b->u.stex; + const QRhiShaderResourceBinding::Data::TextureAndOrSamplerData *data = &b->u.stex; if (bd.stex.count != data->count) { bd.stex.count = data->count; rewriteDescSet = true; @@ -4593,21 +5161,32 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin for (int elem = 0; elem < data->count; ++elem) { QVkTexture *texD = QRHI_RES(QVkTexture, data->texSamplers[elem].tex); QVkSampler *samplerD = QRHI_RES(QVkSampler, data->texSamplers[elem].sampler); - texD->lastActiveFrameSlot = currentFrameSlot; - samplerD->lastActiveFrameSlot = currentFrameSlot; - trackedRegisterTexture(&passResTracker, texD, - QRhiPassResourceTracker::TexSample, - QRhiPassResourceTracker::toPassTrackerTextureStage(b->stage)); - if (texD->generation != bd.stex.d[elem].texGeneration - || texD->m_id != bd.stex.d[elem].texId - || samplerD->generation != bd.stex.d[elem].samplerGeneration - || samplerD->m_id != bd.stex.d[elem].samplerId) + // We use the same code path for both combined and separate + // images and samplers, so tex or sampler (but not both) can be + // null here. + Q_ASSERT(texD || samplerD); + if (texD) { + texD->lastActiveFrameSlot = currentFrameSlot; + trackedRegisterTexture(&passResTracker, texD, + QRhiPassResourceTracker::TexSample, + QRhiPassResourceTracker::toPassTrackerTextureStage(b->stage)); + } + if (samplerD) + samplerD->lastActiveFrameSlot = currentFrameSlot; + const quint64 texId = texD ? texD->m_id : 0; + const uint texGen = texD ? texD->generation : 0; + const quint64 samplerId = samplerD ? samplerD->m_id : 0; + const uint samplerGen = samplerD ? samplerD->generation : 0; + if (texGen != bd.stex.d[elem].texGeneration + || texId != bd.stex.d[elem].texId + || samplerGen != bd.stex.d[elem].samplerGeneration + || samplerId != bd.stex.d[elem].samplerId) { rewriteDescSet = true; - bd.stex.d[elem].texId = texD->m_id; - bd.stex.d[elem].texGeneration = texD->generation; - bd.stex.d[elem].samplerId = samplerD->m_id; - bd.stex.d[elem].samplerGeneration = samplerD->generation; + bd.stex.d[elem].texId = texId; + bd.stex.d[elem].texGeneration = texGen; + bd.stex.d[elem].samplerId = samplerId; + bd.stex.d[elem].samplerGeneration = samplerGen; } } } @@ -4689,8 +5268,8 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin // because dynOfs has to be ordered based on the binding numbers, // and neither srb nor dynamicOffsets has any such ordering // requirement. - for (const QRhiShaderResourceBinding &binding : qAsConst(srbD->sortedBindings)) { - const QRhiShaderResourceBinding::Data *b = binding.data(); + for (const QRhiShaderResourceBinding &binding : std::as_const(srbD->sortedBindings)) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(binding); if (b->type == QRhiShaderResourceBinding::UniformBuffer && b->u.ubuf.hasDynamicOffset) { uint32_t offset = 0; for (int i = 0; i < dynamicOffsetCount; ++i) { @@ -4710,8 +5289,8 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin gfxPsD ? VK_PIPELINE_BIND_POINT_GRAPHICS : VK_PIPELINE_BIND_POINT_COMPUTE, gfxPsD ? gfxPsD->layout : compPsD->layout, 0, 1, &srbD->descSets[descSetIdx], - uint32_t(dynOfs.count()), - dynOfs.count() ? dynOfs.constData() : nullptr); + uint32_t(dynOfs.size()), + dynOfs.size() ? dynOfs.constData() : nullptr); } else { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::BindDescriptorSet; @@ -4719,9 +5298,9 @@ void QRhiVulkan::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBin : VK_PIPELINE_BIND_POINT_COMPUTE; cmd.args.bindDescriptorSet.pipelineLayout = gfxPsD ? gfxPsD->layout : compPsD->layout; cmd.args.bindDescriptorSet.descSet = srbD->descSets[descSetIdx]; - cmd.args.bindDescriptorSet.dynamicOffsetCount = dynOfs.count(); - cmd.args.bindDescriptorSet.dynamicOffsetIndex = cbD->pools.dynamicOffset.count(); - cbD->pools.dynamicOffset.append(dynOfs.constData(), dynOfs.count()); + cmd.args.bindDescriptorSet.dynamicOffsetCount = dynOfs.size(); + cmd.args.bindDescriptorSet.dynamicOffsetIndex = cbD->pools.dynamicOffset.size(); + cbD->pools.dynamicOffset.append(dynOfs.constData(), dynOfs.size()); } if (gfxPsD) { @@ -4780,16 +5359,16 @@ void QRhiVulkan::setVertexInput(QRhiCommandBuffer *cb, if (cbD->passUsesSecondaryCb) { df->vkCmdBindVertexBuffers(cbD->activeSecondaryCbStack.last(), uint32_t(startBinding), - uint32_t(bufs.count()), bufs.constData(), ofs.constData()); + uint32_t(bufs.size()), bufs.constData(), ofs.constData()); } else { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::BindVertexBuffer; cmd.args.bindVertexBuffer.startBinding = startBinding; - cmd.args.bindVertexBuffer.count = bufs.count(); - cmd.args.bindVertexBuffer.vertexBufferIndex = cbD->pools.vertexBuffer.count(); - cbD->pools.vertexBuffer.append(bufs.constData(), bufs.count()); - cmd.args.bindVertexBuffer.vertexBufferOffsetIndex = cbD->pools.vertexBufferOffset.count(); - cbD->pools.vertexBufferOffset.append(ofs.constData(), ofs.count()); + cmd.args.bindVertexBuffer.count = bufs.size(); + cmd.args.bindVertexBuffer.vertexBufferIndex = cbD->pools.vertexBuffer.size(); + cbD->pools.vertexBuffer.append(bufs.constData(), bufs.size()); + cmd.args.bindVertexBuffer.vertexBufferOffsetIndex = cbD->pools.vertexBufferOffset.size(); + cbD->pools.vertexBufferOffset.append(ofs.constData(), ofs.size()); } } @@ -4838,7 +5417,7 @@ void QRhiVulkan::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport // x,y is top-left in VkViewport but bottom-left in QRhiViewport float x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, viewport.viewport(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<UnBounded>(outputSize, viewport.viewport(), &x, &y, &w, &h)) return; QVkCommandBuffer::Command &cmd(cbD->commands.get()); @@ -4857,9 +5436,12 @@ void QRhiVulkan::setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport cmd.cmd = QVkCommandBuffer::Command::SetViewport; } - if (!QRHI_RES(QVkGraphicsPipeline, cbD->currentGraphicsPipeline)->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)) { + if (cbD->currentGraphicsPipeline + && !QRHI_RES(QVkGraphicsPipeline, cbD->currentGraphicsPipeline) + ->m_flags.testFlag(QRhiGraphicsPipeline::UsesScissor)) { QVkCommandBuffer::Command &cmd(cbD->commands.get()); VkRect2D *s = &cmd.args.setScissor.scissor; + qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, viewport.viewport(), &x, &y, &w, &h); s->offset.x = int32_t(x); s->offset.y = int32_t(y); s->extent.width = uint32_t(w); @@ -4882,7 +5464,7 @@ void QRhiVulkan::setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) // x,y is top-left in VkRect2D but bottom-left in QRhiScissor int x, y, w, h; - if (!qrhi_toTopLeftRenderTargetRect(outputSize, scissor.scissor(), &x, &y, &w, &h)) + if (!qrhi_toTopLeftRenderTargetRect<Bounded>(outputSize, scissor.scissor(), &x, &y, &w, &h)) return; QVkCommandBuffer::Command &cmd(cbD->commands.get()); @@ -4972,60 +5554,72 @@ void QRhiVulkan::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, void QRhiVulkan::debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) { - if (!debugMarkers || !caps.debugMarkers) +#ifdef VK_EXT_debug_utils + if (!debugMarkers || !caps.debugUtils) return; - VkDebugMarkerMarkerInfoEXT marker; - memset(&marker, 0, sizeof(marker)); - marker.sType = VK_STRUCTURE_TYPE_DEBUG_MARKER_MARKER_INFO_EXT; + VkDebugUtilsLabelEXT label = {}; + label.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT; QVkCommandBuffer *cbD = QRHI_RES(QVkCommandBuffer, cb); if (cbD->recordingPass != QVkCommandBuffer::NoPass && cbD->passUsesSecondaryCb) { - marker.pMarkerName = name.constData(); - vkCmdDebugMarkerBegin(cbD->activeSecondaryCbStack.last(), &marker); + label.pLabelName = name.constData(); + vkCmdBeginDebugUtilsLabelEXT(cbD->activeSecondaryCbStack.last(), &label); } else { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::DebugMarkerBegin; - cmd.args.debugMarkerBegin.marker = marker; - cmd.args.debugMarkerBegin.markerNameIndex = cbD->pools.debugMarkerData.count(); + cmd.args.debugMarkerBegin.label = label; + cmd.args.debugMarkerBegin.labelNameIndex = cbD->pools.debugMarkerData.size(); cbD->pools.debugMarkerData.append(name); } +#else + Q_UNUSED(cb); + Q_UNUSED(name); +#endif } void QRhiVulkan::debugMarkEnd(QRhiCommandBuffer *cb) { - if (!debugMarkers || !caps.debugMarkers) +#ifdef VK_EXT_debug_utils + if (!debugMarkers || !caps.debugUtils) return; QVkCommandBuffer *cbD = QRHI_RES(QVkCommandBuffer, cb); if (cbD->recordingPass != QVkCommandBuffer::NoPass && cbD->passUsesSecondaryCb) { - vkCmdDebugMarkerEnd(cbD->activeSecondaryCbStack.last()); + vkCmdEndDebugUtilsLabelEXT(cbD->activeSecondaryCbStack.last()); } else { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::DebugMarkerEnd; } +#else + Q_UNUSED(cb); +#endif } void QRhiVulkan::debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) { - if (!debugMarkers || !caps.debugMarkers) +#ifdef VK_EXT_debug_utils + if (!debugMarkers || !caps.debugUtils) return; - VkDebugMarkerMarkerInfoEXT marker; - memset(&marker, 0, sizeof(marker)); - marker.sType = VK_STRUCTURE_TYPE_DEBUG_MARKER_MARKER_INFO_EXT; + VkDebugUtilsLabelEXT label = {}; + label.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT; QVkCommandBuffer *cbD = QRHI_RES(QVkCommandBuffer, cb); if (cbD->recordingPass != QVkCommandBuffer::NoPass && cbD->passUsesSecondaryCb) { - marker.pMarkerName = msg.constData(); - vkCmdDebugMarkerInsert(cbD->activeSecondaryCbStack.last(), &marker); + label.pLabelName = msg.constData(); + vkCmdInsertDebugUtilsLabelEXT(cbD->activeSecondaryCbStack.last(), &label); } else { QVkCommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QVkCommandBuffer::Command::DebugMarkerInsert; - cmd.args.debugMarkerInsert.marker = marker; - cmd.args.debugMarkerInsert.markerNameIndex = cbD->pools.debugMarkerData.count(); + cmd.args.debugMarkerInsert.label = label; + cmd.args.debugMarkerInsert.labelNameIndex = cbD->pools.debugMarkerData.size(); cbD->pools.debugMarkerData.append(msg); } +#else + Q_UNUSED(cb); + Q_UNUSED(msg); +#endif } const QRhiNativeHandles *QRhiVulkan::nativeHandles(QRhiCommandBuffer *cb) @@ -5039,8 +5633,8 @@ static inline QVkRenderTargetData *maybeRenderTargetData(QVkCommandBuffer *cbD) QVkRenderTargetData *rtD = nullptr; if (cbD->recordingPass == QVkCommandBuffer::RenderPass) { switch (cbD->currentTarget->resourceType()) { - case QRhiResource::RenderTarget: - rtD = &QRHI_RES(QVkReferenceRenderTarget, cbD->currentTarget)->d; + case QRhiResource::SwapChainRenderTarget: + rtD = &QRHI_RES(QVkSwapChainRenderTarget, cbD->currentTarget)->d; break; case QRhiResource::TextureRenderTarget: rtD = &QRHI_RES(QVkTextureRenderTarget, cbD->currentTarget)->d; @@ -5106,23 +5700,35 @@ void QRhiVulkan::endExternal(QRhiCommandBuffer *cb) cbD->resetCachedState(); } -void QRhiVulkan::setObjectName(uint64_t object, VkDebugReportObjectTypeEXT type, const QByteArray &name, int slot) +double QRhiVulkan::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QVkCommandBuffer *cbD = QRHI_RES(QVkCommandBuffer, cb); + return cbD->lastGpuTime; +} + +void QRhiVulkan::setObjectName(uint64_t object, VkObjectType type, const QByteArray &name, int slot) { - if (!debugMarkers || !caps.debugMarkers || name.isEmpty()) +#ifdef VK_EXT_debug_utils + if (!debugMarkers || !caps.debugUtils || name.isEmpty()) return; - VkDebugMarkerObjectNameInfoEXT nameInfo; - memset(&nameInfo, 0, sizeof(nameInfo)); - nameInfo.sType = VK_STRUCTURE_TYPE_DEBUG_MARKER_OBJECT_NAME_INFO_EXT; + VkDebugUtilsObjectNameInfoEXT nameInfo = {}; + nameInfo.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT; nameInfo.objectType = type; - nameInfo.object = object; + nameInfo.objectHandle = object; QByteArray decoratedName = name; if (slot >= 0) { decoratedName += '/'; decoratedName += QByteArray::number(slot); } nameInfo.pObjectName = decoratedName.constData(); - vkDebugMarkerSetObjectName(dev, &nameInfo); + vkSetDebugUtilsObjectNameEXT(dev, &nameInfo); +#else + Q_UNUSED(object); + Q_UNUSED(type); + Q_UNUSED(name); + Q_UNUSED(slot); +#endif } static inline VkBufferUsageFlagBits toVkBufferUsage(QRhiBuffer::UsageFlags usage) @@ -5147,8 +5753,7 @@ static inline VkFilter toVkFilter(QRhiSampler::Filter f) case QRhiSampler::Linear: return VK_FILTER_LINEAR; default: - Q_UNREACHABLE(); - return VK_FILTER_NEAREST; + Q_UNREACHABLE_RETURN(VK_FILTER_NEAREST); } } @@ -5162,8 +5767,7 @@ static inline VkSamplerMipmapMode toVkMipmapMode(QRhiSampler::Filter f) case QRhiSampler::Linear: return VK_SAMPLER_MIPMAP_MODE_LINEAR; default: - Q_UNREACHABLE(); - return VK_SAMPLER_MIPMAP_MODE_NEAREST; + Q_UNREACHABLE_RETURN(VK_SAMPLER_MIPMAP_MODE_NEAREST); } } @@ -5177,8 +5781,7 @@ static inline VkSamplerAddressMode toVkAddressMode(QRhiSampler::AddressMode m) case QRhiSampler::Mirror: return VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT; default: - Q_UNREACHABLE(); - return VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE; + Q_UNREACHABLE_RETURN(VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE); } } @@ -5187,13 +5790,18 @@ static inline VkShaderStageFlagBits toVkShaderStage(QRhiShaderStage::Type type) switch (type) { case QRhiShaderStage::Vertex: return VK_SHADER_STAGE_VERTEX_BIT; + case QRhiShaderStage::TessellationControl: + return VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT; + case QRhiShaderStage::TessellationEvaluation: + return VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT; case QRhiShaderStage::Fragment: return VK_SHADER_STAGE_FRAGMENT_BIT; case QRhiShaderStage::Compute: return VK_SHADER_STAGE_COMPUTE_BIT; + case QRhiShaderStage::Geometry: + return VK_SHADER_STAGE_GEOMETRY_BIT; default: - Q_UNREACHABLE(); - return VK_SHADER_STAGE_VERTEX_BIT; + Q_UNREACHABLE_RETURN(VK_SHADER_STAGE_VERTEX_BIT); } } @@ -5230,9 +5838,32 @@ static inline VkFormat toVkAttributeFormat(QRhiVertexInputAttribute::Format form return VK_FORMAT_R32G32_SINT; case QRhiVertexInputAttribute::SInt: return VK_FORMAT_R32_SINT; + case QRhiVertexInputAttribute::Half4: + return VK_FORMAT_R16G16B16A16_SFLOAT; + case QRhiVertexInputAttribute::Half3: + return VK_FORMAT_R16G16B16_SFLOAT; + case QRhiVertexInputAttribute::Half2: + return VK_FORMAT_R16G16_SFLOAT; + case QRhiVertexInputAttribute::Half: + return VK_FORMAT_R16_SFLOAT; + case QRhiVertexInputAttribute::UShort4: + return VK_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort3: + return VK_FORMAT_R16G16B16_UINT; + case QRhiVertexInputAttribute::UShort2: + return VK_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return VK_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + return VK_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort3: + return VK_FORMAT_R16G16B16_SINT; + case QRhiVertexInputAttribute::SShort2: + return VK_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return VK_FORMAT_R16_SINT; default: - Q_UNREACHABLE(); - return VK_FORMAT_R32G32B32A32_SFLOAT; + Q_UNREACHABLE_RETURN(VK_FORMAT_R32G32B32A32_SFLOAT); } } @@ -5251,9 +5882,10 @@ static inline VkPrimitiveTopology toVkTopology(QRhiGraphicsPipeline::Topology t) return VK_PRIMITIVE_TOPOLOGY_LINE_STRIP; case QRhiGraphicsPipeline::Points: return VK_PRIMITIVE_TOPOLOGY_POINT_LIST; + case QRhiGraphicsPipeline::Patches: + return VK_PRIMITIVE_TOPOLOGY_PATCH_LIST; default: - Q_UNREACHABLE(); - return VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST; + Q_UNREACHABLE_RETURN(VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST); } } @@ -5267,8 +5899,7 @@ static inline VkCullModeFlags toVkCullMode(QRhiGraphicsPipeline::CullMode c) case QRhiGraphicsPipeline::Back: return VK_CULL_MODE_BACK_BIT; default: - Q_UNREACHABLE(); - return VK_CULL_MODE_NONE; + Q_UNREACHABLE_RETURN(VK_CULL_MODE_NONE); } } @@ -5280,8 +5911,7 @@ static inline VkFrontFace toVkFrontFace(QRhiGraphicsPipeline::FrontFace f) case QRhiGraphicsPipeline::CW: return VK_FRONT_FACE_CLOCKWISE; default: - Q_UNREACHABLE(); - return VK_FRONT_FACE_COUNTER_CLOCKWISE; + Q_UNREACHABLE_RETURN(VK_FRONT_FACE_COUNTER_CLOCKWISE); } } @@ -5341,8 +5971,7 @@ static inline VkBlendFactor toVkBlendFactor(QRhiGraphicsPipeline::BlendFactor f) case QRhiGraphicsPipeline::OneMinusSrc1Alpha: return VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA; default: - Q_UNREACHABLE(); - return VK_BLEND_FACTOR_ZERO; + Q_UNREACHABLE_RETURN(VK_BLEND_FACTOR_ZERO); } } @@ -5360,8 +5989,7 @@ static inline VkBlendOp toVkBlendOp(QRhiGraphicsPipeline::BlendOp op) case QRhiGraphicsPipeline::Max: return VK_BLEND_OP_MAX; default: - Q_UNREACHABLE(); - return VK_BLEND_OP_ADD; + Q_UNREACHABLE_RETURN(VK_BLEND_OP_ADD); } } @@ -5385,8 +6013,7 @@ static inline VkCompareOp toVkCompareOp(QRhiGraphicsPipeline::CompareOp op) case QRhiGraphicsPipeline::Always: return VK_COMPARE_OP_ALWAYS; default: - Q_UNREACHABLE(); - return VK_COMPARE_OP_ALWAYS; + Q_UNREACHABLE_RETURN(VK_COMPARE_OP_ALWAYS); } } @@ -5410,8 +6037,19 @@ static inline VkStencilOp toVkStencilOp(QRhiGraphicsPipeline::StencilOp op) case QRhiGraphicsPipeline::DecrementAndWrap: return VK_STENCIL_OP_DECREMENT_AND_WRAP; default: - Q_UNREACHABLE(); - return VK_STENCIL_OP_KEEP; + Q_UNREACHABLE_RETURN(VK_STENCIL_OP_KEEP); + } +} + +static inline VkPolygonMode toVkPolygonMode(QRhiGraphicsPipeline::PolygonMode mode) +{ + switch (mode) { + case QRhiGraphicsPipeline::Fill: + return VK_POLYGON_MODE_FILL; + case QRhiGraphicsPipeline::Line: + return VK_POLYGON_MODE_LINE; + default: + Q_UNREACHABLE_RETURN(VK_POLYGON_MODE_FILL); } } @@ -5433,6 +6071,12 @@ static inline VkDescriptorType toVkDescriptorType(const QRhiShaderResourceBindin case QRhiShaderResourceBinding::SampledTexture: return VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER; + case QRhiShaderResourceBinding::Texture: + return VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE; + + case QRhiShaderResourceBinding::Sampler: + return VK_DESCRIPTOR_TYPE_SAMPLER; + case QRhiShaderResourceBinding::ImageLoad: case QRhiShaderResourceBinding::ImageStore: case QRhiShaderResourceBinding::ImageLoadStore: @@ -5444,8 +6088,7 @@ static inline VkDescriptorType toVkDescriptorType(const QRhiShaderResourceBindin return VK_DESCRIPTOR_TYPE_STORAGE_BUFFER; default: - Q_UNREACHABLE(); - return VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER; + Q_UNREACHABLE_RETURN(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER); } } @@ -5458,6 +6101,12 @@ static inline VkShaderStageFlags toVkShaderStageFlags(QRhiShaderResourceBinding: s |= VK_SHADER_STAGE_FRAGMENT_BIT; if (stage.testFlag(QRhiShaderResourceBinding::ComputeStage)) s |= VK_SHADER_STAGE_COMPUTE_BIT; + if (stage.testFlag(QRhiShaderResourceBinding::TessellationControlStage)) + s |= VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT; + if (stage.testFlag(QRhiShaderResourceBinding::TessellationEvaluationStage)) + s |= VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT; + if (stage.testFlag(QRhiShaderResourceBinding::GeometryStage)) + s |= VK_SHADER_STAGE_GEOMETRY_BIT; return VkShaderStageFlags(s); } @@ -5481,12 +6130,11 @@ static inline VkCompareOp toVkTextureCompareOp(QRhiSampler::CompareOp op) case QRhiSampler::Always: return VK_COMPARE_OP_ALWAYS; default: - Q_UNREACHABLE(); - return VK_COMPARE_OP_NEVER; + Q_UNREACHABLE_RETURN(VK_COMPARE_OP_NEVER); } } -QVkBuffer::QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size) +QVkBuffer::QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size) : QRhiBuffer(rhi, type, usage, size) { for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { @@ -5523,12 +6171,13 @@ void QVkBuffer::destroy() } QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - QRHI_PROF; - QRHI_PROF_F(releaseBuffer(this)); - - rhiD->unregisterResource(this); + // destroy() implementations, unlike other functions, are expected to test + // for m_rhi being null, to allow surviving in case one attempts to destroy + // a (leaked) resource after the QRhi. + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkBuffer::create() @@ -5541,16 +6190,14 @@ bool QVkBuffer::create() return false; } - const int nonZeroSize = m_size <= 0 ? 256 : m_size; + const quint32 nonZeroSize = m_size <= 0 ? 256 : m_size; - VkBufferCreateInfo bufferInfo; - memset(&bufferInfo, 0, sizeof(bufferInfo)); + VkBufferCreateInfo bufferInfo = {}; bufferInfo.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO; - bufferInfo.size = uint32_t(nonZeroSize); + bufferInfo.size = nonZeroSize; bufferInfo.usage = toVkBufferUsage(m_usage); - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; if (m_type == Dynamic) { #ifndef Q_OS_DARWIN // not for MoltenVK @@ -5578,19 +6225,17 @@ bool QVkBuffer::create() if (err != VK_SUCCESS) break; allocations[i] = allocation; - rhiD->setObjectName(uint64_t(buffers[i]), VK_DEBUG_REPORT_OBJECT_TYPE_BUFFER_EXT, m_objectName, + rhiD->setObjectName(uint64_t(buffers[i]), VK_OBJECT_TYPE_BUFFER, m_objectName, m_type == Dynamic ? i : -1); } } if (err != VK_SUCCESS) { - qWarning("Failed to create buffer: %d", err); + qWarning("Failed to create buffer of size %u: %d", nonZeroSize, err); + rhiD->printExtraErrorInfo(err); return false; } - QRHI_PROF; - QRHI_PROF_F(newBuffer(this, uint(nonZeroSize), m_type != Dynamic ? 1 : QVK_FRAMES_IN_FLIGHT, 0)); - lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -5639,8 +6284,8 @@ void QVkBuffer::endFullDynamicBufferUpdateForCurrentFrame() QRHI_RES_RHI(QRhiVulkan); const int slot = rhiD->currentFrameSlot; VmaAllocation a = toVmaAllocation(allocations[slot]); - vmaUnmapMemory(toVmaAllocator(rhiD->allocator), a); vmaFlushAllocation(toVmaAllocator(rhiD->allocator), a, 0, m_size); + vmaUnmapMemory(toVmaAllocator(rhiD->allocator), a); } QVkRenderBuffer::QVkRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, @@ -5680,12 +6325,10 @@ void QVkRenderBuffer::destroy() } QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - QRHI_PROF; - QRHI_PROF_F(releaseRenderBuffer(this)); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkRenderBuffer::create() @@ -5697,8 +6340,7 @@ bool QVkRenderBuffer::create() return false; QRHI_RES_RHI(QRhiVulkan); - QRHI_PROF; - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); switch (m_type) { case QRhiRenderBuffer::Color: @@ -5707,6 +6349,7 @@ bool QVkRenderBuffer::create() backingTexture = QRHI_RES(QVkTexture, rhiD->createTexture(backingFormat(), m_pixelSize, 1, + 0, m_sampleCount, QRhiTexture::RenderTarget | QRhiTexture::UsedAsTransferSource)); } else { @@ -5717,7 +6360,6 @@ bool QVkRenderBuffer::create() if (!backingTexture->create()) return false; vkformat = backingTexture->vkformat; - QRHI_PROF_F(newRenderBuffer(this, false, false, samples)); } break; case QRhiRenderBuffer::DepthStencil: @@ -5734,8 +6376,7 @@ bool QVkRenderBuffer::create() { return false; } - rhiD->setObjectName(uint64_t(image), VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_EXT, m_objectName); - QRHI_PROF_F(newRenderBuffer(this, true, false, samples)); + rhiD->setObjectName(uint64_t(image), VK_OBJECT_TYPE_IMAGE, m_objectName); break; default: Q_UNREACHABLE(); @@ -5743,6 +6384,7 @@ bool QVkRenderBuffer::create() } lastActiveFrameSlot = -1; + generation += 1; rhiD->registerResource(this); return true; } @@ -5756,8 +6398,8 @@ QRhiTexture::Format QVkRenderBuffer::backingFormat() const } QVkTexture::QVkTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags) - : QRhiTexture(rhi, format, pixelSize, depth, sampleCount, flags) + int arraySize, int sampleCount, Flags flags) + : QRhiTexture(rhi, format, pixelSize, depth, arraySize, sampleCount, flags) { for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { stagingBuffers[i] = VK_NULL_HANDLE; @@ -5803,12 +6445,10 @@ void QVkTexture::destroy() imageAlloc = nullptr; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - QRHI_PROF; - QRHI_PROF_F(releaseTexture(this)); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkTexture::prepareCreate(QSize *adjustedSize) @@ -5818,6 +6458,15 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) QRHI_RES_RHI(QRhiVulkan); vkformat = toVkTextureFormat(m_format, m_flags); + if (m_writeViewFormat.format != UnknownFormat) + viewFormat = toVkTextureFormat(m_writeViewFormat.format, m_writeViewFormat.srgb ? sRGB : Flags()); + else + viewFormat = vkformat; + if (m_readViewFormat.format != UnknownFormat) + viewFormatForSampling = toVkTextureFormat(m_readViewFormat.format, m_readViewFormat.srgb ? sRGB : Flags()); + else + viewFormatForSampling = vkformat; + VkFormatProperties props; rhiD->f->vkGetPhysicalDeviceFormatProperties(rhiD->physDev, vkformat, &props); const bool canSampleOptimal = (props.optimalTilingFeatures & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT); @@ -5826,18 +6475,22 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) return false; } - const QSize size = m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize; const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool is1D = m_flags.testFlag(OneDimensional); const bool hasMipMaps = m_flags.testFlag(MipMapped); + const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) + : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); + mipLevelCount = uint(hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1); const int maxLevels = QRhi::MAX_MIP_LEVELS; if (mipLevelCount > maxLevels) { qWarning("Too many mip levels (%d, max is %d), truncating mip chain", mipLevelCount, maxLevels); mipLevelCount = maxLevels; } - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); if (samples > VK_SAMPLE_COUNT_1_BIT) { if (isCube) { qWarning("Cubemap texture cannot be multisample"); @@ -5856,11 +6509,30 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both cube and 3D"); return false; } - m_depth = qMax(1, m_depth); + if (isArray && is3D) { + qWarning("Texture cannot be both array and 3D"); + return false; + } + if (isCube && is1D) { + qWarning("Texture cannot be both cube and 1D"); + return false; + } + if (is1D && is3D) { + qWarning("Texture cannot be both 1D and 3D"); + return false; + } if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } + if (m_arraySize > 0 && !isArray) { + qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); + return false; + } + if (m_arraySize < 1 && isArray) { + qWarning("Texture is an array but array size is %d", m_arraySize); + return false; + } usageState.layout = VK_IMAGE_LAYOUT_PREINITIALIZED; usageState.access = 0; @@ -5878,21 +6550,31 @@ bool QVkTexture::finishCreate() const auto aspectMask = aspectMaskForTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool is1D = m_flags.testFlag(OneDimensional); - VkImageViewCreateInfo viewInfo; - memset(&viewInfo, 0, sizeof(viewInfo)); + VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = image; - viewInfo.viewType = isCube ? VK_IMAGE_VIEW_TYPE_CUBE : (is3D ? VK_IMAGE_VIEW_TYPE_3D : VK_IMAGE_VIEW_TYPE_2D); - viewInfo.format = vkformat; + viewInfo.viewType = isCube + ? VK_IMAGE_VIEW_TYPE_CUBE + : (is3D ? VK_IMAGE_VIEW_TYPE_3D + : (is1D ? (isArray ? VK_IMAGE_VIEW_TYPE_1D_ARRAY : VK_IMAGE_VIEW_TYPE_1D) + : (isArray ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D))); + viewInfo.format = viewFormatForSampling; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; viewInfo.components.a = VK_COMPONENT_SWIZZLE_A; viewInfo.subresourceRange.aspectMask = aspectMask; viewInfo.subresourceRange.levelCount = mipLevelCount; - viewInfo.subresourceRange.layerCount = isCube ? 6 : 1; + if (isArray && m_arrayRangeStart >= 0 && m_arrayRangeLength >= 0) { + viewInfo.subresourceRange.baseArrayLayer = uint32_t(m_arrayRangeStart); + viewInfo.subresourceRange.layerCount = uint32_t(m_arrayRangeLength); + } else { + viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); + } VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &imageView); if (err != VK_SUCCESS) { @@ -5916,10 +6598,11 @@ bool QVkTexture::create() const bool isRenderTarget = m_flags.testFlag(QRhiTexture::RenderTarget); const bool isDepth = isDepthTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool is1D = m_flags.testFlag(OneDimensional); - VkImageCreateInfo imageInfo; - memset(&imageInfo, 0, sizeof(imageInfo)); + VkImageCreateInfo imageInfo = {}; imageInfo.sType = VK_STRUCTURE_TYPE_IMAGE_CREATE_INFO; imageInfo.flags = 0; if (isCube) @@ -5940,13 +6623,13 @@ bool QVkTexture::create() #endif } - imageInfo.imageType = is3D ? VK_IMAGE_TYPE_3D : VK_IMAGE_TYPE_2D; + imageInfo.imageType = is1D ? VK_IMAGE_TYPE_1D : is3D ? VK_IMAGE_TYPE_3D : VK_IMAGE_TYPE_2D; imageInfo.format = vkformat; imageInfo.extent.width = uint32_t(size.width()); imageInfo.extent.height = uint32_t(size.height()); - imageInfo.extent.depth = is3D ? m_depth : 1; + imageInfo.extent.depth = is3D ? qMax(1, m_depth) : 1; imageInfo.mipLevels = mipLevelCount; - imageInfo.arrayLayers = isCube ? 6 : 1; + imageInfo.arrayLayers = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); imageInfo.samples = samples; imageInfo.tiling = VK_IMAGE_TILING_OPTIMAL; imageInfo.initialLayout = VK_IMAGE_LAYOUT_PREINITIALIZED; @@ -5965,14 +6648,20 @@ bool QVkTexture::create() if (m_flags.testFlag(QRhiTexture::UsedWithLoadStore)) imageInfo.usage |= VK_IMAGE_USAGE_STORAGE_BIT; - VmaAllocationCreateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VmaAllocationCreateInfo allocInfo = {}; allocInfo.usage = VMA_MEMORY_USAGE_GPU_ONLY; VmaAllocation allocation; VkResult err = vmaCreateImage(toVmaAllocator(rhiD->allocator), &imageInfo, &allocInfo, &image, &allocation, nullptr); if (err != VK_SUCCESS) { - qWarning("Failed to create image: %d", err); + qWarning("Failed to create image (with VkImageCreateInfo %ux%u depth %u vkformat 0x%X mips %u layers %u vksamples 0x%X): %d", + imageInfo.extent.width, imageInfo.extent.height, imageInfo.extent.depth, + int(imageInfo.format), + imageInfo.mipLevels, + imageInfo.arrayLayers, + int(imageInfo.samples), + err); + rhiD->printExtraErrorInfo(err); return false; } imageAlloc = allocation; @@ -5980,10 +6669,7 @@ bool QVkTexture::create() if (!finishCreate()) return false; - rhiD->setObjectName(uint64_t(image), VK_DEBUG_REPORT_OBJECT_TYPE_IMAGE_EXT, m_objectName); - - QRHI_PROF; - QRHI_PROF_F(newTexture(this, true, int(mipLevelCount), isCube ? 6 : 1, samples)); + rhiD->setObjectName(uint64_t(image), VK_OBJECT_TYPE_IMAGE, m_objectName); owns = true; rhiD->registerResource(this); @@ -6004,9 +6690,6 @@ bool QVkTexture::createFrom(QRhiTexture::NativeTexture src) if (!finishCreate()) return false; - QRHI_PROF; - QRHI_PROF_F(newTexture(this, false, int(mipLevelCount), m_flags.testFlag(CubeMap) ? 6 : 1, samples)); - usageState.layout = VkImageLayout(src.layout); owns = false; @@ -6025,7 +6708,7 @@ void QVkTexture::setNativeLayout(int layout) usageState.layout = VkImageLayout(layout); } -VkImageView QVkTexture::imageViewForLevel(int level) +VkImageView QVkTexture::perLevelImageViewForLoadStore(int level) { Q_ASSERT(level >= 0 && level < int(mipLevelCount)); if (perLevelImageViews[level] != VK_NULL_HANDLE) @@ -6033,14 +6716,19 @@ VkImageView QVkTexture::imageViewForLevel(int level) const VkImageAspectFlags aspectMask = aspectMaskForTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); + const bool isArray = m_flags.testFlag(TextureArray); const bool is3D = m_flags.testFlag(ThreeDimensional); + const bool is1D = m_flags.testFlag(OneDimensional); - VkImageViewCreateInfo viewInfo; - memset(&viewInfo, 0, sizeof(viewInfo)); + VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = image; - viewInfo.viewType = isCube ? VK_IMAGE_VIEW_TYPE_CUBE : (is3D ? VK_IMAGE_VIEW_TYPE_3D : VK_IMAGE_VIEW_TYPE_2D); - viewInfo.format = vkformat; + viewInfo.viewType = isCube + ? VK_IMAGE_VIEW_TYPE_CUBE + : (is3D ? VK_IMAGE_VIEW_TYPE_3D + : (is1D ? (isArray ? VK_IMAGE_VIEW_TYPE_1D_ARRAY : VK_IMAGE_VIEW_TYPE_1D) + : (isArray ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D))); + viewInfo.format = viewFormat; // this is writeViewFormat, regardless of Load, Store, or LoadStore; intentional viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6049,7 +6737,7 @@ VkImageView QVkTexture::imageViewForLevel(int level) viewInfo.subresourceRange.baseMipLevel = uint32_t(level); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = 0; - viewInfo.subresourceRange.layerCount = isCube ? 