diff options
Diffstat (limited to 'src/gui/rhi')
34 files changed, 16453 insertions, 10008 deletions
diff --git a/src/gui/rhi/qrhi.cpp b/src/gui/rhi/qrhi.cpp index 3b9c150166..a39709c726 100644 --- a/src/gui/rhi/qrhi.cpp +++ b/src/gui/rhi/qrhi.cpp @@ -1,23 +1,23 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhi_p_p.h" +#include "qrhi_p.h" #include <qmath.h> #include <QLoggingCategory> -#include "qrhinull_p_p.h" +#include "qrhinull_p.h" #ifndef QT_NO_OPENGL -#include "qrhigles2_p_p.h" +#include "qrhigles2_p.h" #endif #if QT_CONFIG(vulkan) -#include "qrhivulkan_p_p.h" +#include "qrhivulkan_p.h" #endif #ifdef Q_OS_WIN -#include "qrhid3d11_p_p.h" -#include "qrhid3d12_p_p.h" +#include "qrhid3d11_p.h" +#include "qrhid3d12_p.h" #endif -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) -#include "qrhimetal_p_p.h" +#if QT_CONFIG(metal) +#include "qrhimetal_p.h" #endif #include <memory> @@ -28,8 +28,10 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") /*! \class QRhi - \internal - \inmodule QtGui + \ingroup painting-3D + \inmodule QtGuiPrivate + \inheaderfile rhi/qrhi.h + \since 6.6 \brief Accelerated 2D/3D graphics API abstraction. @@ -40,52 +42,55 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \l{https://developer.apple.com/metal/}{Metal}, and \l{https://www.khronos.org/vulkan/}{Vulkan}. - Some of the main design goals are: - - \list - - \li Simple, minimal, understandable, extensible. Follow the proven path of the - Qt Quick scenegraph. - - \li Aim to be a product - and in the bigger picture, part of a product (Qt) - - that is usable out of the box both by internal (such as, Qt Quick) and, - eventually, external users. - - \li Not a complete 1:1 wrapper for any of the underlying APIs. The feature set - is tuned towards the needs of Qt's 2D and 3D offering (QPainter, Qt Quick, Qt - 3D Studio). Iterate and evolve in a sustainable manner. - - \li Intrinsically cross-platform, without reinventing: abstracting - cross-platform aspects of certain APIs (such as, OpenGL context creation and - windowing system interfaces, Vulkan instance and surface management) is not in - scope here. These are delegated to the existing QtGui facilities (QWindow, - QOpenGLContext, QVulkanInstance) and its backing QPA architecture. - - \endlist + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qrhi.h>}. Each QRhi instance is backed by a backend for a specific graphics API. The selection of the backend is a run time choice and is up to the application or library that creates the QRhi instance. Some backends are available on multiple platforms (OpenGL, Vulkan, Null), while APIs specific to a given platform are only available when running on the platform in question (Metal - on macOS/iOS/tvOS, Direct3D on Windows). + on macOS/iOS, Direct3D on Windows). The available backends currently are: \list - \li OpenGL 2.1 or OpenGL ES 2.0 or newer. Some extensions are utilized when - present, for example to enable multisample framebuffers. + \li OpenGL 2.1 / OpenGL ES 2.0 or newer. Some extensions and newer core + specification features are utilized when present, for example to enable + multisample framebuffers or compute shaders. Operating in core profile + contexts is supported as well. If necessary, applications can query the + \l{QRhi::Feature}{feature flags} at runtime to check for features that are + not supported in the OpenGL context backing the QRhi. The OpenGL backend + builds on QOpenGLContext, QOpenGLFunctions, and the related cross-platform + infrastructure of the Qt GUI module. - \li Direct3D 11.1 + \li Direct3D 11.1 or newer, with Shader Model 5.0 or newer. When the D3D + runtime has no support for 11.1 features or Shader Model 5.0, + initialization using an accelerated graphics device will fail, but using + the + \l{https://learn.microsoft.com/en-us/windows/win32/direct3darticles/directx-warp}{software + adapter} is still an option. - \li Direct3D 12 + \li Direct3D 12 on Windows 10 version 1703 and newer, with Shader Model 5.0 + or newer. Qt requires ID3D12Device2 to be present, hence the requirement + for at least version 1703 of Windows 10. The D3D12 device is by default + created with specifying a minimum feature level of + \c{D3D_FEATURE_LEVEL_11_0}. - \li Metal + \li Metal 1.2 or newer. - \li Vulkan 1.0, optionally with some extensions that are part of Vulkan 1.1 + \li Vulkan 1.0 or newer, optionally utilizing some Vulkan 1.1 level + features. - \li Null - A "dummy" backend that issues no graphics calls at all. + \li Null, a "dummy" backend that issues no graphics calls at all. \endlist @@ -95,15 +100,60 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") are then generated from that, together with reflection information (inputs, outputs, shader resources). This is then packed into easily and efficiently serializable QShader instances. The compilers and tools to generate such - shaders are not part of QRhi, but the core classes for using such shaders, - QShader and QShaderDescription, are. + shaders are not part of QRhi and the Qt GUI module, but the core classes + for using such shaders, QShader and QShaderDescription, are. The APIs and + tools for performing compilation and translation are part of the Qt Shader + Tools module. + + See the \l{RHI Window Example} for an introductory example of creating a + portable, cross-platform application that performs accelerated 3D rendering + onto a QWindow using QRhi. + + \section1 An Impression of the API - \section2 Design Fundamentals + To provide a quick look at the API with a short yet complete example that + does not involve window-related setup, the following is a complete, + runnable cross-platform application that renders 20 frames off-screen, and + then saves the generated images to files after reading back the texture + contents from the GPU. For an example that renders on-screen, which then + involves setting up a QWindow and a swapchain, refer to the + \l{RHI Window Example}. + + For brevity, the initialization of the QRhi is done based on the platform: + the sample code here chooses Direct 3D 12 on Windows, Metal on macOS and + iOS, and Vulkan otherwise. OpenGL and Direct 3D 11 are never used by this + application, but support for those could be introduced with a few + additional lines. + + \snippet rhioffscreen/main.cpp 0 + + The result of the application is 20 \c PNG images (frame0.png - + frame19.png). These contain a rotating triangle with varying opacity over a + green background. + + The vertex and fragment shaders are expected to be processed and packaged + into \c{.qsb} files. The Vulkan-compatible GLSL source code is the + following: + + \e color.vert + \snippet rhioffscreen/color.vert 0 + + \e color.frag + \snippet rhioffscreen/color.frag 0 + + To manually compile and transpile these shaders to a number of targets + (SPIR-V, HLSL, MSL, GLSL) and generate the \c{.qsb} files the application + loads at run time, run \c{qsb --qt6 color.vert -o color.vert.qsb} and + \c{qsb --qt6 color.frag -o color.frag.qsb}. Alternatively, the Qt Shader + Tools module offers build system integration for CMake, the + \c qt_add_shaders() CMake function, that can achieve the same at build time. + + \section1 Design Fundamentals A QRhi cannot be instantiated directly. Instead, use the create() function. Delete the QRhi instance normally to release the graphics device. - \section3 Resources + \section2 Resources Instances of classes deriving from QRhiResource, such as, QRhiBuffer, QRhiTexture, etc., encapsulate zero, one, or more native graphics @@ -111,10 +161,10 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") functions of the QRhi, such as, newBuffer(), newTexture(), newTextureRenderTarget(), newSwapChain(). - \badcode - vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertexData)); - if (!vbuf->create()) { error } - ... + \code + QRhiBuffer *vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertexData)); + if (!vbuf->create()) { error(); } + // ... delete vbuf; \endcode @@ -123,9 +173,10 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \li The returned value from functions like newBuffer() is always owned by the caller. - \li Just creating a QRhiResource subclass never allocates or initializes any - native resources. That is only done when calling the \c create() function of a - subclass, for example, QRhiBuffer::create() or QRhiTexture::create(). + \li Just creating an instance of a QRhiResource subclass never allocates or + initializes any native resources. That is only done when calling the + \c create() function of a subclass, for example, QRhiBuffer::create() or + QRhiTexture::create(). \li The exceptions are QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor(), @@ -149,15 +200,15 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \li Note that this does not mean that a QRhiResource can freely be destroy()'ed or deleted within a frame (that is, in a - \l{QRhiCommandBuffer::beginFrame()}{beginFrame()} - - \l{QRhiCommandBuffer::endFrame()}{endFrame()} section). As a general rule, - all referenced QRhiResource objects must stay unchanged until the frame is - submitted by calling \l{QRhiCommandBuffer::endFrame()}{endFrame()}. To ease - this, QRhiResource::deleteLater() is provided as a convenience. + \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} + section). As a general rule, all referenced QRhiResource objects must stay + unchanged until the frame is submitted by calling + \l{QRhi::endFrame()}{endFrame()}. To ease this, + QRhiResource::deleteLater() is provided as a convenience. \endlist - \section3 Command buffers and deferred command execution + \section2 Command buffers and deferred command execution Regardless of the design and capabilities of the underlying graphics API, all QRhi backends implement some level of command buffers. No @@ -184,7 +235,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") As a general rule, all referenced QRhiResource objects must stay valid and unmodified until the frame is submitted by calling - \l{QRhiCommandBuffer::endFrame()}{endFrame()}. On the other hand, calling + \l{QRhi::endFrame()}{endFrame()}. On the other hand, calling \l{QRhiResource::destroy()}{destroy()} or deleting the QRhiResource are always safe once the frame is submitted, regardless of the status of the underlying native resources (which may still be in use by the GPU - but @@ -192,10 +243,20 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") Unlike APIs like OpenGL, upload and copy type of commands cannot be mixed with draw commands. The typical renderer will involve a sequence similar to - the following: \c{(re)create resources} - \c{begin frame} - \c{record - uploads and copies} - \c{start renderpass} - \c{record draw calls} - \c{end - renderpass} - \c{end frame}. Recording copy type of operations happens via - QRhiResourceUpdateBatch. Such operations are committed typically on + the following: + + \list + \li (re)create resources + \li begin frame + \li record/issue uploads and copies + \li start recording a render pass + \li record draw calls + \li end render pass + \li end frame + \endlist + + Recording copy type of operations happens via QRhiResourceUpdateBatch. Such + operations are committed typically on \l{QRhiCommandBuffer::beginPass()}{beginPass()}. When working with legacy rendering engines designed for OpenGL, the @@ -214,7 +275,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") remain the primary way of operating since this is what fits Qt's various UI technologies best. - \section3 Threading + \section2 