diff options
Diffstat (limited to 'src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11')
58 files changed, 20541 insertions, 0 deletions
diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.cpp new file mode 100644 index 0000000000..e38b61709f --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.cpp @@ -0,0 +1,1059 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Blit11.cpp: Texture copy utility class. + +#include "libANGLE/renderer/d3d/d3d11/Blit11.h" + +#include <float.h> + +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/formatutils.h" + +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthrough2d11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughdepth2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba2dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba2di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb2dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb2di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg2dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg2di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr2dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr2di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughlum2d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughlumalpha2d11ps.h" + +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthrough3d11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthrough3d11gs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba3d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba3dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba3di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb3d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb3dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgb3di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg3d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg3dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrg3di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr3d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr3dui11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughr3di11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughlum3d11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughlumalpha3d11ps.h" + +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlef2dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlei2dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzleui2dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlef3dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlei3dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzleui3dps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlef2darrayps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzlei2darrayps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/swizzleui2darrayps.h" + +namespace rx +{ + +static DXGI_FORMAT GetTextureFormat(ID3D11Resource *resource) +{ + ID3D11Texture2D *texture = d3d11::DynamicCastComObject<ID3D11Texture2D>(resource); + if (!texture) + { + return DXGI_FORMAT_UNKNOWN; + } + + D3D11_TEXTURE2D_DESC desc; + texture->GetDesc(&desc); + + SafeRelease(texture); + + return desc.Format; +} + +static ID3D11Resource *CreateStagingTexture(ID3D11Device *device, ID3D11DeviceContext *context, + ID3D11Resource *source, unsigned int subresource, + const gl::Extents &size, unsigned int cpuAccessFlags) +{ + D3D11_TEXTURE2D_DESC stagingDesc; + stagingDesc.Width = size.width; + stagingDesc.Height = size.height; + stagingDesc.MipLevels = 1; + stagingDesc.ArraySize = 1; + stagingDesc.Format = GetTextureFormat(source); + stagingDesc.SampleDesc.Count = 1; + stagingDesc.SampleDesc.Quality = 0; + stagingDesc.Usage = D3D11_USAGE_STAGING; + stagingDesc.CPUAccessFlags = cpuAccessFlags; + stagingDesc.MiscFlags = 0; + stagingDesc.BindFlags = 0; + + ID3D11Texture2D *stagingTexture = NULL; + HRESULT result = device->CreateTexture2D(&stagingDesc, NULL, &stagingTexture); + if (FAILED(result)) + { + ERR("Failed to create staging texture for depth stencil blit. HRESULT: 0x%X.", result); + return NULL; + } + + context->CopySubresourceRegion(stagingTexture, 0, 0, 0, 0, source, subresource, NULL); + + return stagingTexture; +} + +inline static void GenerateVertexCoords(const gl::Box &sourceArea, const gl::Extents &sourceSize, + const gl::Box &destArea, const gl::Extents &destSize, + float *x1, float *y1, float *x2, float *y2, + float *u1, float *v1, float *u2, float *v2) +{ + *x1 = (destArea.x / float(destSize.width)) * 2.0f - 1.0f; + *y1 = ((destSize.height - destArea.y - destArea.height) / float(destSize.height)) * 2.0f - 1.0f; + *x2 = ((destArea.x + destArea.width) / float(destSize.width)) * 2.0f - 1.0f; + *y2 = ((destSize.height - destArea.y) / float(destSize.height)) * 2.0f - 1.0f; + + *u1 = sourceArea.x / float(sourceSize.width); + *v1 = sourceArea.y / float(sourceSize.height); + *u2 = (sourceArea.x + sourceArea.width) / float(sourceSize.width); + *v2 = (sourceArea.y + sourceArea.height) / float(sourceSize.height); +} + +static void Write2DVertices(const gl::Box &sourceArea, const gl::Extents &sourceSize, + const gl::Box &destArea, const gl::Extents &destSize, + void *outVertices, unsigned int *outStride, unsigned int *outVertexCount, + D3D11_PRIMITIVE_TOPOLOGY *outTopology) +{ + float x1, y1, x2, y2, u1, v1, u2, v2; + GenerateVertexCoords(sourceArea, sourceSize, destArea, destSize, &x1, &y1, &x2, &y2, &u1, &v1, &u2, &v2); + + d3d11::PositionTexCoordVertex *vertices = static_cast<d3d11::PositionTexCoordVertex*>(outVertices); + + d3d11::SetPositionTexCoordVertex(&vertices[0], x1, y1, u1, v2); + d3d11::SetPositionTexCoordVertex(&vertices[1], x1, y2, u1, v1); + d3d11::SetPositionTexCoordVertex(&vertices[2], x2, y1, u2, v2); + d3d11::SetPositionTexCoordVertex(&vertices[3], x2, y2, u2, v1); + + *outStride = sizeof(d3d11::PositionTexCoordVertex); + *outVertexCount = 4; + *outTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; +} + +static void Write3DVertices(const gl::Box &sourceArea, const gl::Extents &sourceSize, + const gl::Box &destArea, const gl::Extents &destSize, + void *outVertices, unsigned int *outStride, unsigned int *outVertexCount, + D3D11_PRIMITIVE_TOPOLOGY *outTopology) +{ + ASSERT(sourceSize.depth > 0 && destSize.depth > 0); + + float x1, y1, x2, y2, u1, v1, u2, v2; + GenerateVertexCoords(sourceArea, sourceSize, destArea, destSize, &x1, &y1, &x2, &y2, &u1, &v1, &u2, &v2); + + d3d11::PositionLayerTexCoord3DVertex *vertices = static_cast<d3d11::PositionLayerTexCoord3DVertex*>(outVertices); + + for (int i = 0; i < destSize.depth; i++) + { + float readDepth = (float)i / std::max(destSize.depth - 1, 1); + + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 0], x1, y1, i, u1, v2, readDepth); + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 1], x1, y2, i, u1, v1, readDepth); + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 2], x2, y1, i, u2, v2, readDepth); + + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 3], x1, y2, i, u1, v1, readDepth); + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 4], x2, y2, i, u2, v1, readDepth); + d3d11::SetPositionLayerTexCoord3DVertex(&vertices[i * 6 + 5], x2, y1, i, u2, v2, readDepth); + } + + *outStride = sizeof(d3d11::PositionLayerTexCoord3DVertex); + *outVertexCount = destSize.depth * 6; + *outTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; +} + +Blit11::Blit11(Renderer11 *renderer) + : mRenderer(renderer), mBlitShaderMap(compareBlitParameters), mSwizzleShaderMap(compareSwizzleParameters), + mVertexBuffer(NULL), mPointSampler(NULL), mLinearSampler(NULL), mScissorEnabledRasterizerState(NULL), + mScissorDisabledRasterizerState(NULL), mDepthStencilState(NULL), + mQuad2DIL(NULL), mQuad2DVS(NULL), mDepthPS(NULL), + mQuad3DIL(NULL), mQuad3DVS(NULL), mQuad3DGS(NULL), + mSwizzleCB(NULL) +{ + HRESULT result; + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = std::max(sizeof(d3d11::PositionLayerTexCoord3DVertex), sizeof(d3d11::PositionTexCoordVertex)) * + 6 * renderer->getRendererCaps().max3DTextureSize; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + result = device->CreateBuffer(&vbDesc, NULL, &mVertexBuffer); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mVertexBuffer, "Blit11 vertex buffer"); + + D3D11_SAMPLER_DESC pointSamplerDesc; + pointSamplerDesc.Filter = D3D11_FILTER_MIN_MAG_POINT_MIP_LINEAR; + pointSamplerDesc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP; + pointSamplerDesc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP; + pointSamplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + pointSamplerDesc.MipLODBias = 0.0f; + pointSamplerDesc.MaxAnisotropy = 0; + pointSamplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + pointSamplerDesc.BorderColor[0] = 0.0f; + pointSamplerDesc.BorderColor[1] = 0.0f; + pointSamplerDesc.BorderColor[2] = 0.0f; + pointSamplerDesc.BorderColor[3] = 0.0f; + pointSamplerDesc.MinLOD = 0.0f; + pointSamplerDesc.MaxLOD = FLT_MAX; + + result = device->CreateSamplerState(&pointSamplerDesc, &mPointSampler); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPointSampler, "Blit11 point sampler"); + + D3D11_SAMPLER_DESC linearSamplerDesc; + linearSamplerDesc.Filter = D3D11_FILTER_MIN_MAG_MIP_LINEAR; + linearSamplerDesc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP; + linearSamplerDesc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP; + linearSamplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + linearSamplerDesc.MipLODBias = 0.0f; + linearSamplerDesc.MaxAnisotropy = 0; + linearSamplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + linearSamplerDesc.BorderColor[0] = 0.0f; + linearSamplerDesc.BorderColor[1] = 0.0f; + linearSamplerDesc.BorderColor[2] = 0.0f; + linearSamplerDesc.BorderColor[3] = 0.0f; + linearSamplerDesc.MinLOD = 0.0f; + linearSamplerDesc.MaxLOD = FLT_MAX; + + result = device->CreateSamplerState(&linearSamplerDesc, &mLinearSampler); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mLinearSampler, "Blit11 linear sampler"); + + // Use a rasterizer state that will not cull so that inverted quads will not be culled + D3D11_RASTERIZER_DESC rasterDesc; + rasterDesc.FillMode = D3D11_FILL_SOLID; + rasterDesc.CullMode = D3D11_CULL_NONE; + rasterDesc.FrontCounterClockwise = FALSE; + rasterDesc.DepthBias = 0; + rasterDesc.SlopeScaledDepthBias = 0.0f; + rasterDesc.DepthBiasClamp = 0.0f; + rasterDesc.DepthClipEnable = TRUE; + rasterDesc.MultisampleEnable = FALSE; + rasterDesc.AntialiasedLineEnable = FALSE; + + rasterDesc.ScissorEnable = TRUE; + result = device->CreateRasterizerState(&rasterDesc, &mScissorEnabledRasterizerState); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mScissorEnabledRasterizerState, "Blit11 scissoring rasterizer state"); + + rasterDesc.ScissorEnable = FALSE; + result = device->CreateRasterizerState(&rasterDesc, &mScissorDisabledRasterizerState); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mScissorDisabledRasterizerState, "Blit11 no scissoring rasterizer state"); + + D3D11_DEPTH_STENCIL_DESC depthStencilDesc; + depthStencilDesc.DepthEnable = true; + depthStencilDesc.DepthWriteMask = D3D11_DEPTH_WRITE_MASK_ALL; + depthStencilDesc.DepthFunc = D3D11_COMPARISON_ALWAYS; + depthStencilDesc.StencilEnable = FALSE; + depthStencilDesc.StencilReadMask = D3D11_DEFAULT_STENCIL_READ_MASK; + depthStencilDesc.StencilWriteMask = D3D11_DEFAULT_STENCIL_WRITE_MASK; + depthStencilDesc.FrontFace.StencilFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilDepthFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilPassOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + depthStencilDesc.BackFace.StencilFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilDepthFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilPassOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + + result = device->CreateDepthStencilState(&depthStencilDesc, &mDepthStencilState); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mDepthStencilState, "Blit11 depth stencil state"); + + D3D11_INPUT_ELEMENT_DESC quad2DLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 8, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = device->CreateInputLayout(quad2DLayout, ArraySize(quad2DLayout), g_VS_Passthrough2D, ArraySize(g_VS_Passthrough2D), &mQuad2DIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuad2DIL, "Blit11 2D input layout"); + + result = device->CreateVertexShader(g_VS_Passthrough2D, ArraySize(g_VS_Passthrough2D), NULL, &mQuad2DVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuad2DVS, "Blit11 2D vertex shader"); + + if (renderer->isES3Capable()) + { + result = device->CreatePixelShader(g_PS_PassthroughDepth2D, ArraySize(g_PS_PassthroughDepth2D), NULL, &mDepthPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mDepthPS, "Blit11 2D depth pixel shader"); + + D3D11_INPUT_ELEMENT_DESC quad3DLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "LAYER", 0, DXGI_FORMAT_R32_UINT, 0, 8, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "TEXCOORD", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = device->CreateInputLayout(quad3DLayout, ArraySize(quad3DLayout), g_VS_Passthrough3D, ArraySize(g_VS_Passthrough3D), &mQuad3DIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuad3DIL, "Blit11 3D input layout"); + + result = device->CreateVertexShader(g_VS_Passthrough3D, ArraySize(g_VS_Passthrough3D), NULL, &mQuad3DVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuad3DVS, "Blit11 3D vertex shader"); + + result = device->CreateGeometryShader(g_GS_Passthrough3D, ArraySize(g_GS_Passthrough3D), NULL, &mQuad3DGS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuad3DGS, "Renderer11 copy 3D texture geometry shader"); + } + + buildShaderMap(); + + D3D11_BUFFER_DESC swizzleBufferDesc; + swizzleBufferDesc.ByteWidth = sizeof(unsigned int) * 4; + swizzleBufferDesc.Usage = D3D11_USAGE_DYNAMIC; + swizzleBufferDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + swizzleBufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + swizzleBufferDesc.MiscFlags = 0; + swizzleBufferDesc.StructureByteStride = 0; + + result = device->CreateBuffer(&swizzleBufferDesc, NULL, &mSwizzleCB); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mSwizzleCB, "Blit11 swizzle constant buffer"); +} + +Blit11::~Blit11() +{ + SafeRelease(mVertexBuffer); + SafeRelease(mPointSampler); + SafeRelease(mLinearSampler); + SafeRelease(mScissorEnabledRasterizerState); + SafeRelease(mScissorDisabledRasterizerState); + SafeRelease(mDepthStencilState); + + SafeRelease(mQuad2DIL); + SafeRelease(mQuad2DVS); + SafeRelease(mDepthPS); + + SafeRelease(mQuad3DIL); + SafeRelease(mQuad3DVS); + SafeRelease(mQuad3DGS); + + SafeRelease(mSwizzleCB); + + clearShaderMap(); +} + +static inline unsigned int GetSwizzleIndex(GLenum swizzle) +{ + unsigned int colorIndex = 0; + + switch (swizzle) + { + case GL_RED: colorIndex = 0; break; + case GL_GREEN: colorIndex = 1; break; + case GL_BLUE: colorIndex = 2; break; + case GL_ALPHA: colorIndex = 3; break; + case GL_ZERO: colorIndex = 4; break; + case GL_ONE: colorIndex = 5; break; + default: UNREACHABLE(); break; + } + + return colorIndex; +} + +gl::Error Blit11::swizzleTexture(ID3D11ShaderResourceView *source, ID3D11RenderTargetView *dest, const gl::Extents &size, + GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha) +{ + HRESULT result; + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + D3D11_SHADER_RESOURCE_VIEW_DESC sourceSRVDesc; + source->GetDesc(&sourceSRVDesc); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(sourceSRVDesc.Format); + const gl::InternalFormat &sourceFormatInfo = gl::GetInternalFormatInfo(dxgiFormatInfo.internalFormat); + + GLenum shaderType = GL_NONE; + switch (sourceFormatInfo.componentType) + { + case GL_UNSIGNED_NORMALIZED: + case GL_SIGNED_NORMALIZED: + case GL_FLOAT: + shaderType = GL_FLOAT; + break; + case GL_INT: + shaderType = GL_INT; + break; + case GL_UNSIGNED_INT: + shaderType = GL_UNSIGNED_INT; + break; + default: + UNREACHABLE(); + break; + } + + SwizzleParameters parameters = { 0 }; + parameters.mDestinationType = shaderType; + parameters.mViewDimension = sourceSRVDesc.ViewDimension; + + SwizzleShaderMap::const_iterator i = mSwizzleShaderMap.find(parameters); + if (i == mSwizzleShaderMap.end()) + { + UNREACHABLE(); + return gl::Error(GL_INVALID_OPERATION, "Internal error, missing swizzle shader."); + } + + const Shader &shader = i->second; + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = deviceContext->Map(mVertexBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal vertex buffer for swizzle, HRESULT: 0x%X.", result); + } + + UINT stride = 0; + UINT startIdx = 0; + UINT drawCount = 0; + D3D11_PRIMITIVE_TOPOLOGY topology; + + gl::Box area(0, 0, 0, size.width, size.height, size.depth); + shader.mVertexWriteFunction(area, size, area, size, mappedResource.pData, &stride, &drawCount, &topology); + + deviceContext->Unmap(mVertexBuffer, 0); + + // Set constant buffer + result = deviceContext->Map(mSwizzleCB, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal constant buffer for swizzle, HRESULT: 0x%X.", result); + } + + unsigned int *swizzleIndices = reinterpret_cast<unsigned int*>(mappedResource.pData); + swizzleIndices[0] = GetSwizzleIndex(swizzleRed); + swizzleIndices[1] = GetSwizzleIndex(swizzleGreen); + swizzleIndices[2] = GetSwizzleIndex(swizzleBlue); + swizzleIndices[3] = GetSwizzleIndex(swizzleAlpha); + + deviceContext->Unmap(mSwizzleCB, 0); + + // Apply vertex buffer + deviceContext->IASetVertexBuffers(0, 1, &mVertexBuffer, &stride, &startIdx); + + // Apply constant buffer + deviceContext->PSSetConstantBuffers(0, 1, &mSwizzleCB); + + // Apply state + deviceContext->OMSetBlendState(NULL, NULL, 0xFFFFFFF); + deviceContext->OMSetDepthStencilState(NULL, 0xFFFFFFFF); + deviceContext->RSSetState(mScissorDisabledRasterizerState); + + // Apply shaders + deviceContext->IASetInputLayout(shader.mInputLayout); + deviceContext->IASetPrimitiveTopology(topology); + deviceContext->VSSetShader(shader.mVertexShader, NULL, 0); + + deviceContext->PSSetShader(shader.mPixelShader, NULL, 0); + deviceContext->GSSetShader(shader.mGeometryShader, NULL, 0); + + // Unset the currently bound shader resource to avoid conflicts + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + // Apply render target + mRenderer->setOneTimeRenderTarget(dest); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = size.width; + viewport.Height = size.height; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + // Apply textures + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, source); + + // Apply samplers + deviceContext->PSSetSamplers(0, 1, &mPointSampler); + + // Draw the quad + deviceContext->Draw(drawCount, 0); + + // Unbind textures and render targets and vertex buffer + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + mRenderer->unapplyRenderTargets(); + + UINT zero = 0; + ID3D11Buffer *const nullBuffer = NULL; + deviceContext->IASetVertexBuffers(0, 1, &nullBuffer, &zero, &zero); + + mRenderer->markAllStateDirty(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Blit11::copyTexture(ID3D11ShaderResourceView *source, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11RenderTargetView *dest, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor, GLenum destFormat, GLenum filter) +{ + HRESULT result; + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + // Determine if the source format is a signed integer format, the destFormat will already + // be GL_XXXX_INTEGER but it does not tell us if it is signed or unsigned. + D3D11_SHADER_RESOURCE_VIEW_DESC sourceSRVDesc; + source->GetDesc(&sourceSRVDesc); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(sourceSRVDesc.Format); + const gl::InternalFormat &internalFormatInfo = gl::GetInternalFormatInfo(dxgiFormatInfo.internalFormat); + + BlitParameters parameters = { 0 }; + parameters.mDestinationFormat = destFormat; + parameters.mSignedInteger = (internalFormatInfo.componentType == GL_INT); + parameters.m3DBlit = sourceSRVDesc.ViewDimension == D3D11_SRV_DIMENSION_TEXTURE3D; + + BlitShaderMap::const_iterator i = mBlitShaderMap.find(parameters); + if (i == mBlitShaderMap.end()) + { + UNREACHABLE(); + return gl::Error(GL_OUT_OF_MEMORY, "Could not find appropriate shader for internal texture blit."); + } + + const Shader& shader = i->second; + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = deviceContext->Map(mVertexBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal vertex buffer for texture copy, HRESULT: 0x%X.", result); + } + + UINT stride = 0; + UINT startIdx = 0; + UINT drawCount = 0; + D3D11_PRIMITIVE_TOPOLOGY topology; + + shader.mVertexWriteFunction(sourceArea, sourceSize, destArea, destSize, mappedResource.pData, + &stride, &drawCount, &topology); + + deviceContext->Unmap(mVertexBuffer, 0); + + // Apply vertex buffer + deviceContext->IASetVertexBuffers(0, 1, &mVertexBuffer, &stride, &startIdx); + + // Apply state + deviceContext->OMSetBlendState(NULL, NULL, 0xFFFFFFF); + deviceContext->OMSetDepthStencilState(NULL, 0xFFFFFFFF); + + if (scissor) + { + D3D11_RECT scissorRect; + scissorRect.left = scissor->x; + scissorRect.right = scissor->x + scissor->width; + scissorRect.top = scissor->y; + scissorRect.bottom = scissor->y + scissor->height; + + deviceContext->RSSetScissorRects(1, &scissorRect); + deviceContext->RSSetState(mScissorEnabledRasterizerState); + } + else + { + deviceContext->RSSetState(mScissorDisabledRasterizerState); + } + + // Apply shaders + deviceContext->IASetInputLayout(shader.mInputLayout); + deviceContext->IASetPrimitiveTopology(topology); + deviceContext->VSSetShader(shader.mVertexShader, NULL, 0); + + deviceContext->PSSetShader(shader.mPixelShader, NULL, 0); + deviceContext->GSSetShader(shader.mGeometryShader, NULL, 0); + + // Unset the currently bound shader resource to avoid conflicts + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + // Apply render target + mRenderer->setOneTimeRenderTarget(dest); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = destSize.width; + viewport.Height = destSize.height; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + // Apply textures + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, source); + + // Apply samplers + ID3D11SamplerState *sampler = NULL; + switch (filter) + { + case GL_NEAREST: sampler = mPointSampler; break; + case GL_LINEAR: sampler = mLinearSampler; break; + + default: + UNREACHABLE(); + return gl::Error(GL_OUT_OF_MEMORY, "Internal error, unknown blit filter mode."); + } + deviceContext->PSSetSamplers(0, 1, &sampler); + + // Draw the quad + deviceContext->Draw(drawCount, 0); + + // Unbind textures and render targets and vertex buffer + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + mRenderer->unapplyRenderTargets(); + + UINT zero = 0; + ID3D11Buffer *const nullBuffer = NULL; + deviceContext->IASetVertexBuffers(0, 1, &nullBuffer, &zero, &zero); + + mRenderer->markAllStateDirty(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Blit11::copyStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor) +{ + return copyDepthStencil(source, sourceSubresource, sourceArea, sourceSize, + dest, destSubresource, destArea, destSize, + scissor, true); +} + +gl::Error Blit11::copyDepth(ID3D11ShaderResourceView *source, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11DepthStencilView *dest, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor) +{ + HRESULT result; + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + result = deviceContext->Map(mVertexBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal vertex buffer for texture copy, HRESULT: 0x%X.", result); + } + + UINT stride = 0; + UINT startIdx = 0; + UINT drawCount = 0; + D3D11_PRIMITIVE_TOPOLOGY topology; + + Write2DVertices(sourceArea, sourceSize, destArea, destSize, mappedResource.pData, + &stride, &drawCount, &topology); + + deviceContext->Unmap(mVertexBuffer, 0); + + // Apply vertex buffer + deviceContext->IASetVertexBuffers(0, 1, &mVertexBuffer, &stride, &startIdx); + + // Apply state + deviceContext->OMSetBlendState(NULL, NULL, 0xFFFFFFF); + deviceContext->OMSetDepthStencilState(mDepthStencilState, 0xFFFFFFFF); + + if (scissor) + { + D3D11_RECT scissorRect; + scissorRect.left = scissor->x; + scissorRect.right = scissor->x + scissor->width; + scissorRect.top = scissor->y; + scissorRect.bottom = scissor->y + scissor->height; + + deviceContext->RSSetScissorRects(1, &scissorRect); + deviceContext->RSSetState(mScissorEnabledRasterizerState); + } + else + { + deviceContext->RSSetState(mScissorDisabledRasterizerState); + } + + // Apply shaders + deviceContext->IASetInputLayout(mQuad2DIL); + deviceContext->IASetPrimitiveTopology(topology); + deviceContext->VSSetShader(mQuad2DVS, NULL, 0); + + deviceContext->PSSetShader(mDepthPS, NULL, 0); + deviceContext->GSSetShader(NULL, NULL, 0); + + // Unset the currently bound shader resource to avoid conflicts + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + // Apply render target + deviceContext->OMSetRenderTargets(0, NULL, dest); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = destSize.width; + viewport.Height = destSize.height; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + // Apply textures + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, source); + + // Apply samplers + deviceContext->PSSetSamplers(0, 1, &mPointSampler); + + // Draw the quad + deviceContext->Draw(drawCount, 0); + + // Unbind textures and render targets and vertex buffer + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + mRenderer->unapplyRenderTargets(); + + UINT zero = 0; + ID3D11Buffer *const nullBuffer = NULL; + deviceContext->IASetVertexBuffers(0, 1, &nullBuffer, &zero, &zero); + + mRenderer->markAllStateDirty(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Blit11::copyDepthStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor) +{ + return copyDepthStencil(source, sourceSubresource, sourceArea, sourceSize, + dest, destSubresource, destArea, destSize, + scissor, false); +} + +gl::Error Blit11::copyDepthStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor, bool stencilOnly) +{ + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + ID3D11Resource *sourceStaging = CreateStagingTexture(device, deviceContext, source, sourceSubresource, sourceSize, D3D11_CPU_ACCESS_READ); + // HACK: Create the destination staging buffer as a read/write texture so ID3D11DevicContext::UpdateSubresource can be called + // using it's mapped data as a source + ID3D11Resource *destStaging = CreateStagingTexture(device, deviceContext, dest, destSubresource, destSize, D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE); + + if (!sourceStaging || !destStaging) + { + SafeRelease(sourceStaging); + SafeRelease(destStaging); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal staging textures for depth stencil blit."); + } + + DXGI_FORMAT format = GetTextureFormat(source); + ASSERT(format == GetTextureFormat(dest)); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(format); + unsigned int pixelSize = dxgiFormatInfo.pixelBytes; + unsigned int copyOffset = 0; + unsigned int copySize = pixelSize; + if (stencilOnly) + { + copyOffset = dxgiFormatInfo.depthBits / 8; + copySize = dxgiFormatInfo.stencilBits / 8; + + // It would be expensive to have non-byte sized stencil sizes since it would + // require reading from the destination, currently there aren't any though. + ASSERT(dxgiFormatInfo.stencilBits % 8 == 0 && + dxgiFormatInfo.depthBits % 8 == 0); + } + + D3D11_MAPPED_SUBRESOURCE sourceMapping; + HRESULT result = deviceContext->Map(sourceStaging, 0, D3D11_MAP_READ, 0, &sourceMapping); + if (FAILED(result)) + { + SafeRelease(sourceStaging); + SafeRelease(destStaging); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal source staging texture for depth stencil blit, HRESULT: 0x%X.", result); + } + + D3D11_MAPPED_SUBRESOURCE destMapping; + result = deviceContext->Map(destStaging, 0, D3D11_MAP_WRITE, 0, &destMapping); + if (FAILED(result)) + { + deviceContext->Unmap(sourceStaging, 0); + SafeRelease(sourceStaging); + SafeRelease(destStaging); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal destination staging texture for depth stencil blit, HRESULT: 0x%X.", result); + } + + gl::Rectangle clippedDestArea(destArea.x, destArea.y, destArea.width, destArea.height); + + // Clip dest area to the destination size + gl::ClipRectangle(clippedDestArea, gl::Rectangle(0, 0, destSize.width, destSize.height), &clippedDestArea); + + // Clip dest area to the scissor + if (scissor) + { + gl::ClipRectangle(clippedDestArea, *scissor, &clippedDestArea); + } + + // Determine if entire rows can be copied at once instead of each individual pixel, requires that there is + // no out of bounds lookups required, the entire pixel is copied and no stretching + bool wholeRowCopy = sourceArea.width == clippedDestArea.width && + sourceArea.x >= 0 && sourceArea.x + sourceArea.width <= sourceSize.width && + copySize == pixelSize; + + for (int y = clippedDestArea.y; y < clippedDestArea.y + clippedDestArea.height; y++) + { + float yPerc = static_cast<float>(y - destArea.y) / (destArea.height - 1); + + // Interpolate using the original source rectangle to determine which row to sample from while clamping to the edges + unsigned int readRow = gl::clamp(sourceArea.y + floor(yPerc * (sourceArea.height - 1) + 0.5f), 0, sourceSize.height - 1); + unsigned int writeRow = y; + + if (wholeRowCopy) + { + void *sourceRow = reinterpret_cast<char*>(sourceMapping.pData) + + readRow * sourceMapping.RowPitch + + sourceArea.x * pixelSize; + + void *destRow = reinterpret_cast<char*>(destMapping.pData) + + writeRow * destMapping.RowPitch + + destArea.x * pixelSize; + + memcpy(destRow, sourceRow, pixelSize * destArea.width); + } + else + { + for (int x = clippedDestArea.x; x < clippedDestArea.x + clippedDestArea.width; x++) + { + float xPerc = static_cast<float>(x - destArea.x) / (destArea.width - 1); + + // Interpolate the original source rectangle to determine which column to sample from while clamping to the edges + unsigned int readColumn = gl::clamp(sourceArea.x + floor(xPerc * (sourceArea.width - 1) + 0.5f), 0, sourceSize.width - 1); + unsigned int writeColumn = x; + + void *sourcePixel = reinterpret_cast<char*>(sourceMapping.pData) + + readRow * sourceMapping.RowPitch + + readColumn * pixelSize + + copyOffset; + + void *destPixel = reinterpret_cast<char*>(destMapping.pData) + + writeRow * destMapping.RowPitch + + writeColumn * pixelSize + + copyOffset; + + memcpy(destPixel, sourcePixel, copySize); + } + } + } + + // HACK: Use ID3D11DevicContext::UpdateSubresource which causes an extra copy compared to ID3D11DevicContext::CopySubresourceRegion + // according to MSDN. + deviceContext->UpdateSubresource(dest, destSubresource, NULL, destMapping.pData, destMapping.RowPitch, destMapping.DepthPitch); + + deviceContext->Unmap(sourceStaging, 0); + deviceContext->Unmap(destStaging, 0); + + // TODO: Determine why this call to ID3D11DevicContext::CopySubresourceRegion causes a TDR timeout on some + // systems when called repeatedly. + // deviceContext->CopySubresourceRegion(dest, destSubresource, 0, 0, 0, destStaging, 0, NULL); + + SafeRelease(sourceStaging); + SafeRelease(destStaging); + + return gl::Error(GL_NO_ERROR); +} + +bool Blit11::compareBlitParameters(const Blit11::BlitParameters &a, const Blit11::BlitParameters &b) +{ + return memcmp(&a, &b, sizeof(Blit11::BlitParameters)) < 0; +} + +bool Blit11::compareSwizzleParameters(const SwizzleParameters &a, const SwizzleParameters &b) +{ + return memcmp(&a, &b, sizeof(Blit11::SwizzleParameters)) < 0; +} + +void Blit11::add2DBlitShaderToMap(GLenum destFormat, bool signedInteger, ID3D11PixelShader *ps) +{ + BlitParameters params = { 0 }; + params.mDestinationFormat = destFormat; + params.mSignedInteger = signedInteger; + params.m3DBlit = false; + + ASSERT(mBlitShaderMap.find(params) == mBlitShaderMap.end()); + ASSERT(ps); + + Shader shader; + shader.mVertexWriteFunction = Write2DVertices; + shader.mInputLayout = mQuad2DIL; + shader.mVertexShader = mQuad2DVS; + shader.mGeometryShader = NULL; + shader.mPixelShader = ps; + + mBlitShaderMap[params] = shader; +} + +void Blit11::add3DBlitShaderToMap(GLenum destFormat, bool signedInteger, ID3D11PixelShader *ps) +{ + BlitParameters params = { 0 }; + params.mDestinationFormat = destFormat; + params.mSignedInteger = signedInteger; + params.m3DBlit = true; + + ASSERT(mBlitShaderMap.find(params) == mBlitShaderMap.end()); + ASSERT(ps); + + Shader shader; + shader.mVertexWriteFunction = Write3DVertices; + shader.mInputLayout = mQuad3DIL; + shader.mVertexShader = mQuad3DVS; + shader.mGeometryShader = mQuad3DGS; + shader.mPixelShader = ps; + + mBlitShaderMap[params] = shader; +} + +void Blit11::addSwizzleShaderToMap(GLenum destType, D3D11_SRV_DIMENSION viewDimension, ID3D11PixelShader *ps) +{ + SwizzleParameters params = { 0 }; + params.mDestinationType = destType; + params.mViewDimension = viewDimension; + + ASSERT(mSwizzleShaderMap.find(params) == mSwizzleShaderMap.end()); + ASSERT(ps); + + Shader shader; + switch (viewDimension) + { + case D3D_SRV_DIMENSION_TEXTURE2D: + shader.mVertexWriteFunction = Write2DVertices; + shader.mInputLayout = mQuad2DIL; + shader.mVertexShader = mQuad2DVS; + shader.mGeometryShader = NULL; + break; + + case D3D_SRV_DIMENSION_TEXTURE3D: + case D3D_SRV_DIMENSION_TEXTURE2DARRAY: + case D3D_SRV_DIMENSION_TEXTURECUBE: + shader.mVertexWriteFunction = Write3DVertices; + shader.mInputLayout = mQuad3DIL; + shader.mVertexShader = mQuad3DVS; + shader.mGeometryShader = mQuad3DGS; + break; + + default: + UNREACHABLE(); + break; + } + shader.mPixelShader = ps; + + mSwizzleShaderMap[params] = shader; +} + +void Blit11::buildShaderMap() +{ + ID3D11Device *device = mRenderer->getDevice(); + + // 2D shaders (OpenGL ES 2+) + add2DBlitShaderToMap(GL_RGBA, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA2D, "Blit11 2D RGBA pixel shader" )); + add2DBlitShaderToMap(GL_BGRA_EXT, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA2D, "Blit11 2D BGRA pixel shader" )); + add2DBlitShaderToMap(GL_RGB, false, d3d11::CompilePS(device, g_PS_PassthroughRGB2D, "Blit11 2D RGB pixel shader" )); + add2DBlitShaderToMap(GL_RG, false, d3d11::CompilePS(device, g_PS_PassthroughRG2D, "Blit11 2D RG pixel shader" )); + add2DBlitShaderToMap(GL_RED, false, d3d11::CompilePS(device, g_PS_PassthroughR2D, "Blit11 2D R pixel shader" )); + add2DBlitShaderToMap(GL_ALPHA, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA2D, "Blit11 2D alpha pixel shader" )); + add2DBlitShaderToMap(GL_LUMINANCE, false, d3d11::CompilePS(device, g_PS_PassthroughLum2D, "Blit11 2D lum pixel shader" )); + add2DBlitShaderToMap(GL_LUMINANCE_ALPHA, false, d3d11::CompilePS(device, g_PS_PassthroughLumAlpha2D, "Blit11 2D luminance alpha pixel shader")); + + // 2D shaders (OpenGL ES 3+) + if (mRenderer->isES3Capable()) + { + add2DBlitShaderToMap(GL_RGBA_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA2DUI, "Blit11 2D RGBA UI pixel shader" )); + add2DBlitShaderToMap(GL_RGBA_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRGBA2DI, "Blit11 2D RGBA I pixel shader" )); + add2DBlitShaderToMap(GL_RGB_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRGB2DUI, "Blit11 2D RGB UI pixel shader" )); + add2DBlitShaderToMap(GL_RGB_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRGB2DI, "Blit11 2D RGB I pixel shader" )); + add2DBlitShaderToMap(GL_RG_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRG2DUI, "Blit11 2D RG UI pixel shader" )); + add2DBlitShaderToMap(GL_RG_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRG2DI, "Blit11 2D RG I pixel shader" )); + add2DBlitShaderToMap(GL_RED_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughR2DUI, "Blit11 2D R UI pixel shader" )); + add2DBlitShaderToMap(GL_RED_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughR2DI, "Blit11 2D R I pixel shader" )); + } + + // 3D shaders (OpenGL ES 3+) + if (mRenderer->isES3Capable()) + { + add3DBlitShaderToMap(GL_RGBA, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA3D, "Blit11 3D RGBA pixel shader" )); + add3DBlitShaderToMap(GL_RGBA_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA3DUI, "Blit11 3D UI RGBA pixel shader" )); + add3DBlitShaderToMap(GL_RGBA_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRGBA3DI, "Blit11 3D I RGBA pixel shader" )); + add3DBlitShaderToMap(GL_BGRA_EXT, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA3D, "Blit11 3D BGRA pixel shader" )); + add3DBlitShaderToMap(GL_RGB, false, d3d11::CompilePS(device, g_PS_PassthroughRGB3D, "Blit11 3D RGB pixel shader" )); + add3DBlitShaderToMap(GL_RGB_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRGB3DUI, "Blit11 3D RGB UI pixel shader" )); + add3DBlitShaderToMap(GL_RGB_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRGB3DI, "Blit11 3D RGB I pixel shader" )); + add3DBlitShaderToMap(GL_RG, false, d3d11::CompilePS(device, g_PS_PassthroughRG3D, "Blit11 3D RG pixel shader" )); + add3DBlitShaderToMap(GL_RG_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughRG3DUI, "Blit11 3D RG UI pixel shader" )); + add3DBlitShaderToMap(GL_RG_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughRG3DI, "Blit11 3D RG I pixel shader" )); + add3DBlitShaderToMap(GL_RED, false, d3d11::CompilePS(device, g_PS_PassthroughR3D, "Blit11 3D R pixel shader" )); + add3DBlitShaderToMap(GL_RED_INTEGER, false, d3d11::CompilePS(device, g_PS_PassthroughR3DUI, "Blit11 3D R UI pixel shader" )); + add3DBlitShaderToMap(GL_RED_INTEGER, true, d3d11::CompilePS(device, g_PS_PassthroughR3DI, "Blit11 3D R I pixel shader" )); + add3DBlitShaderToMap(GL_ALPHA, false, d3d11::CompilePS(device, g_PS_PassthroughRGBA3D, "Blit11 3D alpha pixel shader" )); + add3DBlitShaderToMap(GL_LUMINANCE, false, d3d11::CompilePS(device, g_PS_PassthroughLum3D, "Blit11 3D luminance pixel shader" )); + add3DBlitShaderToMap(GL_LUMINANCE_ALPHA, false, d3d11::CompilePS(device, g_PS_PassthroughLumAlpha3D, "Blit11 3D luminance alpha pixel shader")); + } + + // Swizzling shaders (OpenGL ES 3+) + if (mRenderer->isES3Capable()) + { + addSwizzleShaderToMap(GL_FLOAT, D3D_SRV_DIMENSION_TEXTURE2D, d3d11::CompilePS(device, g_PS_SwizzleF2D, "Blit11 2D F swizzle pixel shader" )); + addSwizzleShaderToMap(GL_UNSIGNED_INT, D3D_SRV_DIMENSION_TEXTURE2D, d3d11::CompilePS(device, g_PS_SwizzleUI2D, "Blit11 2D UI swizzle pixel shader")); + addSwizzleShaderToMap(GL_INT, D3D_SRV_DIMENSION_TEXTURE2D, d3d11::CompilePS(device, g_PS_SwizzleI2D, "Blit11 2D I swizzle pixel shader" )); + + addSwizzleShaderToMap(GL_FLOAT, D3D_SRV_DIMENSION_TEXTURECUBE, d3d11::CompilePS(device, g_PS_SwizzleF2DArray, "Blit11 2D Cube F swizzle pixel shader" )); + addSwizzleShaderToMap(GL_UNSIGNED_INT, D3D_SRV_DIMENSION_TEXTURECUBE, d3d11::CompilePS(device, g_PS_SwizzleUI2DArray, "Blit11 2D Cube UI swizzle pixel shader")); + addSwizzleShaderToMap(GL_INT, D3D_SRV_DIMENSION_TEXTURECUBE, d3d11::CompilePS(device, g_PS_SwizzleI2DArray, "Blit11 2D Cube I swizzle pixel shader" )); + + addSwizzleShaderToMap(GL_FLOAT, D3D_SRV_DIMENSION_TEXTURE3D, d3d11::CompilePS(device, g_PS_SwizzleF3D, "Blit11 3D F swizzle pixel shader" )); + addSwizzleShaderToMap(GL_UNSIGNED_INT, D3D_SRV_DIMENSION_TEXTURE3D, d3d11::CompilePS(device, g_PS_SwizzleUI3D, "Blit11 3D UI swizzle pixel shader")); + addSwizzleShaderToMap(GL_INT, D3D_SRV_DIMENSION_TEXTURE3D, d3d11::CompilePS(device, g_PS_SwizzleI3D, "Blit11 3D I swizzle pixel shader" )); + + addSwizzleShaderToMap(GL_FLOAT, D3D_SRV_DIMENSION_TEXTURE2DARRAY, d3d11::CompilePS(device, g_PS_SwizzleF2DArray, "Blit11 2D Array F swizzle pixel shader" )); + addSwizzleShaderToMap(GL_UNSIGNED_INT, D3D_SRV_DIMENSION_TEXTURE2DARRAY, d3d11::CompilePS(device, g_PS_SwizzleUI2DArray, "Blit11 2D Array UI swizzle pixel shader")); + addSwizzleShaderToMap(GL_INT, D3D_SRV_DIMENSION_TEXTURE2DARRAY, d3d11::CompilePS(device, g_PS_SwizzleI2DArray, "Blit11 2D Array I swizzle pixel shader" )); + } +} + +void Blit11::clearShaderMap() +{ + for (BlitShaderMap::iterator i = mBlitShaderMap.begin(); i != mBlitShaderMap.end(); ++i) + { + Shader &shader = i->second; + SafeRelease(shader.mPixelShader); + } + mBlitShaderMap.clear(); + + for (SwizzleShaderMap::iterator i = mSwizzleShaderMap.begin(); i != mSwizzleShaderMap.end(); ++i) + { + Shader &shader = i->second; + SafeRelease(shader.mPixelShader); + } + mSwizzleShaderMap.clear(); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.h new file mode 100644 index 0000000000..d3a8c2c8a3 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Blit11.h @@ -0,0 +1,121 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Blit11.cpp: Texture copy utility class. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_BLIT11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_BLIT11_H_ + +#include "common/angleutils.h" +#include "libANGLE/angletypes.h" +#include "libANGLE/Error.h" + +#include <map> + +namespace rx +{ +class Renderer11; + +class Blit11 : angle::NonCopyable +{ + public: + explicit Blit11(Renderer11 *renderer); + ~Blit11(); + + gl::Error swizzleTexture(ID3D11ShaderResourceView *source, ID3D11RenderTargetView *dest, const gl::Extents &size, + GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha); + + gl::Error copyTexture(ID3D11ShaderResourceView *source, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11RenderTargetView *dest, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor, GLenum destFormat, GLenum filter); + + gl::Error copyStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor); + + gl::Error copyDepth(ID3D11ShaderResourceView *source, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11DepthStencilView *dest, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor); + + gl::Error copyDepthStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor); + + private: + Renderer11 *mRenderer; + + struct BlitParameters + { + GLenum mDestinationFormat; + bool mSignedInteger; + bool m3DBlit; + }; + + gl::Error copyDepthStencil(ID3D11Resource *source, unsigned int sourceSubresource, const gl::Box &sourceArea, const gl::Extents &sourceSize, + ID3D11Resource *dest, unsigned int destSubresource, const gl::Box &destArea, const gl::Extents &destSize, + const gl::Rectangle *scissor, bool stencilOnly); + + static bool compareBlitParameters(const BlitParameters &a, const BlitParameters &b); + + typedef void (*WriteVertexFunction)(const gl::Box &sourceArea, const gl::Extents &sourceSize, + const gl::Box &destArea, const gl::Extents &destSize, + void *outVertices, unsigned int *outStride, unsigned int *outVertexCount, + D3D11_PRIMITIVE_TOPOLOGY *outTopology); + + struct Shader + { + WriteVertexFunction mVertexWriteFunction; + ID3D11InputLayout *mInputLayout; + ID3D11VertexShader *mVertexShader; + ID3D11GeometryShader *mGeometryShader; + ID3D11PixelShader *mPixelShader; + }; + + typedef bool (*BlitParametersComparisonFunction)(const BlitParameters&, const BlitParameters &); + typedef std::map<BlitParameters, Shader, BlitParametersComparisonFunction> BlitShaderMap; + BlitShaderMap mBlitShaderMap; + + void add2DBlitShaderToMap(GLenum destFormat, bool signedInteger, ID3D11PixelShader *ps); + void add3DBlitShaderToMap(GLenum destFormat, bool signedInteger, ID3D11PixelShader *ps); + + struct SwizzleParameters + { + GLenum mDestinationType; + D3D11_SRV_DIMENSION mViewDimension; + }; + + static bool compareSwizzleParameters(const SwizzleParameters &a, const SwizzleParameters &b); + + typedef bool (*SwizzleParametersComparisonFunction)(const SwizzleParameters&, const SwizzleParameters &); + typedef std::map<SwizzleParameters, Shader, SwizzleParametersComparisonFunction> SwizzleShaderMap; + SwizzleShaderMap mSwizzleShaderMap; + + void addSwizzleShaderToMap(GLenum destType, D3D11_SRV_DIMENSION viewDimension, ID3D11PixelShader *ps); + + void buildShaderMap(); + void clearShaderMap(); + + ID3D11Buffer *mVertexBuffer; + ID3D11SamplerState *mPointSampler; + ID3D11SamplerState *mLinearSampler; + ID3D11RasterizerState *mScissorEnabledRasterizerState; + ID3D11RasterizerState *mScissorDisabledRasterizerState; + ID3D11DepthStencilState *mDepthStencilState; + + ID3D11InputLayout *mQuad2DIL; + ID3D11VertexShader *mQuad2DVS; + ID3D11PixelShader *mDepthPS; + + ID3D11InputLayout *mQuad3DIL; + ID3D11VertexShader *mQuad3DVS; + ID3D11GeometryShader *mQuad3DGS; + + ID3D11Buffer *mSwizzleCB; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_BLIT11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.cpp new file mode 100644 index 0000000000..d56b0ea7ad --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.cpp @@ -0,0 +1,1070 @@ +// +// Copyright 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Buffer11.cpp Defines the Buffer11 class. + +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" + +#include "common/MemoryBuffer.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" + +#if defined(ANGLE_MINGW32_COMPAT) +typedef enum D3D11_MAP_FLAG { + D3D11_MAP_FLAG_DO_NOT_WAIT = 0x100000 +} D3D11_MAP_FLAG; +#endif + +namespace rx +{ + +PackPixelsParams::PackPixelsParams() + : format(GL_NONE), + type(GL_NONE), + outputPitch(0), + packBuffer(NULL), + offset(0) +{} + +PackPixelsParams::PackPixelsParams(const gl::Rectangle &areaIn, GLenum formatIn, GLenum typeIn, GLuint outputPitchIn, + const gl::PixelPackState &packIn, ptrdiff_t offsetIn) + : area(areaIn), + format(formatIn), + type(typeIn), + outputPitch(outputPitchIn), + packBuffer(packIn.pixelBuffer.get()), + pack(packIn.alignment, packIn.reverseRowOrder), + offset(offsetIn) +{} + +namespace gl_d3d11 +{ + +D3D11_MAP GetD3DMapTypeFromBits(GLbitfield access) +{ + bool readBit = ((access & GL_MAP_READ_BIT) != 0); + bool writeBit = ((access & GL_MAP_WRITE_BIT) != 0); + + ASSERT(readBit || writeBit); + + // Note : we ignore the discard bit, because in D3D11, staging buffers + // don't accept the map-discard flag (discard only works for DYNAMIC usage) + + if (readBit && !writeBit) + { + return D3D11_MAP_READ; + } + else if (writeBit && !readBit) + { + return D3D11_MAP_WRITE; + } + else if (writeBit && readBit) + { + return D3D11_MAP_READ_WRITE; + } + else + { + UNREACHABLE(); + return D3D11_MAP_READ; + } +} + +} + +// Each instance of Buffer11::BufferStorage is specialized for a class of D3D binding points +// - vertex/transform feedback buffers +// - index buffers +// - pixel unpack buffers +// - uniform buffers +class Buffer11::BufferStorage : angle::NonCopyable +{ + public: + virtual ~BufferStorage() {} + + DataRevision getDataRevision() const { return mRevision; } + BufferUsage getUsage() const { return mUsage; } + size_t getSize() const { return mBufferSize; } + void setDataRevision(DataRevision rev) { mRevision = rev; } + + virtual bool isMappable() const = 0; + + virtual bool copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) = 0; + virtual gl::Error resize(size_t size, bool preserveData) = 0; + + virtual uint8_t *map(size_t offset, size_t length, GLbitfield access) = 0; + virtual void unmap() = 0; + + gl::Error setData(const uint8_t *data, size_t offset, size_t size); + + protected: + BufferStorage(Renderer11 *renderer, BufferUsage usage); + + Renderer11 *mRenderer; + DataRevision mRevision; + const BufferUsage mUsage; + size_t mBufferSize; +}; + +// A native buffer storage represents an underlying D3D11 buffer for a particular +// type of storage. +class Buffer11::NativeStorage : public Buffer11::BufferStorage +{ + public: + NativeStorage(Renderer11 *renderer, BufferUsage usage); + ~NativeStorage() override; + + bool isMappable() const override { return mUsage == BUFFER_USAGE_STAGING; } + + ID3D11Buffer *getNativeStorage() const { return mNativeStorage; } + + bool copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) override; + gl::Error resize(size_t size, bool preserveData) override; + + uint8_t *map(size_t offset, size_t length, GLbitfield access) override; + void unmap() override; + + private: + static void fillBufferDesc(D3D11_BUFFER_DESC* bufferDesc, Renderer11 *renderer, BufferUsage usage, unsigned int bufferSize); + + ID3D11Buffer *mNativeStorage; +}; + +// Pack storage represents internal storage for pack buffers. We implement pack buffers +// as CPU memory, tied to a staging texture, for asynchronous texture readback. +class Buffer11::PackStorage : public Buffer11::BufferStorage +{ + public: + explicit PackStorage(Renderer11 *renderer); + ~PackStorage() override; + + bool isMappable() const override { return true; } + + bool copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) override; + gl::Error resize(size_t size, bool preserveData) override; + + uint8_t *map(size_t offset, size_t length, GLbitfield access) override; + void unmap() override; + + gl::Error packPixels(ID3D11Texture2D *srcTexure, UINT srcSubresource, const PackPixelsParams ¶ms); + + private: + gl::Error flushQueuedPackCommand(); + + ID3D11Texture2D *mStagingTexture; + DXGI_FORMAT mTextureFormat; + gl::Extents mTextureSize; + MemoryBuffer mMemoryBuffer; + PackPixelsParams *mQueuedPackCommand; + PackPixelsParams mPackParams; + bool mDataModified; +}; + +// System memory storage stores a CPU memory buffer with our buffer data. +// For dynamic data, it's much faster to update the CPU memory buffer than +// it is to update a D3D staging buffer and read it back later. +class Buffer11::SystemMemoryStorage : public Buffer11::BufferStorage +{ + public: + explicit SystemMemoryStorage(Renderer11 *renderer); + ~SystemMemoryStorage() override {} + + bool isMappable() const override { return true; } + + bool copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) override; + gl::Error resize(size_t size, bool preserveData) override; + + uint8_t *map(size_t offset, size_t length, GLbitfield access) override; + void unmap() override; + + MemoryBuffer *getSystemCopy() { return &mSystemCopy; } + + protected: + MemoryBuffer mSystemCopy; +}; + +Buffer11::Buffer11(Renderer11 *renderer) + : BufferD3D(renderer), + mRenderer(renderer), + mSize(0), + mMappedStorage(NULL), + mReadUsageCount(0), + mHasSystemMemoryStorage(false) +{} + +Buffer11::~Buffer11() +{ + for (auto it = mBufferStorages.begin(); it != mBufferStorages.end(); it++) + { + SafeDelete(it->second); + } +} + +Buffer11 *Buffer11::makeBuffer11(BufferImpl *buffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(Buffer11*, buffer)); + return static_cast<Buffer11*>(buffer); +} + +gl::Error Buffer11::setData(const void *data, size_t size, GLenum usage) +{ + gl::Error error = setSubData(data, size, 0); + if (error.isError()) + { + return error; + } + + if (usage == GL_STATIC_DRAW) + { + initializeStaticData(); + } + + return error; +} + +gl::Error Buffer11::getData(const uint8_t **outData) +{ + SystemMemoryStorage *systemMemoryStorage = nullptr; + gl::Error error = getSystemMemoryStorage(&systemMemoryStorage); + + if (error.isError()) + { + *outData = nullptr; + return error; + } + + mReadUsageCount = 0; + + ASSERT(systemMemoryStorage->getSize() >= mSize); + + *outData = systemMemoryStorage->getSystemCopy()->data(); + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::getSystemMemoryStorage(SystemMemoryStorage **storageOut) +{ + BufferStorage *memStorageUntyped = getBufferStorage(BUFFER_USAGE_SYSTEM_MEMORY); + + if (memStorageUntyped == nullptr) + { + // TODO(jmadill): convert all to errors + return gl::Error(GL_OUT_OF_MEMORY); + } + + *storageOut = GetAs<SystemMemoryStorage>(memStorageUntyped); + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::setSubData(const void *data, size_t size, size_t offset) +{ + size_t requiredSize = size + offset; + + if (data && size > 0) + { + // Use system memory storage for dynamic buffers. + + BufferStorage *writeBuffer = nullptr; + if (supportsDirectBinding()) + { + writeBuffer = getStagingStorage(); + + if (!writeBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate internal buffer."); + } + } + else + { + SystemMemoryStorage *systemMemoryStorage = nullptr; + gl::Error error = getSystemMemoryStorage(&systemMemoryStorage); + if (error.isError()) + { + return error; + } + + writeBuffer = systemMemoryStorage; + } + + ASSERT(writeBuffer); + + // Explicitly resize the staging buffer, preserving data if the new data will not + // completely fill the buffer + if (writeBuffer->getSize() < requiredSize) + { + bool preserveData = (offset > 0); + gl::Error error = writeBuffer->resize(requiredSize, preserveData); + if (error.isError()) + { + return error; + } + } + + writeBuffer->setData(static_cast<const uint8_t *>(data), offset, size); + writeBuffer->setDataRevision(writeBuffer->getDataRevision() + 1); + } + + mSize = std::max(mSize, requiredSize); + invalidateStaticData(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::copySubData(BufferImpl* source, GLintptr sourceOffset, GLintptr destOffset, GLsizeiptr size) +{ + Buffer11 *sourceBuffer = makeBuffer11(source); + ASSERT(sourceBuffer != NULL); + + BufferStorage *copyDest = getLatestBufferStorage(); + if (!copyDest) + { + copyDest = getStagingStorage(); + } + + BufferStorage *copySource = sourceBuffer->getLatestBufferStorage(); + + if (!copySource || !copyDest) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate internal staging buffer."); + } + + // If copying to/from a pixel pack buffer, we must have a staging or + // pack buffer partner, because other native buffers can't be mapped + if (copyDest->getUsage() == BUFFER_USAGE_PIXEL_PACK && !copySource->isMappable()) + { + copySource = sourceBuffer->getStagingStorage(); + } + else if (copySource->getUsage() == BUFFER_USAGE_PIXEL_PACK && !copyDest->isMappable()) + { + copyDest = getStagingStorage(); + } + + // D3D11 does not allow overlapped copies until 11.1, and only if the + // device supports D3D11_FEATURE_DATA_D3D11_OPTIONS::CopyWithOverlap + // Get around this via a different source buffer + if (copySource == copyDest) + { + if (copySource->getUsage() == BUFFER_USAGE_STAGING) + { + copySource = getBufferStorage(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK); + } + else + { + copySource = getStagingStorage(); + } + } + + copyDest->copyFromStorage(copySource, sourceOffset, size, destOffset); + copyDest->setDataRevision(copyDest->getDataRevision() + 1); + + mSize = std::max<size_t>(mSize, destOffset + size); + invalidateStaticData(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::map(size_t offset, size_t length, GLbitfield access, GLvoid **mapPtr) +{ + ASSERT(!mMappedStorage); + + BufferStorage *latestStorage = getLatestBufferStorage(); + if (latestStorage && + (latestStorage->getUsage() == BUFFER_USAGE_PIXEL_PACK || + latestStorage->getUsage() == BUFFER_USAGE_STAGING)) + { + // Latest storage is mappable. + mMappedStorage = latestStorage; + } + else + { + // Fall back to using the staging buffer if the latest storage does + // not exist or is not CPU-accessible. + mMappedStorage = getStagingStorage(); + } + + if (!mMappedStorage) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate mappable internal buffer."); + } + + if ((access & GL_MAP_WRITE_BIT) > 0) + { + // Update the data revision immediately, since the data might be changed at any time + mMappedStorage->setDataRevision(mMappedStorage->getDataRevision() + 1); + } + + uint8_t *mappedBuffer = mMappedStorage->map(offset, length, access); + if (!mappedBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal buffer."); + } + + *mapPtr = static_cast<GLvoid *>(mappedBuffer); + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::unmap() +{ + ASSERT(mMappedStorage); + mMappedStorage->unmap(); + mMappedStorage = NULL; + return gl::Error(GL_NO_ERROR); +} + +void Buffer11::markTransformFeedbackUsage() +{ + BufferStorage *transformFeedbackStorage = getBufferStorage(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK); + + if (transformFeedbackStorage) + { + transformFeedbackStorage->setDataRevision(transformFeedbackStorage->getDataRevision() + 1); + } + + invalidateStaticData(); +} + +void Buffer11::markBufferUsage() +{ + mReadUsageCount++; + + // Free the system memory storage if we decide it isn't being used very often. + const unsigned int usageLimit = 5; + + if (mReadUsageCount > usageLimit && mHasSystemMemoryStorage) + { + auto systemMemoryStorageIt = mBufferStorages.find(BUFFER_USAGE_SYSTEM_MEMORY); + ASSERT(systemMemoryStorageIt != mBufferStorages.end()); + + SafeDelete(systemMemoryStorageIt->second); + mBufferStorages.erase(systemMemoryStorageIt); + mHasSystemMemoryStorage = false; + } +} + +ID3D11Buffer *Buffer11::getBuffer(BufferUsage usage) +{ + markBufferUsage(); + + BufferStorage *bufferStorage = getBufferStorage(usage); + + if (!bufferStorage) + { + // Storage out-of-memory + return NULL; + } + + ASSERT(HAS_DYNAMIC_TYPE(NativeStorage*, bufferStorage)); + + return static_cast<NativeStorage*>(bufferStorage)->getNativeStorage(); +} + +ID3D11ShaderResourceView *Buffer11::getSRV(DXGI_FORMAT srvFormat) +{ + BufferStorage *storage = getBufferStorage(BUFFER_USAGE_PIXEL_UNPACK); + + if (!storage) + { + // Storage out-of-memory + return NULL; + } + + ASSERT(HAS_DYNAMIC_TYPE(NativeStorage*, storage)); + ID3D11Buffer *buffer = static_cast<NativeStorage*>(storage)->getNativeStorage(); + + auto bufferSRVIt = mBufferResourceViews.find(srvFormat); + + if (bufferSRVIt != mBufferResourceViews.end()) + { + if (bufferSRVIt->second.first == buffer) + { + return bufferSRVIt->second.second; + } + else + { + // The underlying buffer has changed since the SRV was created: recreate the SRV. + SafeRelease(bufferSRVIt->second.second); + } + } + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11ShaderResourceView *bufferSRV = NULL; + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(srvFormat); + + D3D11_SHADER_RESOURCE_VIEW_DESC bufferSRVDesc; + bufferSRVDesc.Buffer.ElementOffset = 0; + bufferSRVDesc.Buffer.ElementWidth = mSize / dxgiFormatInfo.pixelBytes; + bufferSRVDesc.ViewDimension = D3D11_SRV_DIMENSION_BUFFER; + bufferSRVDesc.Format = srvFormat; + + HRESULT result = device->CreateShaderResourceView(buffer, &bufferSRVDesc, &bufferSRV); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + + mBufferResourceViews[srvFormat] = BufferSRVPair(buffer, bufferSRV); + + return bufferSRV; +} + +gl::Error Buffer11::packPixels(ID3D11Texture2D *srcTexture, UINT srcSubresource, const PackPixelsParams ¶ms) +{ + PackStorage *packStorage = getPackStorage(); + BufferStorage *latestStorage = getLatestBufferStorage(); + + if (packStorage) + { + gl::Error error = packStorage->packPixels(srcTexture, srcSubresource, params); + if (error.isError()) + { + return error; + } + packStorage->setDataRevision(latestStorage ? latestStorage->getDataRevision() + 1 : 1); + } + + return gl::Error(GL_NO_ERROR); +} + +Buffer11::BufferStorage *Buffer11::getBufferStorage(BufferUsage usage) +{ + BufferStorage *newStorage = NULL; + auto directBufferIt = mBufferStorages.find(usage); + if (directBufferIt != mBufferStorages.end()) + { + newStorage = directBufferIt->second; + } + + if (!newStorage) + { + if (usage == BUFFER_USAGE_PIXEL_PACK) + { + newStorage = new PackStorage(mRenderer); + } + else if (usage == BUFFER_USAGE_SYSTEM_MEMORY) + { + newStorage = new SystemMemoryStorage(mRenderer); + mHasSystemMemoryStorage = true; + } + else + { + // buffer is not allocated, create it + newStorage = new NativeStorage(mRenderer, usage); + } + + mBufferStorages.insert(std::make_pair(usage, newStorage)); + } + + // resize buffer + if (newStorage->getSize() < mSize) + { + if (newStorage->resize(mSize, true).isError()) + { + // Out of memory error + return NULL; + } + } + + BufferStorage *latestBuffer = getLatestBufferStorage(); + if (latestBuffer && latestBuffer->getDataRevision() > newStorage->getDataRevision()) + { + // Copy through a staging buffer if we're copying from or to a non-staging, mappable + // buffer storage. This is because we can't map a GPU buffer, and copy CPU + // data directly. If we're already using a staging buffer we're fine. + if (latestBuffer->getUsage() != BUFFER_USAGE_STAGING && + newStorage->getUsage() != BUFFER_USAGE_STAGING && + (!latestBuffer->isMappable() || !newStorage->isMappable())) + { + NativeStorage *stagingBuffer = getStagingStorage(); + + stagingBuffer->copyFromStorage(latestBuffer, 0, latestBuffer->getSize(), 0); + stagingBuffer->setDataRevision(latestBuffer->getDataRevision()); + + latestBuffer = stagingBuffer; + } + + // if copyFromStorage returns true, the D3D buffer has been recreated + // and we should update our serial + if (newStorage->copyFromStorage(latestBuffer, 0, latestBuffer->getSize(), 0)) + { + updateSerial(); + } + newStorage->setDataRevision(latestBuffer->getDataRevision()); + } + + return newStorage; +} + +Buffer11::BufferStorage *Buffer11::getLatestBufferStorage() const +{ + // Even though we iterate over all the direct buffers, it is expected that only + // 1 or 2 will be present. + BufferStorage *latestStorage = NULL; + DataRevision latestRevision = 0; + for (auto it = mBufferStorages.begin(); it != mBufferStorages.end(); it++) + { + BufferStorage *storage = it->second; + if (!latestStorage || storage->getDataRevision() > latestRevision) + { + latestStorage = storage; + latestRevision = storage->getDataRevision(); + } + } + + // resize buffer + if (latestStorage && latestStorage->getSize() < mSize) + { + if (latestStorage->resize(mSize, true).isError()) + { + // Out of memory error + return NULL; + } + } + + return latestStorage; +} + +Buffer11::NativeStorage *Buffer11::getStagingStorage() +{ + BufferStorage *stagingStorage = getBufferStorage(BUFFER_USAGE_STAGING); + + if (!stagingStorage) + { + // Out-of-memory + return NULL; + } + + ASSERT(HAS_DYNAMIC_TYPE(NativeStorage*, stagingStorage)); + return static_cast<NativeStorage*>(stagingStorage); +} + +Buffer11::PackStorage *Buffer11::getPackStorage() +{ + BufferStorage *packStorage = getBufferStorage(BUFFER_USAGE_PIXEL_PACK); + + if (!packStorage) + { + // Out-of-memory + return NULL; + } + + ASSERT(HAS_DYNAMIC_TYPE(PackStorage*, packStorage)); + return static_cast<PackStorage*>(packStorage); +} + +bool Buffer11::supportsDirectBinding() const +{ + // Do not support direct buffers for dynamic data. The streaming buffer + // offers better performance for data which changes every frame. + // Check for absence of static buffer interfaces to detect dynamic data. + return (mStaticVertexBuffer && mStaticIndexBuffer); +} + +Buffer11::BufferStorage::BufferStorage(Renderer11 *renderer, BufferUsage usage) + : mRenderer(renderer), + mUsage(usage), + mRevision(0), + mBufferSize(0) +{ +} + +gl::Error Buffer11::BufferStorage::setData(const uint8_t *data, size_t offset, size_t size) +{ + ASSERT(isMappable()); + + uint8_t *writePointer = map(offset, size, GL_MAP_WRITE_BIT); + if (!writePointer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal buffer."); + } + + memcpy(writePointer, data, size); + + unmap(); + + return gl::Error(GL_NO_ERROR); +} + +Buffer11::NativeStorage::NativeStorage(Renderer11 *renderer, BufferUsage usage) + : BufferStorage(renderer, usage), + mNativeStorage(NULL) +{ +} + +Buffer11::NativeStorage::~NativeStorage() +{ + SafeRelease(mNativeStorage); +} + +// Returns true if it recreates the direct buffer +bool Buffer11::NativeStorage::copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) +{ + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + size_t requiredSize = sourceOffset + size; + bool createBuffer = !mNativeStorage || mBufferSize < requiredSize; + + // (Re)initialize D3D buffer if needed + if (createBuffer) + { + bool preserveData = (destOffset > 0); + resize(source->getSize(), preserveData); + } + + if (source->getUsage() == BUFFER_USAGE_PIXEL_PACK || + source->getUsage() == BUFFER_USAGE_SYSTEM_MEMORY) + { + ASSERT(source->isMappable()); + + uint8_t *sourcePointer = source->map(sourceOffset, size, GL_MAP_READ_BIT); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT hr = context->Map(mNativeStorage, 0, D3D11_MAP_WRITE, 0, &mappedResource); + UNUSED_ASSERTION_VARIABLE(hr); + ASSERT(SUCCEEDED(hr)); + + uint8_t *destPointer = static_cast<uint8_t *>(mappedResource.pData) + destOffset; + + // Offset bounds are validated at the API layer + ASSERT(sourceOffset + size <= destOffset + mBufferSize); + memcpy(destPointer, sourcePointer, size); + + context->Unmap(mNativeStorage, 0); + source->unmap(); + } + else + { + ASSERT(HAS_DYNAMIC_TYPE(NativeStorage*, source)); + + D3D11_BOX srcBox; + srcBox.left = sourceOffset; + srcBox.right = sourceOffset + size; + srcBox.top = 0; + srcBox.bottom = 1; + srcBox.front = 0; + srcBox.back = 1; + + ASSERT(HAS_DYNAMIC_TYPE(NativeStorage*, source)); + ID3D11Buffer *sourceBuffer = static_cast<NativeStorage*>(source)->getNativeStorage(); + + context->CopySubresourceRegion(mNativeStorage, 0, destOffset, 0, 0, sourceBuffer, 0, &srcBox); + } + + return createBuffer; +} + +gl::Error Buffer11::NativeStorage::resize(size_t size, bool preserveData) +{ + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + D3D11_BUFFER_DESC bufferDesc; + fillBufferDesc(&bufferDesc, mRenderer, mUsage, size); + + ID3D11Buffer *newBuffer; + HRESULT result = device->CreateBuffer(&bufferDesc, NULL, &newBuffer); + + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal buffer, result: 0x%X.", result); + } + + if (mNativeStorage && preserveData) + { + // We don't call resize if the buffer is big enough already. + ASSERT(mBufferSize <= size); + + D3D11_BOX srcBox; + srcBox.left = 0; + srcBox.right = mBufferSize; + srcBox.top = 0; + srcBox.bottom = 1; + srcBox.front = 0; + srcBox.back = 1; + + context->CopySubresourceRegion(newBuffer, 0, 0, 0, 0, mNativeStorage, 0, &srcBox); + } + + // No longer need the old buffer + SafeRelease(mNativeStorage); + mNativeStorage = newBuffer; + + mBufferSize = bufferDesc.ByteWidth; + + return gl::Error(GL_NO_ERROR); +} + +void Buffer11::NativeStorage::fillBufferDesc(D3D11_BUFFER_DESC* bufferDesc, Renderer11 *renderer, + BufferUsage usage, unsigned int bufferSize) +{ + bufferDesc->ByteWidth = bufferSize; + bufferDesc->MiscFlags = 0; + bufferDesc->StructureByteStride = 0; + + switch (usage) + { + case BUFFER_USAGE_STAGING: + bufferDesc->Usage = D3D11_USAGE_STAGING; + bufferDesc->BindFlags = 0; + bufferDesc->CPUAccessFlags = D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE; + break; + + case BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK: + bufferDesc->Usage = D3D11_USAGE_DEFAULT; + bufferDesc->BindFlags = D3D11_BIND_VERTEX_BUFFER; + + if (renderer->isES3Capable()) + { + bufferDesc->BindFlags |= D3D11_BIND_STREAM_OUTPUT; + } + + bufferDesc->CPUAccessFlags = 0; + break; + + case BUFFER_USAGE_INDEX: + bufferDesc->Usage = D3D11_USAGE_DEFAULT; + bufferDesc->BindFlags = D3D11_BIND_INDEX_BUFFER; + bufferDesc->CPUAccessFlags = 0; + break; + + case BUFFER_USAGE_PIXEL_UNPACK: + bufferDesc->Usage = D3D11_USAGE_DEFAULT; + bufferDesc->BindFlags = D3D11_BIND_SHADER_RESOURCE; + bufferDesc->CPUAccessFlags = 0; + break; + + case BUFFER_USAGE_UNIFORM: + bufferDesc->Usage = D3D11_USAGE_DYNAMIC; + bufferDesc->BindFlags = D3D11_BIND_CONSTANT_BUFFER; + bufferDesc->CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + + // Constant buffers must be of a limited size, and aligned to 16 byte boundaries + // For our purposes we ignore any buffer data past the maximum constant buffer size + bufferDesc->ByteWidth = roundUp(bufferDesc->ByteWidth, 16u); + bufferDesc->ByteWidth = std::min<UINT>(bufferDesc->ByteWidth, renderer->getRendererCaps().maxUniformBlockSize); + break; + + default: + UNREACHABLE(); + } +} + +uint8_t *Buffer11::NativeStorage::map(size_t offset, size_t length, GLbitfield access) +{ + ASSERT(mUsage == BUFFER_USAGE_STAGING); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + D3D11_MAP d3dMapType = gl_d3d11::GetD3DMapTypeFromBits(access); + UINT d3dMapFlag = ((access & GL_MAP_UNSYNCHRONIZED_BIT) != 0 ? D3D11_MAP_FLAG_DO_NOT_WAIT : 0); + + HRESULT result = context->Map(mNativeStorage, 0, d3dMapType, d3dMapFlag, &mappedResource); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + + return static_cast<uint8_t*>(mappedResource.pData) + offset; +} + +void Buffer11::NativeStorage::unmap() +{ + ASSERT(mUsage == BUFFER_USAGE_STAGING); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + context->Unmap(mNativeStorage, 0); +} + +Buffer11::PackStorage::PackStorage(Renderer11 *renderer) + : BufferStorage(renderer, BUFFER_USAGE_PIXEL_PACK), + mStagingTexture(NULL), + mTextureFormat(DXGI_FORMAT_UNKNOWN), + mQueuedPackCommand(NULL), + mDataModified(false) +{ +} + +Buffer11::PackStorage::~PackStorage() +{ + SafeRelease(mStagingTexture); + SafeDelete(mQueuedPackCommand); +} + +bool Buffer11::PackStorage::copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) +{ + // We copy through a staging buffer when drawing with a pack buffer, + // or for other cases where we access the pack buffer + UNREACHABLE(); + return false; +} + +gl::Error Buffer11::PackStorage::resize(size_t size, bool preserveData) +{ + if (size != mBufferSize) + { + if (!mMemoryBuffer.resize(size)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to resize internal buffer storage."); + } + mBufferSize = size; + } + + return gl::Error(GL_NO_ERROR); +} + +uint8_t *Buffer11::PackStorage::map(size_t offset, size_t length, GLbitfield access) +{ + ASSERT(offset + length <= getSize()); + // TODO: fast path + // We might be able to optimize out one or more memcpy calls by detecting when + // and if D3D packs the staging texture memory identically to how we would fill + // the pack buffer according to the current pack state. + + gl::Error error = flushQueuedPackCommand(); + if (error.isError()) + { + return NULL; + } + + mDataModified = (mDataModified || (access & GL_MAP_WRITE_BIT) != 0); + + return mMemoryBuffer.data() + offset; +} + +void Buffer11::PackStorage::unmap() +{ + // No-op +} + +gl::Error Buffer11::PackStorage::packPixels(ID3D11Texture2D *srcTexure, UINT srcSubresource, const PackPixelsParams ¶ms) +{ + gl::Error error = flushQueuedPackCommand(); + if (error.isError()) + { + return error; + } + + mQueuedPackCommand = new PackPixelsParams(params); + + D3D11_TEXTURE2D_DESC textureDesc; + srcTexure->GetDesc(&textureDesc); + + if (mStagingTexture != NULL && + (mTextureFormat != textureDesc.Format || + mTextureSize.width != params.area.width || + mTextureSize.height != params.area.height)) + { + SafeRelease(mStagingTexture); + mTextureSize.width = 0; + mTextureSize.height = 0; + mTextureFormat = DXGI_FORMAT_UNKNOWN; + } + + if (mStagingTexture == NULL) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT hr; + + mTextureSize.width = params.area.width; + mTextureSize.height = params.area.height; + mTextureFormat = textureDesc.Format; + + D3D11_TEXTURE2D_DESC stagingDesc; + stagingDesc.Width = params.area.width; + stagingDesc.Height = params.area.height; + stagingDesc.MipLevels = 1; + stagingDesc.ArraySize = 1; + stagingDesc.Format = mTextureFormat; + stagingDesc.SampleDesc.Count = 1; + stagingDesc.SampleDesc.Quality = 0; + stagingDesc.Usage = D3D11_USAGE_STAGING; + stagingDesc.BindFlags = 0; + stagingDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ; + stagingDesc.MiscFlags = 0; + + hr = device->CreateTexture2D(&stagingDesc, NULL, &mStagingTexture); + if (FAILED(hr)) + { + ASSERT(hr == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate internal staging texture."); + } + } + + // ReadPixels from multisampled FBOs isn't supported in current GL + ASSERT(textureDesc.SampleDesc.Count <= 1); + + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + D3D11_BOX srcBox; + srcBox.left = params.area.x; + srcBox.right = params.area.x + params.area.width; + srcBox.top = params.area.y; + srcBox.bottom = params.area.y + params.area.height; + srcBox.front = 0; + srcBox.back = 1; + + // Asynchronous copy + immediateContext->CopySubresourceRegion(mStagingTexture, 0, 0, 0, 0, srcTexure, srcSubresource, &srcBox); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Buffer11::PackStorage::flushQueuedPackCommand() +{ + ASSERT(mMemoryBuffer.size() > 0); + + if (mQueuedPackCommand) + { + gl::Error error = mRenderer->packPixels(mStagingTexture, *mQueuedPackCommand, mMemoryBuffer.data()); + SafeDelete(mQueuedPackCommand); + if (error.isError()) + { + return error; + } + } + + return gl::Error(GL_NO_ERROR); +} + +Buffer11::SystemMemoryStorage::SystemMemoryStorage(Renderer11 *renderer) + : Buffer11::BufferStorage(renderer, BUFFER_USAGE_SYSTEM_MEMORY) +{} + +bool Buffer11::SystemMemoryStorage::copyFromStorage(BufferStorage *source, size_t sourceOffset, + size_t size, size_t destOffset) +{ + ASSERT(source->isMappable()); + const uint8_t *sourceData = source->map(sourceOffset, size, GL_MAP_READ_BIT); + ASSERT(destOffset + size <= mSystemCopy.size()); + memcpy(mSystemCopy.data() + destOffset, sourceData, size); + source->unmap(); + return true; +} + +gl::Error Buffer11::SystemMemoryStorage::resize(size_t size, bool preserveData) +{ + if (mSystemCopy.size() < size) + { + if (!mSystemCopy.resize(size)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to resize SystemMemoryStorage"); + } + mBufferSize = size; + } + + return gl::Error(GL_NO_ERROR); +} + +uint8_t *Buffer11::SystemMemoryStorage::map(size_t offset, size_t length, GLbitfield access) +{ + ASSERT(!mSystemCopy.empty() && offset + length <= mSystemCopy.size()); + return mSystemCopy.data() + offset; +} + +void Buffer11::SystemMemoryStorage::unmap() +{ + // No-op +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.h new file mode 100644 index 0000000000..39bafe880e --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Buffer11.h @@ -0,0 +1,103 @@ +// +// Copyright 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Buffer11.h: Defines the rx::Buffer11 class which implements rx::BufferImpl via rx::BufferD3D. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_BUFFER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_BUFFER11_H_ + +#include "libANGLE/angletypes.h" +#include "libANGLE/renderer/d3d/BufferD3D.h" + +namespace rx +{ +class Renderer11; + +enum BufferUsage +{ + BUFFER_USAGE_STAGING, + BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK, + BUFFER_USAGE_INDEX, + BUFFER_USAGE_PIXEL_UNPACK, + BUFFER_USAGE_PIXEL_PACK, + BUFFER_USAGE_UNIFORM, + BUFFER_USAGE_SYSTEM_MEMORY, +}; + +struct PackPixelsParams +{ + PackPixelsParams(); + PackPixelsParams(const gl::Rectangle &area, GLenum format, GLenum type, GLuint outputPitch, + const gl::PixelPackState &pack, ptrdiff_t offset); + + gl::Rectangle area; + GLenum format; + GLenum type; + GLuint outputPitch; + gl::Buffer *packBuffer; + gl::PixelPackState pack; + ptrdiff_t offset; +}; + +typedef size_t DataRevision; + +class Buffer11 : public BufferD3D +{ + public: + Buffer11(Renderer11 *renderer); + virtual ~Buffer11(); + + static Buffer11 *makeBuffer11(BufferImpl *buffer); + + ID3D11Buffer *getBuffer(BufferUsage usage); + ID3D11ShaderResourceView *getSRV(DXGI_FORMAT srvFormat); + bool isMapped() const { return mMappedStorage != NULL; } + gl::Error packPixels(ID3D11Texture2D *srcTexure, UINT srcSubresource, const PackPixelsParams ¶ms); + + // BufferD3D implementation + virtual size_t getSize() const { return mSize; } + virtual bool supportsDirectBinding() const; + + // BufferImpl implementation + virtual gl::Error setData(const void* data, size_t size, GLenum usage); + gl::Error getData(const uint8_t **outData) override; + virtual gl::Error setSubData(const void* data, size_t size, size_t offset); + virtual gl::Error copySubData(BufferImpl* source, GLintptr sourceOffset, GLintptr destOffset, GLsizeiptr size); + virtual gl::Error map(size_t offset, size_t length, GLbitfield access, GLvoid **mapPtr); + virtual gl::Error unmap(); + virtual void markTransformFeedbackUsage(); + + private: + class BufferStorage; + class NativeStorage; + class PackStorage; + class SystemMemoryStorage; + + Renderer11 *mRenderer; + size_t mSize; + + BufferStorage *mMappedStorage; + + std::map<BufferUsage, BufferStorage*> mBufferStorages; + + typedef std::pair<ID3D11Buffer *, ID3D11ShaderResourceView *> BufferSRVPair; + std::map<DXGI_FORMAT, BufferSRVPair> mBufferResourceViews; + + unsigned int mReadUsageCount; + bool mHasSystemMemoryStorage; + + void markBufferUsage(); + NativeStorage *getStagingStorage(); + PackStorage *getPackStorage(); + gl::Error getSystemMemoryStorage(SystemMemoryStorage **storageOut); + + BufferStorage *getBufferStorage(BufferUsage usage); + BufferStorage *getLatestBufferStorage() const; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_BUFFER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.cpp new file mode 100644 index 0000000000..057c3bed42 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.cpp @@ -0,0 +1,614 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Clear11.cpp: Framebuffer clear utility class. + +#include "libANGLE/renderer/d3d/d3d11/Clear11.h" + +#include <algorithm> + +#include "libANGLE/FramebufferAttachment.h" +#include "libANGLE/formatutils.h" +#include "libANGLE/renderer/d3d/FramebufferD3D.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" + +// Precompiled shaders +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearfloat11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearfloat11ps.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearfloat11_fl9ps.h" + +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearuint11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearuint11ps.h" + +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearsint11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/clearsint11ps.h" + +namespace rx +{ + +template <typename T> +static void ApplyVertices(const gl::Extents &framebufferSize, const gl::Rectangle *scissor, const gl::Color<T> &color, float depth, void *buffer) +{ + d3d11::PositionDepthColorVertex<T> *vertices = reinterpret_cast<d3d11::PositionDepthColorVertex<T>*>(buffer); + + float depthClear = gl::clamp01(depth); + float left = -1.0f; + float right = 1.0f; + float top = -1.0f; + float bottom = 1.0f; + + // Clip the quad coordinates to the scissor if needed + if (scissor != NULL) + { + left = std::max(left, (scissor->x / float(framebufferSize.width)) * 2.0f - 1.0f); + right = std::min(right, ((scissor->x + scissor->width) / float(framebufferSize.width)) * 2.0f - 1.0f); + top = std::max(top, ((framebufferSize.height - scissor->y - scissor->height) / float(framebufferSize.height)) * 2.0f - 1.0f); + bottom = std::min(bottom, ((framebufferSize.height - scissor->y) / float(framebufferSize.height)) * 2.0f - 1.0f); + } + + d3d11::SetPositionDepthColorVertex<T>(vertices + 0, left, bottom, depthClear, color); + d3d11::SetPositionDepthColorVertex<T>(vertices + 1, left, top, depthClear, color); + d3d11::SetPositionDepthColorVertex<T>(vertices + 2, right, bottom, depthClear, color); + d3d11::SetPositionDepthColorVertex<T>(vertices + 3, right, top, depthClear, color); +} + +template <unsigned int vsSize, unsigned int psSize> +Clear11::ClearShader Clear11::CreateClearShader(ID3D11Device *device, DXGI_FORMAT colorType, const BYTE (&vsByteCode)[vsSize], const BYTE (&psByteCode)[psSize]) +{ + HRESULT result; + + ClearShader shader = { 0 }; + + D3D11_INPUT_ELEMENT_DESC quadLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32B32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "COLOR", 0, colorType, 0, 12, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = device->CreateInputLayout(quadLayout, ArraySize(quadLayout), vsByteCode, vsSize, &shader.inputLayout); + ASSERT(SUCCEEDED(result)); + + result = device->CreateVertexShader(vsByteCode, vsSize, NULL, &shader.vertexShader); + ASSERT(SUCCEEDED(result)); + + result = device->CreatePixelShader(psByteCode, psSize, NULL, &shader.pixelShader); + ASSERT(SUCCEEDED(result)); + + return shader; +} + +Clear11::Clear11(Renderer11 *renderer) + : mRenderer(renderer), mClearBlendStates(StructLessThan<ClearBlendInfo>), mClearDepthStencilStates(StructLessThan<ClearDepthStencilInfo>), + mVertexBuffer(NULL), mRasterizerState(NULL), mSupportsClearView(false) +{ + HRESULT result; + ID3D11Device *device = renderer->getDevice(); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = sizeof(d3d11::PositionDepthColorVertex<float>) * 4; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + result = device->CreateBuffer(&vbDesc, NULL, &mVertexBuffer); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mVertexBuffer, "Clear11 masked clear vertex buffer"); + + D3D11_RASTERIZER_DESC rsDesc; + rsDesc.FillMode = D3D11_FILL_SOLID; + rsDesc.CullMode = D3D11_CULL_NONE; + rsDesc.FrontCounterClockwise = FALSE; + rsDesc.DepthBias = 0; + rsDesc.DepthBiasClamp = 0.0f; + rsDesc.SlopeScaledDepthBias = 0.0f; + rsDesc.DepthClipEnable = TRUE; + rsDesc.ScissorEnable = FALSE; + rsDesc.MultisampleEnable = FALSE; + rsDesc.AntialiasedLineEnable = FALSE; + + result = device->CreateRasterizerState(&rsDesc, &mRasterizerState); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mRasterizerState, "Clear11 masked clear rasterizer state"); + + if (renderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3) + { + mFloatClearShader = CreateClearShader(device, DXGI_FORMAT_R32G32B32A32_FLOAT, g_VS_ClearFloat, g_PS_ClearFloat_FL9); + } + else + { + mFloatClearShader = CreateClearShader(device, DXGI_FORMAT_R32G32B32A32_FLOAT, g_VS_ClearFloat, g_PS_ClearFloat); + } + + if (renderer->isES3Capable()) + { + mUintClearShader = CreateClearShader(device, DXGI_FORMAT_R32G32B32A32_UINT, g_VS_ClearUint, g_PS_ClearUint ); + mIntClearShader = CreateClearShader(device, DXGI_FORMAT_R32G32B32A32_SINT, g_VS_ClearSint, g_PS_ClearSint ); + } + +#if defined(ANGLE_ENABLE_D3D11_1) + if (renderer->getDeviceContext1IfSupported()) + { + D3D11_FEATURE_DATA_D3D11_OPTIONS d3d11Options; + device->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &d3d11Options, sizeof(D3D11_FEATURE_DATA_D3D11_OPTIONS)); + mSupportsClearView = (d3d11Options.ClearView != FALSE); + } +#endif +} + +Clear11::~Clear11() +{ + for (ClearBlendStateMap::iterator i = mClearBlendStates.begin(); i != mClearBlendStates.end(); i++) + { + SafeRelease(i->second); + } + mClearBlendStates.clear(); + + SafeRelease(mFloatClearShader.inputLayout); + SafeRelease(mFloatClearShader.vertexShader); + SafeRelease(mFloatClearShader.pixelShader); + + if (mRenderer->isES3Capable()) + { + SafeRelease(mUintClearShader.inputLayout); + SafeRelease(mUintClearShader.vertexShader); + SafeRelease(mUintClearShader.pixelShader); + + SafeRelease(mIntClearShader.inputLayout); + SafeRelease(mIntClearShader.vertexShader); + SafeRelease(mIntClearShader.pixelShader); + } + + for (ClearDepthStencilStateMap::iterator i = mClearDepthStencilStates.begin(); i != mClearDepthStencilStates.end(); i++) + { + SafeRelease(i->second); + } + mClearDepthStencilStates.clear(); + + SafeRelease(mVertexBuffer); + SafeRelease(mRasterizerState); +} + +gl::Error Clear11::clearFramebuffer(const ClearParameters &clearParams, const gl::Framebuffer::Data &fboData) +{ + const auto &colorAttachments = fboData.mColorAttachments; + const auto &drawBufferStates = fboData.mDrawBufferStates; + const auto *depthAttachment = fboData.mDepthAttachment; + const auto *stencilAttachment = fboData.mStencilAttachment; + + ASSERT(colorAttachments.size() == drawBufferStates.size()); + + // Iterate over the color buffers which require clearing and determine if they can be + // cleared with ID3D11DeviceContext::ClearRenderTargetView or ID3D11DeviceContext1::ClearView. + // This requires: + // 1) The render target is being cleared to a float value (will be cast to integer when clearing integer + // render targets as expected but does not work the other way around) + // 2) The format of the render target has no color channels that are currently masked out. + // Clear the easy-to-clear buffers on the spot and accumulate the ones that require special work. + // + // If these conditions are met, and: + // - No scissored clear is needed, then clear using ID3D11DeviceContext::ClearRenderTargetView. + // - A scissored clear is needed then clear using ID3D11DeviceContext1::ClearView if available. + // Otherwise draw a quad. + // + // Also determine if the depth stencil can be cleared with ID3D11DeviceContext::ClearDepthStencilView + // by checking if the stencil write mask covers the entire stencil. + // + // To clear the remaining buffers, quads must be drawn containing an int, uint or float vertex color + // attribute. + + gl::Extents framebufferSize; + + auto iter = std::find_if(colorAttachments.begin(), colorAttachments.end(), [](const gl::FramebufferAttachment *attachment) { return attachment != nullptr; }); + if (iter != colorAttachments.end()) + { + framebufferSize.width = (*iter)->getWidth(); + framebufferSize.height = (*iter)->getHeight(); + framebufferSize.depth = 1; + } + else if (depthAttachment != nullptr) + { + framebufferSize.width = depthAttachment->getWidth(); + framebufferSize.height = depthAttachment->getHeight(); + framebufferSize.depth = 1; + } + else if (stencilAttachment != nullptr) + { + framebufferSize.width = stencilAttachment->getWidth(); + framebufferSize.height = stencilAttachment->getHeight(); + framebufferSize.depth = 1; + } + else + { + UNREACHABLE(); + return gl::Error(GL_INVALID_OPERATION); + } + + if (clearParams.scissorEnabled && (clearParams.scissor.x >= framebufferSize.width || + clearParams.scissor.y >= framebufferSize.height || + clearParams.scissor.x + clearParams.scissor.width <= 0 || + clearParams.scissor.y + clearParams.scissor.height <= 0)) + { + // Scissor is enabled and the scissor rectangle is outside the renderbuffer + return gl::Error(GL_NO_ERROR); + } + + bool needScissoredClear = clearParams.scissorEnabled && (clearParams.scissor.x > 0 || clearParams.scissor.y > 0 || + clearParams.scissor.x + clearParams.scissor.width < framebufferSize.width || + clearParams.scissor.y + clearParams.scissor.height < framebufferSize.height); + + std::vector<MaskedRenderTarget> maskedClearRenderTargets; + RenderTarget11* maskedClearDepthStencil = NULL; + + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + ID3D11DeviceContext1 *deviceContext1 = mRenderer->getDeviceContext1IfSupported(); + + for (size_t colorAttachment = 0; colorAttachment < colorAttachments.size(); colorAttachment++) + { + if (clearParams.clearColor[colorAttachment] && + colorAttachments[colorAttachment] != nullptr && + drawBufferStates[colorAttachment] != GL_NONE) + { + const gl::FramebufferAttachment *attachment = colorAttachments[colorAttachment]; + + RenderTarget11 *renderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(attachment, &renderTarget); + if (error.isError()) + { + return error; + } + + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(attachment->getInternalFormat()); + + if (clearParams.colorClearType == GL_FLOAT && + !