6 : 1; + viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); VkImageView v = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); @@ -6087,8 +6775,10 @@ void QVkSampler::destroy() sampler = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkSampler::create() @@ -6096,8 +6786,7 @@ bool QVkSampler::create() if (sampler) destroy(); - VkSamplerCreateInfo samplerInfo; - memset(&samplerInfo, 0, sizeof(samplerInfo)); + VkSamplerCreateInfo samplerInfo = {}; samplerInfo.sType = VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO; samplerInfo.magFilter = toVkFilter(m_magFilter); samplerInfo.minFilter = toVkFilter(m_minFilter); @@ -6126,6 +6815,7 @@ bool QVkSampler::create() QVkRenderPassDescriptor::QVkRenderPassDescriptor(QRhiImplementation *rhi) : QRhiRenderPassDescriptor(rhi) { + serializedFormatData.reserve(64); } QVkRenderPassDescriptor::~QVkRenderPassDescriptor() @@ -6152,9 +6842,10 @@ void QVkRenderPassDescriptor::destroy() rp = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } static inline bool attachmentDescriptionEquals(const VkAttachmentDescription &a, const VkAttachmentDescription &b) @@ -6179,16 +6870,20 @@ bool QVkRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other const QVkRenderPassDescriptor *o = QRHI_RES(const QVkRenderPassDescriptor, other); - if (attDescs.count() != o->attDescs.count()) + if (attDescs.size() != o->attDescs.size()) return false; - if (colorRefs.count() != o->colorRefs.count()) + if (colorRefs.size() != o->colorRefs.size()) return false; - if (resolveRefs.count() != o->resolveRefs.count()) + if (resolveRefs.size() != o->resolveRefs.size()) return false; if (hasDepthStencil != o->hasDepthStencil) return false; + if (hasDepthStencilResolve != o->hasDepthStencilResolve) + return false; + if (multiViewCount != o->multiViewCount) + return false; - for (int i = 0, ie = colorRefs.count(); i != ie; ++i) { + for (int i = 0, ie = colorRefs.size(); i != ie; ++i) { const uint32_t attIdx = colorRefs[i].attachment; if (attIdx != o->colorRefs[i].attachment) return false; @@ -6204,7 +6899,7 @@ bool QVkRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other return false; } - for (int i = 0, ie = resolveRefs.count(); i != ie; ++i) { + for (int i = 0, ie = resolveRefs.size(); i != ie; ++i) { const uint32_t attIdx = resolveRefs[i].attachment; if (attIdx != o->resolveRefs[i].attachment) return false; @@ -6212,11 +6907,69 @@ bool QVkRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other return false; } + if (hasDepthStencilResolve) { + const uint32_t attIdx = dsResolveRef.attachment; + if (attIdx != o->dsResolveRef.attachment) + return false; + if (attIdx != VK_ATTACHMENT_UNUSED && !attachmentDescriptionEquals(attDescs[attIdx], o->attDescs[attIdx])) + return false; + } + // subpassDeps is not included return true; } +void QVkRenderPassDescriptor::updateSerializedFormat() +{ + serializedFormatData.clear(); + auto p = std::back_inserter(serializedFormatData); + + *p++ = attDescs.size(); + *p++ = colorRefs.size(); + *p++ = resolveRefs.size(); + *p++ = hasDepthStencil; + *p++ = hasDepthStencilResolve; + *p++ = multiViewCount; + + auto serializeAttachmentData = [this, &p](uint32_t attIdx) { + const bool used = attIdx != VK_ATTACHMENT_UNUSED; + const VkAttachmentDescription *a = used ? &attDescs[attIdx] : nullptr; + *p++ = used ? a->format : 0; + *p++ = used ? a->samples : 0; + *p++ = used ? a->loadOp : 0; + *p++ = used ? a->storeOp : 0; + *p++ = used ? a->stencilLoadOp : 0; + *p++ = used ? a->stencilStoreOp : 0; + *p++ = used ? a->initialLayout : 0; + *p++ = used ? a->finalLayout : 0; + }; + + for (int i = 0, ie = colorRefs.size(); i != ie; ++i) { + const uint32_t attIdx = colorRefs[i].attachment; + *p++ = attIdx; + serializeAttachmentData(attIdx); + } + + if (hasDepthStencil) { + const uint32_t attIdx = dsRef.attachment; + *p++ = attIdx; + serializeAttachmentData(attIdx); + } + + for (int i = 0, ie = resolveRefs.size(); i != ie; ++i) { + const uint32_t attIdx = resolveRefs[i].attachment; + *p++ = attIdx; + serializeAttachmentData(attIdx); + } + + if (hasDepthStencilResolve) { + const uint32_t attIdx = dsResolveRef.attachment; + *p++ = attIdx; + serializeAttachmentData(attIdx); + } +} + QRhiRenderPassDescriptor *QVkRenderPassDescriptor::newCompatibleRenderPassDescriptor() const { QVkRenderPassDescriptor *rpD = new QVkRenderPassDescriptor(m_rhi); @@ -6227,13 +6980,22 @@ QRhiRenderPassDescriptor *QVkRenderPassDescriptor::newCompatibleRenderPassDescri rpD->resolveRefs = resolveRefs; rpD->subpassDeps = subpassDeps; rpD->hasDepthStencil = hasDepthStencil; + rpD->hasDepthStencilResolve = hasDepthStencilResolve; + rpD->multiViewCount = multiViewCount; rpD->dsRef = dsRef; + rpD->dsResolveRef = dsResolveRef; VkRenderPassCreateInfo rpInfo; VkSubpassDescription subpassDesc; fillRenderPassCreateInfo(&rpInfo, &subpassDesc, rpD); QRHI_RES_RHI(QRhiVulkan); + MultiViewRenderPassSetupHelper multiViewHelper; + if (!multiViewHelper.prepare(&rpInfo, multiViewCount, rhiD->caps.multiView)) { + delete rpD; + return nullptr; + } + VkResult err = rhiD->df->vkCreateRenderPass(rhiD->dev, &rpInfo, nullptr, &rpD->rp); if (err != VK_SUCCESS) { qWarning("Failed to create renderpass: %d", err); @@ -6241,42 +7003,48 @@ QRhiRenderPassDescriptor *QVkRenderPassDescriptor::newCompatibleRenderPassDescri return nullptr; } + rpD->updateSerializedFormat(); rhiD->registerResource(rpD); return rpD; } +QVector<quint32> QVkRenderPassDescriptor::serializedFormat() const +{ + return serializedFormatData; +} + const QRhiNativeHandles *QVkRenderPassDescriptor::nativeHandles() { nativeHandlesStruct.renderPass = rp; return &nativeHandlesStruct; } -QVkReferenceRenderTarget::QVkReferenceRenderTarget(QRhiImplementation *rhi) - : QRhiRenderTarget(rhi) +QVkSwapChainRenderTarget::QVkSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain) + : QRhiSwapChainRenderTarget(rhi, swapchain) { } -QVkReferenceRenderTarget::~QVkReferenceRenderTarget() +QVkSwapChainRenderTarget::~QVkSwapChainRenderTarget() { destroy(); } -void QVkReferenceRenderTarget::destroy() +void QVkSwapChainRenderTarget::destroy() { // nothing to do here } -QSize QVkReferenceRenderTarget::pixelSize() const +QSize QVkSwapChainRenderTarget::pixelSize() const { return d.pixelSize; } -float QVkReferenceRenderTarget::devicePixelRatio() const +float QVkSwapChainRenderTarget::devicePixelRatio() const { return d.dpr; } -int QVkReferenceRenderTarget::sampleCount() const +int QVkSwapChainRenderTarget::sampleCount() const { return d.sampleCount; } @@ -6316,10 +7084,16 @@ void QVkTextureRenderTarget::destroy() resrtv[att] = VK_NULL_HANDLE; } - QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); + e.textureRenderTarget.dsv = dsv; + dsv = VK_NULL_HANDLE; + e.textureRenderTarget.resdsv = resdsv; + resdsv = VK_NULL_HANDLE; - rhiD->unregisterResource(this); + QRHI_RES_RHI(QRhiVulkan); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } QRhiRenderPassDescriptor *QVkTextureRenderTarget::newCompatibleRenderPassDescriptor() @@ -6333,14 +7107,17 @@ QRhiRenderPassDescriptor *QVkTextureRenderTarget::newCompatibleRenderPassDescrip m_desc.cendColorAttachments(), m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents), m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents), + m_desc.depthTexture() && !m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture(), m_desc.depthStencilBuffer(), - m_desc.depthTexture())) + m_desc.depthTexture(), + m_desc.depthResolveTexture())) { delete rp; return nullptr; } rp->ownsRp = true; + rp->updateSerializedFormat(); rhiD->registerResource(rp); return rp; } @@ -6350,13 +7127,13 @@ bool QVkTextureRenderTarget::create() if (d.fb) destroy(); - const bool hasColorAttachments = m_desc.cbeginColorAttachments() != m_desc.cendColorAttachments(); - Q_ASSERT(hasColorAttachments || m_desc.depthTexture()); + Q_ASSERT(m_desc.colorAttachmentCount() > 0 || m_desc.depthTexture()); Q_ASSERT(!m_desc.depthStencilBuffer() || !m_desc.depthTexture()); const bool hasDepthStencil = m_desc.depthStencilBuffer() || m_desc.depthTexture(); QRHI_RES_RHI(QRhiVulkan); QVarLengthArray<VkImageView, 8> views; + d.multiViewCount = 0; d.colorAttCount = 0; int attIndex = 0; @@ -6367,12 +7144,17 @@ bool QVkTextureRenderTarget::create() Q_ASSERT(texD || rbD); if (texD) { Q_ASSERT(texD->flags().testFlag(QRhiTexture::RenderTarget)); - VkImageViewCreateInfo viewInfo; - memset(&viewInfo, 0, sizeof(viewInfo)); + const bool is1D = texD->flags().testFlag(QRhiTexture::OneDimensional); + const bool isMultiView = it->multiViewCount() >= 2; + if (isMultiView && d.multiViewCount == 0) + d.multiViewCount = it->multiViewCount(); + VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = texD->image; - viewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; - viewInfo.format = texD->vkformat; + viewInfo.viewType = is1D ? VK_IMAGE_VIEW_TYPE_1D + : (isMultiView ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D); + viewInfo.format = texD->viewFormat; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6381,7 +7163,7 @@ bool QVkTextureRenderTarget::create() viewInfo.subresourceRange.baseMipLevel = uint32_t(it->level()); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = uint32_t(it->layer()); - viewInfo.subresourceRange.layerCount = 1; + viewInfo.subresourceRange.layerCount = uint32_t(isMultiView ? it->multiViewCount() : 1); VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &rtv[attIndex]); if (err != VK_SUCCESS) { qWarning("Failed to create render target image view: %d", err); @@ -6406,7 +7188,25 @@ bool QVkTextureRenderTarget::create() if (hasDepthStencil) { if (m_desc.depthTexture()) { QVkTexture *depthTexD = QRHI_RES(QVkTexture, m_desc.depthTexture()); - views.append(depthTexD->imageView); + // need a dedicated view just because viewFormat may differ from vkformat + VkImageViewCreateInfo viewInfo = {}; + viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + viewInfo.image = depthTexD->image; + viewInfo.viewType = d.multiViewCount > 1 ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = depthTexD->viewFormat; + viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + viewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + viewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + viewInfo.subresourceRange.levelCount = 1; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); + VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &dsv); + if (err != VK_SUCCESS) { + qWarning("Failed to create depth-stencil image view for rt: %d", err); + return false; + } + views.append(dsv); if (d.colorAttCount == 0) { d.pixelSize = depthTexD->pixelSize(); d.sampleCount = depthTexD->samples; @@ -6426,18 +7226,19 @@ bool QVkTextureRenderTarget::create() d.resolveAttCount = 0; attIndex = 0; + Q_ASSERT(d.multiViewCount == 0 || d.multiViewCount >= 2); for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it, ++attIndex) { if (it->resolveTexture()) { QVkTexture *resTexD = QRHI_RES(QVkTexture, it->resolveTexture()); Q_ASSERT(resTexD->flags().testFlag(QRhiTexture::RenderTarget)); d.resolveAttCount += 1; - VkImageViewCreateInfo viewInfo; - memset(&viewInfo, 0, sizeof(viewInfo)); + VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = resTexD->image; - viewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; - viewInfo.format = resTexD->vkformat; + viewInfo.viewType = d.multiViewCount ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = resTexD->viewFormat; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6446,7 +7247,7 @@ bool QVkTextureRenderTarget::create() viewInfo.subresourceRange.baseMipLevel = uint32_t(it->resolveLevel()); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = uint32_t(it->resolveLayer()); - viewInfo.subresourceRange.layerCount = 1; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &resrtv[attIndex]); if (err != VK_SUCCESS) { qWarning("Failed to create render target resolve image view: %d", err); @@ -6456,17 +7257,46 @@ bool QVkTextureRenderTarget::create() } } + if (m_desc.depthResolveTexture()) { + QVkTexture *resTexD = QRHI_RES(QVkTexture, m_desc.depthResolveTexture()); + Q_ASSERT(resTexD->flags().testFlag(QRhiTexture::RenderTarget)); + + VkImageViewCreateInfo viewInfo = {}; + viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + viewInfo.image = resTexD->image; + viewInfo.viewType = d.multiViewCount ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = resTexD->viewFormat; + viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + viewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + viewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + viewInfo.subresourceRange.baseMipLevel = 0; + viewInfo.subresourceRange.levelCount = 1; + viewInfo.subresourceRange.baseArrayLayer = 0; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); + VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &resdsv); + if (err != VK_SUCCESS) { + qWarning("Failed to create render target depth resolve image view: %d", err); + return false; + } + views.append(resdsv); + d.dsResolveAttCount = 1; + } else { + d.dsResolveAttCount = 0; + } + if (!m_renderPassDesc) qWarning("QVkTextureRenderTarget: No renderpass descriptor set. See newCompatibleRenderPassDescriptor() and setRenderPassDescriptor()."); d.rp = QRHI_RES(QVkRenderPassDescriptor, m_renderPassDesc); Q_ASSERT(d.rp && d.rp->rp); - VkFramebufferCreateInfo fbInfo; - memset(&fbInfo, 0, sizeof(fbInfo)); + VkFramebufferCreateInfo fbInfo = {}; fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; fbInfo.renderPass = d.rp->rp; - fbInfo.attachmentCount = uint32_t(d.colorAttCount + d.dsAttCount + d.resolveAttCount); + fbInfo.attachmentCount = uint32_t(d.colorAttCount + d.dsAttCount + d.resolveAttCount + d.dsResolveAttCount); fbInfo.pAttachments = views.constData(); fbInfo.width = uint32_t(d.pixelSize.width()); fbInfo.height = uint32_t(d.pixelSize.height()); @@ -6478,6 +7308,8 @@ bool QVkTextureRenderTarget::create() return false; } + QRhiRenderTargetAttachmentTracker::updateResIdList<QVkTexture, QVkRenderBuffer>(m_desc, &d.currentResIdList); + lastActiveFrameSlot = -1; rhiD->registerResource(this); return true; @@ -6485,6 +7317,9 @@ bool QVkTextureRenderTarget::create() QSize QVkTextureRenderTarget::pixelSize() const { + if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QVkTexture, QVkRenderBuffer>(m_desc, d.currentResIdList)) + const_cast<QVkTextureRenderTarget *>(this)->create(); + return d.pixelSize; } @@ -6528,9 +7363,10 @@ void QVkShaderResourceBindings::destroy() descSets[i] = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkShaderResourceBindings::create() @@ -6549,16 +7385,12 @@ bool QVkShaderResourceBindings::create() sortedBindings.clear(); std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); - std::sort(sortedBindings.begin(), sortedBindings.end(), - [](const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) - { - return a.data()->binding < b.data()->binding; - }); + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); hasSlottedResource = false; hasDynamicOffset = false; - for (const QRhiShaderResourceBinding &binding : qAsConst(sortedBindings)) { - const QRhiShaderResourceBinding::Data *b = binding.data(); + for (const QRhiShaderResourceBinding &binding : std::as_const(sortedBindings)) { + const QRhiShaderResourceBinding::Data *b = QRhiImplementation::shaderResourceBindingData(binding); if (b->type == QRhiShaderResourceBinding::UniformBuffer && b->u.ubuf.buf) { if (QRHI_RES(QVkBuffer, b->u.ubuf.buf)->type() == QRhiBuffer::Dynamic) hasSlottedResource = true; @@ -6568,13 +7400,12 @@ bool QVkShaderResourceBindings::create() } QVarLengthArray<VkDescriptorSetLayoutBinding, 4> vkbindings; - for (const QRhiShaderResourceBinding &binding : qAsConst(sortedBindings)) { - const QRhiShaderResourceBinding::Data *b = binding.data(); - VkDescriptorSetLayoutBinding vkbinding; - memset(&vkbinding, 0, sizeof(vkbinding)); + for (const QRhiShaderResourceBinding &binding : std::as_const(sortedBindings)) { + const QRhiShaderResourceBinding::Data *b = QRhiImplementation::shaderResourceBindingData(binding); + VkDescriptorSetLayoutBinding vkbinding = {}; vkbinding.binding = uint32_t(b->binding); vkbinding.descriptorType = toVkDescriptorType(b); - if (b->type == QRhiShaderResourceBinding::SampledTexture) + if (b->type == QRhiShaderResourceBinding::SampledTexture || b->type == QRhiShaderResourceBinding::Texture) vkbinding.descriptorCount = b->u.stex.count; else vkbinding.descriptorCount = 1; @@ -6582,10 +7413,9 @@ bool QVkShaderResourceBindings::create() vkbindings.append(vkbinding); } - VkDescriptorSetLayoutCreateInfo layoutInfo; - memset(&layoutInfo, 0, sizeof(layoutInfo)); + VkDescriptorSetLayoutCreateInfo layoutInfo = {}; layoutInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO; - layoutInfo.bindingCount = uint32_t(vkbindings.count()); + layoutInfo.bindingCount = uint32_t(vkbindings.size()); layoutInfo.pBindings = vkbindings.constData(); VkResult err = rhiD->df->vkCreateDescriptorSetLayout(rhiD->dev, &layoutInfo, nullptr, &layout); @@ -6594,8 +7424,7 @@ bool QVkShaderResourceBindings::create() return false; } - VkDescriptorSetAllocateInfo allocInfo; - memset(&allocInfo, 0, sizeof(allocInfo)); + VkDescriptorSetAllocateInfo allocInfo = {}; allocInfo.sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO; allocInfo.descriptorSetCount = QVK_FRAMES_IN_FLIGHT; VkDescriptorSetLayout layouts[QVK_FRAMES_IN_FLIGHT]; @@ -6606,7 +7435,7 @@ bool QVkShaderResourceBindings::create() return false; for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { - boundResourceData[i].resize(sortedBindings.count()); + boundResourceData[i].resize(sortedBindings.size()); for (BoundResourceData &bd : boundResourceData[i]) memset(&bd, 0, sizeof(BoundResourceData)); } @@ -6617,6 +7446,31 @@ bool QVkShaderResourceBindings::create() return true; } +void QVkShaderResourceBindings::updateResources(UpdateFlags flags) +{ + sortedBindings.clear(); + std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); + if (!flags.testFlag(BindingsAreSorted)) + std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); + + // Reset the state tracking table too - it can deal with assigning a + // different QRhiBuffer/Texture/Sampler for a binding point, but it cannot + // detect changes in the associated data, such as the buffer offset. And + // just like after a create(), a call to updateResources() may lead to now + // specifying a different offset for the same QRhiBuffer for a given binding + // point. The same applies to other type of associated data that is not part + // of the layout, such as the mip level for a StorageImage. Instead of + // complicating the checks in setShaderResources(), reset the table here + // just like we do in create(). + for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { + Q_ASSERT(boundResourceData[i].size() == sortedBindings.size()); + for (BoundResourceData &bd : boundResourceData[i]) + memset(&bd, 0, sizeof(BoundResourceData)); + } + + generation += 1; +} + QVkGraphicsPipeline::QVkGraphicsPipeline(QRhiImplementation *rhi) : QRhiGraphicsPipeline(rhi) { @@ -6643,9 +7497,10 @@ void QVkGraphicsPipeline::destroy() layout = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkGraphicsPipeline::create() @@ -6654,14 +7509,14 @@ bool QVkGraphicsPipeline::create() destroy(); QRHI_RES_RHI(QRhiVulkan); + rhiD->pipelineCreationStart(); if (!rhiD->sanityCheckGraphicsPipeline(this)) return false; if (!rhiD->ensurePipelineCache()) return false; - VkPipelineLayoutCreateInfo pipelineLayoutInfo; - memset(&pipelineLayoutInfo, 0, sizeof(pipelineLayoutInfo)); + VkPipelineLayoutCreateInfo pipelineLayoutInfo = {}; pipelineLayoutInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; pipelineLayoutInfo.setLayoutCount = 1; QVkShaderResourceBindings *srbD = QRHI_RES(QVkShaderResourceBindings, m_shaderResourceBindings); @@ -6673,8 +7528,7 @@ bool QVkGraphicsPipeline::create() return false; } - VkGraphicsPipelineCreateInfo pipelineInfo; - memset(&pipelineInfo, 0, sizeof(pipelineInfo)); + VkGraphicsPipelineCreateInfo pipelineInfo = {}; pipelineInfo.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO; QVarLengthArray<VkShaderModule, 4> shaders; @@ -6689,8 +7543,7 @@ bool QVkGraphicsPipeline::create() VkShaderModule shader = rhiD->createShader(spirv.shader()); if (shader) { shaders.append(shader); - VkPipelineShaderStageCreateInfo shaderInfo; - memset(&shaderInfo, 0, sizeof(shaderInfo)); + VkPipelineShaderStageCreateInfo shaderInfo = {}; shaderInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; shaderInfo.stage = toVkShaderStage(shaderStage.type()); shaderInfo.module = shader; @@ -6698,11 +7551,13 @@ bool QVkGraphicsPipeline::create() shaderStageCreateInfos.append(shaderInfo); } } - pipelineInfo.stageCount = uint32_t(shaderStageCreateInfos.count()); + pipelineInfo.stageCount = uint32_t(shaderStageCreateInfos.size()); pipelineInfo.pStages = shaderStageCreateInfos.constData(); QVarLengthArray<VkVertexInputBindingDescription, 4> vertexBindings; +#ifdef VK_EXT_vertex_attribute_divisor QVarLengthArray<VkVertexInputBindingDivisorDescriptionEXT> nonOneStepRates; +#endif int bindingIndex = 0; for (auto it = m_vertexInputLayout.cbeginBindings(), itEnd = m_vertexInputLayout.cendBindings(); it != itEnd; ++it, ++bindingIndex) @@ -6714,9 +7569,12 @@ bool QVkGraphicsPipeline::create() ? VK_VERTEX_INPUT_RATE_VERTEX : VK_VERTEX_INPUT_RATE_INSTANCE }; if (it->classification() == QRhiVertexInputBinding::PerInstance && it->instanceStepRate() != 1) { +#ifdef VK_EXT_vertex_attribute_divisor if (rhiD->caps.vertexAttribDivisor) { - nonOneStepRates.append({ uint32_t(bindingIndex), uint32_t(it->instanceStepRate()) }); - } else { + nonOneStepRates.append({ uint32_t(bindingIndex), it->instanceStepRate() }); + } else +#endif + { qWarning("QRhiVulkan: Instance step rates other than 1 not supported without " "VK_EXT_vertex_attribute_divisor on the device and " "VK_KHR_get_physical_device_properties2 on the instance"); @@ -6736,21 +7594,21 @@ bool QVkGraphicsPipeline::create() }; vertexAttributes.append(attributeInfo); } - VkPipelineVertexInputStateCreateInfo vertexInputInfo; - memset(&vertexInputInfo, 0, sizeof(vertexInputInfo)); + VkPipelineVertexInputStateCreateInfo vertexInputInfo = {}; vertexInputInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO; - vertexInputInfo.vertexBindingDescriptionCount = uint32_t(vertexBindings.count()); + vertexInputInfo.vertexBindingDescriptionCount = uint32_t(vertexBindings.size()); vertexInputInfo.pVertexBindingDescriptions = vertexBindings.constData(); - vertexInputInfo.vertexAttributeDescriptionCount = uint32_t(vertexAttributes.count()); + vertexInputInfo.vertexAttributeDescriptionCount = uint32_t(vertexAttributes.size()); vertexInputInfo.pVertexAttributeDescriptions = vertexAttributes.constData(); - VkPipelineVertexInputDivisorStateCreateInfoEXT divisorInfo; +#ifdef VK_EXT_vertex_attribute_divisor + VkPipelineVertexInputDivisorStateCreateInfoEXT divisorInfo = {}; if (!nonOneStepRates.isEmpty()) { - memset(&divisorInfo, 0, sizeof(divisorInfo)); - divisorInfo.sType = VkStructureType(1000190001); // VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT - divisorInfo.vertexBindingDivisorCount = uint32_t(nonOneStepRates.count()); + divisorInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT; + divisorInfo.vertexBindingDivisorCount = uint32_t(nonOneStepRates.size()); divisorInfo.pVertexBindingDivisors = nonOneStepRates.constData(); vertexInputInfo.pNext = &divisorInfo; } +#endif pipelineInfo.pVertexInputState = &vertexInputInfo; QVarLengthArray<VkDynamicState, 8> dynEnable; @@ -6761,28 +7619,52 @@ bool QVkGraphicsPipeline::create() if (m_flags.testFlag(QRhiGraphicsPipeline::UsesStencilRef)) dynEnable << VK_DYNAMIC_STATE_STENCIL_REFERENCE; - VkPipelineDynamicStateCreateInfo dynamicInfo; - memset(&dynamicInfo, 0, sizeof(dynamicInfo)); + VkPipelineDynamicStateCreateInfo dynamicInfo = {}; dynamicInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO; - dynamicInfo.dynamicStateCount = uint32_t(dynEnable.count()); + dynamicInfo.dynamicStateCount = uint32_t(dynEnable.size()); dynamicInfo.pDynamicStates = dynEnable.constData(); pipelineInfo.pDynamicState = &dynamicInfo; - VkPipelineViewportStateCreateInfo viewportInfo; - memset(&viewportInfo, 0, sizeof(viewportInfo)); + VkPipelineViewportStateCreateInfo viewportInfo = {}; viewportInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO; viewportInfo.viewportCount = viewportInfo.scissorCount = 1; pipelineInfo.pViewportState = &viewportInfo; - VkPipelineInputAssemblyStateCreateInfo inputAsmInfo; - memset(&inputAsmInfo, 0, sizeof(inputAsmInfo)); + VkPipelineInputAssemblyStateCreateInfo inputAsmInfo = {}; inputAsmInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO; inputAsmInfo.topology = toVkTopology(m_topology); inputAsmInfo.primitiveRestartEnable = (m_topology == TriangleStrip || m_topology == LineStrip); pipelineInfo.pInputAssemblyState = &inputAsmInfo; - VkPipelineRasterizationStateCreateInfo rastInfo; - memset(&rastInfo, 0, sizeof(rastInfo)); + VkPipelineTessellationStateCreateInfo tessInfo = {}; +#ifdef VK_VERSION_1_1 + VkPipelineTessellationDomainOriginStateCreateInfo originInfo = {}; +#endif + if (m_topology == Patches) { + tessInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_STATE_CREATE_INFO; + tessInfo.patchControlPoints = uint32_t(qMax(1, m_patchControlPointCount)); + + // To be able to use the same tess.evaluation shader with both OpenGL + // and Vulkan, flip the tessellation domain origin to be lower left. + // This allows declaring the winding order in the shader to be CCW and + // still have it working with both APIs. This requires Vulkan 1.1 (or + // VK_KHR_maintenance2 but don't bother with that). +#ifdef VK_VERSION_1_1 + if (rhiD->caps.apiVersion >= QVersionNumber(1, 1)) { + originInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_TESSELLATION_DOMAIN_ORIGIN_STATE_CREATE_INFO; + originInfo.domainOrigin = VK_TESSELLATION_DOMAIN_ORIGIN_LOWER_LEFT; + tessInfo.pNext = &originInfo; + } else { + qWarning("Proper tessellation support requires Vulkan 1.1 or newer, leaving domain origin unset"); + } +#else + qWarning("QRhi was built without Vulkan 1.1 headers, this is not sufficient for proper tessellation support"); +#endif + + pipelineInfo.pTessellationState = &tessInfo; + } + + VkPipelineRasterizationStateCreateInfo rastInfo = {}; rastInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO; rastInfo.cullMode = toVkCullMode(m_cullMode); rastInfo.frontFace = toVkFrontFace(m_frontFace); @@ -6792,16 +7674,15 @@ bool QVkGraphicsPipeline::create() rastInfo.depthBiasSlopeFactor = m_slopeScaledDepthBias; } rastInfo.lineWidth = rhiD->caps.wideLines ? m_lineWidth : 1.0f; + rastInfo.polygonMode = toVkPolygonMode(m_polygonMode); pipelineInfo.pRasterizationState = &rastInfo; - VkPipelineMultisampleStateCreateInfo msInfo; - memset(&msInfo, 0, sizeof(msInfo)); + VkPipelineMultisampleStateCreateInfo msInfo = {}; msInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; - msInfo.rasterizationSamples = rhiD->effectiveSampleCount(m_sampleCount); + msInfo.rasterizationSamples = rhiD->effectiveSampleCountBits(m_sampleCount); pipelineInfo.pMultisampleState = &msInfo; - VkPipelineDepthStencilStateCreateInfo dsInfo; - memset(&dsInfo, 0, sizeof(dsInfo)); + VkPipelineDepthStencilStateCreateInfo dsInfo = {}; dsInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO; dsInfo.depthTestEnable = m_depthTest; dsInfo.depthWriteEnable = m_depthWrite; @@ -6817,13 +7698,11 @@ bool QVkGraphicsPipeline::create() } pipelineInfo.pDepthStencilState = &dsInfo; - VkPipelineColorBlendStateCreateInfo blendInfo; - memset(&blendInfo, 0, sizeof(blendInfo)); + VkPipelineColorBlendStateCreateInfo blendInfo = {}; blendInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO; QVarLengthArray<VkPipelineColorBlendAttachmentState, 4> vktargetBlends; - for (const QRhiGraphicsPipeline::TargetBlend &b : qAsConst(m_targetBlends)) { - VkPipelineColorBlendAttachmentState blend; - memset(&blend, 0, sizeof(blend)); + for (const QRhiGraphicsPipeline::TargetBlend &b : std::as_const(m_targetBlends)) { + VkPipelineColorBlendAttachmentState blend = {}; blend.blendEnable = b.enable; blend.srcColorBlendFactor = toVkBlendFactor(b.srcColor); blend.dstColorBlendFactor = toVkBlendFactor(b.dstColor); @@ -6835,13 +7714,12 @@ bool QVkGraphicsPipeline::create() vktargetBlends.append(blend); } if (vktargetBlends.isEmpty()) { - VkPipelineColorBlendAttachmentState blend; - memset(&blend, 0, sizeof(blend)); + VkPipelineColorBlendAttachmentState blend = {}; blend.colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT | VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT; vktargetBlends.append(blend); } - blendInfo.attachmentCount = uint32_t(vktargetBlends.count()); + blendInfo.attachmentCount = uint32_t(vktargetBlends.size()); blendInfo.pAttachments = vktargetBlends.constData(); pipelineInfo.pColorBlendState = &blendInfo; @@ -6860,6 +7738,7 @@ bool QVkGraphicsPipeline::create() return false; } + rhiD->pipelineCreationEnd(); lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -6892,9 +7771,10 @@ void QVkComputePipeline::destroy() layout = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); - rhiD->releaseQueue.append(e); - - rhiD->unregisterResource(this); + if (rhiD) { + rhiD->releaseQueue.append(e); + rhiD->unregisterResource(this); + } } bool QVkComputePipeline::create() @@ -6903,11 +7783,11 @@ bool QVkComputePipeline::create() destroy(); QRHI_RES_RHI(QRhiVulkan); + rhiD->pipelineCreationStart(); if (!rhiD->ensurePipelineCache()) return false; - VkPipelineLayoutCreateInfo pipelineLayoutInfo; - memset(&pipelineLayoutInfo, 0, sizeof(pipelineLayoutInfo)); + VkPipelineLayoutCreateInfo pipelineLayoutInfo = {}; pipelineLayoutInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO; pipelineLayoutInfo.setLayoutCount = 1; QVkShaderResourceBindings *srbD = QRHI_RES(QVkShaderResourceBindings, m_shaderResourceBindings); @@ -6919,8 +7799,7 @@ bool QVkComputePipeline::create() return false; } - VkComputePipelineCreateInfo pipelineInfo; - memset(&pipelineInfo, 0, sizeof(pipelineInfo)); + VkComputePipelineCreateInfo pipelineInfo = {}; pipelineInfo.