Threading A QRhi instance and the associated resources can be created and used on any thread but all usage must be limited to that one single thread. When @@ -243,7 +304,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") (gui) thread, but becomes important when a separate, dedicated render thread is used. - \section3 Resource synchronization + \section2 Resource synchronization QRhi does not expose APIs for resource barriers or image layout transitions. Such synchronization is done implicitly by the backends, where @@ -263,7 +324,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") different access (one for load, one for store) is supported even within the same pass. - \section3 Resource reuse + \section2 Resource reuse From the user's point of view a QRhiResource is reusable immediately after calling QRhiResource::destroy(). With the exception of swapchains, calling @@ -283,24 +344,24 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") though there is a good chance that under the hood the QRhiBuffer is now backed by a whole new native buffer. - \badcode - ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 256); + \code + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 256); ubuf->create(); - srb = rhi->newShaderResourceBindings() + QRhiShaderResourceBindings *srb = rhi->newShaderResourceBindings() srb->setBindings({ QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage | QRhiShaderResourceBinding::FragmentStage, ubuf) }); srb->create(); - ... + // ... // now in a later frame we need to grow the buffer to a larger size ubuf->setSize(512); ubuf->create(); // same as ubuf->destroy(); ubuf->create(); - // That's it, srb needs no changes whatsoever, any references in it to - // ubuf stay valid. When it comes to internal details, such as that + // srb needs no changes whatsoever, any references in it to ubuf + // stay valid. When it comes to internal details, such as that // ubuf may now be backed by a completely different native buffer // resource, that is is recognized and handled automatically by the // next setShaderResources(). @@ -315,7 +376,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") resource underneath, without having to update the QRhiTextureRenderTarget as that will be done implicitly in beginPass(). - \section3 Pooled objects + \section2 Pooled objects In addition to resources, there are pooled objects as well, such as, QRhiResourceUpdateBatch. An instance is retrieved via a \c next function, @@ -325,24 +386,25 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") QRhiCommandBuffer::beginPass() or QRhiCommandBuffer::endPass(). These functions take care of returning the batch to the pool. Alternatively, a batch can be "canceled" and returned to the pool without processing by - calling QRhiResourceUpdateBatch::destroy(). + calling QRhiResourceUpdateBatch::release(). A typical pattern is thus: - \badcode + \code QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); - ... + // ... resUpdates->updateDynamicBuffer(ubuf, 0, 64, mvp.constData()); if (!image.isNull()) { resUpdates->uploadTexture(texture, image); image = QImage(); } - ... + // ... QRhiCommandBuffer *cb = m_sc->currentFrameCommandBuffer(); + // note the last argument cb->beginPass(swapchain->currentFrameRenderTarget(), clearCol, clearDs, resUpdates); \endcode - \section3 Swapchain specifics + \section2 Swapchain specifics QRhiSwapChain features some special semantics due to the peculiar nature of swapchains. @@ -373,30 +435,174 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \endlist - \section3 Ownership + \section2 Ownership The general rule is no ownership transfer. Creating a QRhi with an already existing graphics device does not mean the QRhi takes ownership of the device object. Similarly, ownership is not given away when a device or texture object is "exported" via QRhi::nativeHandles() or - QRhiTexture::nativeHandles(). Most importantly, passing pointers in structs + QRhiTexture::nativeTexture(). Most importantly, passing pointers in structs and via setters does not transfer ownership. - \section2 Troubleshooting + \section1 Troubleshooting and Profiling + + \section2 Error reporting + + Functions such as \l QRhi::create() and the resource classes' \c create() + member functions (e.g., \l QRhiBuffer::create()) indicate failure with the + return value (\nullptr or + \c false, respectively). When working with QShader, \l QShader::fromSerialized() + returns an invalid QShader (for which \l{QShader::isValid()}{isValid()} returns + \c false) when the data passed to the function cannot be successfully deserialized. + Some functions, beginFrame() in particular, may also sometimes report "soft failures", + such as \l FrameOpSwapChainOutOfDate, which do not indicate an unrecoverable error, + but rather should be seen as a "try again later" response. - Errors are printed to the output via qWarning(). Additional debug messages - can be enabled via the following logging categories. Messages from these - categories are not printed by default unless explicitly enabled via - QLoggingCategory or the \c QT_LOGGING_RULES environment variable. + Warnings and errors may get printed at any time to the debug output via + qWarning(). It is therefore always advisable to inspect the output of the + application. + + Additional debug messages can be enabled via the following logging + categories. Messages from these categories are not printed by default + unless explicitly enabled via QLoggingCategory or the \c QT_LOGGING_RULES + environment variable. For better interoperation with Qt Quick, the + environment variable \c{QSG_INFO} also enables these debug prints. \list \li \c{qt.rhi.general} \endlist - It is strongly advised to inspect the output with the logging categories - (\c{qt.rhi.*}) enabled whenever a QRhi-based application is not behaving as - expected. For better interoperation with Qt Quick, the environment variable - \c{QSG_INFO} also enables these debug prints. + Additionally, applications can query the \l{QRhi::backendName()}{QRhi + backend name} and + \l{QRhi::driverInfo()}{graphics device information} from a successfully + initialized QRhi. This can then be printed to the user or stored in the + application logs even in production builds, if desired. + + \section2 Investigating rendering problems + + When the rendering results are not as expected, or the application is + experiencing problems, always consider checking with the the native 3D + APIs' debug and validation facilities. QRhi itself features limited error + checking since replicating the already existing, vast amount of + functionality in the underlying layers is not reasonable. + + \list + + \li For Vulkan, controlling the + \l{https://github.com/KhronosGroup/Vulkan-ValidationLayers}{Vulkan + Validation Layers} is not in the scope of the QRhi, but rather can be + achieved by configuring the \l QVulkanInstance with the appropriate layers. + For example, call \c{instance.setLayers({ "VK_LAYER_KHRONOS_validation" });} + before invoking \l{QVulkanInstance::create()}{create()} on the QVulkanInstance. + (note that this assumes that the validation layers are actually installed + and available, e.g. from the Vulkan SDK) By default, QVulkanInstance conveniently + redirects the Vulkan debug messages to qDebug, meaning the validation messages get + printed just like other Qt warnings. + + \li With Direct 3D 11 and 12, a graphics device with the debug layer + enabled can be requested by toggling the \c enableDebugLayer flag in the + appropriate \l{QRhiD3D11InitParams}{init params struct}. The messages appear on the + debug output, which is visible in Qt Creator's messages panel or via a tool + such as \l{https://learn.microsoft.com/en-us/sysinternals/downloads/debugview}{DebugView}. + + \li For Metal, controlling Metal Validation is outside of QRhi's scope. + Rather, to enable validation, run the application with the environment + variable \c{METAL_DEVICE_WRAPPER_TYPE=1} set, or run the application within + XCode. There may also be further settings and environment variable in modern + XCode and macOS versions. See for instance + \l{https://developer.apple.com/documentation/metal/diagnosing_metal_programming_issues_early}{this + page}. + + \endlist + + \section2 Frame captures and performance profiling + + A Qt application rendering with QRhi to a window while relying on a 3D API + under the hood, is, from the windowing and graphics pipeline perspective at + least, no different from any other (non-Qt) applications using the same 3D + API. This means that tools and practices for debugging and profiling + applications involving 3D graphics, such as games, all apply to such a Qt + application as well. + + A few examples of tools that can provide insights into the rendering + internals of Qt applications that use QRhi, which includes Qt Quick and Qt + Quick 3D based projects as well: + + \list + + \li \l{https://renderdoc.org/}{RenderDoc} allows taking frame captures and + introspecting the recorded commands and pipeline state on Windows and Linux + for applications using OpenGL, Vulkan, D3D11, or D3D12. When trying to + figure out why some parts of the 3D scene do not show up as expected, + RenderDoc is often a fast and efficient way to check the pipeline stages + and the related state and discover the missing or incorrect value. It is + also a tool that is actively used when developing Qt itself. + + \li For NVIDIA-based systems, + \l{https://developer.nvidia.com/nsight-graphics}{Nsight Graphics} provides + a graphics debugger tool on Windows and Linux. In addition to investigating the commands + in the frame and the pipeline, the vendor-specific tools allow looking at timings and + hardware performance information, which is not something simple frame captures can provide. + + \li For AMD-based systems, the \l{https://gpuopen.com/rgp/}{Radeon GPU + Profiler} can be used to gain deeper insights into the application's + rendering and its performance. + + \li As QRhi supports Direct 3D 12, using + \l{https://devblogs.microsoft.com/pix/download/}{PIX}, a performance tuning + and debugging tool for DirectX 12 games on Windows is an option as well. + + \li On macOS, + \l{https://developer.apple.com/documentation/metal/debugging_tools/viewing_your_gpu_workload_with_the_metal_debugger}{the + XCode Metal debugger} can be used to take and introspect frame + captures, to investigate performance details, and debug shaders. In macOS 13 it is also possible + to enable an overlay that displays frame rate and other information for any Metal-based window by + setting the environment variable \c{MTL_HUD_ENABLED=1}. + + \endlist + + On mobile and embedded platforms, there may be vendor and platform-specific + tools, provided by the GPU or SoC vendor, available to perform performance + profiling of application using OpenGL ES or Vulkan. + + When capturing frames, remember that objects and groups of commands can be + named via debug markers, as long as \l{QRhi::EnableDebugMarkers}{debug + markers were enabled} for the QRhi, and the graphics API in use supports + this. To annotate the command stream, call + \l{QRhiCommandBuffer::debugMarkBegin()}{debugMarkBegin()}, + \l{QRhiCommandBuffer::debugMarkEnd()}{debugMarkEnd()} and/or + \l{QRhiCommandBuffer::debugMarkMsg()}{debugMarkMsg()}. + This can be particularly useful in larger frames with multiple render passes. + Resources are named by calling \l{QRhiResource::setName()}{setName()} before create(). + + To perform basic timing measurements on the CPU