(formatInfo.componentType == GL_FLOAT || formatInfo.componentType == GL_UNSIGNED_NORMALIZED || formatInfo.componentType == GL_SIGNED_NORMALIZED)) + { + ERR("It is undefined behaviour to clear a render buffer which is not normalized fixed point or floating-" + "point to floating point values (color attachment %u has internal format 0x%X).", colorAttachment, + attachment->getInternalFormat()); + } + + if ((formatInfo.redBits == 0 || !clearParams.colorMaskRed) && + (formatInfo.greenBits == 0 || !clearParams.colorMaskGreen) && + (formatInfo.blueBits == 0 || !clearParams.colorMaskBlue) && + (formatInfo.alphaBits == 0 || !clearParams.colorMaskAlpha)) + { + // Every channel either does not exist in the render target or is masked out + continue; + } + else if ((!mSupportsClearView && needScissoredClear) || clearParams.colorClearType != GL_FLOAT || + (formatInfo.redBits > 0 && !clearParams.colorMaskRed) || + (formatInfo.greenBits > 0 && !clearParams.colorMaskGreen) || + (formatInfo.blueBits > 0 && !clearParams.colorMaskBlue) || + (formatInfo.alphaBits > 0 && !clearParams.colorMaskAlpha)) + { + // A masked clear is required, or a scissored clear is required and ID3D11DeviceContext1::ClearView is unavailable + MaskedRenderTarget maskAndRt; + bool clearColor = clearParams.clearColor[colorAttachment]; + maskAndRt.colorMask[0] = (clearColor && clearParams.colorMaskRed); + maskAndRt.colorMask[1] = (clearColor && clearParams.colorMaskGreen); + maskAndRt.colorMask[2] = (clearColor && clearParams.colorMaskBlue); + maskAndRt.colorMask[3] = (clearColor && clearParams.colorMaskAlpha); + maskAndRt.renderTarget = renderTarget; + maskedClearRenderTargets.push_back(maskAndRt); + } + else + { + // ID3D11DeviceContext::ClearRenderTargetView or ID3D11DeviceContext1::ClearView is possible + + ID3D11RenderTargetView *framebufferRTV = renderTarget->getRenderTargetView(); + if (!framebufferRTV) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal render target view pointer unexpectedly null."); + } + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(renderTarget->getDXGIFormat()); + + // Check if the actual format has a channel that the internal format does not and set them to the + // default values + const float clearValues[4] = + { + ((formatInfo.redBits == 0 && dxgiFormatInfo.redBits > 0) ? 0.0f : clearParams.colorFClearValue.red), + ((formatInfo.greenBits == 0 && dxgiFormatInfo.greenBits > 0) ? 0.0f : clearParams.colorFClearValue.green), + ((formatInfo.blueBits == 0 && dxgiFormatInfo.blueBits > 0) ? 0.0f : clearParams.colorFClearValue.blue), + ((formatInfo.alphaBits == 0 && dxgiFormatInfo.alphaBits > 0) ? 1.0f : clearParams.colorFClearValue.alpha), + }; + + if (needScissoredClear) + { +#if defined(ANGLE_ENABLE_D3D11_1) + // We shouldn't reach here if deviceContext1 is unavailable. + ASSERT(deviceContext1); + + D3D11_RECT rect; + rect.left = clearParams.scissor.x; + rect.right = clearParams.scissor.x + clearParams.scissor.width; + rect.top = clearParams.scissor.y; + rect.bottom = clearParams.scissor.y + clearParams.scissor.height; + + deviceContext1->ClearView(framebufferRTV, clearValues, &rect, 1); +#endif + } + else + { + deviceContext->ClearRenderTargetView(framebufferRTV, clearValues); + } + } + } + } + + if (clearParams.clearDepth || clearParams.clearStencil) + { + const gl::FramebufferAttachment *attachment = (depthAttachment != nullptr) ? depthAttachment : stencilAttachment; + ASSERT(attachment != nullptr); + + RenderTarget11 *renderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(attachment, &renderTarget); + if (error.isError()) + { + return error; + } + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(renderTarget->getDXGIFormat()); + + unsigned int stencilUnmasked = (stencilAttachment != nullptr) ? (1 << dxgiFormatInfo.stencilBits) - 1 : 0; + bool needMaskedStencilClear = clearParams.clearStencil && (clearParams.stencilWriteMask & stencilUnmasked) != stencilUnmasked; + + if (needScissoredClear || needMaskedStencilClear) + { + maskedClearDepthStencil = renderTarget; + } + else + { + ID3D11DepthStencilView *framebufferDSV = renderTarget->getDepthStencilView(); + if (!framebufferDSV) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal depth stencil view pointer unexpectedly null."); + } + + UINT clearFlags = (clearParams.clearDepth ? D3D11_CLEAR_DEPTH : 0) | + (clearParams.clearStencil ? D3D11_CLEAR_STENCIL : 0); + FLOAT depthClear = gl::clamp01(clearParams.depthClearValue); + UINT8 stencilClear = clearParams.stencilClearValue & 0xFF; + + deviceContext->ClearDepthStencilView(framebufferDSV, clearFlags, depthClear, stencilClear); + } + } + + if (maskedClearRenderTargets.size() > 0 || maskedClearDepthStencil) + { + // To clear the render targets and depth stencil in one pass: + // + // Render a quad clipped to the scissor rectangle which draws the clear color and a blend + // state that will perform the required color masking. + // + // The quad's depth is equal to the depth clear value with a depth stencil state that + // will enable or disable depth test/writes if the depth buffer should be cleared or not. + // + // The rasterizer state's stencil is set to always pass or fail based on if the stencil + // should be cleared or not with a stencil write mask of the stencil clear value. + // + // ====================================================================================== + // + // Luckily, the gl spec (ES 3.0.2 pg 183) states that the results of clearing a render- + // buffer that is not normalized fixed point or floating point with floating point values + // are undefined so we can just write floats to them and D3D11 will bit cast them to + // integers. + // + // Also, we don't have to worry about attempting to clear a normalized fixed/floating point + // buffer with integer values because there is no gl API call which would allow it, + // glClearBuffer* calls only clear a single renderbuffer at a time which is verified to + // be a compatible clear type. + + // Bind all the render targets which need clearing + ASSERT(maskedClearRenderTargets.size() <= mRenderer->getRendererCaps().maxDrawBuffers); + std::vector<ID3D11RenderTargetView*> rtvs(maskedClearRenderTargets.size()); + for (unsigned int i = 0; i < maskedClearRenderTargets.size(); i++) + { + RenderTarget11 *renderTarget = maskedClearRenderTargets[i].renderTarget; + ID3D11RenderTargetView *rtv = renderTarget->getRenderTargetView(); + if (!rtv) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal render target view pointer unexpectedly null."); + } + + rtvs[i] = rtv; + } + ID3D11DepthStencilView *dsv = maskedClearDepthStencil ? maskedClearDepthStencil->getDepthStencilView() : NULL; + + ID3D11BlendState *blendState = getBlendState(maskedClearRenderTargets); + const FLOAT blendFactors[4] = { 1.0f, 1.0f, 1.0f, 1.0f }; + const UINT sampleMask = 0xFFFFFFFF; + + ID3D11DepthStencilState *dsState = getDepthStencilState(clearParams); + const UINT stencilClear = clearParams.stencilClearValue & 0xFF; + + // Set the vertices + UINT vertexStride = 0; + const UINT startIdx = 0; + const ClearShader* shader = NULL; + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = deviceContext->Map(mVertexBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal masked clear vertex buffer, HRESULT: 0x%X.", result); + } + + const gl::Rectangle *scissorPtr = clearParams.scissorEnabled ? &clearParams.scissor : NULL; + switch (clearParams.colorClearType) + { + case GL_FLOAT: + ApplyVertices(framebufferSize, scissorPtr, clearParams.colorFClearValue, clearParams.depthClearValue, mappedResource.pData); + vertexStride = sizeof(d3d11::PositionDepthColorVertex<float>); + shader = &mFloatClearShader; + break; + + case GL_UNSIGNED_INT: + ApplyVertices(framebufferSize, scissorPtr, clearParams.colorUIClearValue, clearParams.depthClearValue, mappedResource.pData); + vertexStride = sizeof(d3d11::PositionDepthColorVertex<unsigned int>); + shader = &mUintClearShader; + break; + + case GL_INT: + ApplyVertices(framebufferSize, scissorPtr, clearParams.colorIClearValue, clearParams.depthClearValue, mappedResource.pData); + vertexStride = sizeof(d3d11::PositionDepthColorVertex<int>); + shader = &mIntClearShader; + break; + + default: + UNREACHABLE(); + break; + } + + deviceContext->Unmap(mVertexBuffer, 0); + + // Set the viewport to be the same size as the framebuffer + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = framebufferSize.width; + viewport.Height = framebufferSize.height; + viewport.MinDepth = 0; + viewport.MaxDepth = 1; + deviceContext->RSSetViewports(1, &viewport); + + // Apply state + deviceContext->OMSetBlendState(blendState, blendFactors, sampleMask); + deviceContext->OMSetDepthStencilState(dsState, stencilClear); + deviceContext->RSSetState(mRasterizerState); + + // Apply shaders + deviceContext->IASetInputLayout(shader->inputLayout); + deviceContext->VSSetShader(shader->vertexShader, NULL, 0); + deviceContext->PSSetShader(shader->pixelShader, NULL, 0); + deviceContext->GSSetShader(NULL, NULL, 0); + + // Apply vertex buffer + deviceContext->IASetVertexBuffers(0, 1, &mVertexBuffer, &vertexStride, &startIdx); + deviceContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + + // Apply render targets + deviceContext->OMSetRenderTargets(rtvs.size(), (rtvs.empty() ? NULL : &rtvs[0]), dsv); + + // Draw the clear quad + deviceContext->Draw(4, 0); + + // Clean up + mRenderer->markAllStateDirty(); + } + + return gl::Error(GL_NO_ERROR); +} + +ID3D11BlendState *Clear11::getBlendState(const std::vector<MaskedRenderTarget>& rts) +{ + ClearBlendInfo blendKey = { 0 }; + for (unsigned int i = 0; i < D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; i++) + { + if (i < rts.size()) + { + RenderTarget11 *rt = rts[i].renderTarget; + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(rt->getInternalFormat()); + + blendKey.maskChannels[i][0] = (rts[i].colorMask[0] && formatInfo.redBits > 0); + blendKey.maskChannels[i][1] = (rts[i].colorMask[1] && formatInfo.greenBits > 0); + blendKey.maskChannels[i][2] = (rts[i].colorMask[2] && formatInfo.blueBits > 0); + blendKey.maskChannels[i][3] = (rts[i].colorMask[3] && formatInfo.alphaBits > 0); + } + else + { + blendKey.maskChannels[i][0] = false; + blendKey.maskChannels[i][1] = false; + blendKey.maskChannels[i][2] = false; + blendKey.maskChannels[i][3] = false; + } + } + + ClearBlendStateMap::const_iterator i = mClearBlendStates.find(blendKey); + if (i != mClearBlendStates.end()) + { + return i->second; + } + else + { + D3D11_BLEND_DESC blendDesc = { 0 }; + blendDesc.AlphaToCoverageEnable = FALSE; + blendDesc.IndependentBlendEnable = (rts.size() > 1) ? TRUE : FALSE; + + for (unsigned int j = 0; j < D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; j++) + { + blendDesc.RenderTarget[j].BlendEnable = FALSE; + blendDesc.RenderTarget[j].RenderTargetWriteMask = gl_d3d11::ConvertColorMask(blendKey.maskChannels[j][0], + blendKey.maskChannels[j][1], + blendKey.maskChannels[j][2], + blendKey.maskChannels[j][3]); + } + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11BlendState* blendState = NULL; + HRESULT result = device->CreateBlendState(&blendDesc, &blendState); + if (FAILED(result) || !blendState) + { + ERR("Unable to create a ID3D11BlendState, HRESULT: 0x%X.", result); + return NULL; + } + + mClearBlendStates[blendKey] = blendState; + + return blendState; + } +} + +ID3D11DepthStencilState *Clear11::getDepthStencilState(const ClearParameters &clearParams) +{ + ClearDepthStencilInfo dsKey = { 0 }; + dsKey.clearDepth = clearParams.clearDepth; + dsKey.clearStencil = clearParams.clearStencil; + dsKey.stencilWriteMask = clearParams.stencilWriteMask & 0xFF; + + ClearDepthStencilStateMap::const_iterator i = mClearDepthStencilStates.find(dsKey); + if (i != mClearDepthStencilStates.end()) + { + return i->second; + } + else + { + D3D11_DEPTH_STENCIL_DESC dsDesc = { 0 }; + dsDesc.DepthEnable = dsKey.clearDepth ? TRUE : FALSE; + dsDesc.DepthWriteMask = dsKey.clearDepth ? D3D11_DEPTH_WRITE_MASK_ALL : D3D11_DEPTH_WRITE_MASK_ZERO; + dsDesc.DepthFunc = D3D11_COMPARISON_ALWAYS; + dsDesc.StencilEnable = dsKey.clearStencil ? TRUE : FALSE; + dsDesc.StencilReadMask = 0; + dsDesc.StencilWriteMask = dsKey.stencilWriteMask; + dsDesc.FrontFace.StencilFailOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.FrontFace.StencilDepthFailOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.FrontFace.StencilPassOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.FrontFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + dsDesc.BackFace.StencilFailOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.BackFace.StencilDepthFailOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.BackFace.StencilPassOp = D3D11_STENCIL_OP_REPLACE; + dsDesc.BackFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DepthStencilState* dsState = NULL; + HRESULT result = device->CreateDepthStencilState(&dsDesc, &dsState); + if (FAILED(result) || !dsState) + { + ERR("Unable to create a ID3D11DepthStencilState, HRESULT: 0x%X.", result); + return NULL; + } + + mClearDepthStencilStates[dsKey] = dsState; + + return dsState; + } +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.h new file mode 100644 index 0000000000..4797ca1aa0 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Clear11.h @@ -0,0 +1,86 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Clear11.h: Framebuffer clear utility class. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_CLEAR11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_CLEAR11_H_ + +#include <map> +#include <vector> + +#include "libANGLE/angletypes.h" +#include "libANGLE/Error.h" +#include "libANGLE/Framebuffer.h" + +namespace rx +{ +class Renderer11; +class RenderTarget11; +struct ClearParameters; + +class Clear11 : angle::NonCopyable +{ + public: + explicit Clear11(Renderer11 *renderer); + ~Clear11(); + + // Clears the framebuffer with the supplied clear parameters, assumes that the framebuffer is currently applied. + gl::Error clearFramebuffer(const ClearParameters &clearParams, const gl::Framebuffer::Data &fboData); + + private: + struct MaskedRenderTarget + { + bool colorMask[4]; + RenderTarget11 *renderTarget; + }; + + ID3D11BlendState *getBlendState(const std::vector<MaskedRenderTarget> &rts); + ID3D11DepthStencilState *getDepthStencilState(const ClearParameters &clearParams); + + struct ClearShader + { + ID3D11InputLayout *inputLayout; + ID3D11VertexShader *vertexShader; + ID3D11PixelShader *pixelShader; + }; + + template <unsigned int vsSize, unsigned int psSize> + static ClearShader CreateClearShader(ID3D11Device *device, DXGI_FORMAT colorType, const BYTE(&vsByteCode)[vsSize], const BYTE(&psByteCode)[psSize]); + + Renderer11 *mRenderer; + + struct ClearBlendInfo + { + bool maskChannels[D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT][4]; + }; + typedef bool(*ClearBlendInfoComparisonFunction)(const ClearBlendInfo&, const ClearBlendInfo &); + typedef std::map<ClearBlendInfo, ID3D11BlendState*, ClearBlendInfoComparisonFunction> ClearBlendStateMap; + ClearBlendStateMap mClearBlendStates; + + ClearShader mFloatClearShader; + ClearShader mUintClearShader; + ClearShader mIntClearShader; + + struct ClearDepthStencilInfo + { + bool clearDepth; + bool clearStencil; + UINT8 stencilWriteMask; + }; + typedef bool (*ClearDepthStencilInfoComparisonFunction)(const ClearDepthStencilInfo&, const ClearDepthStencilInfo &); + typedef std::map<ClearDepthStencilInfo, ID3D11DepthStencilState*, ClearDepthStencilInfoComparisonFunction> ClearDepthStencilStateMap; + ClearDepthStencilStateMap mClearDepthStencilStates; + + ID3D11Buffer *mVertexBuffer; + ID3D11RasterizerState *mRasterizerState; + + bool mSupportsClearView; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_CLEAR11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.cpp new file mode 100644 index 0000000000..f1fe2bb2c7 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.cpp @@ -0,0 +1,119 @@ +// +// Copyright 2015 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// +// DebugAnnotator11.cpp: D3D11 helpers for adding trace annotations. +// + +#include "libANGLE/renderer/d3d/d3d11/DebugAnnotator11.h" + +#include "common/debug.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" + +namespace rx +{ + +DebugAnnotator11::DebugAnnotator11() + : mInitialized(false), + mD3d11Module(nullptr), + mUserDefinedAnnotation(nullptr) +{ + // D3D11 devices can't be created during DllMain. + // We defer device creation until the object is actually used. +} + +DebugAnnotator11::~DebugAnnotator11() +{ + if (mInitialized) + { +#if defined(ANGLE_ENABLE_D3D11_1) + SafeRelease(mUserDefinedAnnotation); +#endif + +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) + FreeLibrary(mD3d11Module); +#endif // !ANGLE_ENABLE_WINDOWS_STORE + } +} + +void DebugAnnotator11::beginEvent(const std::wstring &eventName) +{ + initializeDevice(); + +#if defined(ANGLE_ENABLE_D3D11_1) + mUserDefinedAnnotation->BeginEvent(eventName.c_str()); +#endif +} + +void DebugAnnotator11::endEvent() +{ + initializeDevice(); + +#if defined(ANGLE_ENABLE_D3D11_1) + mUserDefinedAnnotation->EndEvent(); +#endif +} + +void DebugAnnotator11::setMarker(const std::wstring &markerName) +{ + initializeDevice(); + +#if defined(ANGLE_ENABLE_D3D11_1) + mUserDefinedAnnotation->SetMarker(markerName.c_str()); +#endif +} + +bool DebugAnnotator11::getStatus() +{ + // ID3DUserDefinedAnnotation::GetStatus doesn't work with the Graphics Diagnostics tools in Visual Studio 2013. + +#if defined(_DEBUG) && defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY != WINAPI_FAMILY_PHONE_APP) + // In the Windows Store, we can use IDXGraphicsAnalysis. The call to GetDebugInterface1 only succeeds if the app is under capture. + // This should only be called in DEBUG mode. + // If an app links against DXGIGetDebugInterface1 in release mode then it will fail Windows Store ingestion checks. + IDXGraphicsAnalysis *graphicsAnalysis; + DXGIGetDebugInterface1(0, IID_PPV_ARGS(&graphicsAnalysis)); + bool underCapture = (graphicsAnalysis != nullptr); + SafeRelease(graphicsAnalysis); + return underCapture; +#endif // _DEBUG && !ANGLE_ENABLE_WINDOWS_STORE + + // Otherwise, we have to return true here. + return true; +} + +void DebugAnnotator11::initializeDevice() +{ + if (!mInitialized) + { +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) + mD3d11Module = LoadLibrary(TEXT("d3d11.dll")); + ASSERT(mD3d11Module); + + PFN_D3D11_CREATE_DEVICE D3D11CreateDevice = (PFN_D3D11_CREATE_DEVICE)GetProcAddress(mD3d11Module, "D3D11CreateDevice"); + ASSERT(D3D11CreateDevice != nullptr); +#endif // !ANGLE_ENABLE_WINDOWS_STORE + + ID3D11Device *device = nullptr; + ID3D11DeviceContext *context = nullptr; + + HRESULT hr = E_FAIL; + + // Create a D3D_DRIVER_TYPE_NULL device, which is much cheaper than other types of device. + hr = D3D11CreateDevice(NULL, D3D_DRIVER_TYPE_NULL, nullptr, 0, nullptr, 0, D3D11_SDK_VERSION, &device, nullptr, &context); + ASSERT(SUCCEEDED(hr)); + +#if defined(ANGLE_ENABLE_D3D11_1) + mUserDefinedAnnotation = d3d11::DynamicCastComObject<ID3DUserDefinedAnnotation>(context); + ASSERT(mUserDefinedAnnotation != nullptr); +#endif + + SafeRelease(device); + SafeRelease(context); + + mInitialized = true; + } +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.h new file mode 100644 index 0000000000..3df62b015c --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/DebugAnnotator11.h @@ -0,0 +1,39 @@ +// +// Copyright 2015 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// +// DebugAnnotator11.h: D3D11 helpers for adding trace annotations. +// + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_DEBUGANNOTATOR11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_DEBUGANNOTATOR11_H_ + +#include "common/debug.h" + +struct ID3DUserDefinedAnnotation; + +namespace rx +{ + +class DebugAnnotator11 : public gl::DebugAnnotator +{ + public: + DebugAnnotator11(); + ~DebugAnnotator11() override; + void beginEvent(const std::wstring &eventName) override; + void endEvent() override; + void setMarker(const std::wstring &markerName) override; + bool getStatus() override; + + private: + void initializeDevice(); + + bool mInitialized; + HMODULE mD3d11Module; + ID3DUserDefinedAnnotation *mUserDefinedAnnotation; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_DEBUGANNOTATOR11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.cpp new file mode 100644 index 0000000000..8552bc2beb --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.cpp @@ -0,0 +1,231 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence11.cpp: Defines the rx::FenceNV11 and rx::FenceSync11 classes which implement rx::FenceNVImpl and rx::FenceSyncImpl. + +#include "libANGLE/renderer/d3d/d3d11/Fence11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" + +#include "common/utilities.h" + +namespace rx +{ + +// +// Template helpers for set and test operations. +// + +template<class FenceClass> +gl::Error FenceSetHelper(FenceClass *fence) +{ + if (!fence->mQuery) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = D3D11_QUERY_EVENT; + queryDesc.MiscFlags = 0; + + HRESULT result = fence->mRenderer->getDevice()->CreateQuery(&queryDesc, &fence->mQuery); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create event query, result: 0x%X.", result); + } + } + + fence->mRenderer->getDeviceContext()->End(fence->mQuery); + return gl::Error(GL_NO_ERROR); +} + +template <class FenceClass> +gl::Error FenceTestHelper(FenceClass *fence, bool flushCommandBuffer, GLboolean *outFinished) +{ + ASSERT(fence->mQuery); + + UINT getDataFlags = (flushCommandBuffer ? 0 : D3D11_ASYNC_GETDATA_DONOTFLUSH); + HRESULT result = fence->mRenderer->getDeviceContext()->GetData(fence->mQuery, NULL, 0, getDataFlags); + + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to get query data, result: 0x%X.", result); + } + else if (fence->mRenderer->isDeviceLost()) + { + return gl::Error(GL_OUT_OF_MEMORY, "Device was lost while querying result of an event query."); + } + + ASSERT(result == S_OK || result == S_FALSE); + *outFinished = ((result == S_OK) ? GL_TRUE : GL_FALSE); + return gl::Error(GL_NO_ERROR); +} + +// +// FenceNV11 +// + +FenceNV11::FenceNV11(Renderer11 *renderer) + : FenceNVImpl(), + mRenderer(renderer), + mQuery(NULL) +{ +} + +FenceNV11::~FenceNV11() +{ + SafeRelease(mQuery); +} + +gl::Error FenceNV11::set() +{ + return FenceSetHelper(this); +} + +gl::Error FenceNV11::test(bool flushCommandBuffer, GLboolean *outFinished) +{ + return FenceTestHelper(this, flushCommandBuffer, outFinished); +} + +gl::Error FenceNV11::finishFence(GLboolean *outFinished) +{ + ASSERT(outFinished); + + while (*outFinished != GL_TRUE) + { + gl::Error error = test(true, outFinished); + if (error.isError()) + { + return error; + } + + ScheduleYield(); + } + + return gl::Error(GL_NO_ERROR); +} + +// +// FenceSync11 +// + +// Important note on accurate timers in Windows: +// +// QueryPerformanceCounter has a few major issues, including being 10x as expensive to call +// as timeGetTime on laptops and "jumping" during certain hardware events. +// +// See the comments at the top of the Chromium source file "chromium/src/base/time/time_win.cc" +// https://code.google.com/p/chromium/codesearch#chromium/src/base/time/time_win.cc +// +// We still opt to use QPC. In the present and moving forward, most newer systems will not suffer +// from buggy implementations. + +FenceSync11::FenceSync11(Renderer11 *renderer) + : FenceSyncImpl(), + mRenderer(renderer), + mQuery(NULL) +{ + LARGE_INTEGER counterFreqency = { 0 }; + BOOL success = QueryPerformanceFrequency(&counterFreqency); + UNUSED_ASSERTION_VARIABLE(success); + ASSERT(success); + + mCounterFrequency = counterFreqency.QuadPart; +} + +FenceSync11::~FenceSync11() +{ + SafeRelease(mQuery); +} + +gl::Error FenceSync11::set() +{ + return FenceSetHelper(this); +} + +gl::Error FenceSync11::clientWait(GLbitfield flags, GLuint64 timeout, GLenum *outResult) +{ + ASSERT(outResult); + + bool flushCommandBuffer = ((flags & GL_SYNC_FLUSH_COMMANDS_BIT) != 0); + + GLboolean result = GL_FALSE; + gl::Error error = FenceTestHelper(this, flushCommandBuffer, &result); + if (error.isError()) + { + *outResult = GL_WAIT_FAILED; + return error; + } + + if (result == GL_TRUE) + { + *outResult = GL_ALREADY_SIGNALED; + return gl::Error(GL_NO_ERROR); + } + + if (timeout == 0) + { + *outResult = GL_TIMEOUT_EXPIRED; + return gl::Error(GL_NO_ERROR); + } + + LARGE_INTEGER currentCounter = { 0 }; + BOOL success = QueryPerformanceCounter(¤tCounter); + UNUSED_ASSERTION_VARIABLE(success); + ASSERT(success); + + LONGLONG timeoutInSeconds = static_cast<LONGLONG>(timeout) * static_cast<LONGLONG>(1000000ll); + LONGLONG endCounter = currentCounter.QuadPart + mCounterFrequency * timeoutInSeconds; + + while (currentCounter.QuadPart < endCounter && !result) + { + ScheduleYield(); + success = QueryPerformanceCounter(¤tCounter); + UNUSED_ASSERTION_VARIABLE(success); + ASSERT(success); + + error = FenceTestHelper(this, flushCommandBuffer, &result); + if (error.isError()) + { + *outResult = GL_WAIT_FAILED; + return error; + } + } + + if (currentCounter.QuadPart >= endCounter) + { + *outResult = GL_TIMEOUT_EXPIRED; + } + else + { + *outResult = GL_CONDITION_SATISFIED; + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error FenceSync11::serverWait(GLbitfield flags, GLuint64 timeout) +{ + // Because our API is currently designed to be called from a single thread, we don't need to do + // extra work for a server-side fence. GPU commands issued after the fence is created will always + // be processed after the fence is signaled. + return gl::Error(GL_NO_ERROR); +} + +gl::Error FenceSync11::getStatus(GLint *outResult) +{ + GLboolean result = GL_FALSE; + gl::Error error = FenceTestHelper(this, false, &result); + if (error.isError()) + { + // The spec does not specify any way to report errors during the status test (e.g. device lost) + // so we report the fence is unblocked in case of error or signaled. + *outResult = GL_SIGNALED; + + return error; + } + + *outResult = (result ? GL_SIGNALED : GL_UNSIGNALED); + return gl::Error(GL_NO_ERROR); +} + +} // namespace rx diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.h new file mode 100644 index 0000000000..2d87f43e76 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Fence11.h @@ -0,0 +1,59 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Fence11.h: Defines the rx::FenceNV11 and rx::FenceSync11 classes which implement rx::FenceNVImpl and rx::FenceSyncImpl. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_FENCE11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_FENCE11_H_ + +#include "libANGLE/renderer/FenceNVImpl.h" +#include "libANGLE/renderer/FenceSyncImpl.h" + +namespace rx +{ +class Renderer11; + +class FenceNV11 : public FenceNVImpl +{ + public: + explicit FenceNV11(Renderer11 *renderer); + virtual ~FenceNV11(); + + gl::Error set(); + gl::Error test(bool flushCommandBuffer, GLboolean *outFinished); + gl::Error finishFence(GLboolean *outFinished); + + private: + template<class T> friend gl::Error FenceSetHelper(T *fence); + template<class T> friend gl::Error FenceTestHelper(T *fence, bool flushCommandBuffer, GLboolean *outFinished); + + Renderer11 *mRenderer; + ID3D11Query *mQuery; +}; + +class FenceSync11 : public FenceSyncImpl +{ + public: + explicit FenceSync11(Renderer11 *renderer); + virtual ~FenceSync11(); + + gl::Error set(); + gl::Error clientWait(GLbitfield flags, GLuint64 timeout, GLenum *outResult); + gl::Error serverWait(GLbitfield flags, GLuint64 timeout); + gl::Error getStatus(GLint *outResult); + + private: + template<class T> friend gl::Error FenceSetHelper(T *fence); + template<class T> friend gl::Error FenceTestHelper(T *fence, bool flushCommandBuffer, GLboolean *outFinished); + + Renderer11 *mRenderer; + ID3D11Query *mQuery; + LONGLONG mCounterFrequency; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_FENCE11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.cpp new file mode 100644 index 0000000000..da01f320c0 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.cpp @@ -0,0 +1,270 @@ +// +// Copyright 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Framebuffer11.cpp: Implements the Framebuffer11 class. + +#include "libANGLE/renderer/d3d/d3d11/Framebuffer11.h" + +#include "common/debug.h" +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Clear11.h" +#include "libANGLE/renderer/d3d/d3d11/TextureStorage11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/TextureD3D.h" +#include "libANGLE/Framebuffer.h" +#include "libANGLE/FramebufferAttachment.h" +#include "libANGLE/Texture.h" + +namespace rx +{ + +Framebuffer11::Framebuffer11(const gl::Framebuffer::Data &data, Renderer11 *renderer) + : FramebufferD3D(data, renderer), + mRenderer(renderer) +{ + ASSERT(mRenderer != nullptr); +} + +Framebuffer11::~Framebuffer11() +{ +} + +static gl::Error InvalidateAttachmentSwizzles(const gl::FramebufferAttachment *attachment) +{ + if (attachment && attachment->type() == GL_TEXTURE) + { + gl::Texture *texture = attachment->getTexture(); + + TextureD3D *textureD3D = GetImplAs<TextureD3D>(texture); + + TextureStorage *texStorage = NULL; + gl::Error error = textureD3D->getNativeTexture(&texStorage); + if (error.isError()) + { + return error; + } + + if (texStorage) + { + TextureStorage11 *texStorage11 = TextureStorage11::makeTextureStorage11(texStorage); + ASSERT(texStorage11); + + texStorage11->invalidateSwizzleCacheLevel(attachment->mipLevel()); + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Framebuffer11::invalidateSwizzles() const +{ + for (auto it = mData.mColorAttachments.cbegin(); it != mData.mColorAttachments.cend(); ++it) + { + gl::FramebufferAttachment *colorAttachment = *it; + gl::Error error = InvalidateAttachmentSwizzles(colorAttachment); + if (error.isError()) + { + return error; + } + } + + gl::Error error = InvalidateAttachmentSwizzles(mData.mDepthAttachment); + if (error.isError()) + { + return error; + } + + error = InvalidateAttachmentSwizzles(mData.mStencilAttachment); + if (error.isError()) + { + return error; + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Framebuffer11::clear(const gl::State &state, const ClearParameters &clearParams) +{ + Clear11 *clearer = mRenderer->getClearer(); + gl::Error error = clearer->clearFramebuffer(clearParams, mData); + if (error.isError()) + { + return error; + } + + error = invalidateSwizzles(); + if (error.isError()) + { + return error; + } + + return gl::Error(GL_NO_ERROR); +} + +static gl::Error getRenderTargetResource(const gl::FramebufferAttachment *colorbuffer, unsigned int *subresourceIndexOut, + ID3D11Texture2D **texture2DOut) +{ + ASSERT(colorbuffer); + + RenderTarget11 *renderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &renderTarget); + if (error.isError()) + { + return error; + } + + ID3D11Resource *renderTargetResource = renderTarget->getTexture(); + ASSERT(renderTargetResource); + + *subresourceIndexOut = renderTarget->getSubresourceIndex(); + *texture2DOut = d3d11::DynamicCastComObject<ID3D11Texture2D>(renderTargetResource); + + if (!(*texture2DOut)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to query the ID3D11Texture2D from a RenderTarget"); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Framebuffer11::readPixels(const gl::Rectangle &area, GLenum format, GLenum type, size_t outputPitch, const gl::PixelPackState &pack, uint8_t *pixels) const +{ + ID3D11Texture2D *colorBufferTexture = NULL; + unsigned int subresourceIndex = 0; + + const gl::FramebufferAttachment *colorbuffer = mData.getReadAttachment(); + ASSERT(colorbuffer); + + gl::Error error = getRenderTargetResource(colorbuffer, &subresourceIndex, &colorBufferTexture); + if (error.isError()) + { + return error; + } + + gl::Buffer *packBuffer = pack.pixelBuffer.get(); + if (packBuffer != NULL) + { + Buffer11 *packBufferStorage = Buffer11::makeBuffer11(packBuffer->getImplementation()); + PackPixelsParams packParams(area, format, type, outputPitch, pack, reinterpret_cast<ptrdiff_t>(pixels)); + + error = packBufferStorage->packPixels(colorBufferTexture, subresourceIndex, packParams); + if (error.isError()) + { + SafeRelease(colorBufferTexture); + return error; + } + + packBuffer->getIndexRangeCache()->clear(); + } + else + { + error = mRenderer->readTextureData(colorBufferTexture, subresourceIndex, area, format, type, outputPitch, pack, pixels); + if (error.isError()) + { + SafeRelease(colorBufferTexture); + return error; + } + } + + SafeRelease(colorBufferTexture); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Framebuffer11::blit(const gl::Rectangle &sourceArea, const gl::Rectangle &destArea, const gl::Rectangle *scissor, + bool blitRenderTarget, bool blitDepth, bool blitStencil, GLenum filter, + const gl::Framebuffer *sourceFramebuffer) +{ + if (blitRenderTarget) + { + const gl::FramebufferAttachment *readBuffer = sourceFramebuffer->getReadColorbuffer(); + ASSERT(readBuffer); + + RenderTargetD3D *readRenderTarget = NULL; + gl::Error error = GetAttachmentRenderTarget(readBuffer, &readRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(readRenderTarget); + + for (size_t colorAttachment = 0; colorAttachment < mData.mColorAttachments.size(); colorAttachment++) + { + if (mData.mColorAttachments[colorAttachment] != nullptr && + mData.mDrawBufferStates[colorAttachment] != GL_NONE) + { + const gl::FramebufferAttachment *drawBuffer = mData.mColorAttachments[colorAttachment]; + + RenderTargetD3D *drawRenderTarget = NULL; + error = GetAttachmentRenderTarget(drawBuffer, &drawRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(drawRenderTarget); + + error = mRenderer->blitRenderbufferRect(sourceArea, destArea, readRenderTarget, drawRenderTarget, + filter, scissor, blitRenderTarget, false, false); + if (error.isError()) + { + return error; + } + } + } + } + + if (blitDepth || blitStencil) + { + gl::FramebufferAttachment *readBuffer = sourceFramebuffer->getDepthOrStencilbuffer(); + ASSERT(readBuffer); + + RenderTargetD3D *readRenderTarget = NULL; + gl::Error error = GetAttachmentRenderTarget(readBuffer, &readRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(readRenderTarget); + + const gl::FramebufferAttachment *drawBuffer = mData.getDepthOrStencilAttachment(); + ASSERT(drawBuffer); + + RenderTargetD3D *drawRenderTarget = NULL; + error = GetAttachmentRenderTarget(drawBuffer, &drawRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(drawRenderTarget); + + error = mRenderer->blitRenderbufferRect(sourceArea, destArea, readRenderTarget, drawRenderTarget, filter, scissor, + false, blitDepth, blitStencil); + if (error.isError()) + { + return error; + } + } + + gl::Error error = invalidateSwizzles(); + if (error.isError()) + { + return error; + } + + return gl::Error(GL_NO_ERROR); +} + +GLenum Framebuffer11::getRenderTargetImplementationFormat(RenderTargetD3D *renderTarget) const +{ + RenderTarget11 *renderTarget11 = RenderTarget11::makeRenderTarget11(renderTarget); + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(renderTarget11->getDXGIFormat()); + return dxgiFormatInfo.internalFormat; +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.h new file mode 100644 index 0000000000..07fa480fa2 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Framebuffer11.h @@ -0,0 +1,45 @@ +// +// Copyright 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Framebuffer11.h: Defines the Framebuffer11 class. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_FRAMBUFFER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_FRAMBUFFER11_H_ + +#include "libANGLE/renderer/d3d/FramebufferD3D.h" + +namespace rx +{ +class Renderer11; + +class Framebuffer11 : public FramebufferD3D +{ + public: + Framebuffer11(const gl::Framebuffer::Data &data, Renderer11 *renderer); + virtual ~Framebuffer11(); + + // Invalidate the cached swizzles of all bound texture attachments. + gl::Error invalidateSwizzles() const; + + private: + gl::Error clear(const gl::State &state, const ClearParameters &clearParams) override; + + gl::Error readPixels(const gl::Rectangle &area, GLenum format, GLenum type, size_t outputPitch, + const gl::PixelPackState &pack, uint8_t *pixels) const override; + + gl::Error blit(const gl::Rectangle &sourceArea, const gl::Rectangle &destArea, const gl::Rectangle *scissor, + bool blitRenderTarget, bool blitDepth, bool blitStencil, GLenum filter, + const gl::Framebuffer *sourceFramebuffer) override; + + + GLenum getRenderTargetImplementationFormat(RenderTargetD3D *renderTarget) const override; + + Renderer11 *const mRenderer; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_FRAMBUFFER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.cpp new file mode 100644 index 0000000000..956b78b5a6 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.cpp @@ -0,0 +1,664 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image11.h: Implements the rx::Image11 class, which acts as the interface to +// the actual underlying resources of a Texture + +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/Image11.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/TextureStorage11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/Framebuffer.h" +#include "libANGLE/FramebufferAttachment.h" +#include "libANGLE/formatutils.h" + +#include "common/utilities.h" + +namespace rx +{ + +Image11::Image11(Renderer11 *renderer) + : mRenderer(renderer), + mDXGIFormat(DXGI_FORMAT_UNKNOWN), + mStagingTexture(NULL), + mStagingSubresource(0), + mRecoverFromStorage(false), + mAssociatedStorage(NULL), + mAssociatedImageIndex(gl::ImageIndex::MakeInvalid()), + mRecoveredFromStorageCount(0) +{ + // mRenderer should remain unchanged during the lifetime of the Image11 object. + // This lets us safely use mRenderer (and its Feature Level) in Image11's methods. + mFeatureLevel = renderer->getFeatureLevel(); +} + +Image11::~Image11() +{ + disassociateStorage(); + releaseStagingTexture(); +} + +Image11 *Image11::makeImage11(ImageD3D *img) +{ + ASSERT(HAS_DYNAMIC_TYPE(Image11*, img)); + return static_cast<Image11*>(img); +} + +gl::Error Image11::generateMipmap(Image11 *dest, Image11 *src) +{ + ASSERT(src->getDXGIFormat() == dest->getDXGIFormat()); + ASSERT(src->getWidth() == 1 || src->getWidth() / 2 == dest->getWidth()); + ASSERT(src->getHeight() == 1 || src->getHeight() / 2 == dest->getHeight()); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(src->getDXGIFormat()); + ASSERT(dxgiFormatInfo.mipGenerationFunction != NULL); + + D3D11_MAPPED_SUBRESOURCE destMapped; + gl::Error error = dest->map(D3D11_MAP_WRITE, &destMapped); + if (error.isError()) + { + return error; + } + + D3D11_MAPPED_SUBRESOURCE srcMapped; + error = src->map(D3D11_MAP_READ, &srcMapped); + if (error.isError()) + { + dest->unmap(); + return error; + } + + const uint8_t *sourceData = reinterpret_cast<const uint8_t*>(srcMapped.pData); + uint8_t *destData = reinterpret_cast<uint8_t*>(destMapped.pData); + + dxgiFormatInfo.mipGenerationFunction(src->getWidth(), src->getHeight(), src->getDepth(), + sourceData, srcMapped.RowPitch, srcMapped.DepthPitch, + destData, destMapped.RowPitch, destMapped.DepthPitch); + + dest->unmap(); + src->unmap(); + + dest->markDirty(); + + return gl::Error(GL_NO_ERROR); +} + +bool Image11::isDirty() const +{ + // If mDirty is true + // AND mStagingTexture doesn't exist AND mStagingTexture doesn't need to be recovered from TextureStorage + // AND the texture doesn't require init data (i.e. a blank new texture will suffice) + // then isDirty should still return false. + if (mDirty && !mStagingTexture && !mRecoverFromStorage && !(d3d11::GetTextureFormatInfo(mInternalFormat, mFeatureLevel).dataInitializerFunction != NULL)) + { + return false; + } + + return mDirty; +} + +gl::Error Image11::copyToStorage(TextureStorage *storage, const gl::ImageIndex &index, const gl::Box ®ion) +{ + TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(storage); + + // If an app's behavior results in an Image11 copying its data to/from to a TextureStorage multiple times, + // then we should just keep the staging texture around to prevent the copying from impacting perf. + // We allow the Image11 to copy its data to/from TextureStorage once. + // This accounts for an app making a late call to glGenerateMipmap. + bool attemptToReleaseStagingTexture = (mRecoveredFromStorageCount < 2); + + if (attemptToReleaseStagingTexture) + { + // If another image is relying on this Storage for its data, then we must let it recover its data before we overwrite it. + gl::Error error = storage11->releaseAssociatedImage(index, this); + if (error.isError()) + { + return error; + } + } + + ID3D11Resource *stagingTexture = NULL; + unsigned int stagingSubresourceIndex = 0; + gl::Error error = getStagingTexture(&stagingTexture, &stagingSubresourceIndex); + if (error.isError()) + { + return error; + } + + error = storage11->updateSubresourceLevel(stagingTexture, stagingSubresourceIndex, index, region); + if (error.isError()) + { + return error; + } + + // Once the image data has been copied into the Storage, we can release it locally. + if (attemptToReleaseStagingTexture) + { + storage11->associateImage(this, index); + releaseStagingTexture(); + mRecoverFromStorage = true; + mAssociatedStorage = storage11; + mAssociatedImageIndex = index; + } + + return gl::Error(GL_NO_ERROR); +} + +bool Image11::isAssociatedStorageValid(TextureStorage11* textureStorage) const +{ + return (mAssociatedStorage == textureStorage); +} + +gl::Error Image11::recoverFromAssociatedStorage() +{ + if (mRecoverFromStorage) + { + gl::Error error = createStagingTexture(); + if (error.isError()) + { + return error; + } + + bool textureStorageCorrect = mAssociatedStorage->isAssociatedImageValid(mAssociatedImageIndex, this); + + // This means that the cached TextureStorage has been modified after this Image11 released its copy of its data. + // This should not have happened. The TextureStorage should have told this Image11 to recover its data before it was overwritten. + ASSERT(textureStorageCorrect); + + if (textureStorageCorrect) + { + // CopySubResource from the Storage to the Staging texture + gl::Box region(0, 0, 0, mWidth, mHeight, mDepth); + error = mAssociatedStorage->copySubresourceLevel(mStagingTexture, mStagingSubresource, mAssociatedImageIndex, region); + if (error.isError()) + { + return error; + } + + mRecoveredFromStorageCount += 1; + } + + // Reset all the recovery parameters, even if the texture storage association is broken. + disassociateStorage(); + } + + return gl::Error(GL_NO_ERROR); +} + +void Image11::disassociateStorage() +{ + if (mRecoverFromStorage) + { + // Make the texturestorage release the Image11 too + mAssociatedStorage->disassociateImage(mAssociatedImageIndex, this); + + mRecoverFromStorage = false; + mAssociatedStorage = NULL; + mAssociatedImageIndex = gl::ImageIndex::MakeInvalid(); + } +} + +bool Image11::redefine(GLenum target, GLenum internalformat, const gl::Extents &size, bool forceRelease) +{ + if (mWidth != size.width || + mHeight != size.height || + mInternalFormat != internalformat || + forceRelease) + { + // End the association with the TextureStorage, since that data will be out of date. + // Also reset mRecoveredFromStorageCount since this Image is getting completely redefined. + disassociateStorage(); + mRecoveredFromStorageCount = 0; + + mWidth = size.width; + mHeight = size.height; + mDepth = size.depth; + mInternalFormat = internalformat; + mTarget = target; + + // compute the d3d format that will be used + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalformat, mFeatureLevel); + mDXGIFormat = formatInfo.texFormat; + mRenderable = (formatInfo.rtvFormat != DXGI_FORMAT_UNKNOWN); + + releaseStagingTexture(); + mDirty = (formatInfo.dataInitializerFunction != NULL); + + return true; + } + + return false; +} + +DXGI_FORMAT Image11::getDXGIFormat() const +{ + // this should only happen if the image hasn't been redefined first + // which would be a bug by the caller + ASSERT(mDXGIFormat != DXGI_FORMAT_UNKNOWN); + + return mDXGIFormat; +} + +// Store the pixel rectangle designated by xoffset,yoffset,width,height with pixels stored as format/type at input +// into the target pixel rectangle. +gl::Error Image11::loadData(const gl::Box &area, const gl::PixelUnpackState &unpack, GLenum type, const void *input) +{ + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(mInternalFormat); + GLsizei inputRowPitch = formatInfo.computeRowPitch(type, area.width, unpack.alignment, unpack.rowLength); + GLsizei inputDepthPitch = formatInfo.computeDepthPitch(type, area.width, area.height, unpack.alignment, unpack.rowLength); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(mDXGIFormat); + GLuint outputPixelSize = dxgiFormatInfo.pixelBytes; + + const d3d11::TextureFormat &d3dFormatInfo = d3d11::GetTextureFormatInfo(mInternalFormat, mFeatureLevel); + LoadImageFunction loadFunction = d3dFormatInfo.loadFunctions.at(type); + + D3D11_MAPPED_SUBRESOURCE mappedImage; + gl::Error error = map(D3D11_MAP_WRITE, &mappedImage); + if (error.isError()) + { + return error; + } + + uint8_t *offsetMappedData = (reinterpret_cast<uint8_t*>(mappedImage.pData) + (area.y * mappedImage.RowPitch + area.x * outputPixelSize + area.z * mappedImage.DepthPitch)); + loadFunction(area.width, area.height, area.depth, + reinterpret_cast<const uint8_t*>(input), inputRowPitch, inputDepthPitch, + offsetMappedData, mappedImage.RowPitch, mappedImage.DepthPitch); + + unmap(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Image11::loadCompressedData(const gl::Box &area, const void *input) +{ + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(mInternalFormat); + GLsizei inputRowPitch = formatInfo.computeRowPitch(GL_UNSIGNED_BYTE, area.width, 1, 0); + GLsizei inputDepthPitch = formatInfo.computeDepthPitch(GL_UNSIGNED_BYTE, area.width, area.height, 1, 0); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(mDXGIFormat); + GLuint outputPixelSize = dxgiFormatInfo.pixelBytes; + GLuint outputBlockWidth = dxgiFormatInfo.blockWidth; + GLuint outputBlockHeight = dxgiFormatInfo.blockHeight; + + ASSERT(area.x % outputBlockWidth == 0); + ASSERT(area.y % outputBlockHeight == 0); + + const d3d11::TextureFormat &d3dFormatInfo = d3d11::GetTextureFormatInfo(mInternalFormat, mFeatureLevel); + LoadImageFunction loadFunction = d3dFormatInfo.loadFunctions.at(GL_UNSIGNED_BYTE); + + D3D11_MAPPED_SUBRESOURCE mappedImage; + gl::Error error = map(D3D11_MAP_WRITE, &mappedImage); + if (error.isError()) + { + return error; + } + + uint8_t* offsetMappedData = reinterpret_cast<uint8_t*>(mappedImage.pData) + ((area.y / outputBlockHeight) * mappedImage.RowPitch + + (area.x / outputBlockWidth) * outputPixelSize + + area.z * mappedImage.DepthPitch); + + loadFunction(area.width, area.height, area.depth, + reinterpret_cast<const uint8_t*>(input), inputRowPitch, inputDepthPitch, + offsetMappedData, mappedImage.RowPitch, mappedImage.DepthPitch); + + unmap(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Image11::copy(const gl::Offset &destOffset, const gl::Rectangle &sourceArea, RenderTargetD3D *source) +{ + RenderTarget11 *sourceRenderTarget = RenderTarget11::makeRenderTarget11(source); + ASSERT(sourceRenderTarget->getTexture()); + + ID3D11Resource *resource = sourceRenderTarget->getTexture(); + UINT subresourceIndex = sourceRenderTarget->getSubresourceIndex(); + + gl::Box sourceBox(sourceArea.x, sourceArea.y, 0, sourceArea.width, sourceArea.height, 1); + gl::Error error = copy(destOffset, sourceBox, resource, subresourceIndex); + + SafeRelease(resource); + + return error; +} + +gl::Error Image11::copy(const gl::Offset &destOffset, const gl::Box &sourceArea, const gl::ImageIndex &sourceIndex, TextureStorage *source) +{ + TextureStorage11 *sourceStorage11 = TextureStorage11::makeTextureStorage11(source); + + UINT subresourceIndex = sourceStorage11->getSubresourceIndex(sourceIndex); + ID3D11Resource *resource = NULL; + gl::Error error = sourceStorage11->getResource(&resource); + if (error.isError()) + { + return error; + } + + error = copy(destOffset, sourceArea, resource, subresourceIndex); + + SafeRelease(resource); + + return error; +} + +gl::Error Image11::copy(const gl::Offset &destOffset, const gl::Box &sourceArea, ID3D11Resource *source, UINT sourceSubResource) +{ + D3D11_RESOURCE_DIMENSION dim; + source->GetType(&dim); + + DXGI_FORMAT format = DXGI_FORMAT_UNKNOWN; + gl::Extents extents; + UINT sampleCount = 0; + + ID3D11Texture2D *source2D = NULL; + + if (dim == D3D11_RESOURCE_DIMENSION_TEXTURE2D) + { + D3D11_TEXTURE2D_DESC textureDesc2D; + source2D = d3d11::DynamicCastComObject<ID3D11Texture2D>(source); + ASSERT(source2D); + source2D->GetDesc(&textureDesc2D); + + format = textureDesc2D.Format; + extents = gl::Extents(textureDesc2D.Width, textureDesc2D.Height, 1); + sampleCount = textureDesc2D.SampleDesc.Count; + } + else if (dim == D3D11_RESOURCE_DIMENSION_TEXTURE3D) + { + D3D11_TEXTURE3D_DESC textureDesc3D; + ID3D11Texture3D *source3D = d3d11::DynamicCastComObject<ID3D11Texture3D>(source); + ASSERT(source3D); + source3D->GetDesc(&textureDesc3D); + + format = textureDesc3D.Format; + extents = gl::Extents(textureDesc3D.Width, textureDesc3D.Height, textureDesc3D.Depth); + sampleCount = 1; + } + else + { + UNREACHABLE(); + } + + if (format == mDXGIFormat) + { + // No conversion needed-- use copyback fastpath + ID3D11Resource *stagingTexture = NULL; + unsigned int stagingSubresourceIndex = 0; + gl::Error error = getStagingTexture(&stagingTexture, &stagingSubresourceIndex); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + UINT subresourceAfterResolve = sourceSubResource; + + ID3D11Resource *srcTex = NULL; + + bool needResolve = (dim == D3D11_RESOURCE_DIMENSION_TEXTURE2D && sampleCount > 1); + + if (needResolve) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = extents.width; + resolveDesc.Height = extents.height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = D3D11_USAGE_DEFAULT; + resolveDesc.BindFlags = 0; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + ID3D11Texture2D *srcTex2D = NULL; + HRESULT result = device->CreateTexture2D(&resolveDesc, NULL, &srcTex2D); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create resolve texture for Image11::copy, HRESULT: 0x%X.", result); + } + srcTex = srcTex2D; + + deviceContext->ResolveSubresource(srcTex, 0, source, sourceSubResource, format); + subresourceAfterResolve = 0; + } + else + { + srcTex = source; + } + + D3D11_BOX srcBox; + srcBox.left = sourceArea.x; + srcBox.right = sourceArea.x + sourceArea.width; + srcBox.top = sourceArea.y; + srcBox.bottom = sourceArea.y + sourceArea.height; + srcBox.front = sourceArea.z; + srcBox.back = sourceArea.z + sourceArea.depth; + + deviceContext->CopySubresourceRegion(stagingTexture, stagingSubresourceIndex, destOffset.x, destOffset.y, + destOffset.z, srcTex, subresourceAfterResolve, &srcBox); + + if (needResolve) + { + SafeRelease(srcTex); + } + } + else + { + // This format requires conversion, so we must copy the texture to staging and manually convert via readPixels + D3D11_MAPPED_SUBRESOURCE mappedImage; + gl::Error error = map(D3D11_MAP_WRITE, &mappedImage); + if (error.isError()) + { + return error; + } + + // determine the offset coordinate into the destination buffer + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(mDXGIFormat); + GLsizei rowOffset = dxgiFormatInfo.pixelBytes * destOffset.x; + uint8_t *dataOffset = static_cast<uint8_t*>(mappedImage.pData) + mappedImage.RowPitch * destOffset.y + rowOffset + destOffset.z * mappedImage.DepthPitch; + + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(mInternalFormat); + + // Currently in ANGLE, the source data may only need to be converted if the source is the current framebuffer + // and OpenGL ES framebuffers must be 2D textures therefore we should not need to convert 3D textures between different formats. + ASSERT(dim == D3D11_RESOURCE_DIMENSION_TEXTURE2D); + ASSERT(sourceArea.z == 0 && sourceArea.depth == 1); + gl::Rectangle sourceRect(sourceArea.x, sourceArea.y, sourceArea.width, sourceArea.height); + error = mRenderer->readTextureData(source2D, sourceSubResource, sourceRect, formatInfo.format, formatInfo.type, mappedImage.RowPitch, gl::PixelPackState(), dataOffset); + + unmap(); + + if (error.isError()) + { + return error; + } + } + + mDirty = true; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Image11::getStagingTexture(ID3D11Resource **outStagingTexture, unsigned int *outSubresourceIndex) +{ + gl::Error error = createStagingTexture(); + if (error.isError()) + { + return error; + } + + *outStagingTexture = mStagingTexture; + *outSubresourceIndex = mStagingSubresource; + return gl::Error(GL_NO_ERROR); +} + +void Image11::releaseStagingTexture() +{ + SafeRelease(mStagingTexture); +} + +gl::Error Image11::createStagingTexture() +{ + if (mStagingTexture) + { + return gl::Error(GL_NO_ERROR); + } + + ASSERT(mWidth > 0 && mHeight > 0 && mDepth > 0); + + const DXGI_FORMAT dxgiFormat = getDXGIFormat(); + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + int lodOffset = 1; + GLsizei width = mWidth; + GLsizei height = mHeight; + + // adjust size if needed for compressed textures + d3d11::MakeValidSize(false, dxgiFormat, &width, &height, &lodOffset); + + if (mTarget == GL_TEXTURE_3D) + { + ID3D11Texture3D *newTexture = NULL; + + D3D11_TEXTURE3D_DESC desc; + desc.Width = width; + desc.Height = height; + desc.Depth = mDepth; + desc.MipLevels = lodOffset + 1; + desc.Format = dxgiFormat; + desc.Usage = D3D11_USAGE_STAGING; + desc.BindFlags = 0; + desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE; + desc.MiscFlags = 0; + + if (d3d11::GetTextureFormatInfo(mInternalFormat, mFeatureLevel).dataInitializerFunction != NULL) + { + std::vector<D3D11_SUBRESOURCE_DATA> initialData; + std::vector< std::vector<BYTE> > textureData; + d3d11::GenerateInitialTextureData(mInternalFormat, mFeatureLevel, width, height, mDepth, + lodOffset + 1, &initialData, &textureData); + + result = device->CreateTexture3D(&desc, initialData.data(), &newTexture); + } + else + { + result = device->CreateTexture3D(&desc, NULL, &newTexture); + } + + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create staging texture, result: 0x%X.", result); + } + + mStagingTexture = newTexture; + mStagingSubresource = D3D11CalcSubresource(lodOffset, 0, lodOffset + 1); + } + else if (mTarget == GL_TEXTURE_2D || mTarget == GL_TEXTURE_2D_ARRAY || mTarget == GL_TEXTURE_CUBE_MAP) + { + ID3D11Texture2D *newTexture = NULL; + + D3D11_TEXTURE2D_DESC desc; + desc.Width = width; + desc.Height = height; + desc.MipLevels = lodOffset + 1; + desc.ArraySize = 1; + desc.Format = dxgiFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_STAGING; + desc.BindFlags = 0; + desc.CPUAccessFlags = D3D11_CPU_ACCESS_READ | D3D11_CPU_ACCESS_WRITE; + desc.MiscFlags = 0; + + if (d3d11::GetTextureFormatInfo(mInternalFormat, mFeatureLevel).dataInitializerFunction != NULL) + { + std::vector<D3D11_SUBRESOURCE_DATA> initialData; + std::vector< std::vector<BYTE> > textureData; + d3d11::GenerateInitialTextureData(mInternalFormat, mFeatureLevel, width, height, 1, + lodOffset + 1, &initialData, &textureData); + + result = device->CreateTexture2D(&desc, initialData.data(), &newTexture); + } + else + { + result = device->CreateTexture2D(&desc, NULL, &newTexture); + } + + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create staging texture, result: 0x%X.", result); + } + + mStagingTexture = newTexture; + mStagingSubresource = D3D11CalcSubresource(lodOffset, 0, lodOffset + 1); + } + else + { + UNREACHABLE(); + } + + mDirty = false; + return gl::Error(GL_NO_ERROR); +} + +gl::Error Image11::map(D3D11_MAP mapType, D3D11_MAPPED_SUBRESOURCE *map) +{ + // We must recover from the TextureStorage if necessary, even for D3D11_MAP_WRITE. + gl::Error error = recoverFromAssociatedStorage(); + if (error.isError()) + { + return error; + } + + ID3D11Resource *stagingTexture = NULL; + unsigned int subresourceIndex = 0; + error = getStagingTexture(&stagingTexture, &subresourceIndex); + if (error.isError()) + { + return error; + } + + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + ASSERT(mStagingTexture); + HRESULT result = deviceContext->Map(stagingTexture, subresourceIndex, mapType, 0, map); + + // this can fail if the device is removed (from TDR) + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + } + else if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map staging texture, result: 0x%X.", result); + } + + mDirty = true; + + return gl::Error(GL_NO_ERROR); +} + +void Image11::unmap() +{ + if (mStagingTexture) + { + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + deviceContext->Unmap(mStagingTexture, mStagingSubresource); + } +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.h new file mode 100644 index 0000000000..5734f73b15 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Image11.h @@ -0,0 +1,84 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Image11.h: Defines the rx::Image11 class, which acts as the interface to +// the actual underlying resources of a Texture + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_IMAGE11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_IMAGE11_H_ + +#include "libANGLE/renderer/d3d/ImageD3D.h" +#include "libANGLE/ImageIndex.h" + +#include "common/debug.h" + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer11; +class TextureStorage11; + +class Image11 : public ImageD3D +{ + public: + Image11(Renderer11 *renderer); + virtual ~Image11(); + + static Image11 *makeImage11(ImageD3D *img); + + static gl::Error generateMipmap(Image11 *dest, Image11 *src); + + virtual bool isDirty() const; + + virtual gl::Error copyToStorage(TextureStorage *storage, const gl::ImageIndex &index, const gl::Box ®ion); + + bool redefine(GLenum target, GLenum internalformat, const gl::Extents &size, bool forceRelease) override; + + DXGI_FORMAT getDXGIFormat() const; + + virtual gl::Error loadData(const gl::Box &area, const gl::PixelUnpackState &unpack, GLenum type, const void *input); + virtual gl::Error loadCompressedData(const gl::Box &area, const void *input); + + virtual gl::Error copy(const gl::Offset &destOffset, const gl::Rectangle &sourceArea, RenderTargetD3D *source); + virtual gl::Error copy(const gl::Offset &destOffset, const gl::Box &sourceArea, + const gl::ImageIndex &sourceIndex, TextureStorage *source); + + gl::Error recoverFromAssociatedStorage(); + bool isAssociatedStorageValid(TextureStorage11* textureStorage) const; + void disassociateStorage(); + + protected: + gl::Error map(D3D11_MAP mapType, D3D11_MAPPED_SUBRESOURCE *map); + void unmap(); + + private: + gl::Error copyToStorageImpl(TextureStorage11 *storage11, const gl::ImageIndex &index, const gl::Box ®ion); + gl::Error copy(const gl::Offset &destOffset, const gl::Box &sourceArea, ID3D11Resource *source, UINT sourceSubResource); + + gl::Error getStagingTexture(ID3D11Resource **outStagingTexture, unsigned int *outSubresourceIndex); + gl::Error createStagingTexture(); + void releaseStagingTexture(); + + Renderer11 *mRenderer; + D3D_FEATURE_LEVEL mFeatureLevel; + + DXGI_FORMAT mDXGIFormat; + ID3D11Resource *mStagingTexture; + unsigned int mStagingSubresource; + + bool mRecoverFromStorage; + TextureStorage11 *mAssociatedStorage; + gl::ImageIndex mAssociatedImageIndex; + unsigned int mRecoveredFromStorageCount; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_IMAGE11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.cpp new file mode 100644 index 0000000000..99c199f2b9 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.cpp @@ -0,0 +1,162 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer11.cpp: Defines the D3D11 IndexBuffer implementation. + +#include "libANGLE/renderer/d3d/d3d11/IndexBuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" + +namespace rx +{ + +IndexBuffer11::IndexBuffer11(Renderer11 *const renderer) : mRenderer(renderer) +{ + mBuffer = NULL; + mBufferSize = 0; + mDynamicUsage = false; +} + +IndexBuffer11::~IndexBuffer11() +{ + SafeRelease(mBuffer); +} + +gl::Error IndexBuffer11::initialize(unsigned int bufferSize, GLenum indexType, bool dynamic) +{ + SafeRelease(mBuffer); + + updateSerial(); + + if (bufferSize > 0) + { + ID3D11Device* dxDevice = mRenderer->getDevice(); + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = bufferSize; + bufferDesc.Usage = D3D11_USAGE_DYNAMIC; + bufferDesc.BindFlags = D3D11_BIND_INDEX_BUFFER; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + HRESULT result = dxDevice->CreateBuffer(&bufferDesc, NULL, &mBuffer); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate internal index buffer of size, %lu.", bufferSize); + } + } + + mBufferSize = bufferSize; + mIndexType = indexType; + mDynamicUsage = dynamic; + + return gl::Error(GL_NO_ERROR); +} + +IndexBuffer11 *IndexBuffer11::makeIndexBuffer11(IndexBuffer *indexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(IndexBuffer11*, indexBuffer)); + return static_cast<IndexBuffer11*>(indexBuffer); +} + +gl::Error IndexBuffer11::mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory) +{ + if (!mBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal index buffer is not initialized."); + } + + // Check for integer overflows and out-out-bounds map requests + if (offset + size < offset || offset + size > mBufferSize) + { + return gl::Error(GL_OUT_OF_MEMORY, "Index buffer map range is not inside the buffer."); + } + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_NO_OVERWRITE, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal index buffer, HRESULT: 0x%08x.", result); + } + + *outMappedMemory = reinterpret_cast<char*>(mappedResource.pData) + offset; + return gl::Error(GL_NO_ERROR); +} + +gl::Error IndexBuffer11::unmapBuffer() +{ + if (!mBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal index buffer is not initialized."); + } + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + dxContext->Unmap(mBuffer, 0); + return gl::Error(GL_NO_ERROR); +} + +GLenum IndexBuffer11::getIndexType() const +{ + return mIndexType; +} + +unsigned int IndexBuffer11::getBufferSize() const +{ + return mBufferSize; +} + +gl::Error IndexBuffer11::setSize(unsigned int bufferSize, GLenum indexType) +{ + if (bufferSize > mBufferSize || indexType != mIndexType) + { + return initialize(bufferSize, indexType, mDynamicUsage); + } + else + { + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error IndexBuffer11::discard() +{ + if (!mBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal index buffer is not initialized."); + } + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal index buffer, HRESULT: 0x%08x.", result); + } + + dxContext->Unmap(mBuffer, 0); + + return gl::Error(GL_NO_ERROR); +} + +DXGI_FORMAT IndexBuffer11::getIndexFormat() const +{ + switch (mIndexType) + { + case GL_UNSIGNED_BYTE: return DXGI_FORMAT_R16_UINT; + case GL_UNSIGNED_SHORT: return DXGI_FORMAT_R16_UINT; + case GL_UNSIGNED_INT: return DXGI_FORMAT_R32_UINT; + default: UNREACHABLE(); return DXGI_FORMAT_UNKNOWN; + } +} + +ID3D11Buffer *IndexBuffer11::getBuffer() const +{ + return mBuffer; +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.h new file mode 100644 index 0000000000..eadd03eb76 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/IndexBuffer11.h @@ -0,0 +1,51 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// IndexBuffer11.h: Defines the D3D11 IndexBuffer implementation. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_INDEXBUFFER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_INDEXBUFFER11_H_ + +#include "libANGLE/renderer/d3d/IndexBuffer.h" + +namespace rx +{ +class Renderer11; + +class IndexBuffer11 : public IndexBuffer +{ + public: + explicit IndexBuffer11(Renderer11 *const renderer); + virtual ~IndexBuffer11(); + + virtual gl::Error initialize(unsigned int bufferSize, GLenum indexType, bool dynamic); + + static IndexBuffer11 *makeIndexBuffer11(IndexBuffer *indexBuffer); + + virtual gl::Error mapBuffer(unsigned int offset, unsigned int size, void** outMappedMemory); + virtual gl::Error unmapBuffer(); + + virtual GLenum getIndexType() const; + virtual unsigned int getBufferSize() const; + virtual gl::Error setSize(unsigned int bufferSize, GLenum indexType); + + virtual gl::Error discard(); + + DXGI_FORMAT getIndexFormat() const; + ID3D11Buffer *getBuffer() const; + + private: + Renderer11 *const mRenderer; + + ID3D11Buffer *mBuffer; + unsigned int mBufferSize; + GLenum mIndexType; + bool mDynamicUsage; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_INDEXBUFFER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.cpp new file mode 100644 index 0000000000..242c09d6ce --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.cpp @@ -0,0 +1,430 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InputLayoutCache.cpp: Defines InputLayoutCache, a class that builds and caches +// D3D11 input layouts. + +#include "libANGLE/renderer/d3d/d3d11/InputLayoutCache.h" +#include "libANGLE/renderer/d3d/d3d11/VertexBuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" +#include "libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/ProgramD3D.h" +#include "libANGLE/renderer/d3d/VertexDataManager.h" +#include "libANGLE/Program.h" +#include "libANGLE/VertexAttribute.h" + +#include "third_party/murmurhash/MurmurHash3.h" + +namespace rx +{ + +static void GetInputLayout(const TranslatedAttribute translatedAttributes[gl::MAX_VERTEX_ATTRIBS], + gl::VertexFormat inputLayout[gl::MAX_VERTEX_ATTRIBS]) +{ + for (unsigned int attributeIndex = 0; attributeIndex < gl::MAX_VERTEX_ATTRIBS; attributeIndex++) + { + const TranslatedAttribute &translatedAttribute = translatedAttributes[attributeIndex]; + + if (translatedAttributes[attributeIndex].active) + { + inputLayout[attributeIndex] = gl::VertexFormat(*translatedAttribute.attribute, + translatedAttribute.currentValueType); + } + } +} + +const unsigned int InputLayoutCache::kMaxInputLayouts = 1024; + +InputLayoutCache::InputLayoutCache() : mInputLayoutMap(kMaxInputLayouts, hashInputLayout, compareInputLayouts) +{ + mCounter = 0; + mDevice = NULL; + mDeviceContext = NULL; + mCurrentIL = NULL; + for (unsigned int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mCurrentBuffers[i] = NULL; + mCurrentVertexStrides[i] = static_cast<UINT>(-1); + mCurrentVertexOffsets[i] = static_cast<UINT>(-1); + } + mPointSpriteVertexBuffer = NULL; + mPointSpriteIndexBuffer = NULL; +} + +InputLayoutCache::~InputLayoutCache() +{ + clear(); +} + +void InputLayoutCache::initialize(ID3D11Device *device, ID3D11DeviceContext *context) +{ + clear(); + mDevice = device; + mDeviceContext = context; + mFeatureLevel = device->GetFeatureLevel(); +} + +void InputLayoutCache::clear() +{ + for (InputLayoutMap::iterator i = mInputLayoutMap.begin(); i != mInputLayoutMap.end(); i++) + { + SafeRelease(i->second.inputLayout); + } + mInputLayoutMap.clear(); + SafeRelease(mPointSpriteVertexBuffer); + SafeRelease(mPointSpriteIndexBuffer); + markDirty(); +} + +void InputLayoutCache::markDirty() +{ + mCurrentIL = NULL; + for (unsigned int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + mCurrentBuffers[i] = NULL; + mCurrentVertexStrides[i] = static_cast<UINT>(-1); + mCurrentVertexOffsets[i] = static_cast<UINT>(-1); + } +} + +gl::Error InputLayoutCache::applyVertexBuffers(TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS], + GLenum mode, gl::Program *program) +{ + ProgramD3D *programD3D = GetImplAs<ProgramD3D>(program); + + int sortedSemanticIndices[gl::MAX_VERTEX_ATTRIBS]; + programD3D->sortAttributesByLayout(attributes, sortedSemanticIndices); + bool programUsesInstancedPointSprites = programD3D->usesPointSize() && programD3D->usesInstancedPointSpriteEmulation(); + bool instancedPointSpritesActive = programUsesInstancedPointSprites && (mode == GL_POINTS); + + if (!mDevice || !mDeviceContext) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal input layout cache is not initialized."); + } + + InputLayoutKey ilKey = { 0 }; + + static const char* semanticName = "TEXCOORD"; + + unsigned int firstIndexedElement = gl::MAX_VERTEX_ATTRIBS; + unsigned int firstInstancedElement = gl::MAX_VERTEX_ATTRIBS; + unsigned int nextAvailableInputSlot = 0; + + for (unsigned int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + if (attributes[i].active) + { + D3D11_INPUT_CLASSIFICATION inputClass = attributes[i].divisor > 0 ? D3D11_INPUT_PER_INSTANCE_DATA : D3D11_INPUT_PER_VERTEX_DATA; + // If rendering points and instanced pointsprite emulation is being used, the inputClass is required to be configured as per instance data + inputClass = instancedPointSpritesActive ? D3D11_INPUT_PER_INSTANCE_DATA : inputClass; + + gl::VertexFormat vertexFormat(*attributes[i].attribute, attributes[i].currentValueType); + const d3d11::VertexFormat &vertexFormatInfo = d3d11::GetVertexFormatInfo(vertexFormat, mFeatureLevel); + + // Record the type of the associated vertex shader vector in our key + // This will prevent mismatched vertex shaders from using the same input layout + GLint attributeSize; + program->getActiveAttribute(ilKey.elementCount, 0, NULL, &attributeSize, &ilKey.elements[ilKey.elementCount].glslElementType, NULL); + + ilKey.elements[ilKey.elementCount].desc.SemanticName = semanticName; + ilKey.elements[ilKey.elementCount].desc.SemanticIndex = sortedSemanticIndices[i]; + ilKey.elements[ilKey.elementCount].desc.Format = vertexFormatInfo.nativeFormat; + ilKey.elements[ilKey.elementCount].desc.InputSlot = i; + ilKey.elements[ilKey.elementCount].desc.AlignedByteOffset = 0; + ilKey.elements[ilKey.elementCount].desc.InputSlotClass = inputClass; + ilKey.elements[ilKey.elementCount].desc.InstanceDataStepRate = instancedPointSpritesActive ? 1 : attributes[i].divisor; + + if (inputClass == D3D11_INPUT_PER_VERTEX_DATA && firstIndexedElement == gl::MAX_VERTEX_ATTRIBS) + { + firstIndexedElement = ilKey.elementCount; + } + else if (inputClass == D3D11_INPUT_PER_INSTANCE_DATA && firstInstancedElement == gl::MAX_VERTEX_ATTRIBS) + { + firstInstancedElement = ilKey.elementCount; + } + + ilKey.elementCount++; + nextAvailableInputSlot = i + 1; + } + } + + // Instanced PointSprite emulation requires additional entries in the + // inputlayout to support the vertices that make up the pointsprite quad. + // We do this even if mode != GL_POINTS, since the shader signature has these inputs, and the input layout must match the shader + if (programUsesInstancedPointSprites) + { + ilKey.elements[ilKey.elementCount].desc.SemanticName = "SPRITEPOSITION"; + ilKey.elements[ilKey.elementCount].desc.SemanticIndex = 0; + ilKey.elements[ilKey.elementCount].desc.Format = DXGI_FORMAT_R32G32B32_FLOAT; + ilKey.elements[ilKey.elementCount].desc.InputSlot = nextAvailableInputSlot; + ilKey.elements[ilKey.elementCount].desc.AlignedByteOffset = 0; + ilKey.elements[ilKey.elementCount].desc.InputSlotClass = D3D11_INPUT_PER_VERTEX_DATA; + ilKey.elements[ilKey.elementCount].desc.InstanceDataStepRate = 0; + + // The new elements are D3D11_INPUT_PER_VERTEX_DATA data so the indexed element + // tracking must be applied. This ensures that the instancing specific + // buffer swapping logic continues to work. + if (firstIndexedElement == gl::MAX_VERTEX_ATTRIBS) + { + firstIndexedElement = ilKey.elementCount; + } + + ilKey.elementCount++; + + ilKey.elements[ilKey.elementCount].desc.SemanticName = "SPRITETEXCOORD"; + ilKey.elements[ilKey.elementCount].desc.SemanticIndex = 0; + ilKey.elements[ilKey.elementCount].desc.Format = DXGI_FORMAT_R32G32_FLOAT; + ilKey.elements[ilKey.elementCount].desc.InputSlot = nextAvailableInputSlot; + ilKey.elements[ilKey.elementCount].desc.AlignedByteOffset = sizeof(float) * 3; + ilKey.elements[ilKey.elementCount].desc.InputSlotClass = D3D11_INPUT_PER_VERTEX_DATA; + ilKey.elements[ilKey.elementCount].desc.InstanceDataStepRate = 0; + + ilKey.elementCount++; + } + + // On 9_3, we must ensure that slot 0 contains non-instanced data. + // If slot 0 currently contains instanced data then we swap it with a non-instanced element. + // Note that instancing is only available on 9_3 via ANGLE_instanced_arrays, since 9_3 doesn't support OpenGL ES 3.0. + // As per the spec for ANGLE_instanced_arrays, not all attributes can be instanced simultaneously, so a non-instanced element must exist. + ASSERT(!(mFeatureLevel <= D3D_FEATURE_LEVEL_9_3 && firstIndexedElement == gl::MAX_VERTEX_ATTRIBS)); + bool moveFirstIndexedIntoSlotZero = mFeatureLevel <= D3D_FEATURE_LEVEL_9_3 && firstInstancedElement == 0 && firstIndexedElement != gl::MAX_VERTEX_ATTRIBS; + + if (moveFirstIndexedIntoSlotZero) + { + ilKey.elements[firstInstancedElement].desc.InputSlot = ilKey.elements[firstIndexedElement].desc.InputSlot; + ilKey.elements[firstIndexedElement].desc.InputSlot = 0; + + // Instanced PointSprite emulation uses multiple layout entries across a single vertex buffer. + // If an index swap is performed, we need to ensure that all elements get the proper InputSlot. + if (programUsesInstancedPointSprites) + { + ilKey.elements[firstIndexedElement + 1].desc.InputSlot = 0; + } + } + + ID3D11InputLayout *inputLayout = NULL; + + InputLayoutMap::iterator keyIter = mInputLayoutMap.find(ilKey); + if (keyIter != mInputLayoutMap.end()) + { + inputLayout = keyIter->second.inputLayout; + keyIter->second.lastUsedTime = mCounter++; + } + else + { + gl::VertexFormat shaderInputLayout[gl::MAX_VERTEX_ATTRIBS]; + GetInputLayout(attributes, shaderInputLayout); + + ShaderExecutableD3D *shader = NULL; + gl::Error error = programD3D->getVertexExecutableForInputLayout(shaderInputLayout, &shader, nullptr); + if (error.isError()) + { + return error; + } + + ShaderExecutableD3D *shader11 = ShaderExecutable11::makeShaderExecutable11(shader); + + D3D11_INPUT_ELEMENT_DESC descs[gl::MAX_VERTEX_ATTRIBS]; + for (unsigned int j = 0; j < ilKey.elementCount; ++j) + { + descs[j] = ilKey.elements[j].desc; + } + + HRESULT result = mDevice->CreateInputLayout(descs, ilKey.elementCount, shader11->getFunction(), shader11->getLength(), &inputLayout); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal input layout, HRESULT: 0x%08x", result); + } + + if (mInputLayoutMap.size() >= kMaxInputLayouts) + { + TRACE("Overflowed the limit of %u input layouts, removing the least recently used " + "to make room.", kMaxInputLayouts); + + InputLayoutMap::iterator leastRecentlyUsed = mInputLayoutMap.begin(); + for (InputLayoutMap::iterator i = mInputLayoutMap.begin(); i != mInputLayoutMap.end(); i++) + { + if (i->second.lastUsedTime < leastRecentlyUsed->second.lastUsedTime) + { + leastRecentlyUsed = i; + } + } + SafeRelease(leastRecentlyUsed->second.inputLayout); + mInputLayoutMap.erase(leastRecentlyUsed); + } + + InputLayoutCounterPair inputCounterPair; + inputCounterPair.inputLayout = inputLayout; + inputCounterPair.lastUsedTime = mCounter++; + + mInputLayoutMap.insert(std::make_pair(ilKey, inputCounterPair)); + } + + if (inputLayout != mCurrentIL) + { + mDeviceContext->IASetInputLayout(inputLayout); + mCurrentIL = inputLayout; + } + + bool dirtyBuffers = false; + size_t minDiff = gl::MAX_VERTEX_ATTRIBS; + size_t maxDiff = 0; + unsigned int nextAvailableIndex = 0; + + for (unsigned int i = 0; i < gl::MAX_VERTEX_ATTRIBS; i++) + { + ID3D11Buffer *buffer = NULL; + + if (attributes[i].active) + { + VertexBuffer11 *vertexBuffer = VertexBuffer11::makeVertexBuffer11(attributes[i].vertexBuffer); + Buffer11 *bufferStorage = attributes[i].storage ? Buffer11::makeBuffer11(attributes[i].storage) : NULL; + + buffer = bufferStorage ? bufferStorage->getBuffer(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK) + : vertexBuffer->getBuffer(); + } + + UINT vertexStride = attributes[i].stride; + UINT vertexOffset = attributes[i].offset; + + if (buffer != mCurrentBuffers[i] || vertexStride != mCurrentVertexStrides[i] || + vertexOffset != mCurrentVertexOffsets[i]) + { + dirtyBuffers = true; + minDiff = std::min(minDiff, static_cast<size_t>(i)); + maxDiff = std::max(maxDiff, static_cast<size_t>(i)); + + mCurrentBuffers[i] = buffer; + mCurrentVertexStrides[i] = vertexStride; + mCurrentVertexOffsets[i] = vertexOffset; + + // If a non null ID3D11Buffer is being assigned to mCurrentBuffers, + // then the next available index needs to be tracked to ensure + // that any instanced pointsprite emulation buffers will be properly packed. + if (buffer) + { + nextAvailableIndex = i + 1; + } + } + } + + // Instanced PointSprite emulation requires two additional ID3D11Buffers. + // A vertex buffer needs to be created and added to the list of current buffers, + // strides and offsets collections. This buffer contains the vertices for a single + // PointSprite quad. + // An index buffer also needs to be created and applied because rendering instanced + // data on D3D11 FL9_3 requires DrawIndexedInstanced() to be used. + if (instancedPointSpritesActive) + { + HRESULT result = S_OK; + const UINT pointSpriteVertexStride = sizeof(float) * 5; + + if (!mPointSpriteVertexBuffer) + { + static const float pointSpriteVertices[] = + { + // Position // TexCoord + -1.0f, -1.0f, 0.0f, 0.0f, 1.0f, + -1.0f, 1.0f, 0.0f, 0.0f, 0.0f, + 1.0f, 1.0f, 0.0f, 1.0f, 0.0f, + 1.0f, -1.0f, 0.0f, 1.0f, 1.0f, + -1.0f, -1.0f, 0.0f, 0.0f, 1.0f, + 1.0f, 1.0f, 0.0f, 1.0f, 0.0f, + }; + + D3D11_SUBRESOURCE_DATA vertexBufferData = { pointSpriteVertices, 0, 0 }; + D3D11_BUFFER_DESC vertexBufferDesc; + vertexBufferDesc.ByteWidth = sizeof(pointSpriteVertices); + vertexBufferDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vertexBufferDesc.Usage = D3D11_USAGE_IMMUTABLE; + vertexBufferDesc.CPUAccessFlags = 0; + vertexBufferDesc.MiscFlags = 0; + vertexBufferDesc.StructureByteStride = 0; + + result = mDevice->CreateBuffer(&vertexBufferDesc, &vertexBufferData, &mPointSpriteVertexBuffer); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create instanced pointsprite emulation vertex buffer, HRESULT: 0x%08x", result); + } + } + + mCurrentBuffers[nextAvailableIndex] = mPointSpriteVertexBuffer; + mCurrentVertexStrides[nextAvailableIndex] = pointSpriteVertexStride; + mCurrentVertexOffsets[nextAvailableIndex] = 0; + + if (!mPointSpriteIndexBuffer) + { + // Create an index buffer and set it for pointsprite rendering + static const unsigned short pointSpriteIndices[] = + { + 0, 1, 2, 3, 4, 5, + }; + + D3D11_SUBRESOURCE_DATA indexBufferData = { pointSpriteIndices, 0, 0 }; + D3D11_BUFFER_DESC indexBufferDesc; + indexBufferDesc.ByteWidth = sizeof(pointSpriteIndices); + indexBufferDesc.BindFlags = D3D11_BIND_INDEX_BUFFER; + indexBufferDesc.Usage = D3D11_USAGE_IMMUTABLE; + indexBufferDesc.CPUAccessFlags = 0; + indexBufferDesc.MiscFlags = 0; + indexBufferDesc.StructureByteStride = 0; + + result = mDevice->CreateBuffer(&indexBufferDesc, &indexBufferData, &mPointSpriteIndexBuffer); + if (FAILED(result)) + { + SafeRelease(mPointSpriteVertexBuffer); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create instanced pointsprite emulation index buffer, HRESULT: 0x%08x", result); + } + } + + // The index buffer is applied here because Instanced PointSprite emulation uses + // the a non-indexed rendering path in ANGLE (DrawArrays). This means that applyIndexBuffer() + // on the renderer will not be called and setting this buffer here ensures that the rendering + // path will contain the correct index buffers. + mDeviceContext->IASetIndexBuffer(mPointSpriteIndexBuffer, DXGI_FORMAT_R16_UINT, 0); + } + + if (moveFirstIndexedIntoSlotZero) + { + // In this case, we swapped the slots of the first instanced element and the first indexed element, to ensure + // that the first slot contains non-instanced data (required by Feature Level 9_3). + // We must also swap the corresponding buffers sent to IASetVertexBuffers so that the correct data is sent to each slot. + std::swap(mCurrentBuffers[firstIndexedElement], mCurrentBuffers[firstInstancedElement]); + std::swap(mCurrentVertexStrides[firstIndexedElement], mCurrentVertexStrides[firstInstancedElement]); + std::swap(mCurrentVertexOffsets[firstIndexedElement], mCurrentVertexOffsets[firstInstancedElement]); + } + + if (dirtyBuffers) + { + ASSERT(minDiff <= maxDiff && maxDiff < gl::MAX_VERTEX_ATTRIBS); + mDeviceContext->IASetVertexBuffers(minDiff, maxDiff - minDiff + 1, mCurrentBuffers + minDiff, + mCurrentVertexStrides + minDiff, mCurrentVertexOffsets + minDiff); + } + + return gl::Error(GL_NO_ERROR); +} + +std::size_t InputLayoutCache::hashInputLayout(const InputLayoutKey &inputLayout) +{ + static const unsigned int seed = 0xDEADBEEF; + + std::size_t hash = 0; + MurmurHash3_x86_32(inputLayout.begin(), inputLayout.end() - inputLayout.begin(), seed, &hash); + return hash; +} + +bool InputLayoutCache::compareInputLayouts(const InputLayoutKey &a, const InputLayoutKey &b) +{ + if (a.elementCount != b.elementCount) + { + return false; + } + + return std::equal(a.begin(), a.end(), b.begin()); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.h new file mode 100644 index 0000000000..2c94c57595 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/InputLayoutCache.h @@ -0,0 +1,103 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InputLayoutCache.h: Defines InputLayoutCache, a class that builds and caches +// D3D11 input layouts. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_INPUTLAYOUTCACHE_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_INPUTLAYOUTCACHE_H_ + +#include "libANGLE/Constants.h" +#include "libANGLE/Error.h" +#include "common/angleutils.h" + +#include <GLES2/gl2.h> + +#include <cstddef> +#include <unordered_map> + +namespace gl +{ +class Program; +} + +namespace rx +{ +struct TranslatedAttribute; + +class InputLayoutCache : angle::NonCopyable +{ + public: + InputLayoutCache(); + virtual ~InputLayoutCache(); + + void initialize(ID3D11Device *device, ID3D11DeviceContext *context); + void clear(); + void markDirty(); + + gl::Error applyVertexBuffers(TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS], + GLenum mode, gl::Program *program); + + private: + struct InputLayoutElement + { + D3D11_INPUT_ELEMENT_DESC desc; + GLenum glslElementType; + }; + + struct InputLayoutKey + { + unsigned int elementCount; + InputLayoutElement elements[gl::MAX_VERTEX_ATTRIBS]; + + const char *begin() const + { + return reinterpret_cast<const char*>(&elementCount); + } + + const char *end() const + { + return reinterpret_cast<const char*>(&elements[elementCount]); + } + }; + + struct InputLayoutCounterPair + { + ID3D11InputLayout *inputLayout; + unsigned long long lastUsedTime; + }; + + ID3D11InputLayout *mCurrentIL; + ID3D11Buffer *mCurrentBuffers[gl::MAX_VERTEX_ATTRIBS]; + UINT mCurrentVertexStrides[gl::MAX_VERTEX_ATTRIBS]; + UINT mCurrentVertexOffsets[gl::MAX_VERTEX_ATTRIBS]; + + ID3D11Buffer *mPointSpriteVertexBuffer; + ID3D11Buffer *mPointSpriteIndexBuffer; + + static std::size_t hashInputLayout(const InputLayoutKey &inputLayout); + static bool compareInputLayouts(const InputLayoutKey &a, const InputLayoutKey &b); + + typedef std::size_t (*InputLayoutHashFunction)(const InputLayoutKey &); + typedef bool (*InputLayoutEqualityFunction)(const InputLayoutKey &, const InputLayoutKey &); + typedef std::unordered_map<InputLayoutKey, + InputLayoutCounterPair, + InputLayoutHashFunction, + InputLayoutEqualityFunction> InputLayoutMap; + InputLayoutMap mInputLayoutMap; + + static const unsigned int kMaxInputLayouts; + + unsigned long long mCounter; + + ID3D11Device *mDevice; + ID3D11DeviceContext *mDeviceContext; + D3D_FEATURE_LEVEL mFeatureLevel; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_INPUTLAYOUTCACHE_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/NativeWindow.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/NativeWindow.h new file mode 100644 index 0000000000..81b9ea748d --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/NativeWindow.h @@ -0,0 +1,80 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// NativeWindow.h: Defines NativeWindow, a class for managing and +// performing operations on an EGLNativeWindowType. +// It is used for HWND (Desktop Windows) and IInspectable objects +//(Windows Store Applications). + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_NATIVEWINDOW_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_NATIVEWINDOW_H_ + +#include "common/debug.h" +#include "common/platform.h" + +#include <EGL/eglplatform.h> + +// DXGISwapChain and DXGIFactory are typedef'd to specific required +// types. The HWND NativeWindow implementation requires IDXGISwapChain +// and IDXGIFactory and the Windows Store NativeWindow +// implementation requires IDXGISwapChain1 and IDXGIFactory2. +#if defined(ANGLE_ENABLE_WINDOWS_STORE) +typedef IDXGISwapChain1 DXGISwapChain; +typedef IDXGIFactory2 DXGIFactory; + +#include <wrl.h> +#include <wrl/wrappers/corewrappers.h> +#include <windows.applicationmodel.core.h> +#include <memory> + +namespace rx +{ +class InspectableNativeWindow; +} + +using namespace Microsoft::WRL; +using namespace Microsoft::WRL::Wrappers; + +#else +typedef IDXGISwapChain DXGISwapChain; +typedef IDXGIFactory DXGIFactory; +#endif + +namespace rx +{ + +class NativeWindow +{ + public: + enum RotationFlags { RotateNone = 0, RotateLeft = 1, RotateRight = 2 }; + explicit NativeWindow(EGLNativeWindowType window); + + bool initialize(); + bool getClientRect(LPRECT rect); + bool isIconic(); +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + RotationFlags rotationFlags() const; +#endif + static bool isValidNativeWindow(EGLNativeWindowType window); + + HRESULT createSwapChain(ID3D11Device* device, DXGIFactory* factory, + DXGI_FORMAT format, UINT width, UINT height, + DXGISwapChain** swapChain); + + inline EGLNativeWindowType getNativeWindow() const { return mWindow; } + + private: + EGLNativeWindowType mWindow; + +#if defined(ANGLE_ENABLE_WINDOWS_STORE) + std::shared_ptr<InspectableNativeWindow> mImpl; +#endif + +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_NATIVEWINDOW_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.cpp new file mode 100644 index 0000000000..5fd5237d90 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.cpp @@ -0,0 +1,302 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// PixelTransfer11.cpp: +// Implementation for buffer-to-texture and texture-to-buffer copies. +// Used to implement pixel transfers from unpack and to pack buffers. +// + +#include "libANGLE/renderer/d3d/d3d11/PixelTransfer11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" +#include "libANGLE/renderer/d3d/d3d11/TextureStorage11.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/formatutils.h" +#include "libANGLE/Texture.h" +#include "libANGLE/Buffer.h" +#include "libANGLE/Context.h" + +// Precompiled shaders +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/buffertotexture11_vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/buffertotexture11_gs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/buffertotexture11_ps_4f.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/buffertotexture11_ps_4i.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/buffertotexture11_ps_4ui.h" + +namespace rx +{ + +PixelTransfer11::PixelTransfer11(Renderer11 *renderer) + : mRenderer(renderer), + mResourcesLoaded(false), + mBufferToTextureVS(NULL), + mBufferToTextureGS(NULL), + mParamsConstantBuffer(NULL), + mCopyRasterizerState(NULL), + mCopyDepthStencilState(NULL) +{ +} + +PixelTransfer11::~PixelTransfer11() +{ + for (auto shaderMapIt = mBufferToTexturePSMap.begin(); shaderMapIt != mBufferToTexturePSMap.end(); shaderMapIt++) + { + SafeRelease(shaderMapIt->second); + } + + mBufferToTexturePSMap.clear(); + + SafeRelease(mBufferToTextureVS); + SafeRelease(mBufferToTextureGS); + SafeRelease(mParamsConstantBuffer); + SafeRelease(mCopyRasterizerState); + SafeRelease(mCopyDepthStencilState); +} + +gl::Error PixelTransfer11::loadResources() +{ + if (mResourcesLoaded) + { + return gl::Error(GL_NO_ERROR); + } + + HRESULT result = S_OK; + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RASTERIZER_DESC rasterDesc; + rasterDesc.FillMode = D3D11_FILL_SOLID; + rasterDesc.CullMode = D3D11_CULL_NONE; + rasterDesc.FrontCounterClockwise = FALSE; + rasterDesc.DepthBias = 0; + rasterDesc.SlopeScaledDepthBias = 0.0f; + rasterDesc.DepthBiasClamp = 0.0f; + rasterDesc.DepthClipEnable = TRUE; + rasterDesc.ScissorEnable = FALSE; + rasterDesc.MultisampleEnable = FALSE; + rasterDesc.AntialiasedLineEnable = FALSE; + + result = device->CreateRasterizerState(&rasterDesc, &mCopyRasterizerState); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal pixel transfer rasterizer state, result: 0x%X.", result); + } + + D3D11_DEPTH_STENCIL_DESC depthStencilDesc; + depthStencilDesc.DepthEnable = true; + depthStencilDesc.DepthWriteMask = D3D11_DEPTH_WRITE_MASK_ALL; + depthStencilDesc.DepthFunc = D3D11_COMPARISON_ALWAYS; + depthStencilDesc.StencilEnable = FALSE; + depthStencilDesc.StencilReadMask = D3D11_DEFAULT_STENCIL_READ_MASK; + depthStencilDesc.StencilWriteMask = D3D11_DEFAULT_STENCIL_WRITE_MASK; + depthStencilDesc.FrontFace.StencilFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilDepthFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilPassOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.FrontFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + depthStencilDesc.BackFace.StencilFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilDepthFailOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilPassOp = D3D11_STENCIL_OP_KEEP; + depthStencilDesc.BackFace.StencilFunc = D3D11_COMPARISON_ALWAYS; + + result = device->CreateDepthStencilState(&depthStencilDesc, &mCopyDepthStencilState); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal pixel transfer depth stencil state, result: 0x%X.", result); + } + + D3D11_BUFFER_DESC constantBufferDesc = { 0 }; + constantBufferDesc.ByteWidth = roundUp<UINT>(sizeof(CopyShaderParams), 32u); + constantBufferDesc.Usage = D3D11_USAGE_DYNAMIC; + constantBufferDesc.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + constantBufferDesc.MiscFlags = 0; + constantBufferDesc.StructureByteStride = 0; + + result = device->CreateBuffer(&constantBufferDesc, NULL, &mParamsConstantBuffer); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal pixel transfer constant buffer, result: 0x%X.", result); + } + d3d11::SetDebugName(mParamsConstantBuffer, "PixelTransfer11 constant buffer"); + + // init shaders + mBufferToTextureVS = d3d11::CompileVS(device, g_VS_BufferToTexture, "BufferToTexture VS"); + if (!mBufferToTextureVS) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal buffer to texture vertex shader."); + } + + mBufferToTextureGS = d3d11::CompileGS(device, g_GS_BufferToTexture, "BufferToTexture GS"); + if (!mBufferToTextureGS) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal buffer to texture geometry shader."); + } + + gl::Error error = buildShaderMap(); + if (error.isError()) + { + return error; + } + + StructZero(&mParamsData); + + mResourcesLoaded = true; + + return gl::Error(GL_NO_ERROR); +} + +void PixelTransfer11::setBufferToTextureCopyParams(const gl::Box &destArea, const gl::Extents &destSize, GLenum internalFormat, + const gl::PixelUnpackState &unpack, unsigned int offset, CopyShaderParams *parametersOut) +{ + StructZero(parametersOut); + + float texelCenterX = 0.5f / static_cast<float>(destSize.width - 1); + float texelCenterY = 0.5f / static_cast<float>(destSize.height - 1); + + unsigned int bytesPerPixel = gl::GetInternalFormatInfo(internalFormat).pixelBytes; + unsigned int alignmentBytes = static_cast<unsigned int>(unpack.alignment); + unsigned int alignmentPixels = (alignmentBytes <= bytesPerPixel ? 1 : alignmentBytes / bytesPerPixel); + + parametersOut->FirstPixelOffset = offset / bytesPerPixel; + parametersOut->PixelsPerRow = static_cast<unsigned int>((unpack.rowLength > 0) ? unpack.rowLength : destArea.width); + parametersOut->RowStride = roundUp(parametersOut->PixelsPerRow, alignmentPixels); + parametersOut->RowsPerSlice = static_cast<unsigned int>(destArea.height); + parametersOut->PositionOffset[0] = texelCenterX + (destArea.x / float(destSize.width)) * 2.0f - 1.0f; + parametersOut->PositionOffset[1] = texelCenterY + ((destSize.height - destArea.y - 1) / float(destSize.height)) * 2.0f - 1.0f; + parametersOut->PositionScale[0] = 2.0f / static_cast<float>(destSize.width); + parametersOut->PositionScale[1] = -2.0f / static_cast<float>(destSize.height); + parametersOut->FirstSlice = destArea.z; +} + +gl::Error PixelTransfer11::copyBufferToTexture(const gl::PixelUnpackState &unpack, unsigned int offset, RenderTargetD3D *destRenderTarget, + GLenum destinationFormat, GLenum sourcePixelsType, const gl::Box &destArea) +{ + gl::Error error = loadResources(); + if (error.isError()) + { + return error; + } + + gl::Extents destSize = destRenderTarget->getExtents(); + + ASSERT(destArea.x >= 0 && destArea.x + destArea.width <= destSize.width && + destArea.y >= 0 && destArea.y + destArea.height <= destSize.height && + destArea.z >= 0 && destArea.z + destArea.depth <= destSize.depth ); + + const gl::Buffer &sourceBuffer = *unpack.pixelBuffer.get(); + + ASSERT(mRenderer->supportsFastCopyBufferToTexture(destinationFormat)); + + ID3D11PixelShader *pixelShader = findBufferToTexturePS(destinationFormat); + ASSERT(pixelShader); + + // The SRV must be in the proper read format, which may be different from the destination format + // EG: for half float data, we can load full precision floats with implicit conversion + GLenum unsizedFormat = gl::GetInternalFormatInfo(destinationFormat).format; + GLenum sourceFormat = gl::GetSizedInternalFormat(unsizedFormat, sourcePixelsType); + + const d3d11::TextureFormat &sourceFormatInfo = d3d11::GetTextureFormatInfo(sourceFormat, mRenderer->getFeatureLevel()); + DXGI_FORMAT srvFormat = sourceFormatInfo.srvFormat; + ASSERT(srvFormat != DXGI_FORMAT_UNKNOWN); + Buffer11 *bufferStorage11 = Buffer11::makeBuffer11(sourceBuffer.getImplementation()); + ID3D11ShaderResourceView *bufferSRV = bufferStorage11->getSRV(srvFormat); + ASSERT(bufferSRV != NULL); + + ID3D11RenderTargetView *textureRTV = RenderTarget11::makeRenderTarget11(destRenderTarget)->getRenderTargetView(); + ASSERT(textureRTV != NULL); + + CopyShaderParams shaderParams; + setBufferToTextureCopyParams(destArea, destSize, sourceFormat, unpack, offset, &shaderParams); + + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + ID3D11Buffer *nullBuffer = NULL; + UINT zero = 0; + + // Are we doing a 2D or 3D copy? + ID3D11GeometryShader *geometryShader = ((destSize.depth > 1) ? mBufferToTextureGS : NULL); + + deviceContext->VSSetShader(mBufferToTextureVS, NULL, 0); + deviceContext->GSSetShader(geometryShader, NULL, 0); + deviceContext->PSSetShader(pixelShader, NULL, 0); + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, bufferSRV); + deviceContext->IASetInputLayout(NULL); + deviceContext->IASetPrimitiveTopology(D3D11_PRIMITIVE_TOPOLOGY_POINTLIST); + + deviceContext->IASetVertexBuffers(0, 1, &nullBuffer, &zero, &zero); + deviceContext->OMSetBlendState(NULL, NULL, 0xFFFFFFF); + deviceContext->OMSetDepthStencilState(mCopyDepthStencilState, 0xFFFFFFFF); + deviceContext->RSSetState(mCopyRasterizerState); + + mRenderer->setOneTimeRenderTarget(textureRTV); + + if (!StructEquals(mParamsData, shaderParams)) + { + d3d11::SetBufferData(deviceContext, mParamsConstantBuffer, shaderParams); + mParamsData = shaderParams; + } + + deviceContext->VSSetConstantBuffers(0, 1, &mParamsConstantBuffer); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; + viewport.Width = destSize.width; + viewport.Height = destSize.height; + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + UINT numPixels = (destArea.width * destArea.height * destArea.depth); + deviceContext->Draw(numPixels, 0); + + // Unbind textures and render targets and vertex buffer + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + deviceContext->VSSetConstantBuffers(0, 1, &nullBuffer); + + mRenderer->markAllStateDirty(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error PixelTransfer11::buildShaderMap() +{ + ID3D11Device *device = mRenderer->getDevice(); + + mBufferToTexturePSMap[GL_FLOAT] = d3d11::CompilePS(device, g_PS_BufferToTexture_4F, "BufferToTexture RGBA ps"); + mBufferToTexturePSMap[GL_INT] = d3d11::CompilePS(device, g_PS_BufferToTexture_4I, "BufferToTexture RGBA-I ps"); + mBufferToTexturePSMap[GL_UNSIGNED_INT] = d3d11::CompilePS(device, g_PS_BufferToTexture_4UI, "BufferToTexture RGBA-UI ps"); + + // Check that all the shaders were created successfully + for (auto shaderMapIt = mBufferToTexturePSMap.begin(); shaderMapIt != mBufferToTexturePSMap.end(); shaderMapIt++) + { + if (shaderMapIt->second == NULL) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal buffer to texture pixel shader."); + } + } + + return gl::Error(GL_NO_ERROR); +} + +ID3D11PixelShader *PixelTransfer11::findBufferToTexturePS(GLenum internalFormat) const +{ + GLenum componentType = gl::GetInternalFormatInfo(internalFormat).componentType; + if (componentType == GL_SIGNED_NORMALIZED || componentType == GL_UNSIGNED_NORMALIZED) + { + componentType = GL_FLOAT; + } + + auto shaderMapIt = mBufferToTexturePSMap.find(componentType); + return (shaderMapIt == mBufferToTexturePSMap.end() ? NULL : shaderMapIt->second); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.h new file mode 100644 index 0000000000..1672121ec7 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/PixelTransfer11.h @@ -0,0 +1,89 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// PixelTransfer11.h: +// Buffer-to-Texture and Texture-to-Buffer data transfers. +// Used to implement pixel unpack and pixel pack buffers in ES3. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_PIXELTRANSFER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_PIXELTRANSFER11_H_ + +#include "libANGLE/Error.h" + +#include "common/platform.h" + +#include <GLES2/gl2.h> + +#include <map> + +namespace gl +{ + +class Buffer; +struct Box; +struct Extents; +struct PixelUnpackState; + +} + +namespace rx +{ +class Renderer11; +class RenderTargetD3D; + +class PixelTransfer11 +{ + public: + explicit PixelTransfer11(Renderer11 *renderer); + ~PixelTransfer11(); + + // unpack: the source buffer is stored in the unpack state, and buffer strides + // offset: the start of the data within the unpack buffer + // destRenderTarget: individual slice/layer of a target texture + // destinationFormat/sourcePixelsType: determines shaders + shader parameters + // destArea: the sub-section of destRenderTarget to copy to + gl::Error copyBufferToTexture(const gl::PixelUnpackState &unpack, unsigned int offset, RenderTargetD3D *destRenderTarget, + GLenum destinationFormat, GLenum sourcePixelsType, const gl::Box &destArea); + + private: + + struct CopyShaderParams + { + unsigned int FirstPixelOffset; + unsigned int PixelsPerRow; + unsigned int RowStride; + unsigned int RowsPerSlice; + float PositionOffset[2]; + float PositionScale[2]; + int TexLocationOffset[2]; + int TexLocationScale[2]; + unsigned int FirstSlice; + }; + + static void setBufferToTextureCopyParams(const gl::Box &destArea, const gl::Extents &destSize, GLenum internalFormat, + const gl::PixelUnpackState &unpack, unsigned int offset, CopyShaderParams *parametersOut); + + gl::Error loadResources(); + gl::Error buildShaderMap(); + ID3D11PixelShader *findBufferToTexturePS(GLenum internalFormat) const; + + Renderer11 *mRenderer; + + bool mResourcesLoaded; + std::map<GLenum, ID3D11PixelShader *> mBufferToTexturePSMap; + ID3D11VertexShader *mBufferToTextureVS; + ID3D11GeometryShader *mBufferToTextureGS; + ID3D11Buffer *mParamsConstantBuffer; + CopyShaderParams mParamsData; + + ID3D11RasterizerState *mCopyRasterizerState; + ID3D11DepthStencilState *mCopyDepthStencilState; + +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_PIXELTRANSFER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.cpp new file mode 100644 index 0000000000..4979ff51a9 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.cpp @@ -0,0 +1,164 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query11.cpp: Defines the rx::Query11 class which implements rx::QueryImpl. + +#include "libANGLE/renderer/d3d/d3d11/Query11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "common/utilities.h" + +#include <GLES2/gl2ext.h> + +#if defined(ANGLE_MINGW32_COMPAT) +typedef struct D3D11_QUERY_DATA_SO_STATISTICS { + UINT64 NumPrimitivesWritten; + UINT64 PrimitivesStorageNeeded; +} D3D11_QUERY_DATA_SO_STATISTICS; +#endif + +namespace rx +{ + +Query11::Query11(Renderer11 *renderer, GLenum type) + : QueryImpl(type), + mResult(0), + mQueryFinished(false), + mRenderer(renderer), + mQuery(NULL) +{ +} + +Query11::~Query11() +{ + SafeRelease(mQuery); +} + +gl::Error Query11::begin() +{ + if (mQuery == NULL) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = gl_d3d11::ConvertQueryType(getType()); + queryDesc.MiscFlags = 0; + + HRESULT result = mRenderer->getDevice()->CreateQuery(&queryDesc, &mQuery); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal query creation failed, result: 0x%X.", result); + } + } + + mRenderer->getDeviceContext()->Begin(mQuery); + return gl::Error(GL_NO_ERROR); +} + +gl::Error Query11::end() +{ + ASSERT(mQuery); + mRenderer->getDeviceContext()->End(mQuery); + + mQueryFinished = false; + mResult = GL_FALSE; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Query11::getResult(GLuint *params) +{ + while (!mQueryFinished) + { + gl::Error error = testQuery(); + if (error.isError()) + { + return error; + } + + if (!mQueryFinished) + { + ScheduleYield(); + } + } + + ASSERT(mQueryFinished); + *params = mResult; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Query11::isResultAvailable(GLuint *available) +{ + gl::Error error = testQuery(); + if (error.isError()) + { + return error; + } + + *available = (mQueryFinished ? GL_TRUE : GL_FALSE); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Query11::testQuery() +{ + if (!mQueryFinished) + { + ASSERT(mQuery); + + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + switch (getType()) + { + case GL_ANY_SAMPLES_PASSED_EXT: + case GL_ANY_SAMPLES_PASSED_CONSERVATIVE_EXT: + { + UINT64 numPixels = 0; + HRESULT result = context->GetData(mQuery, &numPixels, sizeof(numPixels), 0); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to get the data of an internal query, result: 0x%X.", result); + } + + if (result == S_OK) + { + mQueryFinished = true; + mResult = (numPixels > 0) ? GL_TRUE : GL_FALSE; + } + } + break; + + case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN: + { + D3D11_QUERY_DATA_SO_STATISTICS soStats = { 0 }; + HRESULT result = context->GetData(mQuery, &soStats, sizeof(soStats), 0); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to get the data of an internal query, result: 0x%X.", result); + } + + if (result == S_OK) + { + mQueryFinished = true; + mResult = static_cast<GLuint>(soStats.NumPrimitivesWritten); + } + } + break; + + default: + UNREACHABLE(); + break; + } + + if (!mQueryFinished && mRenderer->testDeviceLost()) + { + mRenderer->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to test get query result, device is lost."); + } + } + + return gl::Error(GL_NO_ERROR); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.h new file mode 100644 index 0000000000..bd53fed250 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Query11.h @@ -0,0 +1,42 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Query11.h: Defines the rx::Query11 class which implements rx::QueryImpl. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_QUERY11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_QUERY11_H_ + +#include "libANGLE/renderer/QueryImpl.h" + +namespace rx +{ +class Renderer11; + +class Query11 : public QueryImpl +{ + public: + Query11(Renderer11 *renderer, GLenum type); + virtual ~Query11(); + + virtual gl::Error begin(); + virtual gl::Error end(); + virtual gl::Error getResult(GLuint *params); + virtual gl::Error isResultAvailable(GLuint *available); + + private: + gl::Error testQuery(); + + GLuint mResult; + + bool mQueryFinished; + + Renderer11 *mRenderer; + ID3D11Query *mQuery; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_QUERY11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.cpp new file mode 100644 index 0000000000..4990e6bc6e --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.cpp @@ -0,0 +1,452 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderStateCache.cpp: Defines rx::RenderStateCache, a cache of Direct3D render +// state objects. + +#include "libANGLE/renderer/d3d/d3d11/RenderStateCache.h" + +#include <float.h> + +#include "common/debug.h" +#include "libANGLE/Framebuffer.h" +#include "libANGLE/FramebufferAttachment.h" +#include "libANGLE/renderer/d3d/FramebufferD3D.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "third_party/murmurhash/MurmurHash3.h" + +namespace rx +{ + +template <typename mapType> +static void ClearStateMap(mapType &map) +{ + for (typename mapType::iterator i = map.begin(); i != map.end(); i++) + { + SafeRelease(i->second.first); + } + map.clear(); +} + +// MSDN's documentation of ID3D11Device::CreateBlendState, ID3D11Device::CreateRasterizerState, +// ID3D11Device::CreateDepthStencilState and ID3D11Device::CreateSamplerState claims the maximum +// number of unique states of each type an application can create is 4096 +const unsigned int RenderStateCache::kMaxBlendStates = 4096; +const unsigned int RenderStateCache::kMaxRasterizerStates = 4096; +const unsigned int RenderStateCache::kMaxDepthStencilStates = 4096; +const unsigned int RenderStateCache::kMaxSamplerStates = 4096; + +RenderStateCache::RenderStateCache(Renderer11 *renderer) + : mRenderer(renderer), + mDevice(NULL), + mCounter(0), + mBlendStateCache(kMaxBlendStates, hashBlendState, compareBlendStates), + mRasterizerStateCache(kMaxRasterizerStates, hashRasterizerState, compareRasterizerStates), + mDepthStencilStateCache(kMaxDepthStencilStates, hashDepthStencilState, compareDepthStencilStates), + mSamplerStateCache(kMaxSamplerStates, hashSamplerState, compareSamplerStates) +{ +} + +RenderStateCache::~RenderStateCache() +{ + clear(); +} + +void RenderStateCache::initialize(ID3D11Device *device) +{ + clear(); + mDevice = device; +} + +void RenderStateCache::clear() +{ + ClearStateMap(mBlendStateCache); + ClearStateMap(mRasterizerStateCache); + ClearStateMap(mDepthStencilStateCache); + ClearStateMap(mSamplerStateCache); +} + +std::size_t RenderStateCache::hashBlendState(const BlendStateKey &blendState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&blendState, sizeof(gl::BlendState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareBlendStates(const BlendStateKey &a, const BlendStateKey &b) +{ + return memcmp(&a, &b, sizeof(BlendStateKey)) == 0; +} + +gl::Error RenderStateCache::getBlendState(const gl::Framebuffer *framebuffer, const gl::BlendState &blendState, + ID3D11BlendState **outBlendState) +{ + if (!mDevice) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal error, RenderStateCache is not initialized."); + } + + bool mrt = false; + + const FramebufferD3D *framebufferD3D = GetImplAs<FramebufferD3D>(framebuffer); + const gl::AttachmentList &colorbuffers = framebufferD3D->getColorAttachmentsForRender(mRenderer->getWorkarounds()); + + BlendStateKey key = { 0 }; + key.blendState = blendState; + for (size_t colorAttachment = 0; colorAttachment < colorbuffers.size(); ++colorAttachment) + { + const gl::FramebufferAttachment *attachment = colorbuffers[colorAttachment]; + + auto rtChannels = key.rtChannels[colorAttachment]; + + if (attachment) + { + if (colorAttachment > 0) + { + mrt = true; + } + + rtChannels[0] = attachment->getRedSize() > 0; + rtChannels[1] = attachment->getGreenSize() > 0; + rtChannels[2] = attachment->getBlueSize() > 0; + rtChannels[3] = attachment->getAlphaSize() > 0; + } + } + + BlendStateMap::iterator keyIter = mBlendStateCache.find(key); + if (keyIter != mBlendStateCache.end()) + { + BlendStateCounterPair &state = keyIter->second; + state.second = mCounter++; + *outBlendState = state.first; + return gl::Error(GL_NO_ERROR); + } + else + { + if (mBlendStateCache.size() >= kMaxBlendStates) + { + TRACE("Overflowed the limit of %u blend states, removing the least recently used " + "to make room.", kMaxBlendStates); + + BlendStateMap::iterator leastRecentlyUsed = mBlendStateCache.begin(); + for (BlendStateMap::iterator i = mBlendStateCache.begin(); i != mBlendStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + SafeRelease(leastRecentlyUsed->second.first); + mBlendStateCache.erase(leastRecentlyUsed); + } + + // Create a new blend state and insert it into the cache + D3D11_BLEND_DESC blendDesc = { 0 }; + blendDesc.AlphaToCoverageEnable = blendState.sampleAlphaToCoverage; + blendDesc.IndependentBlendEnable = mrt ? TRUE : FALSE; + + for (unsigned int i = 0; i < D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; i++) + { + D3D11_RENDER_TARGET_BLEND_DESC &rtBlend = blendDesc.RenderTarget[i]; + + rtBlend.BlendEnable = blendState.blend; + if (blendState.blend) + { + rtBlend.SrcBlend = gl_d3d11::ConvertBlendFunc(blendState.sourceBlendRGB, false); + rtBlend.DestBlend = gl_d3d11::ConvertBlendFunc(blendState.destBlendRGB, false); + rtBlend.BlendOp = gl_d3d11::ConvertBlendOp(blendState.blendEquationRGB); + + rtBlend.SrcBlendAlpha = gl_d3d11::ConvertBlendFunc(blendState.sourceBlendAlpha, true); + rtBlend.DestBlendAlpha = gl_d3d11::ConvertBlendFunc(blendState.destBlendAlpha, true); + rtBlend.BlendOpAlpha = gl_d3d11::ConvertBlendOp(blendState.blendEquationAlpha); + } + + rtBlend.RenderTargetWriteMask = gl_d3d11::ConvertColorMask(key.rtChannels[i][0] && blendState.colorMaskRed, + key.rtChannels[i][1] && blendState.colorMaskGreen, + key.rtChannels[i][2] && blendState.colorMaskBlue, + key.rtChannels[i][3] && blendState.colorMaskAlpha); + } + + ID3D11BlendState *dx11BlendState = NULL; + HRESULT result = mDevice->CreateBlendState(&blendDesc, &dx11BlendState); + if (FAILED(result) || !dx11BlendState) + { + return gl::Error(GL_OUT_OF_MEMORY, "Unable to create a ID3D11BlendState, HRESULT: 0x%X.", result); + } + + mBlendStateCache.insert(std::make_pair(key, std::make_pair(dx11BlendState, mCounter++))); + + *outBlendState = dx11BlendState; + return gl::Error(GL_NO_ERROR); + } +} + +std::size_t RenderStateCache::hashRasterizerState(const RasterizerStateKey &rasterState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&rasterState, sizeof(RasterizerStateKey), seed, &hash); + return hash; +} + +bool RenderStateCache::compareRasterizerStates(const RasterizerStateKey &a, const RasterizerStateKey &b) +{ + return memcmp(&a, &b, sizeof(RasterizerStateKey)) == 0; +} + +gl::Error RenderStateCache::getRasterizerState(const gl::RasterizerState &rasterState, bool scissorEnabled, + ID3D11RasterizerState **outRasterizerState) +{ + if (!mDevice) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal error, RenderStateCache is not initialized."); + } + + RasterizerStateKey key = { 0 }; + key.rasterizerState = rasterState; + key.scissorEnabled = scissorEnabled; + + RasterizerStateMap::iterator keyIter = mRasterizerStateCache.find(key); + if (keyIter != mRasterizerStateCache.end()) + { + RasterizerStateCounterPair &state = keyIter->second; + state.second = mCounter++; + *outRasterizerState = state.first; + return gl::Error(GL_NO_ERROR); + } + else + { + if (mRasterizerStateCache.size() >= kMaxRasterizerStates) + { + TRACE("Overflowed the limit of %u rasterizer states, removing the least recently used " + "to make room.", kMaxRasterizerStates); + + RasterizerStateMap::iterator leastRecentlyUsed = mRasterizerStateCache.begin(); + for (RasterizerStateMap::iterator i = mRasterizerStateCache.begin(); i != mRasterizerStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + SafeRelease(leastRecentlyUsed->second.first); + mRasterizerStateCache.erase(leastRecentlyUsed); + } + + D3D11_CULL_MODE cullMode = gl_d3d11::ConvertCullMode(rasterState.cullFace, rasterState.cullMode); + + // Disable culling if drawing points + if (rasterState.pointDrawMode) + { + cullMode = D3D11_CULL_NONE; + } + + D3D11_RASTERIZER_DESC rasterDesc; + rasterDesc.FillMode = D3D11_FILL_SOLID; + rasterDesc.CullMode = cullMode; + rasterDesc.FrontCounterClockwise = (rasterState.frontFace == GL_CCW) ? FALSE: TRUE; + rasterDesc.DepthBiasClamp = 0.0f; // MSDN documentation of DepthBiasClamp implies a value of zero will preform no clamping, must be tested though. + rasterDesc.DepthClipEnable = TRUE; + rasterDesc.ScissorEnable = scissorEnabled ? TRUE : FALSE; + rasterDesc.MultisampleEnable = rasterState.multiSample; + rasterDesc.AntialiasedLineEnable = FALSE; + + if (rasterState.polygonOffsetFill) + { + rasterDesc.SlopeScaledDepthBias = rasterState.polygonOffsetFactor; + rasterDesc.DepthBias = (INT)rasterState.polygonOffsetUnits; + } + else + { + rasterDesc.SlopeScaledDepthBias = 0.0f; + rasterDesc.DepthBias = 0; + } + + ID3D11RasterizerState *dx11RasterizerState = NULL; + HRESULT result = mDevice->CreateRasterizerState(&rasterDesc, &dx11RasterizerState); + if (FAILED(result) || !dx11RasterizerState) + { + return gl::Error(GL_OUT_OF_MEMORY, "Unable to create a ID3D11RasterizerState, HRESULT: 0x%X.", result); + } + + mRasterizerStateCache.insert(std::make_pair(key, std::make_pair(dx11RasterizerState, mCounter++))); + + *outRasterizerState = dx11RasterizerState; + return gl::Error(GL_NO_ERROR); + } +} + +std::size_t RenderStateCache::hashDepthStencilState(const gl::DepthStencilState &dsState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&dsState, sizeof(gl::DepthStencilState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareDepthStencilStates(const gl::DepthStencilState &a, const gl::DepthStencilState &b) +{ + return memcmp(&a, &b, sizeof(gl::DepthStencilState)) == 0; +} + +gl::Error RenderStateCache::getDepthStencilState(const gl::DepthStencilState &dsState, ID3D11DepthStencilState **outDSState) +{ + if (!mDevice) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal error, RenderStateCache is not initialized."); + } + + DepthStencilStateMap::iterator keyIter = mDepthStencilStateCache.find(dsState); + if (keyIter != mDepthStencilStateCache.end()) + { + DepthStencilStateCounterPair &state = keyIter->second; + state.second = mCounter++; + *outDSState = state.first; + return gl::Error(GL_NO_ERROR); + } + else + { + if (mDepthStencilStateCache.size() >= kMaxDepthStencilStates) + { + TRACE("Overflowed the limit of %u depth stencil states, removing the least recently used " + "to make room.", kMaxDepthStencilStates); + + DepthStencilStateMap::iterator leastRecentlyUsed = mDepthStencilStateCache.begin(); + for (DepthStencilStateMap::iterator i = mDepthStencilStateCache.begin(); i != mDepthStencilStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + SafeRelease(leastRecentlyUsed->second.first); + mDepthStencilStateCache.erase(leastRecentlyUsed); + } + + D3D11_DEPTH_STENCIL_DESC dsDesc = { 0 }; + dsDesc.DepthEnable = dsState.depthTest ? TRUE : FALSE; + dsDesc.DepthWriteMask = gl_d3d11::ConvertDepthMask(dsState.depthMask); + dsDesc.DepthFunc = gl_d3d11::ConvertComparison(dsState.depthFunc); + dsDesc.StencilEnable = dsState.stencilTest ? TRUE : FALSE; + dsDesc.StencilReadMask = gl_d3d11::ConvertStencilMask(dsState.stencilMask); + dsDesc.StencilWriteMask = gl_d3d11::ConvertStencilMask(dsState.stencilWritemask); + dsDesc.FrontFace.StencilFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilFail); + dsDesc.FrontFace.StencilDepthFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilPassDepthFail); + dsDesc.FrontFace.StencilPassOp = gl_d3d11::ConvertStencilOp(dsState.stencilPassDepthPass); + dsDesc.FrontFace.StencilFunc = gl_d3d11::ConvertComparison(dsState.stencilFunc); + dsDesc.BackFace.StencilFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackFail); + dsDesc.BackFace.StencilDepthFailOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackPassDepthFail); + dsDesc.BackFace.StencilPassOp = gl_d3d11::ConvertStencilOp(dsState.stencilBackPassDepthPass); + dsDesc.BackFace.StencilFunc = gl_d3d11::ConvertComparison(dsState.stencilBackFunc); + + ID3D11DepthStencilState *dx11DepthStencilState = NULL; + HRESULT result = mDevice->CreateDepthStencilState(&dsDesc, &dx11DepthStencilState); + if (FAILED(result) || !dx11DepthStencilState) + { + return gl::Error(GL_OUT_OF_MEMORY, "Unable to create a ID3D11DepthStencilState, HRESULT: 0x%X.", result); + } + + mDepthStencilStateCache.insert(std::make_pair(dsState, std::make_pair(dx11DepthStencilState, mCounter++))); + + *outDSState = dx11DepthStencilState; + return gl::Error(GL_NO_ERROR); + } +} + +std::size_t RenderStateCache::hashSamplerState(const gl::SamplerState &samplerState) +{ + static const unsigned int seed = 0xABCDEF98; + + std::size_t hash = 0; + MurmurHash3_x86_32(&samplerState, sizeof(gl::SamplerState), seed, &hash); + return hash; +} + +bool RenderStateCache::compareSamplerStates(const gl::SamplerState &a, const gl::SamplerState &b) +{ + return memcmp(&a, &b, sizeof(gl::SamplerState)) == 0; +} + +gl::Error RenderStateCache::getSamplerState(const gl::SamplerState &samplerState, ID3D11SamplerState **outSamplerState) +{ + if (!mDevice) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal error, RenderStateCache is not initialized."); + } + + SamplerStateMap::iterator keyIter = mSamplerStateCache.find(samplerState); + if (keyIter != mSamplerStateCache.end()) + { + SamplerStateCounterPair &state = keyIter->second; + state.second = mCounter++; + *outSamplerState = state.first; + return gl::Error(GL_NO_ERROR); + } + else + { + if (mSamplerStateCache.size() >= kMaxSamplerStates) + { + TRACE("Overflowed the limit of %u sampler states, removing the least recently used " + "to make room.", kMaxSamplerStates); + + SamplerStateMap::iterator leastRecentlyUsed = mSamplerStateCache.begin(); + for (SamplerStateMap::iterator i = mSamplerStateCache.begin(); i != mSamplerStateCache.end(); i++) + { + if (i->second.second < leastRecentlyUsed->second.second) + { + leastRecentlyUsed = i; + } + } + SafeRelease(leastRecentlyUsed->second.first); + mSamplerStateCache.erase(leastRecentlyUsed); + } + + D3D11_SAMPLER_DESC samplerDesc; + samplerDesc.Filter = gl_d3d11::ConvertFilter(samplerState.minFilter, samplerState.magFilter, + samplerState.maxAnisotropy, samplerState.compareMode); + samplerDesc.AddressU = gl_d3d11::ConvertTextureWrap(samplerState.wrapS); + samplerDesc.AddressV = gl_d3d11::ConvertTextureWrap(samplerState.wrapT); + samplerDesc.AddressW = gl_d3d11::ConvertTextureWrap(samplerState.wrapR); + samplerDesc.MipLODBias = 0; + samplerDesc.MaxAnisotropy = samplerState.maxAnisotropy; + samplerDesc.ComparisonFunc = gl_d3d11::ConvertComparison(samplerState.compareFunc); + samplerDesc.BorderColor[0] = 0.0f; + samplerDesc.BorderColor[1] = 0.0f; + samplerDesc.BorderColor[2] = 0.0f; + samplerDesc.BorderColor[3] = 0.0f; + samplerDesc.MinLOD = samplerState.minLod; + samplerDesc.MaxLOD = samplerState.maxLod; + + if (mRenderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3) + { + // Check that maxLOD is nearly FLT_MAX (1000.0f is the default), since 9_3 doesn't support anything other than FLT_MAX. + // Note that Feature Level 9_* only supports GL ES 2.0, so the consumer of ANGLE can't modify the Max LOD themselves. + ASSERT(samplerState.maxLod >= 999.9f); + + // Now just set MaxLOD to FLT_MAX. Other parts of the renderer (e.g. the non-zero max LOD workaround) should take account of this. + samplerDesc.MaxLOD = FLT_MAX; + } + + ID3D11SamplerState *dx11SamplerState = NULL; + HRESULT result = mDevice->CreateSamplerState(&samplerDesc, &dx11SamplerState); + if (FAILED(result) || !dx11SamplerState) + { + return gl::Error(GL_OUT_OF_MEMORY, "Unable to create a ID3D11SamplerState, HRESULT: 0x%X.", result); + } + + mSamplerStateCache.insert(std::make_pair(samplerState, std::make_pair(dx11SamplerState, mCounter++))); + + *outSamplerState = dx11SamplerState; + return gl::Error(GL_NO_ERROR); + } +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.h new file mode 100644 index 0000000000..0099b94a04 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderStateCache.h @@ -0,0 +1,111 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderStateCache.h: Defines rx::RenderStateCache, a cache of Direct3D render +// state objects. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_RENDERSTATECACHE_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_RENDERSTATECACHE_H_ + +#include "libANGLE/angletypes.h" +#include "libANGLE/Error.h" +#include "common/angleutils.h" + +#include <unordered_map> + +namespace gl +{ +class Framebuffer; +} + +namespace rx +{ +class Renderer11; + +class RenderStateCache : angle::NonCopyable +{ + public: + RenderStateCache(Renderer11 *renderer); + virtual ~RenderStateCache(); + + void initialize(ID3D11Device *device); + void clear(); + + gl::Error getBlendState(const gl::Framebuffer *framebuffer, const gl::BlendState &blendState, ID3D11BlendState **outBlendState); + gl::Error getRasterizerState(const gl::RasterizerState &rasterState, bool scissorEnabled, ID3D11RasterizerState **outRasterizerState); + gl::Error getDepthStencilState(const gl::DepthStencilState &dsState, ID3D11DepthStencilState **outDSState); + gl::Error getSamplerState(const gl::SamplerState &samplerState, ID3D11SamplerState **outSamplerState); + + private: + Renderer11 *mRenderer; + unsigned long long mCounter; + + // Blend state cache + struct BlendStateKey + { + gl::BlendState blendState; + bool rtChannels[D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT][4]; + }; + static std::size_t hashBlendState(const BlendStateKey &blendState); + static bool compareBlendStates(const BlendStateKey &a, const BlendStateKey &b); + static const unsigned int kMaxBlendStates; + + typedef std::size_t (*BlendStateHashFunction)(const BlendStateKey &); + typedef bool (*BlendStateEqualityFunction)(const BlendStateKey &, const BlendStateKey &); + typedef std::pair<ID3D11BlendState*, unsigned long long> BlendStateCounterPair; + typedef std::unordered_map<BlendStateKey, BlendStateCounterPair, BlendStateHashFunction, BlendStateEqualityFunction> BlendStateMap; + BlendStateMap mBlendStateCache; + + // Rasterizer state cache + struct RasterizerStateKey + { + gl::RasterizerState rasterizerState; + bool scissorEnabled; + }; + static std::size_t hashRasterizerState(const RasterizerStateKey &rasterState); + static bool compareRasterizerStates(const RasterizerStateKey &a, const RasterizerStateKey &b); + static const unsigned int kMaxRasterizerStates; + + typedef std::size_t (*RasterizerStateHashFunction)(const RasterizerStateKey &); + typedef bool (*RasterizerStateEqualityFunction)(const RasterizerStateKey &, const RasterizerStateKey &); + typedef std::pair<ID3D11RasterizerState*, unsigned long long> RasterizerStateCounterPair; + typedef std::unordered_map<RasterizerStateKey, RasterizerStateCounterPair, RasterizerStateHashFunction, RasterizerStateEqualityFunction> RasterizerStateMap; + RasterizerStateMap mRasterizerStateCache; + + // Depth stencil state cache + static std::size_t hashDepthStencilState(const gl::DepthStencilState &dsState); + static bool compareDepthStencilStates(const gl::DepthStencilState &a, const gl::DepthStencilState &b); + static const unsigned int kMaxDepthStencilStates; + + typedef std::size_t (*DepthStencilStateHashFunction)(const gl::DepthStencilState &); + typedef bool (*DepthStencilStateEqualityFunction)(const gl::DepthStencilState &, const gl::DepthStencilState &); + typedef std::pair<ID3D11DepthStencilState*, unsigned long long> DepthStencilStateCounterPair; + typedef std::unordered_map<gl::DepthStencilState, + DepthStencilStateCounterPair, + DepthStencilStateHashFunction, + DepthStencilStateEqualityFunction> DepthStencilStateMap; + DepthStencilStateMap mDepthStencilStateCache; + + // Sample state cache + static std::size_t hashSamplerState(const gl::SamplerState &samplerState); + static bool compareSamplerStates(const gl::SamplerState &a, const gl::SamplerState &b); + static const unsigned int kMaxSamplerStates; + + typedef std::size_t (*SamplerStateHashFunction)(const gl::SamplerState &); + typedef bool (*SamplerStateEqualityFunction)(const gl::SamplerState &, const gl::SamplerState &); + typedef std::pair<ID3D11SamplerState*, unsigned long long> SamplerStateCounterPair; + typedef std::unordered_map<gl::SamplerState, + SamplerStateCounterPair, + SamplerStateHashFunction, + SamplerStateEqualityFunction> SamplerStateMap; + SamplerStateMap mSamplerStateCache; + + ID3D11Device *mDevice; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_RENDERSTATECACHE_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.cpp new file mode 100644 index 0000000000..ecd9e13c90 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.cpp @@ -0,0 +1,394 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget11.cpp: Implements a DX11-specific wrapper for ID3D11View pointers +// retained by Renderbuffers. + +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/SwapChain11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" + +namespace rx +{ + +static bool getTextureProperties(ID3D11Resource *resource, unsigned int *mipLevels, unsigned int *samples) +{ + ID3D11Texture1D *texture1D = d3d11::DynamicCastComObject<ID3D11Texture1D>(resource); + if (texture1D) + { + D3D11_TEXTURE1D_DESC texDesc; + texture1D->GetDesc(&texDesc); + SafeRelease(texture1D); + + *mipLevels = texDesc.MipLevels; + *samples = 0; + + return true; + } + + ID3D11Texture2D *texture2D = d3d11::DynamicCastComObject<ID3D11Texture2D>(resource); + if (texture2D) + { + D3D11_TEXTURE2D_DESC texDesc; + texture2D->GetDesc(&texDesc); + SafeRelease(texture2D); + + *mipLevels = texDesc.MipLevels; + *samples = texDesc.SampleDesc.Count > 1 ? texDesc.SampleDesc.Count : 0; + + return true; + } + + ID3D11Texture3D *texture3D = d3d11::DynamicCastComObject<ID3D11Texture3D>(resource); + if (texture3D) + { + D3D11_TEXTURE3D_DESC texDesc; + texture3D->GetDesc(&texDesc); + SafeRelease(texture3D); + + *mipLevels = texDesc.MipLevels; + *samples = 0; + + return true; + } + + return false; +} + +static unsigned int getRTVSubresourceIndex(ID3D11Resource *resource, ID3D11RenderTargetView *view) +{ + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + view->GetDesc(&rtvDesc); + + unsigned int mipSlice = 0; + unsigned int arraySlice = 0; + + switch (rtvDesc.ViewDimension) + { + case D3D11_RTV_DIMENSION_TEXTURE1D: + mipSlice = rtvDesc.Texture1D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE1DARRAY: + mipSlice = rtvDesc.Texture1DArray.MipSlice; + arraySlice = rtvDesc.Texture1DArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2D: + mipSlice = rtvDesc.Texture2D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DARRAY: + mipSlice = rtvDesc.Texture2DArray.MipSlice; + arraySlice = rtvDesc.Texture2DArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DMS: + mipSlice = 0; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_TEXTURE2DMSARRAY: + mipSlice = 0; + arraySlice = rtvDesc.Texture2DMSArray.FirstArraySlice; + break; + + case D3D11_RTV_DIMENSION_TEXTURE3D: + mipSlice = rtvDesc.Texture3D.MipSlice; + arraySlice = 0; + break; + + case D3D11_RTV_DIMENSION_UNKNOWN: + case D3D11_RTV_DIMENSION_BUFFER: + UNIMPLEMENTED(); + break; + + default: + UNREACHABLE(); + break; + } + + unsigned int mipLevels, samples; + getTextureProperties(resource, &mipLevels, &samples); + + return D3D11CalcSubresource(mipSlice, arraySlice, mipLevels); +} + +static unsigned int getDSVSubresourceIndex(ID3D11Resource *resource, ID3D11DepthStencilView *view) +{ + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + view->GetDesc(&dsvDesc); + + unsigned int mipSlice = 0; + unsigned int arraySlice = 0; + + switch (dsvDesc.ViewDimension) + { + case D3D11_DSV_DIMENSION_TEXTURE1D: + mipSlice = dsvDesc.Texture1D.MipSlice; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE1DARRAY: + mipSlice = dsvDesc.Texture1DArray.MipSlice; + arraySlice = dsvDesc.Texture1DArray.FirstArraySlice; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2D: + mipSlice = dsvDesc.Texture2D.MipSlice; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DARRAY: + mipSlice = dsvDesc.Texture2DArray.MipSlice; + arraySlice = dsvDesc.Texture2DArray.FirstArraySlice; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DMS: + mipSlice = 0; + arraySlice = 0; + break; + + case D3D11_DSV_DIMENSION_TEXTURE2DMSARRAY: + mipSlice = 0; + arraySlice = dsvDesc.Texture2DMSArray.FirstArraySlice; + break; + + case D3D11_DSV_DIMENSION_UNKNOWN: + UNIMPLEMENTED(); + break; + + default: + UNREACHABLE(); + break; + } + + unsigned int mipLevels, samples; + getTextureProperties(resource, &mipLevels, &samples); + + return D3D11CalcSubresource(mipSlice, arraySlice, mipLevels); +} + +RenderTarget11 *RenderTarget11::makeRenderTarget11(RenderTargetD3D *target) +{ + ASSERT(HAS_DYNAMIC_TYPE(RenderTarget11*, target)); + return static_cast<RenderTarget11*>(target); +} + +TextureRenderTarget11::TextureRenderTarget11(ID3D11RenderTargetView *rtv, ID3D11Resource *resource, ID3D11ShaderResourceView *srv, + GLenum internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLsizei samples) + : mWidth(width), + mHeight(height), + mDepth(depth), + mInternalFormat(internalFormat), + mDXGIFormat(DXGI_FORMAT_UNKNOWN), + mSamples(samples), + mSubresourceIndex(0), + mTexture(resource), + mRenderTarget(rtv), + mDepthStencil(NULL), + mShaderResource(srv) +{ + if (mTexture) + { + mTexture->AddRef(); + } + + if (mRenderTarget) + { + mRenderTarget->AddRef(); + } + + if (mShaderResource) + { + mShaderResource->AddRef(); + } + + if (mRenderTarget && mTexture) + { + mSubresourceIndex = getRTVSubresourceIndex(mTexture, mRenderTarget); + + D3D11_RENDER_TARGET_VIEW_DESC desc; + mRenderTarget->GetDesc(&desc); + mDXGIFormat = desc.Format; + } +} + +TextureRenderTarget11::TextureRenderTarget11(ID3D11DepthStencilView *dsv, ID3D11Resource *resource, ID3D11ShaderResourceView *srv, + GLenum internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLsizei samples) + : mWidth(width), + mHeight(height), + mDepth(depth), + mInternalFormat(internalFormat), + mDXGIFormat(DXGI_FORMAT_UNKNOWN), + mSamples(samples), + mSubresourceIndex(0), + mTexture(resource), + mRenderTarget(NULL), + mDepthStencil(dsv), + mShaderResource(srv) +{ + if (mTexture) + { + mTexture->AddRef(); + } + + if (mDepthStencil) + { + mDepthStencil->AddRef(); + } + + if (mShaderResource) + { + mShaderResource->AddRef(); + } + + if (mDepthStencil && mTexture) + { + mSubresourceIndex = getDSVSubresourceIndex(mTexture, mDepthStencil); + + D3D11_DEPTH_STENCIL_VIEW_DESC desc; + mDepthStencil->GetDesc(&desc); + mDXGIFormat = desc.Format; + } +} + +TextureRenderTarget11::~TextureRenderTarget11() +{ + SafeRelease(mTexture); + SafeRelease(mRenderTarget); + SafeRelease(mDepthStencil); + SafeRelease(mShaderResource); +} + +ID3D11Resource *TextureRenderTarget11::getTexture() const +{ + return mTexture; +} + +ID3D11RenderTargetView *TextureRenderTarget11::getRenderTargetView() const +{ + return mRenderTarget; +} + +ID3D11DepthStencilView *TextureRenderTarget11::getDepthStencilView() const +{ + return mDepthStencil; +} + +ID3D11ShaderResourceView *TextureRenderTarget11::getShaderResourceView() const +{ + return mShaderResource; +} + +GLsizei TextureRenderTarget11::getWidth() const +{ + return mWidth; +} + +GLsizei TextureRenderTarget11::getHeight() const +{ + return mHeight; +} + +GLsizei TextureRenderTarget11::getDepth() const +{ + return mDepth; +} + +GLenum TextureRenderTarget11::getInternalFormat() const +{ + return mInternalFormat; +} + +GLsizei TextureRenderTarget11::getSamples() const +{ + return mSamples; +} + +unsigned int TextureRenderTarget11::getSubresourceIndex() const +{ + return mSubresourceIndex; +} + +DXGI_FORMAT TextureRenderTarget11::getDXGIFormat() const +{ + return mDXGIFormat; +} + +SurfaceRenderTarget11::SurfaceRenderTarget11(SwapChain11 *swapChain, Renderer11 *renderer, bool depth) + : mSwapChain(swapChain), + mRenderer(renderer), + mDepth(depth) +{ + ASSERT(mSwapChain); +} + +SurfaceRenderTarget11::~SurfaceRenderTarget11() +{ +} + +GLsizei SurfaceRenderTarget11::getWidth() const +{ + return mSwapChain->getWidth(); +} + +GLsizei SurfaceRenderTarget11::getHeight() const +{ + return mSwapChain->getHeight(); +} + +GLsizei SurfaceRenderTarget11::getDepth() const +{ + return 1; +} + +GLenum SurfaceRenderTarget11::getInternalFormat() const +{ + return (mDepth ? mSwapChain->GetDepthBufferInternalFormat() : mSwapChain->GetBackBufferInternalFormat()); +} + +GLsizei SurfaceRenderTarget11::getSamples() const +{ + // Our EGL surfaces do not support multisampling. + return 0; +} + +ID3D11Resource *SurfaceRenderTarget11::getTexture() const +{ + return (mDepth ? mSwapChain->getDepthStencilTexture() : mSwapChain->getOffscreenTexture()); +} + +ID3D11RenderTargetView *SurfaceRenderTarget11::getRenderTargetView() const +{ + return (mDepth ? NULL : mSwapChain->getRenderTarget()); +} + +ID3D11DepthStencilView *SurfaceRenderTarget11::getDepthStencilView() const +{ + return (mDepth ? mSwapChain->getDepthStencil() : NULL); +} + +ID3D11ShaderResourceView *SurfaceRenderTarget11::getShaderResourceView() const +{ + return (mDepth ? mSwapChain->getDepthStencilShaderResource() : mSwapChain->getRenderTargetShaderResource()); +} + +unsigned int SurfaceRenderTarget11::getSubresourceIndex() const +{ + return 0; +} + +DXGI_FORMAT SurfaceRenderTarget11::getDXGIFormat() const +{ + return d3d11::GetTextureFormatInfo(getInternalFormat(), mRenderer->getFeatureLevel()).texFormat; +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.h new file mode 100644 index 0000000000..4472a56175 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/RenderTarget11.h @@ -0,0 +1,110 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// RenderTarget11.h: Defines a DX11-specific wrapper for ID3D11View pointers +// retained by Renderbuffers. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_RENDERTARGET11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_RENDERTARGET11_H_ + +#include "libANGLE/renderer/d3d/RenderTargetD3D.h" + +namespace rx +{ +class SwapChain11; +class Renderer11; + +class RenderTarget11 : public RenderTargetD3D +{ + public: + RenderTarget11() { } + virtual ~RenderTarget11() { } + + static RenderTarget11 *makeRenderTarget11(RenderTargetD3D *renderTarget); + + virtual ID3D11Resource *getTexture() const = 0; + virtual ID3D11RenderTargetView *getRenderTargetView() const = 0; + virtual ID3D11DepthStencilView *getDepthStencilView() const = 0; + virtual ID3D11ShaderResourceView *getShaderResourceView() const = 0; + + virtual unsigned int getSubresourceIndex() const = 0; + + virtual DXGI_FORMAT getDXGIFormat() const = 0; + + private: + D3D_FEATURE_LEVEL mFeatureLevel; +}; + +class TextureRenderTarget11 : public RenderTarget11 +{ + public: + // TextureRenderTarget11 takes ownership of any D3D11 resources it is given and will AddRef them + TextureRenderTarget11(ID3D11RenderTargetView *rtv, ID3D11Resource *resource, ID3D11ShaderResourceView *srv, + GLenum internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLsizei samples); + TextureRenderTarget11(ID3D11DepthStencilView *dsv, ID3D11Resource *resource, ID3D11ShaderResourceView *srv, + GLenum internalFormat, GLsizei width, GLsizei height, GLsizei depth, GLsizei samples); + virtual ~TextureRenderTarget11(); + + GLsizei getWidth() const override; + GLsizei getHeight() const override; + GLsizei getDepth() const override; + GLenum getInternalFormat() const override; + GLsizei getSamples() const override; + + ID3D11Resource *getTexture() const override; + ID3D11RenderTargetView *getRenderTargetView() const override; + ID3D11DepthStencilView *getDepthStencilView() const override; + ID3D11ShaderResourceView *getShaderResourceView() const override; + + unsigned int getSubresourceIndex() const override; + + DXGI_FORMAT getDXGIFormat() const override; + + private: + GLsizei mWidth; + GLsizei mHeight; + GLsizei mDepth; + GLenum mInternalFormat; + DXGI_FORMAT mDXGIFormat; + GLsizei mSamples; + + unsigned int mSubresourceIndex; + ID3D11Resource *mTexture; + ID3D11RenderTargetView *mRenderTarget; + ID3D11DepthStencilView *mDepthStencil; + ID3D11ShaderResourceView *mShaderResource; +}; + +class SurfaceRenderTarget11 : public RenderTarget11 +{ + public: + SurfaceRenderTarget11(SwapChain11 *swapChain, Renderer11 *renderer, bool depth); + virtual ~SurfaceRenderTarget11(); + + GLsizei getWidth() const override; + GLsizei getHeight() const override; + GLsizei getDepth() const override; + GLenum getInternalFormat() const override; + GLsizei getSamples() const override; + + ID3D11Resource *getTexture() const override; + ID3D11RenderTargetView *getRenderTargetView() const override; + ID3D11DepthStencilView *getDepthStencilView() const override; + ID3D11ShaderResourceView *getShaderResourceView() const override; + + unsigned int getSubresourceIndex() const override; + + DXGI_FORMAT getDXGIFormat() const override; + + private: + SwapChain11 *mSwapChain; + Renderer11 *mRenderer; + bool mDepth; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_RENDERTARGET11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.cpp new file mode 100644 index 0000000000..5291a3a086 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.cpp @@ -0,0 +1,3585 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer11.cpp: Implements a back-end specific class for the D3D11 renderer. + +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" + +#include "common/utilities.h" +#include "common/tls.h" +#include "libANGLE/Buffer.h" +#include "libANGLE/Display.h" +#include "libANGLE/Framebuffer.h" +#include "libANGLE/FramebufferAttachment.h" +#include "libANGLE/Program.h" +#include "libANGLE/State.h" +#include "libANGLE/Surface.h" +#include "libANGLE/formatutils.h" +#include "libANGLE/renderer/d3d/CompilerD3D.h" +#include "libANGLE/renderer/d3d/FramebufferD3D.h" +#include "libANGLE/renderer/d3d/IndexDataManager.h" +#include "libANGLE/renderer/d3d/ProgramD3D.h" +#include "libANGLE/renderer/d3d/RenderbufferD3D.h" +#include "libANGLE/renderer/d3d/ShaderD3D.h" +#include "libANGLE/renderer/d3d/SurfaceD3D.h" +#include "libANGLE/renderer/d3d/TextureD3D.h" +#include "libANGLE/renderer/d3d/TransformFeedbackD3D.h" +#include "libANGLE/renderer/d3d/VertexDataManager.h" +#include "libANGLE/renderer/d3d/d3d11/Blit11.h" +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Clear11.h" +#include "libANGLE/renderer/d3d/d3d11/Fence11.h" +#include "libANGLE/renderer/d3d/d3d11/Framebuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Image11.h" +#include "libANGLE/renderer/d3d/d3d11/IndexBuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/PixelTransfer11.h" +#include "libANGLE/renderer/d3d/d3d11/Query11.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h" +#include "libANGLE/renderer/d3d/d3d11/SwapChain11.h" +#include "libANGLE/renderer/d3d/d3d11/TextureStorage11.h" +#include "libANGLE/renderer/d3d/d3d11/Trim11.h" +#include "libANGLE/renderer/d3d/d3d11/VertexArray11.h" +#include "libANGLE/renderer/d3d/d3d11/VertexBuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" + +#include <sstream> +#include <EGL/eglext.h> + +// Enable ANGLE_SKIP_DXGI_1_2_CHECK if there is not a possibility of using cross-process +// HWNDs or the Windows 7 Platform Update (KB2670838) is expected to be installed. +#ifndef ANGLE_SKIP_DXGI_1_2_CHECK +#define ANGLE_SKIP_DXGI_1_2_CHECK 0 +#endif + +#ifdef _DEBUG +// this flag enables suppressing some spurious warnings that pop up in certain WebGL samples +// and conformance tests. to enable all warnings, remove this define. +#define ANGLE_SUPPRESS_D3D11_HAZARD_WARNINGS 1 +#endif + +#ifndef __d3d11sdklayers_h__ +#define D3D11_MESSAGE_CATEGORY UINT +#define D3D11_MESSAGE_SEVERITY UINT +#define D3D11_MESSAGE_ID UINT +struct D3D11_MESSAGE; +typedef struct D3D11_INFO_QUEUE_FILTER_DESC +{ + UINT NumCategories; + D3D11_MESSAGE_CATEGORY *pCategoryList; + UINT NumSeverities; + D3D11_MESSAGE_SEVERITY *pSeverityList; + UINT NumIDs; + D3D11_MESSAGE_ID *pIDList; +} D3D11_INFO_QUEUE_FILTER_DESC; +typedef struct D3D11_INFO_QUEUE_FILTER +{ + D3D11_INFO_QUEUE_FILTER_DESC AllowList; + D3D11_INFO_QUEUE_FILTER_DESC DenyList; +} D3D11_INFO_QUEUE_FILTER; +static const IID IID_ID3D11InfoQueue = { 0x6543dbb6, 0x1b48, 0x42f5, 0xab, 0x82, 0xe9, 0x7e, 0xc7, 0x43, 0x26, 0xf6 }; +MIDL_INTERFACE("6543dbb6-1b48-42f5-ab82-e97ec74326f6") ID3D11InfoQueue : public IUnknown +{ +public: + virtual HRESULT __stdcall SetMessageCountLimit(UINT64) = 0; + virtual void __stdcall ClearStoredMessages() = 0; + virtual HRESULT __stdcall GetMessage(UINT64, D3D11_MESSAGE *, SIZE_T *) = 0; + virtual UINT64 __stdcall GetNumMessagesAllowedByStorageFilter() = 0; + virtual UINT64 __stdcall GetNumMessagesDeniedByStorageFilter() = 0; + virtual UINT64 __stdcall GetNumStoredMessages() = 0; + virtual UINT64 __stdcall GetNumStoredMessagesAllowedByRetrievalFilter() = 0; + virtual UINT64 __stdcall GetNumMessagesDiscardedByMessageCountLimit() = 0; + virtual UINT64 __stdcall GetMessageCountLimit() = 0; + virtual HRESULT __stdcall AddStorageFilterEntries(D3D11_INFO_QUEUE_FILTER *) = 0; + virtual HRESULT __stdcall GetStorageFilter(D3D11_INFO_QUEUE_FILTER *, SIZE_T *) = 0; + virtual void __stdcall ClearStorageFilter() = 0; + virtual HRESULT __stdcall PushEmptyStorageFilter() = 0; + virtual HRESULT __stdcall PushCopyOfStorageFilter() = 0; + virtual HRESULT __stdcall PushStorageFilter(D3D11_INFO_QUEUE_FILTER *) = 0; + virtual void __stdcall PopStorageFilter() = 0; + virtual UINT __stdcall GetStorageFilterStackSize() = 0; + virtual HRESULT __stdcall AddRetrievalFilterEntries(D3D11_INFO_QUEUE_FILTER *) = 0; + virtual HRESULT __stdcall GetRetrievalFilter(D3D11_INFO_QUEUE_FILTER *, SIZE_T *) = 0; + virtual void __stdcall ClearRetrievalFilter() = 0; + virtual HRESULT __stdcall PushEmptyRetrievalFilter() = 0; + virtual HRESULT __stdcall PushCopyOfRetrievalFilter() = 0; + virtual HRESULT __stdcall PushRetrievalFilter(D3D11_INFO_QUEUE_FILTER *) = 0; + virtual void __stdcall PopRetrievalFilter() = 0; + virtual UINT __stdcall GetRetrievalFilterStackSize() = 0; + virtual HRESULT __stdcall AddMessage(D3D11_MESSAGE_CATEGORY, D3D11_MESSAGE_SEVERITY, D3D11_MESSAGE_ID, LPCSTR) = 0; + virtual HRESULT __stdcall AddApplicationMessage(D3D11_MESSAGE_SEVERITY, LPCSTR) = 0; + virtual HRESULT __stdcall SetBreakOnCategory(D3D11_MESSAGE_CATEGORY, BOOL) = 0; + virtual HRESULT __stdcall SetBreakOnSeverity(D3D11_MESSAGE_SEVERITY, BOOL) = 0; + virtual HRESULT __stdcall SetBreakOnID(D3D11_MESSAGE_ID, BOOL) = 0; + virtual BOOL __stdcall GetBreakOnCategory(D3D11_MESSAGE_CATEGORY) = 0; + virtual BOOL __stdcall GetBreakOnSeverity(D3D11_MESSAGE_SEVERITY) = 0; + virtual BOOL __stdcall GetBreakOnID(D3D11_MESSAGE_ID) = 0; + virtual void __stdcall SetMuteDebugOutput(BOOL) = 0; + virtual BOOL __stdcall GetMuteDebugOutput() = 0; +}; +#endif + +namespace rx +{ + +namespace +{ + +enum +{ + MAX_TEXTURE_IMAGE_UNITS_VTF_SM4 = 16 +}; + +// dirtyPointer is a special value that will make the comparison with any valid pointer fail and force the renderer to re-apply the state. +static const uintptr_t DirtyPointer = static_cast<uintptr_t>(-1); + +static bool ImageIndexConflictsWithSRV(const gl::ImageIndex *index, D3D11_SHADER_RESOURCE_VIEW_DESC desc) +{ + unsigned mipLevel = index->mipIndex; + unsigned layerIndex = index->layerIndex; + GLenum type = index->type; + + switch (desc.ViewDimension) + { + case D3D11_SRV_DIMENSION_TEXTURE2D: + { + unsigned maxSrvMip = desc.Texture2D.MipLevels + desc.Texture2D.MostDetailedMip; + maxSrvMip = (desc.Texture2D.MipLevels == -1) ? INT_MAX : maxSrvMip; + + unsigned mipMin = index->mipIndex; + unsigned mipMax = (layerIndex == -1) ? INT_MAX : layerIndex; + + return type == GL_TEXTURE_2D && RangeUI(mipMin, mipMax).intersects(RangeUI(desc.Texture2D.MostDetailedMip, maxSrvMip)); + } + + case D3D11_SRV_DIMENSION_TEXTURE2DARRAY: + { + unsigned maxSrvMip = desc.Texture2DArray.MipLevels + desc.Texture2DArray.MostDetailedMip; + maxSrvMip = (desc.Texture2DArray.MipLevels == -1) ? INT_MAX : maxSrvMip; + + unsigned maxSlice = desc.Texture2DArray.FirstArraySlice + desc.Texture2DArray.ArraySize; + + // Cube maps can be mapped to Texture2DArray SRVs + return (type == GL_TEXTURE_2D_ARRAY || gl::IsCubeMapTextureTarget(type)) && + desc.Texture2DArray.MostDetailedMip <= mipLevel && mipLevel < maxSrvMip && + desc.Texture2DArray.FirstArraySlice <= layerIndex && layerIndex < maxSlice; + } + + case D3D11_SRV_DIMENSION_TEXTURECUBE: + { + unsigned maxSrvMip = desc.TextureCube.MipLevels + desc.TextureCube.MostDetailedMip; + maxSrvMip = (desc.TextureCube.MipLevels == -1) ? INT_MAX : maxSrvMip; + + return gl::IsCubeMapTextureTarget(type) && + desc.TextureCube.MostDetailedMip <= mipLevel && mipLevel < maxSrvMip; + } + + case D3D11_SRV_DIMENSION_TEXTURE3D: + { + unsigned maxSrvMip = desc.Texture3D.MipLevels + desc.Texture3D.MostDetailedMip; + maxSrvMip = (desc.Texture3D.MipLevels == -1) ? INT_MAX : maxSrvMip; + + return type == GL_TEXTURE_3D && + desc.Texture3D.MostDetailedMip <= mipLevel && mipLevel < maxSrvMip; + } + default: + // We only handle the cases corresponding to valid image indexes + UNIMPLEMENTED(); + } + + return false; +} + +// Does *not* increment the resource ref count!! +ID3D11Resource *GetViewResource(ID3D11View *view) +{ + ID3D11Resource *resource = NULL; + ASSERT(view); + view->GetResource(&resource); + resource->Release(); + return resource; +} + +void CalculateConstantBufferParams(GLintptr offset, GLsizeiptr size, UINT *outFirstConstant, UINT *outNumConstants) +{ + // The offset must be aligned to 256 bytes (should have been enforced by glBindBufferRange). + ASSERT(offset % 256 == 0); + + // firstConstant and numConstants are expressed in constants of 16-bytes. Furthermore they must be a multiple of 16 constants. + *outFirstConstant = offset / 16; + + // The GL size is not required to be aligned to a 256 bytes boundary. + // Round the size up to a 256 bytes boundary then express the results in constants of 16-bytes. + *outNumConstants = rx::roundUp(size, static_cast<GLsizeiptr>(256)) / 16; + + // Since the size is rounded up, firstConstant + numConstants may be bigger than the actual size of the buffer. + // This behaviour is explictly allowed according to the documentation on ID3D11DeviceContext1::PSSetConstantBuffers1 + // https://msdn.microsoft.com/en-us/library/windows/desktop/hh404649%28v=vs.85%29.aspx +} + +} + +Renderer11::Renderer11(egl::Display *display) + : RendererD3D(display), + mStateCache(this) +{ + // Initialize global annotator + gl::InitializeDebugAnnotations(&mAnnotator); + + mVertexDataManager = NULL; + mIndexDataManager = NULL; + + mLineLoopIB = NULL; + mTriangleFanIB = NULL; + + mBlit = NULL; + mPixelTransfer = NULL; + + mClear = NULL; + + mTrim = NULL; + + mSyncQuery = NULL; + + mSupportsConstantBufferOffsets = false; + + mD3d11Module = NULL; + mDxgiModule = NULL; + + mDevice = NULL; + mDeviceContext = NULL; + mDeviceContext1 = NULL; + mDxgiAdapter = NULL; + mDxgiFactory = NULL; + + mDriverConstantBufferVS = NULL; + mDriverConstantBufferPS = NULL; + + mAppliedVertexShader = NULL; + mAppliedGeometryShader = NULL; + mAppliedPixelShader = NULL; + + mAppliedNumXFBBindings = static_cast<size_t>(-1); + + const auto &attributes = mDisplay->getAttributeMap(); + + EGLint requestedMajorVersion = attributes.get(EGL_PLATFORM_ANGLE_MAX_VERSION_MAJOR_ANGLE, EGL_DONT_CARE); + EGLint requestedMinorVersion = attributes.get(EGL_PLATFORM_ANGLE_MAX_VERSION_MINOR_ANGLE, EGL_DONT_CARE); + + if (requestedMajorVersion == EGL_DONT_CARE || requestedMajorVersion >= 11) + { + if (requestedMinorVersion == EGL_DONT_CARE || requestedMinorVersion >= 0) + { + mAvailableFeatureLevels.push_back(D3D_FEATURE_LEVEL_11_0); + } + } + + if (requestedMajorVersion == EGL_DONT_CARE || requestedMajorVersion >= 10) + { + if (requestedMinorVersion == EGL_DONT_CARE || requestedMinorVersion >= 1) + { + mAvailableFeatureLevels.push_back(D3D_FEATURE_LEVEL_10_1); + } + if (requestedMinorVersion == EGL_DONT_CARE || requestedMinorVersion >= 0) + { + mAvailableFeatureLevels.push_back(D3D_FEATURE_LEVEL_10_0); + } + } + +#if defined(ANGLE_ENABLE_WINDOWS_STORE) + if (requestedMajorVersion == EGL_DONT_CARE || requestedMajorVersion >= 9) +#else + if (requestedMajorVersion == 9 && requestedMinorVersion == 3) +#endif + { + mAvailableFeatureLevels.push_back(D3D_FEATURE_LEVEL_9_3); + } + + EGLint requestedDeviceType = attributes.get(EGL_PLATFORM_ANGLE_DEVICE_TYPE_ANGLE, + EGL_PLATFORM_ANGLE_DEVICE_TYPE_HARDWARE_ANGLE); + switch (requestedDeviceType) + { + case EGL_PLATFORM_ANGLE_DEVICE_TYPE_HARDWARE_ANGLE: + mDriverType = D3D_DRIVER_TYPE_HARDWARE; + break; + + case EGL_PLATFORM_ANGLE_DEVICE_TYPE_WARP_ANGLE: + mDriverType = D3D_DRIVER_TYPE_WARP; + break; + + case EGL_PLATFORM_ANGLE_DEVICE_TYPE_REFERENCE_ANGLE: + mDriverType = D3D_DRIVER_TYPE_REFERENCE; + break; + + case EGL_PLATFORM_ANGLE_DEVICE_TYPE_NULL_ANGLE: + mDriverType = D3D_DRIVER_TYPE_NULL; + break; + + default: + UNREACHABLE(); + } +} + +Renderer11::~Renderer11() +{ + release(); + + gl::UninitializeDebugAnnotations(); +} + +Renderer11 *Renderer11::makeRenderer11(Renderer *renderer) +{ + ASSERT(HAS_DYNAMIC_TYPE(Renderer11*, renderer)); + return static_cast<Renderer11*>(renderer); +} + +#ifndef __d3d11_1_h__ +#define D3D11_MESSAGE_ID_DEVICE_DRAW_RENDERTARGETVIEW_NOT_SET ((D3D11_MESSAGE_ID)3146081) +#endif + +egl::Error Renderer11::initialize() +{ + if (!mCompiler.initialize()) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_COMPILER_ERROR, + "Failed to initialize compiler."); + } + +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) + mDxgiModule = LoadLibrary(TEXT("dxgi.dll")); + mD3d11Module = LoadLibrary(TEXT("d3d11.dll")); + + if (mD3d11Module == NULL || mDxgiModule == NULL) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_MISSING_DEP, + "Could not load D3D11 or DXGI library."); + } + + // create the D3D11 device + ASSERT(mDevice == NULL); + PFN_D3D11_CREATE_DEVICE D3D11CreateDevice = (PFN_D3D11_CREATE_DEVICE)GetProcAddress(mD3d11Module, "D3D11CreateDevice"); + + if (D3D11CreateDevice == NULL) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_MISSING_DEP, + "Could not retrieve D3D11CreateDevice address."); + } +#endif + + HRESULT result = S_OK; +#ifdef _DEBUG + result = D3D11CreateDevice(NULL, + mDriverType, + NULL, + D3D11_CREATE_DEVICE_DEBUG, + mAvailableFeatureLevels.data(), + mAvailableFeatureLevels.size(), + D3D11_SDK_VERSION, + &mDevice, + &mFeatureLevel, + &mDeviceContext); + + if (!mDevice || FAILED(result)) + { + ERR("Failed creating Debug D3D11 device - falling back to release runtime.\n"); + } + + if (!mDevice || FAILED(result)) +#endif + { + result = D3D11CreateDevice(NULL, + mDriverType, + NULL, + 0, + mAvailableFeatureLevels.data(), + mAvailableFeatureLevels.size(), + D3D11_SDK_VERSION, + &mDevice, + &mFeatureLevel, + &mDeviceContext); + + if (result == E_INVALIDARG) + { + // Cleanup done by destructor through glDestroyRenderer + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_CREATEDEVICE_INVALIDARG, + "Could not create D3D11 device."); + } + + if (!mDevice || FAILED(result)) + { + // Cleanup done by destructor through glDestroyRenderer + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_CREATEDEVICE_ERROR, + "Could not create D3D11 device."); + } + } + +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) +#if !ANGLE_SKIP_DXGI_1_2_CHECK + // In order to create a swap chain for an HWND owned by another process, DXGI 1.2 is required. + // The easiest way to check is to query for a IDXGIDevice2. + bool requireDXGI1_2 = false; + HWND hwnd = WindowFromDC(mDisplay->getNativeDisplayId()); + if (hwnd) + { + DWORD currentProcessId = GetCurrentProcessId(); + DWORD wndProcessId; + GetWindowThreadProcessId(hwnd, &wndProcessId); + requireDXGI1_2 = (currentProcessId != wndProcessId); + } + else + { + requireDXGI1_2 = true; + } + + if (requireDXGI1_2) + { + IDXGIDevice2 *dxgiDevice2 = NULL; + result = mDevice->QueryInterface(__uuidof(IDXGIDevice2), (void**)&dxgiDevice2); + if (FAILED(result)) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_INCOMPATIBLE_DXGI, + "DXGI 1.2 required to present to HWNDs owned by another process."); + } + SafeRelease(dxgiDevice2); + } +#endif +#endif + + // Cast the DeviceContext to a DeviceContext1. + // This could fail on Windows 7 without the Platform Update. + // Don't error in this case- just don't use mDeviceContext1. +#if defined(ANGLE_ENABLE_D3D11_1) + mDeviceContext1 = d3d11::DynamicCastComObject<ID3D11DeviceContext1>(mDeviceContext); +#endif + + IDXGIDevice *dxgiDevice = NULL; + result = mDevice->QueryInterface(__uuidof(IDXGIDevice), (void**)&dxgiDevice); + + if (FAILED(result)) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_OTHER_ERROR, + "Could not query DXGI device."); + } + + result = dxgiDevice->GetParent(__uuidof(IDXGIAdapter), (void**)&mDxgiAdapter); + + if (FAILED(result)) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_OTHER_ERROR, + "Could not retrieve DXGI adapter"); + } + + SafeRelease(dxgiDevice); + +#if defined(ANGLE_ENABLE_D3D11_1) + IDXGIAdapter2 *dxgiAdapter2 = d3d11::DynamicCastComObject<IDXGIAdapter2>(mDxgiAdapter); + + // On D3D_FEATURE_LEVEL_9_*, IDXGIAdapter::GetDesc returns "Software Adapter" for the description string. + // If DXGI1.2 is available then IDXGIAdapter2::GetDesc2 can be used to get the actual hardware values. + if (mFeatureLevel <= D3D_FEATURE_LEVEL_9_3 && dxgiAdapter2 != NULL) + { + DXGI_ADAPTER_DESC2 adapterDesc2 = {0}; + dxgiAdapter2->GetDesc2(&adapterDesc2); + + // Copy the contents of the DXGI_ADAPTER_DESC2 into mAdapterDescription (a DXGI_ADAPTER_DESC). + memcpy(mAdapterDescription.Description, adapterDesc2.Description, sizeof(mAdapterDescription.Description)); + mAdapterDescription.VendorId = adapterDesc2.VendorId; + mAdapterDescription.DeviceId = adapterDesc2.DeviceId; + mAdapterDescription.SubSysId = adapterDesc2.SubSysId; + mAdapterDescription.Revision = adapterDesc2.Revision; + mAdapterDescription.DedicatedVideoMemory = adapterDesc2.DedicatedVideoMemory; + mAdapterDescription.DedicatedSystemMemory = adapterDesc2.DedicatedSystemMemory; + mAdapterDescription.SharedSystemMemory = adapterDesc2.SharedSystemMemory; + mAdapterDescription.AdapterLuid = adapterDesc2.AdapterLuid; + } + else + { + mDxgiAdapter->GetDesc(&mAdapterDescription); + } + + SafeRelease(dxgiAdapter2); +#endif + + memset(mDescription, 0, sizeof(mDescription)); + wcstombs(mDescription, mAdapterDescription.Description, sizeof(mDescription) - 1); + + result = mDxgiAdapter->GetParent(__uuidof(IDXGIFactory), (void**)&mDxgiFactory); + + if (!mDxgiFactory || FAILED(result)) + { + return egl::Error(EGL_NOT_INITIALIZED, + D3D11_INIT_OTHER_ERROR, + "Could not create DXGI factory."); + } + + // Disable some spurious D3D11 debug warnings to prevent them from flooding the output log +#if defined(ANGLE_SUPPRESS_D3D11_HAZARD_WARNINGS) && defined(_DEBUG) + ID3D11InfoQueue *infoQueue; + result = mDevice->QueryInterface(IID_ID3D11InfoQueue, (void **)&infoQueue); + + if (SUCCEEDED(result)) + { + D3D11_MESSAGE_ID hideMessages[] = + { + D3D11_MESSAGE_ID_DEVICE_DRAW_RENDERTARGETVIEW_NOT_SET + }; + + D3D11_INFO_QUEUE_FILTER filter = {0}; + filter.DenyList.NumIDs = ArraySize(hideMessages); + filter.DenyList.pIDList = hideMessages; + + infoQueue->AddStorageFilterEntries(&filter); + SafeRelease(infoQueue); + } +#endif + + initializeDevice(); + + return egl::Error(EGL_SUCCESS); +} + +// do any one-time device initialization +// NOTE: this is also needed after a device lost/reset +// to reset the scene status and ensure the default states are reset. +void Renderer11::initializeDevice() +{ + mStateCache.initialize(mDevice); + mInputLayoutCache.initialize(mDevice, mDeviceContext); + + ASSERT(!mVertexDataManager && !mIndexDataManager); + mVertexDataManager = new VertexDataManager(this); + mIndexDataManager = new IndexDataManager(this, getRendererClass()); + + ASSERT(!mBlit); + mBlit = new Blit11(this); + + ASSERT(!mClear); + mClear = new Clear11(this); + + const auto &attributes = mDisplay->getAttributeMap(); + // If automatic trim is enabled, DXGIDevice3::Trim( ) is called for the application + // automatically when an application is suspended by the OS. This feature is currently + // only supported for Windows Store applications. + EGLint enableAutoTrim = attributes.get(EGL_PLATFORM_ANGLE_ENABLE_AUTOMATIC_TRIM_ANGLE, EGL_FALSE); + + if (enableAutoTrim == EGL_TRUE) + { + ASSERT(!mTrim); + mTrim = new Trim11(this); + } + + ASSERT(!mPixelTransfer); + mPixelTransfer = new PixelTransfer11(this); + + const gl::Caps &rendererCaps = getRendererCaps(); + +#if defined(ANGLE_ENABLE_D3D11_1) + if (getDeviceContext1IfSupported()) + { + D3D11_FEATURE_DATA_D3D11_OPTIONS d3d11Options; + mDevice->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &d3d11Options, sizeof(D3D11_FEATURE_DATA_D3D11_OPTIONS)); + mSupportsConstantBufferOffsets = (d3d11Options.ConstantBufferOffsetting != FALSE); + } +#endif + + mForceSetVertexSamplerStates.resize(rendererCaps.maxVertexTextureImageUnits); + mCurVertexSamplerStates.resize(rendererCaps.maxVertexTextureImageUnits); + + mForceSetPixelSamplerStates.resize(rendererCaps.maxTextureImageUnits); + mCurPixelSamplerStates.resize(rendererCaps.maxTextureImageUnits); + + mCurVertexSRVs.resize(rendererCaps.maxVertexTextureImageUnits); + mCurPixelSRVs.resize(rendererCaps.maxTextureImageUnits); + + markAllStateDirty(); +} + +egl::ConfigSet Renderer11::generateConfigs() const +{ + static const GLenum colorBufferFormats[] = + { + GL_BGRA8_EXT, + GL_RGBA8_OES, + }; + + static const GLenum depthStencilBufferFormats[] = + { + GL_NONE, + GL_DEPTH24_STENCIL8_OES, + GL_DEPTH_COMPONENT16, + }; + + const gl::Caps &rendererCaps = getRendererCaps(); + const gl::TextureCapsMap &rendererTextureCaps = getRendererTextureCaps(); + + egl::ConfigSet configs; + for (size_t formatIndex = 0; formatIndex < ArraySize(colorBufferFormats); formatIndex++) + { + GLenum colorBufferInternalFormat = colorBufferFormats[formatIndex]; + const gl::TextureCaps &colorBufferFormatCaps = rendererTextureCaps.get(colorBufferInternalFormat); + if (colorBufferFormatCaps.renderable) + { + for (size_t depthStencilIndex = 0; depthStencilIndex < ArraySize(depthStencilBufferFormats); depthStencilIndex++) + { + GLenum depthStencilBufferInternalFormat = depthStencilBufferFormats[depthStencilIndex]; + const gl::TextureCaps &depthStencilBufferFormatCaps = rendererTextureCaps.get(depthStencilBufferInternalFormat); + if (depthStencilBufferFormatCaps.renderable || depthStencilBufferInternalFormat == GL_NONE) + { + const gl::InternalFormat &colorBufferFormatInfo = gl::GetInternalFormatInfo(colorBufferInternalFormat); + const gl::InternalFormat &depthStencilBufferFormatInfo = gl::GetInternalFormatInfo(depthStencilBufferInternalFormat); + + egl::Config config; + config.renderTargetFormat = colorBufferInternalFormat; + config.depthStencilFormat = depthStencilBufferInternalFormat; + config.bufferSize = colorBufferFormatInfo.pixelBytes * 8; + config.redSize = colorBufferFormatInfo.redBits; + config.greenSize = colorBufferFormatInfo.greenBits; + config.blueSize = colorBufferFormatInfo.blueBits; + config.luminanceSize = colorBufferFormatInfo.luminanceBits; + config.alphaSize = colorBufferFormatInfo.alphaBits; + config.alphaMaskSize = 0; + config.bindToTextureRGB = (colorBufferFormatInfo.format == GL_RGB); + config.bindToTextureRGBA = (colorBufferFormatInfo.format == GL_RGBA || colorBufferFormatInfo.format == GL_BGRA_EXT); + config.colorBufferType = EGL_RGB_BUFFER; + config.configID = static_cast<EGLint>(configs.size() + 1); + // Can only support a conformant ES2 with feature level greater than 10.0. + config.conformant = (mFeatureLevel >= D3D_FEATURE_LEVEL_10_0) ? (EGL_OPENGL_ES2_BIT | EGL_OPENGL_ES3_BIT_KHR) : EGL_NONE; + config.configCaveat = config.conformant == EGL_NONE ? EGL_NON_CONFORMANT_CONFIG : EGL_NONE; + config.depthSize = depthStencilBufferFormatInfo.depthBits; + config.level = 0; + config.matchNativePixmap = EGL_NONE; + config.maxPBufferWidth = rendererCaps.max2DTextureSize; + config.maxPBufferHeight = rendererCaps.max2DTextureSize; + config.maxPBufferPixels = rendererCaps.max2DTextureSize * rendererCaps.max2DTextureSize; + config.maxSwapInterval = 4; + config.minSwapInterval = 0; + config.nativeRenderable = EGL_FALSE; + config.nativeVisualID = 0; + config.nativeVisualType = EGL_NONE; + // Can't support ES3 at all without feature level 10.0 + config.renderableType = EGL_OPENGL_ES2_BIT | ((mFeatureLevel >= D3D_FEATURE_LEVEL_10_0) ? EGL_OPENGL_ES3_BIT_KHR : 0); + config.sampleBuffers = 0; // FIXME: enumerate multi-sampling + config.samples = 0; + config.stencilSize = depthStencilBufferFormatInfo.stencilBits; + config.surfaceType = EGL_PBUFFER_BIT | EGL_WINDOW_BIT | EGL_SWAP_BEHAVIOR_PRESERVED_BIT; + config.transparentType = EGL_NONE; + config.transparentRedValue = 0; + config.transparentGreenValue = 0; + config.transparentBlueValue = 0; + + configs.add(config); + } + } + } + } + + ASSERT(configs.size() > 0); + return configs; +} + +gl::Error Renderer11::flush() +{ + mDeviceContext->Flush(); + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::finish() +{ + HRESULT result; + + if (!mSyncQuery) + { + D3D11_QUERY_DESC queryDesc; + queryDesc.Query = D3D11_QUERY_EVENT; + queryDesc.MiscFlags = 0; + + result = mDevice->CreateQuery(&queryDesc, &mSyncQuery); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create event query, result: 0x%X.", result); + } + } + + mDeviceContext->End(mSyncQuery); + mDeviceContext->Flush(); + + do + { + result = mDeviceContext->GetData(mSyncQuery, NULL, 0, D3D11_ASYNC_GETDATA_DONOTFLUSH); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to get event query data, result: 0x%X.", result); + } + + // Keep polling, but allow other threads to do something useful first + ScheduleYield(); + + if (testDeviceLost()) + { + mDisplay->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Device was lost while waiting for sync."); + } + } + while (result == S_FALSE); + + return gl::Error(GL_NO_ERROR); +} + +SwapChainD3D *Renderer11::createSwapChain(NativeWindow nativeWindow, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat) +{ + return new SwapChain11(this, nativeWindow, shareHandle, backBufferFormat, depthBufferFormat); +} + +gl::Error Renderer11::generateSwizzle(gl::Texture *texture) +{ + if (texture) + { + TextureD3D *textureD3D = GetImplAs<TextureD3D>(texture); + ASSERT(textureD3D); + + TextureStorage *texStorage = nullptr; + gl::Error error = textureD3D->getNativeTexture(&texStorage); + if (error.isError()) + { + return error; + } + + if (texStorage) + { + TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(texStorage); + error = storage11->generateSwizzles(texture->getSamplerState().swizzleRed, + texture->getSamplerState().swizzleGreen, + texture->getSamplerState().swizzleBlue, + texture->getSamplerState().swizzleAlpha); + if (error.isError()) + { + return error; + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setSamplerState(gl::SamplerType type, int index, gl::Texture *texture, const gl::SamplerState &samplerStateParam) +{ + // Make sure to add the level offset for our tiny compressed texture workaround + TextureD3D *textureD3D = GetImplAs<TextureD3D>(texture); + gl::SamplerState samplerStateInternal = samplerStateParam; + + TextureStorage *storage = nullptr; + gl::Error error = textureD3D->getNativeTexture(&storage); + if (error.isError()) + { + return error; + } + + // Storage should exist, texture should be complete + ASSERT(storage); + + samplerStateInternal.baseLevel += storage->getTopLevel(); + + if (type == gl::SAMPLER_PIXEL) + { + ASSERT(static_cast<unsigned int>(index) < getRendererCaps().maxTextureImageUnits); + + if (mForceSetPixelSamplerStates[index] || memcmp(&samplerStateInternal, &mCurPixelSamplerStates[index], sizeof(gl::SamplerState)) != 0) + { + ID3D11SamplerState *dxSamplerState = NULL; + error = mStateCache.getSamplerState(samplerStateInternal, &dxSamplerState); + if (error.isError()) + { + return error; + } + + ASSERT(dxSamplerState != NULL); + mDeviceContext->PSSetSamplers(index, 1, &dxSamplerState); + + mCurPixelSamplerStates[index] = samplerStateInternal; + } + + mForceSetPixelSamplerStates[index] = false; + } + else if (type == gl::SAMPLER_VERTEX) + { + ASSERT(static_cast<unsigned int>(index) < getRendererCaps().maxVertexTextureImageUnits); + + if (mForceSetVertexSamplerStates[index] || memcmp(&samplerStateInternal, &mCurVertexSamplerStates[index], sizeof(gl::SamplerState)) != 0) + { + ID3D11SamplerState *dxSamplerState = NULL; + error = mStateCache.getSamplerState(samplerStateInternal, &dxSamplerState); + if (error.isError()) + { + return error; + } + + ASSERT(dxSamplerState != NULL); + mDeviceContext->VSSetSamplers(index, 1, &dxSamplerState); + + mCurVertexSamplerStates[index] = samplerStateInternal; + } + + mForceSetVertexSamplerStates[index] = false; + } + else UNREACHABLE(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setTexture(gl::SamplerType type, int index, gl::Texture *texture) +{ + ID3D11ShaderResourceView *textureSRV = NULL; + + if (texture) + { + TextureD3D *textureImpl = GetImplAs<TextureD3D>(texture); + + TextureStorage *texStorage = nullptr; + gl::Error error = textureImpl->getNativeTexture(&texStorage); + if (error.isError()) + { + return error; + } + + // Texture should be complete and have a storage + ASSERT(texStorage); + + TextureStorage11 *storage11 = TextureStorage11::makeTextureStorage11(texStorage); + + // Make sure to add the level offset for our tiny compressed texture workaround + gl::SamplerState samplerState = texture->getSamplerState(); + samplerState.baseLevel += storage11->getTopLevel(); + + error = storage11->getSRV(samplerState, &textureSRV); + if (error.isError()) + { + return error; + } + + // If we get NULL back from getSRV here, something went wrong in the texture class and we're unexpectedly + // missing the shader resource view + ASSERT(textureSRV != NULL); + + textureImpl->resetDirty(); + } + + ASSERT((type == gl::SAMPLER_PIXEL && static_cast<unsigned int>(index) < getRendererCaps().maxTextureImageUnits) || + (type == gl::SAMPLER_VERTEX && static_cast<unsigned int>(index) < getRendererCaps().maxVertexTextureImageUnits)); + + setShaderResource(type, index, textureSRV); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setUniformBuffers(const gl::Data &data, + const GLint vertexUniformBuffers[], + const GLint fragmentUniformBuffers[]) +{ + for (unsigned int uniformBufferIndex = 0; uniformBufferIndex < data.caps->maxVertexUniformBlocks; uniformBufferIndex++) + { + GLint binding = vertexUniformBuffers[uniformBufferIndex]; + + if (binding == -1) + { + continue; + } + + gl::Buffer *uniformBuffer = data.state->getIndexedUniformBuffer(binding); + GLintptr uniformBufferOffset = data.state->getIndexedUniformBufferOffset(binding); + GLsizeiptr uniformBufferSize = data.state->getIndexedUniformBufferSize(binding); + + if (uniformBuffer) + { + Buffer11 *bufferStorage = Buffer11::makeBuffer11(uniformBuffer->getImplementation()); + ID3D11Buffer *constantBuffer = bufferStorage->getBuffer(BUFFER_USAGE_UNIFORM); + + if (!constantBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY); + } + + if (mCurrentConstantBufferVS[uniformBufferIndex] != bufferStorage->getSerial() || + mCurrentConstantBufferVSOffset[uniformBufferIndex] != uniformBufferOffset || + mCurrentConstantBufferVSSize[uniformBufferIndex] != uniformBufferSize) + { +#if defined(ANGLE_ENABLE_D3D11_1) + if (mSupportsConstantBufferOffsets && uniformBufferSize != 0) + { + UINT firstConstant = 0, numConstants = 0; + CalculateConstantBufferParams(uniformBufferOffset, uniformBufferSize, &firstConstant, &numConstants); + mDeviceContext1->VSSetConstantBuffers1(getReservedVertexUniformBuffers() + uniformBufferIndex, + 1, &constantBuffer, &firstConstant, &numConstants); + } + else +#endif + { + ASSERT(uniformBufferOffset == 0); + mDeviceContext->VSSetConstantBuffers(getReservedVertexUniformBuffers() + uniformBufferIndex, + 1, &constantBuffer); + } + + mCurrentConstantBufferVS[uniformBufferIndex] = bufferStorage->getSerial(); + mCurrentConstantBufferVSOffset[uniformBufferIndex] = uniformBufferOffset; + mCurrentConstantBufferVSSize[uniformBufferIndex] = uniformBufferSize; + } + } + } + + for (unsigned int uniformBufferIndex = 0; uniformBufferIndex < data.caps->maxFragmentUniformBlocks; uniformBufferIndex++) + { + GLint binding = fragmentUniformBuffers[uniformBufferIndex]; + + if (binding == -1) + { + continue; + } + + gl::Buffer *uniformBuffer = data.state->getIndexedUniformBuffer(binding); + GLintptr uniformBufferOffset = data.state->getIndexedUniformBufferOffset(binding); + GLsizeiptr uniformBufferSize = data.state->getIndexedUniformBufferSize(binding); + + if (uniformBuffer) + { + Buffer11 *bufferStorage = Buffer11::makeBuffer11(uniformBuffer->getImplementation()); + ID3D11Buffer *constantBuffer = bufferStorage->getBuffer(BUFFER_USAGE_UNIFORM); + + if (!constantBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY); + } + + if (mCurrentConstantBufferPS[uniformBufferIndex] != bufferStorage->getSerial() || + mCurrentConstantBufferPSOffset[uniformBufferIndex] != uniformBufferOffset || + mCurrentConstantBufferPSSize[uniformBufferIndex] != uniformBufferSize) + { +#if defined(ANGLE_ENABLE_D3D11_1) + if (mSupportsConstantBufferOffsets && uniformBufferSize != 0) + { + UINT firstConstant = 0, numConstants = 0; + CalculateConstantBufferParams(uniformBufferOffset, uniformBufferSize, &firstConstant, &numConstants); + mDeviceContext1->PSSetConstantBuffers1(getReservedFragmentUniformBuffers() + uniformBufferIndex, + 1, &constantBuffer, &firstConstant, &numConstants); + } + else +#endif + { + ASSERT(uniformBufferOffset == 0); + mDeviceContext->PSSetConstantBuffers(getReservedFragmentUniformBuffers() + uniformBufferIndex, + 1, &constantBuffer); + } + + mCurrentConstantBufferPS[uniformBufferIndex] = bufferStorage->getSerial(); + mCurrentConstantBufferPSOffset[uniformBufferIndex] = uniformBufferOffset; + mCurrentConstantBufferPSSize[uniformBufferIndex] = uniformBufferSize; + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setRasterizerState(const gl::RasterizerState &rasterState) +{ + if (mForceSetRasterState || memcmp(&rasterState, &mCurRasterState, sizeof(gl::RasterizerState)) != 0) + { + ID3D11RasterizerState *dxRasterState = NULL; + gl::Error error = mStateCache.getRasterizerState(rasterState, mScissorEnabled, &dxRasterState); + if (error.isError()) + { + return error; + } + + mDeviceContext->RSSetState(dxRasterState); + + mCurRasterState = rasterState; + } + + mForceSetRasterState = false; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setBlendState(const gl::Framebuffer *framebuffer, const gl::BlendState &blendState, const gl::ColorF &blendColor, + unsigned int sampleMask) +{ + if (mForceSetBlendState || + memcmp(&blendState, &mCurBlendState, sizeof(gl::BlendState)) != 0 || + memcmp(&blendColor, &mCurBlendColor, sizeof(gl::ColorF)) != 0 || + sampleMask != mCurSampleMask) + { + ID3D11BlendState *dxBlendState = NULL; + gl::Error error = mStateCache.getBlendState(framebuffer, blendState, &dxBlendState); + if (error.isError()) + { + return error; + } + + ASSERT(dxBlendState != NULL); + + float blendColors[4] = {0.0f}; + if (blendState.sourceBlendRGB != GL_CONSTANT_ALPHA && blendState.sourceBlendRGB != GL_ONE_MINUS_CONSTANT_ALPHA && + blendState.destBlendRGB != GL_CONSTANT_ALPHA && blendState.destBlendRGB != GL_ONE_MINUS_CONSTANT_ALPHA) + { + blendColors[0] = blendColor.red; + blendColors[1] = blendColor.green; + blendColors[2] = blendColor.blue; + blendColors[3] = blendColor.alpha; + } + else + { + blendColors[0] = blendColor.alpha; + blendColors[1] = blendColor.alpha; + blendColors[2] = blendColor.alpha; + blendColors[3] = blendColor.alpha; + } + + mDeviceContext->OMSetBlendState(dxBlendState, blendColors, sampleMask); + + mCurBlendState = blendState; + mCurBlendColor = blendColor; + mCurSampleMask = sampleMask; + } + + mForceSetBlendState = false; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW) +{ + if (mForceSetDepthStencilState || + memcmp(&depthStencilState, &mCurDepthStencilState, sizeof(gl::DepthStencilState)) != 0 || + stencilRef != mCurStencilRef || stencilBackRef != mCurStencilBackRef) + { + ASSERT(depthStencilState.stencilWritemask == depthStencilState.stencilBackWritemask); + ASSERT(stencilRef == stencilBackRef); + ASSERT(depthStencilState.stencilMask == depthStencilState.stencilBackMask); + + ID3D11DepthStencilState *dxDepthStencilState = NULL; + gl::Error error = mStateCache.getDepthStencilState(depthStencilState, &dxDepthStencilState); + if (error.isError()) + { + return error; + } + + ASSERT(dxDepthStencilState); + + // Max D3D11 stencil reference value is 0xFF, corresponding to the max 8 bits in a stencil buffer + // GL specifies we should clamp the ref value to the nearest bit depth when doing stencil ops + static_assert(D3D11_DEFAULT_STENCIL_READ_MASK == 0xFF, "Unexpected value of D3D11_DEFAULT_STENCIL_READ_MASK"); + static_assert(D3D11_DEFAULT_STENCIL_WRITE_MASK == 0xFF, "Unexpected value of D3D11_DEFAULT_STENCIL_WRITE_MASK"); + UINT dxStencilRef = std::min<UINT>(stencilRef, 0xFFu); + + mDeviceContext->OMSetDepthStencilState(dxDepthStencilState, dxStencilRef); + + mCurDepthStencilState = depthStencilState; + mCurStencilRef = stencilRef; + mCurStencilBackRef = stencilBackRef; + } + + mForceSetDepthStencilState = false; + + return gl::Error(GL_NO_ERROR); +} + +void Renderer11::setScissorRectangle(const gl::Rectangle &scissor, bool enabled) +{ + if (mForceSetScissor || memcmp(&scissor, &mCurScissor, sizeof(gl::Rectangle)) != 0 || + enabled != mScissorEnabled) + { + if (enabled) + { + D3D11_RECT rect; + rect.left = std::max(0, scissor.x); + rect.top = std::max(0, scissor.y); + rect.right = scissor.x + std::max(0, scissor.width); + rect.bottom = scissor.y + std::max(0, scissor.height); + + mDeviceContext->RSSetScissorRects(1, &rect); + } + + if (enabled != mScissorEnabled) + { + mForceSetRasterState = true; + } + + mCurScissor = scissor; + mScissorEnabled = enabled; + } + + mForceSetScissor = false; +} + +void Renderer11::setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport) +{ + gl::Rectangle actualViewport = viewport; + float actualZNear = gl::clamp01(zNear); + float actualZFar = gl::clamp01(zFar); + if (ignoreViewport) + { + actualViewport.x = 0; + actualViewport.y = 0; + actualViewport.width = mRenderTargetDesc.width; + actualViewport.height = mRenderTargetDesc.height; + actualZNear = 0.0f; + actualZFar = 1.0f; + } + + bool viewportChanged = mForceSetViewport || memcmp(&actualViewport, &mCurViewport, sizeof(gl::Rectangle)) != 0 || + actualZNear != mCurNear || actualZFar != mCurFar; + + if (viewportChanged) + { + const gl::Caps& caps = getRendererCaps(); + + int dxMaxViewportBoundsX = static_cast<int>(caps.maxViewportWidth); + int dxMaxViewportBoundsY = static_cast<int>(caps.maxViewportHeight); + int dxMinViewportBoundsX = -dxMaxViewportBoundsX; + int dxMinViewportBoundsY = -dxMaxViewportBoundsY; + + if (mFeatureLevel <= D3D_FEATURE_LEVEL_9_3) + { + // Feature Level 9 viewports shouldn't exceed the dimensions of the rendertarget. + dxMaxViewportBoundsX = mRenderTargetDesc.width; + dxMaxViewportBoundsY = mRenderTargetDesc.height; + dxMinViewportBoundsX = 0; + dxMinViewportBoundsY = 0; + } + + int dxViewportTopLeftX = gl::clamp(actualViewport.x, dxMinViewportBoundsX, dxMaxViewportBoundsX); + int dxViewportTopLeftY = gl::clamp(actualViewport.y, dxMinViewportBoundsY, dxMaxViewportBoundsY); + int dxViewportWidth = gl::clamp(actualViewport.width, 0, dxMaxViewportBoundsX - dxViewportTopLeftX); + int dxViewportHeight = gl::clamp(actualViewport.height, 0, dxMaxViewportBoundsY - dxViewportTopLeftY); + + D3D11_VIEWPORT dxViewport; + dxViewport.TopLeftX = static_cast<float>(dxViewportTopLeftX); + dxViewport.TopLeftY = static_cast<float>(dxViewportTopLeftY); + dxViewport.Width = static_cast<float>(dxViewportWidth); + dxViewport.Height = static_cast<float>(dxViewportHeight); + dxViewport.MinDepth = actualZNear; + dxViewport.MaxDepth = actualZFar; + + mDeviceContext->RSSetViewports(1, &dxViewport); + + mCurViewport = actualViewport; + mCurNear = actualZNear; + mCurFar = actualZFar; + + // On Feature Level 9_*, we must emulate large and/or negative viewports in the shaders using viewAdjust (like the D3D9 renderer). + if (mFeatureLevel <= D3D_FEATURE_LEVEL_9_3) + { + mVertexConstants.viewAdjust[0] = static_cast<float>((actualViewport.width - dxViewportWidth) + 2 * (actualViewport.x - dxViewportTopLeftX)) / dxViewport.Width; + mVertexConstants.viewAdjust[1] = static_cast<float>((actualViewport.height - dxViewportHeight) + 2 * (actualViewport.y - dxViewportTopLeftY)) / dxViewport.Height; + mVertexConstants.viewAdjust[2] = static_cast<float>(actualViewport.width) / dxViewport.Width; + mVertexConstants.viewAdjust[3] = static_cast<float>(actualViewport.height) / dxViewport.Height; + } + + mPixelConstants.viewCoords[0] = actualViewport.width * 0.5f; + mPixelConstants.viewCoords[1] = actualViewport.height * 0.5f; + mPixelConstants.viewCoords[2] = actualViewport.x + (actualViewport.width * 0.5f); + mPixelConstants.viewCoords[3] = actualViewport.y + (actualViewport.height * 0.5f); + + // Instanced pointsprite emulation requires ViewCoords to be defined in the + // the vertex shader. + mVertexConstants.viewCoords[0] = mPixelConstants.viewCoords[0]; + mVertexConstants.viewCoords[1] = mPixelConstants.viewCoords[1]; + mVertexConstants.viewCoords[2] = mPixelConstants.viewCoords[2]; + mVertexConstants.viewCoords[3] = mPixelConstants.viewCoords[3]; + + mPixelConstants.depthFront[0] = (actualZFar - actualZNear) * 0.5f; + mPixelConstants.depthFront[1] = (actualZNear + actualZFar) * 0.5f; + + mVertexConstants.depthRange[0] = actualZNear; + mVertexConstants.depthRange[1] = actualZFar; + mVertexConstants.depthRange[2] = actualZFar - actualZNear; + + mPixelConstants.depthRange[0] = actualZNear; + mPixelConstants.depthRange[1] = actualZFar; + mPixelConstants.depthRange[2] = actualZFar - actualZNear; + } + + mForceSetViewport = false; +} + +bool Renderer11::applyPrimitiveType(GLenum mode, GLsizei count, bool usesPointSize) +{ + D3D11_PRIMITIVE_TOPOLOGY primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_UNDEFINED; + + GLsizei minCount = 0; + + switch (mode) + { + case GL_POINTS: primitiveTopology = D3D11_PRIMITIVE_TOPOLOGY_POINTLIST; minCount = 1; break; + case GL_LINES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINELIST; minCount = 2; break; + case GL_LINE_LOOP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; minCount = 2; break; + case GL_LINE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_LINESTRIP; minCount = 2; break; + case GL_TRIANGLES: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; minCount = 3; break; + case GL_TRIANGLE_STRIP: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP; minCount = 3; break; + // emulate fans via rewriting index buffer + case GL_TRIANGLE_FAN: primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; minCount = 3; break; + default: + UNREACHABLE(); + return false; + } + + // If instanced pointsprite emulation is being used and If gl_PointSize is used in the shader, + // GL_POINTS mode is expected to render pointsprites. + // Instanced PointSprite emulation requires that the topology to be D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST. + if (mode == GL_POINTS && usesPointSize && getWorkarounds().useInstancedPointSpriteEmulation) + { + primitiveTopology = D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST; + } + + if (primitiveTopology != mCurrentPrimitiveTopology) + { + mDeviceContext->IASetPrimitiveTopology(primitiveTopology); + mCurrentPrimitiveTopology = primitiveTopology; + } + + return count >= minCount; +} + +void Renderer11::unsetConflictingSRVs(gl::SamplerType samplerType, uintptr_t resource, const gl::ImageIndex *index) +{ + auto ¤tSRVs = (samplerType == gl::SAMPLER_VERTEX ? mCurVertexSRVs : mCurPixelSRVs); + + for (size_t resourceIndex = 0; resourceIndex < currentSRVs.size(); ++resourceIndex) + { + auto &record = currentSRVs[resourceIndex]; + + if (record.srv && record.resource == resource && ImageIndexConflictsWithSRV(index, record.desc)) + { + setShaderResource(samplerType, static_cast<UINT>(resourceIndex), NULL); + } + } +} + +gl::Error Renderer11::applyRenderTarget(const gl::Framebuffer *framebuffer) +{ + // Get the color render buffer and serial + // Also extract the render target dimensions and view + unsigned int renderTargetWidth = 0; + unsigned int renderTargetHeight = 0; + DXGI_FORMAT renderTargetFormat = DXGI_FORMAT_UNKNOWN; + ID3D11RenderTargetView* framebufferRTVs[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS] = {NULL}; + bool missingColorRenderTarget = true; + + const FramebufferD3D *framebufferD3D = GetImplAs<FramebufferD3D>(framebuffer); + const gl::AttachmentList &colorbuffers = framebufferD3D->getColorAttachmentsForRender(getWorkarounds()); + + for (size_t colorAttachment = 0; colorAttachment < colorbuffers.size(); ++colorAttachment) + { + gl::FramebufferAttachment *colorbuffer = colorbuffers[colorAttachment]; + + if (colorbuffer) + { + // the draw buffer must be either "none", "back" for the default buffer or the same index as this color (in order) + + // check for zero-sized default framebuffer, which is a special case. + // in this case we do not wish to modify any state and just silently return false. + // this will not report any gl error but will cause the calling method to return. + if (colorbuffer->getWidth() == 0 || colorbuffer->getHeight() == 0) + { + return gl::Error(GL_NO_ERROR); + } + + // Extract the render target dimensions and view + RenderTarget11 *renderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &renderTarget); + if (error.isError()) + { + return error; + } + ASSERT(renderTarget); + + framebufferRTVs[colorAttachment] = renderTarget->getRenderTargetView(); + ASSERT(framebufferRTVs[colorAttachment]); + + if (missingColorRenderTarget) + { + renderTargetWidth = renderTarget->getWidth(); + renderTargetHeight = renderTarget->getHeight(); + renderTargetFormat = renderTarget->getDXGIFormat(); + missingColorRenderTarget = false; + } + + // Unbind render target SRVs from the shader here to prevent D3D11 warnings. + if (colorbuffer->type() == GL_TEXTURE) + { + uintptr_t rtResource = reinterpret_cast<uintptr_t>(GetViewResource(framebufferRTVs[colorAttachment])); + const gl::ImageIndex *index = colorbuffer->getTextureImageIndex(); + ASSERT(index); + // The index doesn't need to be corrected for the small compressed texture workaround + // because a rendertarget is never compressed. + unsetConflictingSRVs(gl::SAMPLER_VERTEX, rtResource, index); + unsetConflictingSRVs(gl::SAMPLER_PIXEL, rtResource, index); + } + } + } + + // Get the depth stencil buffers + ID3D11DepthStencilView* framebufferDSV = NULL; + gl::FramebufferAttachment *depthStencil = framebuffer->getDepthOrStencilbuffer(); + if (depthStencil) + { + RenderTarget11 *depthStencilRenderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(depthStencil, &depthStencilRenderTarget); + if (error.isError()) + { + SafeRelease(framebufferRTVs); + return error; + } + ASSERT(depthStencilRenderTarget); + + framebufferDSV = depthStencilRenderTarget->getDepthStencilView(); + ASSERT(framebufferDSV); + + // If there is no render buffer, the width, height and format values come from + // the depth stencil + if (missingColorRenderTarget) + { + renderTargetWidth = depthStencilRenderTarget->getWidth(); + renderTargetHeight = depthStencilRenderTarget->getHeight(); + renderTargetFormat = depthStencilRenderTarget->getDXGIFormat(); + } + + // Unbind render target SRVs from the shader here to prevent D3D11 warnings. + if (depthStencil->type() == GL_TEXTURE) + { + uintptr_t depthStencilResource = reinterpret_cast<uintptr_t>(GetViewResource(framebufferDSV)); + const gl::ImageIndex *index = depthStencil->getTextureImageIndex(); + ASSERT(index); + // The index doesn't need to be corrected for the small compressed texture workaround + // because a rendertarget is never compressed. + unsetConflictingSRVs(gl::SAMPLER_VERTEX, depthStencilResource, index); + unsetConflictingSRVs(gl::SAMPLER_PIXEL, depthStencilResource, index); + } + } + + // Apply the render target and depth stencil + if (!mRenderTargetDescInitialized || !mDepthStencilInitialized || + memcmp(framebufferRTVs, mAppliedRTVs, sizeof(framebufferRTVs)) != 0 || + reinterpret_cast<uintptr_t>(framebufferDSV) != mAppliedDSV) + { + mDeviceContext->OMSetRenderTargets(getRendererCaps().maxDrawBuffers, framebufferRTVs, framebufferDSV); + + mRenderTargetDesc.width = renderTargetWidth; + mRenderTargetDesc.height = renderTargetHeight; + mRenderTargetDesc.format = renderTargetFormat; + mForceSetViewport = true; + mForceSetScissor = true; + mForceSetBlendState = true; + + if (!mDepthStencilInitialized) + { + mForceSetRasterState = true; + } + + for (size_t rtIndex = 0; rtIndex < ArraySize(framebufferRTVs); rtIndex++) + { + mAppliedRTVs[rtIndex] = reinterpret_cast<uintptr_t>(framebufferRTVs[rtIndex]); + } + mAppliedDSV = reinterpret_cast<uintptr_t>(framebufferDSV); + mRenderTargetDescInitialized = true; + mDepthStencilInitialized = true; + } + + const Framebuffer11 *framebuffer11 = GetImplAs<Framebuffer11>(framebuffer); + gl::Error error = framebuffer11->invalidateSwizzles(); + if (error.isError()) + { + return error; + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::applyVertexBuffer(const gl::State &state, GLenum mode, GLint first, GLsizei count, GLsizei instances) +{ + TranslatedAttribute attributes[gl::MAX_VERTEX_ATTRIBS]; + gl::Error error = mVertexDataManager->prepareVertexData(state, first, count, attributes, instances); + if (error.isError()) + { + return error; + } + + return mInputLayoutCache.applyVertexBuffers(attributes, mode, state.getProgram()); +} + +gl::Error Renderer11::applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo) +{ + gl::Error error = mIndexDataManager->prepareIndexData(type, count, elementArrayBuffer, indices, indexInfo); + if (error.isError()) + { + return error; + } + + ID3D11Buffer *buffer = NULL; + DXGI_FORMAT bufferFormat = (indexInfo->indexType == GL_UNSIGNED_INT) ? DXGI_FORMAT_R32_UINT : DXGI_FORMAT_R16_UINT; + + if (indexInfo->storage) + { + Buffer11 *storage = Buffer11::makeBuffer11(indexInfo->storage); + buffer = storage->getBuffer(BUFFER_USAGE_INDEX); + } + else + { + IndexBuffer11* indexBuffer = IndexBuffer11::makeIndexBuffer11(indexInfo->indexBuffer); + buffer = indexBuffer->getBuffer(); + } + + if (buffer != mAppliedIB || bufferFormat != mAppliedIBFormat || indexInfo->startOffset != mAppliedIBOffset) + { + mDeviceContext->IASetIndexBuffer(buffer, bufferFormat, indexInfo->startOffset); + + mAppliedIB = buffer; + mAppliedIBFormat = bufferFormat; + mAppliedIBOffset = indexInfo->startOffset; + } + + return gl::Error(GL_NO_ERROR); +} + +void Renderer11::applyTransformFeedbackBuffers(const gl::State &state) +{ + size_t numXFBBindings = 0; + bool requiresUpdate = false; + + if (state.isTransformFeedbackActiveUnpaused()) + { + numXFBBindings = state.getTransformFeedbackBufferIndexRange(); + ASSERT(numXFBBindings <= gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS); + + for (size_t i = 0; i < numXFBBindings; i++) + { + gl::Buffer *curXFBBuffer = state.getIndexedTransformFeedbackBuffer(i); + GLintptr curXFBOffset = state.getIndexedTransformFeedbackBufferOffset(i); + ID3D11Buffer *d3dBuffer = NULL; + if (curXFBBuffer) + { + Buffer11 *storage = Buffer11::makeBuffer11(curXFBBuffer->getImplementation()); + d3dBuffer = storage->getBuffer(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK); + } + + // TODO: mAppliedTFBuffers and friends should also be kept in a vector. + if (d3dBuffer != mAppliedTFBuffers[i] || curXFBOffset != mAppliedTFOffsets[i]) + { + requiresUpdate = true; + } + } + } + + if (requiresUpdate || numXFBBindings != mAppliedNumXFBBindings) + { + for (size_t i = 0; i < numXFBBindings; ++i) + { + gl::Buffer *curXFBBuffer = state.getIndexedTransformFeedbackBuffer(i); + GLintptr curXFBOffset = state.getIndexedTransformFeedbackBufferOffset(i); + + if (curXFBBuffer) + { + Buffer11 *storage = Buffer11::makeBuffer11(curXFBBuffer->getImplementation()); + ID3D11Buffer *d3dBuffer = storage->getBuffer(BUFFER_USAGE_VERTEX_OR_TRANSFORM_FEEDBACK); + + mCurrentD3DOffsets[i] = (mAppliedTFBuffers[i] != d3dBuffer || mAppliedTFOffsets[i] != curXFBOffset) ? + static_cast<UINT>(curXFBOffset) : -1; + mAppliedTFBuffers[i] = d3dBuffer; + } + else + { + mAppliedTFBuffers[i] = NULL; + mCurrentD3DOffsets[i] = 0; + } + mAppliedTFOffsets[i] = curXFBOffset; + } + + mAppliedNumXFBBindings = numXFBBindings; + + mDeviceContext->SOSetTargets(numXFBBindings, mAppliedTFBuffers, mCurrentD3DOffsets); + } +} + +gl::Error Renderer11::drawArrays(const gl::Data &data, GLenum mode, GLsizei count, GLsizei instances, bool usesPointSize) +{ + bool useInstancedPointSpriteEmulation = usesPointSize && getWorkarounds().useInstancedPointSpriteEmulation; + if (mode == GL_POINTS && data.state->isTransformFeedbackActiveUnpaused()) + { + // Since point sprites are generated with a geometry shader, too many vertices will + // be written if transform feedback is active. To work around this, draw only the points + // with the stream out shader and no pixel shader to feed the stream out buffers and then + // draw again with the point sprite geometry shader to rasterize the point sprites. + + mDeviceContext->PSSetShader(NULL, NULL, 0); + + if (instances > 0) + { + mDeviceContext->DrawInstanced(count, instances, 0, 0); + } + else + { + mDeviceContext->Draw(count, 0); + } + + ProgramD3D *programD3D = GetImplAs<ProgramD3D>(data.state->getProgram()); + + rx::ShaderExecutableD3D *pixelExe = NULL; + gl::Error error = programD3D->getPixelExecutableForFramebuffer(data.state->getDrawFramebuffer(), &pixelExe); + if (error.isError()) + { + return error; + } + + // Skip this step if we're doing rasterizer discard. + if (pixelExe && !data.state->getRasterizerState().rasterizerDiscard && usesPointSize) + { + ID3D11PixelShader *pixelShader = ShaderExecutable11::makeShaderExecutable11(pixelExe)->getPixelShader(); + ASSERT(reinterpret_cast<uintptr_t>(pixelShader) == mAppliedPixelShader); + mDeviceContext->PSSetShader(pixelShader, NULL, 0); + + // Retrieve the point sprite geometry shader + rx::ShaderExecutableD3D *geometryExe = programD3D->getGeometryExecutable(); + ID3D11GeometryShader *geometryShader = (geometryExe ? ShaderExecutable11::makeShaderExecutable11(geometryExe)->getGeometryShader() : NULL); + mAppliedGeometryShader = reinterpret_cast<uintptr_t>(geometryShader); + ASSERT(geometryShader); + mDeviceContext->GSSetShader(geometryShader, NULL, 0); + + if (instances > 0) + { + mDeviceContext->DrawInstanced(count, instances, 0, 0); + } + else + { + mDeviceContext->Draw(count, 0); + } + } + + return gl::Error(GL_NO_ERROR); + } + else if (mode == GL_LINE_LOOP) + { + return drawLineLoop(count, GL_NONE, NULL, 0, NULL); + } + else if (mode == GL_TRIANGLE_FAN) + { + return drawTriangleFan(count, GL_NONE, NULL, 0, NULL, instances); + } + else if (instances > 0) + { + mDeviceContext->DrawInstanced(count, instances, 0, 0); + return gl::Error(GL_NO_ERROR); + } + else + { + // If gl_PointSize is used and GL_POINTS is specified, then it is expected to render pointsprites. + // If instanced pointsprite emulation is being used the topology is expexted to be + // D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST and DrawIndexedInstanced must be used. + if (mode == GL_POINTS && useInstancedPointSpriteEmulation) + { + mDeviceContext->DrawIndexedInstanced(6, count, 0, 0, 0); + } + else + { + mDeviceContext->Draw(count, 0); + } + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error Renderer11::drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, + gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances) +{ + int minIndex = static_cast<int>(indexInfo.indexRange.start); + + if (mode == GL_LINE_LOOP) + { + return drawLineLoop(count, type, indices, minIndex, elementArrayBuffer); + } + else if (mode == GL_TRIANGLE_FAN) + { + return drawTriangleFan(count, type, indices, minIndex, elementArrayBuffer, instances); + } + else if (instances > 0) + { + mDeviceContext->DrawIndexedInstanced(count, instances, 0, -minIndex, 0); + return gl::Error(GL_NO_ERROR); + } + else + { + mDeviceContext->DrawIndexed(count, 0, -minIndex); + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error Renderer11::drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer) +{ + // Get the raw indices for an indexed draw + if (type != GL_NONE && elementArrayBuffer) + { + BufferD3D *storage = GetImplAs<BufferD3D>(elementArrayBuffer); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + + const uint8_t *bufferData = NULL; + gl::Error error = storage->getData(&bufferData); + if (error.isError()) + { + return error; + } + + indices = bufferData + offset; + } + + if (!mLineLoopIB) + { + mLineLoopIB = new StreamingIndexBufferInterface(this); + gl::Error error = mLineLoopIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT); + if (error.isError()) + { + SafeDelete(mLineLoopIB); + return error; + } + } + + // Checked by Renderer11::applyPrimitiveType + ASSERT(count >= 0); + + if (static_cast<unsigned int>(count) + 1 > (std::numeric_limits<unsigned int>::max() / sizeof(unsigned int))) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create a 32-bit looping index buffer for GL_LINE_LOOP, too many indices required."); + } + + const unsigned int spaceNeeded = (static_cast<unsigned int>(count) + 1) * sizeof(unsigned int); + gl::Error error = mLineLoopIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT); + if (error.isError()) + { + return error; + } + + void* mappedMemory = NULL; + unsigned int offset; + error = mLineLoopIB->mapBuffer(spaceNeeded, &mappedMemory, &offset); + if (error.isError()) + { + return error; + } + + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + unsigned int indexBufferOffset = offset; + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (int i = 0; i < count; i++) + { + data[i] = i; + } + data[count] = 0; + break; + case GL_UNSIGNED_BYTE: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLubyte*>(indices)[i]; + } + data[count] = static_cast<const GLubyte*>(indices)[0]; + break; + case GL_UNSIGNED_SHORT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLushort*>(indices)[i]; + } + data[count] = static_cast<const GLushort*>(indices)[0]; + break; + case GL_UNSIGNED_INT: + for (int i = 0; i < count; i++) + { + data[i] = static_cast<const GLuint*>(indices)[i]; + } + data[count] = static_cast<const GLuint*>(indices)[0]; + break; + default: UNREACHABLE(); + } + + error = mLineLoopIB->unmapBuffer(); + if (error.isError()) + { + return error; + } + + IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mLineLoopIB->getIndexBuffer()); + ID3D11Buffer *d3dIndexBuffer = indexBuffer->getBuffer(); + DXGI_FORMAT indexFormat = indexBuffer->getIndexFormat(); + + if (mAppliedIB != d3dIndexBuffer || mAppliedIBFormat != indexFormat || mAppliedIBOffset != indexBufferOffset) + { + mDeviceContext->IASetIndexBuffer(d3dIndexBuffer, indexFormat, indexBufferOffset); + mAppliedIB = d3dIndexBuffer; + mAppliedIBFormat = indexFormat; + mAppliedIBOffset = indexBufferOffset; + } + + mDeviceContext->DrawIndexed(count + 1, 0, -minIndex); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::drawTriangleFan(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer, int instances) +{ + // Get the raw indices for an indexed draw + if (type != GL_NONE && elementArrayBuffer) + { + BufferD3D *storage = GetImplAs<BufferD3D>(elementArrayBuffer); + intptr_t offset = reinterpret_cast<intptr_t>(indices); + + const uint8_t *bufferData = NULL; + gl::Error error = storage->getData(&bufferData); + if (error.isError()) + { + return error; + } + + indices = bufferData + offset; + } + + if (!mTriangleFanIB) + { + mTriangleFanIB = new StreamingIndexBufferInterface(this); + gl::Error error = mTriangleFanIB->reserveBufferSpace(INITIAL_INDEX_BUFFER_SIZE, GL_UNSIGNED_INT); + if (error.isError()) + { + SafeDelete(mTriangleFanIB); + return error; + } + } + + // Checked by Renderer11::applyPrimitiveType + ASSERT(count >= 3); + + const unsigned int numTris = count - 2; + + if (numTris > (std::numeric_limits<unsigned int>::max() / (sizeof(unsigned int) * 3))) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create a scratch index buffer for GL_TRIANGLE_FAN, too many indices required."); + } + + const unsigned int spaceNeeded = (numTris * 3) * sizeof(unsigned int); + gl::Error error = mTriangleFanIB->reserveBufferSpace(spaceNeeded, GL_UNSIGNED_INT); + if (error.isError()) + { + return error; + } + + void* mappedMemory = NULL; + unsigned int offset; + error = mTriangleFanIB->mapBuffer(spaceNeeded, &mappedMemory, &offset); + if (error.isError()) + { + return error; + } + + unsigned int *data = reinterpret_cast<unsigned int*>(mappedMemory); + unsigned int indexBufferOffset = offset; + + switch (type) + { + case GL_NONE: // Non-indexed draw + for (unsigned int i = 0; i < numTris; i++) + { + data[i*3 + 0] = 0; + data[i*3 + 1] = i + 1; + data[i*3 + 2] = i + 2; + } + break; + case GL_UNSIGNED_BYTE: + for (unsigned int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLubyte*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLubyte*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLubyte*>(indices)[i + 2]; + } + break; + case GL_UNSIGNED_SHORT: + for (unsigned int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLushort*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLushort*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLushort*>(indices)[i + 2]; + } + break; + case GL_UNSIGNED_INT: + for (unsigned int i = 0; i < numTris; i++) + { + data[i*3 + 0] = static_cast<const GLuint*>(indices)[0]; + data[i*3 + 1] = static_cast<const GLuint*>(indices)[i + 1]; + data[i*3 + 2] = static_cast<const GLuint*>(indices)[i + 2]; + } + break; + default: UNREACHABLE(); + } + + error = mTriangleFanIB->unmapBuffer(); + if (error.isError()) + { + return error; + } + + IndexBuffer11 *indexBuffer = IndexBuffer11::makeIndexBuffer11(mTriangleFanIB->getIndexBuffer()); + ID3D11Buffer *d3dIndexBuffer = indexBuffer->getBuffer(); + DXGI_FORMAT indexFormat = indexBuffer->getIndexFormat(); + + if (mAppliedIB != d3dIndexBuffer || mAppliedIBFormat != indexFormat || mAppliedIBOffset != indexBufferOffset) + { + mDeviceContext->IASetIndexBuffer(d3dIndexBuffer, indexFormat, indexBufferOffset); + mAppliedIB = d3dIndexBuffer; + mAppliedIBFormat = indexFormat; + mAppliedIBOffset = indexBufferOffset; + } + + if (instances > 0) + { + mDeviceContext->DrawIndexedInstanced(numTris * 3, instances, 0, -minIndex, 0); + } + else + { + mDeviceContext->DrawIndexed(numTris * 3, 0, -minIndex); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::applyShaders(gl::Program *program, const gl::VertexFormat inputLayout[], const gl::Framebuffer *framebuffer, + bool rasterizerDiscard, bool transformFeedbackActive) +{ + ProgramD3D *programD3D = GetImplAs<ProgramD3D>(program); + + ShaderExecutableD3D *vertexExe = NULL; + gl::Error error = programD3D->getVertexExecutableForInputLayout(inputLayout, &vertexExe, nullptr); + if (error.isError()) + { + return error; + } + + ShaderExecutableD3D *pixelExe = NULL; + error = programD3D->getPixelExecutableForFramebuffer(framebuffer, &pixelExe); + if (error.isError()) + { + return error; + } + + ShaderExecutableD3D *geometryExe = programD3D->getGeometryExecutable(); + + ID3D11VertexShader *vertexShader = (vertexExe ? ShaderExecutable11::makeShaderExecutable11(vertexExe)->getVertexShader() : NULL); + + ID3D11PixelShader *pixelShader = NULL; + // Skip pixel shader if we're doing rasterizer discard. + if (!rasterizerDiscard) + { + pixelShader = (pixelExe ? ShaderExecutable11::makeShaderExecutable11(pixelExe)->getPixelShader() : NULL); + } + + ID3D11GeometryShader *geometryShader = NULL; + if (transformFeedbackActive) + { + geometryShader = (vertexExe ? ShaderExecutable11::makeShaderExecutable11(vertexExe)->getStreamOutShader() : NULL); + } + else if (mCurRasterState.pointDrawMode) + { + geometryShader = (geometryExe ? ShaderExecutable11::makeShaderExecutable11(geometryExe)->getGeometryShader() : NULL); + } + + bool dirtyUniforms = false; + + if (reinterpret_cast<uintptr_t>(vertexShader) != mAppliedVertexShader) + { + mDeviceContext->VSSetShader(vertexShader, NULL, 0); + mAppliedVertexShader = reinterpret_cast<uintptr_t>(vertexShader); + dirtyUniforms = true; + } + + if (reinterpret_cast<uintptr_t>(geometryShader) != mAppliedGeometryShader) + { + mDeviceContext->GSSetShader(geometryShader, NULL, 0); + mAppliedGeometryShader = reinterpret_cast<uintptr_t>(geometryShader); + dirtyUniforms = true; + } + + if (reinterpret_cast<uintptr_t>(pixelShader) != mAppliedPixelShader) + { + mDeviceContext->PSSetShader(pixelShader, NULL, 0); + mAppliedPixelShader = reinterpret_cast<uintptr_t>(pixelShader); + dirtyUniforms = true; + } + + if (dirtyUniforms) + { + programD3D->dirtyAllUniforms(); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::applyUniforms(const ProgramImpl &program, const std::vector<gl::LinkedUniform*> &uniformArray) +{ + unsigned int totalRegisterCountVS = 0; + unsigned int totalRegisterCountPS = 0; + + bool vertexUniformsDirty = false; + bool pixelUniformsDirty = false; + + for (size_t uniformIndex = 0; uniformIndex < uniformArray.size(); uniformIndex++) + { + const gl::LinkedUniform &uniform = *uniformArray[uniformIndex]; + + if (uniform.isReferencedByVertexShader() && !uniform.isSampler()) + { + totalRegisterCountVS += uniform.registerCount; + vertexUniformsDirty = (vertexUniformsDirty || uniform.dirty); + } + + if (uniform.isReferencedByFragmentShader() && !uniform.isSampler()) + { + totalRegisterCountPS += uniform.registerCount; + pixelUniformsDirty = (pixelUniformsDirty || uniform.dirty); + } + } + + const ProgramD3D *programD3D = GetAs<ProgramD3D>(&program); + const UniformStorage11 *vertexUniformStorage = UniformStorage11::makeUniformStorage11(&programD3D->getVertexUniformStorage()); + const UniformStorage11 *fragmentUniformStorage = UniformStorage11::makeUniformStorage11(&programD3D->getFragmentUniformStorage()); + ASSERT(vertexUniformStorage); + ASSERT(fragmentUniformStorage); + + ID3D11Buffer *vertexConstantBuffer = vertexUniformStorage->getConstantBuffer(); + ID3D11Buffer *pixelConstantBuffer = fragmentUniformStorage->getConstantBuffer(); + + float (*mapVS)[4] = NULL; + float (*mapPS)[4] = NULL; + + if (totalRegisterCountVS > 0 && vertexUniformsDirty) + { + D3D11_MAPPED_SUBRESOURCE map = {0}; + HRESULT result = mDeviceContext->Map(vertexConstantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &map); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + mapVS = (float(*)[4])map.pData; + } + + if (totalRegisterCountPS > 0 && pixelUniformsDirty) + { + D3D11_MAPPED_SUBRESOURCE map = {0}; + HRESULT result = mDeviceContext->Map(pixelConstantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &map); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + mapPS = (float(*)[4])map.pData; + } + + for (size_t uniformIndex = 0; uniformIndex < uniformArray.size(); uniformIndex++) + { + gl::LinkedUniform *uniform = uniformArray[uniformIndex]; + + if (!uniform->isSampler()) + { + unsigned int componentCount = (4 - uniform->registerElement); + + // we assume that uniforms from structs are arranged in struct order in our uniforms list. otherwise we would + // overwrite previously written regions of memory. + + if (uniform->isReferencedByVertexShader() && mapVS) + { + memcpy(&mapVS[uniform->vsRegisterIndex][uniform->registerElement], uniform->data, uniform->registerCount * sizeof(float) * componentCount); + } + + if (uniform->isReferencedByFragmentShader() && mapPS) + { + memcpy(&mapPS[uniform->psRegisterIndex][uniform->registerElement], uniform->data, uniform->registerCount * sizeof(float) * componentCount); + } + } + } + + if (mapVS) + { + mDeviceContext->Unmap(vertexConstantBuffer, 0); + } + + if (mapPS) + { + mDeviceContext->Unmap(pixelConstantBuffer, 0); + } + + if (mCurrentVertexConstantBuffer != vertexConstantBuffer) + { + mDeviceContext->VSSetConstantBuffers(0, 1, &vertexConstantBuffer); + mCurrentVertexConstantBuffer = vertexConstantBuffer; + } + + if (mCurrentPixelConstantBuffer != pixelConstantBuffer) + { + mDeviceContext->PSSetConstantBuffers(0, 1, &pixelConstantBuffer); + mCurrentPixelConstantBuffer = pixelConstantBuffer; + } + + // Driver uniforms + if (!mDriverConstantBufferVS) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = sizeof(dx_VertexConstants); + constantBufferDescription.Usage = D3D11_USAGE_DEFAULT; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = 0; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = mDevice->CreateBuffer(&constantBufferDescription, NULL, &mDriverConstantBufferVS); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + + mDeviceContext->VSSetConstantBuffers(1, 1, &mDriverConstantBufferVS); + } + + if (!mDriverConstantBufferPS) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = sizeof(dx_PixelConstants); + constantBufferDescription.Usage = D3D11_USAGE_DEFAULT; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = 0; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = mDevice->CreateBuffer(&constantBufferDescription, NULL, &mDriverConstantBufferPS); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + + mDeviceContext->PSSetConstantBuffers(1, 1, &mDriverConstantBufferPS); + } + + if (memcmp(&mVertexConstants, &mAppliedVertexConstants, sizeof(dx_VertexConstants)) != 0) + { + mDeviceContext->UpdateSubresource(mDriverConstantBufferVS, 0, NULL, &mVertexConstants, 16, 0); + memcpy(&mAppliedVertexConstants, &mVertexConstants, sizeof(dx_VertexConstants)); + } + + if (memcmp(&mPixelConstants, &mAppliedPixelConstants, sizeof(dx_PixelConstants)) != 0) + { + mDeviceContext->UpdateSubresource(mDriverConstantBufferPS, 0, NULL, &mPixelConstants, 16, 0); + memcpy(&mAppliedPixelConstants, &mPixelConstants, sizeof(dx_PixelConstants)); + } + + // GSSetConstantBuffers triggers device removal on 9_3, so we should only call it if necessary + if (programD3D->usesGeometryShader()) + { + // needed for the point sprite geometry shader + if (mCurrentGeometryConstantBuffer != mDriverConstantBufferPS) + { + mDeviceContext->GSSetConstantBuffers(0, 1, &mDriverConstantBufferPS); + mCurrentGeometryConstantBuffer = mDriverConstantBufferPS; + } + } + + return gl::Error(GL_NO_ERROR); +} + +void Renderer11::markAllStateDirty() +{ + for (size_t rtIndex = 0; rtIndex < ArraySize(mAppliedRTVs); rtIndex++) + { + mAppliedRTVs[rtIndex] = DirtyPointer; + } + mAppliedDSV = DirtyPointer; + mDepthStencilInitialized = false; + mRenderTargetDescInitialized = false; + + // We reset the current SRV data because it might not be in sync with D3D's state + // anymore. For example when a currently used SRV is used as an RTV, D3D silently + // remove it from its state. + memset(mCurVertexSRVs.data(), 0, sizeof(SRVRecord) * mCurVertexSRVs.size()); + memset(mCurPixelSRVs.data(), 0, sizeof(SRVRecord) * mCurPixelSRVs.size()); + + ASSERT(mForceSetVertexSamplerStates.size() == mCurVertexSRVs.size()); + for (size_t vsamplerId = 0; vsamplerId < mForceSetVertexSamplerStates.size(); ++vsamplerId) + { + mForceSetVertexSamplerStates[vsamplerId] = true; + } + + ASSERT(mForceSetPixelSamplerStates.size() == mCurPixelSRVs.size()); + for (size_t fsamplerId = 0; fsamplerId < mForceSetPixelSamplerStates.size(); ++fsamplerId) + { + mForceSetPixelSamplerStates[fsamplerId] = true; + } + + mForceSetBlendState = true; + mForceSetRasterState = true; + mForceSetDepthStencilState = true; + mForceSetScissor = true; + mForceSetViewport = true; + + mAppliedIB = NULL; + mAppliedIBFormat = DXGI_FORMAT_UNKNOWN; + mAppliedIBOffset = 0; + + mAppliedVertexShader = DirtyPointer; + mAppliedGeometryShader = DirtyPointer; + mAppliedPixelShader = DirtyPointer; + + mAppliedNumXFBBindings = static_cast<size_t>(-1); + + for (size_t i = 0; i < gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS; i++) + { + mAppliedTFBuffers[i] = NULL; + mAppliedTFOffsets[i] = 0; + } + + memset(&mAppliedVertexConstants, 0, sizeof(dx_VertexConstants)); + memset(&mAppliedPixelConstants, 0, sizeof(dx_PixelConstants)); + + mInputLayoutCache.markDirty(); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS; i++) + { + mCurrentConstantBufferVS[i] = static_cast<unsigned int>(-1); + mCurrentConstantBufferVSOffset[i] = 0; + mCurrentConstantBufferVSSize[i] = 0; + mCurrentConstantBufferPS[i] = static_cast<unsigned int>(-1); + mCurrentConstantBufferPSOffset[i] = 0; + mCurrentConstantBufferPSSize[i] = 0; + } + + mCurrentVertexConstantBuffer = NULL; + mCurrentPixelConstantBuffer = NULL; + mCurrentGeometryConstantBuffer = NULL; + + mCurrentPrimitiveTopology = D3D_PRIMITIVE_TOPOLOGY_UNDEFINED; +} + +void Renderer11::releaseDeviceResources() +{ + mStateCache.clear(); + mInputLayoutCache.clear(); + + SafeDelete(mVertexDataManager); + SafeDelete(mIndexDataManager); + SafeDelete(mLineLoopIB); + SafeDelete(mTriangleFanIB); + SafeDelete(mBlit); + SafeDelete(mClear); + SafeDelete(mTrim); + SafeDelete(mPixelTransfer); + + SafeRelease(mDriverConstantBufferVS); + SafeRelease(mDriverConstantBufferPS); + SafeRelease(mSyncQuery); +} + +// set notify to true to broadcast a message to all contexts of the device loss +bool Renderer11::testDeviceLost() +{ + bool isLost = false; + + // GetRemovedReason is used to test if the device is removed + HRESULT result = mDevice->GetDeviceRemovedReason(); + isLost = d3d11::isDeviceLostError(result); + + if (isLost) + { + // Log error if this is a new device lost event + if (mDeviceLost == false) + { + ERR("The D3D11 device was removed: 0x%08X", result); + } + + // ensure we note the device loss -- + // we'll probably get this done again by notifyDeviceLost + // but best to remember it! + // Note that we don't want to clear the device loss status here + // -- this needs to be done by resetDevice + mDeviceLost = true; + } + + return isLost; +} + +bool Renderer11::testDeviceResettable() +{ + // determine if the device is resettable by creating a dummy device + PFN_D3D11_CREATE_DEVICE D3D11CreateDevice = (PFN_D3D11_CREATE_DEVICE)GetProcAddress(mD3d11Module, "D3D11CreateDevice"); + + if (D3D11CreateDevice == NULL) + { + return false; + } + + ID3D11Device* dummyDevice; + D3D_FEATURE_LEVEL dummyFeatureLevel; + ID3D11DeviceContext* dummyContext; + + HRESULT result = D3D11CreateDevice(NULL, + mDriverType, + NULL, + #if defined(_DEBUG) + D3D11_CREATE_DEVICE_DEBUG, + #else + 0, + #endif + mAvailableFeatureLevels.data(), + mAvailableFeatureLevels.size(), + D3D11_SDK_VERSION, + &dummyDevice, + &dummyFeatureLevel, + &dummyContext); + + if (!mDevice || FAILED(result)) + { + return false; + } + + SafeRelease(dummyContext); + SafeRelease(dummyDevice); + + return true; +} + +void Renderer11::release() +{ + RendererD3D::cleanup(); + + releaseDeviceResources(); + + SafeRelease(mDxgiFactory); + SafeRelease(mDxgiAdapter); + +#if defined(ANGLE_ENABLE_D3D11_1) + SafeRelease(mDeviceContext1); +#endif + + if (mDeviceContext) + { + mDeviceContext->ClearState(); + mDeviceContext->Flush(); + SafeRelease(mDeviceContext); + } + + SafeRelease(mDevice); + + if (mD3d11Module) + { + FreeLibrary(mD3d11Module); + mD3d11Module = NULL; + } + + if (mDxgiModule) + { + FreeLibrary(mDxgiModule); + mDxgiModule = NULL; + } + + mCompiler.release(); +} + +bool Renderer11::resetDevice() +{ + // recreate everything + release(); + egl::Error result = initialize(); + + if (result.isError()) + { + ERR("Could not reinitialize D3D11 device: %08X", result.getCode()); + return false; + } + + mDeviceLost = false; + + return true; +} + +VendorID Renderer11::getVendorId() const +{ + return static_cast<VendorID>(mAdapterDescription.VendorId); +} + +std::string Renderer11::getRendererDescription() const +{ + std::ostringstream rendererString; + + rendererString << mDescription; + rendererString << " Direct3D11"; + + rendererString << " vs_" << getMajorShaderModel() << "_" << getMinorShaderModel() << getShaderModelSuffix(); + rendererString << " ps_" << getMajorShaderModel() << "_" << getMinorShaderModel() << getShaderModelSuffix(); + + return rendererString.str(); +} + +GUID Renderer11::getAdapterIdentifier() const +{ + // Use the adapter LUID as our adapter ID + // This number is local to a machine is only guaranteed to be unique between restarts + static_assert(sizeof(LUID) <= sizeof(GUID), "Size of GUID must be at least as large as LUID."); + GUID adapterId = {0}; + memcpy(&adapterId, &mAdapterDescription.AdapterLuid, sizeof(LUID)); + return adapterId; +} + +unsigned int Renderer11::getReservedVertexUniformVectors() const +{ + return 0; // Driver uniforms are stored in a separate constant buffer +} + +unsigned int Renderer11::getReservedFragmentUniformVectors() const +{ + return 0; // Driver uniforms are stored in a separate constant buffer +} + +unsigned int Renderer11::getReservedVertexUniformBuffers() const +{ + // we reserve one buffer for the application uniforms, and one for driver uniforms + return 2; +} + +unsigned int Renderer11::getReservedFragmentUniformBuffers() const +{ + // we reserve one buffer for the application uniforms, and one for driver uniforms + return 2; +} + +bool Renderer11::getShareHandleSupport() const +{ + // We only currently support share handles with BGRA surfaces, because + // chrome needs BGRA. Once chrome fixes this, we should always support them. + // PIX doesn't seem to support using share handles, so disable them. + // Also disable share handles on Feature Level 9_3, since it doesn't support share handles on RGBA8 textures/swapchains. + return getRendererExtensions().textureFormatBGRA8888 && !gl::DebugAnnotationsActive();// && !(mFeatureLevel <= D3D_FEATURE_LEVEL_9_3); Qt: we don't care about the 9_3 limitation +} + +bool Renderer11::getPostSubBufferSupport() const +{ + // D3D11 does not support present with dirty rectangles until D3D11.1 and DXGI 1.2. + return false; +} + +int Renderer11::getMajorShaderModel() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_SHADER_MAJOR_VERSION; // 5 + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_SHADER_MAJOR_VERSION; // 4 + case D3D_FEATURE_LEVEL_10_0: return D3D10_SHADER_MAJOR_VERSION; // 4 + case D3D_FEATURE_LEVEL_9_3: return D3D10_SHADER_MAJOR_VERSION; // 4 + default: UNREACHABLE(); return 0; + } +} + +int Renderer11::getMinorShaderModel() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return D3D11_SHADER_MINOR_VERSION; // 0 + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_SHADER_MINOR_VERSION; // 1 + case D3D_FEATURE_LEVEL_10_0: return D3D10_SHADER_MINOR_VERSION; // 0 + case D3D_FEATURE_LEVEL_9_3: return D3D10_SHADER_MINOR_VERSION; // 0 + default: UNREACHABLE(); return 0; + } +} + +std::string Renderer11::getShaderModelSuffix() const +{ + switch (mFeatureLevel) + { + case D3D_FEATURE_LEVEL_11_0: return ""; + case D3D_FEATURE_LEVEL_10_1: return ""; + case D3D_FEATURE_LEVEL_10_0: return ""; + case D3D_FEATURE_LEVEL_9_3: return "_level_9_3"; + default: UNREACHABLE(); return ""; + } +} + +gl::Error Renderer11::copyImage2D(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level) +{ + gl::FramebufferAttachment *colorbuffer = framebuffer->getReadColorbuffer(); + ASSERT(colorbuffer); + + RenderTarget11 *sourceRenderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &sourceRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(sourceRenderTarget); + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + ASSERT(source); + + TextureStorage11_2D *storage11 = TextureStorage11_2D::makeTextureStorage11_2D(storage); + ASSERT(storage11); + + gl::ImageIndex index = gl::ImageIndex::Make2D(level); + RenderTargetD3D *destRenderTarget = NULL; + error = storage11->getRenderTarget(index, &destRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(destRenderTarget); + + ID3D11RenderTargetView *dest = RenderTarget11::makeRenderTarget11(destRenderTarget)->getRenderTargetView(); + ASSERT(dest); + + gl::Box sourceArea(sourceRect.x, sourceRect.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents sourceSize(sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), 1); + + gl::Box destArea(destOffset.x, destOffset.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents destSize(destRenderTarget->getWidth(), destRenderTarget->getHeight(), 1); + + // Use nearest filtering because source and destination are the same size for the direct + // copy + mBlit->copyTexture(source, sourceArea, sourceSize, dest, destArea, destSize, NULL, destFormat, GL_NEAREST); + if (error.isError()) + { + return error; + } + + storage11->invalidateSwizzleCacheLevel(level); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::copyImageCube(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLenum target, GLint level) +{ + gl::FramebufferAttachment *colorbuffer = framebuffer->getReadColorbuffer(); + ASSERT(colorbuffer); + + RenderTarget11 *sourceRenderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &sourceRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(sourceRenderTarget); + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + ASSERT(source); + + TextureStorage11_Cube *storage11 = TextureStorage11_Cube::makeTextureStorage11_Cube(storage); + ASSERT(storage11); + + gl::ImageIndex index = gl::ImageIndex::MakeCube(target, level); + RenderTargetD3D *destRenderTarget = NULL; + error = storage11->getRenderTarget(index, &destRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(destRenderTarget); + + ID3D11RenderTargetView *dest = RenderTarget11::makeRenderTarget11(destRenderTarget)->getRenderTargetView(); + ASSERT(dest); + + gl::Box sourceArea(sourceRect.x, sourceRect.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents sourceSize(sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), 1); + + gl::Box destArea(destOffset.x, destOffset.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents destSize(destRenderTarget->getWidth(), destRenderTarget->getHeight(), 1); + + // Use nearest filtering because source and destination are the same size for the direct + // copy + error = mBlit->copyTexture(source, sourceArea, sourceSize, dest, destArea, destSize, NULL, destFormat, GL_NEAREST); + if (error.isError()) + { + return error; + } + + storage11->invalidateSwizzleCacheLevel(level); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::copyImage3D(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level) +{ + gl::FramebufferAttachment *colorbuffer = framebuffer->getReadColorbuffer(); + ASSERT(colorbuffer); + + RenderTarget11 *sourceRenderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &sourceRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(sourceRenderTarget); + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + ASSERT(source); + + TextureStorage11_3D *storage11 = TextureStorage11_3D::makeTextureStorage11_3D(storage); + ASSERT(storage11); + + gl::ImageIndex index = gl::ImageIndex::Make3D(level, destOffset.z); + RenderTargetD3D *destRenderTarget = NULL; + error = storage11->getRenderTarget(index, &destRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(destRenderTarget); + + ID3D11RenderTargetView *dest = RenderTarget11::makeRenderTarget11(destRenderTarget)->getRenderTargetView(); + ASSERT(dest); + + gl::Box sourceArea(sourceRect.x, sourceRect.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents sourceSize(sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), 1); + + gl::Box destArea(destOffset.x, destOffset.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents destSize(destRenderTarget->getWidth(), destRenderTarget->getHeight(), 1); + + // Use nearest filtering because source and destination are the same size for the direct + // copy + error = mBlit->copyTexture(source, sourceArea, sourceSize, dest, destArea, destSize, NULL, destFormat, GL_NEAREST); + if (error.isError()) + { + return error; + } + + storage11->invalidateSwizzleCacheLevel(level); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::copyImage2DArray(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level) +{ + gl::FramebufferAttachment *colorbuffer = framebuffer->getReadColorbuffer(); + ASSERT(colorbuffer); + + RenderTarget11 *sourceRenderTarget = NULL; + gl::Error error = d3d11::GetAttachmentRenderTarget(colorbuffer, &sourceRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(sourceRenderTarget); + + ID3D11ShaderResourceView *source = sourceRenderTarget->getShaderResourceView(); + ASSERT(source); + + TextureStorage11_2DArray *storage11 = TextureStorage11_2DArray::makeTextureStorage11_2DArray(storage); + ASSERT(storage11); + + gl::ImageIndex index = gl::ImageIndex::Make2DArray(level, destOffset.z); + RenderTargetD3D *destRenderTarget = NULL; + error = storage11->getRenderTarget(index, &destRenderTarget); + if (error.isError()) + { + return error; + } + ASSERT(destRenderTarget); + + ID3D11RenderTargetView *dest = RenderTarget11::makeRenderTarget11(destRenderTarget)->getRenderTargetView(); + ASSERT(dest); + + gl::Box sourceArea(sourceRect.x, sourceRect.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents sourceSize(sourceRenderTarget->getWidth(), sourceRenderTarget->getHeight(), 1); + + gl::Box destArea(destOffset.x, destOffset.y, 0, sourceRect.width, sourceRect.height, 1); + gl::Extents destSize(destRenderTarget->getWidth(), destRenderTarget->getHeight(), 1); + + // Use nearest filtering because source and destination are the same size for the direct + // copy + error = mBlit->copyTexture(source, sourceArea, sourceSize, dest, destArea, destSize, NULL, destFormat, GL_NEAREST); + if (error.isError()) + { + return error; + } + + storage11->invalidateSwizzleCacheLevel(level); + + return gl::Error(GL_NO_ERROR); +} + +void Renderer11::unapplyRenderTargets() +{ + setOneTimeRenderTarget(NULL); +} + +// When finished with this rendertarget, markAllStateDirty must be called. +void Renderer11::setOneTimeRenderTarget(ID3D11RenderTargetView *renderTargetView) +{ + ID3D11RenderTargetView *rtvArray[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS] = {NULL}; + + rtvArray[0] = renderTargetView; + + mDeviceContext->OMSetRenderTargets(getRendererCaps().maxDrawBuffers, rtvArray, NULL); + + // Do not preserve the serial for this one-time-use render target + for (size_t rtIndex = 0; rtIndex < ArraySize(mAppliedRTVs); rtIndex++) + { + mAppliedRTVs[rtIndex] = DirtyPointer; + } + mAppliedDSV = DirtyPointer; +} + +gl::Error Renderer11::createRenderTarget(int width, int height, GLenum format, GLsizei samples, RenderTargetD3D **outRT) +{ + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(format, mFeatureLevel); + + const gl::TextureCaps &textureCaps = getRendererTextureCaps().get(format); + GLuint supportedSamples = textureCaps.getNearestSamples(samples); + + if (width > 0 && height > 0) + { + // Create texture resource + D3D11_TEXTURE2D_DESC desc; + desc.Width = width; + desc.Height = height; + desc.MipLevels = 1; + desc.ArraySize = 1; + desc.Format = formatInfo.texFormat; + desc.SampleDesc.Count = (supportedSamples == 0) ? 1 : supportedSamples; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + // If a rendertarget or depthstencil format exists for this texture format, + // we'll flag it to allow binding that way. Shader resource views are a little + // more complicated. + bool bindRTV = false, bindDSV = false, bindSRV = false; + bindRTV = (formatInfo.rtvFormat != DXGI_FORMAT_UNKNOWN); + bindDSV = (formatInfo.dsvFormat != DXGI_FORMAT_UNKNOWN); + if (formatInfo.srvFormat != DXGI_FORMAT_UNKNOWN) + { + // Multisample targets flagged for binding as depth stencil cannot also be + // flagged for binding as SRV, so make certain not to add the SRV flag for + // these targets. + bindSRV = !(formatInfo.dsvFormat != DXGI_FORMAT_UNKNOWN && desc.SampleDesc.Count > 1); + } + + desc.BindFlags = (bindRTV ? D3D11_BIND_RENDER_TARGET : 0) | + (bindDSV ? D3D11_BIND_DEPTH_STENCIL : 0) | + (bindSRV ? D3D11_BIND_SHADER_RESOURCE : 0); + + // The format must be either an RTV or a DSV + ASSERT(bindRTV != bindDSV); + + ID3D11Texture2D *texture = NULL; + HRESULT result = mDevice->CreateTexture2D(&desc, NULL, &texture); + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create render target texture, result: 0x%X.", result); + } + + ID3D11ShaderResourceView *srv = NULL; + if (bindSRV) + { + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = formatInfo.srvFormat; + srvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_SRV_DIMENSION_TEXTURE2D : D3D11_SRV_DIMENSION_TEXTURE2DMS; + srvDesc.Texture2D.MostDetailedMip = 0; + srvDesc.Texture2D.MipLevels = 1; + + result = mDevice->CreateShaderResourceView(texture, &srvDesc, &srv); + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + SafeRelease(texture); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create render target shader resource view, result: 0x%X.", result); + } + } + + if (bindDSV) + { + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = formatInfo.dsvFormat; + dsvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_DSV_DIMENSION_TEXTURE2D : D3D11_DSV_DIMENSION_TEXTURE2DMS; + dsvDesc.Texture2D.MipSlice = 0; + dsvDesc.Flags = 0; + + ID3D11DepthStencilView *dsv = NULL; + result = mDevice->CreateDepthStencilView(texture, &dsvDesc, &dsv); + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + SafeRelease(texture); + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create render target depth stencil view, result: 0x%X.", result); + } + + *outRT = new TextureRenderTarget11(dsv, texture, srv, format, width, height, 1, supportedSamples); + + SafeRelease(dsv); + } + else if (bindRTV) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = formatInfo.rtvFormat; + rtvDesc.ViewDimension = (supportedSamples == 0) ? D3D11_RTV_DIMENSION_TEXTURE2D : D3D11_RTV_DIMENSION_TEXTURE2DMS; + rtvDesc.Texture2D.MipSlice = 0; + + ID3D11RenderTargetView *rtv = NULL; + result = mDevice->CreateRenderTargetView(texture, &rtvDesc, &rtv); + if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + SafeRelease(texture); + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create render target render target view, result: 0x%X.", result); + } + + if (formatInfo.dataInitializerFunction != NULL) + { + const float clearValues[4] = { 0.0f, 0.0f, 0.0f, 1.0f }; + mDeviceContext->ClearRenderTargetView(rtv, clearValues); + } + + *outRT = new TextureRenderTarget11(rtv, texture, srv, format, width, height, 1, supportedSamples); + + SafeRelease(rtv); + } + else + { + UNREACHABLE(); + } + + SafeRelease(texture); + SafeRelease(srv); + } + else + { + *outRT = new TextureRenderTarget11(reinterpret_cast<ID3D11RenderTargetView*>(NULL), NULL, NULL, format, width, height, 1, supportedSamples); + } + + return gl::Error(GL_NO_ERROR); +} + +FramebufferImpl *Renderer11::createDefaultFramebuffer(const gl::Framebuffer::Data &data) +{ + return createFramebuffer(data); +} + +FramebufferImpl *Renderer11::createFramebuffer(const gl::Framebuffer::Data &data) +{ + return new Framebuffer11(data, this); +} + +CompilerImpl *Renderer11::createCompiler(const gl::Data &data) +{ + return new CompilerD3D(data, SH_HLSL11_OUTPUT); +} + +ShaderImpl *Renderer11::createShader(GLenum type) +{ + return new ShaderD3D(type); +} + +ProgramImpl *Renderer11::createProgram() +{ + return new ProgramD3D(this); +} + +gl::Error Renderer11::loadExecutable(const void *function, size_t length, ShaderType type, + const std::vector<gl::LinkedVarying> &transformFeedbackVaryings, + bool separatedOutputBuffers, ShaderExecutableD3D **outExecutable) +{ + switch (type) + { + case SHADER_VERTEX: + { + ID3D11VertexShader *vertexShader = NULL; + ID3D11GeometryShader *streamOutShader = NULL; + + HRESULT result = mDevice->CreateVertexShader(function, length, NULL, &vertexShader); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create vertex shader, result: 0x%X.", result); + } + + if (transformFeedbackVaryings.size() > 0) + { + std::vector<D3D11_SO_DECLARATION_ENTRY> soDeclaration; + for (size_t i = 0; i < transformFeedbackVaryings.size(); i++) + { + const gl::LinkedVarying &varying = transformFeedbackVaryings[i]; + GLenum transposedType = gl::TransposeMatrixType(varying.type); + + for (size_t j = 0; j < varying.semanticIndexCount; j++) + { + D3D11_SO_DECLARATION_ENTRY entry = { 0 }; + entry.Stream = 0; + entry.SemanticName = varying.semanticName.c_str(); + entry.SemanticIndex = varying.semanticIndex + j; + entry.StartComponent = 0; + entry.ComponentCount = gl::VariableColumnCount(transposedType); + entry.OutputSlot = (separatedOutputBuffers ? i : 0); + soDeclaration.push_back(entry); + } + } + + result = mDevice->CreateGeometryShaderWithStreamOutput(function, length, soDeclaration.data(), soDeclaration.size(), + NULL, 0, 0, NULL, &streamOutShader); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create steam output shader, result: 0x%X.", result); + } + } + + *outExecutable = new ShaderExecutable11(function, length, vertexShader, streamOutShader); + } + break; + case SHADER_PIXEL: + { + ID3D11PixelShader *pixelShader = NULL; + + HRESULT result = mDevice->CreatePixelShader(function, length, NULL, &pixelShader); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create pixel shader, result: 0x%X.", result); + } + + *outExecutable = new ShaderExecutable11(function, length, pixelShader); + } + break; + case SHADER_GEOMETRY: + { + ID3D11GeometryShader *geometryShader = NULL; + + HRESULT result = mDevice->CreateGeometryShader(function, length, NULL, &geometryShader); + ASSERT(SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create geometry shader, result: 0x%X.", result); + } + + *outExecutable = new ShaderExecutable11(function, length, geometryShader); + } + break; + default: + UNREACHABLE(); + return gl::Error(GL_INVALID_OPERATION); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::compileToExecutable(gl::InfoLog &infoLog, const std::string &shaderHLSL, ShaderType type, + const std::vector<gl::LinkedVarying> &transformFeedbackVaryings, + bool separatedOutputBuffers, const D3DCompilerWorkarounds &workarounds, + ShaderExecutableD3D **outExectuable) +{ + const char *profileType = NULL; + switch (type) + { + case SHADER_VERTEX: + profileType = "vs"; + break; + case SHADER_PIXEL: + profileType = "ps"; + break; + case SHADER_GEOMETRY: + profileType = "gs"; + break; + default: + UNREACHABLE(); + return gl::Error(GL_INVALID_OPERATION); + } + + std::string profile = FormatString("%s_%d_%d%s", profileType, getMajorShaderModel(), getMinorShaderModel(), getShaderModelSuffix().c_str()); + + UINT flags = D3DCOMPILE_OPTIMIZATION_LEVEL2; + + if (gl::DebugAnnotationsActive()) + { +#ifndef NDEBUG + flags = D3DCOMPILE_SKIP_OPTIMIZATION; +#endif + + flags |= D3DCOMPILE_DEBUG; + } + + if (workarounds.enableIEEEStrictness) + flags |= D3DCOMPILE_IEEE_STRICTNESS; + + // Sometimes D3DCompile will fail with the default compilation flags for complicated shaders when it would otherwise pass with alternative options. + // Try the default flags first and if compilation fails, try some alternatives. + std::vector<CompileConfig> configs; + configs.push_back(CompileConfig(flags, "default" )); + configs.push_back(CompileConfig(flags | D3DCOMPILE_SKIP_VALIDATION, "skip validation" )); + configs.push_back(CompileConfig(flags | D3DCOMPILE_SKIP_OPTIMIZATION, "skip optimization")); + + D3D_SHADER_MACRO loopMacros[] = { {"ANGLE_ENABLE_LOOP_FLATTEN", "1"}, {0, 0} }; + + ID3DBlob *binary = NULL; + std::string debugInfo; + gl::Error error = mCompiler.compileToBinary(infoLog, shaderHLSL, profile, configs, loopMacros, &binary, &debugInfo); + if (error.isError()) + { + return error; + } + + // It's possible that binary is NULL if the compiler failed in all configurations. Set the executable to NULL + // and return GL_NO_ERROR to signify that there was a link error but the internal state is still OK. + if (!binary) + { + *outExectuable = NULL; + return gl::Error(GL_NO_ERROR); + } + + error = loadExecutable(binary->GetBufferPointer(), binary->GetBufferSize(), type, + transformFeedbackVaryings, separatedOutputBuffers, outExectuable); + + SafeRelease(binary); + if (error.isError()) + { + return error; + } + + if (!debugInfo.empty()) + { + (*outExectuable)->appendDebugInfo(debugInfo); + } + + return gl::Error(GL_NO_ERROR); +} + +UniformStorageD3D *Renderer11::createUniformStorage(size_t storageSize) +{ + return new UniformStorage11(this, storageSize); +} + +VertexBuffer *Renderer11::createVertexBuffer() +{ + return new VertexBuffer11(this); +} + +IndexBuffer *Renderer11::createIndexBuffer() +{ + return new IndexBuffer11(this); +} + +BufferImpl *Renderer11::createBuffer() +{ + return new Buffer11(this); +} + +VertexArrayImpl *Renderer11::createVertexArray() +{ + return new VertexArray11(this); +} + +QueryImpl *Renderer11::createQuery(GLenum type) +{ + return new Query11(this, type); +} + +FenceNVImpl *Renderer11::createFenceNV() +{ + return new FenceNV11(this); +} + +FenceSyncImpl *Renderer11::createFenceSync() +{ + return new FenceSync11(this); +} + +TransformFeedbackImpl* Renderer11::createTransformFeedback() +{ + return new TransformFeedbackD3D(); +} + +bool Renderer11::supportsFastCopyBufferToTexture(GLenum internalFormat) const +{ + ASSERT(getRendererExtensions().pixelBufferObject); + + const gl::InternalFormat &internalFormatInfo = gl::GetInternalFormatInfo(internalFormat); + const d3d11::TextureFormat &d3d11FormatInfo = d3d11::GetTextureFormatInfo(internalFormat, mFeatureLevel); + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(d3d11FormatInfo.texFormat); + + // sRGB formats do not work with D3D11 buffer SRVs + if (internalFormatInfo.colorEncoding == GL_SRGB) + { + return false; + } + + // We cannot support direct copies to non-color-renderable formats + if (d3d11FormatInfo.rtvFormat == DXGI_FORMAT_UNKNOWN) + { + return false; + } + + // We skip all 3-channel formats since sometimes format support is missing + if (internalFormatInfo.componentCount == 3) + { + return false; + } + + // We don't support formats which we can't represent without conversion + if (dxgiFormatInfo.internalFormat != internalFormat) + { + return false; + } + + return true; +} + +gl::Error Renderer11::fastCopyBufferToTexture(const gl::PixelUnpackState &unpack, unsigned int offset, RenderTargetD3D *destRenderTarget, + GLenum destinationFormat, GLenum sourcePixelsType, const gl::Box &destArea) +{ + ASSERT(supportsFastCopyBufferToTexture(destinationFormat)); + return mPixelTransfer->copyBufferToTexture(unpack, offset, destRenderTarget, destinationFormat, sourcePixelsType, destArea); +} + +ImageD3D *Renderer11::createImage() +{ + return new Image11(this); +} + +gl::Error Renderer11::generateMipmap(ImageD3D *dest, ImageD3D *src) +{ + Image11 *dest11 = Image11::makeImage11(dest); + Image11 *src11 = Image11::makeImage11(src); + return Image11::generateMipmap(dest11, src11); +} + +TextureStorage *Renderer11::createTextureStorage2D(SwapChainD3D *swapChain) +{ + SwapChain11 *swapChain11 = SwapChain11::makeSwapChain11(swapChain); + return new TextureStorage11_2D(this, swapChain11); +} + +TextureStorage *Renderer11::createTextureStorage2D(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, int levels, bool hintLevelZeroOnly) +{ + return new TextureStorage11_2D(this, internalformat, renderTarget, width, height, levels, hintLevelZeroOnly); +} + +TextureStorage *Renderer11::createTextureStorageCube(GLenum internalformat, bool renderTarget, int size, int levels, bool hintLevelZeroOnly) +{ + return new TextureStorage11_Cube(this, internalformat, renderTarget, size, levels, hintLevelZeroOnly); +} + +TextureStorage *Renderer11::createTextureStorage3D(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, GLsizei depth, int levels) +{ + return new TextureStorage11_3D(this, internalformat, renderTarget, width, height, depth, levels); +} + +TextureStorage *Renderer11::createTextureStorage2DArray(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, GLsizei depth, int levels) +{ + return new TextureStorage11_2DArray(this, internalformat, renderTarget, width, height, depth, levels); +} + +TextureImpl *Renderer11::createTexture(GLenum target) +{ + switch(target) + { + case GL_TEXTURE_2D: return new TextureD3D_2D(this); + case GL_TEXTURE_CUBE_MAP: return new TextureD3D_Cube(this); + case GL_TEXTURE_3D: return new TextureD3D_3D(this); + case GL_TEXTURE_2D_ARRAY: return new TextureD3D_2DArray(this); + default: + UNREACHABLE(); + } + + return NULL; +} + +RenderbufferImpl *Renderer11::createRenderbuffer() +{ + RenderbufferD3D *renderbuffer = new RenderbufferD3D(this); + return renderbuffer; +} + +gl::Error Renderer11::readTextureData(ID3D11Texture2D *texture, unsigned int subResource, const gl::Rectangle &area, GLenum format, + GLenum type, GLuint outputPitch, const gl::PixelPackState &pack, uint8_t *pixels) +{ + ASSERT(area.width >= 0); + ASSERT(area.height >= 0); + + D3D11_TEXTURE2D_DESC textureDesc; + texture->GetDesc(&textureDesc); + + // Clamp read region to the defined texture boundaries, preventing out of bounds reads + // and reads of uninitialized data. + gl::Rectangle safeArea; + safeArea.x = gl::clamp(area.x, 0, static_cast<int>(textureDesc.Width)); + safeArea.y = gl::clamp(area.y, 0, static_cast<int>(textureDesc.Height)); + safeArea.width = gl::clamp(area.width + std::min(area.x, 0), 0, + static_cast<int>(textureDesc.Width) - safeArea.x); + safeArea.height = gl::clamp(area.height + std::min(area.y, 0), 0, + static_cast<int>(textureDesc.Height) - safeArea.y); + + ASSERT(safeArea.x >= 0 && safeArea.y >= 0); + ASSERT(safeArea.x + safeArea.width <= static_cast<int>(textureDesc.Width)); + ASSERT(safeArea.y + safeArea.height <= static_cast<int>(textureDesc.Height)); + + if (safeArea.width == 0 || safeArea.height == 0) + { + // no work to do + return gl::Error(GL_NO_ERROR); + } + + D3D11_TEXTURE2D_DESC stagingDesc; + stagingDesc.Width = safeArea.width; + stagingDesc.Height = safeArea.height; + stagingDesc.MipLevels = 1; + stagingDesc.ArraySize = 1; + stagingDesc.Format = textureDesc.Format; + stagingDesc.SampleDesc.Count = 1; + stagingDesc.SampleDesc.Quality = 0; + stagingDesc.Usage = D3D11_USAGE_STAGING; + stagingDesc.BindFlags = 0; + stagingDesc.CPUAccessFlags = D3D11_CPU_ACCESS_READ; + stagingDesc.MiscFlags = 0; + + ID3D11Texture2D* stagingTex = NULL; + HRESULT result = mDevice->CreateTexture2D(&stagingDesc, NULL, &stagingTex); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal staging texture for ReadPixels, HRESULT: 0x%X.", result); + } + + ID3D11Texture2D* srcTex = NULL; + if (textureDesc.SampleDesc.Count > 1) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = textureDesc.Width; + resolveDesc.Height = textureDesc.Height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = textureDesc.Format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = D3D11_USAGE_DEFAULT; + resolveDesc.BindFlags = 0; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + result = mDevice->CreateTexture2D(&resolveDesc, NULL, &srcTex); + if (FAILED(result)) + { + SafeRelease(stagingTex); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal resolve texture for ReadPixels, HRESULT: 0x%X.", result); + } + + mDeviceContext->ResolveSubresource(srcTex, 0, texture, subResource, textureDesc.Format); + subResource = 0; + } + else + { + srcTex = texture; + srcTex->AddRef(); + } + + D3D11_BOX srcBox; + srcBox.left = static_cast<UINT>(safeArea.x); + srcBox.right = static_cast<UINT>(safeArea.x + safeArea.width); + srcBox.top = static_cast<UINT>(safeArea.y); + srcBox.bottom = static_cast<UINT>(safeArea.y + safeArea.height); + srcBox.front = 0; + srcBox.back = 1; + + mDeviceContext->CopySubresourceRegion(stagingTex, 0, 0, 0, 0, srcTex, subResource, &srcBox); + + SafeRelease(srcTex); + + PackPixelsParams packParams(safeArea, format, type, outputPitch, pack, 0); + gl::Error error = packPixels(stagingTex, packParams, pixels); + + SafeRelease(stagingTex); + + return error; +} + +gl::Error Renderer11::packPixels(ID3D11Texture2D *readTexture, const PackPixelsParams ¶ms, uint8_t *pixelsOut) +{ + D3D11_TEXTURE2D_DESC textureDesc; + readTexture->GetDesc(&textureDesc); + + D3D11_MAPPED_SUBRESOURCE mapping; + HRESULT hr = mDeviceContext->Map(readTexture, 0, D3D11_MAP_READ, 0, &mapping); + if (FAILED(hr)) + { + ASSERT(hr == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal texture for reading, result: 0x%X.", hr); + } + + uint8_t *source; + int inputPitch; + if (params.pack.reverseRowOrder) + { + source = static_cast<uint8_t*>(mapping.pData) + mapping.RowPitch * (params.area.height - 1); + inputPitch = -static_cast<int>(mapping.RowPitch); + } + else + { + source = static_cast<uint8_t*>(mapping.pData); + inputPitch = static_cast<int>(mapping.RowPitch); + } + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(textureDesc.Format); + const gl::InternalFormat &sourceFormatInfo = gl::GetInternalFormatInfo(dxgiFormatInfo.internalFormat); + if (sourceFormatInfo.format == params.format && sourceFormatInfo.type == params.type) + { + uint8_t *dest = pixelsOut + params.offset; + for (int y = 0; y < params.area.height; y++) + { + memcpy(dest + y * params.outputPitch, source + y * inputPitch, params.area.width * sourceFormatInfo.pixelBytes); + } + } + else + { + const d3d11::DXGIFormat &sourceDXGIFormatInfo = d3d11::GetDXGIFormatInfo(textureDesc.Format); + ColorCopyFunction fastCopyFunc = sourceDXGIFormatInfo.getFastCopyFunction(params.format, params.type); + + GLenum sizedDestInternalFormat = gl::GetSizedInternalFormat(params.format, params.type); + const gl::InternalFormat &destFormatInfo = gl::GetInternalFormatInfo(sizedDestInternalFormat); + + if (fastCopyFunc) + { + // Fast copy is possible through some special function + for (int y = 0; y < params.area.height; y++) + { + for (int x = 0; x < params.area.width; x++) + { + uint8_t *dest = pixelsOut + params.offset + y * params.outputPitch + x * destFormatInfo.pixelBytes; + const uint8_t *src = source + y * inputPitch + x * sourceFormatInfo.pixelBytes; + + fastCopyFunc(src, dest); + } + } + } + else + { + ColorReadFunction colorReadFunction = sourceDXGIFormatInfo.colorReadFunction; + ColorWriteFunction colorWriteFunction = GetColorWriteFunction(params.format, params.type); + + uint8_t temp[16]; // Maximum size of any Color<T> type used. + static_assert(sizeof(temp) >= sizeof(gl::ColorF) && + sizeof(temp) >= sizeof(gl::ColorUI) && + sizeof(temp) >= sizeof(gl::ColorI), + "Unexpected size of gl::Color struct."); + + for (int y = 0; y < params.area.height; y++) + { + for (int x = 0; x < params.area.width; x++) + { + uint8_t *dest = pixelsOut + params.offset + y * params.outputPitch + x * destFormatInfo.pixelBytes; + const uint8_t *src = source + y * inputPitch + x * sourceFormatInfo.pixelBytes; + + // readFunc and writeFunc will be using the same type of color, CopyTexImage + // will not allow the copy otherwise. + colorReadFunction(src, temp); + colorWriteFunction(temp, dest); + } + } + } + } + + mDeviceContext->Unmap(readTexture, 0); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error Renderer11::blitRenderbufferRect(const gl::Rectangle &readRect, const gl::Rectangle &drawRect, RenderTargetD3D *readRenderTarget, + RenderTargetD3D *drawRenderTarget, GLenum filter, const gl::Rectangle *scissor, + bool colorBlit, bool depthBlit, bool stencilBlit) +{ + // Since blitRenderbufferRect is called for each render buffer that needs to be blitted, + // it should never be the case that both color and depth/stencil need to be blitted at + // at the same time. + ASSERT(colorBlit != (depthBlit || stencilBlit)); + + RenderTarget11 *drawRenderTarget11 = RenderTarget11::makeRenderTarget11(drawRenderTarget); + if (!drawRenderTarget) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to retrieve the internal draw render target from the draw framebuffer."); + } + + ID3D11Resource *drawTexture = drawRenderTarget11->getTexture(); + unsigned int drawSubresource = drawRenderTarget11->getSubresourceIndex(); + ID3D11RenderTargetView *drawRTV = drawRenderTarget11->getRenderTargetView(); + ID3D11DepthStencilView *drawDSV = drawRenderTarget11->getDepthStencilView(); + + RenderTarget11 *readRenderTarget11 = RenderTarget11::makeRenderTarget11(readRenderTarget); + if (!readRenderTarget) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to retrieve the internal read render target from the read framebuffer."); + } + + ID3D11Resource *readTexture = NULL; + ID3D11ShaderResourceView *readSRV = NULL; + unsigned int readSubresource = 0; + if (readRenderTarget->getSamples() > 0) + { + ID3D11Resource *unresolvedResource = readRenderTarget11->getTexture(); + ID3D11Texture2D *unresolvedTexture = d3d11::DynamicCastComObject<ID3D11Texture2D>(unresolvedResource); + + if (unresolvedTexture) + { + readTexture = resolveMultisampledTexture(unresolvedTexture, readRenderTarget11->getSubresourceIndex()); + readSubresource = 0; + + SafeRelease(unresolvedTexture); + + HRESULT hresult = mDevice->CreateShaderResourceView(readTexture, NULL, &readSRV); + if (FAILED(hresult)) + { + SafeRelease(readTexture); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create shader resource view to resolve multisampled framebuffer."); + } + } + } + else + { + readTexture = readRenderTarget11->getTexture(); + readTexture->AddRef(); + readSubresource = readRenderTarget11->getSubresourceIndex(); + readSRV = readRenderTarget11->getShaderResourceView(); + readSRV->AddRef(); + } + + if (!readTexture || !readSRV) + { + SafeRelease(readTexture); + SafeRelease(readSRV); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to retrieve the internal read render target view from the read render target."); + } + + gl::Extents readSize(readRenderTarget->getWidth(), readRenderTarget->getHeight(), 1); + gl::Extents drawSize(drawRenderTarget->getWidth(), drawRenderTarget->getHeight(), 1); + + bool scissorNeeded = scissor && gl::ClipRectangle(drawRect, *scissor, NULL); + + bool wholeBufferCopy = !scissorNeeded && + readRect.x == 0 && readRect.width == readSize.width && + readRect.y == 0 && readRect.height == readSize.height && + drawRect.x == 0 && drawRect.width == drawSize.width && + drawRect.y == 0 && drawRect.height == drawSize.height; + + bool stretchRequired = readRect.width != drawRect.width || readRect.height != drawRect.height; + + bool flipRequired = readRect.width < 0 || readRect.height < 0 || drawRect.width < 0 || drawRect.height < 0; + + bool outOfBounds = readRect.x < 0 || readRect.x + readRect.width > readSize.width || + readRect.y < 0 || readRect.y + readRect.height > readSize.height || + drawRect.x < 0 || drawRect.x + drawRect.width > drawSize.width || + drawRect.y < 0 || drawRect.y + drawRect.height > drawSize.height; + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(drawRenderTarget11->getDXGIFormat()); + bool partialDSBlit = (dxgiFormatInfo.depthBits > 0 && depthBlit) != (dxgiFormatInfo.stencilBits > 0 && stencilBlit); + + gl::Error result(GL_NO_ERROR); + + if (readRenderTarget11->getDXGIFormat() == drawRenderTarget11->getDXGIFormat() && + !stretchRequired && !outOfBounds && !flipRequired && !partialDSBlit && + (!