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO; pipelineInfo.layout = layout; @@ -6939,8 +7818,7 @@ bool QVkComputePipeline::create() return false; } VkShaderModule shader = rhiD->createShader(spirv.shader()); - VkPipelineShaderStageCreateInfo shaderInfo; - memset(&shaderInfo, 0, sizeof(shaderInfo)); + VkPipelineShaderStageCreateInfo shaderInfo = {}; shaderInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO; shaderInfo.stage = VK_SHADER_STAGE_COMPUTE_BIT; shaderInfo.module = shader; @@ -6954,6 +7832,7 @@ bool QVkComputePipeline::create() return false; } + rhiD->pipelineCreationEnd(); lastActiveFrameSlot = -1; generation += 1; rhiD->registerResource(this); @@ -6998,7 +7877,8 @@ const QRhiNativeHandles *QVkCommandBuffer::nativeHandles() QVkSwapChain::QVkSwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), - rtWrapper(rhi), + rtWrapper(rhi, this), + rtWrapperRight(rhi, this), cbWrapper(rhi) { } @@ -7014,8 +7894,10 @@ void QVkSwapChain::destroy() return; QRHI_RES_RHI(QRhiVulkan); - rhiD->swapchains.remove(this); - rhiD->releaseSwapChainResources(this); + if (rhiD) { + rhiD->swapchains.remove(this); + rhiD->releaseSwapChainResources(this); + } for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) { QVkSwapChain::FrameResources &frame(frameRes[i]); @@ -7025,10 +7907,8 @@ void QVkSwapChain::destroy() surface = lastConnectedSurface = VK_NULL_HANDLE; - QRHI_PROF; - QRHI_PROF_F(releaseSwapChain(this)); - - rhiD->unregisterResource(this); + if (rhiD) + rhiD->unregisterResource(this); } QRhiCommandBuffer *QVkSwapChain::currentFrameCommandBuffer() @@ -7041,6 +7921,11 @@ QRhiRenderTarget *QVkSwapChain::currentFrameRenderTarget() return &rtWrapper; } +QRhiRenderTarget *QVkSwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + return !stereo || targetBuffer == StereoTargetBuffer::LeftBuffer ? &rtWrapper : &rtWrapperRight; +} + QSize QVkSwapChain::surfacePixelSize() { if (!ensureSurface()) @@ -7048,8 +7933,7 @@ QSize QVkSwapChain::surfacePixelSize() // The size from the QWindow may not exactly match the surface... so if a // size is reported from the surface, use that. - VkSurfaceCapabilitiesKHR surfaceCaps; - memset(&surfaceCaps, 0, sizeof(surfaceCaps)); + VkSurfaceCapabilitiesKHR surfaceCaps = {}; QRHI_RES_RHI(QRhiVulkan); rhiD->vkGetPhysicalDeviceSurfaceCapabilitiesKHR(rhiD->physDev, surface, &surfaceCaps); VkExtent2D bufferSize = surfaceCaps.currentExtent; @@ -7060,6 +7944,52 @@ QSize QVkSwapChain::surfacePixelSize() return QSize(int(bufferSize.width), int(bufferSize.height)); } +static inline bool hdrFormatMatchesVkSurfaceFormat(QRhiSwapChain::Format f, const VkSurfaceFormatKHR &s) +{ + switch (f) { + case QRhiSwapChain::HDRExtendedSrgbLinear: + return s.format == VK_FORMAT_R16G16B16A16_SFLOAT + && s.colorSpace == VK_COLOR_SPACE_EXTENDED_SRGB_LINEAR_EXT; + case QRhiSwapChain::HDR10: + return (s.format == VK_FORMAT_A2B10G10R10_UNORM_PACK32 || s.format == VK_FORMAT_A2R10G10B10_UNORM_PACK32) + && s.colorSpace == VK_COLOR_SPACE_HDR10_ST2084_EXT; + case QRhiSwapChain::HDRExtendedDisplayP3Linear: + return s.format == VK_FORMAT_R16G16B16A16_SFLOAT + && s.colorSpace == VK_COLOR_SPACE_DISPLAY_P3_LINEAR_EXT; + default: + break; + } + return false; +} + +bool QVkSwapChain::isFormatSupported(Format f) +{ + if (f == SDR) + return true; + + if (!m_window) { + qWarning("Attempted to call isFormatSupported() without a window set"); + return false; + } + + // we may be called before create so query the surface + VkSurfaceKHR surf = QVulkanInstance::surfaceForWindow(m_window); + + QRHI_RES_RHI(QRhiVulkan); + uint32_t formatCount = 0; + rhiD->vkGetPhysicalDeviceSurfaceFormatsKHR(rhiD->physDev, surf, &formatCount, nullptr); + QVarLengthArray<VkSurfaceFormatKHR, 8> formats(formatCount); + if (formatCount) { + rhiD->vkGetPhysicalDeviceSurfaceFormatsKHR(rhiD->physDev, surf, &formatCount, formats.data()); + for (uint32_t i = 0; i < formatCount; ++i) { + if (hdrFormatMatchesVkSurfaceFormat(f, formats[i])) + return true; + } + } + + return false; +} + QRhiRenderPassDescriptor *QVkSwapChain::newCompatibleRenderPassDescriptor() { // not yet built so cannot rely on data computed in createOrResize() @@ -7079,6 +8009,7 @@ QRhiRenderPassDescriptor *QVkSwapChain::newCompatibleRenderPassDescriptor() } rp->ownsRp = true; + rp->updateSerializedFormat(); rhiD->registerResource(rp); return rp; } @@ -7118,27 +8049,9 @@ bool QVkSwapChain::ensureSurface() surface = surf; QRHI_RES_RHI(QRhiVulkan); - if (rhiD->gfxQueueFamilyIdx != -1) { - if (!rhiD->inst->supportsPresent(rhiD->physDev, uint32_t(rhiD->gfxQueueFamilyIdx), m_window)) { - qWarning("Presenting not supported on this window"); - return false; - } - } - - if (!rhiD->vkGetPhysicalDeviceSurfaceCapabilitiesKHR) { - rhiD->vkGetPhysicalDeviceSurfaceCapabilitiesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR>( - rhiD->inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfaceCapabilitiesKHR")); - rhiD->vkGetPhysicalDeviceSurfaceFormatsKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfaceFormatsKHR>( - rhiD->inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfaceFormatsKHR")); - rhiD->vkGetPhysicalDeviceSurfacePresentModesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfacePresentModesKHR>( - rhiD->inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfacePresentModesKHR")); - if (!rhiD->vkGetPhysicalDeviceSurfaceCapabilitiesKHR - || !rhiD->vkGetPhysicalDeviceSurfaceFormatsKHR - || !rhiD->vkGetPhysicalDeviceSurfacePresentModesKHR) - { - qWarning("Physical device surface queries not available"); - return false; - } + if (!rhiD->inst->supportsPresent(rhiD->physDev, rhiD->gfxQueueFamilyIdx, m_window)) { + qWarning("Presenting not supported on this window"); + return false; } quint32 formatCount = 0; @@ -7147,16 +8060,23 @@ bool QVkSwapChain::ensureSurface() if (formatCount) rhiD->vkGetPhysicalDeviceSurfaceFormatsKHR(rhiD->physDev, surface, &formatCount, formats.data()); + // See if there is a better match than the default BGRA8 format. (but if + // not, we will stick to the default) const bool srgbRequested = m_flags.testFlag(sRGB); for (int i = 0; i < int(formatCount); ++i) { - if (formats[i].format != VK_FORMAT_UNDEFINED && srgbRequested == isSrgbFormat(formats[i].format)) { - colorFormat = formats[i].format; - colorSpace = formats[i].colorSpace; - break; + if (formats[i].format != VK_FORMAT_UNDEFINED) { + bool ok = srgbRequested == isSrgbFormat(formats[i].format); + if (m_format != SDR) + ok &= hdrFormatMatchesVkSurfaceFormat(m_format, formats[i]); + if (ok) { + colorFormat = formats[i].format; + colorSpace = formats[i].colorSpace; + break; + } } } - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); quint32 presModeCount = 0; rhiD->vkGetPhysicalDeviceSurfacePresentModesKHR(rhiD->physDev, surface, &presModeCount, nullptr); @@ -7212,6 +8132,7 @@ bool QVkSwapChain::createOrResize() if (!m_renderPassDesc) qWarning("QVkSwapChain: No renderpass descriptor set. See newCompatibleRenderPassDescriptor() and setRenderPassDescriptor()."); + rtWrapper.setRenderPassDescriptor(m_renderPassDesc); // for the public getter in QRhiRenderTarget rtWrapper.d.rp = QRHI_RES(QVkRenderPassDescriptor, m_renderPassDesc); Q_ASSERT(rtWrapper.d.rp && rtWrapper.d.rp->rp); @@ -7226,6 +8147,7 @@ bool QVkSwapChain::createOrResize() rtWrapper.d.dsAttCount = 0; ds = nullptr; } + rtWrapper.d.dsResolveAttCount = 0; if (samples > VK_SAMPLE_COUNT_1_BIT) rtWrapper.d.resolveAttCount = 1; else @@ -7239,11 +8161,10 @@ bool QVkSwapChain::createOrResize() samples > VK_SAMPLE_COUNT_1_BIT ? image.imageView : VK_NULL_HANDLE }; - VkFramebufferCreateInfo fbInfo; - memset(&fbInfo, 0, sizeof(fbInfo)); + VkFramebufferCreateInfo fbInfo = {}; fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; fbInfo.renderPass = rtWrapper.d.rp->rp; - fbInfo.attachmentCount = uint32_t(rtWrapper.d.colorAttCount + rtWrapper.d.dsAttCount + rtWrapper.d.resolveAttCount); + fbInfo.attachmentCount = uint32_t(rtWrapper.d.colorAttCount + rtWrapper.d.dsAttCount + rtWrapper.d.resolveAttCount + rtWrapper.d.dsResolveAttCount); fbInfo.pAttachments = views; fbInfo.width = uint32_t(pixelSize.width()); fbInfo.height = uint32_t(pixelSize.height()); @@ -7256,10 +8177,57 @@ bool QVkSwapChain::createOrResize() } } - frameCount = 0; + if (stereo) { + rtWrapperRight.setRenderPassDescriptor( + m_renderPassDesc); // for the public getter in QRhiRenderTarget + rtWrapperRight.d.rp = QRHI_RES(QVkRenderPassDescriptor, m_renderPassDesc); + Q_ASSERT(rtWrapperRight.d.rp && rtWrapperRight.d.rp->rp); - QRHI_PROF; - QRHI_PROF_F(resizeSwapChain(this, QVK_FRAMES_IN_FLIGHT, samples > VK_SAMPLE_COUNT_1_BIT ? QVK_FRAMES_IN_FLIGHT : 0, samples)); + rtWrapperRight.d.pixelSize = pixelSize; + rtWrapperRight.d.dpr = float(window->devicePixelRatio()); + rtWrapperRight.d.sampleCount = samples; + rtWrapperRight.d.colorAttCount = 1; + if (m_depthStencil) { + rtWrapperRight.d.dsAttCount = 1; + ds = QRHI_RES(QVkRenderBuffer, m_depthStencil); + } else { + rtWrapperRight.d.dsAttCount = 0; + ds = nullptr; + } + rtWrapperRight.d.dsResolveAttCount = 0; + if (samples > VK_SAMPLE_COUNT_1_BIT) + rtWrapperRight.d.resolveAttCount = 1; + else + rtWrapperRight.d.resolveAttCount = 0; + + for (int i = 0; i < bufferCount; ++i) { + QVkSwapChain::ImageResources &image(imageRes[i + bufferCount]); + VkImageView views[3] = { + // color, ds, resolve + samples > VK_SAMPLE_COUNT_1_BIT ? image.msaaImageView : image.imageView, + ds ? ds->imageView : VK_NULL_HANDLE, + samples > VK_SAMPLE_COUNT_1_BIT ? image.imageView : VK_NULL_HANDLE + }; + + VkFramebufferCreateInfo fbInfo = {}; + fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; + fbInfo.renderPass = rtWrapperRight.d.rp->rp; + fbInfo.attachmentCount = uint32_t(rtWrapperRight.d.colorAttCount + rtWrapperRight.d.dsAttCount + + rtWrapperRight.d.resolveAttCount + rtWrapperRight.d.dsResolveAttCount); + fbInfo.pAttachments = views; + fbInfo.width = uint32_t(pixelSize.width()); + fbInfo.height = uint32_t(pixelSize.height()); + fbInfo.layers = 1; + + VkResult err = rhiD->df->vkCreateFramebuffer(rhiD->dev, &fbInfo, nullptr, &image.fb); + if (err != VK_SUCCESS) { + qWarning("Failed to create framebuffer: %d", err); + return false; + } + } + } + + frameCount = 0; if (needsRegistration) rhiD->registerResource(this); diff --git a/src/gui/rhi/qrhivulkan_p.h b/src/gui/rhi/qrhivulkan_p.h index 4d20b93911..ad8687de5d 100644 --- a/src/gui/rhi/qrhivulkan_p.h +++ b/src/gui/rhi/qrhivulkan_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIVULKAN_H -#define QRHIVULKAN_H +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHIVULKAN_P_H +#define QRHIVULKAN_P_H // // W A R N I N G @@ -51,44 +15,1025 @@ // We mean it. // -#include <private/qrhi_p.h> -#include <QtGui/qvulkaninstance.h> // this is where vulkan.h gets pulled in +#include "qrhi_p.h" QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiVulkanInitParams : public QRhiInitParams +class QVulkanFunctions; +class QVulkanDeviceFunctions; + +static const int QVK_FRAMES_IN_FLIGHT = 2; + +static const int QVK_DESC_SETS_PER_POOL = 128; +static const int QVK_UNIFORM_BUFFERS_PER_POOL = 256; +static const int QVK_COMBINED_IMAGE_SAMPLERS_PER_POOL = 256; +static const int QVK_STORAGE_BUFFERS_PER_POOL = 128; +static const int QVK_STORAGE_IMAGES_PER_POOL = 128; + +static const int QVK_MAX_ACTIVE_TIMESTAMP_PAIRS = 16; + +// no vk_mem_alloc.h available here, void* is good enough +typedef void * QVkAlloc; +typedef void * QVkAllocator; + +struct QVkBuffer : public QRhiBuffer { - QVulkanInstance *inst = nullptr; - QWindow *window = nullptr; - QByteArrayList deviceExtensions; + QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QVkBuffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; - static QByteArrayList preferredInstanceExtensions(); - static QByteArrayList preferredExtensionsForImportedDevice(); + VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; + struct DynamicUpdate { + quint32 offset; + QRhiBufferData data; + }; + QVarLengthArray<DynamicUpdate, 16> pendingDynamicUpdates[QVK_FRAMES_IN_FLIGHT]; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + struct UsageState { + VkAccessFlags access = 0; + VkPipelineStageFlags stage = 0; + }; + UsageState usageState[QVK_FRAMES_IN_FLIGHT]; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; }; -struct Q_GUI_EXPORT QRhiVulkanNativeHandles : public QRhiNativeHandles +Q_DECLARE_TYPEINFO(QVkBuffer::DynamicUpdate, Q_RELOCATABLE_TYPE); + +struct QVkTexture; + +struct QVkRenderBuffer : public QRhiRenderBuffer { - // to import a physical device (always required) - VkPhysicalDevice physDev = VK_NULL_HANDLE; - // to import a device and queue - VkDevice dev = VK_NULL_HANDLE; - int gfxQueueFamilyIdx = -1; - int gfxQueueIdx = 0; - VkQueue gfxQueue = VK_NULL_HANDLE; - // and optionally, the mem allocator - void *vmemAllocator = nullptr; + QVkRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, Flags flags, + QRhiTexture::Format backingFormatHint); + ~QVkRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + VkDeviceMemory memory = VK_NULL_HANDLE; + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + VkSampleCountFlagBits samples; + QVkTexture *backingTexture = nullptr; + VkFormat vkformat; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkTexture : public QRhiTexture +{ + QVkTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QVkTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + void setNativeLayout(int layout) override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + VkImageView perLevelImageViewForLoadStore(int level); + + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + QVkAlloc imageAlloc = nullptr; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + VkImageView perLevelImageViews[QRhi::MAX_MIP_LEVELS]; + bool owns = true; + struct UsageState { + // no tracking of subresource layouts (some operations can keep + // subresources in different layouts for some time, but that does not + // need to be kept track of) + VkImageLayout layout; + VkAccessFlags access; + VkPipelineStageFlags stage; + }; + UsageState usageState; + VkFormat vkformat; + uint mipLevelCount = 0; + VkSampleCountFlagBits samples; + VkFormat viewFormat; + VkFormat viewFormatForSampling; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkSampler : public QRhiSampler +{ + QVkSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QVkSampler(); + void destroy() override; + bool create() override; + + VkSampler sampler = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkRenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QVkRenderPassDescriptor(QRhiImplementation *rhi); + ~QVkRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + const QRhiNativeHandles *nativeHandles() override; + + void updateSerializedFormat(); + + VkRenderPass rp = VK_NULL_HANDLE; + bool ownsRp = false; + QVarLengthArray<VkAttachmentDescription, 8> attDescs; + QVarLengthArray<VkAttachmentReference, 8> colorRefs; + QVarLengthArray<VkAttachmentReference, 8> resolveRefs; + QVarLengthArray<VkSubpassDependency, 2> subpassDeps; + bool hasDepthStencil = false; + bool hasDepthStencilResolve = false; + uint32_t multiViewCount = 0; + VkAttachmentReference dsRef; + VkAttachmentReference dsResolveRef; + QVector<quint32> serializedFormatData; + QRhiVulkanRenderPassNativeHandles nativeHandlesStruct; + int lastActiveFrameSlot = -1; +}; + +struct QVkRenderTargetData +{ + VkFramebuffer fb = VK_NULL_HANDLE; + QVkRenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + int resolveAttCount = 0; + int dsResolveAttCount = 0; + int multiViewCount = 0; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + static const int MAX_COLOR_ATTACHMENTS = 8; +}; + +struct QVkSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QVkSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QVkSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QVkRenderTargetData d; +}; + +struct QVkTextureRenderTarget : public QRhiTextureRenderTarget +{ + QVkTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QVkTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QVkRenderTargetData d; + VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView dsv = VK_NULL_HANDLE; + VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView resdsv = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + friend class QRhiVulkan; +}; + +struct QVkShaderResourceBindings : public QRhiShaderResourceBindings +{ + QVkShaderResourceBindings(QRhiImplementation *rhi); + ~QVkShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + bool hasSlottedResource = false; + bool hasDynamicOffset = false; + int poolIndex = -1; + VkDescriptorSetLayout layout = VK_NULL_HANDLE; + VkDescriptorSet descSets[QVK_FRAMES_IN_FLIGHT]; // multiple sets to support dynamic buffers + int lastActiveFrameSlot = -1; + uint generation = 0; + + // Keep track of the generation number of each referenced QRhi* to be able + // to detect that the underlying descriptor set became out of date and they + // need to be written again with the up-to-date VkBuffer etc. objects. + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData[QVK_FRAMES_IN_FLIGHT]; + + friend class QRhiVulkan; +}; + +Q_DECLARE_TYPEINFO(QVkShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); + +struct QVkGraphicsPipeline : public QRhiGraphicsPipeline +{ + QVkGraphicsPipeline(QRhiImplementation *rhi); + ~QVkGraphicsPipeline(); + void destroy() override; + bool create() override; + + VkPipelineLayout layout = VK_NULL_HANDLE; + VkPipeline pipeline = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkComputePipeline : public QRhiComputePipeline +{ + QVkComputePipeline(QRhiImplementation *rhi); + ~QVkComputePipeline(); + void destroy() override; + bool create() override; + + VkPipelineLayout layout = VK_NULL_HANDLE; + VkPipeline pipeline = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkCommandBuffer : public QRhiCommandBuffer +{ + QVkCommandBuffer(QRhiImplementation *rhi); + ~QVkCommandBuffer(); + void destroy() override; + + const QRhiNativeHandles *nativeHandles(); + + VkCommandBuffer cb = VK_NULL_HANDLE; // primary + QRhiVulkanCommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + void resetState() { + recordingPass = NoPass; + passUsesSecondaryCb = false; + lastGpuTime = 0; + currentTarget = nullptr; + activeSecondaryCbStack.clear(); + resetCommands(); + resetCachedState(); + } + + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentDescSetSlot = -1; + currentIndexBuffer = VK_NULL_HANDLE; + currentIndexOffset = 0; + currentIndexFormat = VK_INDEX_TYPE_UINT16; + memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); + memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); + inExternal = false; + } + + PassType recordingPass; + bool passUsesSecondaryCb; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + int currentDescSetSlot; + VkBuffer currentIndexBuffer; + quint32 currentIndexOffset; + VkIndexType currentIndexFormat; + static const int VERTEX_INPUT_RESOURCE_SLOT_COUNT = 32; + VkBuffer currentVertexBuffers[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + quint32 currentVertexOffsets[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + QVarLengthArray<VkCommandBuffer, 4> activeSecondaryCbStack; + bool inExternal; + + struct { + QHash<QRhiResource *, QPair<VkAccessFlags, bool> > writtenResources; + void reset() { + writtenResources.clear(); + } + } computePassState; + + struct Command { + enum Cmd { + CopyBuffer, + CopyBufferToImage, + CopyImage, + CopyImageToBuffer, + ImageBarrier, + BufferBarrier, + BlitImage, + BeginRenderPass, + EndRenderPass, + BindPipeline, + BindDescriptorSet, + BindVertexBuffer, + BindIndexBuffer, + SetViewport, + SetScissor, + SetBlendConstants, + SetStencilRef, + Draw, + DrawIndexed, + DebugMarkerBegin, + DebugMarkerEnd, + DebugMarkerInsert, + TransitionPassResources, + Dispatch, + ExecuteSecondary + }; + Cmd cmd; + + union Args { + struct { + VkBuffer src; + VkBuffer dst; + VkBufferCopy desc; + } copyBuffer; + struct { + VkBuffer src; + VkImage dst; + VkImageLayout dstLayout; + int count; + int bufferImageCopyIndex; + } copyBufferToImage; + struct { + VkImage src; + VkImageLayout srcLayout; + VkImage dst; + VkImageLayout dstLayout; + VkImageCopy desc; + } copyImage; + struct { + VkImage src; + VkImageLayout srcLayout; + VkBuffer dst; + VkBufferImageCopy desc; + } copyImageToBuffer; + struct { + VkPipelineStageFlags srcStageMask; + VkPipelineStageFlags dstStageMask; + int count; + int index; + } imageBarrier; + struct { + VkPipelineStageFlags srcStageMask; + VkPipelineStageFlags dstStageMask; + int count; + int index; + } bufferBarrier; + struct { + VkImage src; + VkImageLayout srcLayout; + VkImage dst; + VkImageLayout dstLayout; + VkFilter filter; + VkImageBlit desc; + } blitImage; + struct { + VkRenderPassBeginInfo desc; + int clearValueIndex; + bool useSecondaryCb; + } beginRenderPass; + struct { + } endRenderPass; + struct { + VkPipelineBindPoint bindPoint; + VkPipeline pipeline; + } bindPipeline; + struct { + VkPipelineBindPoint bindPoint; + VkPipelineLayout pipelineLayout; + VkDescriptorSet descSet; + int dynamicOffsetCount; + int dynamicOffsetIndex; + } bindDescriptorSet; + struct { + int startBinding; + int count; + int vertexBufferIndex; + int vertexBufferOffsetIndex; + } bindVertexBuffer; + struct { + VkBuffer buf; + VkDeviceSize ofs; + VkIndexType type; + } bindIndexBuffer; + struct { + VkViewport viewport; + } setViewport; + struct { + VkRect2D scissor; + } setScissor; + struct { + float c[4]; + } setBlendConstants; + struct { + uint32_t ref; + } setStencilRef; + struct { + uint32_t vertexCount; + uint32_t instanceCount; + uint32_t firstVertex; + uint32_t firstInstance; + } draw; + struct { + uint32_t indexCount; + uint32_t instanceCount; + uint32_t firstIndex; + int32_t vertexOffset; + uint32_t firstInstance; + } drawIndexed; + struct { +#ifdef VK_EXT_debug_utils + VkDebugUtilsLabelEXT label; + int labelNameIndex; +#endif + } debugMarkerBegin; + struct { + } debugMarkerEnd; + struct { +#ifdef VK_EXT_debug_utils + VkDebugUtilsLabelEXT label; + int labelNameIndex; +#endif + } debugMarkerInsert; + struct { + int trackerIndex; + } transitionResources; + struct { + int x, y, z; + } dispatch; + struct { + VkCommandBuffer cb; + } executeSecondary; + } args; + }; + + QRhiBackendCommandList<Command> commands; + QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; + int currentPassResTrackerIndex; + + void resetCommands() { + commands.reset(); + resetPools(); + + passResTrackers.clear(); + currentPassResTrackerIndex = -1; + } + + void resetPools() { + pools.clearValue.clear(); + pools.bufferImageCopy.clear(); + pools.dynamicOffset.clear(); + pools.vertexBuffer.clear(); + pools.vertexBufferOffset.clear(); + pools.debugMarkerData.clear(); + pools.imageBarrier.clear(); + pools.bufferBarrier.clear(); + } + + struct { + QVarLengthArray<VkClearValue, 4> clearValue; + QVarLengthArray<VkBufferImageCopy, 16> bufferImageCopy; + QVarLengthArray<uint32_t, 4> dynamicOffset; + QVarLengthArray<VkBuffer, 4> vertexBuffer; + QVarLengthArray<VkDeviceSize, 4> vertexBufferOffset; + QVarLengthArray<QByteArray, 4> debugMarkerData; + QVarLengthArray<VkImageMemoryBarrier, 8> imageBarrier; + QVarLengthArray<VkBufferMemoryBarrier, 8> bufferBarrier; + } pools; + + friend class QRhiVulkan; }; -struct Q_GUI_EXPORT QRhiVulkanCommandBufferNativeHandles : public QRhiNativeHandles +struct QVkSwapChain : public QRhiSwapChain { - VkCommandBuffer commandBuffer = VK_NULL_HANDLE; + QVkSwapChain(QRhiImplementation *rhi); + ~QVkSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + bool ensureSurface(); + + static const quint32 EXPECTED_MAX_BUFFER_COUNT = 4; + + QWindow *window = nullptr; + QSize pixelSize; + bool supportsReadback = false; + bool stereo = false; + VkSwapchainKHR sc = VK_NULL_HANDLE; + int bufferCount = 0; + VkSurfaceKHR surface = VK_NULL_HANDLE; + VkSurfaceKHR lastConnectedSurface = VK_NULL_HANDLE; + VkFormat colorFormat = VK_FORMAT_B8G8R8A8_UNORM; + VkColorSpaceKHR colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; + QVkRenderBuffer *ds = nullptr; + VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT; + QVarLengthArray<VkPresentModeKHR, 8> supportedPresentationModes; + VkDeviceMemory msaaImageMem = VK_NULL_HANDLE; + QVkSwapChainRenderTarget rtWrapper; + QVkSwapChainRenderTarget rtWrapperRight; + QVkCommandBuffer cbWrapper; + + struct ImageResources { + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + VkFramebuffer fb = VK_NULL_HANDLE; + VkImage msaaImage = VK_NULL_HANDLE; + VkImageView msaaImageView = VK_NULL_HANDLE; + enum LastUse { + ScImageUseNone, + ScImageUseRender, + ScImageUseTransferSource + }; + LastUse lastUse = ScImageUseNone; + }; + QVarLengthArray<ImageResources, EXPECTED_MAX_BUFFER_COUNT> imageRes; + + struct FrameResources { + VkFence imageFence = VK_NULL_HANDLE; + bool imageFenceWaitable = false; + VkSemaphore imageSem = VK_NULL_HANDLE; + VkSemaphore drawSem = VK_NULL_HANDLE; + bool imageAcquired = false; + bool imageSemWaitable = false; + VkFence cmdFence = VK_NULL_HANDLE; + bool cmdFenceWaitable = false; + VkCommandBuffer cmdBuf = VK_NULL_HANDLE; // primary + int timestampQueryIndex = -1; + } frameRes[QVK_FRAMES_IN_FLIGHT]; + + quint32 currentImageIndex = 0; // index in imageRes + quint32 currentFrameSlot = 0; // index in frameRes + int frameCount = 0; + + friend class QRhiVulkan; }; -struct Q_GUI_EXPORT QRhiVulkanRenderPassNativeHandles : public QRhiNativeHandles +class QRhiVulkan : public QRhiImplementation { - VkRenderPass renderPass = VK_NULL_HANDLE; +public: + QRhiVulkan(QRhiVulkanInitParams *params, QRhiVulkanNativeHandles *importParams = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + VkResult createDescriptorPool(VkDescriptorPool *pool); + bool allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, VkDescriptorSet *result, int *resultPoolIndex); + uint32_t chooseTransientImageMemType(VkImage img, uint32_t startIndex); + bool createTransientImage(VkFormat format, const QSize &pixelSize, VkImageUsageFlags usage, + VkImageAspectFlags aspectMask, VkSampleCountFlagBits samples, + VkDeviceMemory *mem, VkImage *images, VkImageView *views, int count); + + bool recreateSwapChain(QRhiSwapChain *swapChain); + void releaseSwapChainResources(QRhiSwapChain *swapChain); + + VkFormat optimalDepthStencilFormat(); + VkSampleCountFlagBits effectiveSampleCountBits(int sampleCount); + bool createDefaultRenderPass(QVkRenderPassDescriptor *rpD, + bool hasDepthStencil, + VkSampleCountFlagBits samples, + VkFormat colorFormat); + bool createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, + const QRhiColorAttachment *firstColorAttachment, + const QRhiColorAttachment *lastColorAttachment, + bool preserveColor, + bool preserveDs, + bool storeDs, + QRhiRenderBuffer *depthStencilBuffer, + QRhiTexture *depthTexture, + QRhiTexture *depthResolveTexture); + bool ensurePipelineCache(const void *initialData = nullptr, size_t initialDataSize = 0); + VkShaderModule createShader(const QByteArray &spirv); + + void prepareNewFrame(QRhiCommandBuffer *cb); + VkCommandBuffer startSecondaryCommandBuffer(QVkRenderTargetData *rtD = nullptr); + void endAndEnqueueSecondaryCommandBuffer(VkCommandBuffer cb, QVkCommandBuffer *cbD); + QRhi::FrameOpResult startPrimaryCommandBuffer(VkCommandBuffer *cb); + QRhi::FrameOpResult endAndSubmitPrimaryCommandBuffer(VkCommandBuffer cb, VkFence cmdFence, + VkSemaphore *waitSem, VkSemaphore *signalSem); + void waitCommandCompletion(int frameSlot); + VkDeviceSize subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; + using BufferImageCopyList = QVarLengthArray<VkBufferImageCopy, 16>; + void prepareUploadSubres(QVkTexture *texD, int layer, int level, + const QRhiTextureSubresourceUploadDescription &subresDesc, + size_t *curOfs, void *mp, + BufferImageCopyList *copyInfos); + void enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); + void executeBufferHostWritesForSlot(QVkBuffer *bufD, int slot); + void enqueueTransitionPassResources(QVkCommandBuffer *cbD); + void recordPrimaryCommandBuffer(QVkCommandBuffer *cbD); + void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, + QVkBuffer *bufD, + int slot, + QRhiPassResourceTracker::BufferAccess access, + QRhiPassResourceTracker::BufferStage stage); + void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, + QVkTexture *texD, + QRhiPassResourceTracker::TextureAccess access, + QRhiPassResourceTracker::TextureStage stage); + void recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhiPassResourceTracker &tracker); + void activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRenderTarget *rtD); + void executeDeferredReleases(bool forced = false); + void finishActiveReadbacks(bool forced = false); + + void setObjectName(uint64_t object, VkObjectType type, const QByteArray &name, int slot = -1); + void trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, int slot, + VkAccessFlags access, VkPipelineStageFlags stage); + void trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, + VkImageLayout layout, VkAccessFlags access, VkPipelineStageFlags stage); + void depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuffer *rbD); + void subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, + VkImageLayout oldLayout, VkImageLayout newLayout, + VkAccessFlags srcAccess, VkAccessFlags dstAccess, + VkPipelineStageFlags srcStage, VkPipelineStageFlags dstStage, + int startLayer, int layerCount, + int startLevel, int levelCount); + void updateShaderResourceBindings(QRhiShaderResourceBindings *srb, int descSetIdx = -1); + void ensureCommandPoolForNewFrame(); + double elapsedSecondsFromTimestamp(quint64 timestamp[2], bool *ok); + void printExtraErrorInfo(VkResult err); + + QVulkanInstance *inst = nullptr; + QWindow *maybeWindow = nullptr; + QByteArrayList requestedDeviceExtensions; + bool importedDevice = false; + VkPhysicalDevice physDev = VK_NULL_HANDLE; + VkDevice dev = VK_NULL_HANDLE; + VkCommandPool cmdPool[QVK_FRAMES_IN_FLIGHT] = {}; + quint32 gfxQueueFamilyIdx = 0; + quint32 gfxQueueIdx = 0; + VkQueue gfxQueue = VK_NULL_HANDLE; + quint32 timestampValidBits = 0; + bool importedAllocator = false; + QVkAllocator allocator = nullptr; + QVulkanFunctions *f = nullptr; + QVulkanDeviceFunctions *df = nullptr; + QRhi::Flags rhiFlags; + VkPhysicalDeviceFeatures physDevFeatures; +#ifdef VK_VERSION_1_1 + VkPhysicalDeviceMultiviewFeatures multiviewFeaturesIfApi11; +#endif +#ifdef VK_VERSION_1_2 + VkPhysicalDeviceVulkan11Features physDevFeatures11IfApi12OrNewer; + VkPhysicalDeviceVulkan12Features physDevFeatures12; +#endif +#ifdef VK_VERSION_1_3 + VkPhysicalDeviceVulkan13Features physDevFeatures13; +#endif + VkPhysicalDeviceProperties physDevProperties; + VkDeviceSize ubufAlign; + VkDeviceSize texbufAlign; + bool deviceLost = false; + bool releaseCachedResourcesCalledBeforeFrameStart = false; + +#ifdef VK_EXT_debug_utils + PFN_vkSetDebugUtilsObjectNameEXT vkSetDebugUtilsObjectNameEXT = nullptr; + PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT = nullptr; + PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT = nullptr; + PFN_vkCmdInsertDebugUtilsLabelEXT vkCmdInsertDebugUtilsLabelEXT = nullptr; +#endif + + PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR = nullptr; + PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR; + PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR; + PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR; + PFN_vkQueuePresentKHR vkQueuePresentKHR; + PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR; + PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR; + PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR; + +#ifdef VK_KHR_create_renderpass2 + PFN_vkCreateRenderPass2KHR vkCreateRenderPass2KHR = nullptr; +#endif + + struct { + bool compute = false; + bool wideLines = false; + bool debugUtils = false; + bool vertexAttribDivisor = false; + bool texture3DSliceAs2D = false; + bool tessellation = false; + bool geometryShader = false; + bool nonFillPolygonMode = false; + bool multiView = false; + bool renderPass2KHR = false; + bool depthStencilResolveKHR = false; + QVersionNumber apiVersion; + } caps; + + VkPipelineCache pipelineCache = VK_NULL_HANDLE; + struct DescriptorPoolData { + DescriptorPoolData() { } + DescriptorPoolData(VkDescriptorPool pool_) + : pool(pool_) + { } + VkDescriptorPool pool = VK_NULL_HANDLE; + int refCount = 0; + int allocedDescSets = 0; + }; + QVarLengthArray<DescriptorPoolData, 8> descriptorPools; + QVarLengthArray<VkCommandBuffer, 4> freeSecondaryCbs[QVK_FRAMES_IN_FLIGHT]; + + VkQueryPool timestampQueryPool = VK_NULL_HANDLE; + QBitArray timestampQueryPoolMap; + + VkFormat optimalDsFormat = VK_FORMAT_UNDEFINED; + QMatrix4x4 clipCorrectMatrix; + + QVkSwapChain *currentSwapChain = nullptr; + QSet<QVkSwapChain *> swapchains; + QRhiVulkanNativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) + { + for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) + cbWrapper[i] = new QVkCommandBuffer(rhi); + } + ~OffscreenFrame() + { + for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) + delete cbWrapper[i]; + } + bool active = false; + QVkCommandBuffer *cbWrapper[QVK_FRAMES_IN_FLIGHT]; + VkFence cmdFence = VK_NULL_HANDLE; + int timestampQueryIndex = -1; + } ofr; + + struct TextureReadback { + int activeFrameSlot = -1; + QRhiReadbackDescription desc; + QRhiReadbackResult *result; + VkBuffer stagingBuf; + QVkAlloc stagingAlloc; + quint32 byteSize; + QSize pixelSize; + QRhiTexture::Format format; + }; + QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; + struct BufferReadback { + int activeFrameSlot = -1; + QRhiReadbackResult *result; + quint32 byteSize; + VkBuffer stagingBuf; + QVkAlloc stagingAlloc; + }; + QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; + + struct DeferredReleaseEntry { + enum Type { + Pipeline, + ShaderResourceBindings, + Buffer, + RenderBuffer, + Texture, + Sampler, + TextureRenderTarget, + RenderPass, + StagingBuffer, + SecondaryCommandBuffer + }; + Type type; + int lastActiveFrameSlot; // -1 if not used otherwise 0..FRAMES_IN_FLIGHT-1 + union { + struct { + VkPipeline pipeline; + VkPipelineLayout layout; + } pipelineState; + struct { + int poolIndex; + VkDescriptorSetLayout layout; + } shaderResourceBindings; + struct { + VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + } buffer; + struct { + VkDeviceMemory memory; + VkImage image; + VkImageView imageView; + } renderBuffer; + struct { + VkImage image; + VkImageView imageView; + QVkAlloc allocation; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + VkImageView extraImageViews[QRhi::MAX_MIP_LEVELS]; + } texture; + struct { + VkSampler sampler; + } sampler; + struct { + VkFramebuffer fb; + VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView dsv; + VkImageView resdsv; + } textureRenderTarget; + struct { + VkRenderPass rp; + } renderPass; + struct { + VkBuffer stagingBuffer; + QVkAlloc stagingAllocation; + } stagingBuffer; + struct { + VkCommandBuffer cb; + } secondaryCommandBuffer; + }; + }; + QList<DeferredReleaseEntry> releaseQueue; }; +Q_DECLARE_TYPEINFO(QRhiVulkan::DescriptorPoolData, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::TextureReadback, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::BufferReadback, Q_RELOCATABLE_TYPE); + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhivulkan_p_p.h b/src/gui/rhi/qrhivulkan_p_p.h deleted file mode 100644 index af647bb508..0000000000 --- a/src/gui/rhi/qrhivulkan_p_p.h +++ /dev/null @@ -1,1018 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIVULKAN_P_H -#define QRHIVULKAN_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhivulkan_p.h" -#include "qrhi_p_p.h" - -QT_BEGIN_NAMESPACE - -class QVulkanFunctions; -class QVulkanDeviceFunctions; - -static const int QVK_FRAMES_IN_FLIGHT = 2; - -static const int QVK_DESC_SETS_PER_POOL = 128; -static const int QVK_UNIFORM_BUFFERS_PER_POOL = 256; -static const int QVK_COMBINED_IMAGE_SAMPLERS_PER_POOL = 256; -static const int QVK_STORAGE_BUFFERS_PER_POOL = 128; -static const int QVK_STORAGE_IMAGES_PER_POOL = 128; - -static const int QVK_MAX_ACTIVE_TIMESTAMP_PAIRS = 16; - -// no vk_mem_alloc.h available here, void* is good enough -typedef void * QVkAlloc; -typedef void * QVkAllocator; - -struct QVkBuffer : public QRhiBuffer -{ - QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, int size); - ~QVkBuffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; - struct DynamicUpdate { - int offset; - QRhiBufferData data; - }; - QVarLengthArray<DynamicUpdate, 16> pendingDynamicUpdates[QVK_FRAMES_IN_FLIGHT]; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - struct UsageState { - VkAccessFlags access = 0; - VkPipelineStageFlags stage = 0; - }; - UsageState usageState[QVK_FRAMES_IN_FLIGHT]; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -Q_DECLARE_TYPEINFO(QVkBuffer::DynamicUpdate, Q_RELOCATABLE_TYPE); - -struct QVkTexture; - -struct QVkRenderBuffer : public QRhiRenderBuffer -{ - QVkRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, Flags flags, - QRhiTexture::Format backingFormatHint); - ~QVkRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - VkDeviceMemory memory = VK_NULL_HANDLE; - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - VkSampleCountFlagBits samples; - QVkTexture *backingTexture = nullptr; - VkFormat vkformat; - int lastActiveFrameSlot = -1; - friend class QRhiVulkan; -}; - -struct QVkTexture : public QRhiTexture -{ - QVkTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int sampleCount, Flags flags); - ~QVkTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - void setNativeLayout(int layout) override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - bool finishCreate(); - VkImageView imageViewForLevel(int level); - - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - QVkAlloc imageAlloc = nullptr; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - VkImageView perLevelImageViews[QRhi::MAX_MIP_LEVELS]; - bool owns = true; - struct UsageState { - // no tracking of subresource layouts (some operations can keep - // subresources in different layouts for some time, but that does not - // need to be kept track of) - VkImageLayout layout; - VkAccessFlags access; - VkPipelineStageFlags stage; - }; - UsageState usageState; - VkFormat vkformat; - uint mipLevelCount = 0; - VkSampleCountFlagBits samples; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkSampler : public QRhiSampler -{ - QVkSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QVkSampler(); - void destroy() override; - bool create() override; - - VkSampler sampler = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QVkRenderPassDescriptor(QRhiImplementation *rhi); - ~QVkRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - const QRhiNativeHandles *nativeHandles() override; - - VkRenderPass rp = VK_NULL_HANDLE; - bool ownsRp = false; - QVarLengthArray<VkAttachmentDescription, 8> attDescs; - QVarLengthArray<VkAttachmentReference, 8> colorRefs; - QVarLengthArray<VkAttachmentReference, 8> resolveRefs; - QVarLengthArray<VkSubpassDependency, 2> subpassDeps; - bool hasDepthStencil = false; - VkAttachmentReference dsRef; - QRhiVulkanRenderPassNativeHandles nativeHandlesStruct; - int lastActiveFrameSlot = -1; -}; - -struct QVkRenderTargetData -{ - VkFramebuffer fb = VK_NULL_HANDLE; - QVkRenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - int resolveAttCount = 0; - static const int MAX_COLOR_ATTACHMENTS = 8; -}; - -struct QVkReferenceRenderTarget : public QRhiRenderTarget -{ - QVkReferenceRenderTarget(QRhiImplementation *rhi); - ~QVkReferenceRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QVkRenderTargetData d; -}; - -struct QVkTextureRenderTarget : public QRhiTextureRenderTarget -{ - QVkTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QVkTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QVkRenderTargetData d; - VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - int lastActiveFrameSlot = -1; - friend class QRhiVulkan; -}; - -struct QVkShaderResourceBindings : public QRhiShaderResourceBindings -{ - QVkShaderResourceBindings(QRhiImplementation *rhi); - ~QVkShaderResourceBindings(); - void destroy() override; - bool create() override; - - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - bool hasSlottedResource = false; - bool hasDynamicOffset = false; - int poolIndex = -1; - VkDescriptorSetLayout layout = VK_NULL_HANDLE; - VkDescriptorSet descSets[QVK_FRAMES_IN_FLIGHT]; // multiple sets to support dynamic buffers - int lastActiveFrameSlot = -1; - uint generation = 0; - - // Keep track of the generation number of each referenced QRhi* to be able - // to detect that the underlying descriptor set became out of date and they - // need to be written again with the up-to-date VkBuffer etc. objects. - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData[QVK_FRAMES_IN_FLIGHT]; - - friend class QRhiVulkan; -}; - -Q_DECLARE_TYPEINFO(QVkShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); - -struct QVkGraphicsPipeline : public QRhiGraphicsPipeline -{ - QVkGraphicsPipeline(QRhiImplementation *rhi); - ~QVkGraphicsPipeline(); - void destroy() override; - bool create() override; - - VkPipelineLayout layout = VK_NULL_HANDLE; - VkPipeline pipeline = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkComputePipeline : public QRhiComputePipeline -{ - QVkComputePipeline(QRhiImplementation *rhi); - ~QVkComputePipeline(); - void destroy() override; - bool create() override; - - VkPipelineLayout layout = VK_NULL_HANDLE; - VkPipeline pipeline = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkCommandBuffer : public QRhiCommandBuffer -{ - QVkCommandBuffer(QRhiImplementation *rhi); - ~QVkCommandBuffer(); - void destroy() override; - - const QRhiNativeHandles *nativeHandles(); - - VkCommandBuffer cb = VK_NULL_HANDLE; // primary - QRhiVulkanCommandBufferNativeHandles nativeHandlesStruct; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - void resetState() { - recordingPass = NoPass; - passUsesSecondaryCb = false; - currentTarget = nullptr; - activeSecondaryCbStack.clear(); - resetCommands(); - resetCachedState(); - } - - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - currentDescSetSlot = -1; - currentIndexBuffer = VK_NULL_HANDLE; - currentIndexOffset = 0; - currentIndexFormat = VK_INDEX_TYPE_UINT16; - memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); - memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); - inExternal = false; - } - - PassType recordingPass; - bool passUsesSecondaryCb; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - int currentDescSetSlot; - VkBuffer currentIndexBuffer; - quint32 currentIndexOffset; - VkIndexType currentIndexFormat; - static const int VERTEX_INPUT_RESOURCE_SLOT_COUNT = 32; - VkBuffer currentVertexBuffers[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - quint32 currentVertexOffsets[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - QVarLengthArray<VkCommandBuffer, 4> activeSecondaryCbStack; - bool inExternal; - - struct { - QHash<QRhiResource *, QPair<VkAccessFlags, bool> > writtenResources; - void reset() { - writtenResources.clear(); - } - } computePassState; - - struct Command { - enum Cmd { - CopyBuffer, - CopyBufferToImage, - CopyImage, - CopyImageToBuffer, - ImageBarrier, - BufferBarrier, - BlitImage, - BeginRenderPass, - EndRenderPass, - BindPipeline, - BindDescriptorSet, - BindVertexBuffer, - BindIndexBuffer, - SetViewport, - SetScissor, - SetBlendConstants, - SetStencilRef, - Draw, - DrawIndexed, - DebugMarkerBegin, - DebugMarkerEnd, - DebugMarkerInsert, - TransitionPassResources, - Dispatch, - ExecuteSecondary - }; - Cmd cmd; - - union Args { - struct { - VkBuffer src; - VkBuffer dst; - VkBufferCopy desc; - } copyBuffer; - struct { - VkBuffer src; - VkImage dst; - VkImageLayout dstLayout; - int count; - int bufferImageCopyIndex; - } copyBufferToImage; - struct { - VkImage src; - VkImageLayout srcLayout; - VkImage dst; - VkImageLayout dstLayout; - VkImageCopy desc; - } copyImage; - struct { - VkImage src; - VkImageLayout srcLayout; - VkBuffer dst; - VkBufferImageCopy desc; - } copyImageToBuffer; - struct { - VkPipelineStageFlags srcStageMask; - VkPipelineStageFlags dstStageMask; - int count; - int index; - } imageBarrier; - struct { - VkPipelineStageFlags srcStageMask; - VkPipelineStageFlags dstStageMask; - int count; - int index; - } bufferBarrier; - struct { - VkImage src; - VkImageLayout srcLayout; - VkImage dst; - VkImageLayout dstLayout; - VkFilter filter; - VkImageBlit desc; - } blitImage; - struct { - VkRenderPassBeginInfo desc; - int clearValueIndex; - bool useSecondaryCb; - } beginRenderPass; - struct { - } endRenderPass; - struct { - VkPipelineBindPoint bindPoint; - VkPipeline pipeline; - } bindPipeline; - struct { - VkPipelineBindPoint bindPoint; - VkPipelineLayout pipelineLayout; - VkDescriptorSet descSet; - int dynamicOffsetCount; - int dynamicOffsetIndex; - } bindDescriptorSet; - struct { - int startBinding; - int count; - int vertexBufferIndex; - int vertexBufferOffsetIndex; - } bindVertexBuffer; - struct { - VkBuffer buf; - VkDeviceSize ofs; - VkIndexType type; - } bindIndexBuffer; - struct { - VkViewport viewport; - } setViewport; - struct { - VkRect2D scissor; - } setScissor; - struct { - float c[4]; - } setBlendConstants; - struct { - uint32_t ref; - } setStencilRef; - struct { - uint32_t vertexCount; - uint32_t instanceCount; - uint32_t firstVertex; - uint32_t firstInstance; - } draw; - struct { - uint32_t indexCount; - uint32_t instanceCount; - uint32_t firstIndex; - int32_t vertexOffset; - uint32_t firstInstance; - } drawIndexed; - struct { - VkDebugMarkerMarkerInfoEXT marker; - int markerNameIndex; - } debugMarkerBegin; - struct { - } debugMarkerEnd; - struct { - VkDebugMarkerMarkerInfoEXT marker; - int markerNameIndex; - } debugMarkerInsert; - struct { - int trackerIndex; - } transitionResources; - struct { - int x, y, z; - } dispatch; - struct { - VkCommandBuffer cb; - } executeSecondary; - } args; - }; - - QRhiBackendCommandList<Command> commands; - QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; - int currentPassResTrackerIndex; - - void resetCommands() { - commands.reset(); - resetPools(); - - passResTrackers.clear(); - currentPassResTrackerIndex = -1; - } - - void resetPools() { - pools.clearValue.clear(); - pools.bufferImageCopy.clear(); - pools.dynamicOffset.clear(); - pools.vertexBuffer.clear(); - pools.vertexBufferOffset.clear(); - pools.debugMarkerData.clear(); - pools.imageBarrier.clear(); - pools.bufferBarrier.clear(); - } - - struct { - QVarLengthArray<VkClearValue, 4> clearValue; - QVarLengthArray<VkBufferImageCopy, 16> bufferImageCopy; - QVarLengthArray<uint32_t, 4> dynamicOffset; - QVarLengthArray<VkBuffer, 4> vertexBuffer; - QVarLengthArray<VkDeviceSize, 4> vertexBufferOffset; - QVarLengthArray<QByteArray, 4> debugMarkerData; - QVarLengthArray<VkImageMemoryBarrier, 8> imageBarrier; - QVarLengthArray<VkBufferMemoryBarrier, 8> bufferBarrier; - } pools; - - friend class QRhiVulkan; -}; - -struct QVkSwapChain : public QRhiSwapChain -{ - QVkSwapChain(QRhiImplementation *rhi); - ~QVkSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - bool ensureSurface(); - - static const quint32 EXPECTED_MAX_BUFFER_COUNT = 4; - - QWindow *window = nullptr; - QSize pixelSize; - bool supportsReadback = false; - VkSwapchainKHR sc = VK_NULL_HANDLE; - int bufferCount = 0; - VkSurfaceKHR surface = VK_NULL_HANDLE; - VkSurfaceKHR lastConnectedSurface = VK_NULL_HANDLE; - VkFormat colorFormat = VK_FORMAT_B8G8R8A8_UNORM; - VkColorSpaceKHR colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; - QVkRenderBuffer *ds = nullptr; - VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT; - QVarLengthArray<VkPresentModeKHR, 8> supportedPresentationModes; - VkDeviceMemory msaaImageMem = VK_NULL_HANDLE; - QVkReferenceRenderTarget rtWrapper; - QVkCommandBuffer cbWrapper; - - struct ImageResources { - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - VkFramebuffer fb = VK_NULL_HANDLE; - VkImage msaaImage = VK_NULL_HANDLE; - VkImageView msaaImageView = VK_NULL_HANDLE; - enum LastUse { - ScImageUseNone, - ScImageUseRender, - ScImageUseTransferSource - }; - LastUse lastUse = ScImageUseNone; - }; - QVarLengthArray<ImageResources, EXPECTED_MAX_BUFFER_COUNT> imageRes; - - struct FrameResources { - VkFence imageFence = VK_NULL_HANDLE; - bool imageFenceWaitable = false; - VkSemaphore imageSem = VK_NULL_HANDLE; - VkSemaphore drawSem = VK_NULL_HANDLE; - bool imageAcquired = false; - bool imageSemWaitable = false; - VkFence cmdFence = VK_NULL_HANDLE; - bool cmdFenceWaitable = false; - VkCommandBuffer cmdBuf = VK_NULL_HANDLE; // primary - int timestampQueryIndex = -1; - } frameRes[QVK_FRAMES_IN_FLIGHT]; - - quint32 currentImageIndex = 0; // index in imageRes - quint32 currentFrameSlot = 0; // index in frameRes - int frameCount = 0; - - friend class QRhiVulkan; -}; - -class QRhiVulkan : public QRhiImplementation -{ -public: - QRhiVulkan(QRhiVulkanInitParams *params, QRhiVulkanNativeHandles *importParams = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - int size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - void sendVMemStatsToProfiler() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - VkResult createDescriptorPool(VkDescriptorPool *pool); - bool allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, VkDescriptorSet *result, int *resultPoolIndex); - uint32_t chooseTransientImageMemType(VkImage img, uint32_t startIndex); - bool createTransientImage(VkFormat format, const QSize &pixelSize, VkImageUsageFlags usage, - VkImageAspectFlags aspectMask, VkSampleCountFlagBits samples, - VkDeviceMemory *mem, VkImage *images, VkImageView *views, int count); - - bool recreateSwapChain(QRhiSwapChain *swapChain); - void releaseSwapChainResources(QRhiSwapChain *swapChain); - - VkFormat optimalDepthStencilFormat(); - VkSampleCountFlagBits effectiveSampleCount(int sampleCount); - bool createDefaultRenderPass(QVkRenderPassDescriptor *rpD, - bool hasDepthStencil, - VkSampleCountFlagBits samples, - VkFormat colorFormat); - bool createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, - const QRhiColorAttachment *firstColorAttachment, - const QRhiColorAttachment *lastColorAttachment, - bool preserveColor, - bool preserveDs, - QRhiRenderBuffer *depthStencilBuffer, - QRhiTexture *depthTexture); - bool ensurePipelineCache(const void *initialData = nullptr, size_t initialDataSize = 0); - VkShaderModule createShader(const QByteArray &spirv); - - void prepareNewFrame(QRhiCommandBuffer *cb); - VkCommandBuffer startSecondaryCommandBuffer(QVkRenderTargetData *rtD = nullptr); - void endAndEnqueueSecondaryCommandBuffer(VkCommandBuffer cb, QVkCommandBuffer *cbD); - QRhi::FrameOpResult startPrimaryCommandBuffer(VkCommandBuffer *cb); - QRhi::FrameOpResult endAndSubmitPrimaryCommandBuffer(VkCommandBuffer cb, VkFence cmdFence, - VkSemaphore *waitSem, VkSemaphore *signalSem); - void waitCommandCompletion(int frameSlot); - VkDeviceSize subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; - using BufferImageCopyList = QVarLengthArray<VkBufferImageCopy, 16>; - void prepareUploadSubres(QVkTexture *texD, int layer, int level, - const QRhiTextureSubresourceUploadDescription &subresDesc, - size_t *curOfs, void *mp, - BufferImageCopyList *copyInfos); - void enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); - void executeBufferHostWritesForSlot(QVkBuffer *bufD, int slot); - void enqueueTransitionPassResources(QVkCommandBuffer *cbD); - void recordPrimaryCommandBuffer(QVkCommandBuffer *cbD); - void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, - QVkBuffer *bufD, - int slot, - QRhiPassResourceTracker::BufferAccess access, - QRhiPassResourceTracker::BufferStage stage); - void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, - QVkTexture *texD, - QRhiPassResourceTracker::TextureAccess access, - QRhiPassResourceTracker::TextureStage stage); - void recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhiPassResourceTracker &tracker); - void activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRenderTarget *rtD); - void executeDeferredReleases(bool forced = false); - void finishActiveReadbacks(bool forced = false); - - void setObjectName(uint64_t object, VkDebugReportObjectTypeEXT type, const QByteArray &name, int slot = -1); - void trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, int slot, - VkAccessFlags access, VkPipelineStageFlags stage); - void trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, - VkImageLayout layout, VkAccessFlags access, VkPipelineStageFlags stage); - void depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuffer *rbD); - void subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, - VkImageLayout oldLayout, VkImageLayout newLayout, - VkAccessFlags srcAccess, VkAccessFlags dstAccess, - VkPipelineStageFlags srcStage, VkPipelineStageFlags dstStage, - int startLayer, int layerCount, - int startLevel, int levelCount); - void updateShaderResourceBindings(QRhiShaderResourceBindings *srb, int descSetIdx = -1); - void ensureCommandPoolForNewFrame(); - - QVulkanInstance *inst = nullptr; - QWindow *maybeWindow = nullptr; - QByteArrayList requestedDeviceExtensions; - bool importedDevice = false; - VkPhysicalDevice physDev = VK_NULL_HANDLE; - VkDevice dev = VK_NULL_HANDLE; - VkCommandPool cmdPool[QVK_FRAMES_IN_FLIGHT] = {}; - int gfxQueueFamilyIdx = -1; - int gfxQueueIdx = 0; - VkQueue gfxQueue = VK_NULL_HANDLE; - quint32 timestampValidBits = 0; - bool importedAllocator = false; - QVkAllocator allocator = nullptr; - QVulkanFunctions *f = nullptr; - QVulkanDeviceFunctions *df = nullptr; - QRhi::Flags rhiFlags; - VkPhysicalDeviceFeatures physDevFeatures; - VkPhysicalDeviceProperties physDevProperties; - VkDeviceSize ubufAlign; - VkDeviceSize texbufAlign; - bool deviceLost = false; - bool releaseCachedResourcesCalledBeforeFrameStart = false; - - PFN_vkCmdDebugMarkerBeginEXT vkCmdDebugMarkerBegin = nullptr; - PFN_vkCmdDebugMarkerEndEXT vkCmdDebugMarkerEnd = nullptr; - PFN_vkCmdDebugMarkerInsertEXT vkCmdDebugMarkerInsert = nullptr; - PFN_vkDebugMarkerSetObjectNameEXT vkDebugMarkerSetObjectName = nullptr; - - PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR = nullptr; - PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR; - PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR; - PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR; - PFN_vkQueuePresentKHR vkQueuePresentKHR; - PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR = nullptr; - PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR; - PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR; - - struct { - bool compute = false; - bool wideLines = false; - bool debugMarkers = false; - bool vertexAttribDivisor = false; - bool texture3DSliceAs2D = false; - } caps; - - VkPipelineCache pipelineCache = VK_NULL_HANDLE; - struct DescriptorPoolData { - DescriptorPoolData() { } - DescriptorPoolData(VkDescriptorPool pool_) - : pool(pool_) - { } - VkDescriptorPool pool = VK_NULL_HANDLE; - int refCount = 0; - int allocedDescSets = 0; - }; - QVarLengthArray<DescriptorPoolData, 8> descriptorPools; - QVarLengthArray<VkCommandBuffer, 4> freeSecondaryCbs[QVK_FRAMES_IN_FLIGHT]; - - VkQueryPool timestampQueryPool = VK_NULL_HANDLE; - QBitArray timestampQueryPoolMap; - - VkFormat optimalDsFormat = VK_FORMAT_UNDEFINED; - QMatrix4x4 clipCorrectMatrix; - - QVkSwapChain *currentSwapChain = nullptr; - QSet<QVkSwapChain *> swapchains; - QRhiVulkanNativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) - { - for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) - cbWrapper[i] = new QVkCommandBuffer(rhi); - } - ~OffscreenFrame() - { - for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) - delete cbWrapper[i]; - } - bool active = false; - QVkCommandBuffer *cbWrapper[QVK_FRAMES_IN_FLIGHT]; - VkFence cmdFence = VK_NULL_HANDLE; - } ofr; - - struct TextureReadback { - int activeFrameSlot = -1; - QRhiReadbackDescription desc; - QRhiReadbackResult *result; - VkBuffer stagingBuf; - QVkAlloc stagingAlloc; - quint32 byteSize; - QSize pixelSize; - QRhiTexture::Format format; - }; - QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; - struct BufferReadback { - int activeFrameSlot = -1; - QRhiBufferReadbackResult *result; - int byteSize; - VkBuffer stagingBuf; - QVkAlloc stagingAlloc; - }; - QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; - - struct DeferredReleaseEntry { - enum Type { - Pipeline, - ShaderResourceBindings, - Buffer, - RenderBuffer, - Texture, - Sampler, - TextureRenderTarget, - RenderPass, - StagingBuffer, - SecondaryCommandBuffer - }; - Type type; - int lastActiveFrameSlot; // -1 if not used otherwise 0..FRAMES_IN_FLIGHT-1 - union { - struct { - VkPipeline pipeline; - VkPipelineLayout layout; - } pipelineState; - struct { - int poolIndex; - VkDescriptorSetLayout layout; - } shaderResourceBindings; - struct { - VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - } buffer; - struct { - VkDeviceMemory memory; - VkImage image; - VkImageView imageView; - } renderBuffer; - struct { - VkImage image; - VkImageView imageView; - QVkAlloc allocation; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - VkImageView extraImageViews[QRhi::MAX_MIP_LEVELS]; - } texture; - struct { - VkSampler sampler; - } sampler; - struct { - VkFramebuffer fb; - VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - } textureRenderTarget; - struct { - VkRenderPass rp; - } renderPass; - struct { - VkBuffer stagingBuffer; - QVkAlloc stagingAllocation; - } stagingBuffer; - struct { - VkCommandBuffer cb; - } secondaryCommandBuffer; - }; - }; - QList<DeferredReleaseEntry> releaseQueue; -}; - -Q_DECLARE_TYPEINFO(QRhiVulkan::DescriptorPoolData, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::TextureReadback, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::BufferReadback, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhivulkanext_p.h b/src/gui/rhi/qrhivulkanext_p.h deleted file mode 100644 index d98067c803..0000000000 --- a/src/gui/rhi/qrhivulkanext_p.h +++ /dev/null @@ -1,84 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2018 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt RHI module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QRHIVULKANEXT_P_H -#define QRHIVULKANEXT_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhivulkan_p.h" - -QT_BEGIN_NAMESPACE - -#ifndef VK_EXT_vertex_attribute_divisor -#define VK_EXT_vertex_attribute_divisor 1 -#define VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_SPEC_VERSION 2 -#define VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME "VK_EXT_vertex_attribute_divisor" - -typedef struct VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT { - VkStructureType sType; - void* pNext; - uint32_t maxVertexAttribDivisor; -} VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT; - -typedef struct VkVertexInputBindingDivisorDescriptionEXT { - uint32_t binding; - uint32_t divisor; -} VkVertexInputBindingDivisorDescriptionEXT; - -typedef struct VkPipelineVertexInputDivisorStateCreateInfoEXT { - VkStructureType sType; - const void* pNext; - uint32_t vertexBindingDivisorCount; - const VkVertexInputBindingDivisorDescriptionEXT* pVertexBindingDivisors; -} VkPipelineVertexInputDivisorStateCreateInfoEXT; -#endif // VK_EXT_vertex_attribute_divisor - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qshader.cpp b/src/gui/rhi/qshader.cpp index 74255dcf5c..d5fb53e7e6 100644 --- a/src/gui/rhi/qshader.cpp +++ b/src/gui/rhi/qshader.cpp @@ -1,43 +1,7 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qshader_p_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qshader_p.h" #include <QDataStream> #include <QBuffer> @@ -45,8 +9,9 @@ QT_BEGIN_NAMESPACE /*! \class QShader - \internal + \ingroup painting-3D \inmodule QtGui + \since 6.6 \brief Contains multiple versions of a shader translated to multiple shading languages, together with reflection metadata. @@ -57,6 +22,16 @@ QT_BEGIN_NAMESPACE as, Vulkan, Metal, Direct3D, and OpenGL, take QShader as their input whenever a shader needs to be specified. + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qshader.h>}. + A QShader instance is empty and thus invalid by default. To get a useful instance, the two typical methods are: @@ -104,8 +79,9 @@ QT_BEGIN_NAMESPACE can be returned or passed by value. Detach happens implicitly when calling a setter. - For reference, QRhi expects that a QShader suitable for all its - backends contains at least the following: + For reference, a typical, portable QRhi expects that a QShader suitable for + all its backends contains at least the following. (this excludes support + for core profile OpenGL contexts, add GLSL 150 or newer for that) \list @@ -113,11 +89,11 @@ QT_BEGIN_NAMESPACE \li GLSL/ES 100 source code suitable for OpenGL ES 2.0 or newer - \li GLSL 120 source code suitable for OpenGL 2.1 + \li GLSL 120 source code suitable for OpenGL 2.1 or newer - \li HLSL Shader Model 5.0 source code or the corresponding DXBC bytecode suitable for Direct3D 11 + \li HLSL Shader Model 5.0 source code or the corresponding DXBC bytecode suitable for Direct3D 11/12 - \li Metal Shading Language 1.2 source code or the corresponding bytecode suitable for Metal + \li Metal Shading Language 1.2 source code or the corresponding bytecode suitable for Metal 1.2 or newer \endlist @@ -138,8 +114,8 @@ QT_BEGIN_NAMESPACE /*! \class QShaderVersion - \internal \inmodule QtGui + \since 6.6 \brief Specifies the shading language version. @@ -164,6 +140,9 @@ QT_BEGIN_NAMESPACE A default constructed QShaderVersion contains a version of 100 and no flags set. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! @@ -176,13 +155,16 @@ QT_BEGIN_NAMESPACE /*! \class QShaderKey - \internal \inmodule QtGui + \since 6.6 \brief Specifies the shading language, the version with flags, and the variant. A default constructed QShaderKey has source set to SpirvShader and sourceVersion set to 100. sourceVariant defaults to StandardShader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! @@ -196,6 +178,7 @@ QT_BEGIN_NAMESPACE \value MslShader Metal Shading Language \value DxilShader Direct3D bytecode (HLSL compiled by \c dxc) \value MetalLibShader Pre-compiled Metal bytecode + \value WgslShader WGSL */ /*! @@ -203,25 +186,74 @@ QT_BEGIN_NAMESPACE Describes what kind of shader code an entry contains. \value StandardShader The normal, unmodified version of the shader code. + \value BatchableVertexShader Vertex shader rewritten to be suitable for Qt Quick scenegraph batching. + + \value UInt16IndexedVertexAsComputeShader A vertex shader meant to be used + in a Metal pipeline with tessellation in combination with indexed draw + calls sourcing index data from a uint16 index buffer. To support the Metal + tessellation pipeline, the vertex shader is translated to a compute shader + that may be dependent on the index buffer usage in the draw calls (e.g. if + the shader is using gl_VertexIndex), hence the need for three dedicated + variants. + + \value UInt32IndexedVertexAsComputeShader A vertex shader meant to be used + in a Metal pipeline with tessellation in combination with indexed draw + calls sourcing index data from a uint32 index buffer. To support the Metal + tessellation pipeline, the vertex shader is translated to a compute shader + that may be dependent on the index buffer usage in the draw calls (e.g. if + the shader is using gl_VertexIndex), hence the need for three dedicated + variants. + + \value NonIndexedVertexAsComputeShader A vertex shader meant to be used in + a Metal pipeline with tessellation in combination with non-indexed draw + calls. To support the Metal tessellation pipeline, the vertex shader is + translated to a compute shader that may be dependent on the index buffer + usage in the draw calls (e.g. if the shader is using gl_VertexIndex), hence + the need for three dedicated variants. + */ + +/*! + \enum QShader::SerializedFormatVersion + Describes the desired output format when serializing the QShader. + + The default value for the \c version argument of serialized() is \c Latest. + This is sufficient in the vast majority of cases. Specifying another value + is needed only when the intention is to generate serialized data that can + be loaded by earlier Qt versions. For example, the \c qsb tool uses these + enum values when the \c{--qsbversion} command-line argument is given. + + \note Targeting earlier versions will make certain features disfunctional + with the generated asset. This is not an issue when using the asset with + the specified, older Qt version, given that that Qt version does not have + the newer features in newer Qt versions that rely on additional data + generated in the QShader and the serialized data stream, but may become a + problem if the generated asset is then used with a newer Qt version. + + \value Latest The current Qt version + \value Qt_6_5 Qt 6.5 + \value Qt_6_4 Qt 6.4 */ /*! \class QShaderCode - \internal \inmodule QtGui + \since 6.6 \brief Contains source or binary code for a shader and additional metadata. When shader() is empty after retrieving a QShaderCode instance from QShader, it indicates no shader code was found for the requested key. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! Constructs a new, empty (and thus invalid) QShader instance. */ QShader::QShader() - : d(new QShaderPrivate) + : d(nullptr) { } @@ -230,42 +262,87 @@ QShader::QShader() */ void QShader::detach() { - qAtomicDetach(d); + if (d) + qAtomicDetach(d); + else + d = new QShaderPrivate; } /*! - \internal + Constructs a copy of \a other. */ QShader::QShader(const QShader &other) : d(other.d) { - d->ref.ref(); + if (d) + d->ref.ref(); } /*! - \internal + Assigns \a other to this object. */ QShader &QShader::operator=(const QShader &other) { - qAtomicAssign(d, other.d); + if (d) { + if (other.d) { + qAtomicAssign(d, other.d); + } else { + if (!d->ref.deref()) + delete d; + d = nullptr; + } + } else if (other.d) { + other.d->ref.ref(); + d = other.d; + } return *this; } /*! + \fn QShader::QShader(QShader &&other) noexcept + \since 6.7 + + Move-constructs a new QShader from \a other. + + \note The moved-from object \a other is placed in a + partially-formed state, in which the only valid operations are + destruction and assignment of a new value. +*/ + +/*! + \fn QShader &QShader::operator=(QShader &&other) + \since 6.7 + + Move-assigns \a other to this QShader instance. + + \note The moved-from object \a other is placed in a + partially-formed state, in which the only valid operations are + destruction and assignment of a new value. +*/ + +/*! Destructor. */ QShader::~QShader() { - if (!d->ref.deref()) + if (d && !d->ref.deref()) delete d; } /*! + \fn void QShader::swap(QShader &other) + \since 6.7 + + Swaps shader \a other with this shader. This operation is very fast and + never fails. +*/ + +/*! \return true if the QShader contains at least one shader version. */ bool QShader::isValid() const { - return !d->shaders.isEmpty(); + return d ? !d->shaders.isEmpty() : false; } /*! @@ -273,7 +350,7 @@ bool QShader::isValid() const */ QShader::Stage QShader::stage() const { - return d->stage; + return d ? d->stage : QShader::VertexStage; } /*! @@ -281,7 +358,7 @@ QShader::Stage QShader::stage() const */ void QShader::setStage(Stage stage) { - if (stage != d->stage) { + if (!d || stage != d->stage) { detach(); d->stage = stage; } @@ -292,7 +369,7 @@ void QShader::setStage(Stage stage) */ QShaderDescription QShader::description() const { - return d->desc; + return d ? d->desc : QShaderDescription(); } /*! @@ -309,7 +386,7 @@ void QShader::setDescription(const QShaderDescription &desc) */ QList<QShaderKey> QShader::availableShaders() const { - return d->shaders.keys().toVector(); + return d ? d->shaders.keys().toVector() : QList<QShaderKey>(); } /*! @@ -317,7 +394,7 @@ QList<QShaderKey> QShader::availableShaders() const */ QShaderCode QShader::shader(const QShaderKey &key) const { - return d->shaders.value(key); + return d ? d->shaders.value(key) : QShaderCode(); } /*! @@ -325,7 +402,7 @@ QShaderCode QShader::shader(const QShaderKey &key) const */ void QShader::setShader(const QShaderKey &key, const QShaderCode &shader) { - if (d->shaders.value(key) == shader) + if (d && d->shaders.value(key) == shader) return; detach(); @@ -338,6 +415,9 @@ void QShader::setShader(const QShaderKey &key, const QShaderCode &shader) */ void QShader::removeShader(const QShaderKey &key) { + if (!d) + return; + auto it = d->shaders.find(key); if (it == d->shaders.end()) return; @@ -358,39 +438,78 @@ static void writeShaderKey(QDataStream *ds, const QShaderKey &k) \return a serialized binary version of all the data held by the QShader, suitable for writing to files or other I/O devices. + By default the latest serialization format is used. Use \a version + parameter to serialize for a compatibility Qt version. Only when it is + known that the generated data stream must be made compatible with an older + Qt version at the expense of making it incompatible with features + introduced since that Qt version, should another value (for example, + \l{SerializedFormatVersion}{Qt_6_5} for Qt 6.5) be used. + \sa fromSerialized() */ -QByteArray QShader::serialized() const +QByteArray QShader::serialized(SerializedFormatVersion version) const { + static QShaderPrivate sd; + QShaderPrivate *dd = d ? d : &sd; + QBuffer buf; QDataStream ds(&buf); ds.setVersion(QDataStream::Qt_5_10); if (!buf.open(QIODevice::WriteOnly)) return QByteArray(); - ds << QShaderPrivate::QSB_VERSION; - ds << int(d->stage); - d->desc.serialize(&ds); - ds << int(d->shaders.count()); - for (auto it = d->shaders.cbegin(), itEnd = d->shaders.cend(); it != itEnd; ++it) { + const int qsbVersion = QShaderPrivate::qtQsbVersion(version); + ds << qsbVersion; + + ds << int(dd->stage); + dd->desc.serialize(&ds, qsbVersion); + ds << int(dd->shaders.size()); + for (auto it = dd->shaders.cbegin(), itEnd = dd->shaders.cend(); it != itEnd; ++it) { const QShaderKey &k(it.key()); writeShaderKey(&ds, k); - const QShaderCode &shader(d->shaders.value(k)); + const QShaderCode &shader(dd->shaders.value(k)); ds << shader.shader(); ds << shader.entryPoint(); } - ds << int(d->bindings.count()); - for (auto it = d->bindings.cbegin(), itEnd = d->bindings.cend(); it != itEnd; ++it) { + ds << int(dd->bindings.size()); + for (auto it = dd->bindings.cbegin(), itEnd = dd->bindings.cend(); it != itEnd; ++it) { const QShaderKey &k(it.key()); writeShaderKey(&ds, k); const NativeResourceBindingMap &map(it.value()); - ds << int(map.count()); + ds << int(map.size()); for (auto mapIt = map.cbegin(), mapItEnd = map.cend(); mapIt != mapItEnd; ++mapIt) { ds << mapIt.key(); ds << mapIt.value().first; ds << mapIt.value().second; } } + ds << int(dd->combinedImageMap.size()); + for (auto it = dd->combinedImageMap.cbegin(), itEnd = dd->combinedImageMap.cend(); it != itEnd; ++it) { + const QShaderKey &k(it.key()); + writeShaderKey(&ds, k); + const SeparateToCombinedImageSamplerMappingList &list(it.value()); + ds << int(list.size()); + for (auto listIt = list.cbegin(), listItEnd = list.cend(); listIt != listItEnd; ++listIt) { + ds << listIt->combinedSamplerName; + ds << listIt->textureBinding; + ds << listIt->samplerBinding; + } + } + if (qsbVersion > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) { + ds << int(dd->nativeShaderInfoMap.size()); + for (auto it = dd->nativeShaderInfoMap.cbegin(), itEnd = dd->nativeShaderInfoMap.cend(); it != itEnd; ++it) { + const QShaderKey &k(it.key()); + writeShaderKey(&ds, k); + ds << it->flags; + ds << int(it->extraBufferBindings.size()); + for (auto mapIt = it->extraBufferBindings.cbegin(), mapItEnd = it->extraBufferBindings.cend(); + mapIt != mapItEnd; ++mapIt) + { + ds << mapIt.key(); + ds << mapIt.value(); + } + } + } return qCompress(buf.buffer()); } @@ -413,6 +532,9 @@ static void readShaderKey(QDataStream *ds, QShaderKey *k) /*! Creates a new QShader instance from the given \a data. + If \a data cannot be deserialized successfully, the result is a default + constructed QShader for which isValid() returns \c false. + \sa serialized() */ QShader QShader::fromSerialized(const QByteArray &data) @@ -425,12 +547,17 @@ QShader QShader::fromSerialized(const QByteArray &data) return QShader(); QShader bs; + bs.detach(); // to get d created QShaderPrivate *d = QShaderPrivate::get(&bs); Q_ASSERT(d->ref.loadRelaxed() == 1); // must be detached int intVal; ds >> intVal; d->qsbVersion = intVal; if (d->qsbVersion != QShaderPrivate::QSB_VERSION + && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS + && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO + && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO + && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITHOUT_VAR_ARRAYDIMS && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITH_CBOR && d->qsbVersion != QShaderPrivate::QSB_VERSION_WITH_BINARY_JSON @@ -486,19 +613,123 @@ QShader QShader::fromSerialized(const QByteArray &data) } } + if (d->qsbVersion > QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS) { + ds >> count; + for (int i = 0; i < count; ++i) { + QShaderKey k; + readShaderKey(&ds, &k); + SeparateToCombinedImageSamplerMappingList list; + int listSize; + ds >> listSize; + for (int b = 0; b < listSize; ++b) { + QByteArray combinedSamplerName; + ds >> combinedSamplerName; + int textureBinding; + ds >> textureBinding; + int samplerBinding; + ds >> samplerBinding; + list.append({ combinedSamplerName, textureBinding, samplerBinding }); + } + d->combinedImageMap.insert(k, list); + } + } + + if (d->qsbVersion > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) { + ds >> count; + for (int i = 0; i < count; ++i) { + QShaderKey k; + readShaderKey(&ds, &k); + int flags; + ds >> flags; + QMap<int, int> extraBufferBindings; + int mapSize; + ds >> mapSize; + for (int b = 0; b < mapSize; ++b) { + int k, v; + ds >> k; + ds >> v; + extraBufferBindings.insert(k, v); + } + d->nativeShaderInfoMap.insert(k, { flags, extraBufferBindings }); + } + } + return bs; } +/*! + \fn QShaderVersion::QShaderVersion() = default + */ + +/*! + Constructs a new QShaderVersion with version \a v and flags \a f. + */ QShaderVersion::QShaderVersion(int v, Flags f) : m_version(v), m_flags(f) { } +/*! + \fn int QShaderVersion::version() const + \return the version. + */ + +/*! + \fn void QShaderVersion::setVersion(int v) + Sets the shading language version to \a v. + */ + +/*! + \fn QShaderVersion::Flags QShaderVersion::flags() const + \return the flags. + */ + +/*! + \fn void QShaderVersion::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QShaderCode::QShaderCode() = default + */ + +/*! + Constructs a new QShaderCode with the specified shader source \a code and + \a entry point name. + */ QShaderCode::QShaderCode(const QByteArray &code, const QByteArray &entry) : m_shader(code), m_entryPoint(entry) { } +/*! + \fn QByteArray QShaderCode::shader() const + \return the shader source or bytecode. + */ + +/*! + \fn void QShaderCode::setShader(const QByteArray &code) + Sets the shader source or byte \a code. + */ + +/*! + \fn QByteArray QShaderCode::entryPoint() const + \return the entry point name. + */ + +/*! + \fn void QShaderCode::setEntryPoint(const QByteArray &entry) + Sets the \a entry point name. + */ + +/*! + \fn QShaderKey::QShaderKey() = default + */ + +/*! + Constructs a new QShaderKey with shader type \a s, version \a sver, and + variant \a svar. + */ QShaderKey::QShaderKey(QShader::Source s, const QShaderVersion &sver, QShader::Variant svar) @@ -509,6 +740,36 @@ QShaderKey::QShaderKey(QShader::Source s, } /*! + \fn QShader::Source QShaderKey::source() const + \return the shader type. + */ + +/*! + \fn void QShaderKey::setSource(QShader::Source s) + Sets the shader type \a s. + */ + +/*! + \fn QShaderVersion QShaderKey::sourceVersion() const + \return the shading language version. + */ + +/*! + \fn void QShaderKey::setSourceVersion(const QShaderVersion &sver) + Sets the shading language version \a sver. + */ + +/*! + \fn QShader::Variant QShaderKey::sourceVariant() const + \return the type of the variant to use. + */ + +/*! + \fn void QShaderKey::setSourceVariant(QShader::Variant svar) + Sets the type of variant to use to \a svar. + */ + +/*! Returns \c true if the two QShader objects \a lhs and \a rhs are equal, meaning they are for the same stage with matching sets of shader source or binary code. @@ -517,16 +778,18 @@ QShaderKey::QShaderKey(QShader::Source s, */ bool operator==(const QShader &lhs, const QShader &rhs) noexcept { + if (!lhs.d || !rhs.d) + return lhs.d == rhs.d; + return lhs.d->stage == rhs.d->stage - && lhs.d->shaders == rhs.d->shaders; - // do not bother with desc and bindings, if the shader code is the same, the description must match too + && lhs.d->shaders == rhs.d->shaders + && lhs.d->bindings == rhs.d->bindings; } /*! - \internal \fn bool operator!=(const QShader &lhs, const QShader &rhs) - Returns \c false if the values in the two QShader objects \a a and \a b + Returns \c false if the values in the two QShader objects \a lhs and \a rhs are equal; otherwise returns \c true. \relates QShader @@ -539,11 +802,14 @@ bool operator==(const QShader &lhs, const QShader &rhs) noexcept */ size_t qHash(const QShader &s, size_t seed) noexcept { - QtPrivate::QHashCombine hash; - seed = hash(seed, s.stage()); - seed = qHashRange(s.d->shaders.keyValueBegin(), - s.d->shaders.keyValueEnd(), - seed); + if (s.d) { + QtPrivate::QHashCombine hash; + seed = hash(seed, s.stage()); + if (!s.d->shaders.isEmpty()) { + seed = hash(seed, s.d->shaders.firstKey()); + seed = hash(seed, std::as_const(s.d->shaders).first()); + } + } return seed; } @@ -566,11 +832,28 @@ size_t qHash(const QShaderVersion &s, size_t seed) noexcept #endif /*! - \internal + \return true if \a lhs is smaller than \a rhs. + + Establishes a sorting order between the two QShaderVersion \a lhs and \a rhs. + + \relates QShaderVersion + */ +bool operator<(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept +{ + if (lhs.version() < rhs.version()) + return true; + + if (lhs.version() == rhs.version()) + return int(lhs.flags()) < int(rhs.flags()); + + return false; +} + +/*! \fn bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) - Returns \c false if the values in the two QShaderVersion objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderVersion objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderVersion */ @@ -587,11 +870,34 @@ bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept } /*! - \internal + \return true if \a lhs is smaller than \a rhs. + + Establishes a sorting order between the two keys \a lhs and \a rhs. + + \relates QShaderKey + */ +bool operator<(const QShaderKey &lhs, const QShaderKey &rhs) noexcept +{ + if (int(lhs.source()) < int(rhs.source())) + return true; + + if (int(lhs.source()) == int(rhs.source())) { + if (lhs.sourceVersion() < rhs.sourceVersion()) + return true; + if (lhs.sourceVersion() == rhs.sourceVersion()) { + if (int(lhs.sourceVariant()) < int(rhs.sourceVariant())) + return true; + } + } + + return false; +} + +/*! \fn bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) - Returns \c false if the values in the two QShaderKey objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderKey objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderKey */ @@ -621,11 +927,10 @@ bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept } /*! - \internal \fn bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) - Returns \c false if the values in the two QShaderCode objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderCode objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderCode */ @@ -646,11 +951,15 @@ QDebug operator<<(QDebug dbg, const QShader &bs) const QShaderPrivate *d = bs.d; QDebugStateSaver saver(dbg); - dbg.nospace() << "QShader(" - << "stage=" << d->stage - << " shaders=" << d->shaders.keys() - << " desc.isValid=" << d->desc.isValid() - << ')'; + if (d) { + dbg.nospace() << "QShader(" + << "stage=" << d->stage + << " shaders=" << d->shaders.keys() + << " desc.isValid=" << d->desc.isValid() + << ')'; + } else { + dbg.nospace() << "QShader()"; + } return dbg; } @@ -675,7 +984,7 @@ QDebug operator<<(QDebug dbg, const QShaderVersion &v) /*! \typedef QShader::NativeResourceBindingMap - Synonym for QHash<int, QPair<int, int>>. + Synonym for QMap<int, QPair<int, int>>. The resource binding model QRhi assumes is based on SPIR-V. This means that uniform buffers, storage buffers, combined image samplers, and storage @@ -684,17 +993,20 @@ QDebug operator<<(QDebug dbg, const QShaderVersion &v) \c binding layout qualifier in the Vulkan-compatible GLSL shader. Graphics APIs other than Vulkan may use a resource binding model that is - not fully compatible with this. In addition, the generator of the shader - code translated from SPIR-V may choose not to take the SPIR-V binding - qualifiers into account, for various reasons. (this is the case with the - Metal backend of SPIRV-Cross, for example). + not fully compatible with this. The generator of the shader code translated + from SPIR-V may choose not to take the SPIR-V binding qualifiers into + account, for various reasons. This is the case with the Metal backend of + SPIRV-Cross, for example. In addition, even when an automatic, implicit + translation is mostly possible (e.g. by using SPIR-V binding points as HLSL + resource register indices), assigning resource bindings without being + constrained by the SPIR-V binding points can lead to better results. Therefore, a QShader may expose an additional map that describes what the - native binding point for a given SPIR-V binding is. The QRhi backends are - expected to use this map automatically, as appropriate. The value is a - pair, because combined image samplers may map to two native resources (a - texture and a sampler) in some shading languages. In that case the second - value refers to the sampler. + native binding point for a given SPIR-V binding is. The QRhi backends, for + which this is relevant, are expected to use this map automatically, as + appropriate. The value is a pair, because combined image samplers may map + to two native resources (a texture and a sampler) in some shading + languages. In that case the second value refers to the sampler. \note The native binding may be -1, in case there is no active binding for the resource in the shader. (for example, there is a uniform block @@ -705,16 +1017,20 @@ QDebug operator<<(QDebug dbg, const QShaderVersion &v) */ /*! - \return the native binding map for \a key or null if no extra mapping is - available, or is not applicable. + \return the native binding map for \a key. The map is empty if no mapping + is available for \a key (for example, because the map is not applicable for + the API and shading language described by \a key). */ -const QShader::NativeResourceBindingMap *QShader::nativeResourceBindingMap(const QShaderKey &key) const +QShader::NativeResourceBindingMap QShader::nativeResourceBindingMap(const QShaderKey &key) const { + if (!d) + return {}; + auto it = d->bindings.constFind(key); if (it == d->bindings.cend()) - return nullptr; + return {}; - return &it.value(); + return it.value(); } /*! @@ -733,6 +1049,9 @@ void QShader::setResourceBindingMap(const QShaderKey &key, const NativeResourceB */ void QShader::removeResourceBindingMap(const QShaderKey &key) { + if (!d) + return; + auto it = d->bindings.find(key); if (it == d->bindings.end()) return; @@ -741,4 +1060,162 @@ void QShader::removeResourceBindingMap(const QShaderKey &key) d->bindings.erase(it); } +/*! + \typedef QShader::SeparateToCombinedImageSamplerMappingList + + Synonym for QList<QShader::SeparateToCombinedImageSamplerMapping>. + */ + +/*! + \struct QShader::SeparateToCombinedImageSamplerMapping + \inmodule QtGui + \brief Mapping metadata for sampler uniforms. + + Describes a mapping from a traditional combined image sampler uniform to + binding points for a separate texture and sampler. + + For example, if \c combinedImageSampler is \c{"_54"}, \c textureBinding is + \c 1, and \c samplerBinding is \c 2, this means that the GLSL shader code + contains a \c sampler2D (or sampler3D, etc.) uniform with the name of + \c{_54} which corresponds to two separate resource bindings (\c 1 and \c 2) + in the original shader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. + */ + +/*! + \variable QShader::SeparateToCombinedImageSamplerMapping::combinedSamplerName +*/ + +/*! + \variable QShader::SeparateToCombinedImageSamplerMapping::textureBinding +*/ + +/*! + \variable QShader::SeparateToCombinedImageSamplerMapping::samplerBinding +*/ + +/*! + \return the combined image sampler mapping list for \a key, or an empty + list if there is no data available for \a key, for example because such a + mapping is not applicable for the shading language. + */ +QShader::SeparateToCombinedImageSamplerMappingList QShader::separateToCombinedImageSamplerMappingList(const QShaderKey &key) const +{ + if (!d) + return {}; + + auto it = d->combinedImageMap.constFind(key); + if (it == d->combinedImageMap.cend()) + return {}; + + return it.value(); +} + +/*! + Stores the given combined image sampler mapping \a list associated with \a key. + + \sa separateToCombinedImageSamplerMappingList() + */ +void QShader::setSeparateToCombinedImageSamplerMappingList(const QShaderKey &key, + const SeparateToCombinedImageSamplerMappingList &list) +{ + detach(); + d->combinedImageMap[key] = list; +} + +/*! + Removes the combined image sampler mapping list for \a key. + */ +void QShader::removeSeparateToCombinedImageSamplerMappingList(const QShaderKey &key) +{ + if (!d) + return; + + auto it = d->combinedImageMap.find(key); + if (it == d->combinedImageMap.end()) + return; + + detach(); + d->combinedImageMap.erase(it); +} + +/*! + \struct QShader::NativeShaderInfo + \inmodule QtGui + \brief Additional metadata about the native shader code. + + Describes information about the native shader code, if applicable. This + becomes relevant with certain shader languages for certain shader stages, + in case the translation from SPIR-V involves the introduction of + additional, "magic" inputs, outputs, or resources in the generated shader. + Such additions may be dependent on the original source code (i.e. the usage + of various GLSL language constructs or built-ins), and therefore it needs + to be indicated in a dynamic manner if certain features got added to the + generated shader code. + + As an example, consider a tessellation control shader with a per-patch (not + per-vertex) output variable. This is translated to a Metal compute shader + outputting (among others) into an spvPatchOut buffer. But this buffer would + not be present at all if per-patch output variables were not used. The fact + that the shader code relies on such a buffer present can be indicated by + the data in this struct. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. + */ + +/*! + \variable QShader::NativeShaderInfo::flags +*/ + +/*! + \variable QShader::NativeShaderInfo::extraBufferBindings +*/ + +/*! + \return the native shader info struct for \a key, or an empty object if + there is no data available for \a key, for example because such a mapping + is not applicable for the shading language or the shader stage. + */ +QShader::NativeShaderInfo QShader::nativeShaderInfo(const QShaderKey &key) const +{ + if (!d) + return {}; + + auto it = d->nativeShaderInfoMap.constFind(key); + if (it == d->nativeShaderInfoMap.cend()) + return {}; + + return it.value(); +} + +/*! + Stores the given native shader \a info associated with \a key. + + \sa nativeShaderInfo() + */ +void QShader::setNativeShaderInfo(const QShaderKey &key, const NativeShaderInfo &info) +{ + detach(); + d->nativeShaderInfoMap[key] = info; +} + +/*! + Removes the native shader information for \a key. + */ +void QShader::removeNativeShaderInfo(const QShaderKey &key) +{ + if (!d) + return; + + auto it = d->nativeShaderInfoMap.find(key); + if (it == d->nativeShaderInfoMap.end()) + return; + + detach(); + d->nativeShaderInfoMap.erase(it); +} + QT_END_NAMESPACE diff --git a/src/gui/rhi/qshader.h b/src/gui/rhi/qshader.h new file mode 100644 index 0000000000..2465081366 --- /dev/null +++ b/src/gui/rhi/qshader.h @@ -0,0 +1,241 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QSHADER_H +#define QSHADER_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qhash.h> +#include <QtCore/qmap.h> +#include <rhi/qshaderdescription.h> + +QT_BEGIN_NAMESPACE + +struct QShaderPrivate; +class QShaderKey; + +#ifdef Q_OS_INTEGRITY + class QShaderVersion; + size_t qHash(const QShaderVersion &, size_t = 0) noexcept; +#endif + +class Q_GUI_EXPORT QShaderVersion +{ +public: + enum Flag { + GlslEs = 0x01 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QShaderVersion() = default; + QShaderVersion(int v, Flags f = Flags()); + + int version() const { return m_version; } + void setVersion(int v) { m_version = v; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + +private: + int m_version = 100; + Flags m_flags; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderVersion::Flags) +Q_DECLARE_TYPEINFO(QShaderVersion, Q_RELOCATABLE_TYPE); + +class QShaderCode; +Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t = 0) noexcept; + +class Q_GUI_EXPORT QShaderCode +{ +public: + QShaderCode() = default; + QShaderCode(const QByteArray &code, const QByteArray &entry = QByteArray()); + + QByteArray shader() const { return m_shader; } + void setShader(const QByteArray &code) { m_shader = code; } + + QByteArray entryPoint() const { return m_entryPoint; } + void setEntryPoint(const QByteArray &entry) { m_entryPoint = entry; } + +private: + friend Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t) noexcept; + + QByteArray m_shader; + QByteArray m_entryPoint; +}; + +Q_DECLARE_TYPEINFO(QShaderCode, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QShader +{ +public: + enum Stage { + VertexStage = 0, + TessellationControlStage, + TessellationEvaluationStage, + GeometryStage, + FragmentStage, + ComputeStage + }; + + enum Source { + SpirvShader = 0, + GlslShader, + HlslShader, + DxbcShader, // fxc + MslShader, + DxilShader, // dxc + MetalLibShader, // xcrun metal + xcrun metallib + WgslShader + }; + + enum Variant { + StandardShader = 0, + BatchableVertexShader, + UInt16IndexedVertexAsComputeShader, + UInt32IndexedVertexAsComputeShader, + NonIndexedVertexAsComputeShader + }; + + enum class SerializedFormatVersion { + Latest = 0, + Qt_6_5, + Qt_6_4 + }; + + QShader(); + QShader(const QShader &other); + QShader &operator=(const QShader &other); + QShader(QShader &&other) noexcept : d(std::exchange(other.d, nullptr)) {} + QT_MOVE_ASSIGNMENT_OPERATOR_IMPL_VIA_PURE_SWAP(QShader) + ~QShader(); + + void swap(QShader &other) noexcept { qt_ptr_swap(d, other.d); } + void detach(); + + bool isValid() const; + + Stage stage() const; + void setStage(Stage stage); + + QShaderDescription description() const; + void setDescription(const QShaderDescription &desc); + + QList<QShaderKey> availableShaders() const; + QShaderCode shader(const QShaderKey &key) const; + void setShader(const QShaderKey &key, const QShaderCode &shader); + void removeShader(const QShaderKey &key); + + QByteArray serialized(SerializedFormatVersion version = SerializedFormatVersion::Latest) const; + static QShader fromSerialized(const QByteArray &data); + + using NativeResourceBindingMap = QMap<int, QPair<int, int> >; // binding -> native_binding[, native_binding] + NativeResourceBindingMap nativeResourceBindingMap(const QShaderKey &key) const; + void setResourceBindingMap(const QShaderKey &key, const NativeResourceBindingMap &map); + void removeResourceBindingMap(const QShaderKey &key); + + struct SeparateToCombinedImageSamplerMapping { + QByteArray combinedSamplerName; + int textureBinding; + int samplerBinding; + }; + using SeparateToCombinedImageSamplerMappingList = QList<SeparateToCombinedImageSamplerMapping>; + SeparateToCombinedImageSamplerMappingList separateToCombinedImageSamplerMappingList(const QShaderKey &key) const; + void setSeparateToCombinedImageSamplerMappingList(const QShaderKey &key, + const SeparateToCombinedImageSamplerMappingList &list); + void removeSeparateToCombinedImageSamplerMappingList(const QShaderKey &key); + + struct NativeShaderInfo { + int flags = 0; + QMap<int, int> extraBufferBindings; + }; + NativeShaderInfo nativeShaderInfo(const QShaderKey &key) const; + void setNativeShaderInfo(const QShaderKey &key, const NativeShaderInfo &info); + void removeNativeShaderInfo(const QShaderKey &key); + +private: + QShaderPrivate *d; + friend struct QShaderPrivate; + friend Q_GUI_EXPORT bool operator==(const QShader &, const QShader &) noexcept; + friend Q_GUI_EXPORT size_t qHash(const QShader &, size_t) noexcept; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); +#endif +}; + +class Q_GUI_EXPORT QShaderKey +{ +public: + QShaderKey() = default; + QShaderKey(QShader::Source s, + const QShaderVersion &sver, + QShader::Variant svar = QShader::StandardShader); + + QShader::Source source() const { return m_source; } + void setSource(QShader::Source s) { m_source = s; } + + QShaderVersion sourceVersion() const { return m_sourceVersion; } + void setSourceVersion(const QShaderVersion &sver) { m_sourceVersion = sver; } + + QShader::Variant sourceVariant() const { return m_sourceVariant; } + void setSourceVariant(QShader::Variant svar) { m_sourceVariant = svar; } + +private: + QShader::Source m_source = QShader::SpirvShader; + QShaderVersion m_sourceVersion; + QShader::Variant m_sourceVariant = QShader::StandardShader; +}; + +Q_DECLARE_TYPEINFO(QShaderKey, Q_RELOCATABLE_TYPE); + +Q_GUI_EXPORT bool operator==(const QShader &lhs, const QShader &rhs) noexcept; +Q_GUI_EXPORT size_t qHash(const QShader &s, size_t seed = 0) noexcept; + +inline bool operator!=(const QShader &lhs, const QShader &rhs) noexcept +{ + return !(lhs == rhs); +} + +Q_GUI_EXPORT bool operator==(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; +Q_GUI_EXPORT bool operator<(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; +Q_GUI_EXPORT bool operator<(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept; + +inline bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) noexcept +{ + return !