and GPU side within the + application, \l QElapsedTimer and + \l QRhiCommandBuffer::lastCompletedGpuTime() can be used. The latter is + only available with select graphics APIs at the moment and requires opting + in via the \l QRhi::EnableTimestamps flag. + + \section2 Resource leak checking + + When destroying a QRhi object without properly destroying all buffers, + textures, and other resources created from it, warnings about this are + printed to the debug output whenever the application is a debug build, or + when the \c QT_RHI_LEAK_CHECK environment variable is set to a non-zero + value. This is a simple way to discover design issues around resource + handling within the application rendering logic. Note however that some + platforms and underlying graphics APIs may perform their own allocation and + resource leak detection as well, over which Qt will have no direct control. + For example, when using Vulkan, the memory allocator may raise failing + assertions in debug builds when resources that own graphics memory + allocations are not destroyed before the QRhi. In addition, the Vulkan + validation layer, when enabled, will issue warnings about native graphics + resources that were not released. Similarly, with Direct 3D warnings may + get printed about unreleased COM objects when the application does not + destroy the QRhi and its resources in the correct order. + + \sa {RHI Window Example}, QRhiCommandBuffer, QRhiResourceUpdateBatch, + QRhiShaderResourceBindings, QShader, QRhiBuffer, QRhiTexture, + QRhiRenderBuffer, QRhiSampler, QRhiTextureRenderTarget, + QRhiGraphicsPipeline, QRhiComputePipeline, QRhiSwapChain */ /*! @@ -415,13 +621,18 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \enum QRhi::Flag Describes what special features to enable. - \value EnableProfiling This flag has currently no effect. - \value EnableDebugMarkers Enables debug marker groups. Without this frame debugging features like making debug groups and custom resource name visible in external GPU debugging tools will not be available and functions - like QRhiCommandBuffer::debugMarkBegin() will become a no-op. Avoid - enabling in production builds as it may involve a performance penalty. + like QRhiCommandBuffer::debugMarkBegin() will become no-ops. Avoid enabling + in production builds as it may involve a small performance impact. Has no + effect when the QRhi::DebugMarkers feature is not reported as supported. + + \value EnableTimestamps Enables GPU timestamp collection. When not set, + QRhiCommandBuffer::lastCompletedGpuTime() always returns 0. Enable this + only when needed since there may be a small amount of extra work involved + (e.g. timestamp queries), depending on the underlying graphics API. Has no + effect when the QRhi::Timestamps feature is not reported as supported. \value PreferSoftwareRenderer Indicates that backends should prefer choosing an adapter or physical device that renders in software on the CPU. @@ -443,7 +654,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") mechanism because the cost of maintaining the related data structures is not insignificant with some backends. With Vulkan this feature maps directly to VkPipelineCache, vkGetPipelineCacheData and - VkPipelineCacheCreateInfo::pInitialData. With D3D11 there is no real + VkPipelineCacheCreateInfo::pInitialData. With Direct3D 11 there is no real pipline cache, but the results of HLSL->DXBC compilations are stored and can be serialized/deserialized via this mechanism. This allows skipping the time consuming D3DCompile() in future runs of the applications for shaders @@ -455,6 +666,17 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") mechanisms for shader/program binaries provided by Qt. Writing to those may get disabled whenever this flag is set since storing program binaries to multiple caches is not sensible. + + \value SuppressSmokeTestWarnings Indicates that, with backends where this + is relevant, certain, non-fatal QRhi::create() failures should not + produce qWarning() calls. For example, with D3D11, passing this flag + makes a number of warning messages (that appear due to QRhi::create() + failing) to become categorized debug prints instead under the commonly used + \c{qt.rhi.general} logging category. This can be used by engines, such as + Qt Quick, that feature fallback logic, i.e. they retry calling create() + with a different set of flags (such as, \l PreferSoftwareRenderer), in order + to hide the unconditional warnings from the output that would be printed + when the first create() attempt had failed. */ /*! @@ -492,8 +714,12 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") QRhiCommandBuffer::debugMarkBegin()) are supported. \value Timestamps Indicates that command buffer timestamps are supported. - Relevant for addGpuFrameTimeCallback(). Can be expected to be supported on - D3D11 and Vulkan, assuming the underlying implementation supports it. + Relevant for QRhiCommandBuffer::lastCompletedGpuTime(). This can be + expected to be supported on Metal, Vulkan, Direct 3D 11 and 12, and OpenGL + contexts of version 3.3 or newer. However, with some of these APIs support + for timestamp queries is technically optional, and therefore it cannot be + guaranteed that this feature is always supported with every implementation + of them. \value Instancing Indicates that instanced drawing is supported. In practice this feature will be unsupported with OpenGL ES 2.0 and OpenGL @@ -563,7 +789,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value WideLines Indicates that lines with a width other than 1 are supported. When reported as not supported, the line width set on the graphics pipeline state is ignored. This can always be false with some - backends (D3D11, Metal). With Vulkan, the value depends on the + backends (D3D11, D3D12, Metal). With Vulkan, the value depends on the implementation. With OpenGL, wide lines are not supported in core profile contexts. @@ -724,7 +950,7 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") with image load/store. This feature is only available with some backends as it does not map well to all graphics APIs, and it is only meant to provide support for special cases anyhow. In practice the feature can be expected to - be supported with Direct3D 11 and Vulkan. + be supported with Direct3D 11/12 and Vulkan. \value NonFillPolygonMode Indicates that setting a PolygonMode other than the default Fill is supported for QRhiGraphicsPipeline. A common use case @@ -745,8 +971,8 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") When not supported, build() will succeed but just show a warning message and the values of the target attributes will be broken. In practice this feature will be unsupported in some OpenGL ES 2.0 and OpenGL 2.x - implementations. Note that while D3D does support half precision input - attributes, it does not support the half3 type. The D3D backends pass + implementations. Note that while Direct3D 11/12 does support half precision + input attributes, it does not support the half3 type. The D3D backends pass half3 attributes as half4. To ensure cross platform compatibility, half3 inputs should be padded to 8 bytes. @@ -758,6 +984,58 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") \value ThreeDimensionalTextureMipmaps Indicates that generating 3D texture mipmaps are supported. In practice this feature will be unsupported with Direct 3D 12. + + \value MultiView Indicates that multiview, see e.g. + \l{https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_multiview.html}{VK_KHR_multiview} + is supported. With OpenGL ES 2.0, Direct 3D 11, and OpenGL (ES) + implementations without \c{GL_OVR_multiview2} this feature will not be + supported. With Vulkan 1.1 and newer, and Direct 3D 12 multiview is + typically supported. When reported as supported, creating a + QRhiTextureRenderTarget with a QRhiColorAttachment that references a texture + array and has \l{QRhiColorAttachment::setMultiViewCount()}{multiViewCount} + set enables recording a render pass that uses multiview rendering. In addition, + any QRhiGraphicsPipeline used in that render pass must have + \l{QRhiGraphicsPipeline::setMultiViewCount()}{the same view count set}. Note that + multiview is only available in combination with 2D texture arrays. It cannot + be used to optimize the rendering into individual textures (e.g. two, for + the left and right eyes). Rather, the target of a multiview render pass is + always a texture array, automatically rendering to the layer (array element) + corresponding to each view. Therefore this feature implies \l TextureArrays + as well. Multiview rendering is not supported in combination with + tessellation or geometry shaders. See QRhiColorAttachment::setMultiViewCount() + for further details on multiview rendering. This enum value has been introduced in Qt 6.7. + + \value TextureViewFormat Indicates that setting a + \l{QRhiTexture::setWriteViewFormat()}{view format} on a QRhiTexture is + effective. When reported as supported, setting the read (sampling) or write + (render target / image load-store) view mode changes the texture's viewing + format. When unsupported, setting a view format has no effect. Note that Qt + has no knowledge or control over format compatibility or resource view rules + in the underlying 3D API and its implementation. Passing in unsuitable, + incompatible formats may lead to errors and unspecified behavior. This is + provided mainly to allow "casting" rendering into a texture created with an + sRGB format to non-sRGB to avoid the unwanted linear->sRGB conversion on + shader writes. Other types of casting may or may not be functional, + depending on the underlying API. Currently implemented for Vulkan and Direct + 3D 12. With D3D12 the feature is available only if + \c CastingFullyTypedFormatSupported is supported, see + \l{https://microsoft.github.io/DirectX-Specs/d3d/RelaxedCasting.html} (and + note that QRhi always uses fully typed formats for textures.) This enum + value has been introduced in Qt 6.8. + + \value ResolveDepthStencil Indicates that resolving a multisample depth or + depth-stencil texture is supported. Otherwise, + \l{QRhiTextureRenderTargetDescription::setDepthResolveTexture()}{setting a + depth resolve texture} is not functional and must be avoided. Direct 3D 11 + and 12 have no support for resolving depth/depth-stencil formats, and + therefore this feature will never be supported with those. Vulkan 1.0 has no + API to request resolving a depth-stencil attachment. Therefore, with Vulkan + this feature will only be supported with Vulkan 1.2 and up, and on 1.1 + implementations with the appropriate extensions present. This feature is + provided for the rare case when resolving into a non-multisample depth + texture becomes necessary, for example when rendering into an + OpenXR-provided depth texture (XR_KHR_composition_layer_depth). This enum + value has been introduced in Qt 6.8. */ /*! @@ -869,22 +1147,28 @@ Q_LOGGING_CATEGORY(QRHI_LOG_INFO, "qt.rhi.general") /*! \class QRhiInitParams - \internal \inmodule QtGui + \since 6.6 \brief Base class for backend-specific initialization parameters. Contains fields that are relevant to all backends. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! \class QRhiDepthStencilClearValue - \internal \inmodule QtGui + \since 6.6 \brief Specifies clear values for a depth or stencil buffer. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiDepthStencilClearValue::QRhiDepthStencilClearValue() + \fn QRhiDepthStencilClearValue::QRhiDepthStencilClearValue() = default Constructs a depth/stencil clear value with depth clear value 1.0f and stencil clear value 0. @@ -901,6 +1185,26 @@ QRhiDepthStencilClearValue::QRhiDepthStencilClearValue(float d, quint32 s) } /*! + \fn float QRhiDepthStencilClearValue::depthClearValue() const + \return the depth clear value. In most cases this is 1.0f. + */ + +/*! + \fn void QRhiDepthStencilClearValue::setDepthClearValue(float d) + Sets the depth clear value to \a d. + */ + +/*! + \fn quint32 QRhiDepthStencilClearValue::stencilClearValue() const + \return the stencil clear value. In most cases this is 0. + */ + +/*! + \fn void QRhiDepthStencilClearValue::setStencilClearValue(quint32 s) + Sets the stencil clear value to \a s. + */ + +/*! \fn bool QRhiDepthStencilClearValue::operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept \return \c true if the values in the two QRhiDepthStencilClearValue objects @@ -934,8 +1238,8 @@ QDebug operator<<(QDebug dbg, const QRhiDepthStencilClearValue &v) /*! \class QRhiViewport - \internal \inmodule QtGui + \since 6.6 \brief Specifies a viewport rectangle. Used with QRhiCommandBuffer::setViewport(). @@ -945,20 +1249,23 @@ QDebug operator<<(QDebug dbg, const QRhiDepthStencilClearValue &v) Typical usage is like the following: - \badcode + \code const QSize outputSizeInPixels = swapchain->currentPixelSize(); const QRhiViewport viewport(0, 0, outputSizeInPixels.width(), outputSizeInPixels.height()); - cb->beginPass(swapchain->currentFrameRenderTarget(), { 0, 0, 0, 1 }, { 1, 0 }); + cb->beginPass(swapchain->currentFrameRenderTarget(), Qt::black, { 1.0f, 0 }); cb->setGraphicsPipeline(ps); cb->setViewport(viewport); - ... + // ... \endcode + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setViewport(), QRhi::clipSpaceCorrMatrix(), QRhiScissor */ /*! - \fn QRhiViewport::QRhiViewport() + \fn QRhiViewport::QRhiViewport() = default Constructs a viewport description with an empty rectangle and a depth range of 0.0f - 1.0f. @@ -984,6 +1291,41 @@ QRhiViewport::QRhiViewport(float x, float y, float w, float h, float minDepth, f } /*! + \fn std::array<float, 4> QRhiViewport::viewport() const + \return the viewport x, y, width, and height. + */ + +/*! + \fn void QRhiViewport::setViewport(float x, float y, float w, float h) + Sets the viewport's position and size to \a x, \a y, \a w, and \a h. + + \note Viewports are specified in a coordinate system that has its origin in + the bottom-left. + */ + +/*! + \fn float QRhiViewport::minDepth() const + \return the minDepth value of the depth range of the viewport. + */ + +/*! + \fn void QRhiViewport::setMinDepth(float minDepth) + Sets the \a minDepth of the depth range of the viewport. + By default this is set to 0.0f. + */ + +/*! + \fn float QRhiViewport::maxDepth() const + \return the maxDepth value of the depth range of the viewport. + */ + +/*! + \fn void QRhiViewport::setMaxDepth(float maxDepth) + Sets the \a maxDepth of the depth range of the viewport. + By default this is set to 1.0f. + */ + +/*! \fn bool QRhiViewport::operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept \return \c true if the values in the two QRhiViewport objects @@ -1021,8 +1363,8 @@ QDebug operator<<(QDebug dbg, const QRhiViewport &v) /*! \class QRhiScissor - \internal \inmodule QtGui + \since 6.6 \brief Specifies a scissor rectangle. Used with QRhiCommandBuffer::setScissor(). Setting a scissor rectangle is @@ -1036,11 +1378,14 @@ QDebug operator<<(QDebug dbg, const QRhiViewport &v) appropriate. Therefore, any rendering logic targeting OpenGL can feed scissor rectangles into QRhiScissor as-is, without any adaptation. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setScissor(), QRhiViewport */ /*! - \fn QRhiScissor::QRhiScissor() + \fn QRhiScissor::QRhiScissor() = default Constructs an empty scissor. */ @@ -1061,6 +1406,19 @@ QRhiScissor::QRhiScissor(int x, int y, int w, int h) } /*! + \fn std::array<int, 4> QRhiScissor::scissor() const + \return the scissor position and size. + */ + +/*! + \fn void QRhiScissor::setScissor(int x, int y, int w, int h) + Sets the scissor position and size to \a x, \a y, \a w, \a h. + + \note The position is always expected to be specified in a coordinate + system that has its origin in the bottom-left corner, like OpenGL. + */ + +/*! \fn bool QRhiScissor::operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept \return \c true if the values in the two QRhiScissor objects @@ -1096,8 +1454,8 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) /*! \class QRhiVertexInputBinding - \internal \inmodule QtGui + \since 6.6 \brief Describes a vertex input binding. Specifies the stride (in bytes, must be a multiple of 4), the @@ -1115,7 +1473,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) format in a buffer (or separate buffers even). Defining two bindings could then be done like this: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 3 * sizeof(float) }, @@ -1132,7 +1490,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) vertices, assuming we have a single buffer with first the positions and then the texture coordinates: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBindings[] = { { cubeBuf, 0 }, { cubeBuf, 36 * 3 * sizeof(float) } @@ -1148,6 +1506,9 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) \note the stride must always be a multiple of 4. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setVertexInput() */ @@ -1160,7 +1521,7 @@ QDebug operator<<(QDebug dbg, const QRhiScissor &s) */ /*! - \fn QRhiVertexInputBinding::QRhiVertexInputBinding() + \fn QRhiVertexInputBinding::QRhiVertexInputBinding() = default Constructs a default vertex input binding description. */ @@ -1180,6 +1541,36 @@ QRhiVertexInputBinding::QRhiVertexInputBinding(quint32 stride, Classification cl } /*! + \fn quint32 QRhiVertexInputBinding::stride() const + \return the stride in bytes. + */ + +/*! + \fn void QRhiVertexInputBinding::setStride(quint32 s) + Sets the stride to \a s. + */ + +/*! + \fn QRhiVertexInputBinding::Classification QRhiVertexInputBinding::classification() const + \return the input data classification. + */ + +/*! + \fn void QRhiVertexInputBinding::setClassification(Classification c) + Sets the input data classification \a c. By default this is set to PerVertex. + */ + +/*! + \fn quint32 QRhiVertexInputBinding::instanceStepRate() const + \return the instance step rate. + */ + +/*! + \fn void QRhiVertexInputBinding::setInstanceStepRate(quint32 rate) + Sets the instance step \a rate. By default this is set to 1. + */ + +/*! \fn bool QRhiVertexInputBinding::operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept \return \c true if the values in the two QRhiVertexInputBinding objects @@ -1213,14 +1604,15 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) /*! \class QRhiVertexInputAttribute - \internal \inmodule QtGui + \since 6.6 \brief Describes a single vertex input element. The members specify the binding number, location, format, and offset for a single vertex input element. - \note For HLSL it is assumed that the vertex shader uses + \note For HLSL it is assumed that the vertex shader translated from SPIR-V + uses \c{TEXCOORD<location>} as the semantic for each input. Hence no separate semantic name and index. @@ -1236,7 +1628,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) non-interleaved format in a buffer (or separate buffers even). Once two bindings are defined, the attributes could be specified as: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 3 * sizeof(float) }, @@ -1253,7 +1645,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) vertices, assuming we have a single buffer with first the positions and then the texture coordinates: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBindings[] = { { cubeBuf, 0 }, { cubeBuf, 36 * 3 * sizeof(float) } @@ -1265,7 +1657,7 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) binding, with multiple attributes referring to that same buffer binding point: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 5 * sizeof(float) } @@ -1278,11 +1670,14 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) and then: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBinding(interleavedCubeBuf, 0); cb->setVertexInput(0, 1, &vbufBinding); \endcode + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiCommandBuffer::setVertexInput() */ @@ -1305,20 +1700,30 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputBinding &b) \value SInt3 Three component signed integer vector \value SInt2 Two component signed integer vector \value SInt Signed integer - \value Half4 Four component half precision (16bit) float vector - \value Half3 Three component half precision (16bit) float vector - \value Half2 Two component half precision (16bit) float vector - \value Half half precision (16bit) float + \value Half4 Four component half precision (16 bit) float vector + \value Half3 Three component half precision (16 bit) float vector + \value Half2 Two component half precision (16 bit) float vector + \value Half Half precision (16 bit) float + \value UShort4 Four component unsigned short (16 bit) integer vector + \value UShort3 Three component unsigned short (16 bit) integer vector + \value UShort2 Two component unsigned short (16 bit) integer vector + \value UShort Unsigned short (16 bit) integer + \value SShort4 Four component signed short (16 bit) integer vector + \value SShort3 Three component signed short (16 bit) integer vector + \value SShort2 Two component signed short (16 bit) integer vector + \value SShort Signed short (16 bit) integer \note Support for half precision floating point attributes is indicated at - run time by the QRhi::Feature::HalfAttributes feature flag. Note that D3D - supports half input attributes, but does not support the Half3 type. The - D3D backends pass through Half3 as Half4. To ensure cross platform - compatibility, Half3 inputs should be padded to 8 bytes. + run time by the QRhi::Feature::HalfAttributes feature flag. + + \note Direct3D 11/12 supports 16 bit input attributes, but does not support + the Half3, UShort3 or SShort3 types. The D3D backends pass through Half3 as + Half4, UShort3 as UShort4, and SShort3 as SShort4. To ensure cross platform + compatibility, 16 bit inputs should be padded to 8 bytes. */ /*! - \fn QRhiVertexInputAttribute::QRhiVertexInputAttribute() + \fn QRhiVertexInputAttribute::QRhiVertexInputAttribute() = default Constructs a default vertex input attribute description. */ @@ -1343,6 +1748,67 @@ QRhiVertexInputAttribute::QRhiVertexInputAttribute(int binding, int location, Fo } /*! + \fn int QRhiVertexInputAttribute::binding() const + \return the binding point index. + */ + +/*! + \fn void QRhiVertexInputAttribute::setBinding(int b) + Sets the binding point index to \a b. + By default this is set to 0. + */ + +/*! + \fn int QRhiVertexInputAttribute::location() const + \return the location of the vertex input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setLocation(int loc) + Sets the location of the vertex input element to \a loc. + By default this is set to 0. + */ + +/*! + \fn QRhiVertexInputAttribute::Format QRhiVertexInputAttribute::format() const + \return the format of the vertex input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setFormat(Format f) + Sets the format of the vertex input element to \a f. + By default this is set to Float4. + */ + +/*! + \fn quint32 QRhiVertexInputAttribute::offset() const + \return the byte offset for the input element. + */ + +/*! + \fn void QRhiVertexInputAttribute::setOffset(quint32 ofs) + Sets the byte offset for the input element to \a ofs. + */ + +/*! + \fn int QRhiVertexInputAttribute::matrixSlice() const + + \return the matrix slice if the input element corresponds to a row or + column of a matrix, or -1 if not relevant. + */ + +/*! + \fn void QRhiVertexInputAttribute::setMatrixSlice(int slice) + + Sets the matrix \a slice. By default this is set to -1, and should be set + to a >= 0 value only when this attribute corresponds to a row or column of + a matrix (for example, a 4x4 matrix becomes 4 vec4s, consuming 4 + consecutive vertex input locations), in which case it is the index of the + row or column. \c{location - matrixSlice} must always be equal to the \c + location for the first row or column of the unrolled matrix. + */ + +/*! \fn bool QRhiVertexInputAttribute::operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept \return \c true if the values in the two QRhiVertexInputAttribute objects @@ -1465,6 +1931,24 @@ quint32 QRhiImplementation::byteSizePerVertexForVertexInputFormat(QRhiVertexInpu case QRhiVertexInputAttribute::Half: return sizeof(qfloat16); + case QRhiVertexInputAttribute::UShort4: + return 4 * sizeof(quint16); + case QRhiVertexInputAttribute::UShort3: + return 4 * sizeof(quint16); // ivec3 still takes 8 bytes + case QRhiVertexInputAttribute::UShort2: + return 2 * sizeof(quint16); + case QRhiVertexInputAttribute::UShort: + return sizeof(quint16); + + case QRhiVertexInputAttribute::SShort4: + return 4 * sizeof(qint16); + case QRhiVertexInputAttribute::SShort3: + return 4 * sizeof(qint16); // uvec3 still takes 8 bytes + case QRhiVertexInputAttribute::SShort2: + return 2 * sizeof(qint16); + case QRhiVertexInputAttribute::SShort: + return sizeof(qint16); + default: Q_UNREACHABLE_RETURN(1); } @@ -1472,21 +1956,102 @@ quint32 QRhiImplementation::byteSizePerVertexForVertexInputFormat(QRhiVertexInpu /*! \class QRhiVertexInputLayout - \internal \inmodule QtGui + \since 6.6 \brief Describes the layout of vertex inputs consumed by a vertex shader. The vertex input layout is defined by the collections of QRhiVertexInputBinding and QRhiVertexInputAttribute. + + As an example, let's assume that we have a single buffer with 3 component + vertex positions and 2 component UV coordinates interleaved (\c x, \c y, \c + z, \c u, \c v), that the position and UV are expected at input locations 0 + and 1 by the vertex shader, and that the vertex buffer will be bound at + binding point 0 using + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()} later on: + + \code + QRhiVertexInputLayout inputLayout; + inputLayout.setBindings({ + { 5 * sizeof(float) } + }); + inputLayout.setAttributes({ + { 0, 0, QRhiVertexInputAttribute::Float3, 0 }, + { 0, 1, QRhiVertexInputAttribute::Float2, 3 * sizeof(float) } + }); + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiVertexInputLayout::QRhiVertexInputLayout() + \fn QRhiVertexInputLayout::QRhiVertexInputLayout() = default Constructs an empty vertex input layout description. */ /*! + \fn void QRhiVertexInputLayout::setBindings(std::initializer_list<QRhiVertexInputBinding> list) + Sets the bindings from the specified \a list. + */ + +/*! + \fn template<typename InputIterator> void QRhiVertexInputLayout::setBindings(InputIterator first, InputIterator last) + Sets the bindings using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::cbeginBindings() const + \return a const iterator pointing to the first item in the binding list. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::cendBindings() const + \return a const iterator pointing just after the last item in the binding list. + */ + +/*! + \fn const QRhiVertexInputBinding *QRhiVertexInputLayout::bindingAt(qsizetype index) const + \return the binding at the given \a index. + */ + +/*! + \fn qsizetype QRhiVertexInputLayout::bindingCount() const + \return the number of bindings. + */ + +/*! + \fn void QRhiVertexInputLayout::setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) + Sets the attributes from the specified \a list. + */ + +/*! + \fn template<typename InputIterator> void QRhiVertexInputLayout::setAttributes(InputIterator first, InputIterator last) + Sets the attributes using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::cbeginAttributes() const + \return a const iterator pointing to the first item in the attribute list. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::cendAttributes() const + \return a const iterator pointing just after the last item in the attribute list. + */ + +/*! + \fn const QRhiVertexInputAttribute *QRhiVertexInputLayout::attributeAt(qsizetype index) const + \return the attribute at the given \a index. + */ + +/*! + \fn qsizetype QRhiVertexInputLayout::attributeCount() const + \return the number of attributes. + */ + +/*! \fn bool QRhiVertexInputLayout::operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept \return \c true if the values in the two QRhiVertexInputLayout objects @@ -1519,9 +2084,39 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputLayout &v) /*! \class QRhiShaderStage - \internal \inmodule QtGui + \since 6.6 \brief Specifies the type and the shader code for a shader stage in the pipeline. + + When setting up a QRhiGraphicsPipeline, a collection of shader stages are + specified. The QRhiShaderStage contains a QShader and some associated + metadata, such as the graphics pipeline stage, and the + \l{QShader::Variant}{shader variant} to select. There is no need to specify + the shader language or version because the QRhi backend in use at runtime + will take care of choosing the appropriate shader version from the + collection within the QShader. + + The typical usage is in combination with + QRhiGraphicsPipeline::setShaderStages(), shown here with a simple approach + to load the QShader from \c{.qsb} files generated offline or at build time: + + \code + QShader getShader(const QString &name) + { + QFile f(name); + return f.open(QIODevice::ReadOnly) ? QShader::fromSerialized(f.readAll()) : QShader(); + } + + QShader vs = getShader("material.vert.qsb"); + QShader fs = getShader("material.frag.qsb"); + pipeline->setShaderStages({ + { QRhiShaderStage::Vertex, vs }, + { QRhiShaderStage::Fragment, fs } + }); + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1546,13 +2141,46 @@ QDebug operator<<(QDebug dbg, const QRhiVertexInputLayout &v) */ /*! - \fn QRhiShaderStage::QRhiShaderStage() + \fn QRhiShaderStage::QRhiShaderStage() = default Constructs a shader stage description for the vertex stage with an empty QShader. */ /*! + \fn QRhiShaderStage::Type QRhiShaderStage::type() const + \return the type of the stage. + */ + +/*! + \fn void QRhiShaderStage::setType(Type t) + + Sets the type of the stage to \a t. Setters should rarely be needed in + pratice. Most applications will likely use the QRhiShaderStage constructor + in most cases. + */ + +/*! + \fn QShader QRhiShaderStage::shader() const + \return the QShader to be used for this stage in the graphics pipeline. + */ + +/*! + \fn void QRhiShaderStage::setShader(const QShader &s) + Sets the shader collection \a s. + */ + +/*! + \fn QShader::Variant QRhiShaderStage::shaderVariant() const + \return the requested shader variant. + */ + +/*! + \fn void QRhiShaderStage::setShaderVariant(QShader::Variant v) + Sets the requested shader variant \a v. + */ + +/*! Constructs a shader stage description with the \a type of the stage and the \a shader. @@ -1602,12 +2230,14 @@ QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) /*! \class QRhiColorAttachment - \internal \inmodule QtGui + \since 6.6 \brief Describes the a single color attachment of a render target. A color attachment is either a QRhiTexture or a QRhiRenderBuffer. The - former, when texture() is set, is used in most cases. + former, i.e. when texture() is set, is used in most cases. + QRhiColorAttachment is commonly used in combination with + QRhiTextureRenderTargetDescription. \note texture() and renderBuffer() cannot be both set (be non-null at the same time). @@ -1634,10 +2264,15 @@ QDebug operator<<(QDebug dbg, const QRhiShaderStage &s) \note when resolving is enabled, the multisample data may not be written out at all. This means that the multisample texture() must not be used afterwards with shaders for sampling when resolveTexture() is set. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiTextureRenderTargetDescription */ /*! - \fn QRhiColorAttachment::QRhiColorAttachment() + \fn QRhiColorAttachment::QRhiColorAttachment() = default Constructs an empty color attachment description. */ @@ -1661,9 +2296,194 @@ QRhiColorAttachment::QRhiColorAttachment(QRhiRenderBuffer *renderBuffer) } /*! + \fn QRhiTexture *QRhiColorAttachment::texture() const + + \return the texture this attachment description references, or \nullptr if + there is none. + */ + +/*! + \fn void QRhiColorAttachment::setTexture(QRhiTexture *tex) + + Sets the texture \a tex. + + \note texture() and renderBuffer() cannot be both set (be non-null at the + same time). + */ + +/*! + \fn QRhiRenderBuffer *QRhiColorAttachment::renderBuffer() const + + \return the renderbuffer this attachment description references, or + \nullptr if there is none. + + In practice associating a QRhiRenderBuffer with a QRhiColorAttachment makes + the most sense when setting up multisample rendering via a multisample + \l{QRhiRenderBuffer::Type}{color} renderbuffer that is then resolved into a + non-multisample texture at the end of the render pass. + */ + +/*! + \fn void QRhiColorAttachment::setRenderBuffer(QRhiRenderBuffer *rb) + + Sets the renderbuffer \a rb. + + \note texture() and renderBuffer() cannot be both set (be non-null at the + same time). + */ + +/*! + \fn int QRhiColorAttachment::layer() const + \return the layer index (cubemap face or array layer). 0 by default. + */ + +/*! + \fn void QRhiColorAttachment::setLayer(int layer) + Sets the \a layer index. + */ + +/*! + \fn int QRhiColorAttachment::level() const + \return the mip level. 