(depthBlit || stencilBlit) || wholeBufferCopy)) + { + UINT dstX = drawRect.x; + UINT dstY = drawRect.y; + + D3D11_BOX readBox; + readBox.left = readRect.x; + readBox.right = readRect.x + readRect.width; + readBox.top = readRect.y; + readBox.bottom = readRect.y + readRect.height; + readBox.front = 0; + readBox.back = 1; + + if (scissorNeeded) + { + // drawRect is guaranteed to have positive width and height because stretchRequired is false. + ASSERT(drawRect.width >= 0 || drawRect.height >= 0); + + if (drawRect.x < scissor->x) + { + dstX = scissor->x; + readBox.left += (scissor->x - drawRect.x); + } + if (drawRect.y < scissor->y) + { + dstY = scissor->y; + readBox.top += (scissor->y - drawRect.y); + } + if (drawRect.x + drawRect.width > scissor->x + scissor->width) + { + readBox.right -= ((drawRect.x + drawRect.width) - (scissor->x + scissor->width)); + } + if (drawRect.y + drawRect.height > scissor->y + scissor->height) + { + readBox.bottom -= ((drawRect.y + drawRect.height) - (scissor->y + scissor->height)); + } + } + + // D3D11 needs depth-stencil CopySubresourceRegions to have a NULL pSrcBox + // We also require complete framebuffer copies for depth-stencil blit. + D3D11_BOX *pSrcBox = wholeBufferCopy ? NULL : &readBox; + + mDeviceContext->CopySubresourceRegion(drawTexture, drawSubresource, dstX, dstY, 0, + readTexture, readSubresource, pSrcBox); + result = gl::Error(GL_NO_ERROR); + } + else + { + gl::Box readArea(readRect.x, readRect.y, 0, readRect.width, readRect.height, 1); + gl::Box drawArea(drawRect.x, drawRect.y, 0, drawRect.width, drawRect.height, 1); + + if (depthBlit && stencilBlit) + { + result = mBlit->copyDepthStencil(readTexture, readSubresource, readArea, readSize, + drawTexture, drawSubresource, drawArea, drawSize, + scissor); + } + else if (depthBlit) + { + result = mBlit->copyDepth(readSRV, readArea, readSize, drawDSV, drawArea, drawSize, + scissor); + } + else if (stencilBlit) + { + result = mBlit->copyStencil(readTexture, readSubresource, readArea, readSize, + drawTexture, drawSubresource, drawArea, drawSize, + scissor); + } + else + { + GLenum format = gl::GetInternalFormatInfo(drawRenderTarget->getInternalFormat()).format; + result = mBlit->copyTexture(readSRV, readArea, readSize, drawRTV, drawArea, drawSize, + scissor, format, filter); + } + } + + SafeRelease(readTexture); + SafeRelease(readSRV); + + return result; +} + +bool Renderer11::isES3Capable() const +{ + return (d3d11_gl::GetMaximumClientVersion(mFeatureLevel) > 2); +}; + +ID3D11Texture2D *Renderer11::resolveMultisampledTexture(ID3D11Texture2D *source, unsigned int subresource) +{ + D3D11_TEXTURE2D_DESC textureDesc; + source->GetDesc(&textureDesc); + + if (textureDesc.SampleDesc.Count > 1) + { + D3D11_TEXTURE2D_DESC resolveDesc; + resolveDesc.Width = textureDesc.Width; + resolveDesc.Height = textureDesc.Height; + resolveDesc.MipLevels = 1; + resolveDesc.ArraySize = 1; + resolveDesc.Format = textureDesc.Format; + resolveDesc.SampleDesc.Count = 1; + resolveDesc.SampleDesc.Quality = 0; + resolveDesc.Usage = textureDesc.Usage; + resolveDesc.BindFlags = textureDesc.BindFlags; + resolveDesc.CPUAccessFlags = 0; + resolveDesc.MiscFlags = 0; + + ID3D11Texture2D *resolveTexture = NULL; + HRESULT result = mDevice->CreateTexture2D(&resolveDesc, NULL, &resolveTexture); + if (FAILED(result)) + { + ERR("Failed to create a multisample resolve texture, HRESULT: 0x%X.", result); + return NULL; + } + + mDeviceContext->ResolveSubresource(resolveTexture, 0, source, subresource, textureDesc.Format); + return resolveTexture; + } + else + { + source->AddRef(); + return source; + } +} + +bool Renderer11::getLUID(LUID *adapterLuid) const +{ + adapterLuid->HighPart = 0; + adapterLuid->LowPart = 0; + + if (!mDxgiAdapter) + { + return false; + } + + DXGI_ADAPTER_DESC adapterDesc; + if (FAILED(mDxgiAdapter->GetDesc(&adapterDesc))) + { + return false; + } + + *adapterLuid = adapterDesc.AdapterLuid; + return true; +} + +VertexConversionType Renderer11::getVertexConversionType(const gl::VertexFormat &vertexFormat) const +{ + return d3d11::GetVertexFormatInfo(vertexFormat, mFeatureLevel).conversionType; +} + +GLenum Renderer11::getVertexComponentType(const gl::VertexFormat &vertexFormat) const +{ + return d3d11::GetDXGIFormatInfo(d3d11::GetVertexFormatInfo(vertexFormat, mFeatureLevel).nativeFormat).componentType; +} + +void Renderer11::generateCaps(gl::Caps *outCaps, gl::TextureCapsMap *outTextureCaps, gl::Extensions *outExtensions) const +{ + d3d11_gl::GenerateCaps(mDevice, mDeviceContext, outCaps, outTextureCaps, outExtensions); +} + +Workarounds Renderer11::generateWorkarounds() const +{ + return d3d11::GenerateWorkarounds(mFeatureLevel); +} + +void Renderer11::setShaderResource(gl::SamplerType shaderType, UINT resourceSlot, ID3D11ShaderResourceView *srv) +{ + auto ¤tSRVs = (shaderType == gl::SAMPLER_VERTEX ? mCurVertexSRVs : mCurPixelSRVs); + + ASSERT(static_cast<size_t>(resourceSlot) < currentSRVs.size()); + auto &record = currentSRVs[resourceSlot]; + + if (record.srv != reinterpret_cast<uintptr_t>(srv)) + { + if (shaderType == gl::SAMPLER_VERTEX) + { + mDeviceContext->VSSetShaderResources(resourceSlot, 1, &srv); + } + else + { + mDeviceContext->PSSetShaderResources(resourceSlot, 1, &srv); + } + + record.srv = reinterpret_cast<uintptr_t>(srv); + if (srv) + { + record.resource = reinterpret_cast<uintptr_t>(GetViewResource(srv)); + srv->GetDesc(&record.desc); + } + else + { + record.resource = 0; + } + } +} +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.h new file mode 100644 index 0000000000..cc7d6c237b --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Renderer11.h @@ -0,0 +1,400 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Renderer11.h: Defines a back-end specific class for the D3D11 renderer. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_H_ + +#include "common/angleutils.h" +#include "common/mathutil.h" +#include "libANGLE/AttributeMap.h" +#include "libANGLE/angletypes.h" +#include "libANGLE/renderer/d3d/HLSLCompiler.h" +#include "libANGLE/renderer/d3d/RendererD3D.h" +#include "libANGLE/renderer/d3d/RenderTargetD3D.h" +#include "libANGLE/renderer/d3d/d3d11/DebugAnnotator11.h" +#include "libANGLE/renderer/d3d/d3d11/InputLayoutCache.h" +#include "libANGLE/renderer/d3d/d3d11/RenderStateCache.h" + +struct ID3D11DeviceContext1; + +namespace gl +{ +class FramebufferAttachment; +struct ImageIndex; +} + +namespace rx +{ + +class VertexDataManager; +class IndexDataManager; +class StreamingIndexBufferInterface; +class Blit11; +class Clear11; +class PixelTransfer11; +class RenderTarget11; +class Trim11; +struct PackPixelsParams; + +enum +{ + MAX_VERTEX_UNIFORM_VECTORS_D3D11 = 1024, + MAX_FRAGMENT_UNIFORM_VECTORS_D3D11 = 1024 +}; + +// Possible reasons RendererD3D initialize can fail +enum D3D11InitError +{ + // The renderer loaded successfully + D3D11_INIT_SUCCESS = 0, + // Failed to load the ANGLE & D3D compiler libraries + D3D11_INIT_COMPILER_ERROR, + // Failed to load a necessary DLL (non-compiler) + D3D11_INIT_MISSING_DEP, + // CreateDevice returned E_INVALIDARG + D3D11_INIT_CREATEDEVICE_INVALIDARG, + // CreateDevice failed with an error other than invalid arg + D3D11_INIT_CREATEDEVICE_ERROR, + // DXGI 1.2 required but not found + D3D11_INIT_INCOMPATIBLE_DXGI, + // Other initialization error + D3D11_INIT_OTHER_ERROR, + NUM_D3D11_INIT_ERRORS +}; + +class Renderer11 : public RendererD3D +{ + public: + explicit Renderer11(egl::Display *display); + virtual ~Renderer11(); + + static Renderer11 *makeRenderer11(Renderer *renderer); + + egl::Error initialize() override; + virtual bool resetDevice(); + + egl::ConfigSet generateConfigs() const override; + + gl::Error flush() override; + gl::Error finish() override; + + virtual SwapChainD3D *createSwapChain(NativeWindow nativeWindow, HANDLE shareHandle, GLenum backBufferFormat, GLenum depthBufferFormat); + + virtual gl::Error generateSwizzle(gl::Texture *texture); + virtual gl::Error setSamplerState(gl::SamplerType type, int index, gl::Texture *texture, const gl::SamplerState &sampler); + virtual gl::Error setTexture(gl::SamplerType type, int index, gl::Texture *texture); + + gl::Error setUniformBuffers(const gl::Data &data, + const GLint vertexUniformBuffers[], + const GLint fragmentUniformBuffers[]) override; + + virtual gl::Error setRasterizerState(const gl::RasterizerState &rasterState); + gl::Error setBlendState(const gl::Framebuffer *framebuffer, const gl::BlendState &blendState, const gl::ColorF &blendColor, + unsigned int sampleMask) override; + virtual gl::Error setDepthStencilState(const gl::DepthStencilState &depthStencilState, int stencilRef, + int stencilBackRef, bool frontFaceCCW); + + virtual void setScissorRectangle(const gl::Rectangle &scissor, bool enabled); + virtual void setViewport(const gl::Rectangle &viewport, float zNear, float zFar, GLenum drawMode, GLenum frontFace, + bool ignoreViewport); + + virtual bool applyPrimitiveType(GLenum mode, GLsizei count, bool usesPointSize); + gl::Error applyRenderTarget(const gl::Framebuffer *frameBuffer) override; + virtual gl::Error applyShaders(gl::Program *program, const gl::VertexFormat inputLayout[], const gl::Framebuffer *framebuffer, + bool rasterizerDiscard, bool transformFeedbackActive); + + virtual gl::Error applyUniforms(const ProgramImpl &program, const std::vector<gl::LinkedUniform*> &uniformArray); + virtual gl::Error applyVertexBuffer(const gl::State &state, GLenum mode, GLint first, GLsizei count, GLsizei instances); + virtual gl::Error applyIndexBuffer(const GLvoid *indices, gl::Buffer *elementArrayBuffer, GLsizei count, GLenum mode, GLenum type, TranslatedIndexData *indexInfo); + void applyTransformFeedbackBuffers(const gl::State &state) override; + + gl::Error drawArrays(const gl::Data &data, GLenum mode, GLsizei count, GLsizei instances, bool usesPointSize) override; + virtual gl::Error drawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices, + gl::Buffer *elementArrayBuffer, const TranslatedIndexData &indexInfo, GLsizei instances); + + virtual void markAllStateDirty(); + + // lost device + bool testDeviceLost() override; + bool testDeviceResettable() override; + + VendorID getVendorId() const override; + std::string getRendererDescription() const override; + GUID getAdapterIdentifier() const override; + + virtual unsigned int getReservedVertexUniformVectors() const; + virtual unsigned int getReservedFragmentUniformVectors() const; + virtual unsigned int getReservedVertexUniformBuffers() const; + virtual unsigned int getReservedFragmentUniformBuffers() const; + virtual bool getShareHandleSupport() const; + virtual bool getPostSubBufferSupport() const; + + virtual int getMajorShaderModel() const; + int getMinorShaderModel() const override; + std::string getShaderModelSuffix() const override; + + // Pixel operations + virtual gl::Error copyImage2D(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level); + virtual gl::Error copyImageCube(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLenum target, GLint level); + virtual gl::Error copyImage3D(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level); + virtual gl::Error copyImage2DArray(const gl::Framebuffer *framebuffer, const gl::Rectangle &sourceRect, GLenum destFormat, + const gl::Offset &destOffset, TextureStorage *storage, GLint level); + + // RenderTarget creation + virtual gl::Error createRenderTarget(int width, int height, GLenum format, GLsizei samples, RenderTargetD3D **outRT); + + // Framebuffer creation + FramebufferImpl *createDefaultFramebuffer(const gl::Framebuffer::Data &data) override; + FramebufferImpl *createFramebuffer(const gl::Framebuffer::Data &data) override; + + // Shader creation + virtual CompilerImpl *createCompiler(const gl::Data &data); + virtual ShaderImpl *createShader(GLenum type); + virtual ProgramImpl *createProgram(); + + // Shader operations + virtual gl::Error loadExecutable(const void *function, size_t length, ShaderType type, + const std::vector<gl::LinkedVarying> &transformFeedbackVaryings, + bool separatedOutputBuffers, ShaderExecutableD3D **outExecutable); + virtual gl::Error compileToExecutable(gl::InfoLog &infoLog, const std::string &shaderHLSL, ShaderType type, + const std::vector<gl::LinkedVarying> &transformFeedbackVaryings, + bool separatedOutputBuffers, const D3DCompilerWorkarounds &workarounds, + ShaderExecutableD3D **outExectuable); + virtual UniformStorageD3D *createUniformStorage(size_t storageSize); + + // Image operations + virtual ImageD3D *createImage(); + gl::Error generateMipmap(ImageD3D *dest, ImageD3D *source) override; + virtual TextureStorage *createTextureStorage2D(SwapChainD3D *swapChain); + virtual TextureStorage *createTextureStorage2D(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, int levels, bool hintLevelZeroOnly); + virtual TextureStorage *createTextureStorageCube(GLenum internalformat, bool renderTarget, int size, int levels, bool hintLevelZeroOnly); + virtual TextureStorage *createTextureStorage3D(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, GLsizei depth, int levels); + virtual TextureStorage *createTextureStorage2DArray(GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, GLsizei depth, int levels); + + // Texture creation + virtual TextureImpl *createTexture(GLenum target); + + // Renderbuffer creation + virtual RenderbufferImpl *createRenderbuffer(); + + // Buffer creation + virtual BufferImpl *createBuffer(); + virtual VertexBuffer *createVertexBuffer(); + virtual IndexBuffer *createIndexBuffer(); + + // Vertex Array creation + virtual VertexArrayImpl *createVertexArray(); + + // Query and Fence creation + virtual QueryImpl *createQuery(GLenum type); + virtual FenceNVImpl *createFenceNV(); + virtual FenceSyncImpl *createFenceSync(); + + // Transform Feedback creation + virtual TransformFeedbackImpl* createTransformFeedback(); + + // D3D11-renderer specific methods + ID3D11Device *getDevice() { return mDevice; } + ID3D11DeviceContext *getDeviceContext() { return mDeviceContext; }; + ID3D11DeviceContext1 *getDeviceContext1IfSupported() { return mDeviceContext1; }; + DXGIFactory *getDxgiFactory() { return mDxgiFactory; }; + + Blit11 *getBlitter() { return mBlit; } + Clear11 *getClearer() { return mClear; } + + // Buffer-to-texture and Texture-to-buffer copies + virtual bool supportsFastCopyBufferToTexture(GLenum internalFormat) const; + virtual gl::Error fastCopyBufferToTexture(const gl::PixelUnpackState &unpack, unsigned int offset, RenderTargetD3D *destRenderTarget, + GLenum destinationFormat, GLenum sourcePixelsType, const gl::Box &destArea); + + void unapplyRenderTargets(); + void setOneTimeRenderTarget(ID3D11RenderTargetView *renderTargetView); + gl::Error packPixels(ID3D11Texture2D *readTexture, const PackPixelsParams ¶ms, uint8_t *pixelsOut); + + bool getLUID(LUID *adapterLuid) const override; + virtual VertexConversionType getVertexConversionType(const gl::VertexFormat &vertexFormat) const; + virtual GLenum getVertexComponentType(const gl::VertexFormat &vertexFormat) const; + + gl::Error readTextureData(ID3D11Texture2D *texture, unsigned int subResource, const gl::Rectangle &area, GLenum format, + GLenum type, GLuint outputPitch, const gl::PixelPackState &pack, uint8_t *pixels); + + void setShaderResource(gl::SamplerType shaderType, UINT resourceSlot, ID3D11ShaderResourceView *srv); + + gl::Error blitRenderbufferRect(const gl::Rectangle &readRect, const gl::Rectangle &drawRect, RenderTargetD3D *readRenderTarget, + RenderTargetD3D *drawRenderTarget, GLenum filter, const gl::Rectangle *scissor, + bool colorBlit, bool depthBlit, bool stencilBlit); + + bool isES3Capable() const; + D3D_FEATURE_LEVEL getFeatureLevel() const { return mFeatureLevel; }; + + RendererClass getRendererClass() const override { return RENDERER_D3D11; } + + private: + void generateCaps(gl::Caps *outCaps, gl::TextureCapsMap *outTextureCaps, gl::Extensions *outExtensions) const override; + Workarounds generateWorkarounds() const override; + + gl::Error drawLineLoop(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer); + gl::Error drawTriangleFan(GLsizei count, GLenum type, const GLvoid *indices, int minIndex, gl::Buffer *elementArrayBuffer, int instances); + + ID3D11Texture2D *resolveMultisampledTexture(ID3D11Texture2D *source, unsigned int subresource); + void unsetConflictingSRVs(gl::SamplerType shaderType, uintptr_t resource, const gl::ImageIndex *index); + + HMODULE mD3d11Module; + HMODULE mDxgiModule; + std::vector<D3D_FEATURE_LEVEL> mAvailableFeatureLevels; + D3D_DRIVER_TYPE mDriverType; + + HLSLCompiler mCompiler; + + void initializeDevice(); + void releaseDeviceResources(); + void release(); + + RenderStateCache mStateCache; + + // current render target states + uintptr_t mAppliedRTVs[gl::IMPLEMENTATION_MAX_DRAW_BUFFERS]; + uintptr_t mAppliedDSV; + bool mDepthStencilInitialized; + bool mRenderTargetDescInitialized; + + struct RenderTargetDesc + { + size_t width; + size_t height; + DXGI_FORMAT format; + }; + RenderTargetDesc mRenderTargetDesc; + + // Currently applied sampler states + std::vector<bool> mForceSetVertexSamplerStates; + std::vector<gl::SamplerState> mCurVertexSamplerStates; + + std::vector<bool> mForceSetPixelSamplerStates; + std::vector<gl::SamplerState> mCurPixelSamplerStates; + + // Currently applied textures + struct SRVRecord + { + uintptr_t srv; + uintptr_t resource; + D3D11_SHADER_RESOURCE_VIEW_DESC desc; + }; + std::vector<SRVRecord> mCurVertexSRVs; + std::vector<SRVRecord> mCurPixelSRVs; + + // Currently applied blend state + bool mForceSetBlendState; + gl::BlendState mCurBlendState; + gl::ColorF mCurBlendColor; + unsigned int mCurSampleMask; + + // Currently applied rasterizer state + bool mForceSetRasterState; + gl::RasterizerState mCurRasterState; + + // Currently applied depth stencil state + bool mForceSetDepthStencilState; + gl::DepthStencilState mCurDepthStencilState; + int mCurStencilRef; + int mCurStencilBackRef; + + // Currently applied scissor rectangle + bool mForceSetScissor; + bool mScissorEnabled; + gl::Rectangle mCurScissor; + + // Currently applied viewport + bool mForceSetViewport; + gl::Rectangle mCurViewport; + float mCurNear; + float mCurFar; + + // Currently applied primitive topology + D3D11_PRIMITIVE_TOPOLOGY mCurrentPrimitiveTopology; + + // Currently applied index buffer + ID3D11Buffer *mAppliedIB; + DXGI_FORMAT mAppliedIBFormat; + unsigned int mAppliedIBOffset; + + // Currently applied transform feedback buffers + size_t mAppliedNumXFBBindings; + ID3D11Buffer *mAppliedTFBuffers[gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS]; // Tracks the current D3D buffers + // in use for streamout + GLintptr mAppliedTFOffsets[gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS]; // Tracks the current GL-specified + // buffer offsets to transform feedback + // buffers + UINT mCurrentD3DOffsets[gl::IMPLEMENTATION_MAX_TRANSFORM_FEEDBACK_BUFFERS]; // Tracks the D3D buffer offsets, + // which may differ from GLs, due + // to different append behavior + + // Currently applied shaders + uintptr_t mAppliedVertexShader; + uintptr_t mAppliedGeometryShader; + uintptr_t mAppliedPixelShader; + + dx_VertexConstants mVertexConstants; + dx_VertexConstants mAppliedVertexConstants; + ID3D11Buffer *mDriverConstantBufferVS; + ID3D11Buffer *mCurrentVertexConstantBuffer; + unsigned int mCurrentConstantBufferVS[gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS]; + GLintptr mCurrentConstantBufferVSOffset[gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS]; + GLsizeiptr mCurrentConstantBufferVSSize[gl::IMPLEMENTATION_MAX_VERTEX_SHADER_UNIFORM_BUFFERS]; + + dx_PixelConstants mPixelConstants; + dx_PixelConstants mAppliedPixelConstants; + ID3D11Buffer *mDriverConstantBufferPS; + ID3D11Buffer *mCurrentPixelConstantBuffer; + unsigned int mCurrentConstantBufferPS[gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS]; + GLintptr mCurrentConstantBufferPSOffset[gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS]; + GLsizeiptr mCurrentConstantBufferPSSize[gl::IMPLEMENTATION_MAX_FRAGMENT_SHADER_UNIFORM_BUFFERS]; + + ID3D11Buffer *mCurrentGeometryConstantBuffer; + + // Vertex, index and input layouts + VertexDataManager *mVertexDataManager; + IndexDataManager *mIndexDataManager; + InputLayoutCache mInputLayoutCache; + + StreamingIndexBufferInterface *mLineLoopIB; + StreamingIndexBufferInterface *mTriangleFanIB; + + // Texture copy resources + Blit11 *mBlit; + PixelTransfer11 *mPixelTransfer; + + // Masked clear resources + Clear11 *mClear; + + // Perform trim for D3D resources + Trim11 *mTrim; + + // Sync query + ID3D11Query *mSyncQuery; + + // Constant buffer offset support + bool mSupportsConstantBufferOffsets; + + ID3D11Device *mDevice; + D3D_FEATURE_LEVEL mFeatureLevel; + ID3D11DeviceContext *mDeviceContext; + ID3D11DeviceContext1 *mDeviceContext1; + IDXGIAdapter *mDxgiAdapter; + DXGI_ADAPTER_DESC mAdapterDescription; + char mDescription[128]; + DXGIFactory *mDxgiFactory; + + DebugAnnotator11 mAnnotator; +}; + +} +#endif // LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.cpp new file mode 100644 index 0000000000..7e64c3183d --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.cpp @@ -0,0 +1,110 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable11.cpp: Implements a D3D11-specific class to contain shader +// executable implementation details. + +#include "libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" + +namespace rx +{ + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11PixelShader *executable) + : ShaderExecutableD3D(function, length) +{ + mPixelExecutable = executable; + mVertexExecutable = NULL; + mGeometryExecutable = NULL; + mStreamOutExecutable = NULL; +} + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11VertexShader *executable, ID3D11GeometryShader *streamOut) + : ShaderExecutableD3D(function, length) +{ + mVertexExecutable = executable; + mPixelExecutable = NULL; + mGeometryExecutable = NULL; + mStreamOutExecutable = streamOut; +} + +ShaderExecutable11::ShaderExecutable11(const void *function, size_t length, ID3D11GeometryShader *executable) + : ShaderExecutableD3D(function, length) +{ + mGeometryExecutable = executable; + mVertexExecutable = NULL; + mPixelExecutable = NULL; + mStreamOutExecutable = NULL; +} + +ShaderExecutable11::~ShaderExecutable11() +{ + SafeRelease(mVertexExecutable); + SafeRelease(mPixelExecutable); + SafeRelease(mGeometryExecutable); + SafeRelease(mStreamOutExecutable); +} + +ShaderExecutable11 *ShaderExecutable11::makeShaderExecutable11(ShaderExecutableD3D *executable) +{ + ASSERT(HAS_DYNAMIC_TYPE(ShaderExecutable11*, executable)); + return static_cast<ShaderExecutable11*>(executable); +} + +ID3D11VertexShader *ShaderExecutable11::getVertexShader() const +{ + return mVertexExecutable; +} + +ID3D11PixelShader *ShaderExecutable11::getPixelShader() const +{ + return mPixelExecutable; +} + +ID3D11GeometryShader *ShaderExecutable11::getGeometryShader() const +{ + return mGeometryExecutable; +} + +ID3D11GeometryShader *ShaderExecutable11::getStreamOutShader() const +{ + return mStreamOutExecutable; +} + +UniformStorage11::UniformStorage11(Renderer11 *renderer, size_t initialSize) + : UniformStorageD3D(initialSize), + mConstantBuffer(NULL) +{ + ID3D11Device *d3d11Device = renderer->getDevice(); + + if (initialSize > 0) + { + D3D11_BUFFER_DESC constantBufferDescription = {0}; + constantBufferDescription.ByteWidth = initialSize; + constantBufferDescription.Usage = D3D11_USAGE_DYNAMIC; + constantBufferDescription.BindFlags = D3D11_BIND_CONSTANT_BUFFER; + constantBufferDescription.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + constantBufferDescription.MiscFlags = 0; + constantBufferDescription.StructureByteStride = 0; + + HRESULT result = d3d11Device->CreateBuffer(&constantBufferDescription, NULL, &mConstantBuffer); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + } +} + +UniformStorage11::~UniformStorage11() +{ + SafeRelease(mConstantBuffer); +} + +const UniformStorage11 *UniformStorage11::makeUniformStorage11(const UniformStorageD3D *uniformStorage) +{ + ASSERT(HAS_DYNAMIC_TYPE(const UniformStorage11*, uniformStorage)); + return static_cast<const UniformStorage11*>(uniformStorage); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h new file mode 100644 index 0000000000..02558ee4dc --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/ShaderExecutable11.h @@ -0,0 +1,59 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// ShaderExecutable11.h: Defines a D3D11-specific class to contain shader +// executable implementation details. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_SHADEREXECUTABLE11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_SHADEREXECUTABLE11_H_ + +#include "libANGLE/renderer/d3d/ShaderExecutableD3D.h" + +namespace rx +{ +class Renderer11; +class UniformStorage11; + +class ShaderExecutable11 : public ShaderExecutableD3D +{ + public: + ShaderExecutable11(const void *function, size_t length, ID3D11PixelShader *executable); + ShaderExecutable11(const void *function, size_t length, ID3D11VertexShader *executable, ID3D11GeometryShader *streamOut); + ShaderExecutable11(const void *function, size_t length, ID3D11GeometryShader *executable); + + virtual ~ShaderExecutable11(); + + static ShaderExecutable11 *makeShaderExecutable11(ShaderExecutableD3D *executable); + + ID3D11PixelShader *getPixelShader() const; + ID3D11VertexShader *getVertexShader() const; + ID3D11GeometryShader *getGeometryShader() const; + ID3D11GeometryShader *getStreamOutShader() const; + + private: + ID3D11PixelShader *mPixelExecutable; + ID3D11VertexShader *mVertexExecutable; + ID3D11GeometryShader *mGeometryExecutable; + ID3D11GeometryShader *mStreamOutExecutable; +}; + +class UniformStorage11 : public UniformStorageD3D +{ + public: + UniformStorage11(Renderer11 *renderer, size_t initialSize); + virtual ~UniformStorage11(); + + static const UniformStorage11 *makeUniformStorage11(const UniformStorageD3D *uniformStorage); + + ID3D11Buffer *getConstantBuffer() const { return mConstantBuffer; } + + private: + ID3D11Buffer *mConstantBuffer; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_SHADEREXECUTABLE11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.cpp new file mode 100644 index 0000000000..298f3ccbd2 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.cpp @@ -0,0 +1,711 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain11.cpp: Implements a back-end specific class for the D3D11 swap chain. + +#include "libANGLE/renderer/d3d/d3d11/SwapChain11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/NativeWindow.h" +#include "libANGLE/features.h" + +// Precompiled shaders +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthrough2d11vs.h" +#include "libANGLE/renderer/d3d/d3d11/shaders/compiled/passthroughrgba2d11ps.h" + + +namespace rx +{ + +SwapChain11::SwapChain11(Renderer11 *renderer, NativeWindow nativeWindow, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat) + : mRenderer(renderer), + SwapChainD3D(nativeWindow, shareHandle, backBufferFormat, depthBufferFormat), + mColorRenderTarget(this, renderer, false), + mDepthStencilRenderTarget(this, renderer, true) +{ + mSwapChain = NULL; + mBackBufferTexture = NULL; + mBackBufferRTView = NULL; + mOffscreenTexture = NULL; + mOffscreenRTView = NULL; + mOffscreenSRView = NULL; + mDepthStencilTexture = NULL; + mDepthStencilDSView = NULL; + mDepthStencilSRView = NULL; + mQuadVB = NULL; + mPassThroughSampler = NULL; + mPassThroughIL = NULL; + mPassThroughVS = NULL; + mPassThroughPS = NULL; + mWidth = -1; + mHeight = -1; + mSwapInterval = 0; + mAppCreatedShareHandle = mShareHandle != NULL; + mPassThroughResourcesInit = false; +} + +SwapChain11::~SwapChain11() +{ + release(); +} + +void SwapChain11::release() +{ + SafeRelease(mSwapChain); + SafeRelease(mBackBufferTexture); + SafeRelease(mBackBufferRTView); + SafeRelease(mOffscreenTexture); + SafeRelease(mOffscreenRTView); + SafeRelease(mOffscreenSRView); + SafeRelease(mDepthStencilTexture); + SafeRelease(mDepthStencilDSView); + SafeRelease(mDepthStencilSRView); + SafeRelease(mQuadVB); + SafeRelease(mPassThroughSampler); + SafeRelease(mPassThroughIL); + SafeRelease(mPassThroughVS); + SafeRelease(mPassThroughPS); + + if (!mAppCreatedShareHandle) + { + mShareHandle = NULL; + } +} + +void SwapChain11::releaseOffscreenTexture() +{ + SafeRelease(mOffscreenTexture); + SafeRelease(mOffscreenRTView); + SafeRelease(mOffscreenSRView); + SafeRelease(mDepthStencilTexture); + SafeRelease(mDepthStencilDSView); + SafeRelease(mDepthStencilSRView); +} + +EGLint SwapChain11::resetOffscreenTexture(int backbufferWidth, int backbufferHeight) +{ + ID3D11Device *device = mRenderer->getDevice(); + + ASSERT(device != NULL); + + // D3D11 does not allow zero size textures + ASSERT(backbufferWidth >= 1); + ASSERT(backbufferHeight >= 1); + + // Preserve the render target content + ID3D11Texture2D *previousOffscreenTexture = mOffscreenTexture; + if (previousOffscreenTexture) + { + previousOffscreenTexture->AddRef(); + } + const int previousWidth = mWidth; + const int previousHeight = mHeight; + + releaseOffscreenTexture(); + + const d3d11::TextureFormat &backbufferFormatInfo = d3d11::GetTextureFormatInfo(mBackBufferFormat, mRenderer->getFeatureLevel()); + + // If the app passed in a share handle, open the resource + // See EGL_ANGLE_d3d_share_handle_client_buffer + if (mAppCreatedShareHandle) + { + ID3D11Resource *tempResource11; + HRESULT result = device->OpenSharedResource(mShareHandle, __uuidof(ID3D11Resource), (void**)&tempResource11); + + if (FAILED(result)) + { + ERR("Failed to open the swap chain pbuffer share handle: %08lX", result); + release(); + return EGL_BAD_PARAMETER; + } + + result = tempResource11->QueryInterface(__uuidof(ID3D11Texture2D), (void**)&mOffscreenTexture); + SafeRelease(tempResource11); + + if (FAILED(result)) + { + ERR("Failed to query texture2d interface in pbuffer share handle: %08lX", result); + release(); + return EGL_BAD_PARAMETER; + } + + // Validate offscreen texture parameters + D3D11_TEXTURE2D_DESC offscreenTextureDesc = {0}; + mOffscreenTexture->GetDesc(&offscreenTextureDesc); + + if (offscreenTextureDesc.Width != (UINT)backbufferWidth || + offscreenTextureDesc.Height != (UINT)backbufferHeight || + offscreenTextureDesc.Format != backbufferFormatInfo.texFormat || + offscreenTextureDesc.MipLevels != 1 || + offscreenTextureDesc.ArraySize != 1) + { + ERR("Invalid texture parameters in the shared offscreen texture pbuffer"); + release(); + return EGL_BAD_PARAMETER; + } + } + else + { + const bool useSharedResource = !mNativeWindow.getNativeWindow() && mRenderer->getShareHandleSupport(); + + D3D11_TEXTURE2D_DESC offscreenTextureDesc = {0}; +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + const int textureLength = std::max(backbufferWidth, backbufferHeight); + offscreenTextureDesc.Width = textureLength; + offscreenTextureDesc.Height = textureLength; +#else + offscreenTextureDesc.Width = backbufferWidth; + offscreenTextureDesc.Height = backbufferHeight; +#endif + offscreenTextureDesc.Format = backbufferFormatInfo.texFormat; + offscreenTextureDesc.MipLevels = 1; + offscreenTextureDesc.ArraySize = 1; + offscreenTextureDesc.SampleDesc.Count = 1; + offscreenTextureDesc.SampleDesc.Quality = 0; + offscreenTextureDesc.Usage = D3D11_USAGE_DEFAULT; + offscreenTextureDesc.BindFlags = D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE; + offscreenTextureDesc.CPUAccessFlags = 0; + offscreenTextureDesc.MiscFlags = useSharedResource ? D3D11_RESOURCE_MISC_SHARED : 0; + + HRESULT result = device->CreateTexture2D(&offscreenTextureDesc, NULL, &mOffscreenTexture); + + if (FAILED(result)) + { + ERR("Could not create offscreen texture: %08lX", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + d3d11::SetDebugName(mOffscreenTexture, "Offscreen back buffer texture"); + + // EGL_ANGLE_surface_d3d_texture_2d_share_handle requires that we store a share handle for the client + if (useSharedResource) + { + IDXGIResource *offscreenTextureResource = NULL; + result = mOffscreenTexture->QueryInterface(__uuidof(IDXGIResource), (void**)&offscreenTextureResource); + + // Fall back to no share handle on failure + if (FAILED(result)) + { + ERR("Could not query offscreen texture resource: %08lX", result); + } + else + { + result = offscreenTextureResource->GetSharedHandle(&mShareHandle); + SafeRelease(offscreenTextureResource); + + if (FAILED(result)) + { + mShareHandle = NULL; + ERR("Could not get offscreen texture shared handle: %08lX", result); + } + } + } + } + + + D3D11_RENDER_TARGET_VIEW_DESC offscreenRTVDesc; + offscreenRTVDesc.Format = backbufferFormatInfo.rtvFormat; + offscreenRTVDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + offscreenRTVDesc.Texture2D.MipSlice = 0; + + HRESULT result = device->CreateRenderTargetView(mOffscreenTexture, &offscreenRTVDesc, &mOffscreenRTView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mOffscreenRTView, "Offscreen back buffer render target"); + + D3D11_SHADER_RESOURCE_VIEW_DESC offscreenSRVDesc; + offscreenSRVDesc.Format = backbufferFormatInfo.srvFormat; + offscreenSRVDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + offscreenSRVDesc.Texture2D.MostDetailedMip = 0; + offscreenSRVDesc.Texture2D.MipLevels = static_cast<UINT>(-1); + + result = device->CreateShaderResourceView(mOffscreenTexture, &offscreenSRVDesc, &mOffscreenSRView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mOffscreenSRView, "Offscreen back buffer shader resource"); + + const d3d11::TextureFormat &depthBufferFormatInfo = d3d11::GetTextureFormatInfo(mDepthBufferFormat, mRenderer->getFeatureLevel()); + + if (mDepthBufferFormat != GL_NONE) + { + D3D11_TEXTURE2D_DESC depthStencilTextureDesc; +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + const int textureLength = std::max(backbufferWidth, backbufferHeight); + depthStencilTextureDesc.Width = textureLength; + depthStencilTextureDesc.Height = textureLength; +#else + depthStencilTextureDesc.Width = backbufferWidth; + depthStencilTextureDesc.Height = backbufferHeight; +#endif + depthStencilTextureDesc.Format = depthBufferFormatInfo.texFormat; + depthStencilTextureDesc.MipLevels = 1; + depthStencilTextureDesc.ArraySize = 1; + depthStencilTextureDesc.SampleDesc.Count = 1; + depthStencilTextureDesc.SampleDesc.Quality = 0; + depthStencilTextureDesc.Usage = D3D11_USAGE_DEFAULT; + depthStencilTextureDesc.BindFlags = D3D11_BIND_DEPTH_STENCIL; + + if (depthBufferFormatInfo.srvFormat != DXGI_FORMAT_UNKNOWN) + { + depthStencilTextureDesc.BindFlags |= D3D11_BIND_SHADER_RESOURCE; + } + + depthStencilTextureDesc.CPUAccessFlags = 0; + depthStencilTextureDesc.MiscFlags = 0; + + result = device->CreateTexture2D(&depthStencilTextureDesc, NULL, &mDepthStencilTexture); + if (FAILED(result)) + { + ERR("Could not create depthstencil surface for new swap chain: 0x%08X", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + d3d11::SetDebugName(mDepthStencilTexture, "Offscreen depth stencil texture"); + + D3D11_DEPTH_STENCIL_VIEW_DESC depthStencilDesc; + depthStencilDesc.Format = depthBufferFormatInfo.dsvFormat; + depthStencilDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D; + depthStencilDesc.Flags = 0; + depthStencilDesc.Texture2D.MipSlice = 0; + + result = device->CreateDepthStencilView(mDepthStencilTexture, &depthStencilDesc, &mDepthStencilDSView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mDepthStencilDSView, "Offscreen depth stencil view"); + + if (depthBufferFormatInfo.srvFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_SHADER_RESOURCE_VIEW_DESC depthStencilSRVDesc; + depthStencilSRVDesc.Format = depthBufferFormatInfo.srvFormat; + depthStencilSRVDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + depthStencilSRVDesc.Texture2D.MostDetailedMip = 0; + depthStencilSRVDesc.Texture2D.MipLevels = static_cast<UINT>(-1); + + result = device->CreateShaderResourceView(mDepthStencilTexture, &depthStencilSRVDesc, &mDepthStencilSRView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mDepthStencilSRView, "Offscreen depth stencil shader resource"); + } + } + + mWidth = backbufferWidth; + mHeight = backbufferHeight; + + if (previousOffscreenTexture != NULL) + { + D3D11_BOX sourceBox = {0}; + sourceBox.left = 0; + sourceBox.right = std::min(previousWidth, mWidth); + sourceBox.top = std::max(previousHeight - mHeight, 0); + sourceBox.bottom = previousHeight; + sourceBox.front = 0; + sourceBox.back = 1; + + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + const int yoffset = std::max(mHeight - previousHeight, 0); + deviceContext->CopySubresourceRegion(mOffscreenTexture, 0, 0, yoffset, 0, previousOffscreenTexture, 0, &sourceBox); + + SafeRelease(previousOffscreenTexture); + + if (mSwapChain) + { + swapRect(0, 0, mWidth, mHeight); + } + } + + return EGL_SUCCESS; +} + +EGLint SwapChain11::resize(EGLint backbufferWidth, EGLint backbufferHeight) +{ + ID3D11Device *device = mRenderer->getDevice(); + + if (device == NULL) + { + return EGL_BAD_ACCESS; + } + + // EGL allows creating a surface with 0x0 dimension, however, DXGI does not like 0x0 swapchains + if (backbufferWidth < 1 || backbufferHeight < 1) + { + return EGL_SUCCESS; + } + +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) || (WINAPI_FAMILY != WINAPI_FAMILY_PHONE_APP) + // Can only call resize if we have already created our swap buffer and resources + ASSERT(mSwapChain && mBackBufferTexture && mBackBufferRTView); + + SafeRelease(mBackBufferTexture); + SafeRelease(mBackBufferRTView); + + // Resize swap chain + DXGI_SWAP_CHAIN_DESC desc; + mSwapChain->GetDesc(&desc); + const d3d11::TextureFormat &backbufferFormatInfo = d3d11::GetTextureFormatInfo(mBackBufferFormat, mRenderer->getFeatureLevel()); + HRESULT result = mSwapChain->ResizeBuffers(desc.BufferCount, backbufferWidth, backbufferHeight, backbufferFormatInfo.texFormat, 0); + + if (FAILED(result)) + { + ERR("Error resizing swap chain buffers: 0x%08X", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + result = mSwapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), (LPVOID*)&mBackBufferTexture); + ASSERT(SUCCEEDED(result)); + if (SUCCEEDED(result)) + { + d3d11::SetDebugName(mBackBufferTexture, "Back buffer texture"); + } + + result = device->CreateRenderTargetView(mBackBufferTexture, NULL, &mBackBufferRTView); + ASSERT(SUCCEEDED(result)); + if (SUCCEEDED(result)) + { + d3d11::SetDebugName(mBackBufferRTView, "Back buffer render target"); + } + + return resetOffscreenTexture(backbufferWidth, backbufferHeight); +#else + // Do nothing on Windows Phone apart from updating the internal buffer/width height + mWidth = backbufferWidth; + mHeight = backbufferHeight; + return EGL_SUCCESS; +#endif +} + +EGLint SwapChain11::reset(int backbufferWidth, int backbufferHeight, EGLint swapInterval) +{ + ID3D11Device *device = mRenderer->getDevice(); + + if (device == NULL) + { + return EGL_BAD_ACCESS; + } + + // Release specific resources to free up memory for the new render target, while the + // old render target still exists for the purpose of preserving its contents. + SafeRelease(mSwapChain); + SafeRelease(mBackBufferTexture); + SafeRelease(mBackBufferRTView); + + mSwapInterval = static_cast<unsigned int>(swapInterval); + if (mSwapInterval > 4) + { + // IDXGISwapChain::Present documentation states that valid sync intervals are in the [0,4] range + return EGL_BAD_PARAMETER; + } + + // EGL allows creating a surface with 0x0 dimension, however, DXGI does not like 0x0 swapchains + if (backbufferWidth < 1 || backbufferHeight < 1) + { + releaseOffscreenTexture(); + return EGL_SUCCESS; + } + + if (mNativeWindow.getNativeWindow()) + { + const d3d11::TextureFormat &backbufferFormatInfo = d3d11::GetTextureFormatInfo(mBackBufferFormat, mRenderer->getFeatureLevel()); + + HRESULT result = mNativeWindow.createSwapChain(device, mRenderer->getDxgiFactory(), + backbufferFormatInfo.texFormat, + backbufferWidth, backbufferHeight, &mSwapChain); + + if (FAILED(result)) + { + ERR("Could not create additional swap chains or offscreen surfaces: %08lX", result); + release(); + + if (d3d11::isDeviceLostError(result)) + { + return EGL_CONTEXT_LOST; + } + else + { + return EGL_BAD_ALLOC; + } + } + + result = mSwapChain->GetBuffer(0, __uuidof(ID3D11Texture2D), (LPVOID*)&mBackBufferTexture); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mBackBufferTexture, "Back buffer texture"); + + result = device->CreateRenderTargetView(mBackBufferTexture, NULL, &mBackBufferRTView); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mBackBufferRTView, "Back buffer render target"); + } + + // If we are resizing the swap chain, we don't wish to recreate all the static resources + if (!mPassThroughResourcesInit) + { + mPassThroughResourcesInit = true; + initPassThroughResources(); + } + + return resetOffscreenTexture(backbufferWidth, backbufferHeight); +} + +void SwapChain11::initPassThroughResources() +{ + ID3D11Device *device = mRenderer->getDevice(); + + ASSERT(device != NULL); + + // Make sure our resources are all not allocated, when we create + ASSERT(mQuadVB == NULL && mPassThroughSampler == NULL); + ASSERT(mPassThroughIL == NULL && mPassThroughVS == NULL && mPassThroughPS == NULL); + + D3D11_BUFFER_DESC vbDesc; + vbDesc.ByteWidth = sizeof(d3d11::PositionTexCoordVertex) * 4; + vbDesc.Usage = D3D11_USAGE_DYNAMIC; + vbDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + vbDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + vbDesc.MiscFlags = 0; + vbDesc.StructureByteStride = 0; + + HRESULT result = device->CreateBuffer(&vbDesc, NULL, &mQuadVB); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mQuadVB, "Swap chain quad vertex buffer"); + + D3D11_SAMPLER_DESC samplerDesc; + samplerDesc.Filter = D3D11_FILTER_MIN_MAG_MIP_POINT; + samplerDesc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP; + samplerDesc.MipLODBias = 0.0f; + samplerDesc.MaxAnisotropy = 0; + samplerDesc.ComparisonFunc = D3D11_COMPARISON_NEVER; + samplerDesc.BorderColor[0] = 0.0f; + samplerDesc.BorderColor[1] = 0.0f; + samplerDesc.BorderColor[2] = 0.0f; + samplerDesc.BorderColor[3] = 0.0f; + samplerDesc.MinLOD = 0; + samplerDesc.MaxLOD = D3D11_FLOAT32_MAX; + + result = device->CreateSamplerState(&samplerDesc, &mPassThroughSampler); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughSampler, "Swap chain pass through sampler"); + + D3D11_INPUT_ELEMENT_DESC quadLayout[] = + { + { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 0, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + { "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT, 0, 8, D3D11_INPUT_PER_VERTEX_DATA, 0 }, + }; + + result = device->CreateInputLayout(quadLayout, 2, g_VS_Passthrough2D, sizeof(g_VS_Passthrough2D), &mPassThroughIL); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughIL, "Swap chain pass through layout"); + + result = device->CreateVertexShader(g_VS_Passthrough2D, sizeof(g_VS_Passthrough2D), NULL, &mPassThroughVS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughVS, "Swap chain pass through vertex shader"); + + result = device->CreatePixelShader(g_PS_PassthroughRGBA2D, sizeof(g_PS_PassthroughRGBA2D), NULL, &mPassThroughPS); + ASSERT(SUCCEEDED(result)); + d3d11::SetDebugName(mPassThroughPS, "Swap chain pass through pixel shader"); +} + +// parameters should be validated/clamped by caller +EGLint SwapChain11::swapRect(EGLint x, EGLint y, EGLint width, EGLint height) +{ + if (!mSwapChain) + { + return EGL_SUCCESS; + } + + ID3D11Device *device = mRenderer->getDevice(); + ID3D11DeviceContext *deviceContext = mRenderer->getDeviceContext(); + + // Set vertices + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = deviceContext->Map(mQuadVB, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return EGL_BAD_ACCESS; + } + + d3d11::PositionTexCoordVertex *vertices = static_cast<d3d11::PositionTexCoordVertex*>(mappedResource.pData); + + // Create a quad in homogeneous coordinates + float x1 = (x / float(mWidth)) * 2.0f - 1.0f; + float y1 = (y / float(mHeight)) * 2.0f - 1.0f; + float x2 = ((x + width) / float(mWidth)) * 2.0f - 1.0f; + float y2 = ((y + height) / float(mHeight)) * 2.0f - 1.0f; + +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + const float dim = std::max(mWidth, mHeight); + float u1 = x / dim; + float v1 = y / dim; + float u2 = (x + width) / dim; + float v2 = (y + height) / dim; + + const NativeWindow::RotationFlags flags = mNativeWindow.rotationFlags(); + const bool rotateL = flags == NativeWindow::RotateLeft; + const bool rotateR = flags == NativeWindow::RotateRight; + d3d11::SetPositionTexCoordVertex(&vertices[0], x1, y1, rotateL ? u2 : u1, rotateR ? v2 : v1); + d3d11::SetPositionTexCoordVertex(&vertices[1], x1, y2, rotateR ? u2 : u1, rotateL ? v1 : v2); + d3d11::SetPositionTexCoordVertex(&vertices[2], x2, y1, rotateR ? u1 : u2, rotateL ? v2 : v1); + d3d11::SetPositionTexCoordVertex(&vertices[3], x2, y2, rotateL ? u1 : u2, rotateR ? v1 : v2); +#else + float u1 = x / float(mWidth); + float v1 = y / float(mHeight); + float u2 = (x + width) / float(mWidth); + float v2 = (y + height) / float(mHeight); + + d3d11::SetPositionTexCoordVertex(&vertices[0], x1, y1, u1, v1); + d3d11::SetPositionTexCoordVertex(&vertices[1], x1, y2, u1, v2); + d3d11::SetPositionTexCoordVertex(&vertices[2], x2, y1, u2, v1); + d3d11::SetPositionTexCoordVertex(&vertices[3], x2, y2, u2, v2); +#endif + + deviceContext->Unmap(mQuadVB, 0); + + static UINT stride = sizeof(d3d11::PositionTexCoordVertex); + static UINT startIdx = 0; + deviceContext->IASetVertexBuffers(0, 1, &mQuadVB, &stride, &startIdx); + + // Apply state + deviceContext->OMSetDepthStencilState(NULL, 0xFFFFFFFF); + + static const float blendFactor[4] = { 1.0f, 1.0f, 1.0f, 1.0f }; + deviceContext->OMSetBlendState(NULL, blendFactor, 0xFFFFFFF); + + deviceContext->RSSetState(NULL); + + // Apply shaders + deviceContext->IASetInputLayout(mPassThroughIL); + deviceContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + deviceContext->VSSetShader(mPassThroughVS, NULL, 0); + deviceContext->PSSetShader(mPassThroughPS, NULL, 0); + deviceContext->GSSetShader(NULL, NULL, 0); + + // Apply render targets + mRenderer->setOneTimeRenderTarget(mBackBufferRTView); + + // Set the viewport + D3D11_VIEWPORT viewport; + viewport.TopLeftX = 0; + viewport.TopLeftY = 0; +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + viewport.Width = (rotateL || rotateR) ? mHeight : mWidth; + viewport.Height = (rotateL || rotateR) ? mWidth : mHeight; +#else + viewport.Width = mWidth; + viewport.Height = mHeight; +#endif + viewport.MinDepth = 0.0f; + viewport.MaxDepth = 1.0f; + deviceContext->RSSetViewports(1, &viewport); + + // Apply textures + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, mOffscreenSRView); + deviceContext->PSSetSamplers(0, 1, &mPassThroughSampler); + + // Draw + deviceContext->Draw(4, 0); + +#if ANGLE_VSYNC == ANGLE_DISABLED + result = mSwapChain->Present(0, 0); +#else + result = mSwapChain->Present(mSwapInterval, 0); +#endif + + if (result == DXGI_ERROR_DEVICE_REMOVED) + { + HRESULT removedReason = device->GetDeviceRemovedReason(); + UNUSED_TRACE_VARIABLE(removedReason); + ERR("Present failed: the D3D11 device was removed: 0x%08X", removedReason); + return EGL_CONTEXT_LOST; + } + else if (result == DXGI_ERROR_DEVICE_RESET) + { + ERR("Present failed: the D3D11 device was reset from a bad command."); + return EGL_CONTEXT_LOST; + } + else if (FAILED(result)) + { + ERR("Present failed with error code 0x%08X", result); + } + + // Unbind + mRenderer->setShaderResource(gl::SAMPLER_PIXEL, 0, NULL); + + mRenderer->unapplyRenderTargets(); + mRenderer->markAllStateDirty(); + + return EGL_SUCCESS; +} + +ID3D11Texture2D *SwapChain11::getOffscreenTexture() +{ + return mOffscreenTexture; +} + +ID3D11RenderTargetView *SwapChain11::getRenderTarget() +{ + return mOffscreenRTView; +} + +ID3D11ShaderResourceView *SwapChain11::getRenderTargetShaderResource() +{ + return mOffscreenSRView; +} + +ID3D11DepthStencilView *SwapChain11::getDepthStencil() +{ + return mDepthStencilDSView; +} + +ID3D11ShaderResourceView * SwapChain11::getDepthStencilShaderResource() +{ + return mDepthStencilSRView; +} + +ID3D11Texture2D *SwapChain11::getDepthStencilTexture() +{ + return mDepthStencilTexture; +} + +SwapChain11 *SwapChain11::makeSwapChain11(SwapChainD3D *swapChain) +{ + ASSERT(HAS_DYNAMIC_TYPE(SwapChain11*, swapChain)); + return static_cast<SwapChain11*>(swapChain); +} + +void SwapChain11::recreate() +{ + // possibly should use this method instead of reset +} + +void *rx::SwapChain11::getDevice() +{ + return mRenderer->getDevice(); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.h new file mode 100644 index 0000000000..48c808a261 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/SwapChain11.h @@ -0,0 +1,87 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChain11.h: Defines a back-end specific class for the D3D11 swap chain. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_SWAPCHAIN11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_SWAPCHAIN11_H_ + +#include "common/angleutils.h" +#include "libANGLE/renderer/d3d/SwapChainD3D.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" + +namespace rx +{ +class Renderer11; + +class SwapChain11 : public SwapChainD3D +{ + public: + SwapChain11(Renderer11 *renderer, NativeWindow nativeWindow, HANDLE shareHandle, + GLenum backBufferFormat, GLenum depthBufferFormat); + virtual ~SwapChain11(); + + EGLint resize(EGLint backbufferWidth, EGLint backbufferHeight); + virtual EGLint reset(EGLint backbufferWidth, EGLint backbufferHeight, EGLint swapInterval); + virtual EGLint swapRect(EGLint x, EGLint y, EGLint width, EGLint height); + virtual void recreate(); + + RenderTargetD3D *getColorRenderTarget() override { return &mColorRenderTarget; } + RenderTargetD3D *getDepthStencilRenderTarget() override { return &mDepthStencilRenderTarget; } + + virtual ID3D11Texture2D *getOffscreenTexture(); + virtual ID3D11RenderTargetView *getRenderTarget(); + virtual ID3D11ShaderResourceView *getRenderTargetShaderResource(); + + virtual ID3D11Texture2D *getDepthStencilTexture(); + virtual ID3D11DepthStencilView *getDepthStencil(); + virtual ID3D11ShaderResourceView *getDepthStencilShaderResource(); + + EGLint getWidth() const { return mWidth; } + EGLint getHeight() const { return mHeight; } + + virtual void *getDevice(); + + static SwapChain11 *makeSwapChain11(SwapChainD3D *swapChain); + + private: + void release(); + void initPassThroughResources(); + void releaseOffscreenTexture(); + EGLint resetOffscreenTexture(int backbufferWidth, int backbufferHeight); + + Renderer11 *mRenderer; + EGLint mHeight; + EGLint mWidth; + bool mAppCreatedShareHandle; + unsigned int mSwapInterval; + bool mPassThroughResourcesInit; + + DXGISwapChain *mSwapChain; + + ID3D11Texture2D *mBackBufferTexture; + ID3D11RenderTargetView *mBackBufferRTView; + + ID3D11Texture2D *mOffscreenTexture; + ID3D11RenderTargetView *mOffscreenRTView; + ID3D11ShaderResourceView *mOffscreenSRView; + + ID3D11Texture2D *mDepthStencilTexture; + ID3D11DepthStencilView *mDepthStencilDSView; + ID3D11ShaderResourceView *mDepthStencilSRView; + + ID3D11Buffer *mQuadVB; + ID3D11SamplerState *mPassThroughSampler; + ID3D11InputLayout *mPassThroughIL; + ID3D11VertexShader *mPassThroughVS; + ID3D11PixelShader *mPassThroughPS; + + SurfaceRenderTarget11 mColorRenderTarget; + SurfaceRenderTarget11 mDepthStencilRenderTarget; +}; + +} +#endif // LIBANGLE_RENDERER_D3D_D3D11_SWAPCHAIN11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.cpp new file mode 100644 index 0000000000..103e90fed3 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.cpp @@ -0,0 +1,2676 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage11.cpp: Implements the abstract rx::TextureStorage11 class and its concrete derived +// classes TextureStorage11_2D and TextureStorage11_Cube, which act as the interface to the D3D11 texture. + +#include "libANGLE/renderer/d3d/d3d11/TextureStorage11.h" + +#include <tuple> + +#include "common/MemoryBuffer.h" +#include "common/utilities.h" +#include "libANGLE/ImageIndex.h" +#include "libANGLE/formatutils.h" +#include "libANGLE/renderer/d3d/TextureD3D.h" +#include "libANGLE/renderer/d3d/d3d11/Blit11.h" +#include "libANGLE/renderer/d3d/d3d11/Image11.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/SwapChain11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" + +namespace rx +{ + +TextureStorage11::SwizzleCacheValue::SwizzleCacheValue() + : swizzleRed(GL_NONE), swizzleGreen(GL_NONE), swizzleBlue(GL_NONE), swizzleAlpha(GL_NONE) +{ +} + +TextureStorage11::SwizzleCacheValue::SwizzleCacheValue(GLenum red, GLenum green, GLenum blue, GLenum alpha) + : swizzleRed(red), swizzleGreen(green), swizzleBlue(blue), swizzleAlpha(alpha) +{ +} + +bool TextureStorage11::SwizzleCacheValue::operator==(const SwizzleCacheValue &other) const +{ + return swizzleRed == other.swizzleRed && + swizzleGreen == other.swizzleGreen && + swizzleBlue == other.swizzleBlue && + swizzleAlpha == other.swizzleAlpha; +} + +bool TextureStorage11::SwizzleCacheValue::operator!=(const SwizzleCacheValue &other) const +{ + return !(*this == other); +} + +TextureStorage11::SRVKey::SRVKey(int baseLevel, int mipLevels, bool swizzle) + : baseLevel(baseLevel), mipLevels(mipLevels), swizzle(swizzle) +{ +} + +bool TextureStorage11::SRVKey::operator<(const SRVKey &rhs) const +{ + return std::tie(baseLevel, mipLevels, swizzle) < std::tie(rhs.baseLevel, rhs.mipLevels, rhs.swizzle); +} + +TextureStorage11::TextureStorage11(Renderer11 *renderer, UINT bindFlags) + : mBindFlags(bindFlags), + mTopLevel(0), + mMipLevels(0), + mInternalFormat(GL_NONE), + mTextureFormat(DXGI_FORMAT_UNKNOWN), + mShaderResourceFormat(DXGI_FORMAT_UNKNOWN), + mRenderTargetFormat(DXGI_FORMAT_UNKNOWN), + mDepthStencilFormat(DXGI_FORMAT_UNKNOWN), + mTextureWidth(0), + mTextureHeight(0), + mTextureDepth(0) +{ + mRenderer = Renderer11::makeRenderer11(renderer); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mLevelSRVs[i] = NULL; + } +} + +TextureStorage11::~TextureStorage11() +{ + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + SafeRelease(mLevelSRVs[level]); + } + + for (SRVCache::iterator i = mSrvCache.begin(); i != mSrvCache.end(); i++) + { + SafeRelease(i->second); + } + mSrvCache.clear(); +} + +TextureStorage11 *TextureStorage11::makeTextureStorage11(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11*, storage)); + return static_cast<TextureStorage11*>(storage); +} + +DWORD TextureStorage11::GetTextureBindFlags(GLenum internalFormat, D3D_FEATURE_LEVEL featureLevel, bool renderTarget) +{ + UINT bindFlags = 0; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalFormat, featureLevel); + if (formatInfo.srvFormat != DXGI_FORMAT_UNKNOWN) + { + bindFlags |= D3D11_BIND_SHADER_RESOURCE; + } + if (formatInfo.dsvFormat != DXGI_FORMAT_UNKNOWN) + { + bindFlags |= D3D11_BIND_DEPTH_STENCIL; + } + if (formatInfo.rtvFormat != DXGI_FORMAT_UNKNOWN && renderTarget) + { + bindFlags |= D3D11_BIND_RENDER_TARGET; + } + + return bindFlags; +} + +UINT TextureStorage11::getBindFlags() const +{ + return mBindFlags; +} + +int TextureStorage11::getTopLevel() const +{ + return mTopLevel; +} + +bool TextureStorage11::isRenderTarget() const +{ + return (mBindFlags & (D3D11_BIND_RENDER_TARGET | D3D11_BIND_DEPTH_STENCIL)) != 0; +} + +bool TextureStorage11::isManaged() const +{ + return false; +} + +int TextureStorage11::getLevelCount() const +{ + return mMipLevels - mTopLevel; +} + +int TextureStorage11::getLevelWidth(int mipLevel) const +{ + return std::max(static_cast<int>(mTextureWidth) >> mipLevel, 1); +} + +int TextureStorage11::getLevelHeight(int mipLevel) const +{ + return std::max(static_cast<int>(mTextureHeight) >> mipLevel, 1); +} + +int TextureStorage11::getLevelDepth(int mipLevel) const +{ + return std::max(static_cast<int>(mTextureDepth) >> mipLevel, 1); +} + +UINT TextureStorage11::getSubresourceIndex(const gl::ImageIndex &index) const +{ + UINT mipSlice = static_cast<UINT>(index.mipIndex + mTopLevel); + UINT arraySlice = static_cast<UINT>(index.hasLayer() ? index.layerIndex : 0); + UINT subresource = D3D11CalcSubresource(mipSlice, arraySlice, mMipLevels); + ASSERT(subresource != std::numeric_limits<UINT>::max()); + return subresource; +} + +gl::Error TextureStorage11::getSRV(const gl::SamplerState &samplerState, ID3D11ShaderResourceView **outSRV) +{ + bool swizzleRequired = samplerState.swizzleRequired(); + bool mipmapping = gl::IsMipmapFiltered(samplerState); + unsigned int mipLevels = mipmapping ? (samplerState.maxLevel - samplerState.baseLevel + 1) : 1; + + // Make sure there's 'mipLevels' mipmap levels below the base level (offset by the top level, which corresponds to GL level 0) + mipLevels = std::min(mipLevels, mMipLevels - mTopLevel - samplerState.baseLevel); + + if (mRenderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3) + { + ASSERT(!swizzleRequired); + ASSERT(mipLevels == 1 || mipLevels == mMipLevels); + } + + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + // We must ensure that the level zero texture is in sync with mipped texture. + gl::Error error = useLevelZeroWorkaroundTexture(mipLevels == 1); + if (error.isError()) + { + return error; + } + } + + if (swizzleRequired) + { + verifySwizzleExists(samplerState.swizzleRed, samplerState.swizzleGreen, samplerState.swizzleBlue, samplerState.swizzleAlpha); + } + + SRVKey key(samplerState.baseLevel, mipLevels, swizzleRequired); + SRVCache::const_iterator iter = mSrvCache.find(key); + if (iter != mSrvCache.end()) + { + *outSRV = iter->second; + } + else + { + ID3D11Resource *texture = NULL; + if (swizzleRequired) + { + gl::Error error = getSwizzleTexture(&texture); + if (error.isError()) + { + return error; + } + } + else + { + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + } + + ID3D11ShaderResourceView *srv = NULL; + DXGI_FORMAT format = (swizzleRequired ? mSwizzleShaderResourceFormat : mShaderResourceFormat); + gl::Error error = createSRV(samplerState.baseLevel, mipLevels, format, texture, &srv); + if (error.isError()) + { + return error; + } + + mSrvCache.insert(std::make_pair(key, srv)); + *outSRV = srv; + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11::getSRVLevel(int mipLevel, ID3D11ShaderResourceView **outSRV) +{ + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + + if (!mLevelSRVs[mipLevel]) + { + ID3D11Resource *resource = NULL; + gl::Error error = getResource(&resource); + if (error.isError()) + { + return error; + } + + error = createSRV(mipLevel, 1, mShaderResourceFormat, resource, &mLevelSRVs[mipLevel]); + if (error.isError()) + { + return error; + } + } + + *outSRV = mLevelSRVs[mipLevel]; + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11::generateSwizzles(GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha) +{ + SwizzleCacheValue swizzleTarget(swizzleRed, swizzleGreen, swizzleBlue, swizzleAlpha); + for (int level = 0; level < getLevelCount(); level++) + { + // Check if the swizzle for this level is out of date + if (mSwizzleCache[level] != swizzleTarget) + { + // Need to re-render the swizzle for this level + ID3D11ShaderResourceView *sourceSRV = NULL; + gl::Error error = getSRVLevel(level, &sourceSRV); + if (error.isError()) + { + return error; + } + + ID3D11RenderTargetView *destRTV = NULL; + error = getSwizzleRenderTarget(level, &destRTV); + if (error.isError()) + { + return error; + } + + gl::Extents size(getLevelWidth(level), getLevelHeight(level), getLevelDepth(level)); + + Blit11 *blitter = mRenderer->getBlitter(); + + error = blitter->swizzleTexture(sourceSRV, destRTV, size, swizzleRed, swizzleGreen, swizzleBlue, swizzleAlpha); + if (error.isError()) + { + return error; + } + + mSwizzleCache[level] = swizzleTarget; + } + } + + return gl::Error(GL_NO_ERROR); +} + +void TextureStorage11::invalidateSwizzleCacheLevel(int mipLevel) +{ + if (mipLevel >= 0 && static_cast<unsigned int>(mipLevel) < ArraySize(mSwizzleCache)) + { + // The default constructor of SwizzleCacheValue has GL_NONE for all channels which is not a + // valid swizzle combination + mSwizzleCache[mipLevel] = SwizzleCacheValue(); + } +} + +void TextureStorage11::invalidateSwizzleCache() +{ + for (unsigned int mipLevel = 0; mipLevel < ArraySize(mSwizzleCache); mipLevel++) + { + invalidateSwizzleCacheLevel(mipLevel); + } +} + +gl::Error TextureStorage11::updateSubresourceLevel(ID3D11Resource *srcTexture, unsigned int sourceSubresource, + const gl::ImageIndex &index, const gl::Box ©Area) +{ + ASSERT(srcTexture); + + GLint level = index.mipIndex; + + invalidateSwizzleCacheLevel(level); + + gl::Extents texSize(getLevelWidth(level), getLevelHeight(level), getLevelDepth(level)); + + bool fullCopy = copyArea.x == 0 && + copyArea.y == 0 && + copyArea.z == 0 && + copyArea.width == texSize.width && + copyArea.height == texSize.height && + copyArea.depth == texSize.depth; + + ID3D11Resource *dstTexture = NULL; + gl::Error error(GL_NO_ERROR); + + // If the zero-LOD workaround is active and we want to update a level greater than zero, then we should + // update the mipmapped texture, even if mapmaps are currently disabled. + if (index.mipIndex > 0 && mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + error = getMippedResource(&dstTexture); + } + else + { + error = getResource(&dstTexture); + } + + if (error.isError()) + { + return error; + } + + unsigned int dstSubresource = getSubresourceIndex(index); + + ASSERT(dstTexture); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(mTextureFormat); + if (!fullCopy && (dxgiFormatInfo.depthBits > 0 || dxgiFormatInfo.stencilBits > 0)) + { + // CopySubresourceRegion cannot copy partial depth stencils, use the blitter instead + Blit11 *blitter = mRenderer->getBlitter(); + + return blitter->copyDepthStencil(srcTexture, sourceSubresource, copyArea, texSize, + dstTexture, dstSubresource, copyArea, texSize, + NULL); + } + else + { + D3D11_BOX srcBox; + srcBox.left = copyArea.x; + srcBox.top = copyArea.y; + srcBox.right = copyArea.x + roundUp(static_cast<UINT>(copyArea.width), dxgiFormatInfo.blockWidth); + srcBox.bottom = copyArea.y + roundUp(static_cast<UINT>(copyArea.height), dxgiFormatInfo.blockHeight); + srcBox.front = copyArea.z; + srcBox.back = copyArea.z + copyArea.depth; + + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + context->CopySubresourceRegion(dstTexture, dstSubresource, copyArea.x, copyArea.y, copyArea.z, + srcTexture, sourceSubresource, fullCopy ? NULL : &srcBox); + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error TextureStorage11::copySubresourceLevel(ID3D11Resource* dstTexture, unsigned int dstSubresource, + const gl::ImageIndex &index, const gl::Box ®ion) +{ + ASSERT(dstTexture); + + ID3D11Resource *srcTexture = NULL; + gl::Error error(GL_NO_ERROR); + + // If the zero-LOD workaround is active and we want to update a level greater than zero, then we should + // update the mipmapped texture, even if mapmaps are currently disabled. + if (index.mipIndex > 0 && mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + error = getMippedResource(&srcTexture); + } + else + { + error = getResource(&srcTexture); + } + + if (error.isError()) + { + return error; + } + + ASSERT(srcTexture); + + unsigned int srcSubresource = getSubresourceIndex(index); + + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + // D3D11 can't perform partial CopySubresourceRegion on depth/stencil textures, so pSrcBox should be NULL. + D3D11_BOX srcBox; + D3D11_BOX *pSrcBox = NULL; + if (mRenderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3) + { + // However, D3D10Level9 doesn't always perform CopySubresourceRegion correctly unless the source box + // is specified. This is okay, since we don't perform CopySubresourceRegion on depth/stencil + // textures on 9_3. + ASSERT(d3d11::GetDXGIFormatInfo(mTextureFormat).depthBits == 0); + ASSERT(d3d11::GetDXGIFormatInfo(mTextureFormat).stencilBits == 0); + srcBox.left = region.x; + srcBox.right = region.x + region.width; + srcBox.top = region.y; + srcBox.bottom = region.y + region.height; + srcBox.front = region.z; + srcBox.back = region.z + region.depth; + pSrcBox = &srcBox; + } + + context->CopySubresourceRegion(dstTexture, dstSubresource, region.x, region.y, region.z, + srcTexture, srcSubresource, pSrcBox); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11::generateMipmap(const gl::ImageIndex &sourceIndex, const gl::ImageIndex &destIndex) +{ + ASSERT(sourceIndex.layerIndex == destIndex.layerIndex); + + invalidateSwizzleCacheLevel(destIndex.mipIndex); + + RenderTargetD3D *source = NULL; + gl::Error error = getRenderTarget(sourceIndex, &source); + if (error.isError()) + { + return error; + } + + RenderTargetD3D *dest = NULL; + error = getRenderTarget(destIndex, &dest); + if (error.isError()) + { + return error; + } + + ID3D11ShaderResourceView *sourceSRV = RenderTarget11::makeRenderTarget11(source)->getShaderResourceView(); + ID3D11RenderTargetView *destRTV = RenderTarget11::makeRenderTarget11(dest)->getRenderTargetView(); + + gl::Box sourceArea(0, 0, 0, source->getWidth(), source->getHeight(), source->getDepth()); + gl::Extents sourceSize(source->getWidth(), source->getHeight(), source->getDepth()); + + gl::Box destArea(0, 0, 0, dest->getWidth(), dest->getHeight(), dest->getDepth()); + gl::Extents destSize(dest->getWidth(), dest->getHeight(), dest->getDepth()); + + Blit11 *blitter = mRenderer->getBlitter(); + return blitter->copyTexture(sourceSRV, sourceArea, sourceSize, destRTV, destArea, destSize, NULL, + gl::GetInternalFormatInfo(source->getInternalFormat()).format, GL_LINEAR); +} + +void TextureStorage11::verifySwizzleExists(GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha) +{ + SwizzleCacheValue swizzleTarget(swizzleRed, swizzleGreen, swizzleBlue, swizzleAlpha); + for (unsigned int level = 0; level < mMipLevels; level++) + { + ASSERT(mSwizzleCache[level] == swizzleTarget); + } +} + +gl::Error TextureStorage11::copyToStorage(TextureStorage *destStorage) +{ + ASSERT(destStorage); + + ID3D11Resource *sourceResouce = NULL; + gl::Error error = getResource(&sourceResouce); + if (error.isError()) + { + return error; + } + + TextureStorage11 *dest11 = TextureStorage11::makeTextureStorage11(destStorage); + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + immediateContext->CopyResource(destResource, sourceResouce); + + dest11->invalidateSwizzleCache(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11::setData(const gl::ImageIndex &index, ImageD3D *image, const gl::Box *destBox, GLenum type, + const gl::PixelUnpackState &unpack, const uint8_t *pixelData) +{ + ASSERT(!image->isDirty()); + + ID3D11Resource *resource = NULL; + gl::Error error = getResource(&resource); + if (error.isError()) + { + return error; + } + ASSERT(resource); + + UINT destSubresource = getSubresourceIndex(index); + + const gl::InternalFormat &internalFormatInfo = gl::GetInternalFormatInfo(image->getInternalFormat()); + + gl::Box levelBox(0, 0, 0, getLevelWidth(index.mipIndex), getLevelHeight(index.mipIndex), getLevelDepth(index.mipIndex)); + bool fullUpdate = (destBox == NULL || *destBox == levelBox); + ASSERT(internalFormatInfo.depthBits == 0 || fullUpdate); + + // TODO(jmadill): Handle compressed formats + // Compressed formats have different load syntax, so we'll have to handle them with slightly + // different logic. Will implemnent this in a follow-up patch, and ensure we do not use SetData + // with compressed formats in the calling logic. + ASSERT(!internalFormatInfo.compressed); + + int width = destBox ? destBox->width : static_cast<int>(image->getWidth()); + int height = destBox ? destBox->height : static_cast<int>(image->getHeight()); + int depth = destBox ? destBox->depth : static_cast<int>(image->getDepth()); + UINT srcRowPitch = internalFormatInfo.computeRowPitch(type, width, unpack.alignment, unpack.rowLength); + UINT srcDepthPitch = internalFormatInfo.computeDepthPitch(type, width, height, unpack.alignment, unpack.rowLength); + + const d3d11::TextureFormat &d3d11Format = d3d11::GetTextureFormatInfo(image->getInternalFormat(), mRenderer->getFeatureLevel()); + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(d3d11Format.texFormat); + + size_t outputPixelSize = dxgiFormatInfo.pixelBytes; + + UINT bufferRowPitch = outputPixelSize * width; + UINT bufferDepthPitch = bufferRowPitch * height; + + size_t neededSize = bufferDepthPitch * depth; + MemoryBuffer *conversionBuffer = NULL; + error = mRenderer->getScratchMemoryBuffer(neededSize, &conversionBuffer); + if (error.isError()) + { + return error; + } + + // TODO: fast path + LoadImageFunction loadFunction = d3d11Format.loadFunctions.at(type); + loadFunction(width, height, depth, + pixelData, srcRowPitch, srcDepthPitch, + conversionBuffer->data(), bufferRowPitch, bufferDepthPitch); + + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + + if (!fullUpdate) + { + ASSERT(destBox); + + D3D11_BOX destD3DBox; + destD3DBox.left = destBox->x; + destD3DBox.right = destBox->x + destBox->width; + destD3DBox.top = destBox->y; + destD3DBox.bottom = destBox->y + destBox->height; + destD3DBox.front = destBox->z; + destD3DBox.back = destBox->z + destBox->depth; + + immediateContext->UpdateSubresource(resource, destSubresource, + &destD3DBox, conversionBuffer->data(), + bufferRowPitch, bufferDepthPitch); + } + else + { + immediateContext->UpdateSubresource(resource, destSubresource, + NULL, conversionBuffer->data(), + bufferRowPitch, bufferDepthPitch); + } + + return gl::Error(GL_NO_ERROR); +} + +TextureStorage11_2D::TextureStorage11_2D(Renderer11 *renderer, SwapChain11 *swapchain) + : TextureStorage11(renderer, D3D11_BIND_RENDER_TARGET | D3D11_BIND_SHADER_RESOURCE), + mTexture(swapchain->getOffscreenTexture()), + mSwizzleTexture(NULL), + mLevelZeroTexture(NULL), + mLevelZeroRenderTarget(NULL), + mUseLevelZeroTexture(false) +{ + mTexture->AddRef(); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mAssociatedImages[i] = NULL; + mRenderTarget[i] = NULL; + mSwizzleRenderTargets[i] = NULL; + } + + D3D11_TEXTURE2D_DESC texDesc; + mTexture->GetDesc(&texDesc); + mMipLevels = texDesc.MipLevels; + mTextureFormat = texDesc.Format; + mTextureWidth = texDesc.Width; + mTextureHeight = texDesc.Height; + mTextureDepth = 1; + + mInternalFormat = swapchain->GetBackBufferInternalFormat(); + + ID3D11ShaderResourceView *srv = swapchain->getRenderTargetShaderResource(); + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srv->GetDesc(&srvDesc); + mShaderResourceFormat = srvDesc.Format; + + ID3D11RenderTargetView* offscreenRTV = swapchain->getRenderTarget(); + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + offscreenRTV->GetDesc(&rtvDesc); + mRenderTargetFormat = rtvDesc.Format; + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(mTextureFormat); + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(dxgiFormatInfo.internalFormat, mRenderer->getFeatureLevel()); + mSwizzleTextureFormat = formatInfo.swizzleTexFormat; + mSwizzleShaderResourceFormat = formatInfo.swizzleSRVFormat; + mSwizzleRenderTargetFormat = formatInfo.swizzleRTVFormat; + + mDepthStencilFormat = DXGI_FORMAT_UNKNOWN; + + initializeSerials(1, 1); +} + +TextureStorage11_2D::TextureStorage11_2D(Renderer11 *renderer, GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, int levels, bool hintLevelZeroOnly) + : TextureStorage11(renderer, GetTextureBindFlags(internalformat, renderer->getFeatureLevel(), renderTarget)), + mTexture(NULL), + mSwizzleTexture(NULL), + mLevelZeroTexture(NULL), + mLevelZeroRenderTarget(NULL), + mUseLevelZeroTexture(false) +{ + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mAssociatedImages[i] = NULL; + mRenderTarget[i] = NULL; + mSwizzleRenderTargets[i] = NULL; + } + + mInternalFormat = internalformat; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalformat, renderer->getFeatureLevel()); + mTextureFormat = formatInfo.texFormat; + mShaderResourceFormat = formatInfo.srvFormat; + mDepthStencilFormat = formatInfo.dsvFormat; + mRenderTargetFormat = formatInfo.rtvFormat; + mSwizzleTextureFormat = formatInfo.swizzleTexFormat; + mSwizzleShaderResourceFormat = formatInfo.swizzleSRVFormat; + mSwizzleRenderTargetFormat = formatInfo.swizzleRTVFormat; + + d3d11::MakeValidSize(false, mTextureFormat, &width, &height, &mTopLevel); + mMipLevels = mTopLevel + levels; + mTextureWidth = width; + mTextureHeight = height; + mTextureDepth = 1; + + if (hintLevelZeroOnly && levels > 1) + { + //The LevelZeroOnly hint should only be true if the zero max LOD workaround is active. + ASSERT(mRenderer->getWorkarounds().zeroMaxLodWorkaround); + mUseLevelZeroTexture = true; + } + + initializeSerials(getLevelCount(), 1); +} + +TextureStorage11_2D::~TextureStorage11_2D() +{ + for (unsigned i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + if (mAssociatedImages[i] != NULL) + { + bool imageAssociationCorrect = mAssociatedImages[i]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // We must let the Images recover their data before we delete it from the TextureStorage. + gl::Error error = mAssociatedImages[i]->recoverFromAssociatedStorage(); + if (error.isError()) + { + // TODO: Find a way to report this back to the context + } + } + } + } + + SafeRelease(mTexture); + SafeRelease(mSwizzleTexture); + + SafeRelease(mLevelZeroTexture); + SafeDelete(mLevelZeroRenderTarget); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + SafeDelete(mRenderTarget[i]); + SafeRelease(mSwizzleRenderTargets[i]); + } +} + +TextureStorage11_2D *TextureStorage11_2D::makeTextureStorage11_2D(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_2D*, storage)); + return static_cast<TextureStorage11_2D*>(storage); +} + +gl::Error TextureStorage11_2D::copyToStorage(TextureStorage *destStorage) +{ + ASSERT(destStorage); + + TextureStorage11_2D *dest11 = TextureStorage11_2D::makeTextureStorage11_2D(destStorage); + + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + + // If either mTexture or mLevelZeroTexture exist, then we need to copy them into the corresponding textures in destStorage. + if (mTexture) + { + gl::Error error = dest11->useLevelZeroWorkaroundTexture(false); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + immediateContext->CopyResource(destResource, mTexture); + } + + if (mLevelZeroTexture) + { + gl::Error error = dest11->useLevelZeroWorkaroundTexture(true); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + immediateContext->CopyResource(destResource, mLevelZeroTexture); + } + } + else + { + ID3D11Resource *sourceResouce = NULL; + gl::Error error = getResource(&sourceResouce); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + immediateContext->CopyResource(destResource, sourceResouce); + } + + dest11->invalidateSwizzleCache(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::useLevelZeroWorkaroundTexture(bool useLevelZeroTexture) +{ + if (useLevelZeroTexture && mMipLevels > 1) + { + if (!mUseLevelZeroTexture && mTexture) + { + gl::Error error = ensureTextureExists(1); + if (error.isError()) + { + return error; + } + + // Pull data back from the mipped texture if necessary. + ASSERT(mLevelZeroTexture); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + context->CopySubresourceRegion(mLevelZeroTexture, 0, 0, 0, 0, mTexture, 0, NULL); + } + + mUseLevelZeroTexture = true; + } + else + { + if (mUseLevelZeroTexture && mLevelZeroTexture) + { + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + // Pull data back from the level zero texture if necessary. + ASSERT(mTexture); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + context->CopySubresourceRegion(mTexture, 0, 0, 0, 0, mLevelZeroTexture, 0, NULL); + } + + mUseLevelZeroTexture = false; + } + + return gl::Error(GL_NO_ERROR); +} + +void TextureStorage11_2D::associateImage(Image11* image, const gl::ImageIndex &index) +{ + GLint level = index.mipIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + mAssociatedImages[level] = image; + } +} + +bool TextureStorage11_2D::isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + // This validation check should never return false. It means the Image/TextureStorage association is broken. + bool retValue = (mAssociatedImages[level] == expectedImage); + ASSERT(retValue); + return retValue; + } + + return false; +} + +// disassociateImage allows an Image to end its association with a Storage. +void TextureStorage11_2D::disassociateImage(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + ASSERT(mAssociatedImages[level] == expectedImage); + + if (mAssociatedImages[level] == expectedImage) + { + mAssociatedImages[level] = NULL; + } + } +} + +// releaseAssociatedImage prepares the Storage for a new Image association. It lets the old Image recover its data before ending the association. +gl::Error TextureStorage11_2D::releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage) +{ + GLint level = index.mipIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + // No need to let the old Image recover its data, if it is also the incoming Image. + if (mAssociatedImages[level] != NULL && mAssociatedImages[level] != incomingImage) + { + // Ensure that the Image is still associated with this TextureStorage. This should be true. + bool imageAssociationCorrect = mAssociatedImages[level]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // Force the image to recover from storage before its data is overwritten. + // This will reset mAssociatedImages[level] to NULL too. + gl::Error error = mAssociatedImages[level]->recoverFromAssociatedStorage(); + if (error.isError()) + { + return error; + } + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::getResource(ID3D11Resource **outResource) +{ + if (mUseLevelZeroTexture && mMipLevels > 1) + { + gl::Error error = ensureTextureExists(1); + if (error.isError()) + { + return error; + } + + *outResource = mLevelZeroTexture; + return gl::Error(GL_NO_ERROR); + } + else + { + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error TextureStorage11_2D::getMippedResource(ID3D11Resource **outResource) +{ + // This shouldn't be called unless the zero max LOD workaround is active. + ASSERT(mRenderer->getWorkarounds().zeroMaxLodWorkaround); + + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::ensureTextureExists(int mipLevels) +{ + // If mMipLevels = 1 then always use mTexture rather than mLevelZeroTexture. + bool useLevelZeroTexture = mRenderer->getWorkarounds().zeroMaxLodWorkaround ? (mipLevels == 1) && (mMipLevels > 1) : false; + ID3D11Texture2D **outputTexture = useLevelZeroTexture ? &mLevelZeroTexture : &mTexture; + + // if the width or height is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (*outputTexture == NULL && mTextureWidth > 0 && mTextureHeight > 0) + { + ASSERT(mipLevels > 0); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; // Compressed texture size constraints? + desc.Height = mTextureHeight; + desc.MipLevels = mipLevels; + desc.ArraySize = 1; + desc.Format = mTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, outputTexture); + + // this can happen from windows TDR + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 2D texture storage, result: 0x%X.", result); + } + else if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 2D texture storage, result: 0x%X.", result); + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT) +{ + ASSERT(!index.hasLayer()); + + int level = index.mipIndex; + ASSERT(level >= 0 && level < getLevelCount()); + + // In GL ES 2.0, the application can only render to level zero of the texture (Section 4.4.3 of the GLES 2.0 spec, page 113 of version 2.0.25). + // Other parts of TextureStorage11_2D could create RTVs on non-zero levels of the texture (e.g. generateMipmap). + // On Feature Level 9_3, this is unlikely to be useful. The renderer can't create SRVs on the individual levels of the texture, + // so methods like generateMipmap can't do anything useful with non-zero-level RTVs. + // Therefore if level > 0 on 9_3 then there's almost certainly something wrong. + ASSERT(!(mRenderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3 && level > 0)); + + if (!mRenderTarget[level]) + { + ID3D11Resource *texture = NULL; + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + + ID3D11ShaderResourceView *srv = NULL; + error = getSRVLevel(level, &srv); + if (error.isError()) + { + return error; + } + + if (mUseLevelZeroTexture) + { + if (!mLevelZeroRenderTarget) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + rtvDesc.Texture2D.MipSlice = mTopLevel + level; + + ID3D11RenderTargetView *rtv; + HRESULT result = device->CreateRenderTargetView(mLevelZeroTexture, &rtvDesc, &rtv); + + if (result == E_OUTOFMEMORY) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + ASSERT(SUCCEEDED(result)); + + mLevelZeroRenderTarget = new TextureRenderTarget11(rtv, mLevelZeroTexture, NULL, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + } + + ASSERT(outRT); + *outRT = mLevelZeroRenderTarget; + return gl::Error(GL_NO_ERROR); + } + + if (mRenderTargetFormat != DXGI_FORMAT_UNKNOWN) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + rtvDesc.Texture2D.MipSlice = mTopLevel + level; + + ID3D11RenderTargetView *rtv; + HRESULT result = device->CreateRenderTargetView(texture, &rtvDesc, &rtv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + + mRenderTarget[level] = new TextureRenderTarget11(rtv, texture, srv, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + } + else if (mDepthStencilFormat != DXGI_FORMAT_UNKNOWN) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = mDepthStencilFormat; + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2D; + dsvDesc.Texture2D.MipSlice = mTopLevel + level; + dsvDesc.Flags = 0; + + ID3D11DepthStencilView *dsv; + HRESULT result = device->CreateDepthStencilView(texture, &dsvDesc, &dsv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY,"Failed to create internal depth stencil view for texture storage, result: 0x%X.", result); + } + + mRenderTarget[level] = new TextureRenderTarget11(dsv, texture, srv, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(dsv); + } + else + { + UNREACHABLE(); + } + } + + ASSERT(outRT); + *outRT = mRenderTarget[level]; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const +{ + ASSERT(outSRV); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = format; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D; + srvDesc.Texture2D.MostDetailedMip = mTopLevel + baseLevel; + srvDesc.Texture2D.MipLevels = mipLevels; + + ID3D11Resource *srvTexture = texture; + + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + ASSERT(mTopLevel == 0); + ASSERT(baseLevel == 0); + // This code also assumes that the incoming texture equals either mLevelZeroTexture or mTexture. + + if (mipLevels == 1 && mMipLevels > 1) + { + // We must use a SRV on the level-zero-only texture. + ASSERT(mLevelZeroTexture != NULL && texture == mLevelZeroTexture); + srvTexture = mLevelZeroTexture; + } + else + { + ASSERT(mipLevels == static_cast<int>(mMipLevels)); + ASSERT(mTexture != NULL && texture == mTexture); + srvTexture = mTexture; + } + } + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result = device->CreateShaderResourceView(srvTexture, &srvDesc, outSRV); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal texture storage SRV, result: 0x%X.", result); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::getSwizzleTexture(ID3D11Resource **outTexture) +{ + ASSERT(outTexture); + + if (!mSwizzleTexture) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.MipLevels = mMipLevels; + desc.ArraySize = 1; + desc.Format = mSwizzleTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mSwizzleTexture); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle texture, result: 0x%X.", result); + } + } + + *outTexture = mSwizzleTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2D::getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV) +{ + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + ASSERT(outRTV); + + if (!mSwizzleRenderTargets[mipLevel]) + { + ID3D11Resource *swizzleTexture = NULL; + gl::Error error = getSwizzleTexture(&swizzleTexture); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mSwizzleRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; + rtvDesc.Texture2D.MipSlice = mTopLevel + mipLevel; + + HRESULT result = device->CreateRenderTargetView(mSwizzleTexture, &rtvDesc, &mSwizzleRenderTargets[mipLevel]); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle render target view, result: 0x%X.", result); + } + } + + *outRTV = mSwizzleRenderTargets[mipLevel]; + return gl::Error(GL_NO_ERROR); +} + +TextureStorage11_Cube::TextureStorage11_Cube(Renderer11 *renderer, GLenum internalformat, bool renderTarget, int size, int levels, bool hintLevelZeroOnly) + : TextureStorage11(renderer, GetTextureBindFlags(internalformat, renderer->getFeatureLevel(), renderTarget)) +{ + mTexture = NULL; + mSwizzleTexture = NULL; + + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + mSwizzleRenderTargets[level] = NULL; + for (unsigned int face = 0; face < CUBE_FACE_COUNT; face++) + { + mAssociatedImages[face][level] = NULL; + mRenderTarget[face][level] = NULL; + } + } + + mLevelZeroTexture = NULL; + mUseLevelZeroTexture = false; + + for (unsigned int face = 0; face < CUBE_FACE_COUNT; face++) + { + mLevelZeroRenderTarget[face] = NULL; + } + + mInternalFormat = internalformat; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalformat, renderer->getFeatureLevel()); + mTextureFormat = formatInfo.texFormat; + mShaderResourceFormat = formatInfo.srvFormat; + mDepthStencilFormat = formatInfo.dsvFormat; + mRenderTargetFormat = formatInfo.rtvFormat; + mSwizzleTextureFormat = formatInfo.swizzleTexFormat; + mSwizzleShaderResourceFormat = formatInfo.swizzleSRVFormat; + mSwizzleRenderTargetFormat = formatInfo.swizzleRTVFormat; + + // adjust size if needed for compressed textures + int height = size; + d3d11::MakeValidSize(false, mTextureFormat, &size, &height, &mTopLevel); + + mMipLevels = mTopLevel + levels; + mTextureWidth = size; + mTextureHeight = size; + mTextureDepth = 1; + + if (hintLevelZeroOnly && levels > 1) + { + //The LevelZeroOnly hint should only be true if the zero max LOD workaround is active. + ASSERT(mRenderer->getWorkarounds().zeroMaxLodWorkaround); + mUseLevelZeroTexture = true; + } + + initializeSerials(getLevelCount() * CUBE_FACE_COUNT, CUBE_FACE_COUNT); +} + +TextureStorage11_Cube::~TextureStorage11_Cube() +{ + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + for (unsigned int face = 0; face < CUBE_FACE_COUNT; face++) + { + if (mAssociatedImages[face][level] != NULL) + { + bool imageAssociationCorrect = mAssociatedImages[face][level]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // We must let the Images recover their data before we delete it from the TextureStorage. + mAssociatedImages[face][level]->recoverFromAssociatedStorage(); + } + } + } + } + + SafeRelease(mTexture); + SafeRelease(mSwizzleTexture); + SafeRelease(mLevelZeroTexture); + + for (unsigned int face = 0; face < CUBE_FACE_COUNT; face++) + { + SafeDelete(mLevelZeroRenderTarget[face]); + } + + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + SafeRelease(mSwizzleRenderTargets[level]); + for (unsigned int face = 0; face < CUBE_FACE_COUNT; face++) + { + SafeDelete(mRenderTarget[face][level]); + } + } +} + +TextureStorage11_Cube *TextureStorage11_Cube::makeTextureStorage11_Cube(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_Cube*, storage)); + return static_cast<TextureStorage11_Cube*>(storage); +} + +UINT TextureStorage11_Cube::getSubresourceIndex(const gl::ImageIndex &index) const +{ + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround && mUseLevelZeroTexture && index.mipIndex == 0) + { + UINT arraySlice = static_cast<UINT>(index.hasLayer() ? index.layerIndex : 0); + UINT subresource = D3D11CalcSubresource(0, arraySlice, 1); + ASSERT(subresource != std::numeric_limits<UINT>::max()); + return subresource; + } + else + { + UINT mipSlice = static_cast<UINT>(index.mipIndex + mTopLevel); + UINT arraySlice = static_cast<UINT>(index.hasLayer() ? index.layerIndex : 0); + UINT subresource = D3D11CalcSubresource(mipSlice, arraySlice, mMipLevels); + ASSERT(subresource != std::numeric_limits<UINT>::max()); + return subresource; + } +} + +gl::Error TextureStorage11_Cube::copyToStorage(TextureStorage *destStorage) +{ + ASSERT(destStorage); + + TextureStorage11_Cube *dest11 = TextureStorage11_Cube::makeTextureStorage11_Cube(destStorage); + + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + + // If either mTexture or mLevelZeroTexture exist, then we need to copy them into the corresponding textures in destStorage. + if (mTexture) + { + gl::Error error = dest11->useLevelZeroWorkaroundTexture(false); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + immediateContext->CopyResource(destResource, mTexture); + } + + if (mLevelZeroTexture) + { + gl::Error error = dest11->useLevelZeroWorkaroundTexture(true); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + immediateContext->CopyResource(destResource, mLevelZeroTexture); + } + } + else + { + ID3D11Resource *sourceResouce = NULL; + gl::Error error = getResource(&sourceResouce); + if (error.isError()) + { + return error; + } + + ID3D11Resource *destResource = NULL; + error = dest11->getResource(&destResource); + if (error.isError()) + { + return error; + } + + ID3D11DeviceContext *immediateContext = mRenderer->getDeviceContext(); + immediateContext->CopyResource(destResource, sourceResouce); + } + + dest11->invalidateSwizzleCache(); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::useLevelZeroWorkaroundTexture(bool useLevelZeroTexture) +{ + if (useLevelZeroTexture && mMipLevels > 1) + { + if (!mUseLevelZeroTexture && mTexture) + { + gl::Error error = ensureTextureExists(1); + if (error.isError()) + { + return error; + } + + // Pull data back from the mipped texture if necessary. + ASSERT(mLevelZeroTexture); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + for (int face = 0; face < 6; face++) + { + context->CopySubresourceRegion(mLevelZeroTexture, D3D11CalcSubresource(0, face, 1), 0, 0, 0, mTexture, face * mMipLevels, NULL); + } + } + + mUseLevelZeroTexture = true; + } + else + { + if (mUseLevelZeroTexture && mLevelZeroTexture) + { + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + // Pull data back from the level zero texture if necessary. + ASSERT(mTexture); + ID3D11DeviceContext *context = mRenderer->getDeviceContext(); + + for (int face = 0; face < 6; face++) + { + context->CopySubresourceRegion(mTexture, D3D11CalcSubresource(0, face, mMipLevels), 0, 0, 0, mLevelZeroTexture, face, NULL); + } + } + + mUseLevelZeroTexture = false; + } + + return gl::Error(GL_NO_ERROR); +} + +void TextureStorage11_Cube::associateImage(Image11* image, const gl::ImageIndex &index) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + ASSERT(0 <= layerTarget && layerTarget < CUBE_FACE_COUNT); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + if (0 <= layerTarget && layerTarget < CUBE_FACE_COUNT) + { + mAssociatedImages[layerTarget][level] = image; + } + } +} + +bool TextureStorage11_Cube::isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + if (0 <= layerTarget && layerTarget < CUBE_FACE_COUNT) + { + // This validation check should never return false. It means the Image/TextureStorage association is broken. + bool retValue = (mAssociatedImages[layerTarget][level] == expectedImage); + ASSERT(retValue); + return retValue; + } + } + + return false; +} + +// disassociateImage allows an Image to end its association with a Storage. +void TextureStorage11_Cube::disassociateImage(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + ASSERT(0 <= layerTarget && layerTarget < CUBE_FACE_COUNT); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + if (0 <= layerTarget && layerTarget < CUBE_FACE_COUNT) + { + ASSERT(mAssociatedImages[layerTarget][level] == expectedImage); + + if (mAssociatedImages[layerTarget][level] == expectedImage) + { + mAssociatedImages[layerTarget][level] = NULL; + } + } + } +} + +// releaseAssociatedImage prepares the Storage for a new Image association. It lets the old Image recover its data before ending the association. +gl::Error TextureStorage11_Cube::releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + ASSERT(0 <= layerTarget && layerTarget < CUBE_FACE_COUNT); + + if ((0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS)) + { + if (0 <= layerTarget && layerTarget < CUBE_FACE_COUNT) + { + // No need to let the old Image recover its data, if it is also the incoming Image. + if (mAssociatedImages[layerTarget][level] != NULL && mAssociatedImages[layerTarget][level] != incomingImage) + { + // Ensure that the Image is still associated with this TextureStorage. This should be true. + bool imageAssociationCorrect = mAssociatedImages[layerTarget][level]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // Force the image to recover from storage before its data is overwritten. + // This will reset mAssociatedImages[level] to NULL too. + gl::Error error = mAssociatedImages[layerTarget][level]->recoverFromAssociatedStorage(); + if (error.isError()) + { + return error; + } + } + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::getResource(ID3D11Resource **outResource) +{ + if (mUseLevelZeroTexture && mMipLevels > 1) + { + gl::Error error = ensureTextureExists(1); + if (error.isError()) + { + return error; + } + + *outResource = mLevelZeroTexture; + return gl::Error(GL_NO_ERROR); + } + else + { + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error TextureStorage11_Cube::getMippedResource(ID3D11Resource **outResource) +{ + // This shouldn't be called unless the zero max LOD workaround is active. + ASSERT(mRenderer->getWorkarounds().zeroMaxLodWorkaround); + + gl::Error error = ensureTextureExists(mMipLevels); + if (error.isError()) + { + return error; + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::ensureTextureExists(int mipLevels) +{ + // If mMipLevels = 1 then always use mTexture rather than mLevelZeroTexture. + bool useLevelZeroTexture = mRenderer->getWorkarounds().zeroMaxLodWorkaround ? (mipLevels == 1) && (mMipLevels > 1) : false; + ID3D11Texture2D **outputTexture = useLevelZeroTexture ? &mLevelZeroTexture : &mTexture; + + // if the size is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (*outputTexture == NULL && mTextureWidth > 0 && mTextureHeight > 0) + { + ASSERT(mMipLevels > 0); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.MipLevels = mipLevels; + desc.ArraySize = CUBE_FACE_COUNT; + desc.Format = mTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = D3D11_RESOURCE_MISC_TEXTURECUBE; + + HRESULT result = device->CreateTexture2D(&desc, NULL, outputTexture); + + // this can happen from windows TDR + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create cube texture storage, result: 0x%X.", result); + } + else if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create cube texture storage, result: 0x%X.", result); + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT) +{ + int faceIndex = index.layerIndex; + int level = index.mipIndex; + + ASSERT(level >= 0 && level < getLevelCount()); + ASSERT(faceIndex >= 0 && faceIndex < CUBE_FACE_COUNT); + + if (!mRenderTarget[faceIndex][level]) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + ID3D11Resource *texture = NULL; + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + + if (mUseLevelZeroTexture) + { + if (!mLevelZeroRenderTarget[faceIndex]) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = mTopLevel + level; + rtvDesc.Texture2DArray.FirstArraySlice = faceIndex; + rtvDesc.Texture2DArray.ArraySize = 1; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(mLevelZeroTexture, &rtvDesc, &rtv); + + if (result == E_OUTOFMEMORY) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + ASSERT(SUCCEEDED(result)); + + mLevelZeroRenderTarget[faceIndex] = new TextureRenderTarget11(rtv, mLevelZeroTexture, NULL, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + } + + ASSERT(outRT); + *outRT = mLevelZeroRenderTarget[faceIndex]; + return gl::Error(GL_NO_ERROR); + } + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.Texture2DArray.MostDetailedMip = mTopLevel + level; + srvDesc.Texture2DArray.MipLevels = 1; + srvDesc.Texture2DArray.FirstArraySlice = faceIndex; + srvDesc.Texture2DArray.ArraySize = 1; + + if (mRenderer->getFeatureLevel() <= D3D_FEATURE_LEVEL_9_3) + { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; + } + else + { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; // Will be used with Texture2D sampler, not TextureCube + } + + ID3D11ShaderResourceView *srv; + result = device->CreateShaderResourceView(texture, &srvDesc, &srv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal shader resource view for texture storage, result: 0x%X.", result); + } + + if (mRenderTargetFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = mTopLevel + level; + rtvDesc.Texture2DArray.FirstArraySlice = faceIndex; + rtvDesc.Texture2DArray.ArraySize = 1; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(texture, &rtvDesc, &rtv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + + mRenderTarget[faceIndex][level] = new TextureRenderTarget11(rtv, texture, srv, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + SafeRelease(srv); + } + else if (mDepthStencilFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_DEPTH_STENCIL_VIEW_DESC dsvDesc; + dsvDesc.Format = mDepthStencilFormat; + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DARRAY; + dsvDesc.Flags = 0; + dsvDesc.Texture2DArray.MipSlice = mTopLevel + level; + dsvDesc.Texture2DArray.FirstArraySlice = faceIndex; + dsvDesc.Texture2DArray.ArraySize = 1; + + ID3D11DepthStencilView *dsv; + result = device->CreateDepthStencilView(texture, &dsvDesc, &dsv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal depth stencil view for texture storage, result: 0x%X.", result); + } + + mRenderTarget[faceIndex][level] = new TextureRenderTarget11(dsv, texture, srv, mInternalFormat, getLevelWidth(level), getLevelHeight(level), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(dsv); + SafeRelease(srv); + } + else + { + UNREACHABLE(); + } + } + + ASSERT(outRT); + *outRT = mRenderTarget[faceIndex][level]; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const +{ + ASSERT(outSRV); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = format; + + // Unnormalized integer cube maps are not supported by DX11; we emulate them as an array of six 2D textures + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(format); + if (dxgiFormatInfo.componentType == GL_INT || dxgiFormatInfo.componentType == GL_UNSIGNED_INT) + { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MostDetailedMip = mTopLevel + baseLevel; + srvDesc.Texture2DArray.MipLevels = 1; + srvDesc.Texture2DArray.FirstArraySlice = 0; + srvDesc.Texture2DArray.ArraySize = CUBE_FACE_COUNT; + } + else + { + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURECUBE; + srvDesc.TextureCube.MipLevels = mipLevels; + srvDesc.TextureCube.MostDetailedMip = mTopLevel + baseLevel; + } + + ID3D11Resource *srvTexture = texture; + + if (mRenderer->getWorkarounds().zeroMaxLodWorkaround) + { + ASSERT(mTopLevel == 0); + ASSERT(baseLevel == 0); + // This code also assumes that the incoming texture equals either mLevelZeroTexture or mTexture. + + if (mipLevels == 1 && mMipLevels > 1) + { + // We must use a SRV on the level-zero-only texture. + ASSERT(mLevelZeroTexture != NULL && texture == mLevelZeroTexture); + srvTexture = mLevelZeroTexture; + } + else + { + ASSERT(mipLevels == static_cast<int>(mMipLevels)); + ASSERT(mTexture != NULL && texture == mTexture); + srvTexture = mTexture; + } + } + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result = device->CreateShaderResourceView(srvTexture, &srvDesc, outSRV); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal texture storage SRV, result: 0x%X.", result); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::getSwizzleTexture(ID3D11Resource **outTexture) +{ + ASSERT(outTexture); + + if (!mSwizzleTexture) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.MipLevels = mMipLevels; + desc.ArraySize = CUBE_FACE_COUNT; + desc.Format = mSwizzleTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + desc.CPUAccessFlags = 0; + desc.MiscFlags = D3D11_RESOURCE_MISC_TEXTURECUBE; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mSwizzleTexture); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle texture, result: 0x%X.", result); + } + } + + *outTexture = mSwizzleTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_Cube::getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV) +{ + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + ASSERT(outRTV); + + if (!mSwizzleRenderTargets[mipLevel]) + { + ID3D11Resource *swizzleTexture = NULL; + gl::Error error = getSwizzleTexture(&swizzleTexture); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mSwizzleRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture2DArray.FirstArraySlice = 0; + rtvDesc.Texture2DArray.ArraySize = CUBE_FACE_COUNT; + + HRESULT result = device->CreateRenderTargetView(mSwizzleTexture, &rtvDesc, &mSwizzleRenderTargets[mipLevel]); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle render target view, result: 0x%X.", result); + } + } + + *outRTV = mSwizzleRenderTargets[mipLevel]; + return gl::Error(GL_NO_ERROR); +} + +TextureStorage11_3D::TextureStorage11_3D(Renderer11 *renderer, GLenum internalformat, bool renderTarget, + GLsizei width, GLsizei height, GLsizei depth, int levels) + : TextureStorage11(renderer, GetTextureBindFlags(internalformat, renderer->getFeatureLevel(), renderTarget)) +{ + mTexture = NULL; + mSwizzleTexture = NULL; + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + mAssociatedImages[i] = NULL; + mLevelRenderTargets[i] = NULL; + mSwizzleRenderTargets[i] = NULL; + } + + mInternalFormat = internalformat; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalformat, renderer->getFeatureLevel()); + mTextureFormat = formatInfo.texFormat; + mShaderResourceFormat = formatInfo.srvFormat; + mDepthStencilFormat = formatInfo.dsvFormat; + mRenderTargetFormat = formatInfo.rtvFormat; + mSwizzleTextureFormat = formatInfo.swizzleTexFormat; + mSwizzleShaderResourceFormat = formatInfo.swizzleSRVFormat; + mSwizzleRenderTargetFormat = formatInfo.swizzleRTVFormat; + + // adjust size if needed for compressed textures + d3d11::MakeValidSize(false, mTextureFormat, &width, &height, &mTopLevel); + + mMipLevels = mTopLevel + levels; + mTextureWidth = width; + mTextureHeight = height; + mTextureDepth = depth; + + initializeSerials(getLevelCount() * depth, depth); +} + +TextureStorage11_3D::~TextureStorage11_3D() +{ + for (unsigned i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + if (mAssociatedImages[i] != NULL) + { + bool imageAssociationCorrect = mAssociatedImages[i]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // We must let the Images recover their data before we delete it from the TextureStorage. + mAssociatedImages[i]->recoverFromAssociatedStorage(); + } + } + } + + SafeRelease(mTexture); + SafeRelease(mSwizzleTexture); + + for (RenderTargetMap::iterator i = mLevelLayerRenderTargets.begin(); i != mLevelLayerRenderTargets.end(); i++) + { + SafeDelete(i->second); + } + mLevelLayerRenderTargets.clear(); + + for (unsigned int i = 0; i < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; i++) + { + SafeDelete(mLevelRenderTargets[i]); + SafeRelease(mSwizzleRenderTargets[i]); + } +} + +TextureStorage11_3D *TextureStorage11_3D::makeTextureStorage11_3D(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_3D*, storage)); + return static_cast<TextureStorage11_3D*>(storage); +} + +void TextureStorage11_3D::associateImage(Image11* image, const gl::ImageIndex &index) +{ + GLint level = index.mipIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + mAssociatedImages[level] = image; + } +} + +bool TextureStorage11_3D::isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + // This validation check should never return false. It means the Image/TextureStorage association is broken. + bool retValue = (mAssociatedImages[level] == expectedImage); + ASSERT(retValue); + return retValue; + } + + return false; +} + +// disassociateImage allows an Image to end its association with a Storage. +void TextureStorage11_3D::disassociateImage(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + + ASSERT(0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + ASSERT(mAssociatedImages[level] == expectedImage); + + if (mAssociatedImages[level] == expectedImage) + { + mAssociatedImages[level] = NULL; + } + } +} + +// releaseAssociatedImage prepares the Storage for a new Image association. It lets the old Image recover its data before ending the association. +gl::Error TextureStorage11_3D::releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage) +{ + GLint level = index.mipIndex; + + ASSERT((0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS)); + + if (0 <= level && level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS) + { + // No need to let the old Image recover its data, if it is also the incoming Image. + if (mAssociatedImages[level] != NULL && mAssociatedImages[level] != incomingImage) + { + // Ensure that the Image is still associated with this TextureStorage. This should be true. + bool imageAssociationCorrect = mAssociatedImages[level]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // Force the image to recover from storage before its data is overwritten. + // This will reset mAssociatedImages[level] to NULL too. + gl::Error error = mAssociatedImages[level]->recoverFromAssociatedStorage(); + if (error.isError()) + { + return error; + } + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_3D::getResource(ID3D11Resource **outResource) +{ + // If the width, height or depth are not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (mTexture == NULL && mTextureWidth > 0 && mTextureHeight > 0 && mTextureDepth > 0) + { + ASSERT(mMipLevels > 0); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE3D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.Depth = mTextureDepth; + desc.MipLevels = mMipLevels; + desc.Format = mTextureFormat; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture3D(&desc, NULL, &mTexture); + + // this can happen from windows TDR + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 3D texture storage, result: 0x%X.", result); + } + else if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 3D texture storage, result: 0x%X.", result); + } + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_3D::createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const +{ + ASSERT(outSRV); + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = format; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE3D; + srvDesc.Texture3D.MostDetailedMip = baseLevel; + srvDesc.Texture3D.MipLevels = mipLevels; + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result = device->CreateShaderResourceView(texture, &srvDesc, outSRV); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal texture storage SRV, result: 0x%X.", result); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_3D::getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT) +{ + int mipLevel = index.mipIndex; + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + + ASSERT(mRenderTargetFormat != DXGI_FORMAT_UNKNOWN); + + if (!index.hasLayer()) + { + if (!mLevelRenderTargets[mipLevel]) + { + ID3D11Resource *texture = NULL; + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + + ID3D11ShaderResourceView *srv = NULL; + error = getSRVLevel(mipLevel, &srv); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE3D; + rtvDesc.Texture3D.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture3D.FirstWSlice = 0; + rtvDesc.Texture3D.WSize = static_cast<UINT>(-1); + + ID3D11RenderTargetView *rtv; + HRESULT result = device->CreateRenderTargetView(texture, &rtvDesc, &rtv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + + mLevelRenderTargets[mipLevel] = new TextureRenderTarget11(rtv, texture, srv, mInternalFormat, getLevelWidth(mipLevel), getLevelHeight(mipLevel), getLevelDepth(mipLevel), 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + } + + ASSERT(outRT); + *outRT = mLevelRenderTargets[mipLevel]; + return gl::Error(GL_NO_ERROR); + } + else + { + int layer = index.layerIndex; + + LevelLayerKey key(mipLevel, layer); + if (mLevelLayerRenderTargets.find(key) == mLevelLayerRenderTargets.end()) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + ID3D11Resource *texture = NULL; + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + + // TODO, what kind of SRV is expected here? + ID3D11ShaderResourceView *srv = NULL; + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE3D; + rtvDesc.Texture3D.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture3D.FirstWSlice = layer; + rtvDesc.Texture3D.WSize = 1; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(texture, &rtvDesc, &rtv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + SafeRelease(srv); return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + ASSERT(SUCCEEDED(result)); + + mLevelLayerRenderTargets[key] = new TextureRenderTarget11(rtv, texture, srv, mInternalFormat, getLevelWidth(mipLevel), getLevelHeight(mipLevel), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + } + + ASSERT(outRT); + *outRT = mLevelLayerRenderTargets[key]; + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error TextureStorage11_3D::getSwizzleTexture(ID3D11Resource **outTexture) +{ + ASSERT(outTexture); + + if (!mSwizzleTexture) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE3D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.Depth = mTextureDepth; + desc.MipLevels = mMipLevels; + desc.Format = mSwizzleTextureFormat; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture3D(&desc, NULL, &mSwizzleTexture); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle texture, result: 0x%X.", result); + } + } + + *outTexture = mSwizzleTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_3D::getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV) +{ + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + ASSERT(outRTV); + + if (!mSwizzleRenderTargets[mipLevel]) + { + ID3D11Resource *swizzleTexture = NULL; + gl::Error error = getSwizzleTexture(&swizzleTexture); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mSwizzleRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE3D; + rtvDesc.Texture3D.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture3D.FirstWSlice = 0; + rtvDesc.Texture3D.WSize = static_cast<UINT>(-1); + + HRESULT result = device->CreateRenderTargetView(mSwizzleTexture, &rtvDesc, &mSwizzleRenderTargets[mipLevel]); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle render target view, result: 0x%X.", result); + } + } + + *outRTV = mSwizzleRenderTargets[mipLevel]; + return gl::Error(GL_NO_ERROR); +} + +TextureStorage11_2DArray::TextureStorage11_2DArray(Renderer11 *renderer, GLenum internalformat, bool renderTarget, + GLsizei width, GLsizei height, GLsizei depth, int levels) + : TextureStorage11(renderer, GetTextureBindFlags(internalformat, renderer->getFeatureLevel(), renderTarget)) +{ + mTexture = NULL; + mSwizzleTexture = NULL; + + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + mSwizzleRenderTargets[level] = NULL; + } + + mInternalFormat = internalformat; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalformat, renderer->getFeatureLevel()); + mTextureFormat = formatInfo.texFormat; + mShaderResourceFormat = formatInfo.srvFormat; + mDepthStencilFormat = formatInfo.dsvFormat; + mRenderTargetFormat = formatInfo.rtvFormat; + mSwizzleTextureFormat = formatInfo.swizzleTexFormat; + mSwizzleShaderResourceFormat = formatInfo.swizzleSRVFormat; + mSwizzleRenderTargetFormat = formatInfo.swizzleRTVFormat; + + // adjust size if needed for compressed textures + d3d11::MakeValidSize(false, mTextureFormat, &width, &height, &mTopLevel); + + mMipLevels = mTopLevel + levels; + mTextureWidth = width; + mTextureHeight = height; + mTextureDepth = depth; + + initializeSerials(getLevelCount() * depth, depth); +} + +TextureStorage11_2DArray::~TextureStorage11_2DArray() +{ + for (ImageMap::iterator i = mAssociatedImages.begin(); i != mAssociatedImages.end(); i++) + { + if (i->second) + { + bool imageAssociationCorrect = i->second->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // We must let the Images recover their data before we delete it from the TextureStorage. + i->second->recoverFromAssociatedStorage(); + } + } + } + mAssociatedImages.clear(); + + SafeRelease(mTexture); + SafeRelease(mSwizzleTexture); + + for (unsigned int level = 0; level < gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS; level++) + { + SafeRelease(mSwizzleRenderTargets[level]); + } + + for (RenderTargetMap::iterator i = mRenderTargets.begin(); i != mRenderTargets.end(); i++) + { + SafeDelete(i->second); + } + mRenderTargets.clear(); +} + +TextureStorage11_2DArray *TextureStorage11_2DArray::makeTextureStorage11_2DArray(TextureStorage *storage) +{ + ASSERT(HAS_DYNAMIC_TYPE(TextureStorage11_2DArray*, storage)); + return static_cast<TextureStorage11_2DArray*>(storage); +} + +void TextureStorage11_2DArray::associateImage(Image11* image, const gl::ImageIndex &index) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + ASSERT(0 <= level && level < getLevelCount()); + + if (0 <= level && level < getLevelCount()) + { + LevelLayerKey key(level, layerTarget); + mAssociatedImages[key] = image; + } +} + +bool TextureStorage11_2DArray::isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + LevelLayerKey key(level, layerTarget); + + // This validation check should never return false. It means the Image/TextureStorage association is broken. + bool retValue = (mAssociatedImages.find(key) != mAssociatedImages.end() && (mAssociatedImages[key] == expectedImage)); + ASSERT(retValue); + return retValue; +} + +// disassociateImage allows an Image to end its association with a Storage. +void TextureStorage11_2DArray::disassociateImage(const gl::ImageIndex &index, Image11* expectedImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + LevelLayerKey key(level, layerTarget); + + bool imageAssociationCorrect = (mAssociatedImages.find(key) != mAssociatedImages.end() && (mAssociatedImages[key] == expectedImage)); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + mAssociatedImages[key] = NULL; + } +} + +// releaseAssociatedImage prepares the Storage for a new Image association. It lets the old Image recover its data before ending the association. +gl::Error TextureStorage11_2DArray::releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage) +{ + GLint level = index.mipIndex; + GLint layerTarget = index.layerIndex; + + LevelLayerKey key(level, layerTarget); + + if (mAssociatedImages.find(key) != mAssociatedImages.end()) + { + if (mAssociatedImages[key] != NULL && mAssociatedImages[key] != incomingImage) + { + // Ensure that the Image is still associated with this TextureStorage. This should be true. + bool imageAssociationCorrect = mAssociatedImages[key]->isAssociatedStorageValid(this); + ASSERT(imageAssociationCorrect); + + if (imageAssociationCorrect) + { + // Force the image to recover from storage before its data is overwritten. + // This will reset mAssociatedImages[level] to NULL too. + gl::Error error = mAssociatedImages[key]->recoverFromAssociatedStorage(); + if (error.isError()) + { + return error; + } + } + } + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2DArray::getResource(ID3D11Resource **outResource) +{ + // if the width, height or depth is not positive this should be treated as an incomplete texture + // we handle that here by skipping the d3d texture creation + if (mTexture == NULL && mTextureWidth > 0 && mTextureHeight > 0 && mTextureDepth > 0) + { + ASSERT(mMipLevels > 0); + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.MipLevels = mMipLevels; + desc.ArraySize = mTextureDepth; + desc.Format = mTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = getBindFlags(); + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mTexture); + + // this can happen from windows TDR + if (d3d11::isDeviceLostError(result)) + { + mRenderer->notifyDeviceLost(); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 2D array texture storage, result: 0x%X.", result); + } + else if (FAILED(result)) + { + ASSERT(result == E_OUTOFMEMORY); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create 2D array texture storage, result: 0x%X.", result); + } + } + + *outResource = mTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2DArray::createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const +{ + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = format; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MostDetailedMip = mTopLevel + baseLevel; + srvDesc.Texture2DArray.MipLevels = mipLevels; + srvDesc.Texture2DArray.FirstArraySlice = 0; + srvDesc.Texture2DArray.ArraySize = mTextureDepth; + + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result = device->CreateShaderResourceView(texture, &srvDesc, outSRV); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal texture storage SRV, result: 0x%X.", result); + } + + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2DArray::getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT) +{ + ASSERT(index.hasLayer()); + + int mipLevel = index.mipIndex; + int layer = index.layerIndex; + + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + + LevelLayerKey key(mipLevel, layer); + if (mRenderTargets.find(key) == mRenderTargets.end()) + { + ID3D11Device *device = mRenderer->getDevice(); + HRESULT result; + + ID3D11Resource *texture = NULL; + gl::Error error = getResource(&texture); + if (error.isError()) + { + return error; + } + + D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc; + srvDesc.Format = mShaderResourceFormat; + srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; + srvDesc.Texture2DArray.MostDetailedMip = mTopLevel + mipLevel; + srvDesc.Texture2DArray.MipLevels = 1; + srvDesc.Texture2DArray.FirstArraySlice = layer; + srvDesc.Texture2DArray.ArraySize = 1; + + ID3D11ShaderResourceView *srv; + result = device->CreateShaderResourceView(texture, &srvDesc, &srv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal shader resource view for texture storage, result: 0x%X.", result); + } + + if (mRenderTargetFormat != DXGI_FORMAT_UNKNOWN) + { + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture2DArray.FirstArraySlice = layer; + rtvDesc.Texture2DArray.ArraySize = 1; + + ID3D11RenderTargetView *rtv; + result = device->CreateRenderTargetView(texture, &rtvDesc, &rtv); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + SafeRelease(srv); + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal render target view for texture storage, result: 0x%X.", result); + } + + mRenderTargets[key] = new TextureRenderTarget11(rtv, texture, srv, mInternalFormat, getLevelWidth(mipLevel), getLevelHeight(mipLevel), 1, 0); + + // RenderTarget will take ownership of these resources + SafeRelease(rtv); + SafeRelease(srv); + } + else + { + UNREACHABLE(); + } + } + + ASSERT(outRT); + *outRT = mRenderTargets[key]; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2DArray::getSwizzleTexture(ID3D11Resource **outTexture) +{ + if (!mSwizzleTexture) + { + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_TEXTURE2D_DESC desc; + desc.Width = mTextureWidth; + desc.Height = mTextureHeight; + desc.MipLevels = mMipLevels; + desc.ArraySize = mTextureDepth; + desc.Format = mSwizzleTextureFormat; + desc.SampleDesc.Count = 1; + desc.SampleDesc.Quality = 0; + desc.Usage = D3D11_USAGE_DEFAULT; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + desc.CPUAccessFlags = 0; + desc.MiscFlags = 0; + + HRESULT result = device->CreateTexture2D(&desc, NULL, &mSwizzleTexture); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle texture, result: 0x%X.", result); + } + } + + *outTexture = mSwizzleTexture; + return gl::Error(GL_NO_ERROR); +} + +gl::Error TextureStorage11_2DArray::getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV) +{ + ASSERT(mipLevel >= 0 && mipLevel < getLevelCount()); + ASSERT(outRTV); + + if (!mSwizzleRenderTargets[mipLevel]) + { + ID3D11Resource *swizzleTexture = NULL; + gl::Error error = getSwizzleTexture(&swizzleTexture); + if (error.isError()) + { + return error; + } + + ID3D11Device *device = mRenderer->getDevice(); + + D3D11_RENDER_TARGET_VIEW_DESC rtvDesc; + rtvDesc.Format = mSwizzleRenderTargetFormat; + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.MipSlice = mTopLevel + mipLevel; + rtvDesc.Texture2DArray.FirstArraySlice = 0; + rtvDesc.Texture2DArray.ArraySize = mTextureDepth; + + HRESULT result = device->CreateRenderTargetView(mSwizzleTexture, &rtvDesc, &mSwizzleRenderTargets[mipLevel]); + + ASSERT(result == E_OUTOFMEMORY || SUCCEEDED(result)); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to create internal swizzle render target view, result: 0x%X.", result); + } + } + + *outRTV = mSwizzleRenderTargets[mipLevel]; + return gl::Error(GL_NO_ERROR); +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.h new file mode 100644 index 0000000000..456e2660f9 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/TextureStorage11.h @@ -0,0 +1,326 @@ +// +// Copyright (c) 2012-2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// TextureStorage11.h: Defines the abstract rx::TextureStorage11 class and its concrete derived +// classes TextureStorage11_2D and TextureStorage11_Cube, which act as the interface to the D3D11 texture. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_TEXTURESTORAGE11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_TEXTURESTORAGE11_H_ + +#include "libANGLE/Texture.h" +#include "libANGLE/Error.h" +#include "libANGLE/renderer/d3d/TextureStorage.h" + +#include <map> + +namespace gl +{ +struct ImageIndex; +} + +namespace rx +{ +class RenderTargetD3D; +class RenderTarget11; +class Renderer11; +class SwapChain11; +class Image11; + +class TextureStorage11 : public TextureStorage +{ + public: + virtual ~TextureStorage11(); + + static TextureStorage11 *makeTextureStorage11(TextureStorage *storage); + + static DWORD GetTextureBindFlags(GLenum internalFormat, D3D_FEATURE_LEVEL featureLevel, bool renderTarget); + + UINT getBindFlags() const; + + virtual gl::Error getResource(ID3D11Resource **outResource) = 0; + virtual gl::Error getSRV(const gl::SamplerState &samplerState, ID3D11ShaderResourceView **outSRV); + virtual gl::Error getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT) = 0; + + virtual gl::Error generateMipmap(const gl::ImageIndex &sourceIndex, const gl::ImageIndex &destIndex); + + virtual int getTopLevel() const; + virtual bool isRenderTarget() const; + virtual bool isManaged() const; + virtual int getLevelCount() const; + virtual UINT getSubresourceIndex(const gl::ImageIndex &index) const; + + gl::Error generateSwizzles(GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha); + void invalidateSwizzleCacheLevel(int mipLevel); + void invalidateSwizzleCache(); + + gl::Error updateSubresourceLevel(ID3D11Resource *texture, unsigned int sourceSubresource, + const gl::ImageIndex &index, const gl::Box ©Area); + + gl::Error copySubresourceLevel(ID3D11Resource* dstTexture, unsigned int dstSubresource, + const gl::ImageIndex &index, const gl::Box ®ion); + + virtual void associateImage(Image11* image, const gl::ImageIndex &index) = 0; + virtual void disassociateImage(const gl::ImageIndex &index, Image11* expectedImage) = 0; + virtual bool isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage) = 0; + virtual gl::Error releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage) = 0; + + virtual gl::Error copyToStorage(TextureStorage *destStorage); + virtual gl::Error setData(const gl::ImageIndex &index, ImageD3D *image, const gl::Box *destBox, GLenum type, + const gl::PixelUnpackState &unpack, const uint8_t *pixelData); + + protected: + TextureStorage11(Renderer11 *renderer, UINT bindFlags); + int getLevelWidth(int mipLevel) const; + int getLevelHeight(int mipLevel) const; + int getLevelDepth(int mipLevel) const; + + // Some classes (e.g. TextureStorage11_2D) will override getMippedResource. + virtual gl::Error getMippedResource(ID3D11Resource **outResource) { return getResource(outResource); } + + virtual gl::Error getSwizzleTexture(ID3D11Resource **outTexture) = 0; + virtual gl::Error getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV) = 0; + gl::Error getSRVLevel(int mipLevel, ID3D11ShaderResourceView **outSRV); + + virtual gl::Error createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const = 0; + + void verifySwizzleExists(GLenum swizzleRed, GLenum swizzleGreen, GLenum swizzleBlue, GLenum swizzleAlpha); + + Renderer11 *mRenderer; + int mTopLevel; + unsigned int mMipLevels; + + GLenum mInternalFormat; + DXGI_FORMAT mTextureFormat; + DXGI_FORMAT mShaderResourceFormat; + DXGI_FORMAT mRenderTargetFormat; + DXGI_FORMAT mDepthStencilFormat; + DXGI_FORMAT mSwizzleTextureFormat; + DXGI_FORMAT mSwizzleShaderResourceFormat; + DXGI_FORMAT mSwizzleRenderTargetFormat; + unsigned int mTextureWidth; + unsigned int mTextureHeight; + unsigned int mTextureDepth; + + struct SwizzleCacheValue + { + GLenum swizzleRed; + GLenum swizzleGreen; + GLenum swizzleBlue; + GLenum swizzleAlpha; + + SwizzleCacheValue(); + SwizzleCacheValue(GLenum red, GLenum green, GLenum blue, GLenum alpha); + + bool operator ==(const SwizzleCacheValue &other) const; + bool operator !=(const SwizzleCacheValue &other) const; + }; + SwizzleCacheValue mSwizzleCache[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + private: + const UINT mBindFlags; + + struct SRVKey + { + SRVKey(int baseLevel = 0, int mipLevels = 0, bool swizzle = false); + + bool operator<(const SRVKey &rhs) const; + + int baseLevel; + int mipLevels; + bool swizzle; + }; + typedef std::map<SRVKey, ID3D11ShaderResourceView *> SRVCache; + + SRVCache mSrvCache; + ID3D11ShaderResourceView *mLevelSRVs[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +class TextureStorage11_2D : public TextureStorage11 +{ + public: + TextureStorage11_2D(Renderer11 *renderer, SwapChain11 *swapchain); + TextureStorage11_2D(Renderer11 *renderer, GLenum internalformat, bool renderTarget, GLsizei width, GLsizei height, int levels, bool hintLevelZeroOnly = false); + virtual ~TextureStorage11_2D(); + + static TextureStorage11_2D *makeTextureStorage11_2D(TextureStorage *storage); + + virtual gl::Error getResource(ID3D11Resource **outResource); + virtual gl::Error getMippedResource(ID3D11Resource **outResource); + virtual gl::Error getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT); + + virtual gl::Error copyToStorage(TextureStorage *destStorage); + + virtual void associateImage(Image11* image, const gl::ImageIndex &index); + virtual void disassociateImage(const gl::ImageIndex &index, Image11* expectedImage); + virtual bool isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage); + virtual gl::Error releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage); + + virtual gl::Error useLevelZeroWorkaroundTexture(bool useLevelZeroTexture); + + protected: + virtual gl::Error getSwizzleTexture(ID3D11Resource **outTexture); + virtual gl::Error getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV); + + gl::Error ensureTextureExists(int mipLevels); + + private: + virtual gl::Error createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const; + + ID3D11Texture2D *mTexture; + RenderTarget11 *mRenderTarget[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + // These are members related to the zero max-LOD workaround. + // D3D11 Feature Level 9_3 can't disable mipmaps on a mipmapped texture (i.e. solely sample from level zero). + // These members are used to work around this limitation. + // Usually only mTexture XOR mLevelZeroTexture will exist. + // For example, if an app creates a texture with only one level, then 9_3 will only create mLevelZeroTexture. + // However, in some scenarios, both textures have to be created. This incurs additional memory overhead. + // One example of this is an application that creates a texture, calls glGenerateMipmap, and then disables mipmaps on the texture. + // A more likely example is an app that creates an empty texture, renders to it, and then calls glGenerateMipmap + // TODO: In this rendering scenario, release the mLevelZeroTexture after mTexture has been created to save memory. + ID3D11Texture2D *mLevelZeroTexture; + RenderTarget11 *mLevelZeroRenderTarget; + bool mUseLevelZeroTexture; + + // Swizzle-related variables + ID3D11Texture2D *mSwizzleTexture; + ID3D11RenderTargetView *mSwizzleRenderTargets[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + Image11 *mAssociatedImages[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +class TextureStorage11_Cube : public TextureStorage11 +{ + public: + TextureStorage11_Cube(Renderer11 *renderer, GLenum internalformat, bool renderTarget, int size, int levels, bool hintLevelZeroOnly); + virtual ~TextureStorage11_Cube(); + + static TextureStorage11_Cube *makeTextureStorage11_Cube(TextureStorage *storage); + + virtual UINT getSubresourceIndex(const gl::ImageIndex &index) const; + + virtual gl::Error getResource(ID3D11Resource **outResource); + virtual gl::Error getMippedResource(ID3D11Resource **outResource); + virtual gl::Error getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT); + + virtual gl::Error copyToStorage(TextureStorage *destStorage); + + virtual void associateImage(Image11* image, const gl::ImageIndex &index); + virtual void disassociateImage(const gl::ImageIndex &index, Image11* expectedImage); + virtual bool isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage); + virtual gl::Error releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage); + + virtual gl::Error useLevelZeroWorkaroundTexture(bool useLevelZeroTexture); + + protected: + virtual gl::Error getSwizzleTexture(ID3D11Resource **outTexture); + virtual gl::Error getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV); + + gl::Error ensureTextureExists(int mipLevels); + + private: + virtual gl::Error createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const; + + static const size_t CUBE_FACE_COUNT = 6; + + ID3D11Texture2D *mTexture; + RenderTarget11 *mRenderTarget[CUBE_FACE_COUNT][gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + // Level-zero workaround members. See TextureStorage11_2D's workaround members for a description. + ID3D11Texture2D *mLevelZeroTexture; + RenderTarget11 *mLevelZeroRenderTarget[CUBE_FACE_COUNT]; + bool mUseLevelZeroTexture; + + ID3D11Texture2D *mSwizzleTexture; + ID3D11RenderTargetView *mSwizzleRenderTargets[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + Image11 *mAssociatedImages[CUBE_FACE_COUNT][gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +class TextureStorage11_3D : public TextureStorage11 +{ + public: + TextureStorage11_3D(Renderer11 *renderer, GLenum internalformat, bool renderTarget, + GLsizei width, GLsizei height, GLsizei depth, int levels); + virtual ~TextureStorage11_3D(); + + static TextureStorage11_3D *makeTextureStorage11_3D(TextureStorage *storage); + + virtual gl::Error getResource(ID3D11Resource **outResource); + + // Handles both layer and non-layer RTs + virtual gl::Error getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT); + + virtual void associateImage(Image11* image, const gl::ImageIndex &index); + virtual void disassociateImage(const gl::ImageIndex &index, Image11* expectedImage); + virtual bool isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage); + virtual gl::Error releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage); + + protected: + virtual gl::Error getSwizzleTexture(ID3D11Resource **outTexture); + virtual gl::Error getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV); + + private: + virtual gl::Error createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const; + + typedef std::pair<int, int> LevelLayerKey; + typedef std::map<LevelLayerKey, RenderTarget11*> RenderTargetMap; + RenderTargetMap mLevelLayerRenderTargets; + + RenderTarget11 *mLevelRenderTargets[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + ID3D11Texture3D *mTexture; + ID3D11Texture3D *mSwizzleTexture; + ID3D11RenderTargetView *mSwizzleRenderTargets[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + Image11 *mAssociatedImages[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; +}; + +class TextureStorage11_2DArray : public TextureStorage11 +{ + public: + TextureStorage11_2DArray(Renderer11 *renderer, GLenum internalformat, bool renderTarget, + GLsizei width, GLsizei height, GLsizei depth, int levels); + virtual ~TextureStorage11_2DArray(); + + static TextureStorage11_2DArray *makeTextureStorage11_2DArray(TextureStorage *storage); + + virtual gl::Error getResource(ID3D11Resource **outResource); + virtual gl::Error getRenderTarget(const gl::ImageIndex &index, RenderTargetD3D **outRT); + + virtual void associateImage(Image11* image, const gl::ImageIndex &index); + virtual void disassociateImage(const gl::ImageIndex &index, Image11* expectedImage); + virtual bool isAssociatedImageValid(const gl::ImageIndex &index, Image11* expectedImage); + virtual gl::Error releaseAssociatedImage(const gl::ImageIndex &index, Image11* incomingImage); + + protected: + virtual gl::Error getSwizzleTexture(ID3D11Resource **outTexture); + virtual gl::Error getSwizzleRenderTarget(int mipLevel, ID3D11RenderTargetView **outRTV); + + private: + virtual gl::Error createSRV(int baseLevel, int mipLevels, DXGI_FORMAT format, ID3D11Resource *texture, + ID3D11ShaderResourceView **outSRV) const; + + typedef std::pair<int, int> LevelLayerKey; + typedef std::map<LevelLayerKey, RenderTarget11*> RenderTargetMap; + RenderTargetMap mRenderTargets; + + ID3D11Texture2D *mTexture; + + ID3D11Texture2D *mSwizzleTexture; + ID3D11RenderTargetView *mSwizzleRenderTargets[gl::IMPLEMENTATION_MAX_TEXTURE_LEVELS]; + + typedef std::map<LevelLayerKey, Image11*> ImageMap; + ImageMap mAssociatedImages; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_TEXTURESTORAGE11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.cpp new file mode 100644 index 0000000000..213ce31817 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.cpp @@ -0,0 +1,95 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Trim11.cpp: Trim support utility class. + +#include "libANGLE/renderer/d3d/d3d11/Trim11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" + +#if defined (ANGLE_ENABLE_WINDOWS_STORE) +using namespace ABI::Windows::Foundation; +using namespace ABI::Windows::ApplicationModel; +using namespace ABI::Windows::ApplicationModel::Core; +#endif + +namespace rx +{ + +Trim11::Trim11(rx::Renderer11 *renderer) + : mRenderer(renderer) +{ + bool result = true; + result = registerForRendererTrimRequest(); + ASSERT(result); +} + +Trim11::~Trim11() +{ + unregisterForRendererTrimRequest(); +} + +void Trim11::trim() +{ + if (!mRenderer) + { + return; + } + +#if defined (ANGLE_ENABLE_WINDOWS_STORE) + ID3D11Device* device = mRenderer->getDevice(); + // IDXGIDevice3 is only supported on Windows 8.1 and Windows Phone 8.1 and above. + IDXGIDevice3 *dxgiDevice3 = d3d11::DynamicCastComObject<IDXGIDevice3>(device); + if (dxgiDevice3) + { + dxgiDevice3->Trim(); + } + SafeRelease(dxgiDevice3); +#endif +} + +bool Trim11::registerForRendererTrimRequest() +{ +#if defined (ANGLE_ENABLE_WINDOWS_STORE) + ICoreApplication* coreApplication = nullptr; + HRESULT result = GetActivationFactory(HStringReference(RuntimeClass_Windows_ApplicationModel_Core_CoreApplication).Get(), &coreApplication); + if (SUCCEEDED(result)) + { + auto suspendHandler = Callback<IEventHandler<SuspendingEventArgs*>>( + [this](IInspectable*, ISuspendingEventArgs*) -> HRESULT + { + trim(); + return S_OK; + }); + result = coreApplication->add_Suspending(suspendHandler.Get(), &mApplicationSuspendedEventToken); + } + SafeRelease(coreApplication); + + if (FAILED(result)) + { + return false; + } +#endif + return true; +} + +void Trim11::unregisterForRendererTrimRequest() +{ +#if defined (ANGLE_ENABLE_WINDOWS_STORE) + if (mApplicationSuspendedEventToken.value != 0) + { + ICoreApplication* coreApplication = nullptr; + if (SUCCEEDED(GetActivationFactory(HStringReference(RuntimeClass_Windows_ApplicationModel_Core_CoreApplication).Get(), &coreApplication))) + { + coreApplication->remove_Suspending(mApplicationSuspendedEventToken); + } + mApplicationSuspendedEventToken.value = 0; + SafeRelease(coreApplication); + } +#endif +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.h new file mode 100644 index 0000000000..f232ad7e8e --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/Trim11.h @@ -0,0 +1,43 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// Trim11.h: Trim support utility class. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_TRIM11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_TRIM11_H_ + +#include "common/angleutils.h" +#include "libANGLE/angletypes.h" +#include "libANGLE/Error.h" + +#if !defined(ANGLE_ENABLE_WINDOWS_STORE) +typedef void* EventRegistrationToken; +#else +#include <EventToken.h> +#endif + +namespace rx +{ +class Renderer11; + +class Trim11 : angle::NonCopyable +{ + public: + explicit Trim11(Renderer11 *renderer); + ~Trim11(); + + private: + Renderer11 *mRenderer; + EventRegistrationToken mApplicationSuspendedEventToken; + + void trim(); + bool registerForRendererTrimRequest(); + void unregisterForRendererTrimRequest(); +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_TRIM11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexArray11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexArray11.h new file mode 100644 index 0000000000..78aad7d106 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexArray11.h @@ -0,0 +1,40 @@ +// +// Copyright 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexArray11.h: Defines the rx::VertexArray11 class which implements rx::VertexArrayImpl. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_VERTEXARRAY11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_VERTEXARRAY11_H_ + +#include "libANGLE/renderer/VertexArrayImpl.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" + +namespace rx +{ +class Renderer11; + +class VertexArray11 : public VertexArrayImpl +{ + public: + VertexArray11(Renderer11 *renderer) + : VertexArrayImpl(), + mRenderer(renderer) + { + } + virtual ~VertexArray11() { } + + virtual void setElementArrayBuffer(const gl::Buffer *buffer) { } + virtual void setAttribute(size_t idx, const gl::VertexAttribute &attr) { } + virtual void setAttributeDivisor(size_t idx, GLuint divisor) { } + virtual void enableAttribute(size_t idx, bool enabledState) { } + + private: + Renderer11 *mRenderer; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_VERTEXARRAY11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.cpp new file mode 100644 index 0000000000..adc64cef5e --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.cpp @@ -0,0 +1,244 @@ +// +// Copyright (c) 2013 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer11.cpp: Defines the D3D11 VertexBuffer implementation. + +#include "libANGLE/renderer/d3d/d3d11/VertexBuffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Buffer11.h" +#include "libANGLE/renderer/d3d/d3d11/Renderer11.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/Buffer.h" +#include "libANGLE/VertexAttribute.h" + +namespace rx +{ + +VertexBuffer11::VertexBuffer11(Renderer11 *const renderer) : mRenderer(renderer) +{ + mBuffer = NULL; + mBufferSize = 0; + mDynamicUsage = false; + mMappedResourceData = NULL; +} + +VertexBuffer11::~VertexBuffer11() +{ + ASSERT(mMappedResourceData == NULL); + SafeRelease(mBuffer); +} + +gl::Error VertexBuffer11::initialize(unsigned int size, bool dynamicUsage) +{ + SafeRelease(mBuffer); + + updateSerial(); + + if (size > 0) + { + ID3D11Device* dxDevice = mRenderer->getDevice(); + + D3D11_BUFFER_DESC bufferDesc; + bufferDesc.ByteWidth = size; + bufferDesc.Usage = D3D11_USAGE_DYNAMIC; + bufferDesc.BindFlags = D3D11_BIND_VERTEX_BUFFER; + bufferDesc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE; + bufferDesc.MiscFlags = 0; + bufferDesc.StructureByteStride = 0; + + HRESULT result = dxDevice->CreateBuffer(&bufferDesc, NULL, &mBuffer); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to allocate internal vertex buffer of size, %lu.", size); + } + } + + mBufferSize = size; + mDynamicUsage = dynamicUsage; + + return gl::Error(GL_NO_ERROR); +} + +VertexBuffer11 *VertexBuffer11::makeVertexBuffer11(VertexBuffer *vetexBuffer) +{ + ASSERT(HAS_DYNAMIC_TYPE(VertexBuffer11*, vetexBuffer)); + return static_cast<VertexBuffer11*>(vetexBuffer); +} + +gl::Error VertexBuffer11::mapResource() +{ + if (mMappedResourceData == NULL) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_NO_OVERWRITE, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal vertex buffer, HRESULT: 0x%08x.", result); + } + + mMappedResourceData = reinterpret_cast<uint8_t*>(mappedResource.pData); + } + + return gl::Error(GL_NO_ERROR); +} + +void VertexBuffer11::hintUnmapResource() +{ + if (mMappedResourceData != NULL) + { + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + dxContext->Unmap(mBuffer, 0); + + mMappedResourceData = NULL; + } +} + +gl::Error VertexBuffer11::storeVertexAttributes(const gl::VertexAttribute &attrib, const gl::VertexAttribCurrentValueData ¤tValue, + GLint start, GLsizei count, GLsizei instances, unsigned int offset) +{ + if (!mBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal vertex buffer is not initialized."); + } + + gl::Buffer *buffer = attrib.buffer.get(); + int inputStride = ComputeVertexAttributeStride(attrib); + + // This will map the resource if it isn't already mapped. + gl::Error error = mapResource(); + if (error.isError()) + { + return error; + } + + uint8_t *output = mMappedResourceData + offset; + + const uint8_t *input = NULL; + if (attrib.enabled) + { + if (buffer) + { + BufferD3D *storage = GetImplAs<BufferD3D>(buffer); + error = storage->getData(&input); + if (error.isError()) + { + return error; + } + input += static_cast<int>(attrib.offset); + } + else + { + input = static_cast<const uint8_t*>(attrib.pointer); + } + } + else + { + input = reinterpret_cast<const uint8_t*>(currentValue.FloatValues); + } + + if (instances == 0 || attrib.divisor == 0) + { + input += inputStride * start; + } + + gl::VertexFormat vertexFormat(attrib, currentValue.Type); + const d3d11::VertexFormat &vertexFormatInfo = d3d11::GetVertexFormatInfo(vertexFormat, mRenderer->getFeatureLevel()); + ASSERT(vertexFormatInfo.copyFunction != NULL); + vertexFormatInfo.copyFunction(input, inputStride, count, output); + + return gl::Error(GL_NO_ERROR); +} + +gl::Error VertexBuffer11::getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, + GLsizei instances, unsigned int *outSpaceRequired) const +{ + unsigned int elementCount = 0; + if (attrib.enabled) + { + if (instances == 0 || attrib.divisor == 0) + { + elementCount = count; + } + else + { + // Round up to divisor, if possible + elementCount = UnsignedCeilDivide(static_cast<unsigned int>(instances), attrib.divisor); + } + + gl::VertexFormat vertexFormat(attrib); + const d3d11::VertexFormat &vertexFormatInfo = d3d11::GetVertexFormatInfo(vertexFormat, mRenderer->getFeatureLevel()); + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(vertexFormatInfo.nativeFormat); + unsigned int elementSize = dxgiFormatInfo.pixelBytes; + if (elementSize <= std::numeric_limits<unsigned int>::max() / elementCount) + { + if (outSpaceRequired) + { + *outSpaceRequired = elementSize * elementCount; + } + return gl::Error(GL_NO_ERROR); + } + else + { + return gl::Error(GL_OUT_OF_MEMORY, "New vertex buffer size would result in an overflow."); + } + } + else + { + const unsigned int elementSize = 4; + if (outSpaceRequired) + { + *outSpaceRequired = elementSize * 4; + } + return gl::Error(GL_NO_ERROR); + } +} + +unsigned int VertexBuffer11::getBufferSize() const +{ + return mBufferSize; +} + +gl::Error VertexBuffer11::setBufferSize(unsigned int size) +{ + if (size > mBufferSize) + { + return initialize(size, mDynamicUsage); + } + else + { + return gl::Error(GL_NO_ERROR); + } +} + +gl::Error VertexBuffer11::discard() +{ + if (!mBuffer) + { + return gl::Error(GL_OUT_OF_MEMORY, "Internal vertex buffer is not initialized."); + } + + ID3D11DeviceContext *dxContext = mRenderer->getDeviceContext(); + + D3D11_MAPPED_SUBRESOURCE mappedResource; + HRESULT result = dxContext->Map(mBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + if (FAILED(result)) + { + return gl::Error(GL_OUT_OF_MEMORY, "Failed to map internal buffer for discarding, HRESULT: 0x%08x", result); + } + + dxContext->Unmap(mBuffer, 0); + + return gl::Error(GL_NO_ERROR); +} + +ID3D11Buffer *VertexBuffer11::getBuffer() const +{ + return mBuffer; +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.h new file mode 100644 index 0000000000..2450e8955c --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/VertexBuffer11.h @@ -0,0 +1,58 @@ +// +// Copyright (c) 2012 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// VertexBuffer11.h: Defines the D3D11 VertexBuffer implementation. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_VERTEXBUFFER11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_VERTEXBUFFER11_H_ + +#include <stdint.h> + +#include "libANGLE/renderer/d3d/VertexBuffer.h" + +namespace rx +{ +class Renderer11; + +class VertexBuffer11 : public VertexBuffer +{ + public: + explicit VertexBuffer11(Renderer11 *const renderer); + virtual ~VertexBuffer11(); + + virtual gl::Error initialize(unsigned int size, bool dynamicUsage); + + static VertexBuffer11 *makeVertexBuffer11(VertexBuffer *vetexBuffer); + + virtual gl::Error storeVertexAttributes(const gl::VertexAttribute &attrib, const gl::VertexAttribCurrentValueData ¤tValue, + GLint start, GLsizei count, GLsizei instances, unsigned int offset); + + virtual gl::Error getSpaceRequired(const gl::VertexAttribute &attrib, GLsizei count, GLsizei instances, + unsigned int *outSpaceRequired) const; + + virtual unsigned int getBufferSize() const; + virtual gl::Error setBufferSize(unsigned int size); + virtual gl::Error discard(); + + virtual void hintUnmapResource(); + + ID3D11Buffer *getBuffer() const; + + private: + gl::Error mapResource(); + + Renderer11 *const mRenderer; + + ID3D11Buffer *mBuffer; + unsigned int mBufferSize; + bool mDynamicUsage; + + uint8_t *mMappedResourceData; +}; + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_VERTEXBUFFER11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.h new file mode 100644 index 0000000000..5501e361f1 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.h @@ -0,0 +1,40 @@ +// +// Copyright (c) 2013-2015 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// copyvertex.h: Defines D3D11 vertex buffer copying and conversion functions + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_COPYVERTEX_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_COPYVERTEX_H_ + +#include "common/mathutil.h" + +namespace rx +{ + +// 'alphaDefaultValueBits' gives the default value for the alpha channel (4th component) +template <typename T, size_t inputComponentCount, size_t outputComponentCount, uint32_t alphaDefaultValueBits> +inline void CopyNativeVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +template <size_t inputComponentCount, size_t outputComponentCount> +inline void Copy8SintTo16SintVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +template <size_t componentCount> +inline void Copy8SnormTo16SnormVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +template <size_t inputComponentCount, size_t outputComponentCount> +inline void Copy32FixedTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +template <typename T, size_t inputComponentCount, size_t outputComponentCount, bool normalized> +inline void CopyTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +template <bool isSigned, bool normalized, bool toFloat> +inline void CopyXYZ10W2ToXYZW32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output); + +} + +#include "copyvertex.inl" + +#endif // LIBANGLE_RENDERER_D3D_D3D11_COPYVERTEX_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.inl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.inl new file mode 100644 index 0000000000..60678d7b9f --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/copyvertex.inl @@ -0,0 +1,377 @@ +// +// Copyright (c) 2014-2015 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +namespace rx +{ + +template <typename T, size_t inputComponentCount, size_t outputComponentCount, uint32_t alphaDefaultValueBits> +inline void CopyNativeVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + const size_t attribSize = sizeof(T)* inputComponentCount; + + if (attribSize == stride && inputComponentCount == outputComponentCount) + { + memcpy(output, input, count * attribSize); + } + else + { + const T defaultAlphaValue = gl::bitCast<T>(alphaDefaultValueBits); + const size_t lastNonAlphaOutputComponent = std::min<size_t>(outputComponentCount, 3); + + for (size_t i = 0; i < count; i++) + { + const T *offsetInput = reinterpret_cast<const T*>(input + (i * stride)); + T *offsetOutput = reinterpret_cast<T*>(output) + i * outputComponentCount; + + for (size_t j = 0; j < inputComponentCount; j++) + { + offsetOutput[j] = offsetInput[j]; + } + + for (size_t j = inputComponentCount; j < lastNonAlphaOutputComponent; j++) + { + // Set the remaining G/B channels to 0. + offsetOutput[j] = 0; + } + + if (inputComponentCount < outputComponentCount && outputComponentCount == 4) + { + // Set the remaining alpha channel to the defaultAlphaValue. + offsetOutput[3] = defaultAlphaValue; + } + } + } +} + +template <size_t inputComponentCount, size_t outputComponentCount> +inline void Copy8SintTo16SintVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + const size_t lastNonAlphaOutputComponent = std::min<size_t>(outputComponentCount, 3); + + for (size_t i = 0; i < count; i++) + { + const GLbyte *offsetInput = reinterpret_cast<const GLbyte*>(input + i * stride); + GLshort *offsetOutput = reinterpret_cast<GLshort*>(output)+i * outputComponentCount; + + for (size_t j = 0; j < inputComponentCount; j++) + { + offsetOutput[j] = static_cast<GLshort>(offsetInput[j]); + } + + for (size_t j = inputComponentCount; j < lastNonAlphaOutputComponent; j++) + { + // Set remaining G/B channels to 0. + offsetOutput[j] = 0; + } + + if (inputComponentCount < outputComponentCount && outputComponentCount == 4) + { + // On integer formats, we must set the Alpha channel to 1 if it's unused. + offsetOutput[3] = 1; + } + } +} + +template <size_t inputComponentCount, size_t outputComponentCount> +inline void Copy8SnormTo16SnormVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + for (size_t i = 0; i < count; i++) + { + const GLbyte *offsetInput = reinterpret_cast<const GLbyte*>(input + i * stride); + GLshort *offsetOutput = reinterpret_cast<GLshort*>(output) + i * outputComponentCount; + + for (size_t j = 0; j < inputComponentCount; j++) + { + // The original GLbyte value ranges from -128 to +127 (INT8_MAX). + // When converted to GLshort, the value must be scaled to between -32768 and +32767 (INT16_MAX). + if (offsetInput[j] > 0) + { + offsetOutput[j] = offsetInput[j] << 8 | offsetInput[j] << 1 | ((offsetInput[j] & 0x40) >> 6); + } + else + { + offsetOutput[j] = offsetInput[j] << 8; + } + } + + for (size_t j = inputComponentCount; j < std::min<size_t>(outputComponentCount, 3); j++) + { + // Set remaining G/B channels to 0. + offsetOutput[j] = 0; + } + + if (inputComponentCount < outputComponentCount && outputComponentCount == 4) + { + // On normalized formats, we must set the Alpha channel to the max value if it's unused. + offsetOutput[3] = INT16_MAX; + } + } +} + +template <size_t inputComponentCount, size_t outputComponentCount> +inline void Copy32FixedTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + static const float divisor = 1.0f / (1 << 16); + + for (size_t i = 0; i < count; i++) + { + const GLfixed* offsetInput = reinterpret_cast<const GLfixed*>(input + (stride * i)); + float* offsetOutput = reinterpret_cast<float*>(output) + i * outputComponentCount; + + for (size_t j = 0; j < inputComponentCount; j++) + { + offsetOutput[j] = static_cast<float>(offsetInput[j]) * divisor; + } + + // 4-component output formats would need special padding in the alpha channel. + static_assert(!(inputComponentCount < 4 && outputComponentCount == 4), + "An inputComponentCount less than 4 and an outputComponentCount equal to 4 is not supported."); + + for (size_t j = inputComponentCount; j < outputComponentCount; j++) + { + offsetOutput[j] = 0.0f; + } + } +} + +template <typename T, size_t inputComponentCount, size_t outputComponentCount, bool normalized> +inline void CopyTo32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + typedef std::numeric_limits<T> NL; + + for (size_t i = 0; i < count; i++) + { + const T *offsetInput = reinterpret_cast<const T*>(input + (stride * i)); + float *offsetOutput = reinterpret_cast<float*>(output) + i * outputComponentCount; + + for (size_t j = 0; j < inputComponentCount; j++) + { + if (normalized) + { + if (NL::is_signed) + { + const float divisor = 1.0f / (2 * static_cast<float>(NL::max()) + 1); + offsetOutput[j] = (2 * static_cast<float>(offsetInput[j]) + 1) * divisor; + } + else + { + offsetOutput[j] = static_cast<float>(offsetInput[j]) / NL::max(); + } + } + else + { + offsetOutput[j] = static_cast<float>(offsetInput[j]); + } + } + + // This would require special padding. + static_assert(!(inputComponentCount < 4 && outputComponentCount == 4), + "An inputComponentCount less than 4 and an outputComponentCount equal to 4 is not supported."); + + for (size_t j = inputComponentCount; j < outputComponentCount; j++) + { + offsetOutput[j] = 0.0f; + } + } +} + +namespace priv +{ + +template <bool isSigned, bool normalized, bool toFloat> +static inline void CopyPackedRGB(uint32_t data, uint8_t *output) +{ + const uint32_t rgbSignMask = 0x200; // 1 set at the 9 bit + const uint32_t negativeMask = 0xFFFFFC00; // All bits from 10 to 31 set to 1 + + if (toFloat) + { + GLfloat *floatOutput = reinterpret_cast<GLfloat*>(output); + if (isSigned) + { + GLfloat finalValue = 0; + if (data & rgbSignMask) + { + int negativeNumber = data | negativeMask; + finalValue = static_cast<GLfloat>(negativeNumber); + } + else + { + finalValue = static_cast<GLfloat>(data); + } + + if (normalized) + { + const int32_t maxValue = 0x1FF; // 1 set in bits 0 through 8 + const int32_t minValue = 0xFFFFFE01; // Inverse of maxValue + + // A 10-bit two's complement number has the possibility of being minValue - 1 but + // OpenGL's normalization rules dictate that it should be clamped to minValue in this + // case. + if (finalValue < minValue) + { + finalValue = minValue; + } + + const int32_t halfRange = (maxValue - minValue) >> 1; + *floatOutput = ((finalValue - minValue) / halfRange) - 1.0f; + } + else + { + *floatOutput = finalValue; + } + } + else + { + if (normalized) + { + const uint32_t maxValue = 0x3FF; // 1 set in bits 0 through 9 + *floatOutput = static_cast<GLfloat>(data) / static_cast<GLfloat>(maxValue); + } + else + { + *floatOutput = static_cast<GLfloat>(data); + } + } + } + else + { + if (isSigned) + { + GLshort *intOutput = reinterpret_cast<GLshort*>(output); + + if (data & rgbSignMask) + { + *intOutput = data | negativeMask; + } + else + { + *intOutput = data; + } + } + else + { + GLushort *uintOutput = reinterpret_cast<GLushort*>(output); + *uintOutput = data; + } + } +} + +template <bool isSigned, bool normalized, bool toFloat> +inline void CopyPackedAlpha(uint32_t data, uint8_t *output) +{ + if (toFloat) + { + GLfloat *floatOutput = reinterpret_cast<GLfloat*>(output); + if (isSigned) + { + if (normalized) + { + switch (data) + { + case 0x0: *floatOutput = 0.0f; break; + case 0x1: *floatOutput = 1.0f; break; + case 0x2: *floatOutput = -1.0f; break; + case 0x3: *floatOutput = -1.0f; break; + default: UNREACHABLE(); + } + } + else + { + switch (data) + { + case 0x0: *floatOutput = 0.0f; break; + case 0x1: *floatOutput = 1.0f; break; + case 0x2: *floatOutput = -2.0f; break; + case 0x3: *floatOutput = -1.0f; break; + default: UNREACHABLE(); + } + } + } + else + { + if (normalized) + { + switch (data) + { + case 0x0: *floatOutput = 0.0f / 3.0f; break; + case 0x1: *floatOutput = 1.0f / 3.0f; break; + case 0x2: *floatOutput = 2.0f / 3.0f; break; + case 0x3: *floatOutput = 3.0f / 3.0f; break; + default: UNREACHABLE(); + } + } + else + { + switch (data) + { + case 0x0: *floatOutput = 0.0f; break; + case 0x1: *floatOutput = 1.0f; break; + case 0x2: *floatOutput = 2.0f; break; + case 0x3: *floatOutput = 3.0f; break; + default: UNREACHABLE(); + } + } + } + } + else + { + if (isSigned) + { + GLshort *intOutput = reinterpret_cast<GLshort*>(output); + switch (data) + { + case 0x0: *intOutput = 0; break; + case 0x1: *intOutput = 1; break; + case 0x2: *intOutput = -2; break; + case 0x3: *intOutput = -1; break; + default: UNREACHABLE(); + } + } + else + { + GLushort *uintOutput = reinterpret_cast<GLushort*>(output); + switch (data) + { + case 0x0: *uintOutput = 0; break; + case 0x1: *uintOutput = 1; break; + case 0x2: *uintOutput = 2; break; + case 0x3: *uintOutput = 3; break; + default: UNREACHABLE(); + } + } + } +} + +} + +template <bool isSigned, bool normalized, bool toFloat> +inline void CopyXYZ10W2ToXYZW32FVertexData(const uint8_t *input, size_t stride, size_t count, uint8_t *output) +{ + const size_t outputComponentSize = toFloat ? 