(lhs == rhs); +} + +Q_GUI_EXPORT size_t qHash(const QShaderKey &k, size_t seed = 0) noexcept; + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); +Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderKey &k); +Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderVersion &v); +#endif + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qshader_p.h b/src/gui/rhi/qshader_p.h index b320340229..f77bcb1259 100644 --- a/src/gui/rhi/qshader_p.h +++ b/src/gui/rhi/qshader_p.h @@ -1,41 +1,5 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only #ifndef QSHADER_P_H #define QSHADER_P_H @@ -51,192 +15,77 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <private/qshaderdescription_p.h> +#include <rhi/qshader.h> +#include <QtCore/QAtomicInt> +#include <QtCore/QMap> +#include <QtCore/QDebug> QT_BEGIN_NAMESPACE -struct QShaderPrivate; -class QShaderKey; - -#ifdef Q_OS_INTEGRITY - class QShaderVersion; - size_t qHash(const QShaderVersion &, size_t = 0) noexcept; -#endif - -class Q_GUI_EXPORT QShaderVersion +struct Q_GUI_EXPORT QShaderPrivate { -public: - enum Flag { - GlslEs = 0x01 + static const int QSB_VERSION = 9; + static const int QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS = 8; + static const int QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO = 7; + static const int QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO = 6; + static const int QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS = 5; + static const int QSB_VERSION_WITHOUT_VAR_ARRAYDIMS = 4; + static const int QSB_VERSION_WITH_CBOR = 3; + static const int QSB_VERSION_WITH_BINARY_JSON = 2; + static const int QSB_VERSION_WITHOUT_BINDINGS = 1; + + enum MslNativeShaderInfoExtraBufferBindings { + MslTessVertIndicesBufferBinding = 0, + MslTessVertTescOutputBufferBinding, + MslTessTescTessLevelBufferBinding, + MslTessTescPatchOutputBufferBinding, + MslTessTescParamsBufferBinding, + MslTessTescInputBufferBinding, + MslBufferSizeBufferBinding, + MslMultiViewMaskBufferBinding }; - Q_DECLARE_FLAGS(Flags, Flag) - - QShaderVersion() = default; - QShaderVersion(int v, Flags f = Flags()); - - int version() const { return m_version; } - void setVersion(int v) { m_version = v; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - -private: - int m_version = 100; - Flags m_flags; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderVersion::Flags) -Q_DECLARE_TYPEINFO(QShaderVersion, Q_RELOCATABLE_TYPE); - -class QShaderCode; -Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t = 0) noexcept; - -class Q_GUI_EXPORT QShaderCode -{ -public: - QShaderCode() = default; - QShaderCode(const QByteArray &code, const QByteArray &entry = QByteArray()); - - QByteArray shader() const { return m_shader; } - void setShader(const QByteArray &code) { m_shader = code; } - - QByteArray entryPoint() const { return m_entryPoint; } - void setEntryPoint(const QByteArray &entry) { m_entryPoint = entry; } - -private: - friend Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t) noexcept; - - QByteArray m_shader; - QByteArray m_entryPoint; -}; - -Q_DECLARE_TYPEINFO(QShaderCode, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QShader -{ -public: - enum Stage { - VertexStage = 0, - TessellationControlStage, - TessellationEvaluationStage, - GeometryStage, - FragmentStage, - ComputeStage - }; - - enum Source { - SpirvShader = 0, - GlslShader, - HlslShader, - DxbcShader, // fxc - MslShader, - DxilShader, // dxc - MetalLibShader // xcrun metal + xcrun metallib - }; - - enum Variant { - StandardShader = 0, - BatchableVertexShader - }; - - QShader(); - QShader(const QShader &other); - QShader &operator=(const QShader &other); - ~QShader(); - void detach(); - - bool isValid() const; - Stage stage() const; - void setStage(Stage stage); - - QShaderDescription description() const; - void setDescription(const QShaderDescription &desc); - - QList<QShaderKey> availableShaders() const; - QShaderCode shader(const QShaderKey &key) const; - void setShader(const QShaderKey &key, const QShaderCode &shader); - void removeShader(const QShaderKey &key); - - QByteArray serialized() const; - static QShader fromSerialized(const QByteArray &data); - - using NativeResourceBindingMap = QHash<int, QPair<int, int> >; // binding -> native_binding[, native_binding] - const NativeResourceBindingMap *nativeResourceBindingMap(const QShaderKey &key) const; - void setResourceBindingMap(const QShaderKey &key, const NativeResourceBindingMap &map); - void removeResourceBindingMap(const QShaderKey &key); - -private: - QShaderPrivate *d; - friend struct QShaderPrivate; - friend Q_GUI_EXPORT bool operator==(const QShader &, const QShader &) noexcept; - friend Q_GUI_EXPORT size_t qHash(const QShader &, size_t) noexcept; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); -#endif -}; - -class Q_GUI_EXPORT QShaderKey -{ -public: - QShaderKey() = default; - QShaderKey(QShader::Source s, - const QShaderVersion &sver, - QShader::Variant svar = QShader::StandardShader); - - QShader::Source source() const { return m_source; } - void setSource(QShader::Source s) { m_source = s; } - - QShaderVersion sourceVersion() const { return m_sourceVersion; } - void setSourceVersion(const QShaderVersion &sver) { m_sourceVersion = sver; } - - QShader::Variant sourceVariant() const { return m_sourceVariant; } - void setSourceVariant(QShader::Variant svar) { m_sourceVariant = svar; } - -private: - QShader::Source m_source = QShader::SpirvShader; - QShaderVersion m_sourceVersion; - QShader::Variant m_sourceVariant = QShader::StandardShader; + QShaderPrivate() + : ref(1) + { + } + + QShaderPrivate(const QShaderPrivate &other) + : ref(1), + qsbVersion(other.qsbVersion), + stage(other.stage), + desc(other.desc), + shaders(other.shaders), + bindings(other.bindings), + combinedImageMap(other.combinedImageMap), + nativeShaderInfoMap(other.nativeShaderInfoMap) + { + } + + static QShaderPrivate *get(QShader *s) { return s->d; } + static const QShaderPrivate *get(const QShader *s) { return s->d; } + static int qtQsbVersion(QShader::SerializedFormatVersion qtVersion) { + switch (qtVersion) { + case QShader::SerializedFormatVersion::Qt_6_4: + return (QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS + 1); + case QShader::SerializedFormatVersion::Qt_6_5: + return (QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO + 1); + default: + return QShaderPrivate::QSB_VERSION; + } + } + + QAtomicInt ref; + int qsbVersion = QSB_VERSION; + QShader::Stage stage = QShader::VertexStage; + QShaderDescription desc; + // QMap not QHash because we need to be able to iterate based on sorted keys + QMap<QShaderKey, QShaderCode> shaders; + QMap<QShaderKey, QShader::NativeResourceBindingMap> bindings; + QMap<QShaderKey, QShader::SeparateToCombinedImageSamplerMappingList> combinedImageMap; + QMap<QShaderKey, QShader::NativeShaderInfo> nativeShaderInfoMap; }; -Q_DECLARE_TYPEINFO(QShaderKey, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QShader &lhs, const QShader &rhs) noexcept; -Q_GUI_EXPORT size_t qHash(const QShader &s, size_t seed = 0) noexcept; - -inline bool operator!=(const QShader &lhs, const QShader &rhs) noexcept -{ - return !(lhs == rhs); -} - -Q_GUI_EXPORT bool operator==(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept; - -inline bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) noexcept -{ - return !(lhs == rhs); -} - -Q_GUI_EXPORT size_t qHash(const QShaderKey &k, size_t seed = 0) noexcept; - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); -Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderKey &k); -Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderVersion &v); -#endif - QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qshader_p_p.h b/src/gui/rhi/qshader_p_p.h deleted file mode 100644 index 2f6afb4a6d..0000000000 --- a/src/gui/rhi/qshader_p_p.h +++ /dev/null @@ -1,97 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QSHADER_P_P_H -#define QSHADER_P_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists for the convenience -// of a number of Qt sources files. This header file may change from -// version to version without notice, or even be removed. -// -// We mean it. -// - -#include "qshader_p.h" -#include <QtCore/QAtomicInt> -#include <QtCore/QHash> -#include <QtCore/QDebug> - -QT_BEGIN_NAMESPACE - -struct Q_GUI_EXPORT QShaderPrivate -{ - static const int QSB_VERSION = 5; - static const int QSB_VERSION_WITHOUT_VAR_ARRAYDIMS = 4; - static const int QSB_VERSION_WITH_CBOR = 3; - static const int QSB_VERSION_WITH_BINARY_JSON = 2; - static const int QSB_VERSION_WITHOUT_BINDINGS = 1; - - QShaderPrivate() - : ref(1) - { - } - - QShaderPrivate(const QShaderPrivate *other) - : ref(1), - qsbVersion(other->qsbVersion), - stage(other->stage), - desc(other->desc), - shaders(other->shaders), - bindings(other->bindings) - { - } - - static QShaderPrivate *get(QShader *s) { return s->d; } - static const QShaderPrivate *get(const QShader *s) { return s->d; } - - QAtomicInt ref; - int qsbVersion = QSB_VERSION; - QShader::Stage stage = QShader::VertexStage; - QShaderDescription desc; - QHash<QShaderKey, QShaderCode> shaders; - QHash<QShaderKey, QShader::NativeResourceBindingMap> bindings; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qshaderdescription.cpp b/src/gui/rhi/qshaderdescription.cpp index 2c79acb1c7..f64daf02ef 100644 --- a/src/gui/rhi/qshaderdescription.cpp +++ b/src/gui/rhi/qshaderdescription.cpp @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#include "qshaderdescription_p_p.h" -#include "qshader_p_p.h" +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qshaderdescription_p.h" +#include "qshader_p.h" #include <QDebug> #include <QDataStream> #include <QJsonObject> @@ -48,11 +12,22 @@ QT_BEGIN_NAMESPACE /*! \class QShaderDescription - \internal + \ingroup painting-3D \inmodule QtGui + \since 6.6 \brief Describes the interface of a shader. + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qshaderdescription.h>}. + A shader typically has a set of inputs and outputs. A vertex shader for example has a number of input variables and may use one or more uniform buffers to access data (e.g. a modelview matrix) provided by the @@ -87,8 +62,6 @@ QT_BEGIN_NAMESPACE float opacity; } ubuf; - out gl_PerVertex { vec4 gl_Position; }; - void main() { v_color = color; @@ -222,6 +195,7 @@ QT_BEGIN_NAMESPACE \value SamplerRect \value SamplerBuffer \value SamplerExternalOES + \value Sampler For separate samplers. \value Image1D \value Image2D \value Image2DMS @@ -235,28 +209,179 @@ QT_BEGIN_NAMESPACE \value ImageRect \value ImageBuffer \value Struct + \value Half + \value Half2 + \value Half3 + \value Half4 */ /*! - \class QShaderDescription::InOutVariable - \internal + \enum QShaderDescription::ImageFormat + Image format. + + \value ImageFormatUnknown + \value ImageFormatRgba32f + \value ImageFormatRgba16f + \value ImageFormatR32f + \value ImageFormatRgba8 + \value ImageFormatRgba8Snorm + \value ImageFormatRg32f + \value ImageFormatRg16f + \value ImageFormatR11fG11fB10f + \value ImageFormatR16f + \value ImageFormatRgba16 + \value ImageFormatRgb10A2 + \value ImageFormatRg16 + \value ImageFormatRg8 + \value ImageFormatR16 + \value ImageFormatR8 + \value ImageFormatRgba16Snorm + \value ImageFormatRg16Snorm + \value ImageFormatRg8Snorm + \value ImageFormatR16Snorm + \value ImageFormatR8Snorm + \value ImageFormatRgba32i + \value ImageFormatRgba16i + \value ImageFormatRgba8i + \value ImageFormatR32i + \value ImageFormatRg32i + \value ImageFormatRg16i + \value ImageFormatRg8i + \value ImageFormatR16i + \value ImageFormatR8i + \value ImageFormatRgba32ui + \value ImageFormatRgba16ui + \value ImageFormatRgba8ui + \value ImageFormatR32ui + \value ImageFormatRgb10a2ui + \value ImageFormatRg32ui + \value ImageFormatRg16ui + \value ImageFormatRg8ui + \value ImageFormatR16ui + \value ImageFormatR8ui + */ + +/*! + \enum QShaderDescription::ImageFlag + Image flags. + + \value ReadOnlyImage + \value WriteOnlyImage + */ + +/*! + \enum QShaderDescription::QualifierFlag + Qualifier flags. + + \value QualifierReadOnly + \value QualifierWriteOnly + \value QualifierCoherent + \value QualifierVolatile + \value QualifierRestrict + */ + +/*! + \struct QShaderDescription::InOutVariable \inmodule QtGui + \since 6.6 \brief Describes an input or output variable in the shader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::BlockVariable - \internal + \variable QShaderDescription::InOutVariable::name + */ + +/*! + \variable QShaderDescription::InOutVariable::type + */ + +/*! + \variable QShaderDescription::InOutVariable::location + */ + +/*! + \variable QShaderDescription::InOutVariable::binding + */ + +/*! + \variable QShaderDescription::InOutVariable::descriptorSet + */ + +/*! + \variable QShaderDescription::InOutVariable::imageFormat + */ + +/*! + \variable QShaderDescription::InOutVariable::imageFlags + */ + +/*! + \variable QShaderDescription::InOutVariable::arrayDims + */ + +/*! + \variable QShaderDescription::InOutVariable::perPatch + */ + +/*! + \variable QShaderDescription::InOutVariable::structMembers + */ + +/*! + \struct QShaderDescription::BlockVariable \inmodule QtGui + \since 6.6 \brief Describes a member of a uniform or push constant block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::UniformBlock - \internal + \variable QShaderDescription::BlockVariable::name + */ + +/*! + \variable QShaderDescription::BlockVariable::type + */ + +/*! + \variable QShaderDescription::BlockVariable::offset + */ + +/*! + \variable QShaderDescription::BlockVariable::size + */ + +/*! + \variable QShaderDescription::BlockVariable::arrayDims + */ + +/*! + \variable QShaderDescription::BlockVariable::arrayStride + */ + +/*! + \variable QShaderDescription::BlockVariable::matrixStride + */ + +/*! + \variable QShaderDescription::BlockVariable::matrixIsRowMajor + */ + +/*! + \variable QShaderDescription::BlockVariable::structMembers + */ + +/*! + \struct QShaderDescription::UniformBlock \inmodule QtGui + \since 6.6 \brief Describes a uniform block. @@ -264,22 +389,157 @@ QT_BEGIN_NAMESPACE (like GLSL 120 or GLSL/ES 100), uniform blocks are replaced with ordinary uniforms in a struct. The name of the struct, and so the prefix for the uniforms generated from the block members, is given by structName. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::PushConstantBlock - \internal + \variable QShaderDescription::UniformBlock::blockName + */ + +/*! + \variable QShaderDescription::UniformBlock::structName + */ + +/*! + \variable QShaderDescription::UniformBlock::size + */ + +/*! + \variable QShaderDescription::UniformBlock::binding + */ + +/*! + \variable QShaderDescription::UniformBlock::descriptorSet + */ + +/*! + \variable QShaderDescription::UniformBlock::members + */ + +/*! + \struct QShaderDescription::PushConstantBlock \inmodule QtGui + \since 6.6 \brief Describes a push constant block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::StorageBlock - \internal + \variable QShaderDescription::PushConstantBlock::name + */ + +/*! + \variable QShaderDescription::PushConstantBlock::size + */ + +/*! + \variable QShaderDescription::PushConstantBlock::members + */ + +/*! + \struct QShaderDescription::StorageBlock \inmodule QtGui + \since 6.6 \brief Describes a shader storage block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. + */ + +/*! + \variable QShaderDescription::StorageBlock::blockName + */ + +/*! + \variable QShaderDescription::StorageBlock::instanceName + */ + +/*! + \variable QShaderDescription::StorageBlock::knownSize + */ + +/*! + \variable QShaderDescription::StorageBlock::binding + */ + +/*! + \variable QShaderDescription::StorageBlock::descriptorSet + */ + +/*! + \variable QShaderDescription::StorageBlock::members + */ + +/*! + \variable QShaderDescription::StorageBlock::runtimeArrayStride + */ + +/*! + \variable QShaderDescription::StorageBlock::qualifierFlags + */ + +/*! + \struct QShaderDescription::BuiltinVariable + \inmodule QtGui + \since 6.6 + + \brief Describes a built-in variable. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. + */ + +/*! + \variable QShaderDescription::BuiltinVariable::type + */ + +/*! + \variable QShaderDescription::BuiltinVariable::varType + */ + +/*! + \variable QShaderDescription::BuiltinVariable::arrayDims + */ + +/*! + \enum QShaderDescription::BuiltinType + Built-in variable type. + + \value PositionBuiltin + \value PointSizeBuiltin + \value ClipDistanceBuiltin + \value CullDistanceBuiltin + \value VertexIdBuiltin + \value InstanceIdBuiltin + \value PrimitiveIdBuiltin + \value InvocationIdBuiltin + \value LayerBuiltin + \value ViewportIndexBuiltin + \value TessLevelOuterBuiltin + \value TessLevelInnerBuiltin + \value TessCoordBuiltin + \value PatchVerticesBuiltin + \value FragCoordBuiltin + \value PointCoordBuiltin + \value FrontFacingBuiltin + \value SampleIdBuiltin + \value SamplePositionBuiltin + \value SampleMaskBuiltin + \value FragDepthBuiltin + \value NumWorkGroupsBuiltin + \value WorkgroupSizeBuiltin + \value WorkgroupIdBuiltin + \value LocalInvocationIdBuiltin + \value GlobalInvocationIdBuiltin + \value LocalInvocationIndexBuiltin + \value VertexIndexBuiltin + \value InstanceIndexBuiltin */ /*! @@ -302,7 +562,7 @@ void QShaderDescription::detach() } /*! - \internal + Constructs a copy of \a other. */ QShaderDescription::QShaderDescription(const QShaderDescription &other) : d(other.d) @@ -311,7 +571,7 @@ QShaderDescription::QShaderDescription(const QShaderDescription &other) } /*! - \internal + Assigns \a other to this object. */ QShaderDescription &QShaderDescription::operator=(const QShaderDescription &other) { @@ -336,7 +596,9 @@ bool QShaderDescription::isValid() const { return !d->inVars.isEmpty() || !d->outVars.isEmpty() || !d->uniformBlocks.isEmpty() || !d->pushConstantBlocks.isEmpty() || !d->storageBlocks.isEmpty() - || !d->combinedImageSamplers.isEmpty() || !d->storageImages.isEmpty(); + || !d->combinedImageSamplers.isEmpty() || !d->storageImages.isEmpty() + || !d->separateImages.isEmpty() || !d->separateSamplers.isEmpty() + || !d->inBuiltins.isEmpty() || !d->outBuiltins.isEmpty(); } /*! @@ -352,13 +614,14 @@ QByteArray QShaderDescription::toJson() const } /*! - Serializes this QShaderDescription to \a stream. + Serializes this QShaderDescription to \a stream. \a version specifies + the qsb version. \sa deserialize(), toJson() */ -void QShaderDescription::serialize(QDataStream *stream) const +void QShaderDescription::serialize(QDataStream *stream, int version) const { - d->writeToStream(stream); + d->writeToStream(stream, version); } /*! @@ -436,6 +699,7 @@ QList<QShaderDescription::PushConstantBlock> QShaderDescription::pushConstantBlo "blockName": "StuffSsbo", "instanceName": "buf", "knownSize": 16, + "runtimeArrayStride": 16 "members": [ { "name": "whatever", @@ -473,7 +737,10 @@ QList<QShaderDescription::PushConstantBlock> QShaderDescription::pushConstantBlo \note The size of the last member in the storage block is undefined. This shows up as \c size 0 and an array dimension of \c{[0]}. The storage block's \c knownSize - excludes the size of the last member since that will only be known at run time. + excludes the size of the last member since that will only be known at run time. The + stride in bytes between array items for a last member with undefined array size is + \c runtimeArrayStride. This value is determined according to the specified buffer + memory layout standard (std140, std430) rules. \note SSBOs are not available with some graphics APIs, such as, OpenGL 2.x or OpenGL ES older than 3.1. @@ -510,6 +777,16 @@ QList<QShaderDescription::InOutVariable> QShaderDescription::combinedImageSample return d->combinedImageSamplers; } +QList<QShaderDescription::InOutVariable> QShaderDescription::separateImages() const +{ + return d->separateImages; +} + +QList<QShaderDescription::InOutVariable> QShaderDescription::separateSamplers() const +{ + return d->separateSamplers; +} + /*! \return the list of image variables. @@ -538,7 +815,26 @@ QList<QShaderDescription::InOutVariable> QShaderDescription::storageImages() con } /*! - Returns the local size of a compute shader. + \return the list of active builtins used as input. For example, a + tessellation evaluation shader reading the value of gl_TessCoord and + gl_Position will have TessCoordBuiltin and PositionBuiltin listed here. + */ +QVector<QShaderDescription::BuiltinVariable> QShaderDescription::inputBuiltinVariables() const +{ + return d->inBuiltins; +} + +/*! + \return the list of active built-in variables used as input. For example, a + vertex shader will very often have PositionBuiltin as an output built-in. + */ +QVector<QShaderDescription::BuiltinVariable> QShaderDescription::outputBuiltinVariables() const +{ + return d->outBuiltins; +} + +/*! + \return the local size of a compute shader. For example, for a compute shader with the following declaration the function returns { 256, 16, 1}. @@ -552,146 +848,346 @@ std::array<uint, 3> QShaderDescription::computeShaderLocalSize() const return d->localSize; } -static struct TypeTab { - QString k; +/*! + \return the number of output vertices. + + For example, for a tessellation control shader with the following + declaration the function returns 3. + + \badcode + layout(vertices = 3) out; + \endcode + */ +uint QShaderDescription::tessellationOutputVertexCount() const +{ + return d->tessOutVertCount; +} + +/*! + \enum QShaderDescription::TessellationMode + + \value UnknownTessellationMode + \value TrianglesTessellationMode + \value QuadTessellationMode + \value IsolineTessellationMode + */ + +/*! + \return the tessellation execution mode for a tessellation control or + evaluation shader. + + When not set, the returned value is UnknownTessellationMode. + + For example, for a tessellation evaluation shader with the following + declaration the function returns TrianglesTessellationMode. + + \badcode + layout(triangles) in; + \endcode + */ +QShaderDescription::TessellationMode QShaderDescription::tessellationMode() const +{ + return d->tessMode; +} + +/*! + \enum QShaderDescription::TessellationWindingOrder + + \value UnknownTessellationWindingOrder + \value CwTessellationWindingOrder + \value CcwTessellationWindingOrder + */ + +/*! + \return the tessellation winding order for a tessellation control or + evaluation shader. + + When not set, the returned value is UnknownTessellationWindingOrder. + + For example, for a tessellation evaluation shader with the following + declaration the function returns CcwTessellationWindingOrder. + + \badcode + layout(triangles, fractional_odd_spacing, ccw) in; + \endcode + */ +QShaderDescription::TessellationWindingOrder QShaderDescription::tessellationWindingOrder() const +{ + return d->tessWind; +} + +/*! + \enum QShaderDescription::TessellationPartitioning + + \value UnknownTessellationPartitioning + \value EqualTessellationPartitioning + \value FractionalEvenTessellationPartitioning + \value FractionalOddTessellationPartitioning + */ + +/*! + \return the tessellation partitioning mode for a tessellation control or + evaluation shader. + + When not set, the returned value is UnknownTessellationPartitioning. + + For example, for a tessellation evaluation shader with the following + declaration the function returns FractionalOddTessellationPartitioning. + + \badcode + layout(triangles, fractional_odd_spacing, ccw) in; + \endcode + */ +QShaderDescription::TessellationPartitioning QShaderDescription::tessellationPartitioning() const +{ + return d->tessPart; +} + +static const struct TypeTab { + const char k[20]; QShaderDescription::VariableType v; } typeTab[] = { - { QLatin1String("float"), QShaderDescription::Float }, - { QLatin1String("vec2"), QShaderDescription::Vec2 }, - { QLatin1String("vec3"), QShaderDescription::Vec3 }, - { QLatin1String("vec4"), QShaderDescription::Vec4 }, - { QLatin1String("mat2"), QShaderDescription::Mat2 }, - { QLatin1String("mat3"), QShaderDescription::Mat3 }, - { QLatin1String("mat4"), QShaderDescription::Mat4 }, - - { QLatin1String("struct"), QShaderDescription::Struct }, - - { QLatin1String("sampler1D"), QShaderDescription::Sampler1D }, - { QLatin1String("sampler2D"), QShaderDescription::Sampler2D }, - { QLatin1String("sampler2DMS"), QShaderDescription::Sampler2DMS }, - { QLatin1String("sampler3D"), QShaderDescription::Sampler3D }, - { QLatin1String("samplerCube"), QShaderDescription::SamplerCube }, - { QLatin1String("sampler1DArray"), QShaderDescription::Sampler1DArray }, - { QLatin1String("sampler2DArray"), QShaderDescription::Sampler2DArray }, - { QLatin1String("sampler2DMSArray"), QShaderDescription::Sampler2DMSArray }, - { QLatin1String("sampler3DArray"), QShaderDescription::Sampler3DArray }, - { QLatin1String("samplerCubeArray"), QShaderDescription::SamplerCubeArray }, - { QLatin1String("samplerRect"), QShaderDescription::SamplerRect }, - { QLatin1String("samplerBuffer"), QShaderDescription::SamplerBuffer }, - { QLatin1String("samplerExternalOES"), QShaderDescription::SamplerExternalOES }, - - { QLatin1String("mat2x3"), QShaderDescription::Mat2x3 }, - { QLatin1String("mat2x4"), QShaderDescription::Mat2x4 }, - { QLatin1String("mat3x2"), QShaderDescription::Mat3x2 }, - { QLatin1String("mat3x4"), QShaderDescription::Mat3x4 }, - { QLatin1String("mat4x2"), QShaderDescription::Mat4x2 }, - { QLatin1String("mat4x3"), QShaderDescription::Mat4x3 }, - - { QLatin1String("int"), QShaderDescription::Int }, - { QLatin1String("ivec2"), QShaderDescription::Int2 }, - { QLatin1String("ivec3"), QShaderDescription::Int3 }, - { QLatin1String("ivec4"), QShaderDescription::Int4 }, - - { QLatin1String("uint"), QShaderDescription::Uint }, - { QLatin1String("uvec2"), QShaderDescription::Uint2 }, - { QLatin1String("uvec3"), QShaderDescription::Uint3 }, - { QLatin1String("uvec4"), QShaderDescription::Uint4 }, - - { QLatin1String("bool"), QShaderDescription::Bool }, - { QLatin1String("bvec2"), QShaderDescription::Bool2 }, - { QLatin1String("bvec3"), QShaderDescription::Bool3 }, - { QLatin1String("bvec4"), QShaderDescription::Bool4 }, - - { QLatin1String("double"), QShaderDescription::Double }, - { QLatin1String("dvec2"), QShaderDescription::Double2 }, - { QLatin1String("dvec3"), QShaderDescription::Double3 }, - { QLatin1String("dvec4"), QShaderDescription::Double4 }, - { QLatin1String("dmat2"), QShaderDescription::DMat2 }, - { QLatin1String("dmat3"), QShaderDescription::DMat3 }, - { QLatin1String("dmat4"), QShaderDescription::DMat4 }, - { QLatin1String("dmat2x3"), QShaderDescription::DMat2x3 }, - { QLatin1String("dmat2x4"), QShaderDescription::DMat2x4 }, - { QLatin1String("dmat3x2"), QShaderDescription::DMat3x2 }, - { QLatin1String("dmat3x4"), QShaderDescription::DMat3x4 }, - { QLatin1String("dmat4x2"), QShaderDescription::DMat4x2 }, - { QLatin1String("dmat4x3"), QShaderDescription::DMat4x3 }, - - { QLatin1String("image1D"), QShaderDescription::Image1D }, - { QLatin1String("image2D"), QShaderDescription::Image2D }, - { QLatin1String("image2DMS"), QShaderDescription::Image2DMS }, - { QLatin1String("image3D"), QShaderDescription::Image3D }, - { QLatin1String("imageCube"), QShaderDescription::ImageCube }, - { QLatin1String("image1DArray"), QShaderDescription::Image1DArray }, - { QLatin1String("image2DArray"), QShaderDescription::Image2DArray }, - { QLatin1String("image2DMSArray"), QShaderDescription::Image2DMSArray }, - { QLatin1String("image3DArray"), QShaderDescription::Image3DArray }, - { QLatin1String("imageCubeArray"), QShaderDescription::ImageCubeArray }, - { QLatin1String("imageRect"), QShaderDescription::ImageRect }, - { QLatin1String("imageBuffer"), QShaderDescription::ImageBuffer } -}; - -static QString typeStr(const QShaderDescription::VariableType &t) + { "float", QShaderDescription::Float }, + { "vec2", QShaderDescription::Vec2 }, + { "vec3", QShaderDescription::Vec3 }, + { "vec4", QShaderDescription::Vec4 }, + { "mat2", QShaderDescription::Mat2 }, + { "mat3", QShaderDescription::Mat3 }, + { "mat4", QShaderDescription::Mat4 }, + + { "struct", QShaderDescription::Struct }, + + { "sampler1D", QShaderDescription::Sampler1D }, + { "sampler2D", QShaderDescription::Sampler2D }, + { "sampler2DMS", QShaderDescription::Sampler2DMS }, + { "sampler3D", QShaderDescription::Sampler3D }, + { "samplerCube", QShaderDescription::SamplerCube }, + { "sampler1DArray", QShaderDescription::Sampler1DArray }, + { "sampler2DArray", QShaderDescription::Sampler2DArray }, + { "sampler2DMSArray", QShaderDescription::Sampler2DMSArray }, + { "sampler3DArray", QShaderDescription::Sampler3DArray }, + { "samplerCubeArray", QShaderDescription::SamplerCubeArray }, + { "samplerRect", QShaderDescription::SamplerRect }, + { "samplerBuffer", QShaderDescription::SamplerBuffer }, + { "samplerExternalOES", QShaderDescription::SamplerExternalOES }, + { "sampler", QShaderDescription::Sampler }, + + { "mat2x3", QShaderDescription::Mat2x3 }, + { "mat2x4", QShaderDescription::Mat2x4 }, + { "mat3x2", QShaderDescription::Mat3x2 }, + { "mat3x4", QShaderDescription::Mat3x4 }, + { "mat4x2", QShaderDescription::Mat4x2 }, + { "mat4x3", QShaderDescription::Mat4x3 }, + + { "int", QShaderDescription::Int }, + { "ivec2", QShaderDescription::Int2 }, + { "ivec3", QShaderDescription::Int3 }, + { "ivec4", QShaderDescription::Int4 }, + + { "uint", QShaderDescription::Uint }, + { "uvec2", QShaderDescription::Uint2 }, + { "uvec3", QShaderDescription::Uint3 }, + { "uvec4", QShaderDescription::Uint4 }, + + { "bool", QShaderDescription::Bool }, + { "bvec2", QShaderDescription::Bool2 }, + { "bvec3", QShaderDescription::Bool3 }, + { "bvec4", QShaderDescription::Bool4 }, + + { "double", QShaderDescription::Double }, + { "dvec2", QShaderDescription::Double2 }, + { "dvec3", QShaderDescription::Double3 }, + { "dvec4", QShaderDescription::Double4 }, + { "dmat2", QShaderDescription::DMat2 }, + { "dmat3", QShaderDescription::DMat3 }, + { "dmat4", QShaderDescription::DMat4 }, + { "dmat2x3", QShaderDescription::DMat2x3 }, + { "dmat2x4", QShaderDescription::DMat2x4 }, + { "dmat3x2", QShaderDescription::DMat3x2 }, + { "dmat3x4", QShaderDescription::DMat3x4 }, + { "dmat4x2", QShaderDescription::DMat4x2 }, + { "dmat4x3", QShaderDescription::DMat4x3 }, + + { "image1D", QShaderDescription::Image1D }, + { "image2D", QShaderDescription::Image2D }, + { "image2DMS", QShaderDescription::Image2DMS }, + { "image3D", QShaderDescription::Image3D }, + { "imageCube", QShaderDescription::ImageCube }, + { "image1DArray", QShaderDescription::Image1DArray }, + { "image2DArray", QShaderDescription::Image2DArray }, + { "image2DMSArray", QShaderDescription::Image2DMSArray }, + { "image3DArray", QShaderDescription::Image3DArray }, + { "imageCubeArray", QShaderDescription::ImageCubeArray }, + { "imageRect", QShaderDescription::ImageRect }, + { "imageBuffer", QShaderDescription::ImageBuffer }, + + { "half", QShaderDescription::Half }, + { "half2", QShaderDescription::Half2 }, + { "half3", QShaderDescription::Half3 }, + { "half4", QShaderDescription::Half4 } }; + +static QLatin1StringView typeStr(QShaderDescription::VariableType t) { for (size_t i = 0; i < sizeof(typeTab) / sizeof(TypeTab); ++i) { if (typeTab[i].v == t) - return typeTab[i].k; + return QLatin1StringView(typeTab[i].k); } - return QString(); + return {}; } -static struct ImageFormatTab { - QString k; +static const struct ImageFormatTab { + const char k[15]; QShaderDescription::ImageFormat v; } imageFormatTab[] { - { QLatin1String("unknown"), QShaderDescription::ImageFormatUnknown }, - { QLatin1String("rgba32f"), QShaderDescription::ImageFormatRgba32f }, - { QLatin1String("rgba16"), QShaderDescription::ImageFormatRgba16f }, - { QLatin1String("r32f"), QShaderDescription::ImageFormatR32f }, - { QLatin1String("rgba8"), QShaderDescription::ImageFormatRgba8 }, - { QLatin1String("rgba8_snorm"), QShaderDescription::ImageFormatRgba8Snorm }, - { QLatin1String("rg32f"), QShaderDescription::ImageFormatRg32f }, - { QLatin1String("rg16f"), QShaderDescription::ImageFormatRg16f }, - { QLatin1String("r11f_g11f_b10f"), QShaderDescription::ImageFormatR11fG11fB10f }, - { QLatin1String("r16f"), QShaderDescription::ImageFormatR16f }, - { QLatin1String("rgba16"), QShaderDescription::ImageFormatRgba16 }, - { QLatin1String("rgb10_a2"), QShaderDescription::ImageFormatRgb10A2 }, - { QLatin1String("rg16"), QShaderDescription::ImageFormatRg16 }, - { QLatin1String("rg8"), QShaderDescription::ImageFormatRg8 }, - { QLatin1String("r16"), QShaderDescription::ImageFormatR16 }, - { QLatin1String("r8"), QShaderDescription::ImageFormatR8 }, - { QLatin1String("rgba16_snorm"), QShaderDescription::ImageFormatRgba16Snorm }, - { QLatin1String("rg16_snorm"), QShaderDescription::ImageFormatRg16Snorm }, - { QLatin1String("rg8_snorm"), QShaderDescription::ImageFormatRg8Snorm }, - { QLatin1String("r16_snorm"), QShaderDescription::ImageFormatR16Snorm }, - { QLatin1String("r8_snorm"), QShaderDescription::ImageFormatR8Snorm }, - { QLatin1String("rgba32i"), QShaderDescription::ImageFormatRgba32i }, - { QLatin1String("rgba16i"), QShaderDescription::ImageFormatRgba16i }, - { QLatin1String("rgba8i"), QShaderDescription::ImageFormatRgba8i }, - { QLatin1String("r32i"), QShaderDescription::ImageFormatR32i }, - { QLatin1String("rg32i"), QShaderDescription::ImageFormatRg32i }, - { QLatin1String("rg16i"), QShaderDescription::ImageFormatRg16i }, - { QLatin1String("rg8i"), QShaderDescription::ImageFormatRg8i }, - { QLatin1String("r16i"), QShaderDescription::ImageFormatR16i }, - { QLatin1String("r8i"), QShaderDescription::ImageFormatR8i }, - { QLatin1String("rgba32ui"), QShaderDescription::ImageFormatRgba32ui }, - { QLatin1String("rgba16ui"), QShaderDescription::ImageFormatRgba16ui }, - { QLatin1String("rgba8ui"), QShaderDescription::ImageFormatRgba8ui }, - { QLatin1String("r32ui"), QShaderDescription::ImageFormatR32ui }, - { QLatin1String("rgb10_a2ui"), QShaderDescription::ImageFormatRgb10a2ui }, - { QLatin1String("rg32ui"), QShaderDescription::ImageFormatRg32ui }, - { QLatin1String("rg16ui"), QShaderDescription::ImageFormatRg16ui }, - { QLatin1String("rg8ui"), QShaderDescription::ImageFormatRg8ui }, - { QLatin1String("r16ui"), QShaderDescription::ImageFormatR16ui }, - { QLatin1String("r8ui"), QShaderDescription::ImageFormatR8ui } + { "unknown", QShaderDescription::ImageFormatUnknown }, + { "rgba32f", QShaderDescription::ImageFormatRgba32f }, + { "rgba16", QShaderDescription::ImageFormatRgba16f }, + { "r32f", QShaderDescription::ImageFormatR32f }, + { "rgba8", QShaderDescription::ImageFormatRgba8 }, + { "rgba8_snorm", QShaderDescription::ImageFormatRgba8Snorm }, + { "rg32f", QShaderDescription::ImageFormatRg32f }, + { "rg16f", QShaderDescription::ImageFormatRg16f }, + { "r11f_g11f_b10f", QShaderDescription::ImageFormatR11fG11fB10f }, + { "r16f", QShaderDescription::ImageFormatR16f }, + { "rgba16", QShaderDescription::ImageFormatRgba16 }, + { "rgb10_a2", QShaderDescription::ImageFormatRgb10A2 }, + { "rg16", QShaderDescription::ImageFormatRg16 }, + { "rg8", QShaderDescription::ImageFormatRg8 }, + { "r16", QShaderDescription::ImageFormatR16 }, + { "r8", QShaderDescription::ImageFormatR8 }, + { "rgba16_snorm", QShaderDescription::ImageFormatRgba16Snorm }, + { "rg16_snorm", QShaderDescription::ImageFormatRg16Snorm }, + { "rg8_snorm", QShaderDescription::ImageFormatRg8Snorm }, + { "r16_snorm", QShaderDescription::ImageFormatR16Snorm }, + { "r8_snorm", QShaderDescription::ImageFormatR8Snorm }, + { "rgba32i", QShaderDescription::ImageFormatRgba32i }, + { "rgba16i", QShaderDescription::ImageFormatRgba16i }, + { "rgba8i", QShaderDescription::ImageFormatRgba8i }, + { "r32i", QShaderDescription::ImageFormatR32i }, + { "rg32i", QShaderDescription::ImageFormatRg32i }, + { "rg16i", QShaderDescription::ImageFormatRg16i }, + { "rg8i", QShaderDescription::ImageFormatRg8i }, + { "r16i", QShaderDescription::ImageFormatR16i }, + { "r8i", QShaderDescription::ImageFormatR8i }, + { "rgba32ui", QShaderDescription::ImageFormatRgba32ui }, + { "rgba16ui", QShaderDescription::ImageFormatRgba16ui }, + { "rgba8ui", QShaderDescription::ImageFormatRgba8ui }, + { "r32ui", QShaderDescription::ImageFormatR32ui }, + { "rgb10_a2ui", QShaderDescription::ImageFormatRgb10a2ui }, + { "rg32ui", QShaderDescription::ImageFormatRg32ui }, + { "rg16ui", QShaderDescription::ImageFormatRg16ui }, + { "rg8ui", QShaderDescription::ImageFormatRg8ui }, + { "r16ui", QShaderDescription::ImageFormatR16ui }, + { "r8ui", QShaderDescription::ImageFormatR8ui } }; -static QString imageFormatStr(const QShaderDescription::ImageFormat &f) +static QLatin1StringView imageFormatStr(QShaderDescription::ImageFormat f) { for (size_t i = 0; i < sizeof(imageFormatTab) / sizeof(ImageFormatTab); ++i) { if (imageFormatTab[i].v == f) - return imageFormatTab[i].k; + return QLatin1StringView(imageFormatTab[i].k); } - return QString(); + return {}; +} + +static const struct BuiltinTypeTab { + const char k[21]; + QShaderDescription::BuiltinType v; +} builtinTypeTab[] = { + { "Position", QShaderDescription::PositionBuiltin }, + { "PointSize", QShaderDescription::PointSizeBuiltin }, + { "ClipDistance", QShaderDescription::ClipDistanceBuiltin }, + { "CullDistance", QShaderDescription::CullDistanceBuiltin }, + { "VertexId", QShaderDescription::VertexIdBuiltin }, + { "InstanceId", QShaderDescription::InstanceIdBuiltin }, + { "PrimitiveId", QShaderDescription::PrimitiveIdBuiltin }, + { "InvocationId", QShaderDescription::InvocationIdBuiltin }, + { "Layer", QShaderDescription::LayerBuiltin }, + { "ViewportIndex", QShaderDescription::ViewportIndexBuiltin }, + { "TessLevelOuter", QShaderDescription::TessLevelOuterBuiltin }, + { "TessLevelInner", QShaderDescription::TessLevelInnerBuiltin }, + { "TessCoord", QShaderDescription::TessCoordBuiltin }, + { "PatchVertices", QShaderDescription::PatchVerticesBuiltin }, + { "FragCoord", QShaderDescription::FragCoordBuiltin }, + { "PointCoord", QShaderDescription::PointCoordBuiltin }, + { "FrontFacing", QShaderDescription::FrontFacingBuiltin }, + { "SampleId", QShaderDescription::SampleIdBuiltin }, + { "SamplePosition", QShaderDescription::SamplePositionBuiltin }, + { "SampleMask", QShaderDescription::SampleMaskBuiltin }, + { "FragDepth", QShaderDescription::FragDepthBuiltin }, + { "NumWorkGroups", QShaderDescription::NumWorkGroupsBuiltin }, + { "WorkgroupSize", QShaderDescription::WorkgroupSizeBuiltin }, + { "WorkgroupId", QShaderDescription::WorkgroupIdBuiltin }, + { "LocalInvocationId", QShaderDescription::LocalInvocationIdBuiltin }, + { "GlobalInvocationId", QShaderDescription::GlobalInvocationIdBuiltin }, + { "LocalInvocationIndex", QShaderDescription::LocalInvocationIndexBuiltin }, + { "VertexIndex", QShaderDescription::VertexIndexBuiltin }, + { "InstanceIndex", QShaderDescription::InstanceIndexBuiltin } +}; + +static QLatin1StringView builtinTypeStr(QShaderDescription::BuiltinType t) +{ + for (size_t i = 0; i < sizeof(builtinTypeTab) / sizeof(BuiltinTypeTab); ++i) { + if (builtinTypeTab[i].v == t) + return QLatin1StringView(builtinTypeTab[i].k); + } + return {}; +} + +static const struct TessellationModeTab { + const char k[10]; + QShaderDescription::TessellationMode v; +} tessellationModeTab[] { + { "unknown", QShaderDescription::UnknownTessellationMode }, + { "triangles", QShaderDescription::TrianglesTessellationMode }, + { "quad", QShaderDescription::QuadTessellationMode }, + { "isoline", QShaderDescription::IsolineTessellationMode } +}; + +static QLatin1StringView tessModeStr(QShaderDescription::TessellationMode mode) +{ + for (size_t i = 0; i < sizeof(tessellationModeTab) / sizeof(TessellationModeTab); ++i) { + if (tessellationModeTab[i].v == mode) + return QLatin1StringView(tessellationModeTab[i].k); + } + return {}; +} + +static const struct TessellationWindingOrderTab { + const char k[8]; + QShaderDescription::TessellationWindingOrder v; +} tessellationWindingOrderTab[] { + { "unknown", QShaderDescription::UnknownTessellationWindingOrder }, + { "cw", QShaderDescription::CwTessellationWindingOrder }, + { "ccw", QShaderDescription::CcwTessellationWindingOrder } +}; + +static QLatin1StringView tessWindStr(QShaderDescription::TessellationWindingOrder w) +{ + for (size_t i = 0; i < sizeof(tessellationWindingOrderTab) / sizeof(TessellationWindingOrderTab); ++i) { + if (tessellationWindingOrderTab[i].v == w) + return QLatin1StringView(tessellationWindingOrderTab[i].k); + } + return {}; +} + +static const struct TessellationPartitioningTab { + const char k[24]; + QShaderDescription::TessellationPartitioning v; +} tessellationPartitioningTab[] { + { "unknown", QShaderDescription::UnknownTessellationPartitioning }, + { "equal_spacing", QShaderDescription::EqualTessellationPartitioning }, + { "fractional_even_spacing", QShaderDescription::FractionalEvenTessellationPartitioning }, + { "fractional_odd_spacing", QShaderDescription::FractionalOddTessellationPartitioning } +}; + +static QLatin1StringView tessPartStr(QShaderDescription::TessellationPartitioning p) +{ + for (size_t i = 0; i < sizeof(tessellationPartitioningTab) / sizeof(TessellationPartitioningTab); ++i) { + if (tessellationPartitioningTab[i].v == p) + return QLatin1StringView(tessellationPartitioningTab[i].k); + } + return {}; } #ifndef QT_NO_DEBUG_STREAM @@ -708,7 +1204,11 @@ QDebug operator<<(QDebug dbg, const QShaderDescription &sd) << " pcBlocks " << d->pushConstantBlocks << " storageBlocks " << d->storageBlocks << " combinedSamplers " << d->combinedImageSamplers - << " images " << d->storageImages + << " storageImages " << d->storageImages + << " separateImages " << d->separateImages + << " separateSamplers " << d->separateSamplers + << " inBuiltins " << d->inBuiltins + << " outBuiltins " << d->outBuiltins << ')'; } else { dbg.nospace() << "QShaderDescription(null)"; @@ -721,6 +1221,8 @@ QDebug operator<<(QDebug dbg, const QShaderDescription::InOutVariable &var) { QDebugStateSaver saver(dbg); dbg.nospace() << "InOutVariable(" << typeStr(var.type) << ' ' << var.name; + if (var.perPatch) + dbg.nospace() << " per-patch"; if (var.location >= 0) dbg.nospace() << " location=" << var.location; if (var.binding >= 0) @@ -733,6 +1235,8 @@ QDebug operator<<(QDebug dbg, const QShaderDescription::InOutVariable &var) dbg.nospace() << " imageFlags=" << var.imageFlags; if (!var.arrayDims.isEmpty()) dbg.nospace() << " array=" << var.arrayDims; + if (!var.structMembers.isEmpty()) + dbg.nospace() << " structMembers=" << var.structMembers; dbg.nospace() << ')'; return dbg; } @@ -740,8 +1244,10 @@ QDebug operator<<(QDebug dbg, const QShaderDescription::InOutVariable &var) QDebug operator<<(QDebug dbg, const QShaderDescription::BlockVariable &var) { QDebugStateSaver saver(dbg); - dbg.nospace() << "BlockVariable(" << typeStr(var.type) << ' ' << var.name - << " offset=" << var.offset << " size=" << var.size; + dbg.nospace() << "BlockVariable(" << typeStr(var.type) << ' ' << var.name; + if (var.offset != -1) + dbg.nospace() << " offset=" << var.offset; + dbg.nospace() << " size=" << var.size; if (!var.arrayDims.isEmpty()) dbg.nospace() << " array=" << var.arrayDims; if (var.arrayStride) @@ -786,111 +1292,169 @@ QDebug operator<<(QDebug dbg, const QShaderDescription::StorageBlock &blk) dbg.nospace() << " binding=" << blk.binding; if (blk.descriptorSet >= 0) dbg.nospace() << " set=" << blk.descriptorSet; + if (blk.runtimeArrayStride) + dbg.nospace() << " runtimeArrayStride=" << blk.runtimeArrayStride; + if (blk.qualifierFlags) + dbg.nospace() << " qualifierFlags=" << blk.qualifierFlags; dbg.nospace() << ' ' << blk.members << ')'; return dbg; } + +QDebug operator<<(QDebug dbg, const QShaderDescription::BuiltinVariable &builtin) +{ + QDebugStateSaver saver(dbg); + dbg.nospace() << "BuiltinVariable(type=" << builtinTypeStr(builtin.type); + dbg.nospace() << " varType=" << typeStr(builtin.varType); + if (!builtin.arrayDims.isEmpty()) + dbg.nospace() << " array=" << builtin.arrayDims; + dbg.nospace() << ")"; + return dbg; +} #endif -static const QString nameKey = QLatin1String("name"); -static const QString typeKey = QLatin1String("type"); -static const QString locationKey = QLatin1String("location"); -static const QString bindingKey = QLatin1String("binding"); -static const QString setKey = QLatin1String("set"); -static const QString imageFormatKey = QLatin1String("imageFormat"); -static const QString imageFlagsKey = QLatin1String("imageFlags"); -static const QString offsetKey = QLatin1String("offset"); -static const QString arrayDimsKey = QLatin1String("arrayDims"); -static const QString arrayStrideKey = QLatin1String("arrayStride"); -static const QString matrixStrideKey = QLatin1String("matrixStride"); -static const QString matrixRowMajorKey = QLatin1String("matrixRowMajor"); -static const QString structMembersKey = QLatin1String("structMembers"); -static const QString membersKey = QLatin1String("members"); -static const QString inputsKey = QLatin1String("inputs"); -static const QString outputsKey = QLatin1String("outputs"); -static const QString uniformBlocksKey = QLatin1String("uniformBlocks"); -static const QString blockNameKey = QLatin1String("blockName"); -static const QString structNameKey = QLatin1String("structName"); -static const QString instanceNameKey = QLatin1String("instanceName"); -static const QString sizeKey = QLatin1String("size"); -static const QString knownSizeKey = QLatin1String("knownSize"); -static const QString pushConstantBlocksKey = QLatin1String("pushConstantBlocks"); -static const QString storageBlocksKey = QLatin1String("storageBlocks"); -static const QString combinedImageSamplersKey = QLatin1String("combinedImageSamplers"); -static const QString storageImagesKey = QLatin1String("storageImages"); -static const QString localSizeKey = QLatin1String("localSize"); +#define JSON_KEY(key) static constexpr QLatin1StringView key ## Key() noexcept { return QLatin1StringView( #key ); } +JSON_KEY(name) +JSON_KEY(type) +JSON_KEY(location) +JSON_KEY(binding) +JSON_KEY(set) +JSON_KEY(perPatch) +JSON_KEY(imageFormat) +JSON_KEY(imageFlags) +JSON_KEY(offset) +JSON_KEY(arrayDims) +JSON_KEY(arrayStride) +JSON_KEY(matrixStride) +JSON_KEY(matrixRowMajor) +JSON_KEY(structMembers) +JSON_KEY(members) +JSON_KEY(inputs) +JSON_KEY(outputs) +JSON_KEY(uniformBlocks) +JSON_KEY(blockName) +JSON_KEY(structName) +JSON_KEY(instanceName) +JSON_KEY(size) +JSON_KEY(knownSize) +JSON_KEY(pushConstantBlocks) +JSON_KEY(storageBlocks) +JSON_KEY(combinedImageSamplers) +JSON_KEY(storageImages) +JSON_KEY(inBuiltins) +JSON_KEY(outBuiltins) +JSON_KEY(computeLocalSize) +JSON_KEY(tessellationOutputVertexCount) +JSON_KEY(tessellationMode) +JSON_KEY(tessellationWindingOrder) +JSON_KEY(tessellationPartitioning) +JSON_KEY(separateImages) +JSON_KEY(separateSamplers) +JSON_KEY(runtimeArrayStride) +JSON_KEY(qualifierFlags) +#undef JSON_KEY static void addDeco(QJsonObject *obj, const QShaderDescription::InOutVariable &v) { if (v.location >= 0) - (*obj)[locationKey] = v.location; + (*obj)[locationKey()] = v.location; if (v.binding >= 0) - (*obj)[bindingKey] = v.binding; + (*obj)[bindingKey()] = v.binding; if (v.descriptorSet >= 0) - (*obj)[setKey] = v.descriptorSet; + (*obj)[setKey()] = v.descriptorSet; + if (v.perPatch) + (*obj)[perPatchKey()] = v.perPatch; if (v.imageFormat != QShaderDescription::ImageFormatUnknown) - (*obj)[imageFormatKey] = imageFormatStr(v.imageFormat); + (*obj)[imageFormatKey()] = imageFormatStr(v.imageFormat); if (v.imageFlags) - (*obj)[imageFlagsKey] = int(v.imageFlags); + (*obj)[imageFlagsKey()] = int(v.imageFlags); if (!v.arrayDims.isEmpty()) { QJsonArray dimArr; for (int dim : v.arrayDims) dimArr.append(dim); - (*obj)[arrayDimsKey] = dimArr; + (*obj)[arrayDimsKey()] = dimArr; } } -static void serializeDecorations(QDataStream *stream, const QShaderDescription::InOutVariable &v) +static void serializeDecorations(QDataStream *stream, const QShaderDescription::InOutVariable &v, int version) { (*stream) << v.location; (*stream) << v.binding; (*stream) << v.descriptorSet; (*stream) << int(v.imageFormat); (*stream) << int(v.imageFlags); - (*stream) << int(v.arrayDims.count()); + (*stream) << int(v.arrayDims.size()); for (int dim : v.arrayDims) (*stream) << dim; + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) + (*stream) << quint8(v.perPatch); } -static QJsonObject inOutObject(const QShaderDescription::InOutVariable &v) +static void serializeBuiltinVar(QDataStream *stream, const QShaderDescription::BuiltinVariable &v, int version) { - QJsonObject obj; - obj[nameKey] = QString::fromUtf8(v.name); - obj[typeKey] = typeStr(v.type); - addDeco(&obj, v); - return obj; -} - -static void serializeInOutVar(QDataStream *stream, const QShaderDescription::InOutVariable &v) -{ - (*stream) << QString::fromUtf8(v.name); (*stream) << int(v.type); - serializeDecorations(stream, v); + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS) { + (*stream) << int(v.varType); + (*stream) << int(v.arrayDims.size()); + for (int dim : v.arrayDims) + (*stream) << dim; + } } static QJsonObject blockMemberObject(const QShaderDescription::BlockVariable &v) { QJsonObject obj; - obj[nameKey] = QString::fromUtf8(v.name); - obj[typeKey] = typeStr(v.type); - obj[offsetKey] = v.offset; - obj[sizeKey] = v.size; + obj[nameKey()] = QString::fromUtf8(v.name); + obj[typeKey()] = typeStr(v.type); + if (v.offset != -1) + obj[offsetKey()] = v.offset; + obj[sizeKey()] = v.size; if (!v.arrayDims.isEmpty()) { QJsonArray dimArr; for (int dim : v.arrayDims) dimArr.append(dim); - obj[arrayDimsKey] = dimArr; + obj[arrayDimsKey()] = dimArr; } if (v.arrayStride) - obj[arrayStrideKey] = v.arrayStride; + obj[arrayStrideKey()] = v.arrayStride; if (v.matrixStride) - obj[matrixStrideKey] = v.matrixStride; + obj[matrixStrideKey()] = v.matrixStride; if (v.matrixIsRowMajor) - obj[matrixRowMajorKey] = true; + obj[matrixRowMajorKey()] = true; + if (!v.structMembers.isEmpty()) { + QJsonArray arr; + for (const QShaderDescription::BlockVariable &sv : v.structMembers) + arr.append(blockMemberObject(sv)); + obj[structMembersKey()] = arr; + } + return obj; +} + +static QJsonObject inOutObject(const QShaderDescription::InOutVariable &v) +{ + QJsonObject obj; + obj[nameKey()] = QString::fromUtf8(v.name); + obj[typeKey()] = typeStr(v.type); + addDeco(&obj, v); if (!v.structMembers.isEmpty()) { QJsonArray arr; for (const QShaderDescription::BlockVariable &sv : v.structMembers) arr.append(blockMemberObject(sv)); - obj[structMembersKey] = arr; + obj[structMembersKey()] = arr; + } + return obj; +} + +static QJsonObject builtinObject(const QShaderDescription::BuiltinVariable &v) +{ + QJsonObject obj; + + obj[nameKey()] = builtinTypeStr(v.type); + obj[typeKey()] = typeStr(v.varType); + if (!v.arrayDims.isEmpty()) { + QJsonArray dimArr; + for (int dim : v.arrayDims) + dimArr.append(dim); + obj[arrayDimsKey()] = dimArr; } return obj; } @@ -901,174 +1465,272 @@ static void serializeBlockMemberVar(QDataStream *stream, const QShaderDescriptio (*stream) << int(v.type); (*stream) << v.offset; (*stream) << v.size; - (*stream) << int(v.arrayDims.count()); + (*stream) << int(v.arrayDims.size()); for (int dim : v.arrayDims) (*stream) << dim; (*stream) << v.arrayStride; (*stream) << v.matrixStride; (*stream) << v.matrixIsRowMajor; - (*stream) << int(v.structMembers.count()); + (*stream) << int(v.structMembers.size()); for (const QShaderDescription::BlockVariable &sv : v.structMembers) serializeBlockMemberVar(stream, sv); } +static void serializeInOutVar(QDataStream *stream, const QShaderDescription::InOutVariable &v, + int version) +{ + (*stream) << QString::fromUtf8(v.name); + (*stream) << int(v.type); + serializeDecorations(stream, v, version); + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS) { + (*stream) << int(v.structMembers.size()); + for (const QShaderDescription::BlockVariable &sv : v.structMembers) + serializeBlockMemberVar(stream, sv); + } +} + QJsonDocument QShaderDescriptionPrivate::makeDoc() { QJsonObject root; QJsonArray jinputs; - for (const QShaderDescription::InOutVariable &v : qAsConst(inVars)) + for (const QShaderDescription::InOutVariable &v : std::as_const(inVars)) jinputs.append(inOutObject(v)); if (!jinputs.isEmpty()) - root[inputsKey] = jinputs; + root[inputsKey()] = jinputs; QJsonArray joutputs; - for (const QShaderDescription::InOutVariable &v : qAsConst(outVars)) + for (const QShaderDescription::InOutVariable &v : std::as_const(outVars)) joutputs.append(inOutObject(v)); if (!joutputs.isEmpty()) - root[outputsKey] = joutputs; + root[outputsKey()] = joutputs; QJsonArray juniformBlocks; for (const QShaderDescription::UniformBlock &b : uniformBlocks) { QJsonObject juniformBlock; - juniformBlock[blockNameKey] = QString::fromUtf8(b.blockName); - juniformBlock[structNameKey] = QString::fromUtf8(b.structName); - juniformBlock[sizeKey] = b.size; + juniformBlock[blockNameKey()] = QString::fromUtf8(b.blockName); + juniformBlock[structNameKey()] = QString::fromUtf8(b.structName); + juniformBlock[sizeKey()] = b.size; if (b.binding >= 0) - juniformBlock[bindingKey] = b.binding; + juniformBlock[bindingKey()] = b.binding; if (b.descriptorSet >= 0) - juniformBlock[setKey] = b.descriptorSet; + juniformBlock[setKey()] = b.descriptorSet; QJsonArray members; for (const QShaderDescription::BlockVariable &v : b.members) members.append(blockMemberObject(v)); - juniformBlock[membersKey] = members; + juniformBlock[membersKey()] = members; juniformBlocks.append(juniformBlock); } if (!juniformBlocks.isEmpty()) - root[uniformBlocksKey] = juniformBlocks; + root[uniformBlocksKey()] = juniformBlocks; QJsonArray jpushConstantBlocks; for (const QShaderDescription::PushConstantBlock &b : pushConstantBlocks) { QJsonObject jpushConstantBlock; - jpushConstantBlock[nameKey] = QString::fromUtf8(b.name); - jpushConstantBlock[sizeKey] = b.size; + jpushConstantBlock[nameKey()] = QString::fromUtf8(b.name); + jpushConstantBlock[sizeKey()] = b.size; QJsonArray members; for (const QShaderDescription::BlockVariable &v : b.members) members.append(blockMemberObject(v)); - jpushConstantBlock[membersKey] = members; + jpushConstantBlock[membersKey()] = members; jpushConstantBlocks.append(jpushConstantBlock); } if (!jpushConstantBlocks.isEmpty()) - root[pushConstantBlocksKey] = jpushConstantBlocks; + root[pushConstantBlocksKey()] = jpushConstantBlocks; QJsonArray jstorageBlocks; for (const QShaderDescription::StorageBlock &b : storageBlocks) { QJsonObject jstorageBlock; - jstorageBlock[blockNameKey] = QString::fromUtf8(b.blockName); - jstorageBlock[instanceNameKey] = QString::fromUtf8(b.instanceName); - jstorageBlock[knownSizeKey] = b.knownSize; + jstorageBlock[blockNameKey()] = QString::fromUtf8(b.blockName); + jstorageBlock[instanceNameKey()] = QString::fromUtf8(b.instanceName); + jstorageBlock[knownSizeKey()] = b.knownSize; if (b.binding >= 0) - jstorageBlock[bindingKey] = b.binding; + jstorageBlock[bindingKey()] = b.binding; if (b.descriptorSet >= 0) - jstorageBlock[setKey] = b.descriptorSet; + jstorageBlock[setKey()] = b.descriptorSet; + if (b.runtimeArrayStride) + jstorageBlock[runtimeArrayStrideKey()] = b.runtimeArrayStride; + if (b.qualifierFlags) + jstorageBlock[qualifierFlagsKey()] = int(b.qualifierFlags); QJsonArray members; for (const QShaderDescription::BlockVariable &v : b.members) members.append(blockMemberObject(v)); - jstorageBlock[membersKey] = members; + jstorageBlock[membersKey()] = members; jstorageBlocks.append(jstorageBlock); } if (!jstorageBlocks.isEmpty()) - root[storageBlocksKey] = jstorageBlocks; + root[storageBlocksKey()] = jstorageBlocks; QJsonArray jcombinedSamplers; - for (const QShaderDescription::InOutVariable &v : qAsConst(combinedImageSamplers)) { + for (const QShaderDescription::InOutVariable &v : std::as_const(combinedImageSamplers)) { QJsonObject sampler; - sampler[nameKey] = QString::fromUtf8(v.name); - sampler[typeKey] = typeStr(v.type); + sampler[nameKey()] = QString::fromUtf8(v.name); + sampler[typeKey()] = typeStr(v.type); addDeco(&sampler, v); jcombinedSamplers.append(sampler); } if (!jcombinedSamplers.isEmpty()) - root[combinedImageSamplersKey] = jcombinedSamplers; + root[combinedImageSamplersKey()] = jcombinedSamplers; QJsonArray jstorageImages; - for (const QShaderDescription::InOutVariable &v : qAsConst(storageImages)) { + for (const QShaderDescription::InOutVariable &v : std::as_const(storageImages)) { QJsonObject image; - image[nameKey] = QString::fromUtf8(v.name); - image[typeKey] = typeStr(v.type); + image[nameKey()] = QString::fromUtf8(v.name); + image[typeKey()] = typeStr(v.type); addDeco(&image, v); jstorageImages.append(image); } if (!jstorageImages.isEmpty()) - root[storageImagesKey] = jstorageImages; + root[storageImagesKey()] = jstorageImages; + + QJsonArray jinBuiltins; + for (const QShaderDescription::BuiltinVariable &v : std::as_const(inBuiltins)) + jinBuiltins.append(builtinObject(v)); + if (!jinBuiltins.isEmpty()) + root[inBuiltinsKey()] = jinBuiltins; + + QJsonArray joutBuiltins; + for (const QShaderDescription::BuiltinVariable &v : std::as_const(outBuiltins)) + joutBuiltins.append(builtinObject(v)); + if (!joutBuiltins.isEmpty()) + root[outBuiltinsKey()] = joutBuiltins; + + if (localSize[0] || localSize[1] || localSize[2]) { + QJsonArray jlocalSize; + for (size_t i = 0; i < 3; ++i) + jlocalSize.append(QJsonValue(int(localSize[i]))); + root[computeLocalSizeKey()] = jlocalSize; + } + + if (tessOutVertCount) + root[tessellationOutputVertexCountKey()] = int(tessOutVertCount); + + if (tessMode != QShaderDescription::UnknownTessellationMode) + root[tessellationModeKey()] = tessModeStr(tessMode); + + if (tessWind != QShaderDescription::UnknownTessellationWindingOrder) + root[tessellationWindingOrderKey()] = tessWindStr(tessWind); - QJsonArray jlocalSize; - for (int i = 0; i < 3; ++i) - jlocalSize.append(QJsonValue(int(localSize[i]))); - root[localSizeKey] = jlocalSize; + if (tessPart != QShaderDescription::UnknownTessellationPartitioning) + root[tessellationPartitioningKey()] = tessPartStr(tessPart); + + QJsonArray jseparateImages; + for (const QShaderDescription::InOutVariable &v : std::as_const(separateImages)) { + QJsonObject image; + image[nameKey()] = QString::fromUtf8(v.name); + image[typeKey()] = typeStr(v.type); + addDeco(&image, v); + jseparateImages.append(image); + } + if (!jseparateImages.isEmpty()) + root[separateImagesKey()] = jseparateImages; + + QJsonArray jseparateSamplers; + for (const QShaderDescription::InOutVariable &v : std::as_const(separateSamplers)) { + QJsonObject sampler; + sampler[nameKey()] = QString::fromUtf8(v.name); + sampler[typeKey()] = typeStr(v.type); + addDeco(&sampler, v); + jseparateSamplers.append(sampler); + } + if (!jseparateSamplers.isEmpty()) + root[separateSamplersKey()] = jseparateSamplers; return QJsonDocument(root); } -void QShaderDescriptionPrivate::writeToStream(QDataStream *stream) +void QShaderDescriptionPrivate::writeToStream(QDataStream *stream, int version) { - (*stream) << int(inVars.count()); - for (const QShaderDescription::InOutVariable &v : qAsConst(inVars)) - serializeInOutVar(stream, v); + (*stream) << int(inVars.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(inVars)) + serializeInOutVar(stream, v, version); - (*stream) << int(outVars.count()); - for (const QShaderDescription::InOutVariable &v : qAsConst(outVars)) - serializeInOutVar(stream, v); + (*stream) << int(outVars.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(outVars)) + serializeInOutVar(stream, v, version); - (*stream) << int(uniformBlocks.count()); + (*stream) << int(uniformBlocks.size()); for (const QShaderDescription::UniformBlock &b : uniformBlocks) { (*stream) << QString::fromUtf8(b.blockName); (*stream) << QString::fromUtf8(b.structName); (*stream) << b.size; (*stream) << b.binding; (*stream) << b.descriptorSet; - (*stream) << int(b.members.count()); + (*stream) << int(b.members.size()); for (const QShaderDescription::BlockVariable &v : b.members) serializeBlockMemberVar(stream, v); } - (*stream) << int(pushConstantBlocks.count()); + (*stream) << int(pushConstantBlocks.size()); for (const QShaderDescription::PushConstantBlock &b : pushConstantBlocks) { (*stream) << QString::fromUtf8(b.name); (*stream) << b.size; - (*stream) << int(b.members.count()); + (*stream) << int(b.members.size()); for (const QShaderDescription::BlockVariable &v : b.members) serializeBlockMemberVar(stream, v); } - (*stream) << int(storageBlocks.count()); + (*stream) << int(storageBlocks.size()); for (const QShaderDescription::StorageBlock &b : storageBlocks) { (*stream) << QString::fromUtf8(b.blockName); (*stream) << QString::fromUtf8(b.instanceName); (*stream) << b.knownSize; (*stream) << b.binding; (*stream) << b.descriptorSet; - (*stream) << int(b.members.count()); + (*stream) << int(b.members.size()); for (const QShaderDescription::BlockVariable &v : b.members) serializeBlockMemberVar(stream, v); + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO) { + (*stream) << b.runtimeArrayStride; + (*stream) << b.qualifierFlags; + } } - (*stream) << int(combinedImageSamplers.count()); - for (const QShaderDescription::InOutVariable &v : qAsConst(combinedImageSamplers)) { + (*stream) << int(combinedImageSamplers.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(combinedImageSamplers)) { (*stream) << QString::fromUtf8(v.name); (*stream) << int(v.type); - serializeDecorations(stream, v); + serializeDecorations(stream, v, version); } - (*stream) << int(storageImages.count()); - for (const QShaderDescription::InOutVariable &v : qAsConst(storageImages)) { + (*stream) << int(storageImages.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(storageImages)) { (*stream) << QString::fromUtf8(v.name); (*stream) << int(v.type); - serializeDecorations(stream, v); + serializeDecorations(stream, v, version); } for (size_t i = 0; i < 3; ++i) - (*stream) << localSize[i]; + (*stream) << quint32(localSize[i]); + + (*stream) << int(separateImages.