0 by default. + */ + +/*! + \fn void QRhiColorAttachment::setLevel(int level) + Sets the mip \a level. + */ + +/*! + \fn QRhiTexture *QRhiColorAttachment::resolveTexture() const + + \return the resolve texture this attachment description references, or + \nullptr if there is none. + + Setting a non-null resolve texture is applicable when the attachment + references a multisample texture or renderbuffer. The QRhiTexture in the + resolveTexture() is then a non-multisample 2D texture (or texture array) + with the same size (but a sample count of 1). The multisample content is + automatically resolved into this texture at the end of each render pass. + */ + +/*! + \fn void QRhiColorAttachment::setResolveTexture(QRhiTexture *tex) + + Sets the resolve texture \a tex. + + \a tex is expected to be a 2D texture or a 2D texture array. In either + case, resolving targets a single mip level of a single layer (array + element) of \a tex. The mip level and array layer are specified by + resolveLevel() and resolveLayer(). + + An exception is \l{setMultiViewCount()}{multiview}: when the color + attachment is associated with a texture array and multiview is enabled, the + resolve texture must also be a texture array with sufficient elements for + all views. In this case all elements that correspond to views are resolved + automatically; the behavior is similar to the following pseudo-code: + \badcode + for (i = 0; i < multiViewCount(); ++i) + resolve texture's layer() + i into resolveTexture's resolveLayer() + i + \endcode + + Setting a non-multisample texture to resolve a multisample texture or + renderbuffer automatically at the end of the render pass is often + preferable to working with multisample textures (and not setting a resolve + texture), because it avoids the need for writing dedicated fragment shaders + that work exclusively with multisample textures (\c sampler2DMS, \c + texelFetch, etc.), and rather allows using the same shader as one would if + the attachment's texture was not multisampled to begin with. This comes at + the expense of an additional resource (the non-multisample \a tex). + */ + +/*! + \fn int QRhiColorAttachment::resolveLayer() const + \return the currently set resolve texture layer. Defaults to 0. + */ + +/*! + \fn void QRhiColorAttachment::setResolveLayer(int layer) + Sets the resolve texture \a layer to use. + */ + +/*! + \fn int QRhiColorAttachment::resolveLevel() const + \return the currently set resolve texture mip level. Defaults to 0. + */ + +/*! + \fn void QRhiColorAttachment::setResolveLevel(int level) + Sets the resolve texture mip \a level to use. + */ + +/*! + \fn int QRhiColorAttachment::multiViewCount() const + + \return the currently set number of views. Defaults to 0 which indicates + the render target with this color attachment is not going to be used with + multiview rendering. + + \since 6.7 + */ + +/*! + \fn void QRhiColorAttachment::setMultiViewCount(int count) + + Sets the view \a count. Setting a value larger than 1 indicates that the + render target with this color attachment is going to be used with multiview + rendering. The default value is 0. Values smaller than 2 indicate no + multiview rendering. + + When \a count is set to \c 2 or greater, the color attachment must be + associated with a 2D texture array. layer() and multiViewCount() together + define the range of texture array elements that are targeted during + multiview rendering. + + For example, if \c layer is \c 0 and \c multiViewCount is \c 2, the texture + array must have 2 (or more) elements, and the multiview rendering will + target elements 0 and 1. The \c{gl_ViewIndex} variable in the shaders has a + value of \c 0 or \c 1 then, where view \c 0 corresponds to the texture array + element \c 0, and view \c 1 to the array element \c 1. + + \note Setting a \a count larger than 1, using a texture array as texture(), + and calling \l{QRhiCommandBuffer::beginPass()}{beginPass()} on a + QRhiTextureRenderTarget with this color attachment implies multiview + rendering for the entire render pass. multiViewCount() should not be set + unless multiview rendering is wanted. Multiview cannot be used with texture + types other than 2D texture arrays. (although 3D textures may work, + depending on the graphics API and backend; applications are nonetheless + advised not to rely on that and only use 2D texture arrays as the render + targets of multiview rendering) + + See + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt}{GL_OVR_multiview} + for more details regarding multiview rendering. Do note that Qt requires + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview2.txt}{GL_OVR_multiview2} + as well, when running on OpenGL (ES). + + Multiview rendering is available only when the + \l{QRhi::MultiView}{MultiView} feature is reported as supported from + \l{QRhi::isFeatureSupported()}{isFeatureSupported()}. + + \note For portability, be aware of limitations that exist for multiview + rendering with some of the graphics APIs. It is recommended that multiview + render passes do not rely on any of the features that + \l{https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt}{GL_OVR_multiview} + declares as unsupported. The one exception is shader stage outputs other + than \c{gl_Position} depending on \c{gl_ViewIndex}: that can be relied on + (even with OpenGL) because QRhi never reports multiview as supported without + \c{GL_OVR_multiview2} also being present. + + \note Multiview rendering is not supported in combination with tessellation + or geometry shaders, even though some implementations of some graphics APIs + may allow this. + + \since 6.7 + */ + +/*! \class QRhiTextureRenderTargetDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes the color and depth or depth/stencil attachments of a render target. A texture render target has zero or more textures as color attachments, @@ -1672,10 +2492,97 @@ QRhiColorAttachment::QRhiColorAttachment(QRhiRenderBuffer *renderBuffer) \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be non-null at the same time). + + Let's look at some example usages in combination with + QRhiTextureRenderTarget. + + Due to the constructors, the targeting a texture (and no depth/stencil + buffer) is simple: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(256, 256), 1, QRhiTexture::RenderTarget); + texture->create(); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ texture })); + \endcode + + The following creates a texture render target that is set up to target mip + level #2 of a texture: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1, QRhiTexture::RenderTarget | QRhiTexture::MipMapped); + texture->create(); + QRhiColorAttachment colorAtt(texture); + colorAtt.setLevel(2); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ colorAtt }); + \endcode + + Another example, this time to render into a depth texture: + + \code + QRhiTexture *shadowMap = rhi->newTexture(QRhiTexture::D32F, QSize(1024, 1024), 1, QRhiTexture::RenderTarget); + shadowMap->create(); + QRhiTextureRenderTargetDescription rtDesc; + rtDesc.setDepthTexture(shadowMap); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget(rtDesc); + \endcode + + A very common case, having a texture as the color attachment and a + renderbuffer as depth/stencil to enable depth testing: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1. QRhiTexture::RenderTarget); + texture->create(); + QRhiRenderBuffer *depthStencil = rhi->newRenderBuffer(QRhiRenderBuffer::DepthStencil, QSize(512, 512)); + depthStencil->create(); + QRhiTextureRenderTargetDescription rtDesc({ texture }, depthStencil); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget(rtDesc); + \endcode + + Finally, to enable multisample rendering in a portable manner (so also + supporting OpenGL ES 3.0), using a QRhiRenderBuffer as the (multisample) + color buffer and then resolving into a regular (non-multisample) 2D + texture. To enable depth testing, a depth-stencil buffer, which also must + use the same sample count, is used as well: + + \code + QRhiRenderBuffer *colorBuffer = rhi->newRenderBuffer(QRhiRenderBuffer::Color, QSize(512, 512), 4); // 4x MSAA + colorBuffer->create(); + QRhiRenderBuffer *depthStencil = rhi->newRenderBuffer(QRhiRenderBuffer::DepthStencil, QSize(512, 512), 4); + depthStencil->create(); + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512), 1, QRhiTexture::RenderTarget); + texture->create(); + QRhiColorAttachment colorAtt(colorBuffer); + colorAtt.setResolveTexture(texture); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ colorAtt, depthStencil }); + \endcode + + \note when multisample resolving is enabled, the multisample data may not be + written out at all. This means that the multisample texture in a color + attachment must not be used afterwards with shaders for sampling (or other + purposes) whenever a resolve texture is set, since the multisample color + buffer is merely an intermediate storage then that gets no data written back + on some GPU architectures at all. See + \l{QRhiTextureRenderTarget::Flag}{PreserveColorContents} for more details. + + \note When using setDepthTexture(), not setDepthStencilBuffer(), and the + depth (stencil) data is not of interest afterwards, set the + DoNotStoreDepthStencilContents flag on the QRhiTextureRenderTarget. This + allows indicating to the underlying 3D API that the depth/stencil data can + be discarded, leading potentially to better performance with tiled GPU + architectures. When the depth-stencil buffer is a QRhiRenderBuffer (and also + for the multisample color texture, see previous note) this is implicit, but + with a depth (stencil) QRhiTexture the intention needs to be declared + explicitly. By default QRhi assumes that the data is of interest (e.g., the + depth texture is sampled in a shader afterwards). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiColorAttachment, QRhiTextureRenderTarget */ /*! - \fn QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription() + \fn QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription() = default Constructs an empty texture render target description. */ @@ -1717,9 +2624,125 @@ QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription(const QRh } /*! + \fn void QRhiTextureRenderTargetDescription::setColorAttachments(std::initializer_list<QRhiColorAttachment> list) + Sets the \a list of color attachments. + */ + +/*! + \fn template<typename InputIterator> void QRhiTextureRenderTargetDescription::setColorAttachments(InputIterator first, InputIterator last) + Sets the list of color attachments via the iterators \a first and \a last. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::cbeginColorAttachments() const + \return a const iterator pointing to the first item in the attachment list. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::cendColorAttachments() const + \return a const iterator pointing just after the last item in the attachment list. + */ + +/*! + \fn const QRhiColorAttachment *QRhiTextureRenderTargetDescription::colorAttachmentAt(qsizetype index) const + \return the color attachment at the specified \a index. + */ + +/*! + \fn qsizetype QRhiTextureRenderTargetDescription::colorAttachmentCount() const + \return the number of currently set color attachments. + */ + +/*! + \fn QRhiRenderBuffer *QRhiTextureRenderTargetDescription::depthStencilBuffer() const + \return the renderbuffer used as depth-stencil buffer, or \nullptr if none was set. + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) + + Sets the \a renderBuffer for depth-stencil. Not mandatory, e.g. when no + depth test/write or stencil-related features are used within any graphics + pipelines in any of the render passes for this render target, it can be + left set to \nullptr. + + \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be + non-null at the same time). + + Using a QRhiRenderBuffer over a 2D QRhiTexture as the depth or + depth/stencil buffer is very common, and is the recommended approach for + applications. Using a QRhiTexture, and so setDepthTexture() becomes + relevant if the depth data is meant to be accessed (e.g. sampled in a + shader) afterwards, or when + \l{QRhiColorAttachment::setMultiViewCount()}{multiview rendering} is + involved (because then the depth texture must be a texture array). + + \sa setDepthTexture() + */ + +/*! + \fn QRhiTexture *QRhiTextureRenderTargetDescription::depthTexture() const + \return the currently referenced depth texture, or \nullptr if none was set. + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthTexture(QRhiTexture *texture) + + Sets the \a texture for depth-stencil. This is an alternative to + setDepthStencilBuffer(), where instead of a QRhiRenderBuffer a QRhiTexture + with a suitable type (e.g., QRhiTexture::D32F) is provided. + + \note depthStencilBuffer() and depthTexture() cannot be both set (cannot be + non-null at the same time). + + \a texture can either be a 2D texture or a 2D texture array (when texture + arrays are supported). Specifying a texture