4 : 2; + const size_t componentCount = 4; + + const uint32_t rgbMask = 0x3FF; // 1 set in bits 0 through 9 + const size_t redShift = 0; // red is bits 0 through 9 + const size_t greenShift = 10; // green is bits 10 through 19 + const size_t blueShift = 20; // blue is bits 20 through 29 + + const uint32_t alphaMask = 0x3; // 1 set in bits 0 and 1 + const size_t alphaShift = 30; // Alpha is the 30 and 31 bits + + for (size_t i = 0; i < count; i++) + { + GLuint packedValue = *reinterpret_cast<const GLuint*>(input + (i * stride)); + uint8_t *offsetOutput = output + (i * outputComponentSize * componentCount); + + priv::CopyPackedRGB<isSigned, normalized, toFloat>( (packedValue >> redShift) & rgbMask, offsetOutput + (0 * outputComponentSize)); + priv::CopyPackedRGB<isSigned, normalized, toFloat>( (packedValue >> greenShift) & rgbMask, offsetOutput + (1 * outputComponentSize)); + priv::CopyPackedRGB<isSigned, normalized, toFloat>( (packedValue >> blueShift) & rgbMask, offsetOutput + (2 * outputComponentSize)); + priv::CopyPackedAlpha<isSigned, normalized, toFloat>((packedValue >> alphaShift) & alphaMask, offsetOutput + (3 * outputComponentSize)); + } +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.cpp new file mode 100644 index 0000000000..2f81d6d608 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.cpp @@ -0,0 +1,1328 @@ +// +// Copyright (c) 2013-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// formatutils11.cpp: Queries for GL image formats and their translations to D3D11 +// formats. + +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" + +#include "libANGLE/formatutils.h" +#include "libANGLE/renderer/Renderer.h" +#include "libANGLE/renderer/d3d/copyimage.h" +#include "libANGLE/renderer/d3d/generatemip.h" +#include "libANGLE/renderer/d3d/loadimage.h" +#include "libANGLE/renderer/d3d/d3d11/copyvertex.h" + +namespace rx +{ + +namespace d3d11 +{ + +typedef std::map<DXGI_FORMAT, GLenum> DXGIToESFormatMap; + +inline void AddDXGIToESEntry(DXGIToESFormatMap *map, DXGI_FORMAT key, GLenum value) +{ + map->insert(std::make_pair(key, value)); +} + +static DXGIToESFormatMap BuildDXGIToESFormatMap() +{ + DXGIToESFormatMap map; + + AddDXGIToESEntry(&map, DXGI_FORMAT_UNKNOWN, GL_NONE); + + AddDXGIToESEntry(&map, DXGI_FORMAT_A8_UNORM, GL_ALPHA8_EXT); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8_UNORM, GL_R8); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8_UNORM, GL_RG8); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8B8A8_UNORM, GL_RGBA8); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, GL_SRGB8_ALPHA8); + AddDXGIToESEntry(&map, DXGI_FORMAT_B8G8R8A8_UNORM, GL_BGRA8_EXT); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R8_SNORM, GL_R8_SNORM); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8_SNORM, GL_RG8_SNORM); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8B8A8_SNORM, GL_RGBA8_SNORM); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R8_UINT, GL_R8UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16_UINT, GL_R16UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32_UINT, GL_R32UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8_UINT, GL_RG8UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16_UINT, GL_RG16UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32_UINT, GL_RG32UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32_UINT, GL_RGB32UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8B8A8_UINT, GL_RGBA8UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16B16A16_UINT, GL_RGBA16UI); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32A32_UINT, GL_RGBA32UI); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R8_SINT, GL_R8I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16_SINT, GL_R16I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32_SINT, GL_R32I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8_SINT, GL_RG8I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16_SINT, GL_RG16I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32_SINT, GL_RG32I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32_SINT, GL_RGB32I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R8G8B8A8_SINT, GL_RGBA8I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16B16A16_SINT, GL_RGBA16I); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32A32_SINT, GL_RGBA32I); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R10G10B10A2_UNORM, GL_RGB10_A2); + AddDXGIToESEntry(&map, DXGI_FORMAT_R10G10B10A2_UINT, GL_RGB10_A2UI); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R16_FLOAT, GL_R16F); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16_FLOAT, GL_RG16F); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16G16B16A16_FLOAT, GL_RGBA16F); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R32_FLOAT, GL_R32F); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32_FLOAT, GL_RG32F); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32_FLOAT, GL_RGB32F); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G32B32A32_FLOAT, GL_RGBA32F); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R9G9B9E5_SHAREDEXP, GL_RGB9_E5); + AddDXGIToESEntry(&map, DXGI_FORMAT_R11G11B10_FLOAT, GL_R11F_G11F_B10F); + + AddDXGIToESEntry(&map, DXGI_FORMAT_R16_TYPELESS, GL_DEPTH_COMPONENT16); + AddDXGIToESEntry(&map, DXGI_FORMAT_R16_UNORM, GL_DEPTH_COMPONENT16); + AddDXGIToESEntry(&map, DXGI_FORMAT_D16_UNORM, GL_DEPTH_COMPONENT16); + AddDXGIToESEntry(&map, DXGI_FORMAT_R24G8_TYPELESS, GL_DEPTH24_STENCIL8_OES); + AddDXGIToESEntry(&map, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, GL_DEPTH24_STENCIL8_OES); + AddDXGIToESEntry(&map, DXGI_FORMAT_D24_UNORM_S8_UINT, GL_DEPTH24_STENCIL8_OES); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32G8X24_TYPELESS, GL_DEPTH32F_STENCIL8); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS, GL_DEPTH32F_STENCIL8); + AddDXGIToESEntry(&map, DXGI_FORMAT_D32_FLOAT_S8X24_UINT, GL_DEPTH32F_STENCIL8); + AddDXGIToESEntry(&map, DXGI_FORMAT_R32_TYPELESS, GL_DEPTH_COMPONENT32F); + AddDXGIToESEntry(&map, DXGI_FORMAT_D32_FLOAT, GL_DEPTH_COMPONENT32F); + + AddDXGIToESEntry(&map, DXGI_FORMAT_BC1_UNORM, GL_COMPRESSED_RGBA_S3TC_DXT1_EXT); + AddDXGIToESEntry(&map, DXGI_FORMAT_BC2_UNORM, GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE); + AddDXGIToESEntry(&map, DXGI_FORMAT_BC3_UNORM, GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE); + + return map; +} + +struct D3D11FastCopyFormat +{ + GLenum destFormat; + GLenum destType; + ColorCopyFunction copyFunction; + + D3D11FastCopyFormat(GLenum destFormat, GLenum destType, ColorCopyFunction copyFunction) + : destFormat(destFormat), destType(destType), copyFunction(copyFunction) + { } + + bool operator<(const D3D11FastCopyFormat& other) const + { + return memcmp(this, &other, sizeof(D3D11FastCopyFormat)) < 0; + } +}; + +typedef std::multimap<DXGI_FORMAT, D3D11FastCopyFormat> D3D11FastCopyMap; + +static D3D11FastCopyMap BuildFastCopyMap() +{ + D3D11FastCopyMap map; + + map.insert(std::make_pair(DXGI_FORMAT_B8G8R8A8_UNORM, D3D11FastCopyFormat(GL_RGBA, GL_UNSIGNED_BYTE, CopyBGRA8ToRGBA8))); + + return map; +} + +struct DXGIColorFormatInfo +{ + size_t redBits; + size_t greenBits; + size_t blueBits; + + size_t luminanceBits; + + size_t alphaBits; + size_t sharedBits; +}; + +typedef std::map<DXGI_FORMAT, DXGIColorFormatInfo> ColorFormatInfoMap; +typedef std::pair<DXGI_FORMAT, DXGIColorFormatInfo> ColorFormatInfoPair; + +static inline void InsertDXGIColorFormatInfo(ColorFormatInfoMap *map, DXGI_FORMAT format, size_t redBits, size_t greenBits, + size_t blueBits, size_t alphaBits, size_t sharedBits) +{ + DXGIColorFormatInfo info; + info.redBits = redBits; + info.greenBits = greenBits; + info.blueBits = blueBits; + info.alphaBits = alphaBits; + info.sharedBits = sharedBits; + + map->insert(std::make_pair(format, info)); +} + +static ColorFormatInfoMap BuildColorFormatInfoMap() +{ + ColorFormatInfoMap map; + + // | DXGI format | R | G | B | A | S | + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_A8_UNORM, 0, 0, 0, 8, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8_UNORM, 8, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8_UNORM, 8, 8, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8B8A8_UNORM, 8, 8, 8, 8, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, 8, 8, 8, 8, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_B8G8R8A8_UNORM, 8, 8, 8, 8, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8_SNORM, 8, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8_SNORM, 8, 8, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8B8A8_SNORM, 8, 8, 8, 8, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8_UINT, 8, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16_UINT, 16, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32_UINT, 32, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8_UINT, 8, 8, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16_UINT, 16, 16, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32_UINT, 32, 32, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32_UINT, 32, 32, 32, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8B8A8_UINT, 8, 8, 8, 8, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16B16A16_UINT, 16, 16, 16, 16, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32A32_UINT, 32, 32, 32, 32, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8_SINT, 8, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16_SINT, 16, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32_SINT, 32, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8_SINT, 8, 8, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16_SINT, 16, 16, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32_SINT, 32, 32, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32_SINT, 32, 32, 32, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R8G8B8A8_SINT, 8, 8, 8, 8, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16B16A16_SINT, 16, 16, 16, 16, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32A32_SINT, 32, 32, 32, 32, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R10G10B10A2_UNORM, 10, 10, 10, 2, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R10G10B10A2_UINT, 10, 10, 10, 2, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16_FLOAT, 16, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16_FLOAT, 16, 16, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R16G16B16A16_FLOAT, 16, 16, 16, 16, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32_FLOAT, 32, 0, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32_FLOAT, 32, 32, 0, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32_FLOAT, 32, 32, 32, 0, 0); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R32G32B32A32_FLOAT, 32, 32, 32, 32, 0); + + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R9G9B9E5_SHAREDEXP, 9, 9, 9, 0, 5); + InsertDXGIColorFormatInfo(&map, DXGI_FORMAT_R11G11B10_FLOAT, 11, 11, 10, 0, 0); + + return map; +} + +struct DXGIDepthStencilInfo +{ + unsigned int depthBits; + unsigned int depthOffset; + unsigned int stencilBits; + unsigned int stencilOffset; +}; + +typedef std::map<DXGI_FORMAT, DXGIDepthStencilInfo> DepthStencilInfoMap; +typedef std::pair<DXGI_FORMAT, DXGIDepthStencilInfo> DepthStencilInfoPair; + +static inline void InsertDXGIDepthStencilInfo(DepthStencilInfoMap *map, DXGI_FORMAT format, unsigned int depthBits, + unsigned int depthOffset, unsigned int stencilBits, unsigned int stencilOffset) +{ + DXGIDepthStencilInfo info; + info.depthBits = depthBits; + info.depthOffset = depthOffset; + info.stencilBits = stencilBits; + info.stencilOffset = stencilOffset; + + map->insert(std::make_pair(format, info)); +} + +static DepthStencilInfoMap BuildDepthStencilInfoMap() +{ + DepthStencilInfoMap map; + + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R16_TYPELESS, 16, 0, 0, 0); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R16_UNORM, 16, 0, 0, 0); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_D16_UNORM, 16, 0, 0, 0); + + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R24G8_TYPELESS, 24, 0, 8, 24); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, 24, 0, 8, 24); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_D24_UNORM_S8_UINT, 24, 0, 8, 24); + + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R32_TYPELESS, 32, 0, 0, 0); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R32_FLOAT, 32, 0, 0, 0); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_D32_FLOAT, 32, 0, 0, 0); + + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R32G8X24_TYPELESS, 32, 0, 8, 32); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS, 32, 0, 8, 32); + InsertDXGIDepthStencilInfo(&map, DXGI_FORMAT_D32_FLOAT_S8X24_UINT, 32, 0, 8, 32); + + return map; +} + +typedef std::map<DXGI_FORMAT, DXGIFormat> DXGIFormatInfoMap; + +DXGIFormat::DXGIFormat() + : pixelBytes(0), + blockWidth(0), + blockHeight(0), + redBits(0), + greenBits(0), + blueBits(0), + alphaBits(0), + sharedBits(0), + depthBits(0), + depthOffset(0), + stencilBits(0), + stencilOffset(0), + internalFormat(GL_NONE), + componentType(GL_NONE), + mipGenerationFunction(NULL), + colorReadFunction(NULL), + fastCopyFunctions() +{ +} + +ColorCopyFunction DXGIFormat::getFastCopyFunction(GLenum format, GLenum type) const +{ + FastCopyFunctionMap::const_iterator iter = fastCopyFunctions.find(std::make_pair(format, type)); + return (iter != fastCopyFunctions.end()) ? iter->second : NULL; +} + +void AddDXGIFormat(DXGIFormatInfoMap *map, DXGI_FORMAT dxgiFormat, GLuint pixelBits, GLuint blockWidth, GLuint blockHeight, + GLenum componentType, MipGenerationFunction mipFunc, ColorReadFunction readFunc) +{ + DXGIFormat info; + info.pixelBytes = pixelBits / 8; + info.blockWidth = blockWidth; + info.blockHeight = blockHeight; + + static const ColorFormatInfoMap colorInfoMap = BuildColorFormatInfoMap(); + ColorFormatInfoMap::const_iterator colorInfoIter = colorInfoMap.find(dxgiFormat); + if (colorInfoIter != colorInfoMap.end()) + { + const DXGIColorFormatInfo &colorInfo = colorInfoIter->second; + info.redBits = colorInfo.redBits; + info.greenBits = colorInfo.greenBits; + info.blueBits = colorInfo.blueBits; + info.alphaBits = colorInfo.alphaBits; + info.sharedBits = colorInfo.sharedBits; + } + + static const DepthStencilInfoMap dsInfoMap = BuildDepthStencilInfoMap(); + DepthStencilInfoMap::const_iterator dsInfoIter = dsInfoMap.find(dxgiFormat); + if (dsInfoIter != dsInfoMap.end()) + { + const DXGIDepthStencilInfo &dsInfo = dsInfoIter->second; + info.depthBits = dsInfo.depthBits; + info.depthOffset = dsInfo.depthOffset; + info.stencilBits = dsInfo.stencilBits; + info.stencilOffset = dsInfo.stencilOffset; + } + + static const DXGIToESFormatMap dxgiToESMap = BuildDXGIToESFormatMap(); + DXGIToESFormatMap::const_iterator dxgiToESIter = dxgiToESMap.find(dxgiFormat); + info.internalFormat = (dxgiToESIter != dxgiToESMap.end()) ? dxgiToESIter->second : GL_NONE; + + info.componentType = componentType; + + info.mipGenerationFunction = mipFunc; + info.colorReadFunction = readFunc; + + static const D3D11FastCopyMap fastCopyMap = BuildFastCopyMap(); + std::pair<D3D11FastCopyMap::const_iterator, D3D11FastCopyMap::const_iterator> fastCopyIter = fastCopyMap.equal_range(dxgiFormat); + for (D3D11FastCopyMap::const_iterator i = fastCopyIter.first; i != fastCopyIter.second; i++) + { + info.fastCopyFunctions.insert(std::make_pair(std::make_pair(i->second.destFormat, i->second.destType), i->second.copyFunction)); + } + + map->insert(std::make_pair(dxgiFormat, info)); +} + +// A map to determine the pixel size and mipmap generation function of a given DXGI format +static DXGIFormatInfoMap BuildDXGIFormatInfoMap() +{ + DXGIFormatInfoMap map; + + // | DXGI format |S |W |H |Component Type | Mip generation function | Color read function + AddDXGIFormat(&map, DXGI_FORMAT_UNKNOWN, 0, 0, 0, GL_NONE, NULL, NULL); + + AddDXGIFormat(&map, DXGI_FORMAT_A8_UNORM, 8, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<A8>, ReadColor<A8, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8_UNORM, 8, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<R8>, ReadColor<R8, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8_UNORM, 16, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<R8G8>, ReadColor<R8G8, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8B8A8_UNORM, 32, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<R8G8B8A8>, ReadColor<R8G8B8A8, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, 32, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<R8G8B8A8>, ReadColor<R8G8B8A8, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_B8G8R8A8_UNORM, 32, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<B8G8R8A8>, ReadColor<B8G8R8A8, GLfloat>); + + AddDXGIFormat(&map, DXGI_FORMAT_R8_SNORM, 8, 1, 1, GL_SIGNED_NORMALIZED, GenerateMip<R8S>, ReadColor<R8S, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8_SNORM, 16, 1, 1, GL_SIGNED_NORMALIZED, GenerateMip<R8G8S>, ReadColor<R8G8S, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8B8A8_SNORM, 32, 1, 1, GL_SIGNED_NORMALIZED, GenerateMip<R8G8B8A8S>, ReadColor<R8G8B8A8S, GLfloat>); + + AddDXGIFormat(&map, DXGI_FORMAT_R8_UINT, 8, 1, 1, GL_UNSIGNED_INT, GenerateMip<R8>, ReadColor<R8, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16_UINT, 16, 1, 1, GL_UNSIGNED_INT, GenerateMip<R16>, ReadColor<R16, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32_UINT, 32, 1, 1, GL_UNSIGNED_INT, GenerateMip<R32>, ReadColor<R32, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8_UINT, 16, 1, 1, GL_UNSIGNED_INT, GenerateMip<R8G8>, ReadColor<R8G8, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16_UINT, 32, 1, 1, GL_UNSIGNED_INT, GenerateMip<R16G16>, ReadColor<R16G16, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32_UINT, 64, 1, 1, GL_UNSIGNED_INT, GenerateMip<R32G32>, ReadColor<R32G32, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32_UINT, 96, 1, 1, GL_UNSIGNED_INT, GenerateMip<R32G32B32>, ReadColor<R32G32B32, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8B8A8_UINT, 32, 1, 1, GL_UNSIGNED_INT, GenerateMip<R8G8B8A8>, ReadColor<R8G8B8A8, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16B16A16_UINT, 64, 1, 1, GL_UNSIGNED_INT, GenerateMip<R16G16B16A16>, ReadColor<R16G16B16A16, GLuint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32A32_UINT, 128, 1, 1, GL_UNSIGNED_INT, GenerateMip<R32G32B32A32>, ReadColor<R32G32B32A32, GLuint>); + + AddDXGIFormat(&map, DXGI_FORMAT_R8_SINT, 8, 1, 1, GL_INT, GenerateMip<R8S>, ReadColor<R8S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16_SINT, 16, 1, 1, GL_INT, GenerateMip<R16S>, ReadColor<R16S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32_SINT, 32, 1, 1, GL_INT, GenerateMip<R32S>, ReadColor<R32S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8_SINT, 16, 1, 1, GL_INT, GenerateMip<R8G8S>, ReadColor<R8G8S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16_SINT, 32, 1, 1, GL_INT, GenerateMip<R16G16S>, ReadColor<R16G16S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32_SINT, 64, 1, 1, GL_INT, GenerateMip<R32G32S>, ReadColor<R32G32S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32_SINT, 96, 1, 1, GL_INT, GenerateMip<R32G32B32S>, ReadColor<R32G32B32S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R8G8B8A8_SINT, 32, 1, 1, GL_INT, GenerateMip<R8G8B8A8S>, ReadColor<R8G8B8A8S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16B16A16_SINT, 64, 1, 1, GL_INT, GenerateMip<R16G16B16A16S>, ReadColor<R16G16B16A16S, GLint>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32A32_SINT, 128, 1, 1, GL_INT, GenerateMip<R32G32B32A32S>, ReadColor<R32G32B32A32S, GLint>); + + AddDXGIFormat(&map, DXGI_FORMAT_R10G10B10A2_UNORM, 32, 1, 1, GL_UNSIGNED_NORMALIZED, GenerateMip<R10G10B10A2>, ReadColor<R10G10B10A2, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R10G10B10A2_UINT, 32, 1, 1, GL_UNSIGNED_INT, GenerateMip<R10G10B10A2>, ReadColor<R10G10B10A2, GLuint>); + + AddDXGIFormat(&map, DXGI_FORMAT_R16_FLOAT, 16, 1, 1, GL_FLOAT, GenerateMip<R16F>, ReadColor<R16F, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16_FLOAT, 32, 1, 1, GL_FLOAT, GenerateMip<R16G16F>, ReadColor<R16G16F, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16B16A16_FLOAT, 64, 1, 1, GL_FLOAT, GenerateMip<R16G16B16A16F>, ReadColor<R16G16B16A16F, GLfloat>); + + AddDXGIFormat(&map, DXGI_FORMAT_R32_FLOAT, 32, 1, 1, GL_FLOAT, GenerateMip<R32F>, ReadColor<R32F, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32_FLOAT, 64, 1, 1, GL_FLOAT, GenerateMip<R32G32F>, ReadColor<R32G32F, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32_FLOAT, 96, 1, 1, GL_FLOAT, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R32G32B32A32_FLOAT, 128, 1, 1, GL_FLOAT, GenerateMip<R32G32B32A32F>, ReadColor<R32G32B32A32F, GLfloat>); + + AddDXGIFormat(&map, DXGI_FORMAT_R9G9B9E5_SHAREDEXP, 32, 1, 1, GL_FLOAT, GenerateMip<R9G9B9E5>, ReadColor<R9G9B9E5, GLfloat>); + AddDXGIFormat(&map, DXGI_FORMAT_R11G11B10_FLOAT, 32, 1, 1, GL_FLOAT, GenerateMip<R11G11B10F>, ReadColor<R11G11B10F, GLfloat>); + + AddDXGIFormat(&map, DXGI_FORMAT_R16_TYPELESS, 16, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16_UNORM, 16, 1, 1, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_D16_UNORM, 16, 1, 1, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R24G8_TYPELESS, 32, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, 32, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_D24_UNORM_S8_UINT, 32, 1, 1, GL_UNSIGNED_INT, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R32G8X24_TYPELESS, 64, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS, 64, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_D32_FLOAT_S8X24_UINT, 64, 1, 1, GL_UNSIGNED_INT, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R32_TYPELESS, 32, 1, 1, GL_NONE, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_D32_FLOAT, 32, 1, 1, GL_FLOAT, NULL, NULL); + + AddDXGIFormat(&map, DXGI_FORMAT_BC1_UNORM, 64, 4, 4, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_BC2_UNORM, 128, 4, 4, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_BC3_UNORM, 128, 4, 4, GL_UNSIGNED_NORMALIZED, NULL, NULL); + + // Useful formats for vertex buffers + AddDXGIFormat(&map, DXGI_FORMAT_R16_UNORM, 16, 1, 1, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16_SNORM, 16, 1, 1, GL_SIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16_UNORM, 32, 1, 1, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16_SNORM, 32, 1, 1, GL_SIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16B16A16_UNORM, 64, 1, 1, GL_UNSIGNED_NORMALIZED, NULL, NULL); + AddDXGIFormat(&map, DXGI_FORMAT_R16G16B16A16_SNORM, 64, 1, 1, GL_SIGNED_NORMALIZED, NULL, NULL); + + return map; +} + +const DXGIFormat &GetDXGIFormatInfo(DXGI_FORMAT format) +{ + static const DXGIFormatInfoMap infoMap = BuildDXGIFormatInfoMap(); + DXGIFormatInfoMap::const_iterator iter = infoMap.find(format); + if (iter != infoMap.end()) + { + return iter->second; + } + else + { + static DXGIFormat defaultInfo; + return defaultInfo; + } +} + +struct SwizzleSizeType +{ + size_t maxComponentSize; + GLenum componentType; + + SwizzleSizeType() + : maxComponentSize(0), componentType(GL_NONE) + { } + + SwizzleSizeType(size_t maxComponentSize, GLenum componentType) + : maxComponentSize(maxComponentSize), componentType(componentType) + { } + + bool operator<(const SwizzleSizeType& other) const + { + return (maxComponentSize != other.maxComponentSize) ? (maxComponentSize < other.maxComponentSize) + : (componentType < other.componentType); + } +}; + +struct SwizzleFormatInfo +{ + DXGI_FORMAT mTexFormat; + DXGI_FORMAT mSRVFormat; + DXGI_FORMAT mRTVFormat; + + SwizzleFormatInfo() + : mTexFormat(DXGI_FORMAT_UNKNOWN), mSRVFormat(DXGI_FORMAT_UNKNOWN), mRTVFormat(DXGI_FORMAT_UNKNOWN) + { } + + SwizzleFormatInfo(DXGI_FORMAT texFormat, DXGI_FORMAT srvFormat, DXGI_FORMAT rtvFormat) + : mTexFormat(texFormat), mSRVFormat(srvFormat), mRTVFormat(rtvFormat) + { } +}; + +typedef std::map<SwizzleSizeType, SwizzleFormatInfo> SwizzleInfoMap; +typedef std::pair<SwizzleSizeType, SwizzleFormatInfo> SwizzleInfoPair; + +static SwizzleInfoMap BuildSwizzleInfoMap() +{ + SwizzleInfoMap map; + + map.insert(SwizzleInfoPair(SwizzleSizeType( 8, GL_UNSIGNED_NORMALIZED), SwizzleFormatInfo(DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM ))); + map.insert(SwizzleInfoPair(SwizzleSizeType(16, GL_UNSIGNED_NORMALIZED), SwizzleFormatInfo(DXGI_FORMAT_R16G16B16A16_UNORM, DXGI_FORMAT_R16G16B16A16_UNORM, DXGI_FORMAT_R16G16B16A16_UNORM))); + map.insert(SwizzleInfoPair(SwizzleSizeType(24, GL_UNSIGNED_NORMALIZED), SwizzleFormatInfo(DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT))); + map.insert(SwizzleInfoPair(SwizzleSizeType(32, GL_UNSIGNED_NORMALIZED), SwizzleFormatInfo(DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT))); + + map.insert(SwizzleInfoPair(SwizzleSizeType( 8, GL_SIGNED_NORMALIZED ), SwizzleFormatInfo(DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM ))); + + map.insert(SwizzleInfoPair(SwizzleSizeType(16, GL_FLOAT ), SwizzleFormatInfo(DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT))); + map.insert(SwizzleInfoPair(SwizzleSizeType(32, GL_FLOAT ), SwizzleFormatInfo(DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT))); + + map.insert(SwizzleInfoPair(SwizzleSizeType( 8, GL_UNSIGNED_INT ), SwizzleFormatInfo(DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT ))); + map.insert(SwizzleInfoPair(SwizzleSizeType(16, GL_UNSIGNED_INT ), SwizzleFormatInfo(DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT ))); + map.insert(SwizzleInfoPair(SwizzleSizeType(32, GL_UNSIGNED_INT ), SwizzleFormatInfo(DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT ))); + + map.insert(SwizzleInfoPair(SwizzleSizeType( 8, GL_INT ), SwizzleFormatInfo(DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT ))); + map.insert(SwizzleInfoPair(SwizzleSizeType(16, GL_INT ), SwizzleFormatInfo(DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT ))); + map.insert(SwizzleInfoPair(SwizzleSizeType(32, GL_INT ), SwizzleFormatInfo(DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT ))); + + return map; +} + +typedef std::pair<GLint, InitializeTextureDataFunction> InternalFormatInitializerPair; +typedef std::map<GLint, InitializeTextureDataFunction> InternalFormatInitializerMap; + +static InternalFormatInitializerMap BuildInternalFormatInitializerMap() +{ + InternalFormatInitializerMap map; + + map.insert(InternalFormatInitializerPair(GL_RGB8, Initialize4ComponentData<GLubyte, 0x00, 0x00, 0x00, 0xFF> )); + map.insert(InternalFormatInitializerPair(GL_RGB565, Initialize4ComponentData<GLubyte, 0x00, 0x00, 0x00, 0xFF> )); + map.insert(InternalFormatInitializerPair(GL_SRGB8, Initialize4ComponentData<GLubyte, 0x00, 0x00, 0x00, 0xFF> )); + map.insert(InternalFormatInitializerPair(GL_RGB16F, Initialize4ComponentData<GLhalf, 0x0000, 0x0000, 0x0000, gl::Float16One>)); + map.insert(InternalFormatInitializerPair(GL_RGB32F, Initialize4ComponentData<GLfloat, 0x00000000, 0x00000000, 0x00000000, gl::Float32One>)); + map.insert(InternalFormatInitializerPair(GL_RGB8UI, Initialize4ComponentData<GLubyte, 0x00, 0x00, 0x00, 0x01> )); + map.insert(InternalFormatInitializerPair(GL_RGB8I, Initialize4ComponentData<GLbyte, 0x00, 0x00, 0x00, 0x01> )); + map.insert(InternalFormatInitializerPair(GL_RGB16UI, Initialize4ComponentData<GLushort, 0x0000, 0x0000, 0x0000, 0x0001> )); + map.insert(InternalFormatInitializerPair(GL_RGB16I, Initialize4ComponentData<GLshort, 0x0000, 0x0000, 0x0000, 0x0001> )); + map.insert(InternalFormatInitializerPair(GL_RGB32UI, Initialize4ComponentData<GLuint, 0x00000000, 0x00000000, 0x00000000, 0x00000001> )); + map.insert(InternalFormatInitializerPair(GL_RGB32I, Initialize4ComponentData<GLint, 0x00000000, 0x00000000, 0x00000000, 0x00000001> )); + + return map; +} + +// ES3 image loading functions vary based on the internal format and data type given, +// this map type determines the loading function from the internal format and type supplied +// to glTex*Image*D and the destination DXGI_FORMAT. Source formats and types are taken from +// Tables 3.2 and 3.3 of the ES 3 spec. +typedef std::pair<GLenum, LoadImageFunction> TypeLoadFunctionPair; +typedef std::map<GLenum, std::vector<TypeLoadFunctionPair> > D3D11LoadFunctionMap; + +static void UnimplementedLoadFunction(size_t width, size_t height, size_t depth, + const uint8_t *input, size_t inputRowPitch, size_t inputDepthPitch, + uint8_t *output, size_t outputRowPitch, size_t outputDepthPitch) +{ + UNIMPLEMENTED(); +} + +static void UnreachableLoadFunction(size_t width, size_t height, size_t depth, + const uint8_t *input, size_t inputRowPitch, size_t inputDepthPitch, + uint8_t *output, size_t outputRowPitch, size_t outputDepthPitch) +{ + UNREACHABLE(); +} + +// A helper function to insert data into the D3D11LoadFunctionMap with fewer characters. +static inline void InsertLoadFunction(D3D11LoadFunctionMap *map, GLenum internalFormat, GLenum type, + LoadImageFunction loadFunc) +{ + (*map)[internalFormat].push_back(TypeLoadFunctionPair(type, loadFunc)); +} + +D3D11LoadFunctionMap BuildD3D11_FL9_3_LoadFunctionMap() +{ + D3D11LoadFunctionMap map; + + // From GL_EXT_texture_storage. Also used by GL_ALPHA8 + // On feature level 9_3, A8_UNORM doesn't support mipmaps, so we must use RGBA8 instead + InsertLoadFunction(&map, GL_ALPHA8_EXT, GL_UNSIGNED_BYTE, LoadA8ToRGBA8); + + return map; +} + +D3D11LoadFunctionMap BuildD3D11_FL10_0Plus_LoadFunctionMap() +{ + D3D11LoadFunctionMap map; + + // From GL_EXT_texture_storage. Also used by GL_ALPHA8 + InsertLoadFunction(&map, GL_ALPHA8_EXT, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 1>); + + return map; +} + +D3D11LoadFunctionMap BuildBaseD3D11LoadFunctionMap() +{ + D3D11LoadFunctionMap map; + + // | Internal format | Type | Load function | + InsertLoadFunction(&map, GL_RGBA8, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_RGB5_A1, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_RGBA4, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_SRGB8_ALPHA8, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_RGBA8_SNORM, GL_BYTE, LoadToNative<GLbyte, 4> ); + InsertLoadFunction(&map, GL_RGBA4, GL_UNSIGNED_SHORT_4_4_4_4, LoadRGBA4ToRGBA8 ); + InsertLoadFunction(&map, GL_RGB10_A2, GL_UNSIGNED_INT_2_10_10_10_REV, LoadToNative<GLuint, 1> ); + InsertLoadFunction(&map, GL_RGB5_A1, GL_UNSIGNED_SHORT_5_5_5_1, LoadRGB5A1ToRGBA8 ); + InsertLoadFunction(&map, GL_RGB5_A1, GL_UNSIGNED_INT_2_10_10_10_REV, LoadRGB10A2ToRGBA8 ); + InsertLoadFunction(&map, GL_RGBA16F, GL_HALF_FLOAT, LoadToNative<GLhalf, 4> ); + InsertLoadFunction(&map, GL_RGBA16F, GL_HALF_FLOAT_OES, LoadToNative<GLhalf, 4> ); + InsertLoadFunction(&map, GL_RGBA32F, GL_FLOAT, LoadToNative<GLfloat, 4> ); + InsertLoadFunction(&map, GL_RGBA16F, GL_FLOAT, Load32FTo16F<4> ); + InsertLoadFunction(&map, GL_RGBA8UI, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_RGBA8I, GL_BYTE, LoadToNative<GLbyte, 4> ); + InsertLoadFunction(&map, GL_RGBA16UI, GL_UNSIGNED_SHORT, LoadToNative<GLushort, 4> ); + InsertLoadFunction(&map, GL_RGBA16I, GL_SHORT, LoadToNative<GLshort, 4> ); + InsertLoadFunction(&map, GL_RGBA32UI, GL_UNSIGNED_INT, LoadToNative<GLuint, 4> ); + InsertLoadFunction(&map, GL_RGBA32I, GL_INT, LoadToNative<GLint, 4> ); + InsertLoadFunction(&map, GL_RGB10_A2UI, GL_UNSIGNED_INT_2_10_10_10_REV, LoadToNative<GLuint, 1> ); + InsertLoadFunction(&map, GL_RGB8, GL_UNSIGNED_BYTE, LoadToNative3To4<GLubyte, 0xFF> ); + InsertLoadFunction(&map, GL_RGB565, GL_UNSIGNED_BYTE, LoadToNative3To4<GLubyte, 0xFF> ); + InsertLoadFunction(&map, GL_SRGB8, GL_UNSIGNED_BYTE, LoadToNative3To4<GLubyte, 0xFF> ); + InsertLoadFunction(&map, GL_RGB8_SNORM, GL_BYTE, LoadToNative3To4<GLbyte, 0x7F> ); + InsertLoadFunction(&map, GL_RGB565, GL_UNSIGNED_SHORT_5_6_5, LoadR5G6B5ToRGBA8 ); + InsertLoadFunction(&map, GL_R11F_G11F_B10F, GL_UNSIGNED_INT_10F_11F_11F_REV, LoadToNative<GLuint, 1> ); + InsertLoadFunction(&map, GL_RGB9_E5, GL_UNSIGNED_INT_5_9_9_9_REV, LoadToNative<GLuint, 1> ); + InsertLoadFunction(&map, GL_RGB16F, GL_HALF_FLOAT, LoadToNative3To4<GLhalf, gl::Float16One>); + InsertLoadFunction(&map, GL_RGB16F, GL_HALF_FLOAT_OES, LoadToNative3To4<GLhalf, gl::Float16One>); + InsertLoadFunction(&map, GL_R11F_G11F_B10F, GL_HALF_FLOAT, LoadRGB16FToRG11B10F ); + InsertLoadFunction(&map, GL_R11F_G11F_B10F, GL_HALF_FLOAT_OES, LoadRGB16FToRG11B10F ); + InsertLoadFunction(&map, GL_RGB9_E5, GL_HALF_FLOAT, LoadRGB16FToRGB9E5 ); + InsertLoadFunction(&map, GL_RGB9_E5, GL_HALF_FLOAT_OES, LoadRGB16FToRGB9E5 ); + InsertLoadFunction(&map, GL_RGB32F, GL_FLOAT, LoadToNative3To4<GLfloat, gl::Float32One>); + InsertLoadFunction(&map, GL_RGB16F, GL_FLOAT, LoadRGB32FToRGBA16F ); + InsertLoadFunction(&map, GL_R11F_G11F_B10F, GL_FLOAT, LoadRGB32FToRG11B10F ); + InsertLoadFunction(&map, GL_RGB9_E5, GL_FLOAT, LoadRGB32FToRGB9E5 ); + InsertLoadFunction(&map, GL_RGB8UI, GL_UNSIGNED_BYTE, LoadToNative3To4<GLubyte, 0x01> ); + InsertLoadFunction(&map, GL_RGB8I, GL_BYTE, LoadToNative3To4<GLbyte, 0x01> ); + InsertLoadFunction(&map, GL_RGB16UI, GL_UNSIGNED_SHORT, LoadToNative3To4<GLushort, 0x0001> ); + InsertLoadFunction(&map, GL_RGB16I, GL_SHORT, LoadToNative3To4<GLshort, 0x0001> ); + InsertLoadFunction(&map, GL_RGB32UI, GL_UNSIGNED_INT, LoadToNative3To4<GLuint, 0x00000001> ); + InsertLoadFunction(&map, GL_RGB32I, GL_INT, LoadToNative3To4<GLint, 0x00000001> ); + InsertLoadFunction(&map, GL_RG8, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 2> ); + InsertLoadFunction(&map, GL_RG8_SNORM, GL_BYTE, LoadToNative<GLbyte, 2> ); + InsertLoadFunction(&map, GL_RG16F, GL_HALF_FLOAT, LoadToNative<GLhalf, 2> ); + InsertLoadFunction(&map, GL_RG16F, GL_HALF_FLOAT_OES, LoadToNative<GLhalf, 2> ); + InsertLoadFunction(&map, GL_RG32F, GL_FLOAT, LoadToNative<GLfloat, 2> ); + InsertLoadFunction(&map, GL_RG16F, GL_FLOAT, Load32FTo16F<2> ); + InsertLoadFunction(&map, GL_RG8UI, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 2> ); + InsertLoadFunction(&map, GL_RG8I, GL_BYTE, LoadToNative<GLbyte, 2> ); + InsertLoadFunction(&map, GL_RG16UI, GL_UNSIGNED_SHORT, LoadToNative<GLushort, 2> ); + InsertLoadFunction(&map, GL_RG16I, GL_SHORT, LoadToNative<GLshort, 2> ); + InsertLoadFunction(&map, GL_RG32UI, GL_UNSIGNED_INT, LoadToNative<GLuint, 2> ); + InsertLoadFunction(&map, GL_RG32I, GL_INT, LoadToNative<GLint, 2> ); + InsertLoadFunction(&map, GL_R8, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 1> ); + InsertLoadFunction(&map, GL_R8_SNORM, GL_BYTE, LoadToNative<GLbyte, 1> ); + InsertLoadFunction(&map, GL_R16F, GL_HALF_FLOAT, LoadToNative<GLhalf, 1> ); + InsertLoadFunction(&map, GL_R16F, GL_HALF_FLOAT_OES, LoadToNative<GLhalf, 1> ); + InsertLoadFunction(&map, GL_R32F, GL_FLOAT, LoadToNative<GLfloat, 1> ); + InsertLoadFunction(&map, GL_R16F, GL_FLOAT, Load32FTo16F<1> ); + InsertLoadFunction(&map, GL_R8UI, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 1> ); + InsertLoadFunction(&map, GL_R8I, GL_BYTE, LoadToNative<GLbyte, 1> ); + InsertLoadFunction(&map, GL_R16UI, GL_UNSIGNED_SHORT, LoadToNative<GLushort, 1> ); + InsertLoadFunction(&map, GL_R16I, GL_SHORT, LoadToNative<GLshort, 1> ); + InsertLoadFunction(&map, GL_R32UI, GL_UNSIGNED_INT, LoadToNative<GLuint, 1> ); + InsertLoadFunction(&map, GL_R32I, GL_INT, LoadToNative<GLint, 1> ); + InsertLoadFunction(&map, GL_DEPTH_COMPONENT16, GL_UNSIGNED_SHORT, LoadToNative<GLushort, 1> ); + InsertLoadFunction(&map, GL_DEPTH_COMPONENT24, GL_UNSIGNED_INT, LoadR32ToR24G8 ); + InsertLoadFunction(&map, GL_DEPTH_COMPONENT16, GL_UNSIGNED_INT, LoadR32ToR16 ); + InsertLoadFunction(&map, GL_DEPTH_COMPONENT32F, GL_FLOAT, LoadToNative<GLfloat, 1> ); + InsertLoadFunction(&map, GL_DEPTH24_STENCIL8, GL_UNSIGNED_INT_24_8, LoadR32ToR24G8 ); + InsertLoadFunction(&map, GL_DEPTH32F_STENCIL8, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, LoadToNative<GLuint, 2> ); + + // Unsized formats + // Load functions are unreachable because they are converted to sized internal formats based on + // the format and type before loading takes place. + InsertLoadFunction(&map, GL_RGBA, GL_UNSIGNED_BYTE, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_RGBA, GL_UNSIGNED_SHORT_5_5_5_1, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_RGB, GL_UNSIGNED_BYTE, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_RGB, GL_UNSIGNED_SHORT_5_6_5, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA, GL_UNSIGNED_BYTE, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_LUMINANCE, GL_UNSIGNED_BYTE, UnreachableLoadFunction ); + InsertLoadFunction(&map, GL_ALPHA, GL_UNSIGNED_BYTE, UnreachableLoadFunction ); + + // From GL_OES_texture_float + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA, GL_FLOAT, LoadLA32FToRGBA32F ); + InsertLoadFunction(&map, GL_LUMINANCE, GL_FLOAT, LoadL32FToRGBA32F ); + InsertLoadFunction(&map, GL_ALPHA, GL_FLOAT, LoadA32FToRGBA32F ); + + // From GL_OES_texture_half_float + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA, GL_HALF_FLOAT, LoadLA16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA, GL_HALF_FLOAT_OES, LoadLA16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE, GL_HALF_FLOAT, LoadL16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE, GL_HALF_FLOAT_OES, LoadL16FToRGBA16F ); + InsertLoadFunction(&map, GL_ALPHA, GL_HALF_FLOAT, LoadA16FToRGBA16F ); + InsertLoadFunction(&map, GL_ALPHA, GL_HALF_FLOAT_OES, LoadA16FToRGBA16F ); + + // From GL_EXT_texture_storage + // GL_ALPHA8_EXT GL_UNSIGNED_BYTE is in the feature-level-specific load function maps, due to differences between 9_3 and 10_0+ + InsertLoadFunction(&map, GL_LUMINANCE8_EXT, GL_UNSIGNED_BYTE, LoadL8ToRGBA8 ); + InsertLoadFunction(&map, GL_LUMINANCE8_ALPHA8_EXT, GL_UNSIGNED_BYTE, LoadLA8ToRGBA8 ); + InsertLoadFunction(&map, GL_ALPHA32F_EXT, GL_FLOAT, LoadA32FToRGBA32F ); + InsertLoadFunction(&map, GL_LUMINANCE32F_EXT, GL_FLOAT, LoadL32FToRGBA32F ); + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA32F_EXT, GL_FLOAT, LoadLA32FToRGBA32F ); + InsertLoadFunction(&map, GL_ALPHA16F_EXT, GL_HALF_FLOAT, LoadA16FToRGBA16F ); + InsertLoadFunction(&map, GL_ALPHA16F_EXT, GL_HALF_FLOAT_OES, LoadA16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE16F_EXT, GL_HALF_FLOAT, LoadL16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE16F_EXT, GL_HALF_FLOAT_OES, LoadL16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA16F_EXT, GL_HALF_FLOAT, LoadLA16FToRGBA16F ); + InsertLoadFunction(&map, GL_LUMINANCE_ALPHA16F_EXT, GL_HALF_FLOAT_OES, LoadLA16FToRGBA16F ); + + // From GL_ANGLE_depth_texture + InsertLoadFunction(&map, GL_DEPTH_COMPONENT32_OES, GL_UNSIGNED_INT, LoadR32ToR24G8 ); + + // From GL_EXT_texture_format_BGRA8888 + InsertLoadFunction(&map, GL_BGRA8_EXT, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_BGRA4_ANGLEX, GL_UNSIGNED_SHORT_4_4_4_4_REV_EXT, LoadRGBA4ToRGBA8 ); + InsertLoadFunction(&map, GL_BGRA4_ANGLEX, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + InsertLoadFunction(&map, GL_BGR5_A1_ANGLEX, GL_UNSIGNED_SHORT_1_5_5_5_REV_EXT, LoadRGB5A1ToRGBA8 ); + InsertLoadFunction(&map, GL_BGR5_A1_ANGLEX, GL_UNSIGNED_BYTE, LoadToNative<GLubyte, 4> ); + + // Compressed formats + // From ES 3.0.1 spec, table 3.16 + // | Internal format | Type | Load function | + InsertLoadFunction(&map, GL_COMPRESSED_R11_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_R11_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_SIGNED_R11_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_RG11_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_SIGNED_RG11_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_RGB8_ETC2, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_SRGB8_ETC2, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_SRGB8_PUNCHTHROUGH_ALPHA1_ETC2, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_RGBA8_ETC2_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + InsertLoadFunction(&map, GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC, GL_UNSIGNED_BYTE, UnimplementedLoadFunction ); + + // From GL_EXT_texture_compression_dxt1 + InsertLoadFunction(&map, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_UNSIGNED_BYTE, LoadCompressedToNative<4, 4, 8>); + InsertLoadFunction(&map, GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, GL_UNSIGNED_BYTE, LoadCompressedToNative<4, 4, 8>); + + // From GL_ANGLE_texture_compression_dxt3 + InsertLoadFunction(&map, GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE, GL_UNSIGNED_BYTE, LoadCompressedToNative<4, 4, 16>); + + // From GL_ANGLE_texture_compression_dxt5 + InsertLoadFunction(&map, GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE, GL_UNSIGNED_BYTE, LoadCompressedToNative<4, 4, 16>); + + return map; +} + +// For sized GL internal formats, there is only one corresponding D3D11 format. This map type allows +// querying for the DXGI texture formats to use for textures, SRVs, RTVs and DSVs given a GL internal +// format. +typedef std::map<GLenum, TextureFormat> D3D11ES3FormatMap; + +TextureFormat::TextureFormat() + : texFormat(DXGI_FORMAT_UNKNOWN), + srvFormat(DXGI_FORMAT_UNKNOWN), + rtvFormat(DXGI_FORMAT_UNKNOWN), + dsvFormat(DXGI_FORMAT_UNKNOWN), + renderFormat(DXGI_FORMAT_UNKNOWN), + swizzleTexFormat(DXGI_FORMAT_UNKNOWN), + swizzleSRVFormat(DXGI_FORMAT_UNKNOWN), + swizzleRTVFormat(DXGI_FORMAT_UNKNOWN), + dataInitializerFunction(NULL), + loadFunctions() +{ +} + +static inline void InsertD3D11FormatInfoBase(D3D11ES3FormatMap *formatMap, const D3D11LoadFunctionMap &flLoadFunctions, GLenum internalFormat, DXGI_FORMAT texFormat, + DXGI_FORMAT srvFormat, DXGI_FORMAT rtvFormat, DXGI_FORMAT dsvFormat) +{ + TextureFormat info; + info.texFormat = texFormat; + info.srvFormat = srvFormat; + info.rtvFormat = rtvFormat; + info.dsvFormat = dsvFormat; + + // Given a GL internal format, the renderFormat is the DSV format if it is depth- or stencil-renderable, + // the RTV format if it is color-renderable, and the (nonrenderable) texture format otherwise. + if (dsvFormat != DXGI_FORMAT_UNKNOWN) + { + info.renderFormat = dsvFormat; + } + else if (rtvFormat != DXGI_FORMAT_UNKNOWN) + { + info.renderFormat = rtvFormat; + } + else if (texFormat != DXGI_FORMAT_UNKNOWN) + { + info.renderFormat = texFormat; + } + else + { + info.renderFormat = DXGI_FORMAT_UNKNOWN; + } + + // Compute the swizzle formats + const gl::InternalFormat &formatInfo = gl::GetInternalFormatInfo(internalFormat); + if (internalFormat != GL_NONE && formatInfo.pixelBytes > 0) + { + if (formatInfo.componentCount != 4 || texFormat == DXGI_FORMAT_UNKNOWN || + srvFormat == DXGI_FORMAT_UNKNOWN || rtvFormat == DXGI_FORMAT_UNKNOWN) + { + // Get the maximum sized component + unsigned int maxBits = 1; + if (formatInfo.compressed) + { + unsigned int compressedBitsPerBlock = formatInfo.pixelBytes * 8; + unsigned int blockSize = formatInfo.compressedBlockWidth * formatInfo.compressedBlockHeight; + maxBits = std::max(compressedBitsPerBlock / blockSize, maxBits); + } + else + { + maxBits = std::max(maxBits, formatInfo.alphaBits); + maxBits = std::max(maxBits, formatInfo.redBits); + maxBits = std::max(maxBits, formatInfo.greenBits); + maxBits = std::max(maxBits, formatInfo.blueBits); + maxBits = std::max(maxBits, formatInfo.luminanceBits); + maxBits = std::max(maxBits, formatInfo.depthBits); + } + + maxBits = roundUp(maxBits, 8U); + + static const SwizzleInfoMap swizzleMap = BuildSwizzleInfoMap(); + SwizzleInfoMap::const_iterator swizzleIter = swizzleMap.find(SwizzleSizeType(maxBits, formatInfo.componentType)); + ASSERT(swizzleIter != swizzleMap.end()); + + const SwizzleFormatInfo &swizzleInfo = swizzleIter->second; + info.swizzleTexFormat = swizzleInfo.mTexFormat; + info.swizzleSRVFormat = swizzleInfo.mSRVFormat; + info.swizzleRTVFormat = swizzleInfo.mRTVFormat; + } + else + { + // The original texture format is suitable for swizzle operations + info.swizzleTexFormat = texFormat; + info.swizzleSRVFormat = srvFormat; + info.swizzleRTVFormat = rtvFormat; + } + } + else + { + // Not possible to swizzle with this texture format since it is either unsized or GL_NONE + info.swizzleTexFormat = DXGI_FORMAT_UNKNOWN; + info.swizzleSRVFormat = DXGI_FORMAT_UNKNOWN; + info.swizzleRTVFormat = DXGI_FORMAT_UNKNOWN; + } + + // Check if there is an initialization function for this texture format + static const InternalFormatInitializerMap initializerMap = BuildInternalFormatInitializerMap(); + InternalFormatInitializerMap::const_iterator initializerIter = initializerMap.find(internalFormat); + info.dataInitializerFunction = (initializerIter != initializerMap.end()) ? initializerIter->second : NULL; + + // Gather all the load functions for this internal format from the base list + static const D3D11LoadFunctionMap loadFunctions = BuildBaseD3D11LoadFunctionMap(); + D3D11LoadFunctionMap::const_iterator loadFunctionIter = loadFunctions.find(internalFormat); + if (loadFunctionIter != loadFunctions.end()) + { + const std::vector<TypeLoadFunctionPair> &loadFunctionVector = loadFunctionIter->second; + for (size_t i = 0; i < loadFunctionVector.size(); i++) + { + GLenum type = loadFunctionVector[i].first; + LoadImageFunction function = loadFunctionVector[i].second; + info.loadFunctions.insert(std::make_pair(type, function)); + } + } + + // Gather load functions for this internal format from the feature-level-specific list + D3D11LoadFunctionMap::const_iterator flLoadFunctionIter = flLoadFunctions.find(internalFormat); + if (flLoadFunctionIter != flLoadFunctions.end()) + { + const std::vector<TypeLoadFunctionPair> &flLoadFunctionVector = flLoadFunctionIter->second; + for (size_t i = 0; i < flLoadFunctionVector.size(); i++) + { + GLenum type = flLoadFunctionVector[i].first; + LoadImageFunction function = flLoadFunctionVector[i].second; + info.loadFunctions.insert(std::make_pair(type, function)); + } + } + + formatMap->insert(std::make_pair(internalFormat, info)); +} + +static inline void InsertD3D11_FL9_3_FormatInfo(D3D11ES3FormatMap *map, GLenum internalFormat, DXGI_FORMAT texFormat, + DXGI_FORMAT srvFormat, DXGI_FORMAT rtvFormat, DXGI_FORMAT dsvFormat) +{ + static const D3D11LoadFunctionMap flLoadFunctions = BuildD3D11_FL9_3_LoadFunctionMap(); + InsertD3D11FormatInfoBase(map, flLoadFunctions, internalFormat, texFormat, srvFormat, rtvFormat, dsvFormat); +} + +static inline void InsertD3D11FormatInfo(D3D11ES3FormatMap *map, GLenum internalFormat, DXGI_FORMAT texFormat, + DXGI_FORMAT srvFormat, DXGI_FORMAT rtvFormat, DXGI_FORMAT dsvFormat) +{ + static const D3D11LoadFunctionMap flLoadFunctions = BuildD3D11_FL10_0Plus_LoadFunctionMap(); + InsertD3D11FormatInfoBase(map, flLoadFunctions, internalFormat, texFormat, srvFormat, rtvFormat, dsvFormat); +} + +static D3D11ES3FormatMap BuildD3D11_FL9_3FormatOverrideMap() +{ + // D3D11 Feature Level 9_3 doesn't support as many texture formats as Feature Level 10_0+. + // In particular, it doesn't support: + // - mipmaps on DXGI_FORMAT_A8_NORM + // - *_TYPELESS formats + // - DXGI_FORMAT_D32_FLOAT_S8X24_UINT or DXGI_FORMAT_D32_FLOAT + + D3D11ES3FormatMap map; + + // | GL internal format | D3D11 texture format | D3D11 SRV format | D3D11 RTV format | D3D11 DSV format + InsertD3D11_FL9_3_FormatInfo(&map, GL_ALPHA, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11_FL9_3_FormatInfo(&map, GL_ALPHA8_EXT, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11_FL9_3_FormatInfo(&map, GL_DEPTH_COMPONENT16, DXGI_FORMAT_D16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D16_UNORM); + InsertD3D11_FL9_3_FormatInfo(&map, GL_DEPTH_COMPONENT24, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT); + InsertD3D11_FL9_3_FormatInfo(&map, GL_DEPTH_COMPONENT32F, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11_FL9_3_FormatInfo(&map, GL_DEPTH24_STENCIL8, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT); + InsertD3D11_FL9_3_FormatInfo(&map, GL_DEPTH32F_STENCIL8, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11_FL9_3_FormatInfo(&map, GL_STENCIL_INDEX8, DXGI_FORMAT_D24_UNORM_S8_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT); + + return map; +} + +static D3D11ES3FormatMap BuildD3D11FormatMap() +{ + D3D11ES3FormatMap map; + + // | GL internal format | D3D11 texture format | D3D11 SRV format | D3D11 RTV format | D3D11 DSV format | + InsertD3D11FormatInfo(&map, GL_NONE, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R8, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_R8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R8_SNORM, DXGI_FORMAT_R8_SNORM, DXGI_FORMAT_R8_SNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG8, DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_R8G8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG8_SNORM, DXGI_FORMAT_R8G8_SNORM, DXGI_FORMAT_R8G8_SNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB8, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB565, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA4, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB5_A1, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA8, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_R8G8B8A8_SNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB10_A2, DXGI_FORMAT_R10G10B10A2_UNORM, DXGI_FORMAT_R10G10B10A2_UNORM, DXGI_FORMAT_R10G10B10A2_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB10_A2UI, DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_R10G10B10A2_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_SRGB8, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_SRGB8_ALPHA8, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, DXGI_FORMAT_R8G8B8A8_UNORM_SRGB, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R16F, DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_R16_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG16F, DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_R16G16_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB16F, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA16F, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R32F, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG32F, DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_R32G32_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB32F, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA32F, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R11F_G11F_B10F, DXGI_FORMAT_R11G11B10_FLOAT, DXGI_FORMAT_R11G11B10_FLOAT, DXGI_FORMAT_R11G11B10_FLOAT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB9_E5, DXGI_FORMAT_R9G9B9E5_SHAREDEXP, DXGI_FORMAT_R9G9B9E5_SHAREDEXP, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R8I, DXGI_FORMAT_R8_SINT, DXGI_FORMAT_R8_SINT, DXGI_FORMAT_R8_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R8UI, DXGI_FORMAT_R8_UINT, DXGI_FORMAT_R8_UINT, DXGI_FORMAT_R8_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R16I, DXGI_FORMAT_R16_SINT, DXGI_FORMAT_R16_SINT, DXGI_FORMAT_R16_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R16UI, DXGI_FORMAT_R16_UINT, DXGI_FORMAT_R16_UINT, DXGI_FORMAT_R16_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R32I, DXGI_FORMAT_R32_SINT, DXGI_FORMAT_R32_SINT, DXGI_FORMAT_R32_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_R32UI, DXGI_FORMAT_R32_UINT, DXGI_FORMAT_R32_UINT, DXGI_FORMAT_R32_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG8I, DXGI_FORMAT_R8G8_SINT, DXGI_FORMAT_R8G8_SINT, DXGI_FORMAT_R8G8_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG8UI, DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_R8G8_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG16I, DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_R16G16_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG16UI, DXGI_FORMAT_R16G16_UINT, DXGI_FORMAT_R16G16_UINT, DXGI_FORMAT_R16G16_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG32I, DXGI_FORMAT_R32G32_SINT, DXGI_FORMAT_R32G32_SINT, DXGI_FORMAT_R32G32_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RG32UI, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_R32G32_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB8I, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB8UI, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB16I, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB16UI, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB32I, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB32UI, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA8I, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_R8G8B8A8_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA8UI, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_R8G8B8A8_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA16I, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_R16G16B16A16_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA16UI, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_R16G16B16A16_UINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA32I, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_R32G32B32A32_SINT, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA32UI, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_R32G32B32A32_UINT, DXGI_FORMAT_UNKNOWN); + + // Unsized formats, TODO: Are types of float and half float allowed for the unsized types? Would it change the DXGI format? + InsertD3D11FormatInfo(&map, GL_ALPHA, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_LUMINANCE, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_LUMINANCE_ALPHA, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGB, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_RGBA, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_BGRA_EXT, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_UNKNOWN); + + // From GL_EXT_texture_storage + // | GL internal format | D3D11 texture format | D3D11 SRV format | D3D11 RTV format | D3D11 DSV format | + InsertD3D11FormatInfo(&map, GL_ALPHA8_EXT, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_A8_UNORM, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE8_EXT, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_ALPHA32F_EXT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE32F_EXT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_ALPHA16F_EXT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE16F_EXT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE8_ALPHA8_EXT, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_R8G8B8A8_UNORM, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE_ALPHA32F_EXT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_R32G32B32A32_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_LUMINANCE_ALPHA16F_EXT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_R16G16B16A16_FLOAT, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_BGRA8_EXT, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_BGRA4_ANGLEX, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_UNKNOWN ); + InsertD3D11FormatInfo(&map, GL_BGR5_A1_ANGLEX, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_B8G8R8A8_UNORM, DXGI_FORMAT_UNKNOWN ); + + // Depth stencil formats + InsertD3D11FormatInfo(&map, GL_DEPTH_COMPONENT16, DXGI_FORMAT_R16_TYPELESS, DXGI_FORMAT_R16_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D16_UNORM ); + InsertD3D11FormatInfo(&map, GL_DEPTH_COMPONENT24, DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT ); + InsertD3D11FormatInfo(&map, GL_DEPTH_COMPONENT32F, DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_R32_FLOAT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D32_FLOAT ); + InsertD3D11FormatInfo(&map, GL_DEPTH24_STENCIL8, DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT ); + InsertD3D11FormatInfo(&map, GL_DEPTH32F_STENCIL8, DXGI_FORMAT_R32G8X24_TYPELESS, DXGI_FORMAT_R32_FLOAT_X8X24_TYPELESS, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D32_FLOAT_S8X24_UINT); + InsertD3D11FormatInfo(&map, GL_STENCIL_INDEX8, DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_X24_TYPELESS_G8_UINT, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT ); + + // From GL_ANGLE_depth_texture + // Since D3D11 doesn't have a D32_UNORM format, use D24S8 which has comparable precision and matches the ES3 format. + InsertD3D11FormatInfo(&map, GL_DEPTH_COMPONENT32_OES, DXGI_FORMAT_R24G8_TYPELESS, DXGI_FORMAT_R24_UNORM_X8_TYPELESS, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_D24_UNORM_S8_UINT); + + // Compressed formats, From ES 3.0.1 spec, table 3.16 + // | GL internal format | D3D11 texture format | D3D11 SRV format | D3D11 RTV format | D3D11 DSV format | + InsertD3D11FormatInfo(&map, GL_COMPRESSED_R11_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_SIGNED_R11_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RG11_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_SIGNED_RG11_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGB8_ETC2, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_SRGB8_ETC2, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGB8_PUNCHTHROUGH_ALPHA1_ETC2, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_SRGB8_PUNCHTHROUGH_ALPHA1_ETC2, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGBA8_ETC2_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_SRGB8_ALPHA8_ETC2_EAC, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + + // From GL_EXT_texture_compression_dxt1 + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGB_S3TC_DXT1_EXT, DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGBA_S3TC_DXT1_EXT, DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_BC1_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + + // From GL_ANGLE_texture_compression_dxt3 + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGBA_S3TC_DXT3_ANGLE, DXGI_FORMAT_BC2_UNORM, DXGI_FORMAT_BC2_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + + // From GL_ANGLE_texture_compression_dxt5 + InsertD3D11FormatInfo(&map, GL_COMPRESSED_RGBA_S3TC_DXT5_ANGLE, DXGI_FORMAT_BC3_UNORM, DXGI_FORMAT_BC3_UNORM, DXGI_FORMAT_UNKNOWN, DXGI_FORMAT_UNKNOWN); + + return map; +} + +const TextureFormat &GetTextureFormatInfo(GLenum internalFormat, D3D_FEATURE_LEVEL featureLevel) +{ + static const D3D11ES3FormatMap formatMap = BuildD3D11FormatMap(); + static const D3D11ES3FormatMap formatMapFL9_3Override = BuildD3D11_FL9_3FormatOverrideMap(); + + if (featureLevel == D3D_FEATURE_LEVEL_9_3) + { + // First see if the internalFormat has a special map for FL9_3 + D3D11ES3FormatMap::const_iterator fl9_3Iter = formatMapFL9_3Override.find(internalFormat); + if (fl9_3Iter != formatMapFL9_3Override.end()) + { + return fl9_3Iter->second; + } + } + + D3D11ES3FormatMap::const_iterator iter = formatMap.find(internalFormat); + if (iter != formatMap.end()) + { + return iter->second; + } + else + { + static const TextureFormat defaultInfo; + return defaultInfo; + } +} + +typedef std::map<gl::VertexFormat, VertexFormat> D3D11VertexFormatInfoMap; +typedef std::pair<gl::VertexFormat, VertexFormat> D3D11VertexFormatPair; + +VertexFormat::VertexFormat() + : conversionType(VERTEX_CONVERT_NONE), + nativeFormat(DXGI_FORMAT_UNKNOWN), + copyFunction(NULL) +{ +} + +static void AddVertexFormatInfo(D3D11VertexFormatInfoMap *map, GLenum inputType, GLboolean normalized, GLuint componentCount, + VertexConversionType conversionType, DXGI_FORMAT nativeFormat, VertexCopyFunction copyFunction) +{ + gl::VertexFormat inputFormat(inputType, normalized, componentCount, false); + + VertexFormat info; + info.conversionType = conversionType; + info.nativeFormat = nativeFormat; + info.copyFunction = copyFunction; + + map->insert(D3D11VertexFormatPair(inputFormat, info)); +} + +static void AddIntegerVertexFormatInfo(D3D11VertexFormatInfoMap *map, GLenum inputType, GLuint componentCount, + VertexConversionType conversionType, DXGI_FORMAT nativeFormat, VertexCopyFunction copyFunction) +{ + gl::VertexFormat inputFormat(inputType, GL_FALSE, componentCount, true); + + VertexFormat info; + info.conversionType = conversionType; + info.nativeFormat = nativeFormat; + info.copyFunction = copyFunction; + + map->insert(D3D11VertexFormatPair(inputFormat, info)); +} + +static D3D11VertexFormatInfoMap BuildD3D11_FL9_3VertexFormatInfoOverrideMap() +{ + // D3D11 Feature Level 9_3 doesn't support as many formats for vertex buffer resource as Feature Level 10_0+. + // http://msdn.microsoft.com/en-us/library/windows/desktop/ff471324(v=vs.85).aspx + + D3D11VertexFormatInfoMap map; + + // GL_BYTE -- unnormalized + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 1, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16_SINT, &Copy8SintTo16SintVertexData<1, 2>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 2, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16_SINT, &Copy8SintTo16SintVertexData<2, 2>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 3, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16B16A16_SINT, &Copy8SintTo16SintVertexData<3, 4>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 4, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16B16A16_SINT, &Copy8SintTo16SintVertexData<4, 4>); + + // GL_BYTE -- normalized + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16_SNORM, &Copy8SnormTo16SnormVertexData<1, 2>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16_SNORM, &Copy8SnormTo16SnormVertexData<2, 2>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_SNORM, &Copy8SnormTo16SnormVertexData<3, 4>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_SNORM, &Copy8SnormTo16SnormVertexData<4, 4>); + + // GL_UNSIGNED_BYTE -- unnormalized + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 1, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 1, 4, 1>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 2, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 2, 4, 1>); + // NOTE: 3 and 4 component unnormalized GL_UNSIGNED_BYTE should use the default format table. + + // GL_UNSIGNED_BYTE -- normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_UNORM, &CopyNativeVertexData<GLubyte, 1, 4, UINT8_MAX>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_UNORM, &CopyNativeVertexData<GLubyte, 2, 4, UINT8_MAX>); + // NOTE: 3 and 4 component normalized GL_UNSIGNED_BYTE should use the default format table. + + // GL_SHORT -- unnormalized + AddVertexFormatInfo(&map, GL_SHORT, GL_FALSE, 1, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16_SINT, &CopyNativeVertexData<GLshort, 1, 2, 0>); + // NOTE: 2, 3 and 4 component unnormalized GL_SHORT should use the default format table. + + // GL_SHORT -- normalized + AddVertexFormatInfo(&map, GL_SHORT, GL_TRUE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16_SNORM, &CopyNativeVertexData<GLshort, 1, 2, 0>); + // NOTE: 2, 3 and 4 component normalized GL_SHORT should use the default format table. + + // GL_UNSIGNED_SHORT -- unnormalized + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLushort, 1, 2, false>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLushort, 2, 2, false>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32_FLOAT, &CopyTo32FVertexData<GLushort, 3, 3, false>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyTo32FVertexData<GLushort, 4, 4, false>); + + // GL_UNSIGNED_SHORT -- normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLushort, 1, 2, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLushort, 2, 2, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32_FLOAT, &CopyTo32FVertexData<GLushort, 3, 3, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyTo32FVertexData<GLushort, 4, 4, true>); + + // GL_FIXED + // TODO: Add test to verify that this works correctly. + AddVertexFormatInfo(&map, GL_FIXED, GL_FALSE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &Copy32FixedTo32FVertexData<1, 2>); + // NOTE: 2, 3 and 4 component GL_FIXED should use the default format table. + + // GL_FLOAT + // TODO: Add test to verify that this works correctly. + AddVertexFormatInfo(&map, GL_FLOAT, GL_FALSE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyNativeVertexData<GLfloat, 1, 2, 0>); + // NOTE: 2, 3 and 4 component GL_FLOAT should use the default format table. + + return map; +} + +static D3D11VertexFormatInfoMap BuildD3D11VertexFormatInfoMap() +{ + D3D11VertexFormatInfoMap map; + + // TODO: column legend + + // + // Float formats + // + + // GL_BYTE -- un-normalized + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8_SINT, &CopyNativeVertexData<GLbyte, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8G8_SINT, &CopyNativeVertexData<GLbyte, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 3, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R8G8B8A8_SINT, &CopyNativeVertexData<GLbyte, 3, 4, 1>); + AddVertexFormatInfo(&map, GL_BYTE, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8G8B8A8_SINT, &CopyNativeVertexData<GLbyte, 4, 4, 0>); + + // GL_BYTE -- normalized + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8_SNORM, &CopyNativeVertexData<GLbyte, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8_SNORM, &CopyNativeVertexData<GLbyte, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_SNORM, &CopyNativeVertexData<GLbyte, 3, 4, INT8_MAX>); + AddVertexFormatInfo(&map, GL_BYTE, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8B8A8_SNORM, &CopyNativeVertexData<GLbyte, 4, 4, 0>); + + // GL_UNSIGNED_BYTE -- un-normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8_UINT, &CopyNativeVertexData<GLubyte, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8G8_UINT, &CopyNativeVertexData<GLubyte, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 3, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 3, 4, 1>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 4, 4, 0>); + + // GL_UNSIGNED_BYTE -- normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8_UNORM, &CopyNativeVertexData<GLubyte, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8_UNORM, &CopyNativeVertexData<GLubyte, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_UNORM, &CopyNativeVertexData<GLubyte, 3, 4, UINT8_MAX>); + AddVertexFormatInfo(&map, GL_UNSIGNED_BYTE, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8B8A8_UNORM, &CopyNativeVertexData<GLubyte, 4, 4, 0>); + + // GL_SHORT -- un-normalized + AddVertexFormatInfo(&map, GL_SHORT, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16_SINT, &CopyNativeVertexData<GLshort, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_SHORT, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16G16_SINT, &CopyNativeVertexData<GLshort, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_SHORT, GL_FALSE, 3, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16B16A16_SINT, &CopyNativeVertexData<GLshort, 3, 4, 1>); + AddVertexFormatInfo(&map, GL_SHORT, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16G16B16A16_SINT, &CopyNativeVertexData<GLshort, 4, 4, 0>); + + // GL_SHORT -- normalized + AddVertexFormatInfo(&map, GL_SHORT, GL_TRUE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16_SNORM, &CopyNativeVertexData<GLshort, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_SHORT, GL_TRUE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16_SNORM, &CopyNativeVertexData<GLshort, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_SHORT, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_SNORM, &CopyNativeVertexData<GLshort, 3, 4, INT16_MAX>); + AddVertexFormatInfo(&map, GL_SHORT, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16B16A16_SNORM, &CopyNativeVertexData<GLshort, 4, 4, 0>); + + // GL_UNSIGNED_SHORT -- un-normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16_UINT, &CopyNativeVertexData<GLushort, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16G16_UINT, &CopyNativeVertexData<GLushort, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 3, VERTEX_CONVERT_BOTH, DXGI_FORMAT_R16G16B16A16_UINT, &CopyNativeVertexData<GLushort, 3, 4, 1>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R16G16B16A16_UINT, &CopyNativeVertexData<GLushort, 4, 4, 0>); + + // GL_UNSIGNED_SHORT -- normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16_UNORM, &CopyNativeVertexData<GLushort, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16_UNORM, &CopyNativeVertexData<GLushort, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_UNORM, &CopyNativeVertexData<GLushort, 3, 4, UINT16_MAX>); + AddVertexFormatInfo(&map, GL_UNSIGNED_SHORT, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16B16A16_UNORM, &CopyNativeVertexData<GLushort, 4, 4, 0>); + + // GL_INT -- un-normalized + AddVertexFormatInfo(&map, GL_INT, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32_SINT, &CopyNativeVertexData<GLint, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_INT, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32_SINT, &CopyNativeVertexData<GLint, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_INT, GL_FALSE, 3, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32B32_SINT, &CopyNativeVertexData<GLint, 3, 3, 0>); + AddVertexFormatInfo(&map, GL_INT, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32B32A32_SINT, &CopyNativeVertexData<GLint, 4, 4, 0>); + + // GL_INT -- normalized + AddVertexFormatInfo(&map, GL_INT, GL_TRUE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32_FLOAT, &CopyTo32FVertexData<GLint, 1, 1, true>); + AddVertexFormatInfo(&map, GL_INT, GL_TRUE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLint, 2, 2, true>); + AddVertexFormatInfo(&map, GL_INT, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32_FLOAT, &CopyTo32FVertexData<GLint, 3, 3, true>); + AddVertexFormatInfo(&map, GL_INT, GL_TRUE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyTo32FVertexData<GLint, 4, 4, true>); + + // GL_UNSIGNED_INT -- un-normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_FALSE, 1, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32_UINT, &CopyNativeVertexData<GLuint, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_FALSE, 2, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32_UINT, &CopyNativeVertexData<GLuint, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_FALSE, 3, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32B32_UINT, &CopyNativeVertexData<GLuint, 3, 3, 0>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_FALSE, 4, VERTEX_CONVERT_GPU, DXGI_FORMAT_R32G32B32A32_UINT, &CopyNativeVertexData<GLuint, 4, 4, 0>); + + // GL_UNSIGNED_INT -- normalized + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_TRUE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32_FLOAT, &CopyTo32FVertexData<GLuint, 1, 1, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_TRUE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32_FLOAT, &CopyTo32FVertexData<GLuint, 2, 2, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_TRUE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32_FLOAT, &CopyTo32FVertexData<GLuint, 3, 3, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyTo32FVertexData<GLuint, 4, 4, true>); + + // GL_FIXED + AddVertexFormatInfo(&map, GL_FIXED, GL_FALSE, 1, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32_FLOAT, &Copy32FixedTo32FVertexData<1, 1>); + AddVertexFormatInfo(&map, GL_FIXED, GL_FALSE, 2, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32_FLOAT, &Copy32FixedTo32FVertexData<2, 2>); + AddVertexFormatInfo(&map, GL_FIXED, GL_FALSE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32_FLOAT, &Copy32FixedTo32FVertexData<3, 3>); + AddVertexFormatInfo(&map, GL_FIXED, GL_FALSE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &Copy32FixedTo32FVertexData<4, 4>); + + // GL_HALF_FLOAT + AddVertexFormatInfo(&map, GL_HALF_FLOAT, GL_FALSE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16_FLOAT, &CopyNativeVertexData<GLhalf, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_HALF_FLOAT, GL_FALSE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16_FLOAT, &CopyNativeVertexData<GLhalf, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_HALF_FLOAT, GL_FALSE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_FLOAT, &CopyNativeVertexData<GLhalf, 3, 4, gl::Float16One>); + AddVertexFormatInfo(&map, GL_HALF_FLOAT, GL_FALSE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16B16A16_FLOAT, &CopyNativeVertexData<GLhalf, 4, 4, 0>); + + // GL_FLOAT + AddVertexFormatInfo(&map, GL_FLOAT, GL_FALSE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32_FLOAT, &CopyNativeVertexData<GLfloat, 1, 1, 0>); + AddVertexFormatInfo(&map, GL_FLOAT, GL_FALSE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32_FLOAT, &CopyNativeVertexData<GLfloat, 2, 2, 0>); + AddVertexFormatInfo(&map, GL_FLOAT, GL_FALSE, 3, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32_FLOAT, &CopyNativeVertexData<GLfloat, 3, 3, 0>); + AddVertexFormatInfo(&map, GL_FLOAT, GL_FALSE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyNativeVertexData<GLfloat, 4, 4, 0>); + + // GL_INT_2_10_10_10_REV + AddVertexFormatInfo(&map, GL_INT_2_10_10_10_REV, GL_FALSE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyXYZ10W2ToXYZW32FVertexData<true, false, true>); + AddVertexFormatInfo(&map, GL_INT_2_10_10_10_REV, GL_TRUE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyXYZ10W2ToXYZW32FVertexData<true, true, true>); + + // GL_UNSIGNED_INT_2_10_10_10_REV + AddVertexFormatInfo(&map, GL_UNSIGNED_INT_2_10_10_10_REV, GL_FALSE, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R32G32B32A32_FLOAT, &CopyXYZ10W2ToXYZW32FVertexData<false, false, true>); + AddVertexFormatInfo(&map, GL_UNSIGNED_INT_2_10_10_10_REV, GL_TRUE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R10G10B10A2_UNORM, &CopyNativeVertexData<GLuint, 1, 1, 0>); + + // + // Integer Formats + // + + // GL_BYTE + AddIntegerVertexFormatInfo(&map, GL_BYTE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8_SINT, &CopyNativeVertexData<GLbyte, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_BYTE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8_SINT, &CopyNativeVertexData<GLbyte, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_BYTE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_SINT, &CopyNativeVertexData<GLbyte, 3, 4, 1>); + AddIntegerVertexFormatInfo(&map, GL_BYTE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8B8A8_SINT, &CopyNativeVertexData<GLbyte, 4, 4, 0>); + + // GL_UNSIGNED_BYTE + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_BYTE, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8_UINT, &CopyNativeVertexData<GLubyte, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_BYTE, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8_UINT, &CopyNativeVertexData<GLubyte, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_BYTE, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 3, 4, 1>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_BYTE, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R8G8B8A8_UINT, &CopyNativeVertexData<GLubyte, 4, 4, 0>); + + // GL_SHORT + AddIntegerVertexFormatInfo(&map, GL_SHORT, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16_SINT, &CopyNativeVertexData<GLshort, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_SHORT, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16_SINT, &CopyNativeVertexData<GLshort, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_SHORT, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_SINT, &CopyNativeVertexData<GLshort, 3, 4, 1>); + AddIntegerVertexFormatInfo(&map, GL_SHORT, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16B16A16_SINT, &CopyNativeVertexData<GLshort, 4, 4, 0>); + + // GL_UNSIGNED_SHORT + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_SHORT, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16_UINT, &CopyNativeVertexData<GLushort, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_SHORT, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16_UINT, &CopyNativeVertexData<GLushort, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_SHORT, 3, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_UINT, &CopyNativeVertexData<GLushort, 3, 4, 1>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_SHORT, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R16G16B16A16_UINT, &CopyNativeVertexData<GLushort, 4, 4, 0>); + + // GL_INT + AddIntegerVertexFormatInfo(&map, GL_INT, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32_SINT, &CopyNativeVertexData<GLint, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_INT, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32_SINT, &CopyNativeVertexData<GLint, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_INT, 3, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32_SINT, &CopyNativeVertexData<GLint, 3, 3, 0>); + AddIntegerVertexFormatInfo(&map, GL_INT, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32A32_SINT, &CopyNativeVertexData<GLint, 4, 4, 0>); + + // GL_UNSIGNED_INT + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_INT, 1, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32_SINT, &CopyNativeVertexData<GLuint, 1, 1, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_INT, 2, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32_SINT, &CopyNativeVertexData<GLuint, 2, 2, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_INT, 3, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32_SINT, &CopyNativeVertexData<GLuint, 3, 3, 0>); + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_INT, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R32G32B32A32_SINT, &CopyNativeVertexData<GLuint, 4, 4, 0>); + + // GL_INT_2_10_10_10_REV + AddIntegerVertexFormatInfo(&map, GL_INT_2_10_10_10_REV, 4, VERTEX_CONVERT_CPU, DXGI_FORMAT_R16G16B16A16_SINT, &CopyXYZ10W2ToXYZW32FVertexData<true, true, false>); + + // GL_UNSIGNED_INT_2_10_10_10_REV + AddIntegerVertexFormatInfo(&map, GL_UNSIGNED_INT_2_10_10_10_REV, 4, VERTEX_CONVERT_NONE, DXGI_FORMAT_R10G10B10A2_UINT, &CopyNativeVertexData<GLuint, 1, 1, 0>); + + return map; +} + +const VertexFormat &GetVertexFormatInfo(const gl::VertexFormat &vertexFormat, D3D_FEATURE_LEVEL featureLevel) +{ + static const D3D11VertexFormatInfoMap vertexFormatMap = BuildD3D11VertexFormatInfoMap(); + static const D3D11VertexFormatInfoMap vertexFormatMapFL9_3Override = BuildD3D11_FL9_3VertexFormatInfoOverrideMap(); + + if (featureLevel == D3D_FEATURE_LEVEL_9_3) + { + // First see if the format has a special mapping for FL9_3 + D3D11VertexFormatInfoMap::const_iterator iter = vertexFormatMapFL9_3Override.find(vertexFormat); + if (iter != vertexFormatMapFL9_3Override.end()) + { + return iter->second; + } + } + + D3D11VertexFormatInfoMap::const_iterator iter = vertexFormatMap.find(vertexFormat); + if (iter != vertexFormatMap.end()) + { + return iter->second; + } + else + { + static const VertexFormat defaultInfo; + return defaultInfo; + } +} + +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.h new file mode 100644 index 0000000000..33fe29dc39 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/formatutils11.h @@ -0,0 +1,93 @@ +// +// Copyright (c) 2013-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// formatutils11.h: Queries for GL image formats and their translations to D3D11 +// formats. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_FORMATUTILS11_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_FORMATUTILS11_H_ + +#include "libANGLE/renderer/d3d/formatutilsD3D.h" +#include "libANGLE/angletypes.h" + +#include "common/platform.h" + +#include <map> + +namespace rx +{ + +namespace d3d11 +{ + +typedef std::map<std::pair<GLenum, GLenum>, ColorCopyFunction> FastCopyFunctionMap; + +struct DXGIFormat +{ + DXGIFormat(); + + GLuint pixelBytes; + GLuint blockWidth; + GLuint blockHeight; + + GLuint redBits; + GLuint greenBits; + GLuint blueBits; + GLuint alphaBits; + GLuint sharedBits; + + GLuint depthBits; + GLuint depthOffset; + GLuint stencilBits; + GLuint stencilOffset; + + GLenum internalFormat; + GLenum componentType; + + MipGenerationFunction mipGenerationFunction; + ColorReadFunction colorReadFunction; + + FastCopyFunctionMap fastCopyFunctions; + ColorCopyFunction getFastCopyFunction(GLenum format, GLenum type) const; +}; +const DXGIFormat &GetDXGIFormatInfo(DXGI_FORMAT format); + +struct TextureFormat +{ + TextureFormat(); + + DXGI_FORMAT texFormat; + DXGI_FORMAT srvFormat; + DXGI_FORMAT rtvFormat; + DXGI_FORMAT dsvFormat; + DXGI_FORMAT renderFormat; + + DXGI_FORMAT swizzleTexFormat; + DXGI_FORMAT swizzleSRVFormat; + DXGI_FORMAT swizzleRTVFormat; + + InitializeTextureDataFunction dataInitializerFunction; + + typedef std::map<GLenum, LoadImageFunction> LoadFunctionMap; + LoadFunctionMap loadFunctions; +}; +const TextureFormat &GetTextureFormatInfo(GLenum internalFormat, D3D_FEATURE_LEVEL featureLevel); + +struct VertexFormat +{ + VertexFormat(); + + VertexConversionType conversionType; + DXGI_FORMAT nativeFormat; + VertexCopyFunction copyFunction; +}; +const VertexFormat &GetVertexFormatInfo(const gl::VertexFormat &vertexFormat, D3D_FEATURE_LEVEL featureLevel); + +} + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_FORMATUTILS11_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.cpp new file mode 100644 index 0000000000..63085f497f --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.cpp @@ -0,0 +1,1378 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer11_utils.cpp: Conversion functions and other utility routines +// specific to the D3D11 renderer. + +#include "libANGLE/renderer/d3d/d3d11/renderer11_utils.h" +#include "libANGLE/renderer/d3d/d3d11/formatutils11.h" +#include "libANGLE/renderer/d3d/d3d11/RenderTarget11.h" +#include "libANGLE/renderer/d3d/FramebufferD3D.h" +#include "libANGLE/renderer/Workarounds.h" +#include "libANGLE/formatutils.h" +#include "libANGLE/Program.h" +#include "libANGLE/Framebuffer.h" + +#include "common/debug.h" + +#include <algorithm> + +#ifndef D3D_FL9_1_DEFAULT_MAX_ANISOTROPY +# define D3D_FL9_1_DEFAULT_MAX_ANISOTROPY 2 +#endif +#ifndef D3D_FL9_1_SIMULTANEOUS_RENDER_TARGET_COUNT +# define D3D_FL9_1_SIMULTANEOUS_RENDER_TARGET_COUNT 1 +#endif +#ifndef D3D_FL9_3_SIMULTANEOUS_RENDER_TARGET_COUNT +# define D3D_FL9_3_SIMULTANEOUS_RENDER_TARGET_COUNT 4 +#endif +#ifndef D3D_FL9_1_IA_PRIMITIVE_MAX_COUNT +# define D3D_FL9_1_IA_PRIMITIVE_MAX_COUNT 65535 +#endif +#ifndef D3D_FL9_2_IA_PRIMITIVE_MAX_COUNT +# define D3D_FL9_2_IA_PRIMITIVE_MAX_COUNT 1048575 +#endif +#ifndef D3D_FL9_1_REQ_TEXTURECUBE_DIMENSION +# define D3D_FL9_1_REQ_TEXTURECUBE_DIMENSION 512 +#endif +#ifndef D3D_FL9_3_REQ_TEXTURECUBE_DIMENSION +# define D3D_FL9_3_REQ_TEXTURECUBE_DIMENSION 4096 +#endif +#ifndef D3D_FL9_1_REQ_TEXTURE2D_U_OR_V_DIMENSION +# define D3D_FL9_1_REQ_TEXTURE2D_U_OR_V_DIMENSION 2048 +#endif +#ifndef D3D_FL9_1_REQ_TEXTURE3D_U_V_OR_W_DIMENSION +# define D3D_FL9_1_REQ_TEXTURE3D_U_V_OR_W_DIMENSION 256 +#endif +#ifndef D3D_FL9_3_REQ_TEXTURE2D_U_OR_V_DIMENSION +# define D3D_FL9_3_REQ_TEXTURE2D_U_OR_V_DIMENSION 4096 +#endif +#ifndef D3D11_REQ_TEXTURECUBE_DIMENSION +# define D3D11_REQ_TEXTURECUBE_DIMENSION 16384 +#endif +#ifndef D3D11_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION +# define D3D11_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION 2048 +#endif +#ifndef D3D11_REQ_TEXTURE3D_U_V_OR_W_DIMENSION +# define D3D11_REQ_TEXTURE3D_U_V_OR_W_DIMENSION 2048 +#endif +#ifndef D3D11_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP +# define D3D11_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP 32 +#endif +#ifndef D3D11_REQ_DRAW_VERTEX_COUNT_2_TO_EXP +# define D3D11_REQ_DRAW_VERTEX_COUNT_2_TO_EXP 32 +#endif +#ifndef D3D10_1_STANDARD_VERTEX_ELEMENT_COUNT +# define D3D10_1_STANDARD_VERTEX_ELEMENT_COUNT 32 +#endif +#ifndef D3D11_STANDARD_VERTEX_ELEMENT_COUNT +# define D3D11_STANDARD_VERTEX_ELEMENT_COUNT 32 +#endif +#ifndef D3D10_1_SO_BUFFER_SLOT_COUNT +# define D3D10_1_SO_BUFFER_SLOT_COUNT 4 +#endif +#ifndef D3D11_SO_BUFFER_SLOT_COUNT +# define D3D11_SO_BUFFER_SLOT_COUNT 4 +#endif +#ifndef D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT +# define D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT 14 +#endif +#ifndef D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT +# define D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT 16 +#endif +#ifndef D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_NEGATIVE +# define D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_NEGATIVE -8 +#endif +#ifndef D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_POSITIVE +# define D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_POSITIVE 7 +#endif +#ifndef D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT +# define D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT 4096 +#endif +#ifndef D3D11_PS_INPUT_REGISTER_COUNT +# define D3D11_PS_INPUT_REGISTER_COUNT 32 +#endif +#ifndef D3D10_1_VS_OUTPUT_REGISTER_COUNT +# define D3D10_1_VS_OUTPUT_REGISTER_COUNT 32 +#endif +#if defined(ANGLE_MINGW32_COMPAT) +static const IID WKPDID_D3DDebugObjectName = { 0x429b8c22, 0x9188, 0x4b0c, 0x87, 0x42, 0xac, 0xb0, 0xbf, 0x85, 0xc2, 0x00 }; +#endif + +namespace rx +{ + +namespace gl_d3d11 +{ + +D3D11_BLEND ConvertBlendFunc(GLenum glBlend, bool isAlpha) +{ + D3D11_BLEND d3dBlend = D3D11_BLEND_ZERO; + + switch (glBlend) + { + case GL_ZERO: d3dBlend = D3D11_BLEND_ZERO; break; + case GL_ONE: d3dBlend = D3D11_BLEND_ONE; break; + case GL_SRC_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_SRC_ALPHA : D3D11_BLEND_SRC_COLOR); break; + case GL_ONE_MINUS_SRC_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_INV_SRC_ALPHA : D3D11_BLEND_INV_SRC_COLOR); break; + case GL_DST_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_DEST_ALPHA : D3D11_BLEND_DEST_COLOR); break; + case GL_ONE_MINUS_DST_COLOR: d3dBlend = (isAlpha ? D3D11_BLEND_INV_DEST_ALPHA : D3D11_BLEND_INV_DEST_COLOR); break; + case GL_SRC_ALPHA: d3dBlend = D3D11_BLEND_SRC_ALPHA; break; + case GL_ONE_MINUS_SRC_ALPHA: d3dBlend = D3D11_BLEND_INV_SRC_ALPHA; break; + case GL_DST_ALPHA: d3dBlend = D3D11_BLEND_DEST_ALPHA; break; + case GL_ONE_MINUS_DST_ALPHA: d3dBlend = D3D11_BLEND_INV_DEST_ALPHA; break; + case GL_CONSTANT_COLOR: d3dBlend = D3D11_BLEND_BLEND_FACTOR; break; + case GL_ONE_MINUS_CONSTANT_COLOR: d3dBlend = D3D11_BLEND_INV_BLEND_FACTOR; break; + case GL_CONSTANT_ALPHA: d3dBlend = D3D11_BLEND_BLEND_FACTOR; break; + case GL_ONE_MINUS_CONSTANT_ALPHA: d3dBlend = D3D11_BLEND_INV_BLEND_FACTOR; break; + case GL_SRC_ALPHA_SATURATE: d3dBlend = D3D11_BLEND_SRC_ALPHA_SAT; break; + default: UNREACHABLE(); + } + + return d3dBlend; +} + +D3D11_BLEND_OP ConvertBlendOp(GLenum glBlendOp) +{ + D3D11_BLEND_OP d3dBlendOp = D3D11_BLEND_OP_ADD; + + switch (glBlendOp) + { + case GL_FUNC_ADD: d3dBlendOp = D3D11_BLEND_OP_ADD; break; + case GL_FUNC_SUBTRACT: d3dBlendOp = D3D11_BLEND_OP_SUBTRACT; break; + case GL_FUNC_REVERSE_SUBTRACT: d3dBlendOp = D3D11_BLEND_OP_REV_SUBTRACT; break; + case GL_MIN: d3dBlendOp = D3D11_BLEND_OP_MIN; break; + case GL_MAX: d3dBlendOp = D3D11_BLEND_OP_MAX; break; + default: UNREACHABLE(); + } + + return d3dBlendOp; +} + +UINT8 ConvertColorMask(bool red, bool green, bool blue, bool alpha) +{ + UINT8 mask = 0; + if (red) + { + mask |= D3D11_COLOR_WRITE_ENABLE_RED; + } + if (green) + { + mask |= D3D11_COLOR_WRITE_ENABLE_GREEN; + } + if (blue) + { + mask |= D3D11_COLOR_WRITE_ENABLE_BLUE; + } + if (alpha) + { + mask |= D3D11_COLOR_WRITE_ENABLE_ALPHA; + } + return mask; +} + +D3D11_CULL_MODE ConvertCullMode(bool cullEnabled, GLenum cullMode) +{ + D3D11_CULL_MODE cull = D3D11_CULL_NONE; + + if (cullEnabled) + { + switch (cullMode) + { + case GL_FRONT: cull = D3D11_CULL_FRONT; break; + case GL_BACK: cull = D3D11_CULL_BACK; break; + case GL_FRONT_AND_BACK: cull = D3D11_CULL_NONE; break; + default: UNREACHABLE(); + } + } + else + { + cull = D3D11_CULL_NONE; + } + + return cull; +} + +D3D11_COMPARISON_FUNC ConvertComparison(GLenum comparison) +{ + D3D11_COMPARISON_FUNC d3dComp = D3D11_COMPARISON_NEVER; + switch (comparison) + { + case GL_NEVER: d3dComp = D3D11_COMPARISON_NEVER; break; + case GL_ALWAYS: d3dComp = D3D11_COMPARISON_ALWAYS; break; + case GL_LESS: d3dComp = D3D11_COMPARISON_LESS; break; + case GL_LEQUAL: d3dComp = D3D11_COMPARISON_LESS_EQUAL; break; + case GL_EQUAL: d3dComp = D3D11_COMPARISON_EQUAL; break; + case GL_GREATER: d3dComp = D3D11_COMPARISON_GREATER; break; + case GL_GEQUAL: d3dComp = D3D11_COMPARISON_GREATER_EQUAL; break; + case GL_NOTEQUAL: d3dComp = D3D11_COMPARISON_NOT_EQUAL; break; + default: UNREACHABLE(); + } + + return d3dComp; +} + +D3D11_DEPTH_WRITE_MASK ConvertDepthMask(bool depthWriteEnabled) +{ + return depthWriteEnabled ? D3D11_DEPTH_WRITE_MASK_ALL : D3D11_DEPTH_WRITE_MASK_ZERO; +} + +UINT8 ConvertStencilMask(GLuint stencilmask) +{ + return static_cast<UINT8>(stencilmask); +} + +D3D11_STENCIL_OP ConvertStencilOp(GLenum stencilOp) +{ + D3D11_STENCIL_OP d3dStencilOp = D3D11_STENCIL_OP_KEEP; + + switch (stencilOp) + { + case GL_ZERO: d3dStencilOp = D3D11_STENCIL_OP_ZERO; break; + case GL_KEEP: d3dStencilOp = D3D11_STENCIL_OP_KEEP; break; + case GL_REPLACE: d3dStencilOp = D3D11_STENCIL_OP_REPLACE; break; + case GL_INCR: d3dStencilOp = D3D11_STENCIL_OP_INCR_SAT; break; + case GL_DECR: d3dStencilOp = D3D11_STENCIL_OP_DECR_SAT; break; + case GL_INVERT: d3dStencilOp = D3D11_STENCIL_OP_INVERT; break; + case GL_INCR_WRAP: d3dStencilOp = D3D11_STENCIL_OP_INCR; break; + case GL_DECR_WRAP: d3dStencilOp = D3D11_STENCIL_OP_DECR; break; + default: UNREACHABLE(); + } + + return d3dStencilOp; +} + +D3D11_FILTER ConvertFilter(GLenum minFilter, GLenum magFilter, float maxAnisotropy, GLenum comparisonMode) +{ + bool comparison = comparisonMode != GL_NONE; + + if (maxAnisotropy > 1.0f) + { + return D3D11_ENCODE_ANISOTROPIC_FILTER(static_cast<D3D11_COMPARISON_FUNC>(comparison)); + } + else + { + D3D11_FILTER_TYPE dxMin = D3D11_FILTER_TYPE_POINT; + D3D11_FILTER_TYPE dxMip = D3D11_FILTER_TYPE_POINT; + switch (minFilter) + { + case GL_NEAREST: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_NEAREST_MIPMAP_NEAREST: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR_MIPMAP_NEAREST: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_POINT; break; + case GL_NEAREST_MIPMAP_LINEAR: dxMin = D3D11_FILTER_TYPE_POINT; dxMip = D3D11_FILTER_TYPE_LINEAR; break; + case GL_LINEAR_MIPMAP_LINEAR: dxMin = D3D11_FILTER_TYPE_LINEAR; dxMip = D3D11_FILTER_TYPE_LINEAR; break; + default: UNREACHABLE(); + } + + D3D11_FILTER_TYPE dxMag = D3D11_FILTER_TYPE_POINT; + switch (magFilter) + { + case GL_NEAREST: dxMag = D3D11_FILTER_TYPE_POINT; break; + case GL_LINEAR: dxMag = D3D11_FILTER_TYPE_LINEAR; break; + default: UNREACHABLE(); + } + + return D3D11_ENCODE_BASIC_FILTER(dxMin, dxMag, dxMip, static_cast<D3D11_COMPARISON_FUNC>(comparison)); + } +} + +D3D11_TEXTURE_ADDRESS_MODE ConvertTextureWrap(GLenum wrap) +{ + switch (wrap) + { + case GL_REPEAT: return D3D11_TEXTURE_ADDRESS_WRAP; + case GL_CLAMP_TO_EDGE: return D3D11_TEXTURE_ADDRESS_CLAMP; + case GL_MIRRORED_REPEAT: return D3D11_TEXTURE_ADDRESS_MIRROR; + default: UNREACHABLE(); + } + + return D3D11_TEXTURE_ADDRESS_WRAP; +} + +D3D11_QUERY ConvertQueryType(GLenum queryType) +{ + switch (queryType) + { + case GL_ANY_SAMPLES_PASSED_EXT: + case GL_ANY_SAMPLES_PASSED_CONSERVATIVE_EXT: return D3D11_QUERY_OCCLUSION; + case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN: return D3D11_QUERY_SO_STATISTICS; + default: UNREACHABLE(); return D3D11_QUERY_EVENT; + } +} + +} + + +namespace d3d11_gl +{ + +GLint GetMaximumClientVersion(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return 3; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 2; + + default: UNREACHABLE(); return 0; + } +} + +static gl::TextureCaps GenerateTextureFormatCaps(GLint maxClientVersion, GLenum internalFormat, ID3D11Device *device) +{ + gl::TextureCaps textureCaps; + + const d3d11::TextureFormat &formatInfo = d3d11::GetTextureFormatInfo(internalFormat, device->GetFeatureLevel()); + + UINT formatSupport; + if (SUCCEEDED(device->CheckFormatSupport(formatInfo.texFormat, &formatSupport))) + { + const gl::InternalFormat &internalFormatInfo = gl::GetInternalFormatInfo(internalFormat); + if (internalFormatInfo.depthBits > 0 || internalFormatInfo.stencilBits > 0) + { + textureCaps.texturable = ((formatSupport & D3D11_FORMAT_SUPPORT_TEXTURE2D) != 0); + } + else + { + UINT formatSupportMask = D3D11_FORMAT_SUPPORT_TEXTURE2D | D3D11_FORMAT_SUPPORT_TEXTURECUBE; + if (maxClientVersion > 2) + { + formatSupportMask |= D3D11_FORMAT_SUPPORT_TEXTURE3D; + } + textureCaps.texturable = ((formatSupport & formatSupportMask) == formatSupportMask); + } + } + + if (SUCCEEDED(device->CheckFormatSupport(formatInfo.renderFormat, &formatSupport)) && + ((formatSupport & D3D11_FORMAT_SUPPORT_MULTISAMPLE_RENDERTARGET) != 0)) + { + for (size_t sampleCount = 1; sampleCount <= D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT; sampleCount++) + { + UINT qualityCount = 0; + if (SUCCEEDED(device->CheckMultisampleQualityLevels(formatInfo.renderFormat, sampleCount, &qualityCount)) && + qualityCount > 0) + { + textureCaps.sampleCounts.insert(sampleCount); + } + } + } + + textureCaps.filterable = SUCCEEDED(device->CheckFormatSupport(formatInfo.srvFormat, &formatSupport)) && + ((formatSupport & D3D11_FORMAT_SUPPORT_SHADER_SAMPLE)) != 0; + textureCaps.renderable = (SUCCEEDED(device->CheckFormatSupport(formatInfo.rtvFormat, &formatSupport)) && + ((formatSupport & D3D11_FORMAT_SUPPORT_RENDER_TARGET)) != 0) || + (SUCCEEDED(device->CheckFormatSupport(formatInfo.dsvFormat, &formatSupport)) && + ((formatSupport & D3D11_FORMAT_SUPPORT_DEPTH_STENCIL) != 0)); + + return textureCaps; +} + +static bool GetNPOTTextureSupport(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476876.aspx + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static float GetMaximumAnisotropy(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_MAX_MAXANISOTROPY; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_MAX_MAXANISOTROPY; + + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476876.aspx + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: return 16; + + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_DEFAULT_MAX_ANISOTROPY; + + default: UNREACHABLE(); return 0; + } +} + +static bool GetOcclusionQuerySupport(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476150.aspx ID3D11Device::CreateQuery + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: return true; + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static bool GetEventQuerySupport(D3D_FEATURE_LEVEL featureLevel) +{ + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476150.aspx ID3D11Device::CreateQuery + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return true; + + default: UNREACHABLE(); return false; + } +} + +static bool GetInstancingSupport(D3D_FEATURE_LEVEL featureLevel) +{ + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476150.aspx ID3D11Device::CreateInputLayout + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + // Feature Level 9_3 supports instancing, but slot 0 in the input layout must not be instanced. + // D3D9 has a similar restriction, where stream 0 must not be instanced. + // This restriction can be worked around by remapping any non-instanced slot to slot 0. + // This works because HLSL uses shader semantics to match the vertex inputs to the elements in the input layout, rather than the slots. + // Note that we only support instancing via ANGLE_instanced_array on 9_3, since 9_3 doesn't support OpenGL ES 3.0 + case D3D_FEATURE_LEVEL_9_3: return true; + + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static bool GetFramebufferMultisampleSupport(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static bool GetFramebufferBlitSupport(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static bool GetDerivativeInstructionSupport(D3D_FEATURE_LEVEL featureLevel) +{ + // http://msdn.microsoft.com/en-us/library/windows/desktop/bb509588.aspx states that shader model + // ps_2_x is required for the ddx (and other derivative functions). + + // http://msdn.microsoft.com/en-us/library/windows/desktop/ff476876.aspx states that feature level + // 9.3 supports shader model ps_2_x. + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: + case D3D_FEATURE_LEVEL_9_3: return true; + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static bool GetShaderTextureLODSupport(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return true; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return false; + + default: UNREACHABLE(); return false; + } +} + +static size_t GetMaximumSimultaneousRenderTargets(D3D_FEATURE_LEVEL featureLevel) +{ + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476150.aspx ID3D11Device::CreateInputLayout + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_SIMULTANEOUS_RENDER_TARGET_COUNT; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_SIMULTANEOUS_RENDER_TARGET_COUNT; + + case D3D_FEATURE_LEVEL_9_3: return D3D_FL9_3_SIMULTANEOUS_RENDER_TARGET_COUNT; + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_SIMULTANEOUS_RENDER_TARGET_COUNT; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximum2DTextureSize(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION; + + case D3D_FEATURE_LEVEL_9_3: return D3D_FL9_3_REQ_TEXTURE2D_U_OR_V_DIMENSION; + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_REQ_TEXTURE2D_U_OR_V_DIMENSION; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumCubeMapTextureSize(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURECUBE_DIMENSION; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURECUBE_DIMENSION; + + case D3D_FEATURE_LEVEL_9_3: return D3D_FL9_3_REQ_TEXTURECUBE_DIMENSION; + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_REQ_TEXTURECUBE_DIMENSION; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximum2DTextureArraySize(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE2D_ARRAY_AXIS_DIMENSION; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximum3DTextureSize(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_TEXTURE3D_U_V_OR_W_DIMENSION; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_TEXTURE3D_U_V_OR_W_DIMENSION; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_REQ_TEXTURE3D_U_V_OR_W_DIMENSION; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumViewportSize(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_VIEWPORT_BOUNDS_MAX; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_VIEWPORT_BOUNDS_MAX; + + // No constants for D3D11 Feature Level 9 viewport size limits, use the maximum texture sizes + case D3D_FEATURE_LEVEL_9_3: return D3D_FL9_3_REQ_TEXTURE2D_U_OR_V_DIMENSION; + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_REQ_TEXTURE2D_U_OR_V_DIMENSION; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumDrawIndexedIndexCount(D3D_FEATURE_LEVEL featureLevel) +{ + // D3D11 allows up to 2^32 elements, but we report max signed int for convenience since that's what's + // returned from glGetInteger + static_assert(D3D11_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP == 32, "Unexpected D3D11 constant value."); + static_assert(D3D10_REQ_DRAWINDEXED_INDEX_COUNT_2_TO_EXP == 32, "Unexpected D3D11 constant value."); + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return std::numeric_limits<GLint>::max(); + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: return D3D_FL9_2_IA_PRIMITIVE_MAX_COUNT; + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_IA_PRIMITIVE_MAX_COUNT; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumDrawVertexCount(D3D_FEATURE_LEVEL featureLevel) +{ + // D3D11 allows up to 2^32 elements, but we report max signed int for convenience since that's what's + // returned from glGetInteger + static_assert(D3D11_REQ_DRAW_VERTEX_COUNT_2_TO_EXP == 32, "Unexpected D3D11 constant value."); + static_assert(D3D10_REQ_DRAW_VERTEX_COUNT_2_TO_EXP == 32, "Unexpected D3D11 constant value."); + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return std::numeric_limits<GLint>::max(); + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: return D3D_FL9_2_IA_PRIMITIVE_MAX_COUNT; + case D3D_FEATURE_LEVEL_9_1: return D3D_FL9_1_IA_PRIMITIVE_MAX_COUNT; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumVertexInputSlots(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_STANDARD_VERTEX_ELEMENT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_STANDARD_VERTEX_ELEMENT_COUNT; + case D3D_FEATURE_LEVEL_10_0: return D3D10_STANDARD_VERTEX_ELEMENT_COUNT; + + // From http://http://msdn.microsoft.com/en-us/library/windows/desktop/ff476876.aspx "Max Input Slots" + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 16; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumVertexUniformVectors(D3D_FEATURE_LEVEL featureLevel) +{ + // TODO(geofflang): Remove hard-coded limit once the gl-uniform-arrays test can pass + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return 1024; // D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return 1024; // D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT; + + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476149.aspx ID3D11DeviceContext::VSSetConstantBuffers + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 255; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetReservedVertexUniformBuffers() +{ + // Reserve one buffer for the application uniforms, and one for driver uniforms + return 2; +} + +static size_t GetMaximumVertexUniformBlocks(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - GetReservedVertexUniformBuffers(); + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - GetReservedVertexUniformBuffers(); + + // Uniform blocks not supported on D3D11 Feature Level 9 + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetReservedVertexOutputVectors(D3D_FEATURE_LEVEL featureLevel) +{ + // According to The OpenGL ES Shading Language specifications + // (Language Version 1.00 section 10.16, Language Version 3.10 section 12.21) + // built-in special variables (e.g. gl_FragCoord, or gl_PointCoord) + // which are statically used in the shader should be included in the variable packing algorithm. + // Therefore, we should not reserve output vectors for them. + + switch (featureLevel) + { + // We must reserve one output vector for dx_Position. + // We also reserve one for gl_Position, which we unconditionally output on Feature Levels 10_0+, + // even if it's unused in the shader (e.g. for transform feedback). TODO: This could be improved. +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return 2; + + // Just reserve dx_Position on Feature Level 9, since we don't ever need to output gl_Position. + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 1; + + default: UNREACHABLE(); return 0; + } + + return 1; +} + +static size_t GetMaximumVertexOutputVectors(D3D_FEATURE_LEVEL featureLevel) +{ + static_assert(gl::IMPLEMENTATION_MAX_VARYING_VECTORS == D3D11_VS_OUTPUT_REGISTER_COUNT, "Unexpected D3D11 constant value."); + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_VS_OUTPUT_REGISTER_COUNT - GetReservedVertexOutputVectors(featureLevel); + + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_VS_OUTPUT_REGISTER_COUNT - GetReservedVertexOutputVectors(featureLevel); + case D3D_FEATURE_LEVEL_10_0: return D3D10_VS_OUTPUT_REGISTER_COUNT - GetReservedVertexOutputVectors(featureLevel); + + // Use Shader Model 2.X limits + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 8 - GetReservedVertexOutputVectors(featureLevel); + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumVertexTextureUnits(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_COMMONSHADER_SAMPLER_SLOT_COUNT; + + // Vertex textures not supported on D3D11 Feature Level 9 according to + // http://msdn.microsoft.com/en-us/library/windows/desktop/ff476149.aspx + // ID3D11DeviceContext::VSSetSamplers and ID3D11DeviceContext::VSSetShaderResources + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumPixelUniformVectors(D3D_FEATURE_LEVEL featureLevel) +{ + // TODO(geofflang): Remove hard-coded limit once the gl-uniform-arrays test can pass + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return 1024; // D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return 1024; // D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT; + + // From http://msdn.microsoft.com/en-us/library/windows/desktop/ff476149.aspx ID3D11DeviceContext::PSSetConstantBuffers + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 32; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetReservedPixelUniformBuffers() +{ + // Reserve one buffer for the application uniforms, and one for driver uniforms + return 2; +} + +static size_t GetMaximumPixelUniformBlocks(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - GetReservedPixelUniformBuffers(); + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_COMMONSHADER_CONSTANT_BUFFER_API_SLOT_COUNT - GetReservedPixelUniformBuffers(); + + // Uniform blocks not supported on D3D11 Feature Level 9 + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumPixelInputVectors(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_PS_INPUT_REGISTER_COUNT - GetReservedVertexOutputVectors(featureLevel); + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_PS_INPUT_REGISTER_COUNT - GetReservedVertexOutputVectors(featureLevel); + + // Use Shader Model 2.X limits + case D3D_FEATURE_LEVEL_9_3: return 8 - GetReservedVertexOutputVectors(featureLevel); + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 8 - GetReservedVertexOutputVectors(featureLevel); + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumPixelTextureUnits(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_SAMPLER_SLOT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_COMMONSHADER_SAMPLER_SLOT_COUNT; + + // http://msdn.microsoft.com/en-us/library/windows/desktop/ff476149.aspx ID3D11DeviceContext::PSSetShaderResources + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 16; + + default: UNREACHABLE(); return 0; + } +} + +static int GetMinimumTexelOffset(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_NEGATIVE; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_COMMONSHADER_TEXEL_OFFSET_MAX_NEGATIVE; + + // Sampling functions with offsets are not available below shader model 4.0. + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static int GetMaximumTexelOffset(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_POSITIVE; + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D11_COMMONSHADER_TEXEL_OFFSET_MAX_POSITIVE; + + // Sampling functions with offsets are not available below shader model 4.0. + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumConstantBufferSize(D3D_FEATURE_LEVEL featureLevel) +{ + // Returns a size_t despite the limit being a GLuint64 because size_t is the maximum size of + // any buffer that could be allocated. + + const size_t bytesPerComponent = 4 * sizeof(float); + + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_REQ_CONSTANT_BUFFER_ELEMENT_COUNT * bytesPerComponent; + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return D3D10_REQ_CONSTANT_BUFFER_ELEMENT_COUNT * bytesPerComponent; + + // Limits from http://msdn.microsoft.com/en-us/library/windows/desktop/ff476501.aspx remarks section + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 4096 * bytesPerComponent; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumStreamOutputBuffers(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return D3D11_SO_BUFFER_SLOT_COUNT; + + case D3D_FEATURE_LEVEL_10_1: return D3D10_1_SO_BUFFER_SLOT_COUNT; + case D3D_FEATURE_LEVEL_10_0: return D3D10_SO_BUFFER_SLOT_COUNT; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumStreamOutputInterleavedComponents(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: + + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return GetMaximumVertexOutputVectors(featureLevel) * 4; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +static size_t GetMaximumStreamOutputSeparateComponents(D3D_FEATURE_LEVEL featureLevel) +{ + switch (featureLevel) + { +#if defined(ANGLE_ENABLE_D3D11_1) + case D3D_FEATURE_LEVEL_11_1: +#endif + case D3D_FEATURE_LEVEL_11_0: return GetMaximumStreamOutputInterleavedComponents(featureLevel) / + GetMaximumStreamOutputBuffers(featureLevel); + + + // D3D 10 and 10.1 only allow one output per output slot if an output slot other than zero is used. + case D3D_FEATURE_LEVEL_10_1: + case D3D_FEATURE_LEVEL_10_0: return 4; + + case D3D_FEATURE_LEVEL_9_3: + case D3D_FEATURE_LEVEL_9_2: + case D3D_FEATURE_LEVEL_9_1: return 0; + + default: UNREACHABLE(); return 0; + } +} + +void GenerateCaps(ID3D11Device *device, ID3D11DeviceContext *deviceContext, gl::Caps *caps, gl::TextureCapsMap *textureCapsMap, gl::Extensions *extensions) +{ + D3D_FEATURE_LEVEL featureLevel = device->GetFeatureLevel(); + + GLuint maxSamples = 0; + const gl::FormatSet &allFormats = gl::GetAllSizedInternalFormats(); + for (gl::FormatSet::const_iterator internalFormat = allFormats.begin(); internalFormat != allFormats.end(); ++internalFormat) + { + gl::TextureCaps textureCaps = GenerateTextureFormatCaps(GetMaximumClientVersion(featureLevel), *internalFormat, device); + textureCapsMap->insert(*internalFormat, textureCaps); + + maxSamples = std::max(maxSamples, textureCaps.getMaxSamples()); + + if (gl::GetInternalFormatInfo(*internalFormat).compressed) + { + caps->compressedTextureFormats.push_back(*internalFormat); + } + } + + // GL core feature limits + caps->maxElementIndex = static_cast<GLint64>(std::numeric_limits<unsigned int>::max()); + caps->max3DTextureSize = GetMaximum3DTextureSize(featureLevel); + caps->max2DTextureSize = GetMaximum2DTextureSize(featureLevel); + caps->maxCubeMapTextureSize = GetMaximumCubeMapTextureSize(featureLevel); + caps->maxArrayTextureLayers = GetMaximum2DTextureArraySize(featureLevel); + + // Unimplemented, set to minimum required + caps->maxLODBias = 2.0f; + + // No specific limits on render target size, maximum 2D texture size is equivalent + caps->maxRenderbufferSize = caps->max2DTextureSize; + + // Maximum draw buffers and color attachments are the same, max color attachments could eventually be + // increased to 16 + caps->maxDrawBuffers = GetMaximumSimultaneousRenderTargets(featureLevel); + caps->maxColorAttachments = GetMaximumSimultaneousRenderTargets(featureLevel); + + // D3D11 has the same limit for viewport width and height + caps->maxViewportWidth = GetMaximumViewportSize(featureLevel); + caps->maxViewportHeight = caps->maxViewportWidth; + + // Choose a reasonable maximum, enforced in the shader. + caps->minAliasedPointSize = 1.0f; + caps->maxAliasedPointSize = 1024.0f; + + // Wide lines not supported + caps->minAliasedLineWidth = 1.0f; + caps->maxAliasedLineWidth = 1.0f; + + // Primitive count limits + caps->maxElementsIndices = GetMaximumDrawIndexedIndexCount(featureLevel); + caps->maxElementsVertices = GetMaximumDrawVertexCount(featureLevel); + + // Program and shader binary formats (no supported shader binary formats) + caps->programBinaryFormats.push_back(GL_PROGRAM_BINARY_ANGLE); + + caps->vertexHighpFloat.setIEEEFloat(); + caps->vertexMediumpFloat.setIEEEFloat(); + caps->vertexLowpFloat.setIEEEFloat(); + caps->fragmentHighpFloat.setIEEEFloat(); + caps->fragmentMediumpFloat.setIEEEFloat(); + caps->fragmentLowpFloat.setIEEEFloat(); + + // 32-bit integers are natively supported + caps->vertexHighpInt.setTwosComplementInt(32); + caps->vertexMediumpInt.setTwosComplementInt(32); + caps->vertexLowpInt.setTwosComplementInt(32); + caps->fragmentHighpInt.setTwosComplementInt(32); + caps->fragmentMediumpInt.setTwosComplementInt(32); + caps->fragmentLowpInt.setTwosComplementInt(32); + + // We do not wait for server fence objects internally, so report a max timeout of zero. + caps->maxServerWaitTimeout = 0; + + // Vertex shader limits + caps->maxVertexAttributes = GetMaximumVertexInputSlots(featureLevel); + caps->maxVertexUniformComponents = GetMaximumVertexUniformVectors(featureLevel) * 4; + caps->maxVertexUniformVectors = GetMaximumVertexUniformVectors(featureLevel); + caps->maxVertexUniformBlocks = GetMaximumVertexUniformBlocks(featureLevel); + caps->maxVertexOutputComponents = GetMaximumVertexOutputVectors(featureLevel) * 4; + caps->maxVertexTextureImageUnits = GetMaximumVertexTextureUnits(featureLevel); + + // Fragment shader limits + caps->maxFragmentUniformComponents = GetMaximumPixelUniformVectors(featureLevel) * 4; + caps->maxFragmentUniformVectors = GetMaximumPixelUniformVectors(featureLevel); + caps->maxFragmentUniformBlocks = GetMaximumPixelUniformBlocks(featureLevel); + caps->maxFragmentInputComponents = GetMaximumPixelInputVectors(featureLevel) * 4; + caps->maxTextureImageUnits = GetMaximumPixelTextureUnits(featureLevel); + caps->minProgramTexelOffset = GetMinimumTexelOffset(featureLevel); + caps->maxProgramTexelOffset = GetMaximumTexelOffset(featureLevel); + + // Aggregate shader limits + caps->maxUniformBufferBindings = caps->maxVertexUniformBlocks + caps->maxFragmentUniformBlocks; + caps->maxUniformBlockSize = GetMaximumConstantBufferSize(featureLevel); + + // Setting a large alignment forces uniform buffers to bind with zero offset + caps->uniformBufferOffsetAlignment = static_cast<GLuint>(std::numeric_limits<GLint>::max()); +#if defined(ANGLE_ENABLE_D3D11_1) + ID3D11DeviceContext1 *deviceContext1 = d3d11::DynamicCastComObject<ID3D11DeviceContext1>(deviceContext); + + if (deviceContext1) + { + D3D11_FEATURE_DATA_D3D11_OPTIONS d3d11Options; + device->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &d3d11Options, sizeof(D3D11_FEATURE_DATA_D3D11_OPTIONS)); + + if (d3d11Options.ConstantBufferOffsetting) + { + // With DirectX 11.1, constant buffer offset and size must be a multiple of 16 constants of 16 bytes each. + // https://msdn.microsoft.com/en-us/library/windows/desktop/hh404649%28v=vs.85%29.aspx + caps->uniformBufferOffsetAlignment = 256; + } + + SafeRelease(deviceContext1); + } +#endif + + caps->maxCombinedUniformBlocks = caps->maxVertexUniformBlocks + caps->maxFragmentUniformBlocks; + caps->maxCombinedVertexUniformComponents = (static_cast<GLint64>(caps->maxVertexUniformBlocks) * static_cast<GLint64>(caps->maxUniformBlockSize / 4)) + + static_cast<GLint64>(caps->maxVertexUniformComponents); + caps->maxCombinedFragmentUniformComponents = (static_cast<GLint64>(caps->maxFragmentUniformBlocks) * static_cast<GLint64>(caps->maxUniformBlockSize / 4)) + + static_cast<GLint64>(caps->maxFragmentUniformComponents); + caps->maxVaryingComponents = GetMaximumVertexOutputVectors(featureLevel) * 4; + caps->maxVaryingVectors = GetMaximumVertexOutputVectors(featureLevel); + caps->maxCombinedTextureImageUnits = caps->maxVertexTextureImageUnits + caps->maxTextureImageUnits; + + // Transform feedback limits + caps->maxTransformFeedbackInterleavedComponents = GetMaximumStreamOutputInterleavedComponents(featureLevel); + caps->maxTransformFeedbackSeparateAttributes = GetMaximumStreamOutputBuffers(featureLevel); + caps->maxTransformFeedbackSeparateComponents = GetMaximumStreamOutputSeparateComponents(featureLevel); + + // GL extension support + extensions->setTextureExtensionSupport(*textureCapsMap); + extensions->elementIndexUint = true; + extensions->packedDepthStencil = true; + extensions->getProgramBinary = true; + extensions->rgb8rgba8 = true; + extensions->readFormatBGRA = true; + extensions->pixelBufferObject = true; + extensions->mapBuffer = true; + extensions->mapBufferRange = true; + extensions->textureNPOT = GetNPOTTextureSupport(featureLevel); + extensions->drawBuffers = GetMaximumSimultaneousRenderTargets(featureLevel) > 1; + extensions->textureStorage = true; + extensions->textureFilterAnisotropic = true; + extensions->maxTextureAnisotropy = GetMaximumAnisotropy(featureLevel); + extensions->occlusionQueryBoolean = GetOcclusionQuerySupport(featureLevel); + extensions->fence = GetEventQuerySupport(featureLevel); + extensions->timerQuery = false; // Unimplemented + extensions->robustness = true; + extensions->blendMinMax = true; + extensions->framebufferBlit = GetFramebufferBlitSupport(featureLevel); + extensions->framebufferMultisample = GetFramebufferMultisampleSupport(featureLevel); + extensions->maxSamples = maxSamples; + extensions->instancedArrays = GetInstancingSupport(featureLevel); + extensions->packReverseRowOrder = true; + extensions->standardDerivatives = GetDerivativeInstructionSupport(featureLevel); + extensions->shaderTextureLOD = GetShaderTextureLODSupport(featureLevel); + extensions->fragDepth = true; + extensions->textureUsage = true; // This could be false since it has no effect in D3D11 + extensions->translatedShaderSource = true; +} + +} + +namespace d3d11 +{ + +void MakeValidSize(bool isImage, DXGI_FORMAT format, GLsizei *requestWidth, GLsizei *requestHeight, int *levelOffset) +{ + const DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(format); + + int upsampleCount = 0; + // Don't expand the size of full textures that are at least (blockWidth x blockHeight) already. + if (isImage || *requestWidth < static_cast<GLsizei>(dxgiFormatInfo.blockWidth) || + *requestHeight < static_cast<GLsizei>(dxgiFormatInfo.blockHeight)) + { + while (*requestWidth % dxgiFormatInfo.blockWidth != 0 || *requestHeight % dxgiFormatInfo.blockHeight != 0) + { + *requestWidth <<= 1; + *requestHeight <<= 1; + upsampleCount++; + } + } + *levelOffset = upsampleCount; +} + +void GenerateInitialTextureData(GLint internalFormat, D3D_FEATURE_LEVEL featureLevel, GLuint width, GLuint height, GLuint depth, + GLuint mipLevels, std::vector<D3D11_SUBRESOURCE_DATA> *outSubresourceData, + std::vector< std::vector<BYTE> > *outData) +{ + const d3d11::TextureFormat &d3dFormatInfo = d3d11::GetTextureFormatInfo(internalFormat, featureLevel); + ASSERT(d3dFormatInfo.dataInitializerFunction != NULL); + + const d3d11::DXGIFormat &dxgiFormatInfo = d3d11::GetDXGIFormatInfo(d3dFormatInfo.texFormat); + + outSubresourceData->resize(mipLevels); + outData->resize(mipLevels); + + for (unsigned int i = 0; i < mipLevels; i++) + { + unsigned int mipWidth = std::max(width >> i, 1U); + unsigned int mipHeight = std::max(height >> i, 1U); + unsigned int mipDepth = std::max(depth >> i, 1U); + + unsigned int rowWidth = dxgiFormatInfo.pixelBytes * mipWidth; + unsigned int imageSize = rowWidth * height; + + outData->at(i).resize(rowWidth * mipHeight * mipDepth); + d3dFormatInfo.dataInitializerFunction(mipWidth, mipHeight, mipDepth, outData->at(i).data(), rowWidth, imageSize); + + outSubresourceData->at(i).pSysMem = outData->at(i).data(); + outSubresourceData->at(i).SysMemPitch = rowWidth; + outSubresourceData->at(i).SysMemSlicePitch = imageSize; + } +} + +void SetPositionTexCoordVertex(PositionTexCoordVertex* vertex, float x, float y, float u, float v) +{ + vertex->x = x; + vertex->y = y; + vertex->u = u; + vertex->v = v; +} + +void SetPositionLayerTexCoord3DVertex(PositionLayerTexCoord3DVertex* vertex, float x, float y, + unsigned int layer, float u, float v, float s) +{ + vertex->x = x; + vertex->y = y; + vertex->l = layer; + vertex->u = u; + vertex->v = v; + vertex->s = s; +} + +HRESULT SetDebugName(ID3D11DeviceChild *resource, const char *name) +{ +#if defined(_DEBUG) + return resource->SetPrivateData(WKPDID_D3DDebugObjectName, strlen(name), name); +#else + return S_OK; +#endif +} + +gl::Error GetAttachmentRenderTarget(const gl::FramebufferAttachment *attachment, RenderTarget11 **outRT) +{ + RenderTargetD3D *renderTarget = NULL; + gl::Error error = rx::GetAttachmentRenderTarget(attachment, &renderTarget); + if (error.isError()) + { + return error; + } + *outRT = RenderTarget11::makeRenderTarget11(renderTarget); + return gl::Error(GL_NO_ERROR); +} + +Workarounds GenerateWorkarounds(D3D_FEATURE_LEVEL featureLevel) +{ + Workarounds workarounds; + workarounds.mrtPerfWorkaround = true; + workarounds.setDataFasterThanImageUpload = true; + workarounds.zeroMaxLodWorkaround = (featureLevel <= D3D_FEATURE_LEVEL_9_3); + workarounds.useInstancedPointSpriteEmulation = (featureLevel <= D3D_FEATURE_LEVEL_9_3); + return workarounds; +} + +} + +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.h new file mode 100644 index 0000000000..207e6b5404 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/renderer11_utils.h @@ -0,0 +1,190 @@ +// +// Copyright (c) 2012-2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// renderer11_utils.h: Conversion functions and other utility routines +// specific to the D3D11 renderer. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_UTILS_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_UTILS_H_ + +#include "libANGLE/angletypes.h" +#include "libANGLE/Caps.h" +#include "libANGLE/Error.h" + +#include <vector> + +namespace gl +{ +class FramebufferAttachment; +} + +namespace rx +{ +class RenderTarget11; +struct Workarounds; + +namespace gl_d3d11 +{ + +D3D11_BLEND ConvertBlendFunc(GLenum glBlend, bool isAlpha); +D3D11_BLEND_OP ConvertBlendOp(GLenum glBlendOp); +UINT8 ConvertColorMask(bool maskRed, bool maskGreen, bool maskBlue, bool maskAlpha); + +D3D11_CULL_MODE ConvertCullMode(bool cullEnabled, GLenum cullMode); + +D3D11_COMPARISON_FUNC ConvertComparison(GLenum comparison); +D3D11_DEPTH_WRITE_MASK ConvertDepthMask(bool depthWriteEnabled); +UINT8 ConvertStencilMask(GLuint stencilmask); +D3D11_STENCIL_OP ConvertStencilOp(GLenum stencilOp); + +D3D11_FILTER ConvertFilter(GLenum minFilter, GLenum magFilter, float maxAnisotropy, GLenum comparisonMode); +D3D11_TEXTURE_ADDRESS_MODE ConvertTextureWrap(GLenum wrap); + +D3D11_QUERY ConvertQueryType(GLenum queryType); + +} + +namespace d3d11_gl +{ + +GLint GetMaximumClientVersion(D3D_FEATURE_LEVEL featureLevel); +void GenerateCaps(ID3D11Device *device, ID3D11DeviceContext *deviceContext, gl::Caps *caps, gl::TextureCapsMap *textureCapsMap, gl::Extensions *extensions); + +} + +namespace d3d11 +{ + +void MakeValidSize(bool isImage, DXGI_FORMAT format, GLsizei *requestWidth, GLsizei *requestHeight, int *levelOffset); + +void GenerateInitialTextureData(GLint internalFormat, D3D_FEATURE_LEVEL featureLevel, GLuint width, GLuint height, GLuint depth, + GLuint mipLevels, std::vector<D3D11_SUBRESOURCE_DATA> *outSubresourceData, + std::vector< std::vector<BYTE> > *outData); + +struct PositionTexCoordVertex +{ + float x, y; + float u, v; +}; +void SetPositionTexCoordVertex(PositionTexCoordVertex* vertex, float x, float y, float u, float v); + +struct PositionLayerTexCoord3DVertex +{ + float x, y; + unsigned int l; + float u, v, s; +}; +void SetPositionLayerTexCoord3DVertex(PositionLayerTexCoord3DVertex* vertex, float x, float y, + unsigned int layer, float u, float v, float s); + +template <typename T> +struct PositionDepthColorVertex +{ + float x, y, z; + T r, g, b, a; +}; + +template <typename T> +void SetPositionDepthColorVertex(PositionDepthColorVertex<T>* vertex, float x, float y, float z, + const gl::Color<T> &color) +{ + vertex->x = x; + vertex->y = y; + vertex->z = z; + vertex->r = color.red; + vertex->g = color.green; + vertex->b = color.blue; + vertex->a = color.alpha; +} + +HRESULT SetDebugName(ID3D11DeviceChild *resource, const char *name); + +template <typename outType> +outType* DynamicCastComObject(IUnknown* object) +{ + outType *outObject = NULL; + HRESULT result = object->QueryInterface(__uuidof(outType), reinterpret_cast<void**>(&outObject)); + if (SUCCEEDED(result)) + { + return outObject; + } + else + { + SafeRelease(outObject); + return NULL; + } +} + +inline bool isDeviceLostError(HRESULT errorCode) +{ + switch (errorCode) + { + case DXGI_ERROR_DEVICE_HUNG: + case DXGI_ERROR_DEVICE_REMOVED: + case DXGI_ERROR_DEVICE_RESET: + case DXGI_ERROR_DRIVER_INTERNAL_ERROR: + case DXGI_ERROR_NOT_CURRENTLY_AVAILABLE: + return true; + default: + return false; + } +} + +template <unsigned int N> +inline ID3D11VertexShader *CompileVS(ID3D11Device *device, const BYTE (&byteCode)[N], const char *name) +{ + ID3D11VertexShader *vs = NULL; + HRESULT result = device->CreateVertexShader(byteCode, N, NULL, &vs); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + SetDebugName(vs, name); + return vs; +} + +template <unsigned int N> +inline ID3D11GeometryShader *CompileGS(ID3D11Device *device, const BYTE (&byteCode)[N], const char *name) +{ + ID3D11GeometryShader *gs = NULL; + HRESULT result = device->CreateGeometryShader(byteCode, N, NULL, &gs); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + SetDebugName(gs, name); + return gs; +} + +template <unsigned int N> +inline ID3D11PixelShader *CompilePS(ID3D11Device *device, const BYTE (&byteCode)[N], const char *name) +{ + ID3D11PixelShader *ps = NULL; + HRESULT result = device->CreatePixelShader(byteCode, N, NULL, &ps); + UNUSED_ASSERTION_VARIABLE(result); + ASSERT(SUCCEEDED(result)); + SetDebugName(ps, name); + return ps; +} + +// Copy data to small D3D11 buffers, such as for small constant buffers, which use one struct to +// represent an entire buffer. +template <class T> +inline void SetBufferData(ID3D11DeviceContext *context, ID3D11Buffer *constantBuffer, const T &value) +{ + D3D11_MAPPED_SUBRESOURCE mappedResource; + context->Map(constantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mappedResource); + + memcpy(mappedResource.pData, &value, sizeof(T)); + + context->Unmap(constantBuffer, 0); +} + +gl::Error GetAttachmentRenderTarget(const gl::FramebufferAttachment *attachment, RenderTarget11 **outRT); + +Workarounds GenerateWorkarounds(D3D_FEATURE_LEVEL featureLevel); + +} + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_RENDERER11_UTILS_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/BufferToTexture11.hlsl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/BufferToTexture11.hlsl new file mode 100644 index 0000000000..c43734f6a3 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/BufferToTexture11.hlsl @@ -0,0 +1,77 @@ +Buffer<float4> Buffer4F : register(t0); +Buffer<int4> Buffer4I : register(t0); +Buffer<uint4> Buffer4UI : register(t0); + +struct VS_OUTPUT +{ + float4 position : SV_Position; + uint index : TEXCOORD0; + uint slice : LAYER; +}; + +struct GS_OUTPUT +{ + float4 position : SV_Position; + uint index : TEXCOORD0; + uint slice : SV_RenderTargetArrayIndex; +}; + +cbuffer BufferCopyParams : register(b0) +{ + uint FirstPixelOffset; + uint PixelsPerRow; + uint RowStride; + uint RowsPerSlice; + float2 PositionOffset; + float2 PositionScale; + int2 TexLocationOffset; + int2 TexLocationScale; + uint FirstSlice; +} + +void ComputePositionAndIndex(uint vertexID, out VS_OUTPUT outVertex) +{ + uint PixelsPerSlice = PixelsPerRow * RowsPerSlice; + uint SliceStride = RowStride * RowsPerSlice; + + uint slice = vertexID / PixelsPerSlice; + uint sliceOffset = slice * PixelsPerSlice; + uint row = (vertexID - sliceOffset) / PixelsPerRow; + uint col = vertexID - sliceOffset - (row * PixelsPerRow); + + float2 coords = float2(float(col), float(row)); + + outVertex.position = float4(PositionOffset + PositionScale * coords, 0.0f, 1.0f); + outVertex.index = FirstPixelOffset + slice * SliceStride + row * RowStride + col; + outVertex.slice = FirstSlice + slice; +} + +void VS_BufferToTexture(in uint vertexID : SV_VertexID, out VS_OUTPUT outVertex) +{ + ComputePositionAndIndex(vertexID, outVertex); +} + +[maxvertexcount(1)] +void GS_BufferToTexture(point VS_OUTPUT inVertex[1], inout PointStream<GS_OUTPUT> outStream) +{ + GS_OUTPUT outVertex; + outVertex.position = inVertex[0].position; + outVertex.index = inVertex[0].index; + outVertex.slice = inVertex[0].slice; + outStream.Append(outVertex); +} + +float4 PS_BufferToTexture_4F(in float4 inPosition : SV_Position, in uint inIndex : TEXCOORD0) : SV_Target +{ + return Buffer4F.Load(inIndex); +} + +int4 PS_BufferToTexture_4I(in float4 inPosition : SV_Position, in uint inIndex : TEXCOORD0) : SV_Target +{ + return Buffer4I.Load(inIndex); +} + +uint4 PS_BufferToTexture_4UI(in float4 inPosition : SV_Position, in uint inIndex : TEXCOORD0) : SV_Target +{ + return Buffer4UI.Load(inIndex); +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Clear11.hlsl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Clear11.hlsl new file mode 100644 index 0000000000..2b3e1ebe4c --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Clear11.hlsl @@ -0,0 +1,119 @@ +// Assume we are in SM4+, which has 8 color outputs + +void VS_ClearFloat( in float3 inPosition : POSITION, in float4 inColor : COLOR, + out float4 outPosition : SV_POSITION, out float4 outColor : COLOR) +{ + outPosition = float4(inPosition, 1.0f); + outColor = inColor; +} + +struct PS_OutputFloat +{ + float4 color0 : SV_TARGET0; + float4 color1 : SV_TARGET1; + float4 color2 : SV_TARGET2; + float4 color3 : SV_TARGET3; + float4 color4 : SV_TARGET4; + float4 color5 : SV_TARGET5; + float4 color6 : SV_TARGET6; + float4 color7 : SV_TARGET7; +}; + +PS_OutputFloat PS_ClearFloat(in float4 inPosition : SV_POSITION, in float4 inColor : COLOR) +{ + PS_OutputFloat outColor; + outColor.color0 = inColor; + outColor.color1 = inColor; + outColor.color2 = inColor; + outColor.color3 = inColor; + outColor.color4 = inColor; + outColor.color5 = inColor; + outColor.color6 = inColor; + outColor.color7 = inColor; + return outColor; +} + +struct PS_OutputFloat_FL9 +{ + float4 color0 : SV_TARGET0; + float4 color1 : SV_TARGET1; + float4 color2 : SV_TARGET2; + float4 color3 : SV_TARGET3; +}; + +PS_OutputFloat_FL9 PS_ClearFloat_FL9(in float4 inPosition : SV_POSITION, in float4 inColor : COLOR) +{ + PS_OutputFloat_FL9 outColor; + outColor.color0 = inColor; + outColor.color1 = inColor; + outColor.color2 = inColor; + outColor.color3 = inColor; + return outColor; +} + +void VS_ClearUint( in float3 inPosition : POSITION, in uint4 inColor : COLOR, + out float4 outPosition : SV_POSITION, out uint4 outColor : COLOR) +{ + outPosition = float4(inPosition, 1.0f); + outColor = inColor; +} + +struct PS_OutputUint +{ + uint4 color0 : SV_TARGET0; + uint4 color1 : SV_TARGET1; + uint4 color2 : SV_TARGET2; + uint4 color3 : SV_TARGET3; + uint4 color4 : SV_TARGET4; + uint4 color5 : SV_TARGET5; + uint4 color6 : SV_TARGET6; + uint4 color7 : SV_TARGET7; +}; + +PS_OutputUint PS_ClearUint(in float4 inPosition : SV_POSITION, in uint4 inColor : COLOR) +{ + PS_OutputUint outColor; + outColor.color0 = inColor; + outColor.color1 = inColor; + outColor.color2 = inColor; + outColor.color3 = inColor; + outColor.color4 = inColor; + outColor.color5 = inColor; + outColor.color6 = inColor; + outColor.color7 = inColor; + return outColor; +} + + +void VS_ClearSint( in float3 inPosition : POSITION, in int4 inColor : COLOR, + out float4 outPosition : SV_POSITION, out int4 outColor : COLOR) +{ + outPosition = float4(inPosition, 1.0f); + outColor = inColor; +} + +struct PS_OutputSint +{ + int4 color0 : SV_TARGET0; + int4 color1 : SV_TARGET1; + int4 color2 : SV_TARGET2; + int4 color3 : SV_TARGET3; + int4 color4 : SV_TARGET4; + int4 color5 : SV_TARGET5; + int4 color6 : SV_TARGET6; + int4 color7 : SV_TARGET7; +}; + +PS_OutputSint PS_ClearSint(in float4 inPosition : SV_POSITION, in int4 inColor : COLOR) +{ + PS_OutputSint outColor; + outColor.color0 = inColor; + outColor.color1 = inColor; + outColor.color2 = inColor; + outColor.color3 = inColor; + outColor.color4 = inColor; + outColor.color5 = inColor; + outColor.color6 = inColor; + outColor.color7 = inColor; + return outColor; +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough2D11.hlsl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough2D11.hlsl new file mode 100644 index 0000000000..8671c39fb7 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough2D11.hlsl @@ -0,0 +1,111 @@ +Texture2D<float4> TextureF : register(t0); +Texture2D<uint4> TextureUI : register(t0); +Texture2D<int4> TextureI : register(t0); + +SamplerState Sampler : register(s0); + +void VS_Passthrough2D( in float2 inPosition : POSITION, in float2 inTexCoord : TEXCOORD0, + out float4 outPosition : SV_POSITION, out float2 outTexCoord : TEXCOORD0) +{ + outPosition = float4(inPosition, 0.0f, 1.0f); + outTexCoord = inTexCoord; +} + +float PS_PassthroughDepth2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_DEPTH +{ + return TextureF.Sample(Sampler, inTexCoord).r; +} + +float4 PS_PassthroughRGBA2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return TextureF.Sample(Sampler, inTexCoord).rgba; +} + +uint4 PS_PassthroughRGBA2DUI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureUI.GetDimensions(size.x, size.y); + + return TextureUI.Load(int3(size * inTexCoord, 0)).rgba; +} + +int4 PS_PassthroughRGBA2DI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureI.GetDimensions(size.x, size.y); + + return TextureI.Load(int3(size * inTexCoord, 0)).rgba; +} + +float4 PS_PassthroughRGB2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, inTexCoord).rgb, 1.0f); +} + +uint4 PS_PassthroughRGB2DUI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureUI.GetDimensions(size.x, size.y); + + return uint4(TextureUI.Load(int3(size * inTexCoord, 0)).rgb, 0); +} + +int4 PS_PassthroughRGB2DI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureI.GetDimensions(size.x, size.y); + + return int4(TextureI.Load(int3(size * inTexCoord, 0)).rgb, 0); +} + +float4 PS_PassthroughRG2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, inTexCoord).rg, 0.0f, 1.0f); +} + +uint4 PS_PassthroughRG2DUI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureUI.GetDimensions(size.x, size.y); + + return uint4(TextureUI.Load(int3(size * inTexCoord, 0)).rg, 0, 0); +} + +int4 PS_PassthroughRG2DI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureI.GetDimensions(size.x, size.y); + + return int4(TextureI.Load(int3(size * inTexCoord, 0)).rg, 0, 0); +} + +float4 PS_PassthroughR2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, inTexCoord).r, 0.0f, 0.0f, 1.0f); +} + +uint4 PS_PassthroughR2DUI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureUI.GetDimensions(size.x, size.y); + + return uint4(TextureUI.Load(int3(size * inTexCoord, 0)).r, 0, 0, 0); +} + +int4 PS_PassthroughR2DI(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureI.GetDimensions(size.x, size.y); + + return int4(TextureI.Load(int3(size * inTexCoord, 0)).r, 0, 0, 0); +} + +float4 PS_PassthroughLum2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, inTexCoord).rrr, 1.0f); +} + +float4 PS_PassthroughLumAlpha2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return TextureF.Sample(Sampler, inTexCoord).rrra; +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough3D11.hlsl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough3D11.hlsl new file mode 100644 index 0000000000..c23c9032ec --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Passthrough3D11.hlsl @@ -0,0 +1,146 @@ +Texture3D<float4> TextureF : register(t0); +Texture3D<uint4> TextureUI : register(t0); +Texture3D<int4> TextureI : register(t0); + +SamplerState Sampler : register(s0); + +struct VS_INPUT +{ + float2 Position : POSITION; + uint Layer : LAYER; + float3 TexCoord : TEXCOORD; +}; + +struct VS_OUTPUT +{ + float4 Position : SV_POSITION; + uint Layer : LAYER; + float3 TexCoord : TEXCOORD; +}; + +struct GS_OUTPUT +{ + float4 Position : SV_POSITION; + uint Layer : SV_RENDERTARGETARRAYINDEX; + float3 TexCoord : TEXCOORD; +}; + +VS_OUTPUT VS_Passthrough3D(VS_INPUT input) +{ + VS_OUTPUT output; + + output.Position = float4(input.Position, 0.0f, 1.0f); + output.Layer = input.Layer; + output.TexCoord = input.TexCoord; + + return output; +} + +[maxvertexcount(3)] +void GS_Passthrough3D(triangle VS_OUTPUT input[3], inout TriangleStream<GS_OUTPUT> outputStream) +{ + GS_OUTPUT output; + + for (int i = 0; i < 3; i++) + { + output.Position = input[i].Position; + output.Layer = input[i].Layer; + output.TexCoord = input[i].TexCoord; + + outputStream.Append(output); + } +} + +float4 PS_PassthroughRGBA3D(GS_OUTPUT input) : SV_TARGET0 +{ + return TextureF.Sample(Sampler, input.TexCoord).rgba; +} + +uint4 PS_PassthroughRGBA3DUI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureUI.GetDimensions(size.x, size.y, size.z); + + return TextureUI.Load(int4(size * input.TexCoord, 0)).rgba; +} + +int4 PS_PassthroughRGBA3DI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureI.GetDimensions(size.x, size.y, size.z); + + return TextureI.Load(int4(size * input.TexCoord, 0)).rgba; +} + +float4 PS_PassthroughRGB3D(GS_OUTPUT input) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, input.TexCoord).rgb, 1.0f); +} + +uint4 PS_PassthroughRGB3DUI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureUI.GetDimensions(size.x, size.y, size.z); + + return uint4(TextureUI.Load(int4(size * input.TexCoord, 0)).rgb, 0); +} + +int4 PS_PassthroughRGB3DI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureI.GetDimensions(size.x, size.y, size.z); + + return int4(TextureI.Load(int4(size * input.TexCoord, 0)).rgb, 0); +} + +float4 PS_PassthroughRG3D(GS_OUTPUT input) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, input.TexCoord).rg, 0.0f, 1.0f); +} + +uint4 PS_PassthroughRG3DUI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureUI.GetDimensions(size.x, size.y, size.z); + + return uint4(TextureUI.Load(int4(size * input.TexCoord, 0)).rg, 0, 0); +} + +int4 PS_PassthroughRG3DI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureI.GetDimensions(size.x, size.y, size.z); + + return int4(TextureI.Load(int4(size * input.TexCoord, 0)).rg, 0, 0); +} + +float4 PS_PassthroughR3D(GS_OUTPUT input) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, input.TexCoord).r, 0.0f, 0.0f, 1.0f); +} + +uint4 PS_PassthroughR3DUI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureUI.GetDimensions(size.x, size.y, size.z); + + return uint4(TextureUI.Load(int4(size * input.TexCoord, 0)).r, 0, 0, 0); +} + +int4 PS_PassthroughR3DI(GS_OUTPUT input) : SV_TARGET0 +{ + uint3 size; + TextureI.GetDimensions(size.x, size.y, size.z); + + return int4(TextureI.Load(int4(size * input.TexCoord, 0)).r, 0, 0, 0); +} + +float4 PS_PassthroughLum3D(GS_OUTPUT input) : SV_TARGET0 +{ + return float4(TextureF.Sample(Sampler, input.TexCoord).rrr, 1.0f); +} + +float4 PS_PassthroughLumAlpha3D(GS_OUTPUT input) : SV_TARGET0 +{ + return TextureF.Sample(Sampler, input.TexCoord).rrra; +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Swizzle11.hlsl b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Swizzle11.hlsl new file mode 100644 index 0000000000..505e222137 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/shaders/Swizzle11.hlsl @@ -0,0 +1,99 @@ +Texture2D<float4> TextureF2D : register(t0); +Texture2D<uint4> TextureUI2D : register(t0); +Texture2D<int4> TextureI2D : register(t0); + +Texture3D<float4> TextureF3D : register(t0); +Texture3D<uint4> TextureUI3D : register(t0); +Texture3D<int4> TextureI3D : register(t0); + +Texture2DArray<float4> TextureF2DArray : register(t0); +Texture2DArray<uint4> TextureUI2DArray : register(t0); +Texture2DArray<int4> TextureI2DArray : register(t0); + +SamplerState Sampler : register(s0); + +cbuffer SwizzleProperties : register(b0) +{ + uint4 SwizzleIndices : packoffset(c0); +} + +float4 SwizzleLookup(in float4 sample) +{ + float lookup[6] = { sample[0], sample[1], sample[2], sample[3], 0.0f, 1.0f }; + return float4(lookup[SwizzleIndices[0]], lookup[SwizzleIndices[1]], lookup[SwizzleIndices[2]], lookup[SwizzleIndices[3]]); +} + +int4 SwizzleLookup(in int4 sample) +{ + int lookup[6] = { sample[0], sample[1], sample[2], sample[3], 0.0f, 1.0f }; + return int4(lookup[SwizzleIndices[0]], lookup[SwizzleIndices[1]], lookup[SwizzleIndices[2]], lookup[SwizzleIndices[3]]); +} + +uint4 SwizzleLookup(in uint4 sample) +{ + uint lookup[6] = { sample[0], sample[1], sample[2], sample[3], 0.0f, 1.0f }; + return uint4(lookup[SwizzleIndices[0]], lookup[SwizzleIndices[1]], lookup[SwizzleIndices[2]], lookup[SwizzleIndices[3]]); +} + +float4 PS_SwizzleF2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return SwizzleLookup(TextureF2D.Sample(Sampler, inTexCoord)); +} + +int4 PS_SwizzleI2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureI2D.GetDimensions(size.x, size.y); + + return SwizzleLookup(TextureI2D.Load(int3(size * inTexCoord, 0))); +} + +uint4 PS_SwizzleUI2D(in float4 inPosition : SV_POSITION, in float2 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint2 size; + TextureUI2D.GetDimensions(size.x, size.y); + + return SwizzleLookup(TextureUI2D.Load(int3(size * inTexCoord, 0))); +} + +float4 PS_SwizzleF3D(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return SwizzleLookup(TextureF3D.Sample(Sampler, inTexCoord)); +} + +int4 PS_SwizzleI3D(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint3 size; + TextureI3D.GetDimensions(size.x, size.y, size.z); + + return SwizzleLookup(TextureI3D.Load(int4(size * inTexCoord, 0))); +} + +uint4 PS_SwizzleUI3D(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint3 size; + TextureUI3D.GetDimensions(size.x, size.y, size.z); + + return SwizzleLookup(TextureUI3D.Load(int4(size * inTexCoord, 0))); +} + +float4 PS_SwizzleF2DArray(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + return SwizzleLookup(TextureF2DArray.Sample(Sampler, float3(inTexCoord.xy, inLayer))); +} + +int4 PS_SwizzleI2DArray(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint3 size; + TextureI2DArray.GetDimensions(size.x, size.y, size.z); + + return SwizzleLookup(TextureI2DArray.Load(int4(size.xy * inTexCoord.xy, inLayer, 0))); +} + +uint4 PS_SwizzleUI2DArray(in float4 inPosition : SV_POSITION, in uint inLayer : SV_RENDERTARGETARRAYINDEX, in float3 inTexCoord : TEXCOORD0) : SV_TARGET0 +{ + uint3 size; + TextureUI2DArray.GetDimensions(size.x, size.y, size.z); + + return SwizzleLookup(TextureUI2DArray.Load(int4(size.xy * inTexCoord.xy, inLayer, 0))); +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/win32/NativeWindow.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/win32/NativeWindow.cpp new file mode 100644 index 0000000000..9d8f0bb96c --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/win32/NativeWindow.cpp @@ -0,0 +1,68 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// NativeWindow.cpp: Handler for managing HWND native window types. + +#include "libANGLE/renderer/d3d/d3d11/NativeWindow.h" + +#include "common/debug.h" + +namespace rx +{ + +NativeWindow::NativeWindow(EGLNativeWindowType window) : mWindow(window) +{ +} + +bool NativeWindow::initialize() +{ + return true; +} + +bool NativeWindow::getClientRect(LPRECT rect) +{ + return GetClientRect(mWindow, rect) == TRUE; +} + +bool NativeWindow::isIconic() +{ + return IsIconic(mWindow) == TRUE; +} + +bool NativeWindow::isValidNativeWindow(EGLNativeWindowType window) +{ + return IsWindow(window) == TRUE; +} + +HRESULT NativeWindow::createSwapChain(ID3D11Device* device, DXGIFactory* factory, + DXGI_FORMAT format, unsigned int width, unsigned int height, + DXGISwapChain** swapChain) +{ + if (device == NULL || factory == NULL || swapChain == NULL || width == 0 || height == 0) + { + return E_INVALIDARG; + } + + DXGI_SWAP_CHAIN_DESC swapChainDesc = { 0 }; + swapChainDesc.BufferCount = 1; + swapChainDesc.BufferDesc.Format = format; + swapChainDesc.BufferDesc.Width = width; + swapChainDesc.BufferDesc.Height = height; + swapChainDesc.BufferDesc.Scaling = DXGI_MODE_SCALING_UNSPECIFIED; + swapChainDesc.BufferDesc.ScanlineOrdering = DXGI_MODE_SCANLINE_ORDER_UNSPECIFIED; + swapChainDesc.BufferDesc.RefreshRate.Numerator = 0; + swapChainDesc.BufferDesc.RefreshRate.Denominator = 1; + swapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT | DXGI_USAGE_BACK_BUFFER; + swapChainDesc.Flags = 0; + swapChainDesc.OutputWindow = mWindow; + swapChainDesc.SampleDesc.Count = 1; + swapChainDesc.SampleDesc.Quality = 0; + swapChainDesc.Windowed = TRUE; + swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_DISCARD; + + return factory->CreateSwapChain(device, &swapChainDesc, swapChain); +} +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.cpp new file mode 100644 index 0000000000..350526c867 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.cpp @@ -0,0 +1,214 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// CoreWindowNativeWindow.cpp: NativeWindow for managing ICoreWindow native window types. + +#include "libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.h" + +using namespace ABI::Windows::Foundation::Collections; + +namespace rx +{ +CoreWindowNativeWindow::~CoreWindowNativeWindow() +{ + unregisterForSizeChangeEvents(); +} + +bool CoreWindowNativeWindow::initialize(EGLNativeWindowType window, IPropertySet *propertySet) +{ + mOrientationChangedEventToken.value = 0; + ComPtr<IPropertySet> props = propertySet; + ComPtr<IInspectable> win = window; + SIZE swapChainSize = {}; + bool swapChainSizeSpecified = false; + HRESULT result = S_OK; + + // IPropertySet is an optional parameter and can be null. + // If one is specified, cache as an IMap and read the properties + // used for initial host initialization. + if (propertySet) + { + result = props.As(&mPropertyMap); + if (SUCCEEDED(result)) + { + // The EGLRenderSurfaceSizeProperty is optional and may be missing. The IPropertySet + // was prevalidated to contain the EGLNativeWindowType before being passed to + // this host. + result = GetOptionalSizePropertyValue(mPropertyMap, EGLRenderSurfaceSizeProperty, &swapChainSize, &swapChainSizeSpecified); + } + } + + if (SUCCEEDED(result)) + { + result = win.As(&mCoreWindow); + } + + if (SUCCEEDED(result)) + { + // If a swapchain size is specfied, then the automatic resize + // behaviors implemented by the host should be disabled. The swapchain + // will be still be scaled when being rendered to fit the bounds + // of the host. + // Scaling of the swapchain output occurs automatically because if + // the scaling mode setting DXGI_SCALING_STRETCH on the swapchain. + if (swapChainSizeSpecified) + { + mClientRect = { 0, 0, swapChainSize.cx, swapChainSize.cy }; + mSupportsSwapChainResize = false; + } + else + { + result = GetCoreWindowSizeInPixels(mCoreWindow, &mClientRect); + } + } + + if (SUCCEEDED(result)) + { + ComPtr<ABI::Windows::Graphics::Display::IDisplayInformationStatics> displayInformation; + result = GetActivationFactory(HStringReference(RuntimeClass_Windows_Graphics_Display_DisplayInformation).Get(), &displayInformation); + if (SUCCEEDED(result)) + { + result = displayInformation->GetForCurrentView(&mDisplayInformation); + } + } + + if (SUCCEEDED(result)) + { + mNewClientRect = mClientRect; + mClientRectChanged = false; + return registerForSizeChangeEvents(); + } + + return false; +} + +bool CoreWindowNativeWindow::registerForSizeChangeEvents() +{ + ComPtr<IWindowSizeChangedEventHandler> sizeChangedHandler; + HRESULT result = Microsoft::WRL::MakeAndInitialize<CoreWindowSizeChangedHandler>(sizeChangedHandler.ReleaseAndGetAddressOf(), this->shared_from_this()); + if (SUCCEEDED(result)) + { + result = mCoreWindow->add_SizeChanged(sizeChangedHandler.Get(), &mSizeChangedEventToken); + } + +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + ComPtr<IDisplayOrientationEventHandler> orientationChangedHandler; + result = sizeChangedHandler.As(&orientationChangedHandler); + if (SUCCEEDED(result)) + { + result = mDisplayInformation->add_OrientationChanged(orientationChangedHandler.Get(), &mOrientationChangedEventToken); + } +#endif + + if (SUCCEEDED(result)) + { + return true; + } + + return false; +} + +void CoreWindowNativeWindow::unregisterForSizeChangeEvents() +{ + if (mCoreWindow) + { + (void)mCoreWindow->remove_SizeChanged(mSizeChangedEventToken); + } +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + if (mDisplayInformation) + { + (void)mDisplayInformation->remove_OrientationChanged(mOrientationChangedEventToken); + } +#endif + mSizeChangedEventToken.value = 0; + mOrientationChangedEventToken.value = 0; +} + +HRESULT CoreWindowNativeWindow::createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain) +{ + if (device == NULL || factory == NULL || swapChain == NULL || width == 0 || height == 0) + { + return E_INVALIDARG; + } + + DXGI_SWAP_CHAIN_DESC1 swapChainDesc = { 0 }; + swapChainDesc.Width = width; + swapChainDesc.Height = height; + swapChainDesc.Format = format; + swapChainDesc.Stereo = FALSE; + swapChainDesc.SampleDesc.Count = 1; + swapChainDesc.SampleDesc.Quality = 0; + swapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT | DXGI_USAGE_BACK_BUFFER; + swapChainDesc.BufferCount = 2; + swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL; + swapChainDesc.Scaling = DXGI_SCALING_STRETCH; + + *swapChain = nullptr; + + ComPtr<IDXGISwapChain1> newSwapChain; + HRESULT result = factory->CreateSwapChainForCoreWindow(device, mCoreWindow.Get(), &swapChainDesc, nullptr, newSwapChain.ReleaseAndGetAddressOf()); + if (SUCCEEDED(result)) + { + +#if 0 //(WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) // Qt: allow Windows Phone to resize, but don't modify the backing texture in the swap chain. + // Test if swapchain supports resize. On Windows Phone devices, this will return DXGI_ERROR_UNSUPPORTED. On + // other devices DXGI_ERROR_INVALID_CALL should be returned because the combination of flags passed + // (DXGI_SWAP_CHAIN_FLAG_NONPREROTATED | DXGI_SWAP_CHAIN_FLAG_GDI_COMPATIBLE) are invalid flag combinations. + if (newSwapChain->ResizeBuffers(swapChainDesc.BufferCount, swapChainDesc.Width, swapChainDesc.Height, swapChainDesc.Format, DXGI_SWAP_CHAIN_FLAG_NONPREROTATED | DXGI_SWAP_CHAIN_FLAG_GDI_COMPATIBLE) == DXGI_ERROR_UNSUPPORTED) + { + mSupportsSwapChainResize = false; + } +#endif // (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + + result = newSwapChain.CopyTo(swapChain); + } + + if (SUCCEEDED(result)) + { + // If automatic swapchain resize behaviors have been disabled, then + // unregister for the resize change events. + if (mSupportsSwapChainResize == false) + { + unregisterForSizeChangeEvents(); + } + } + + return result; +} + +HRESULT GetCoreWindowSizeInPixels(const ComPtr<ABI::Windows::UI::Core::ICoreWindow>& coreWindow, RECT *windowSize) +{ + ABI::Windows::Foundation::Rect bounds; + HRESULT result = coreWindow->get_Bounds(&bounds); + if (SUCCEEDED(result)) + { + *windowSize = { 0, 0, ConvertDipsToPixels(bounds.Width), ConvertDipsToPixels(bounds.Height) }; + } + + return result; +} + +static float GetLogicalDpi() +{ + ComPtr<ABI::Windows::Graphics::Display::IDisplayPropertiesStatics> displayProperties; + float dpi = 96.0f; + + if (SUCCEEDED(GetActivationFactory(HStringReference(RuntimeClass_Windows_Graphics_Display_DisplayProperties).Get(), displayProperties.GetAddressOf()))) + { + if (SUCCEEDED(displayProperties->get_LogicalDpi(&dpi))) + { + return dpi; + } + } + return dpi; +} + +long ConvertDipsToPixels(float dips) +{ + static const float dipsPerInch = 96.0f; + return lround((dips * GetLogicalDpi() / dipsPerInch)); +} +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.h new file mode 100644 index 0000000000..59df9d5a6c --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.h @@ -0,0 +1,110 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// CoreWindowNativeWindow.h: NativeWindow for managing ICoreWindow native window types. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_WINRT_COREWINDOWNATIVEWINDOW_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_WINRT_COREWINDOWNATIVEWINDOW_H_ + +#include "libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.h" + +#include <memory> +#include <windows.graphics.display.h> + +typedef ABI::Windows::Foundation::__FITypedEventHandler_2_Windows__CUI__CCore__CCoreWindow_Windows__CUI__CCore__CWindowSizeChangedEventArgs_t IWindowSizeChangedEventHandler; +typedef ABI::Windows::Foundation::__FITypedEventHandler_2_Windows__CGraphics__CDisplay__CDisplayInformation_IInspectable_t IDisplayOrientationEventHandler; + +namespace rx +{ +long ConvertDipsToPixels(float dips); + +class CoreWindowNativeWindow : public InspectableNativeWindow, public std::enable_shared_from_this<CoreWindowNativeWindow> +{ + public: + ~CoreWindowNativeWindow(); + + bool initialize(EGLNativeWindowType window, IPropertySet *propertySet); + bool registerForSizeChangeEvents(); + void unregisterForSizeChangeEvents(); + HRESULT createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain); + + private: + ComPtr<ABI::Windows::UI::Core::ICoreWindow> mCoreWindow; + ComPtr<IMap<HSTRING, IInspectable*>> mPropertyMap; + ComPtr<ABI::Windows::Graphics::Display::IDisplayInformation> mDisplayInformation; + EventRegistrationToken mOrientationChangedEventToken; +}; + +[uuid(7F924F66-EBAE-40E5-A10B-B8F35E245190)] +class CoreWindowSizeChangedHandler : + public Microsoft::WRL::RuntimeClass<Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>, IWindowSizeChangedEventHandler, IDisplayOrientationEventHandler> +{ + public: + CoreWindowSizeChangedHandler() { } + HRESULT RuntimeClassInitialize(std::shared_ptr<InspectableNativeWindow> host) + { + if (!host) + { + return E_INVALIDARG; + } + + mHost = host; + return S_OK; + } + + // IWindowSizeChangedEventHandler + IFACEMETHOD(Invoke)(ABI::Windows::UI::Core::ICoreWindow *sender, ABI::Windows::UI::Core::IWindowSizeChangedEventArgs *sizeChangedEventArgs) + { + std::shared_ptr<InspectableNativeWindow> host = mHost.lock(); + if (host) + { + ABI::Windows::Foundation::Size windowSize; + if (SUCCEEDED(sizeChangedEventArgs->get_Size(&windowSize))) + { + SIZE windowSizeInPixels = { ConvertDipsToPixels(windowSize.Width), ConvertDipsToPixels(windowSize.Height) }; + host->setNewClientSize(windowSizeInPixels); + } + } + + return S_OK; + } + + IFACEMETHOD(Invoke)(ABI::Windows::Graphics::Display::IDisplayInformation *displayInformation, IInspectable *) + { +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) + NativeWindow::RotationFlags flags = NativeWindow::RotateNone; + ABI::Windows::Graphics::Display::DisplayOrientations orientation; + if (SUCCEEDED(displayInformation->get_CurrentOrientation(&orientation))) + { + switch (orientation) + { + case ABI::Windows::Graphics::Display::DisplayOrientations_Landscape: + flags = NativeWindow::RotateLeft; + break; + case ABI::Windows::Graphics::Display::DisplayOrientations_LandscapeFlipped: + flags = NativeWindow::RotateRight; + break; + default: + break; + } + } + std::shared_ptr<InspectableNativeWindow> host = mHost.lock(); + if (host) + { + host->setRotationFlags(flags); + } +#endif + return S_OK; + } + + private: + std::weak_ptr<InspectableNativeWindow> mHost; +}; + +HRESULT GetCoreWindowSizeInPixels(const ComPtr<ABI::Windows::UI::Core::ICoreWindow>& coreWindow, RECT *windowSize); +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_WINRT_COREWINDOWNATIVEWINDOW_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.cpp new file mode 100644 index 0000000000..2bf48c5d94 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.cpp @@ -0,0 +1,291 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InspectableNativeWindow.cpp: NativeWindow base class for managing IInspectable native window types. + +#include "libANGLE/renderer/d3d/d3d11/winrt/CoreWindowNativeWindow.h" +#include "libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.h" + +namespace rx +{ +NativeWindow::NativeWindow(EGLNativeWindowType window) +{ + mWindow = window; +} + +bool NativeWindow::initialize() +{ + // If the native window type is a IPropertySet, extract the + // EGLNativeWindowType (IInspectable) and initialize the + // proper host with this IPropertySet. + ComPtr<ABI::Windows::Foundation::Collections::IPropertySet> propertySet; + ComPtr<IInspectable> eglNativeWindow; + if (IsEGLConfiguredPropertySet(mWindow, &propertySet, &eglNativeWindow)) + { + // A property set was found and the EGLNativeWindowType was + // retrieved. The mWindow member of the host to must be updated + // to use the EGLNativeWindowType specified in the property set. + // mWindow is treated as a raw pointer not an AddRef'd interface, so + // the old mWindow does not need a Release() before this assignment. + mWindow = eglNativeWindow.Get(); + } + + ComPtr<ABI::Windows::UI::Core::ICoreWindow> coreWindow; + ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> swapChainPanel; + if (IsCoreWindow(mWindow, &coreWindow)) + { + mImpl = std::make_shared<CoreWindowNativeWindow>(); + if (mImpl) + { + return mImpl->initialize(mWindow, propertySet.Get()); + } + } + else if (IsSwapChainPanel(mWindow, &swapChainPanel)) + { + mImpl = std::make_shared<SwapChainPanelNativeWindow>(); + if (mImpl) + { + return mImpl->initialize(mWindow, propertySet.Get()); + } + } + else + { + ERR("Invalid IInspectable EGLNativeWindowType detected. Valid IInspectables include ICoreWindow, ISwapChainPanel and IPropertySet"); + } + + return false; +} + +bool NativeWindow::getClientRect(RECT *rect) +{ + if (mImpl) + { + return mImpl->getClientRect(rect); + } + + return false; +} + +#if defined(ANGLE_ENABLE_WINDOWS_STORE) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP) +NativeWindow::RotationFlags NativeWindow::rotationFlags() const +{ + if (mImpl) + { + return mImpl->rotationFlags(); + } + + return NativeWindow::RotateNone; +} +#endif + +bool NativeWindow::isIconic() +{ + return false; +} + +bool NativeWindow::isValidNativeWindow(EGLNativeWindowType window) +{ + return IsValidEGLNativeWindowType(window); +} + +HRESULT NativeWindow::createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain) +{ + if (mImpl) + { + return mImpl->createSwapChain(device, factory, format, width, height, swapChain); + } + + return E_UNEXPECTED; +} + +bool IsCoreWindow(EGLNativeWindowType window, ComPtr<ABI::Windows::UI::Core::ICoreWindow> *coreWindow) +{ + if (!window) + { + return false; + } + + ComPtr<IInspectable> win = window; + ComPtr<ABI::Windows::UI::Core::ICoreWindow> coreWin; + if (SUCCEEDED(win.As(&coreWin))) + { + if (coreWindow != nullptr) + { + *coreWindow = coreWin.Detach(); + } + return true; + } + + return false; +} + +bool IsSwapChainPanel(EGLNativeWindowType window, ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> *swapChainPanel) +{ + if (!window) + { + return false; + } + + ComPtr<IInspectable> win = window; + ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> panel; + if (SUCCEEDED(win.As(&panel))) + { + if (swapChainPanel != nullptr) + { + *swapChainPanel = panel.Detach(); + } + return true; + } + + return false; +} + +bool IsEGLConfiguredPropertySet(EGLNativeWindowType window, ABI::Windows::Foundation::Collections::IPropertySet **propertySet, IInspectable **eglNativeWindow) +{ + if (!window) + { + return false; + } + + ComPtr<IInspectable> props = window; + ComPtr<IPropertySet> propSet; + ComPtr<IInspectable> nativeWindow; + ComPtr<ABI::Windows::Foundation::Collections::IMap<HSTRING, IInspectable*>> propMap; + boolean hasEglNativeWindowPropertyKey = false; + + HRESULT result = props.As(&propSet); + if (SUCCEEDED(result)) + { + result = propSet.As(&propMap); + } + + // Look for the presence of the EGLNativeWindowType in the property set + if (SUCCEEDED(result)) + { + result = propMap->HasKey(HStringReference(EGLNativeWindowTypeProperty).Get(), &hasEglNativeWindowPropertyKey); + } + + // If the IPropertySet does not contain the required EglNativeWindowType key, the property set is + // considered invalid. + if (SUCCEEDED(result) && !hasEglNativeWindowPropertyKey) + { + ERR("Could not find EGLNativeWindowTypeProperty in IPropertySet. Valid EGLNativeWindowTypeProperty values include ICoreWindow"); + return false; + } + + // The EglNativeWindowType property exists, so retreive the IInspectable that represents the EGLNativeWindowType + if (SUCCEEDED(result) && hasEglNativeWindowPropertyKey) + { + result = propMap->Lookup(HStringReference(EGLNativeWindowTypeProperty).Get(), &nativeWindow); + } + + if (SUCCEEDED(result)) + { + if (propertySet != nullptr) + { + result = propSet.CopyTo(propertySet); + } + } + + if (SUCCEEDED(result)) + { + if (eglNativeWindow != nullptr) + { + result = nativeWindow.CopyTo(eglNativeWindow); + } + } + + if (SUCCEEDED(result)) + { + return true; + } + + return false; +} + +// A Valid EGLNativeWindowType IInspectable can only be: +// +// ICoreWindow +// IPropertySet +// +// Anything else will be rejected as an invalid IInspectable. +bool IsValidEGLNativeWindowType(EGLNativeWindowType window) +{ + return IsCoreWindow(window) || IsSwapChainPanel(window) || IsEGLConfiguredPropertySet(window); +} + +// Attempts to read an optional SIZE property value that is assumed to be in the form of +// an ABI::Windows::Foundation::Size. This function validates the Size value before returning +// it to the caller. +// +// Possible return values are: +// S_OK, valueExists == true - optional SIZE value was successfully retrieved and validated +// S_OK, valueExists == false - optional SIZE value was not found +// E_INVALIDARG, valueExists = false - optional SIZE value was malformed in the property set. +// * Incorrect property type ( must be PropertyType_Size) +// * Invalid property value (width/height must be > 0) +// Additional errors may be returned from IMap or IPropertyValue +// +HRESULT GetOptionalSizePropertyValue(const ComPtr<ABI::Windows::Foundation::Collections::IMap<HSTRING, IInspectable*>>& propertyMap, const wchar_t *propertyName, SIZE *value, bool *valueExists) +{ + if (!propertyMap || !propertyName || !value || !valueExists) + { + return false; + } + + // Assume that the value does not exist + *valueExists = false; + *value = { 0, 0 }; + + ComPtr<ABI::Windows::Foundation::IPropertyValue> propertyValue; + ABI::Windows::Foundation::PropertyType propertyType = ABI::Windows::Foundation::PropertyType::PropertyType_Empty; + Size sizeValue = { 0, 0 }; + boolean hasKey = false; + + HRESULT result = propertyMap->HasKey(HStringReference(propertyName).Get(), &hasKey); + if (SUCCEEDED(result) && !hasKey) + { + // Value does not exist, so return S_OK and set the exists parameter to false to indicate + // that a the optional property does not exist. + *valueExists = false; + return S_OK; + } + + if (SUCCEEDED(result)) + { + result = propertyMap->Lookup(HStringReference(propertyName).Get(), &propertyValue); + } + + if (SUCCEEDED(result)) + { + result = propertyValue->get_Type(&propertyType); + } + + // Check if the expected Size property is of PropertyType_Size type. + if (SUCCEEDED(result) && propertyType == ABI::Windows::Foundation::PropertyType::PropertyType_Size) + { + if (SUCCEEDED(propertyValue->GetSize(&sizeValue)) && (sizeValue.Width > 0 && sizeValue.Height > 0)) + { + // A valid property value exists + *value = { static_cast<long>(sizeValue.Width), static_cast<long>(sizeValue.Height) }; + *valueExists = true; + result = S_OK; + } + else + { + // An invalid Size property was detected. Width/Height values must > 0 + result = E_INVALIDARG; + } + } + else + { + // An invalid property type was detected. Size property must be of PropertyType_Size + result = E_INVALIDARG; + } + + return result; +} +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.h new file mode 100644 index 0000000000..575bdf8a58 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.h @@ -0,0 +1,105 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// InspectableNativeWindow.h: Host specific implementation interface for +// managing IInspectable native window types. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_WINRT_INSPECTABLENATIVEWINDOW_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_WINRT_INSPECTABLENATIVEWINDOW_H_ + +#include "libANGLE/renderer/d3d/d3d11/NativeWindow.h" + +#include "common/platform.h" + +#include "angle_windowsstore.h" + +#include <windows.ui.xaml.h> +#include <windows.ui.xaml.media.dxinterop.h> + +using namespace Microsoft::WRL; +using namespace Microsoft::WRL::Wrappers; +using namespace ABI::Windows::Foundation; +using namespace ABI::Windows::Foundation::Collections; + +namespace rx +{ +class InspectableNativeWindow +{ + public: + InspectableNativeWindow() : + mSupportsSwapChainResize(true), + mRequiresSwapChainScaling(false), + mClientRectChanged(false), + mClientRect({0,0,0,0}), + mNewClientRect({0,0,0,0}), + mRotationFlags(NativeWindow::RotateNone) + { + mSizeChangedEventToken.value = 0; + } + virtual ~InspectableNativeWindow(){} + + virtual bool initialize(EGLNativeWindowType window, IPropertySet *propertySet) = 0; + virtual HRESULT createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain) = 0; + virtual bool registerForSizeChangeEvents() = 0; + virtual void unregisterForSizeChangeEvents() = 0; + virtual HRESULT scaleSwapChain(const SIZE& newSize) { return S_OK; } + + bool getClientRect(RECT *rect) + { + if (mClientRectChanged && mSupportsSwapChainResize) + { + mClientRect = mNewClientRect; + } + + *rect = mClientRect; + + return true; + } + + void setNewClientSize(const SIZE &newSize) + { + if (mSupportsSwapChainResize && !mRequiresSwapChainScaling) + { + mNewClientRect = { 0, 0, newSize.cx, newSize.cy }; + mClientRectChanged = true; + } + + if (mRequiresSwapChainScaling) + { + scaleSwapChain(newSize); + } + } + + NativeWindow::RotationFlags rotationFlags() const + { + return mRotationFlags; + } + + void setRotationFlags(NativeWindow::RotationFlags flags) + { + mRotationFlags = flags; + } + +protected: + bool mSupportsSwapChainResize; + bool mRequiresSwapChainScaling; + RECT mClientRect; + RECT mNewClientRect; + bool mClientRectChanged; + NativeWindow::RotationFlags mRotationFlags; + + EventRegistrationToken mSizeChangedEventToken; +}; + +bool IsValidEGLNativeWindowType(EGLNativeWindowType window); +bool IsCoreWindow(EGLNativeWindowType window, ComPtr<ABI::Windows::UI::Core::ICoreWindow> *coreWindow = nullptr); +bool IsSwapChainPanel(EGLNativeWindowType window, ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> *swapChainPanel = nullptr); +bool IsEGLConfiguredPropertySet(EGLNativeWindowType window, ABI::Windows::Foundation::Collections::IPropertySet **propertySet = nullptr, IInspectable **inspectable = nullptr); +HRESULT GetOptionalSizePropertyValue(const ComPtr<ABI::Windows::Foundation::Collections::IMap<HSTRING, IInspectable*>>& propertyMap, const wchar_t *propertyName, SIZE *value, bool *valueExists); + +} + +#endif // LIBANGLE_RENDERER_D3D_D3D11_WINRT_INSPECTABLENATIVEWINDOW_H_ diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.cpp b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.cpp new file mode 100644 index 0000000000..53899dbb30 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.cpp @@ -0,0 +1,228 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChainPanelNativeWindow.cpp: NativeWindow for managing ISwapChainPanel native window types. + +#include "libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.h" + +#include <algorithm> +#include <math.h> + +using namespace ABI::Windows::Foundation::Collections; + +namespace rx +{ +SwapChainPanelNativeWindow::~SwapChainPanelNativeWindow() +{ + unregisterForSizeChangeEvents(); +} + +bool SwapChainPanelNativeWindow::initialize(EGLNativeWindowType window, IPropertySet *propertySet) +{ + ComPtr<IPropertySet> props = propertySet; + ComPtr<IInspectable> win = window; + SIZE swapChainSize = {}; + bool swapChainSizeSpecified = false; + HRESULT result = S_OK; + + // IPropertySet is an optional parameter and can be null. + // If one is specified, cache as an IMap and read the properties + // used for initial host initialization. + if (propertySet) + { + result = props.As(&mPropertyMap); + if (SUCCEEDED(result)) + { + // The EGLRenderSurfaceSizeProperty is optional and may be missing. The IPropertySet + // was prevalidated to contain the EGLNativeWindowType before being passed to + // this host. + result = GetOptionalSizePropertyValue(mPropertyMap, EGLRenderSurfaceSizeProperty, &swapChainSize, &swapChainSizeSpecified); + } + } + + if (SUCCEEDED(result)) + { + result = win.As(&mSwapChainPanel); + } + + if (SUCCEEDED(result)) + { + // If a swapchain size is specfied, then the automatic resize + // behaviors implemented by the host should be disabled. The swapchain + // will be still be scaled when being rendered to fit the bounds + // of the host. + // Scaling of the swapchain output needs to be handled by the + // host for swapchain panels even though the scaling mode setting + // DXGI_SCALING_STRETCH is configured on the swapchain. + if (swapChainSizeSpecified) + { + mClientRect = { 0, 0, swapChainSize.cx, swapChainSize.cy }; + + // Enable host swapchain scaling + mRequiresSwapChainScaling = true; + } + else + { + result = GetSwapChainPanelSize(mSwapChainPanel, &mClientRect); + } + } + + if (SUCCEEDED(result)) + { + mNewClientRect = mClientRect; + mClientRectChanged = false; + return registerForSizeChangeEvents(); + } + + return false; +} + +bool SwapChainPanelNativeWindow::registerForSizeChangeEvents() +{ + ComPtr<ABI::Windows::UI::Xaml::ISizeChangedEventHandler> sizeChangedHandler; + ComPtr<ABI::Windows::UI::Xaml::IFrameworkElement> frameworkElement; + HRESULT result = Microsoft::WRL::MakeAndInitialize<SwapChainPanelSizeChangedHandler>(sizeChangedHandler.ReleaseAndGetAddressOf(), this->shared_from_this()); + + if (SUCCEEDED(result)) + { + result = mSwapChainPanel.As(&frameworkElement); + } + + if (SUCCEEDED(result)) + { + result = frameworkElement->add_SizeChanged(sizeChangedHandler.Get(), &mSizeChangedEventToken); + } + + if (SUCCEEDED(result)) + { + return true; + } + + return false; +} + +void SwapChainPanelNativeWindow::unregisterForSizeChangeEvents() +{ + ComPtr<ABI::Windows::UI::Xaml::IFrameworkElement> frameworkElement; + if (SUCCEEDED(mSwapChainPanel.As(&frameworkElement))) + { + (void)frameworkElement->remove_SizeChanged(mSizeChangedEventToken); + } + + mSizeChangedEventToken.value = 0; +} + +HRESULT SwapChainPanelNativeWindow::createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain) +{ + if (device == NULL || factory == NULL || swapChain == NULL || width == 0 || height == 0) + { + return E_INVALIDARG; + } + + DXGI_SWAP_CHAIN_DESC1 swapChainDesc = { 0 }; + swapChainDesc.Width = width; + swapChainDesc.Height = height; + swapChainDesc.Format = format; + swapChainDesc.Stereo = FALSE; + swapChainDesc.SampleDesc.Count = 1; + swapChainDesc.SampleDesc.Quality = 0; + swapChainDesc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT | DXGI_USAGE_BACK_BUFFER; + swapChainDesc.BufferCount = 2; + swapChainDesc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_SEQUENTIAL; + swapChainDesc.Scaling = DXGI_SCALING_STRETCH; + swapChainDesc.AlphaMode = DXGI_ALPHA_MODE_IGNORE; + + *swapChain = nullptr; + + ComPtr<IDXGISwapChain1> newSwapChain; + ComPtr<ISwapChainPanelNative> swapChainPanelNative; + RECT currentPanelSize = {}; + + HRESULT result = factory->CreateSwapChainForComposition(device, &swapChainDesc, nullptr, newSwapChain.ReleaseAndGetAddressOf()); + + if (SUCCEEDED(result)) + { + result = mSwapChainPanel.As(&swapChainPanelNative); + } + + if (SUCCEEDED(result)) + { + result = swapChainPanelNative->SetSwapChain(newSwapChain.Get()); + } + + if (SUCCEEDED(result)) + { + // The swapchain panel host requires an instance of the swapchain set on the SwapChainPanel + // to perform the runtime-scale behavior. This swapchain is cached here because there are + // no methods for retreiving the currently configured on from ISwapChainPanelNative. + mSwapChain = newSwapChain; + result = newSwapChain.CopyTo(swapChain); + } + + // If the host is responsible for scaling the output of the swapchain, then + // scale it now before returning an instance to the caller. This is done by + // first reading the current size of the swapchain panel, then scaling + if (SUCCEEDED(result) && mRequiresSwapChainScaling) + { + result = GetSwapChainPanelSize(mSwapChainPanel, ¤tPanelSize); + } + + // Scale the swapchain to fit inside the contents of the panel. + if (SUCCEEDED(result) && mRequiresSwapChainScaling) + { + SIZE currentSize = { currentPanelSize.right, currentPanelSize.bottom }; + result = scaleSwapChain(currentSize); + } + + if (SUCCEEDED(result)) + { + // If automatic swapchain resize behaviors have been disabled, then + // unregister for the resize change events. + if (mSupportsSwapChainResize == false) + { + unregisterForSizeChangeEvents(); + } + } + + return result; +} + +HRESULT SwapChainPanelNativeWindow::scaleSwapChain(const SIZE &newSize) +{ + ABI::Windows::Foundation::Size renderScale = { (float)newSize.cx/(float)mClientRect.right, (float)newSize.cy/(float)mClientRect.bottom }; + // Setup a scale matrix for the swap chain + DXGI_MATRIX_3X2_F scaleMatrix = {}; + scaleMatrix._11 = renderScale.Width; + scaleMatrix._22 = renderScale.Height; + + ComPtr<IDXGISwapChain2> swapChain2; + HRESULT result = mSwapChain.As(&swapChain2); + if (SUCCEEDED(result)) + { + result = swapChain2->SetMatrixTransform(&scaleMatrix); + } + + return result; +} + +HRESULT GetSwapChainPanelSize(const ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> &swapChainPanel, RECT *windowSize) +{ + ComPtr<ABI::Windows::UI::Xaml::IUIElement> uiElement; + ABI::Windows::Foundation::Size renderSize = { 0, 0 }; + HRESULT result = swapChainPanel.As(&uiElement); + if (SUCCEEDED(result)) + { + result = uiElement->get_RenderSize(&renderSize); + } + + if (SUCCEEDED(result)) + { + *windowSize = { 0, 0, lround(renderSize.Width), lround(renderSize.Height) }; + } + + return result; +} +} diff --git a/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.h b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.h new file mode 100644 index 0000000000..caf327d913 --- /dev/null +++ b/src/3rdparty/angle/src/libANGLE/renderer/d3d/d3d11/winrt/SwapChainPanelNativeWindow.h @@ -0,0 +1,79 @@ +// +// Copyright (c) 2014 The ANGLE Project Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. +// + +// SwapChainPanelNativeWindow.h: NativeWindow for managing ISwapChainPanel native window types. + +#ifndef LIBANGLE_RENDERER_D3D_D3D11_WINRT_SWAPCHAINPANELNATIVEWINDOW_H_ +#define LIBANGLE_RENDERER_D3D_D3D11_WINRT_SWAPCHAINPANELNATIVEWINDOW_H_ + +#include "libANGLE/renderer/d3d/d3d11/winrt/InspectableNativeWindow.h" + +namespace rx +{ +class SwapChainPanelNativeWindow : public InspectableNativeWindow, public std::enable_shared_from_this<SwapChainPanelNativeWindow> +{ + public: + ~SwapChainPanelNativeWindow(); + + bool initialize(EGLNativeWindowType window, IPropertySet *propertySet); + bool registerForSizeChangeEvents(); + void unregisterForSizeChangeEvents(); + HRESULT createSwapChain(ID3D11Device *device, DXGIFactory *factory, DXGI_FORMAT format, unsigned int width, unsigned int height, DXGISwapChain **swapChain); + HRESULT scaleSwapChain(const SIZE &newSize); + + private: + ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> mSwapChainPanel; + ComPtr<IMap<HSTRING, IInspectable*>> mPropertyMap; + ComPtr<DXGISwapChain> mSwapChain; +}; + +[uuid(8ACBD974-8187-4508-AD80-AEC77F93CF36)] +class SwapChainPanelSizeChangedHandler : + public Microsoft::WRL::RuntimeClass<Microsoft::WRL::RuntimeClassFlags<Microsoft::WRL::ClassicCom>, ABI::Windows::UI::Xaml::ISizeChangedEventHandler> +{ + public: + SwapChainPanelSizeChangedHandler() { } + HRESULT RuntimeClassInitialize(std::shared_ptr<InspectableNativeWindow> host) + { + if (!host) + { + return E_INVALIDARG; + } + + mHost = host; + return S_OK; + } + + // ISizeChangedEventHandler + IFACEMETHOD(Invoke)(IInspectable *sender, ABI::Windows::UI::Xaml::ISizeChangedEventArgs *sizeChangedEventArgs) + { + std::shared_ptr<InspectableNativeWindow> host = mHost.lock(); + if (host) + { + // The size of the ISwapChainPanel control is returned in DIPs. + // We are keeping these in dips because the swapchain created for composition + // also uses dip units. This keeps dimensions, viewports, etc in the same unit. + // XAML Clients of the ISwapChainPanel are required to use dips to define their + // layout sizes as well. + ABI::Windows::Foundation::Size newSize; + HRESULT result = sizeChangedEventArgs->get_NewSize(&newSize); + if (SUCCEEDED(result)) + { + SIZE windowSize = { lround(newSize.Width), lround(newSize.Height) }; + host->setNewClientSize(windowSize); + } + } + + return S_OK; + } + + private: + std::weak_ptr<InspectableNativeWindow> mHost; +}; + +HRESULT GetSwapChainPanelSize(const ComPtr<ABI::Windows::UI::Xaml::Controls::ISwapChainPanel> &swapChainPanel, RECT *windowSize); +} +#endif // LIBANGLE_RENDERER_D3D_D3D11_WINRT_SWAPCHAINPANELNATIVEWINDOW_H_ |