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(separateImages)) { + (*stream) << QString::fromUtf8(v.name); + (*stream) << int(v.type); + serializeDecorations(stream, v, version); + } + + (*stream) << int(separateSamplers.size()); + for (const QShaderDescription::InOutVariable &v : std::as_const(separateSamplers)) { + (*stream) << QString::fromUtf8(v.name); + (*stream) << int(v.type); + serializeDecorations(stream, v, version); + } + + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) { + (*stream) << quint32(tessOutVertCount); + (*stream) << quint32(tessMode); + (*stream) << quint32(tessWind); + (*stream) << quint32(tessPart); + + (*stream) << int(inBuiltins.size()); + for (const QShaderDescription::BuiltinVariable &v : std::as_const(inBuiltins)) + serializeBuiltinVar(stream, v, version); + + (*stream) << int(outBuiltins.size()); + for (const QShaderDescription::BuiltinVariable &v : std::as_const(outBuiltins)) + serializeBuiltinVar(stream, v, version); + } } static void deserializeDecorations(QDataStream *stream, int version, QShaderDescription::InOutVariable *v) @@ -1088,18 +1750,29 @@ static void deserializeDecorations(QDataStream *stream, int version, QShaderDesc for (int i = 0; i < f; ++i) (*stream) >> v->arrayDims[i]; } + + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) { + quint8 b; + (*stream) >> b; + v->perPatch = b; + } } -static QShaderDescription::InOutVariable deserializeInOutVar(QDataStream *stream, int version) +static QShaderDescription::BuiltinVariable deserializeBuiltinVar(QDataStream *stream, int version) { - QShaderDescription::InOutVariable var; - QString tmp; - (*stream) >> tmp; - var.name = tmp.toUtf8(); + QShaderDescription::BuiltinVariable var; int t; (*stream) >> t; - var.type = QShaderDescription::VariableType(t); - deserializeDecorations(stream, version, &var); + var.type = QShaderDescription::BuiltinType(t); + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS) { + (*stream) >> t; + var.varType = QShaderDescription::VariableType(t); + int count; + (*stream) >> count; + var.arrayDims.resize(count); + for (int i = 0; i < count; ++i) + (*stream) >> var.arrayDims[i]; + } return var; } @@ -1129,6 +1802,26 @@ static QShaderDescription::BlockVariable deserializeBlockMemberVar(QDataStream * return var; } +static QShaderDescription::InOutVariable deserializeInOutVar(QDataStream *stream, int version) +{ + QShaderDescription::InOutVariable var; + QString tmp; + (*stream) >> tmp; + var.name = tmp.toUtf8(); + int t; + (*stream) >> t; + var.type = QShaderDescription::VariableType(t); + deserializeDecorations(stream, version, &var); + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS) { + int count; + (*stream) >> count; + var.structMembers.resize(count); + for (int i = 0; i < count; ++i) + var.structMembers[i] = deserializeBlockMemberVar(stream, version); + } + return var; +} + void QShaderDescriptionPrivate::loadFromStream(QDataStream *stream, int version) { Q_ASSERT(ref.loadRelaxed() == 1); // must be detached @@ -1192,6 +1885,11 @@ void QShaderDescriptionPrivate::loadFromStream(QDataStream *stream, int version) storageBlocks[i].members.resize(memberCount); for (int memberIdx = 0; memberIdx < memberCount; ++memberIdx) storageBlocks[i].members[memberIdx] = deserializeBlockMemberVar(stream, version); + + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO) { + (*stream) >> storageBlocks[i].runtimeArrayStride; + (*stream) >> storageBlocks[i].qualifierFlags; + } } (*stream) >> count; @@ -1218,8 +1916,59 @@ void QShaderDescriptionPrivate::loadFromStream(QDataStream *stream, int version) deserializeDecorations(stream, version, &storageImages[i]); } - for (size_t i = 0; i < 3; ++i) - (*stream) >> localSize[i]; + for (size_t i = 0; i < 3; ++i) { + quint32 v; + (*stream) >> v; + localSize[i] = v; + } + + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS) { + (*stream) >> count; + separateImages.resize(count); + for (int i = 0; i < count; ++i) { + QString tmp; + (*stream) >> tmp; + separateImages[i].name = tmp.toUtf8(); + int t; + (*stream) >> t; + separateImages[i].type = QShaderDescription::VariableType(t); + deserializeDecorations(stream, version, &separateImages[i]); + } + + (*stream) >> count; + separateSamplers.resize(count); + for (int i = 0; i < count; ++i) { + QString tmp; + (*stream) >> tmp; + separateSamplers[i].name = tmp.toUtf8(); + int t; + (*stream) >> t; + separateSamplers[i].type = QShaderDescription::VariableType(t); + deserializeDecorations(stream, version, &separateSamplers[i]); + } + } + + if (version > QShaderPrivate::QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO) { + quint32 v; + (*stream) >> v; + tessOutVertCount = v; + (*stream) >> v; + tessMode = QShaderDescription::TessellationMode(v); + (*stream) >> v; + tessWind = QShaderDescription::TessellationWindingOrder(v); + (*stream) >> v; + tessPart = QShaderDescription::TessellationPartitioning(v); + + (*stream) >> count; + inBuiltins.resize(count); + for (int i = 0; i < count; ++i) + inBuiltins[i] = deserializeBuiltinVar(stream, version); + + (*stream) >> count; + outBuiltins.resize(count); + for (int i = 0; i < count; ++i) + outBuiltins[i] = deserializeBuiltinVar(stream, version); + } } /*! @@ -1239,8 +1988,16 @@ bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) no && lhs.d->pushConstantBlocks == rhs.d->pushConstantBlocks && lhs.d->storageBlocks == rhs.d->storageBlocks && lhs.d->combinedImageSamplers == rhs.d->combinedImageSamplers + && lhs.d->separateImages == rhs.d->separateImages + && lhs.d->separateSamplers == rhs.d->separateSamplers && lhs.d->storageImages == rhs.d->storageImages - && lhs.d->localSize == rhs.d->localSize; + && lhs.d->inBuiltins == rhs.d->inBuiltins + && lhs.d->outBuiltins == rhs.d->outBuiltins + && lhs.d->localSize == rhs.d->localSize + && lhs.d->tessOutVertCount == rhs.d->tessOutVertCount + && lhs.d->tessMode == rhs.d->tessMode + && lhs.d->tessWind == rhs.d->tessWind + && lhs.d->tessPart == rhs.d->tessPart; } /*! @@ -1258,7 +2015,9 @@ bool operator==(const QShaderDescription::InOutVariable &lhs, const QShaderDescr && lhs.descriptorSet == rhs.descriptorSet && lhs.imageFormat == rhs.imageFormat && lhs.imageFlags == rhs.imageFlags - && lhs.arrayDims == rhs.arrayDims; + && lhs.arrayDims == rhs.arrayDims + && lhs.perPatch == rhs.perPatch + && lhs.structMembers == rhs.structMembers; } /*! @@ -1322,7 +2081,22 @@ bool operator==(const QShaderDescription::StorageBlock &lhs, const QShaderDescri && lhs.knownSize == rhs.knownSize && lhs.binding == rhs.binding && lhs.descriptorSet == rhs.descriptorSet + && lhs.runtimeArrayStride == rhs.runtimeArrayStride + && lhs.qualifierFlags == rhs.qualifierFlags && lhs.members == rhs.members; } +/*! + Returns \c true if the two BuiltinVariable objects \a lhs and \a rhs are + equal. + + \relates QShaderDescription::BuiltinVariable + */ +bool operator==(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept +{ + return lhs.type == rhs.type + && lhs.varType == rhs.varType + && lhs.arrayDims == rhs.arrayDims; +} + QT_END_NAMESPACE diff --git a/src/gui/rhi/qshaderdescription.h b/src/gui/rhi/qshaderdescription.h new file mode 100644 index 0000000000..02492a1598 --- /dev/null +++ b/src/gui/rhi/qshaderdescription.h @@ -0,0 +1,386 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QSHADERDESCRIPTION_H +#define QSHADERDESCRIPTION_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qstring.h> +#include <QtCore/qlist.h> +#include <array> + +QT_BEGIN_NAMESPACE + +struct QShaderDescriptionPrivate; +class QDataStream; + +class Q_GUI_EXPORT QShaderDescription +{ +public: + QShaderDescription(); + QShaderDescription(const QShaderDescription &other); + QShaderDescription &operator=(const QShaderDescription &other); + ~QShaderDescription(); + void detach(); + + bool isValid() const; + + void serialize(QDataStream *stream, int version) const; + QByteArray toJson() const; + + static QShaderDescription deserialize(QDataStream *stream, int version); + + enum VariableType { + Unknown = 0, + + // do not reorder + Float, + Vec2, + Vec3, + Vec4, + Mat2, + Mat2x3, + Mat2x4, + Mat3, + Mat3x2, + Mat3x4, + Mat4, + Mat4x2, + Mat4x3, + + Int, + Int2, + Int3, + Int4, + + Uint, + Uint2, + Uint3, + Uint4, + + Bool, + Bool2, + Bool3, + Bool4, + + Double, + Double2, + Double3, + Double4, + DMat2, + DMat2x3, + DMat2x4, + DMat3, + DMat3x2, + DMat3x4, + DMat4, + DMat4x2, + DMat4x3, + + Sampler1D, + Sampler2D, + Sampler2DMS, + Sampler3D, + SamplerCube, + Sampler1DArray, + Sampler2DArray, + Sampler2DMSArray, + Sampler3DArray, + SamplerCubeArray, + SamplerRect, + SamplerBuffer, + SamplerExternalOES, + Sampler, + + Image1D, + Image2D, + Image2DMS, + Image3D, + ImageCube, + Image1DArray, + Image2DArray, + Image2DMSArray, + Image3DArray, + ImageCubeArray, + ImageRect, + ImageBuffer, + + Struct, + + Half, + Half2, + Half3, + Half4 + }; + + enum ImageFormat { + // must match SPIR-V's ImageFormat + ImageFormatUnknown = 0, + ImageFormatRgba32f = 1, + ImageFormatRgba16f = 2, + ImageFormatR32f = 3, + ImageFormatRgba8 = 4, + ImageFormatRgba8Snorm = 5, + ImageFormatRg32f = 6, + ImageFormatRg16f = 7, + ImageFormatR11fG11fB10f = 8, + ImageFormatR16f = 9, + ImageFormatRgba16 = 10, + ImageFormatRgb10A2 = 11, + ImageFormatRg16 = 12, + ImageFormatRg8 = 13, + ImageFormatR16 = 14, + ImageFormatR8 = 15, + ImageFormatRgba16Snorm = 16, + ImageFormatRg16Snorm = 17, + ImageFormatRg8Snorm = 18, + ImageFormatR16Snorm = 19, + ImageFormatR8Snorm = 20, + ImageFormatRgba32i = 21, + ImageFormatRgba16i = 22, + ImageFormatRgba8i = 23, + ImageFormatR32i = 24, + ImageFormatRg32i = 25, + ImageFormatRg16i = 26, + ImageFormatRg8i = 27, + ImageFormatR16i = 28, + ImageFormatR8i = 29, + ImageFormatRgba32ui = 30, + ImageFormatRgba16ui = 31, + ImageFormatRgba8ui = 32, + ImageFormatR32ui = 33, + ImageFormatRgb10a2ui = 34, + ImageFormatRg32ui = 35, + ImageFormatRg16ui = 36, + ImageFormatRg8ui = 37, + ImageFormatR16ui = 38, + ImageFormatR8ui = 39 + }; + + enum ImageFlag { + ReadOnlyImage = 1 << 0, + WriteOnlyImage = 1 << 1 + }; + Q_DECLARE_FLAGS(ImageFlags, ImageFlag) + + enum QualifierFlag { + QualifierReadOnly = 1 << 0, + QualifierWriteOnly = 1 << 1, + QualifierCoherent = 1 << 2, + QualifierVolatile = 1 << 3, + QualifierRestrict = 1 << 4, + }; + Q_DECLARE_FLAGS(QualifierFlags, QualifierFlag) + + // Optional data (like decorations) usually default to an otherwise invalid value (-1 or 0). This is intentional. + + struct BlockVariable { + QByteArray name; + VariableType type = Unknown; + int offset = 0; + int size = 0; + QList<int> arrayDims; + int arrayStride = 0; + int matrixStride = 0; + bool matrixIsRowMajor = false; + QList<BlockVariable> structMembers; + }; + + struct InOutVariable { + QByteArray name; + VariableType type = Unknown; + int location = -1; + int binding = -1; + int descriptorSet = -1; + ImageFormat imageFormat = ImageFormatUnknown; + ImageFlags imageFlags; + QList<int> arrayDims; + bool perPatch = false; + QList<BlockVariable> structMembers; + }; + + struct UniformBlock { + QByteArray blockName; + QByteArray structName; // instanceName + int size = 0; + int binding = -1; + int descriptorSet = -1; + QList<BlockVariable> members; + }; + + struct PushConstantBlock { + QByteArray name; + int size = 0; + QList<BlockVariable> members; + }; + + struct StorageBlock { + QByteArray blockName; + QByteArray instanceName; + int knownSize = 0; + int binding = -1; + int descriptorSet = -1; + QList<BlockVariable> members; + int runtimeArrayStride = 0; + QualifierFlags qualifierFlags; + }; + + QList<InOutVariable> inputVariables() const; + QList<InOutVariable> outputVariables() const; + QList<UniformBlock> uniformBlocks() const; + QList<PushConstantBlock> pushConstantBlocks() const; + QList<StorageBlock> storageBlocks() const; + QList<InOutVariable> combinedImageSamplers() const; + QList<InOutVariable> separateImages() const; + QList<InOutVariable> separateSamplers() const; + QList<InOutVariable> storageImages() const; + + enum BuiltinType { + // must match SpvBuiltIn + PositionBuiltin = 0, + PointSizeBuiltin = 1, + ClipDistanceBuiltin = 3, + CullDistanceBuiltin = 4, + VertexIdBuiltin = 5, + InstanceIdBuiltin = 6, + PrimitiveIdBuiltin = 7, + InvocationIdBuiltin = 8, + LayerBuiltin = 9, + ViewportIndexBuiltin = 10, + TessLevelOuterBuiltin = 11, + TessLevelInnerBuiltin = 12, + TessCoordBuiltin = 13, + PatchVerticesBuiltin = 14, + FragCoordBuiltin = 15, + PointCoordBuiltin = 16, + FrontFacingBuiltin = 17, + SampleIdBuiltin = 18, + SamplePositionBuiltin = 19, + SampleMaskBuiltin = 20, + FragDepthBuiltin = 22, + NumWorkGroupsBuiltin = 24, + WorkgroupSizeBuiltin = 25, + WorkgroupIdBuiltin = 26, + LocalInvocationIdBuiltin = 27, + GlobalInvocationIdBuiltin = 28, + LocalInvocationIndexBuiltin = 29, + VertexIndexBuiltin = 42, + InstanceIndexBuiltin = 43 + }; + + struct BuiltinVariable { + BuiltinType type; + VariableType varType; + QList<int> arrayDims; + }; + + QList<BuiltinVariable> inputBuiltinVariables() const; + QList<BuiltinVariable> outputBuiltinVariables() const; + + std::array<uint, 3> computeShaderLocalSize() const; + + uint tessellationOutputVertexCount() const; + + enum TessellationMode { + UnknownTessellationMode, + TrianglesTessellationMode, + QuadTessellationMode, + IsolineTessellationMode + }; + + TessellationMode tessellationMode() const; + + enum TessellationWindingOrder { + UnknownTessellationWindingOrder, + CwTessellationWindingOrder, + CcwTessellationWindingOrder + }; + + TessellationWindingOrder tessellationWindingOrder() const; + + enum TessellationPartitioning { + UnknownTessellationPartitioning, + EqualTessellationPartitioning, + FractionalEvenTessellationPartitioning, + FractionalOddTessellationPartitioning + }; + + TessellationPartitioning tessellationPartitioning() const; + +private: + QShaderDescriptionPrivate *d; + friend struct QShaderDescriptionPrivate; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); +#endif + friend Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::ImageFlags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::QualifierFlags) + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::InOutVariable &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BlockVariable &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::UniformBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::PushConstantBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::StorageBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BuiltinVariable &); +#endif + +Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept; + +inline bool operator!=(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qshaderdescription_p.h b/src/gui/rhi/qshaderdescription_p.h index edaa964527..df694bbf40 100644 --- a/src/gui/rhi/qshaderdescription_p.h +++ b/src/gui/rhi/qshaderdescription_p.h @@ -1,44 +1,8 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QSHADERDESCRIPTION_H -#define QSHADERDESCRIPTION_H +#ifndef QSHADERDESCRIPTION_P_H +#define QSHADERDESCRIPTION_P_H // // W A R N I N G @@ -51,275 +15,67 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <QtCore/QString> +#include <rhi/qshaderdescription.h> #include <QtCore/QList> -#include <array> +#include <QtCore/QAtomicInt> +#include <QtCore/QJsonDocument> QT_BEGIN_NAMESPACE -struct QShaderDescriptionPrivate; -class QDataStream; - -class Q_GUI_EXPORT QShaderDescription +struct Q_GUI_EXPORT QShaderDescriptionPrivate { -public: - QShaderDescription(); - QShaderDescription(const QShaderDescription &other); - QShaderDescription &operator=(const QShaderDescription &other); - ~QShaderDescription(); - void detach(); - - bool isValid() const; - - void serialize(QDataStream *stream) const; - QByteArray toJson() const; - - static QShaderDescription deserialize(QDataStream *stream, int version); - - enum VariableType { - Unknown = 0, - - // do not reorder - Float, - Vec2, - Vec3, - Vec4, - Mat2, - Mat2x3, - Mat2x4, - Mat3, - Mat3x2, - Mat3x4, - Mat4, - Mat4x2, - Mat4x3, - - Int, - Int2, - Int3, - Int4, - - Uint, - Uint2, - Uint3, - Uint4, - - Bool, - Bool2, - Bool3, - Bool4, - - Double, - Double2, - Double3, - Double4, - DMat2, - DMat2x3, - DMat2x4, - DMat3, - DMat3x2, - DMat3x4, - DMat4, - DMat4x2, - DMat4x3, - - Sampler1D, - Sampler2D, - Sampler2DMS, - Sampler3D, - SamplerCube, - Sampler1DArray, - Sampler2DArray, - Sampler2DMSArray, - Sampler3DArray, - SamplerCubeArray, - SamplerRect, - SamplerBuffer, - SamplerExternalOES, - - Image1D, - Image2D, - Image2DMS, - Image3D, - ImageCube, - Image1DArray, - Image2DArray, - Image2DMSArray, - Image3DArray, - ImageCubeArray, - ImageRect, - ImageBuffer, - - Struct - }; - - enum ImageFormat { - // must match SPIR-V's ImageFormat - ImageFormatUnknown = 0, - ImageFormatRgba32f = 1, - ImageFormatRgba16f = 2, - ImageFormatR32f = 3, - ImageFormatRgba8 = 4, - ImageFormatRgba8Snorm = 5, - ImageFormatRg32f = 6, - ImageFormatRg16f = 7, - ImageFormatR11fG11fB10f = 8, - ImageFormatR16f = 9, - ImageFormatRgba16 = 10, - ImageFormatRgb10A2 = 11, - ImageFormatRg16 = 12, - ImageFormatRg8 = 13, - ImageFormatR16 = 14, - ImageFormatR8 = 15, - ImageFormatRgba16Snorm = 16, - ImageFormatRg16Snorm = 17, - ImageFormatRg8Snorm = 18, - ImageFormatR16Snorm = 19, - ImageFormatR8Snorm = 20, - ImageFormatRgba32i = 21, - ImageFormatRgba16i = 22, - ImageFormatRgba8i = 23, - ImageFormatR32i = 24, - ImageFormatRg32i = 25, - ImageFormatRg16i = 26, - ImageFormatRg8i = 27, - ImageFormatR16i = 28, - ImageFormatR8i = 29, - ImageFormatRgba32ui = 30, - ImageFormatRgba16ui = 31, - ImageFormatRgba8ui = 32, - ImageFormatR32ui = 33, - ImageFormatRgb10a2ui = 34, - ImageFormatRg32ui = 35, - ImageFormatRg16ui = 36, - ImageFormatRg8ui = 37, - ImageFormatR16ui = 38, - ImageFormatR8ui = 39 - }; - - enum ImageFlag { - ReadOnlyImage = 1 << 0, - WriteOnlyImage = 1 << 1 - }; - Q_DECLARE_FLAGS(ImageFlags, ImageFlag) - - // Optional data (like decorations) usually default to an otherwise invalid value (-1 or 0). This is intentional. - - struct InOutVariable { - QByteArray name; - VariableType type = Unknown; - int location = -1; - int binding = -1; - int descriptorSet = -1; - ImageFormat imageFormat = ImageFormatUnknown; - ImageFlags imageFlags; - QList<int> arrayDims; - }; - - struct BlockVariable { - QByteArray name; - VariableType type = Unknown; - int offset = 0; - int size = 0; - QList<int> arrayDims; - int arrayStride = 0; - int matrixStride = 0; - bool matrixIsRowMajor = false; - QList<BlockVariable> structMembers; - }; - - struct UniformBlock { - QByteArray blockName; - QByteArray structName; // instanceName - int size = 0; - int binding = -1; - int descriptorSet = -1; - QList<BlockVariable> members; - }; - - struct PushConstantBlock { - QByteArray name; - int size = 0; - QList<BlockVariable> members; - }; - - struct StorageBlock { - QByteArray blockName; - QByteArray instanceName; - int knownSize = 0; - int binding = -1; - int descriptorSet = -1; - QList<BlockVariable> members; - }; - - QList<InOutVariable> inputVariables() const; - QList<InOutVariable> outputVariables() const; - QList<UniformBlock> uniformBlocks() const; - QList<PushConstantBlock> pushConstantBlocks() const; - QList<StorageBlock> storageBlocks() const; - QList<InOutVariable> combinedImageSamplers() const; - QList<InOutVariable> storageImages() const; - - std::array<uint, 3> computeShaderLocalSize() const; - -private: - QShaderDescriptionPrivate *d; - friend struct QShaderDescriptionPrivate; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); -#endif - friend Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; + QShaderDescriptionPrivate() + : ref(1) + { + } + + QShaderDescriptionPrivate(const QShaderDescriptionPrivate &other) + : ref(1), + inVars(other.inVars), + outVars(other.outVars), + uniformBlocks(other.uniformBlocks), + pushConstantBlocks(other.pushConstantBlocks), + storageBlocks(other.storageBlocks), + combinedImageSamplers(other.combinedImageSamplers), + separateImages(other.separateImages), + separateSamplers(other.separateSamplers), + storageImages(other.storageImages), + inBuiltins(other.inBuiltins), + outBuiltins(other.outBuiltins), + localSize(other.localSize), + tessOutVertCount(other.tessOutVertCount), + tessMode(other.tessMode), + tessWind(other.tessWind), + tessPart(other.tessPart) + { + } + + static QShaderDescriptionPrivate *get(QShaderDescription *desc) { return desc->d; } + static const QShaderDescriptionPrivate *get(const QShaderDescription *desc) { return desc->d; } + + QJsonDocument makeDoc(); + void writeToStream(QDataStream *stream, int version); + void loadFromStream(QDataStream *stream, int version); + + QAtomicInt ref; + QList<QShaderDescription::InOutVariable> inVars; + QList<QShaderDescription::InOutVariable> outVars; + QList<QShaderDescription::UniformBlock> uniformBlocks; + QList<QShaderDescription::PushConstantBlock> pushConstantBlocks; + QList<QShaderDescription::StorageBlock> storageBlocks; + QList<QShaderDescription::InOutVariable> combinedImageSamplers; + QList<QShaderDescription::InOutVariable> separateImages; + QList<QShaderDescription::InOutVariable> separateSamplers; + QList<QShaderDescription::InOutVariable> storageImages; + QList<QShaderDescription::BuiltinVariable> inBuiltins; + QList<QShaderDescription::BuiltinVariable> outBuiltins; + std::array<uint, 3> localSize = {}; + uint tessOutVertCount = 0; + QShaderDescription::TessellationMode tessMode = QShaderDescription::UnknownTessellationMode; + QShaderDescription::TessellationWindingOrder tessWind = QShaderDescription::UnknownTessellationWindingOrder; + QShaderDescription::TessellationPartitioning tessPart = QShaderDescription::UnknownTessellationPartitioning; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::ImageFlags) - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::InOutVariable &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BlockVariable &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::UniformBlock &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::PushConstantBlock &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::StorageBlock &); -#endif - -Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept; - -inline bool operator!=(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qshaderdescription_p_p.h b/src/gui/rhi/qshaderdescription_p_p.h deleted file mode 100644 index e0bed856b7..0000000000 --- a/src/gui/rhi/qshaderdescription_p_p.h +++ /dev/null @@ -1,102 +0,0 @@ -/**************************************************************************** -** -** Copyright (C) 2019 The Qt Company Ltd. -** Contact: https://www.qt.io/licensing/ -** -** This file is part of the Qt Gui module -** -** $QT_BEGIN_LICENSE:LGPL$ -** Commercial License Usage -** Licensees holding valid commercial Qt licenses may use this file in -** accordance with the commercial license agreement provided with the -** Software or, alternatively, in accordance with the terms contained in -** a written agreement between you and The Qt Company. For licensing terms -** and conditions see https://www.qt.io/terms-conditions. For further -** information use the contact form at https://www.qt.io/contact-us. -** -** GNU Lesser General Public License Usage -** Alternatively, this file may be used under the terms of the GNU Lesser -** General Public License version 3 as published by the Free Software -** Foundation and appearing in the file LICENSE.LGPL3 included in the -** packaging of this file. Please review the following information to -** ensure the GNU Lesser General Public License version 3 requirements -** will be met: https://www.gnu.org/licenses/lgpl-3.0.html. -** -** GNU General Public License Usage -** Alternatively, this file may be used under the terms of the GNU -** General Public License version 2.0 or (at your option) the GNU General -** Public license version 3 or any later version approved by the KDE Free -** Qt Foundation. The licenses are as published by the Free Software -** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3 -** included in the packaging of this file. Please review the following -** information to ensure the GNU General Public License requirements will -** be met: https://www.gnu.org/licenses/gpl-2.0.html and -** https://www.gnu.org/licenses/gpl-3.0.html. -** -** $QT_END_LICENSE$ -** -****************************************************************************/ - -#ifndef QSHADERDESCRIPTION_P_H -#define QSHADERDESCRIPTION_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists for the convenience -// of a number of Qt sources files. This header file may change from -// version to version without notice, or even be removed. -// -// We mean it. -// - -#include "qshaderdescription_p.h" -#include <QtCore/QList> -#include <QtCore/QAtomicInt> -#include <QtCore/QJsonDocument> - -QT_BEGIN_NAMESPACE - -struct Q_GUI_EXPORT QShaderDescriptionPrivate -{ - QShaderDescriptionPrivate() - : ref(1) - { - localSize[0] = localSize[1] = localSize[2] = 0; - } - - QShaderDescriptionPrivate(const QShaderDescriptionPrivate *other) - : ref(1), - inVars(other->inVars), - outVars(other->outVars), - uniformBlocks(other->uniformBlocks), - pushConstantBlocks(other->pushConstantBlocks), - storageBlocks(other->storageBlocks), - combinedImageSamplers(other->combinedImageSamplers), - storageImages(other->storageImages), - localSize(other->localSize) - { - } - - static QShaderDescriptionPrivate *get(QShaderDescription *desc) { return desc->d; } - static const QShaderDescriptionPrivate *get(const QShaderDescription *desc) { return desc->d; } - - QJsonDocument makeDoc(); - void writeToStream(QDataStream *stream); - void loadFromStream(QDataStream *stream, int version); - - QAtomicInt ref; - QList<QShaderDescription::InOutVariable> inVars; - QList<QShaderDescription::InOutVariable> outVars; - QList<QShaderDescription::UniformBlock> uniformBlocks; - QList<QShaderDescription::PushConstantBlock> pushConstantBlocks; - QList<QShaderDescription::StorageBlock> storageBlocks; - QList<QShaderDescription::InOutVariable> combinedImageSamplers; - QList<QShaderDescription::InOutVariable> storageImages; - std::array<uint, 3> localSize; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qt_attribution.json b/src/gui/rhi/qt_attribution.json new file mode 100644 index 0000000000..c356f5f087 --- /dev/null +++ b/src/gui/rhi/qt_attribution.json @@ -0,0 +1,16 @@ +[ + { + "Id": "rhi-miniengine-d3d12-mipmap", + "Name": "Mipmap generator for D3D12", + "QDocModule": "qtgui", + "Description": "Compute shader for mipmap generation from MiniEngine in DirectX-Graphics-Samples", + "QtUsage": "Compute shader for mipmap generation with Direct 3D 12", + + "Homepage": "https://github.com/microsoft/DirectX-Graphics-Samples", + "Version": "0aa79bad78992da0b6a8279ddb9002c1753cb849", + "License": "MIT License", + "LicenseId": "MIT", + "LicenseFile": "MiniEngine_LICENSE.txt", + "Copyright": "Copyright (c) 2015 Microsoft" + } +] diff --git a/src/gui/rhi/tdr.hlsl b/src/gui/rhi/tdr.hlsl deleted file mode 100644 index f79de91c4a..0000000000 --- a/src/gui/rhi/tdr.hlsl +++ /dev/null @@ -1,9 +0,0 @@ -RWBuffer<uint> uav; -cbuffer ConstantBuffer { uint zero; } - -[numthreads(256, 1, 1)] -void killDeviceByTimingOut(uint3 id: SV_DispatchThreadID) -{ - while (zero == 0) - uav[id.x] = zero; -} diff --git a/src/gui/rhi/test.hlsl b/src/gui/rhi/test.hlsl new file mode 100644 index 0000000000..2417762f83 --- /dev/null +++ b/src/gui/rhi/test.hlsl @@ -0,0 +1,24 @@ +struct Input +{ + float4 position : TEXCOORD0; + float3 color : TEXCOORD1; +}; + +struct Output +{ + float4 position : SV_Position; + float3 color : TEXCOORD0; +}; + +cbuffer buf : register(b0) +{ + row_major float4x4 ubuf_mvp; +}; + +Output main(Input input) +{ + Output output; + output.position = mul(input.position, ubuf_mvp); + output.color = input.color; + return output; +} diff --git a/src/gui/rhi/vs_test_p.h b/src/gui/rhi/vs_test_p.h new file mode 100644 index 0000000000..5feaef7d38 --- /dev/null +++ b/src/gui/rhi/vs_test_p.h @@ -0,0 +1,237 @@ +// Copyright (C) 2022 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef VS_TEST_P_H +#define VS_TEST_P_H + +// +// W A R N I N G +// ------------- +// +// This file is not part of the Qt API. It exists for the convenience +// of other Qt classes. This header file may change from version to +// version without notice, or even be removed. +// +// We mean it. +// + +#include <QtCore/private/qglobal_p.h> + +#ifdef Q_OS_WIN + +#include <qt_windows.h> + +#if 0 +// +// Generated by Microsoft (R) HLSL Shader Compiler 10.1 +// +// +// Buffer Definitions: +// +// cbuffer buf +// { +// +// row_major float4x4 ubuf_mvp; // Offset: 0 Size: 64 +// +// } +// +// +// Resource Bindings: +// +// Name Type Format Dim HLSL Bind Count +// ------------------------------ ---------- ------- ----------- -------------- ------ +// buf cbuffer NA NA cb0 1 +// +// +// +// Input signature: +// +// Name Index Mask Register SysValue Format Used +// -------------------- ----- ------ -------- -------- ------- ------ +// TEXCOORD 0 xyzw 0 NONE float xyzw +// TEXCOORD 1 xyz 1 NONE float xyz +// +// +// Output signature: +// +// Name Index Mask Register SysValue Format Used +// -------------------- ----- ------ -------- -------- ------- ------ +// SV_Position 0 xyzw 0 POS float xyzw +// TEXCOORD 0 xyz 1 NONE float xyz +// +vs_5_0 +dcl_globalFlags refactoringAllowed +dcl_constantbuffer CB0[4], immediateIndexed +dcl_input v0.xyzw +dcl_input v1.xyz +dcl_output_siv o0.xyzw, position +dcl_output o1.xyz +dcl_temps 1 +mul r0.xyzw, v0.yyyy, cb0[1].xyzw +mad r0.xyzw, v0.xxxx, cb0[0].xyzw, r0.xyzw +mad r0.xyzw, v0.zzzz, cb0[2].xyzw, r0.xyzw +mad o0.xyzw, v0.wwww, cb0[3].xyzw, r0.xyzw +mov o1.xyz, v1.xyzx +ret +// Approximately 6 instruction slots used +#endif + +inline constexpr BYTE g_testVertexShader[] = +{ + 68, 88, 66, 67, 75, 198, + 18, 149, 172, 244, 247, 123, + 98, 31, 128, 185, 22, 199, + 182, 233, 1, 0, 0, 0, + 140, 3, 0, 0, 5, 0, + 0, 0, 52, 0, 0, 0, + 60, 1, 0, 0, 136, 1, + 0, 0, 224, 1, 0, 0, + 240, 2, 0, 0, 82, 68, + 69, 70, 0, 1, 0, 0, + 1, 0, 0, 0, 96, 0, + 0, 0, 1, 0, 0, 0, + 60, 0, 0, 0, 0, 5, + 254, 255, 0, 1, 0, 0, + 216, 0, 0, 0, 82, 68, + 49, 49, 60, 0, 0, 0, + 24, 0, 0, 0, 32, 0, + 0, 0, 40, 0, 0, 0, + 36, 0, 0, 0, 12, 0, + 0, 0, 0, 0, 0, 0, + 92, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 1, 0, 0, 0, 1, 0, + 0, 0, 98, 117, 102, 0, + 92, 0, 0, 0, 1, 0, + 0, 0, 120, 0, 0, 0, + 64, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 160, 0, 0, 0, 0, 0, + 0, 0, 64, 0, 0, 0, + 2, 0, 0, 0, 180, 0, + 0, 0, 0, 0, 0, 0, + 255, 255, 255, 255, 0, 0, + 0, 0, 255, 255, 255, 255, + 0, 0, 0, 0, 117, 98, + 117, 102, 95, 109, 118, 112, + 0, 102, 108, 111, 97, 116, + 52, 120, 52, 0, 171, 171, + 2, 0, 3, 0, 4, 0, + 4, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 169, 0, 0, 0, + 77, 105, 99, 114, 111, 115, + 111, 102, 116, 32, 40, 82, + 41, 32, 72, 76, 83, 76, + 32, 83, 104, 97, 100, 101, + 114, 32, 67, 111, 109, 112, + 105, 108, 101, 114, 32, 49, + 48, 46, 49, 0, 73, 83, + 71, 78, 68, 0, 0, 0, + 2, 0, 0, 0, 8, 0, + 0, 0, 56, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 3, 0, 0, 0, + 0, 0, 0, 0, 15, 15, + 0, 0, 56, 0, 0, 0, + 1, 0, 0, 0, 0, 0, + 0, 0, 3, 0, 0, 0, + 1, 0, 0, 0, 7, 7, + 0, 0, 84, 69, 88, 67, + 79, 79, 82, 68, 0, 171, + 171, 171, 79, 83, 71, 78, + 80, 0, 0, 0, 2, 0, + 0, 0, 8, 0, 0, 0, + 56, 0, 0, 0, 0, 0, + 0, 0, 1, 0, 0, 0, + 3, 0, 0, 0, 0, 0, + 0, 0, 15, 0, 0, 0, + 68, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 3, 0, 0, 0, 1, 0, + 0, 0, 7, 8, 0, 0, + 83, 86, 95, 80, 111, 115, + 105, 116, 105, 111, 110, 0, + 84, 69, 88, 67, 79, 79, + 82, 68, 0, 171, 171, 171, + 83, 72, 69, 88, 8, 1, + 0, 0, 80, 0, 1, 0, + 66, 0, 0, 0, 106, 8, + 0, 1, 89, 0, 0, 4, + 70, 142, 32, 0, 0, 0, + 0, 0, 4, 0, 0, 0, + 95, 0, 0, 3, 242, 16, + 16, 0, 0, 0, 0, 0, + 95, 0, 0, 3, 114, 16, + 16, 0, 1, 0, 0, 0, + 103, 0, 0, 4, 242, 32, + 16, 0, 0, 0, 0, 0, + 1, 0, 0, 0, 101, 0, + 0, 3, 114, 32, 16, 0, + 1, 0, 0, 0, 104, 0, + 0, 2, 1, 0, 0, 0, + 56, 0, 0, 8, 242, 0, + 16, 0, 0, 0, 0, 0, + 86, 21, 16, 0, 0, 0, + 0, 0, 70, 142, 32, 0, + 0, 0, 0, 0, 1, 0, + 0, 0, 50, 0, 0, 10, + 242, 0, 16, 0, 0, 0, + 0, 0, 6, 16, 16, 0, + 0, 0, 0, 0, 70, 142, + 32, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 70, 14, + 16, 0, 0, 0, 0, 0, + 50, 0, 0, 10, 242, 0, + 16, 0, 0, 0, 0, 0, + 166, 26, 16, 0, 0, 0, + 0, 0, 70, 142, 32, 0, + 0, 0, 0, 0, 2, 0, + 0, 0, 70, 14, 16, 0, + 0, 0, 0, 0, 50, 0, + 0, 10, 242, 32, 16, 0, + 0, 0, 0, 0, 246, 31, + 16, 0, 0, 0, 0, 0, + 70, 142, 32, 0, 0, 0, + 0, 0, 3, 0, 0, 0, + 70, 14, 16, 0, 0, 0, + 0, 0, 54, 0, 0, 5, + 114, 32, 16, 0, 1, 0, + 0, 0, 70, 18, 16, 0, + 1, 0, 0, 0, 62, 0, + 0, 1, 83, 84, 65, 84, + 148, 0, 0, 0, 6, 0, + 0, 0, 1, 0, 0, 0, + 0, 0, 0, 0, 4, 0, + 0, 0, 4, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 1, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 1, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, + 0, 0 +}; + +#endif // Q_OS_WIN + +#endif // VS_TEST_P_H |