array is relevant in particular + with + \l{QRhiColorAttachment::setMultiViewCount()}{multiview rendering}. + + \note If \a texture is a format with a stencil component, such as + \l QRhiTexture::D24S8, it will serve as the stencil buffer as well. + + \sa setDepthStencilBuffer() + */ + +/*! + \fn QRhiTexture *QRhiTextureRenderTargetDescription::depthResolveTexture() const + + \return the texture to which a multisample depth (or depth-stencil) texture + (or texture array) is resolved to. \nullptr if there is none, which is the + most common case. + + \since 6.8 + \sa QRhiColorAttachment::resolveTexture(), depthTexture() + */ + +/*! + \fn void QRhiTextureRenderTargetDescription::setDepthResolveTexture(QRhiTexture *tex) + + Sets the depth (or depth-stencil) resolve texture \a tex. + + \a tex is expected to be a 2D texture or a 2D texture array with a format + matching the texture set via setDepthTexture(). + + \note Resolving depth (or depth-stencil) data is only functional when the + \l ResolveDepthStencil feature is reported as supported at run time. Support + for depth-stencil resolve is not universally available among the graphics + APIs. Designs assuming unconditional availability of depth-stencil resolve + are therefore non-portable, and should be avoided. + + \note As an additional limitation for OpenGL ES in particular, setting a + depth resolve texture may only be functional in combination with + setDepthTexture(), not with setDepthStencilBuffer(). + + \since 6.8 + \sa QRhiColorAttachment::setResolveTexture(), setDepthTexture() + */ + +/*! \class QRhiTextureSubresourceUploadDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes the source for one mip level in a layer in a texture upload operation. The source content is specified either as a QImage or as a raw blob. The @@ -1776,10 +2799,15 @@ QRhiTextureRenderTargetDescription::QRhiTextureRenderTargetDescription(const QRh in order to be portable across all backends) If this cannot be ensured, the caller is strongly encouraged to call QImage::detach() on the image before passing it to uploadTexture(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiTextureUploadDescription */ /*! - \fn QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription() + \fn QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription() = default Constructs an empty subresource description. @@ -1827,12 +2855,109 @@ QRhiTextureSubresourceUploadDescription::QRhiTextureSubresourceUploadDescription } /*! + \fn QImage QRhiTextureSubresourceUploadDescription::image() const + \return the currently set QImage. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setImage(const QImage &image) + + Sets \a image. + Upon textures loading, the image data will be read as is, with no formats conversions. + + \note image() and data() cannot be both set at the same time. + */ + +/*! + \fn QByteArray QRhiTextureSubresourceUploadDescription::data() const + \return the currently set raw pixel data. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setData(const QByteArray &data) + + Sets \a data. + + \note image() and data() cannot be both set at the same time. + */ + +/*! + \fn quint32 QRhiTextureSubresourceUploadDescription::dataStride() const + \return the currently set data stride. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setDataStride(quint32 stride) + + Sets the data \a stride in bytes. By default this is 0 and not always + relevant. When providing raw data(), and the stride is not specified via + setDataStride(), the stride (row pitch, row length in bytes) of the + provided data must be equal to \c{width * pixelSize} where \c pixelSize is + the number of bytes used for one pixel, and there must be no additional + padding between rows. Otherwise, if there is additional space between the + lines, set a non-zero \a stride. All this is applicable only when raw image + data is provided, and is not necessary when working QImage since that has + its own \l{QImage::bytesPerLine()}{stride} value. + + \note Setting the stride via setDataStride() is only functional when + QRhi::ImageDataStride is reported as + \l{QRhi::isFeatureSupported()}{supported}. + + \note When a QImage is given, the stride returned from + QImage::bytesPerLine() is taken into account automatically and therefore + there is no need to set the data stride manually. + */ + +/*! + \fn QPoint QRhiTextureSubresourceUploadDescription::destinationTopLeft() const + \return the currently set destination top-left position. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setDestinationTopLeft(const QPoint &p) + Sets the destination top-left position \a p. + */ + +/*! + \fn QSize QRhiTextureSubresourceUploadDescription::sourceSize() const + + \return the source size in pixels. Defaults to a default-constructed QSize, + which indicates the entire subresource. + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setSourceSize(const QSize &size) + + Sets the source \a size in pixels. + + \note Setting sourceSize() or sourceTopLeft() may trigger a QImage copy + internally, depending on the format and the backend. + */ + +/*! + \fn QPoint QRhiTextureSubresourceUploadDescription::sourceTopLeft() const + \return the currently set source top-left position. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureSubresourceUploadDescription::setSourceTopLeft(const QPoint &p) + + Sets the source top-left position \a p. + + \note Setting sourceSize() or sourceTopLeft() may trigger a QImage copy + internally, depending on the format and the backend. + */ + +/*! \class QRhiTextureUploadEntry - \internal \inmodule QtGui + \since 6.6 \brief Describes one layer (face for cubemaps, slice for 3D textures, element for texture arrays) in a texture upload operation. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1858,18 +2983,61 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, } /*! + \fn int QRhiTextureUploadEntry::layer() const + \return the currently set layer index (cubemap face, array layer). Defaults to 0. + */ + +/*! + \fn void QRhiTextureUploadEntry::setLayer(int layer) + Sets the \a layer. + */ + +/*! + \fn int QRhiTextureUploadEntry::level() const + \return the currently set mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureUploadEntry::setLevel(int level) + Sets the mip \a level. + */ + +/*! + \fn QRhiTextureSubresourceUploadDescription QRhiTextureUploadEntry::description() const + \return the currently set subresource description. + */ + +/*! + \fn void QRhiTextureUploadEntry::setDescription(const QRhiTextureSubresourceUploadDescription &desc) + Sets the subresource description \a desc. + */ + +/*! \class QRhiTextureUploadDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a texture upload operation. Used with QRhiResourceUpdateBatch::uploadTexture(). That function has two variants: one taking a QImage and one taking a QRhiTextureUploadDescription. The former is a convenience version, internally creating a QRhiTextureUploadDescription with a single image - targeting level 0 for layer 0. However, when cubemaps, pre-generated mip - images, or compressed textures are involved, applications will have to work - directly with this class instead. + targeting level 0 for layer 0. + + An example of the the common, simple case of wanting to upload the contents + of a QImage to a QRhiTexture with a matching pixel size: + + \code + QImage image(256, 256, QImage::Format_RGBA8888); + image.fill(Qt::green); // or could use a QPainter targeting image + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(256, 256)); + texture->create(); + QRhiResourceUpdateBatch *u = rhi->nextResourceUpdateBatch(); + u->uploadTexture(texture, image); + \endcode + + When cubemaps, pre-generated mip images, compressed textures, or partial + uploads are involved, applications will have to use this class instead. QRhiTextureUploadDescription also enables specifying batched uploads, which are useful for example when generating an atlas or glyph cache texture: @@ -1885,10 +3053,10 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, For example, specifying the faces of a cubemap could look like the following: - \badcode + \code QImage faces[6]; - ... - QList<QRhiTextureUploadEntry> entries; + // ... + QVarLengthArray<QRhiTextureUploadEntry, 6> entries; for (int i = 0; i < 6; ++i) entries.append(QRhiTextureUploadEntry(i, 0, faces[i])); QRhiTextureUploadDescription desc; @@ -1898,7 +3066,7 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, Another example that specifies mip images for a compressed texture: - \badcode + \code QList<QRhiTextureUploadEntry> entries; const int mipCount = rhi->mipLevelsForSize(compressedTexture->pixelSize()); for (int level = 0; level < mipCount; ++level) { @@ -1913,7 +3081,7 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, With partial uploads targeting the same subresource, it is recommended to batch them into a single upload request, whenever possible: - \badcode + \code QRhiTextureSubresourceUploadDescription subresDesc(image); subresDesc.setSourceSize(QSize(10, 10)); subResDesc.setDestinationTopLeft(QPoint(50, 40)); @@ -1927,6 +3095,11 @@ QRhiTextureUploadEntry::QRhiTextureUploadEntry(int layer, int level, QRhiTextureUploadDescription desc({ entry, entry2}); resourceUpdates->uploadTexture(texture, desc); \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiResourceUpdateBatch */ /*! @@ -1959,9 +3132,39 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list } /*! + \fn void QRhiTextureUploadDescription::setEntries(std::initializer_list<QRhiTextureUploadEntry> list) + Sets the \a list of entries. + */ + +/*! + \fn template<typename InputIterator> void QRhiTextureUploadDescription::setEntries(InputIterator first, InputIterator last) + Sets the list of entries using the iterators \a first and \a last. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::cbeginEntries() const + \return a const iterator pointing to the first item in the entry list. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::cendEntries() const + \return a const iterator pointing just after the last item in the entry list. + */ + +/*! + \fn const QRhiTextureUploadEntry *QRhiTextureUploadDescription::entryAt(qsizetype index) const + \return the entry at \a index. + */ + +/*! + \fn qsizetype QRhiTextureUploadDescription::entryCount() const + \return the number of entries. + */ + +/*! \class QRhiTextureCopyDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a texture-to-texture copy operation. An empty pixelSize() indicates that the entire subresource is to be copied. @@ -1981,6 +3184,9 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list copied at a time. The source and destination layer and mip level indices can differ, but the size and position must be carefully controlled to avoid out of bounds copies, in which case the behavior is undefined. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -1990,9 +3196,83 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list */ /*! + \fn QSize QRhiTextureCopyDescription::pixelSize() const + \return the size of the region to copy. + + \note An empty pixelSize() indicates that the entire subresource is to be + copied. A default constructed copy description therefore leads to copying + the entire subresource at level 0 of layer 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setPixelSize(const QSize &sz) + Sets the size of the region to copy to \a sz. + */ + +/*! + \fn int QRhiTextureCopyDescription::sourceLayer() const + \return the source array layer (cubemap face or array layer index). Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceLayer(int layer) + Sets the source array \a layer. + */ + +/*! + \fn int QRhiTextureCopyDescription::sourceLevel() const + \return the source mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceLevel(int level) + Sets the source mip \a level. + */ + +/*! + \fn QPoint QRhiTextureCopyDescription::sourceTopLeft() const + \return the source top-left position (in pixels). Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureCopyDescription::setSourceTopLeft(const QPoint &p) + Sets the source top-left position to \a p. + */ + +/*! + \fn int QRhiTextureCopyDescription::destinationLayer() const + \return the destination array layer (cubemap face or array layer index). Default to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationLayer(int layer) + Sets the destination array \a layer. + */ + +/*! + \fn int QRhiTextureCopyDescription::destinationLevel() const + \return the destionation mip level. Defaults to 0. + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationLevel(int level) + Sets the destination mip \a level. + */ + +/*! + \fn QPoint QRhiTextureCopyDescription::destinationTopLeft() const + \return the destionation top-left position in pixels. Defaults to (0, 0). + */ + +/*! + \fn void QRhiTextureCopyDescription::setDestinationTopLeft(const QPoint &p) + Sets the destination top-left position \a p. + */ + +/*! \class QRhiReadbackDescription - \internal \inmodule QtGui + \since 6.6 \brief Describes a readback (reading back texture contents from possibly GPU-only memory) operation. The source of the readback operation is either a QRhiTexture or the @@ -2010,10 +3290,13 @@ QRhiTextureUploadDescription::QRhiTextureUploadDescription(std::initializer_list \note Multisample textures cannot be read back. Readbacks are supported for multisample swapchain buffers however. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \fn QRhiReadbackDescription::QRhiReadbackDescription() + \fn QRhiReadbackDescription::QRhiReadbackDescription() = default Constructs an empty texture readback description. @@ -2037,32 +3320,140 @@ QRhiReadbackDescription::QRhiReadbackDescription(QRhiTexture *texture) } /*! + \fn QRhiTexture *QRhiReadbackDescription::texture() const + + \return the QRhiTexture that is read back. Can be left set to \nullptr + which indicates that the backbuffer of the current swapchain is to be used + instead. + */ + +/*! + \fn void QRhiReadbackDescription::setTexture(QRhiTexture *tex) + + Sets the texture \a tex as the source of the readback operation. + + Setting \nullptr is valid too, in which case the current swapchain's + current backbuffer is used. (but then the readback cannot be issued in a + non-swapchain-based frame) + + \note Multisample textures cannot be read back. Readbacks are supported for + multisample swapchain buffers however. + + \note Textures used in readbacks must be created with + QRhiTexture::UsedAsTransferSource. + + \note Swapchains used in readbacks must be created with + QRhiSwapChain::UsedAsTransferSource. + */ + +/*! + \fn int QRhiReadbackDescription::layer() const + + \return the currently set array layer (cubemap face, array index). Defaults to 0. + + Applicable only when the source of the readback is a QRhiTexture. + */ + +/*! + \fn void QRhiReadbackDescription::setLayer(int layer) + Sets the array \a layer to read back. + */ + +/*! + \fn int QRhiReadbackDescription::level() const + + \return the currently set mip level. Defaults to 0. + + Applicable only when the source of the readback is a QRhiTexture. + */ + +/*! + \fn void QRhiReadbackDescription::setLevel(int level) + Sets the mip \a level to read back. + */ + +/*! \class QRhiReadbackResult - \internal \inmodule QtGui - \brief Describes the results of a potentially asynchronous readback operation. + \since 6.6 + \brief Describes the results of a potentially asynchronous buffer or texture readback operation. When \l completed is set, the function is invoked when the \l data is available. \l format and \l pixelSize are set upon completion together with \l data. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiReadbackResult::completed + + Callback that is invoked upon completion, on the thread the QRhi operates + on. Can be left set to \nullptr, in which case no callback is invoked. */ /*! + \variable QRhiReadbackResult::format + + Valid only for textures, the texture format. + */ + +/*! + \variable QRhiReadbackResult::pixelSize + + Valid only for textures, the size in pixels. + */ + +/*! + \variable QRhiReadbackResult::data + + The buffer or image data. + + \sa QRhiResourceUpdateBatch::readBackTexture(), QRhiResourceUpdateBatch::readBackBuffer() + */ + + +/*! \class QRhiNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Base class for classes exposing backend-specific collections of native resource objects. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! \class QRhiResource - \internal \inmodule QtGui + \since 6.6 \brief Base class for classes encapsulating native resource objects. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \enum QRhiResource::Type + Specifies type of the resource. + + \value Buffer + \value Texture + \value Sampler + \value RenderBuffer + \value RenderPassDescriptor + \value SwapChainRenderTarget + \value TextureRenderTarget + \value ShaderResourceBindings + \value GraphicsPipeline + \value SwapChain + \value ComputePipeline + \value CommandBuffer */ /*! - \fn QRhiResource::Type QRhiResource::resourceType() const + \fn virtual QRhiResource::Type QRhiResource::resourceType() const = 0 \return the type of the resource. */ @@ -2094,7 +3485,7 @@ QRhiResource::~QRhiResource() } /*! - \fn void QRhiResource::destroy() + \fn virtual void QRhiResource::destroy() = 0 Releases (or requests deferred releasing of) the underlying native graphics resources. Safe to call multiple times, subsequent invocations will be a @@ -2108,7 +3499,7 @@ QRhiResource::~QRhiResource() released until the frame is submitted by QRhi::endFrame(). The QRhiResource destructor also performs the same task, so calling this - function is not necessary before destroying a QRhiResource. + function is not necessary before deleting a QRhiResource. \sa deleteLater() */ @@ -2121,11 +3512,42 @@ QRhiResource::~QRhiResource() requirement of not altering QRhiResource objects that are referenced by the frame being recorded. + If the QRhi that created this object is already destroyed, the object is + deleted immediately. + + Using deleteLater() can be a useful convenience in many cases, and it + complements the low-level guarantee (that the underlying native graphics + objects are never destroyed until it is safe to do so and it is known for + sure that they are not used by the GPU in an still in-flight frame), by + offering a way to make sure the C++ object instances (of QRhiBuffer, + QRhiTexture, etc.) themselves also stay valid until the end of the current + frame. + + The following example shows a convenient way of creating a throwaway buffer + that is only used in one frame and gets automatically released in + endFrame(). (when it comes to the underlying native buffer(s), the usual + guarantee applies: the QRhi backend defers the releasing of those until it + is guaranteed that the frame in which the buffer is accessed by the GPU has + completed) + + \code + rhi->beginFrame(swapchain); + QRhiBuffer *buf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, 256); + buf->deleteLater(); // ! + u = rhi->nextResourceUpdateBatch(); + u->uploadStaticBuffer(buf, data); + // ... draw with buf + rhi->endFrame(); + \endcode + \sa destroy() */ void QRhiResource::deleteLater() { - m_rhi->addDeleteLater(this); + if (m_rhi) + m_rhi->addDeleteLater(this); + else + delete this; } /*! @@ -2139,7 +3561,7 @@ QByteArray QRhiResource::name() const /*! Sets a \a name for the object. - This has two uses: to get descriptive names for the native graphics + This allows getting descriptive names for the native graphics resources visible in graphics debugging tools, such as \l{https://renderdoc.org/}{RenderDoc} and \l{https://developer.apple.com/xcode/}{XCode}. @@ -2174,17 +3596,140 @@ quint64 QRhiResource::globalResourceId() const /*! \return the QRhi that created this resource. + + If the QRhi that created this object is already destroyed, the result is + \nullptr. */ QRhi *QRhiResource::rhi() const { - return m_rhi->q; + return m_rhi ? m_rhi->q : nullptr; } /*! \class QRhiBuffer - \internal \inmodule QtGui + \since 6.6 \brief Vertex, index, or uniform (constant) buffer resource. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + A QRhiBuffer encapsulates zero, one, or more native buffer objects (such as + a \c VkBuffer or \c MTLBuffer). With some graphics APIs and backends + certain types of buffers may not use a native buffer object at all (e.g. + OpenGL if uniform buffer objects are not used), but this is transparent to + the user of the QRhiBuffer API. Similarly, the fact that some types of + buffers may use two or three native buffers underneath, in order to allow + efficient per-frame content update without stalling the GPU pipeline, is + mostly invisible to the applications and libraries. + + A QRhiBuffer instance is always created by calling + \l{QRhi::newBuffer()}{the QRhi's newBuffer() function}. This creates no + native graphics resources. To do that, call create() after setting the + appropriate options, such as the type, usage flags, size, although in most cases these + are already set based on the arguments passed to + \l{QRhi::newBuffer()}{newBuffer()}. + + \section2 Example usage + + To create a uniform buffer for a shader where the GLSL uniform block + contains a single \c mat4 member, and update the contents: + + \code + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, 64); + if (!ubuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + QMatrix4x4 mvp; + // ... set up the modelview-projection matrix + batch->updateDynamicBuffer(ubuf, 0, 64, mvp.constData()); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + An example of creating a buffer with vertex data: + + \code + const float vertices[] = { -1.0f, -1.0f, 1.0f, -1.0f, 0.0f, 1.0f }; + QRhiBuffer *vbuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::VertexBuffer, sizeof(vertices)); + if (!vbuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + batch->uploadStaticBuffer(vbuf, vertices); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + An index buffer: + + \code + static const quint16 indices[] = { 0, 1, 2 }; + QRhiBuffer *ibuf = rhi->newBuffer(QRhiBuffer::Immutable, QRhiBuffer::IndexBuffer, sizeof(indices)); + if (!ibuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + batch->uploadStaticBuffer(ibuf, indices); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + \section2 Common patterns + + A call to create() destroys any existing native resources if create() was + successfully called before. If those native resources are still in use by + an in-flight frame (i.e., there's a chance they are still read by the GPU), + the destroying of those resources is deferred automatically. Thus a very + common and convenient pattern to safely increase the size of an already + initialized buffer is the following. In practice this drops and creates a + whole new set of native resources underneath, so it is not necessarily a + cheap operation, but is more convenient and still faster than the + alternatives, because by not destroying the \c buf object itself, all + references to it stay valid in other data structures (e.g., in any + QRhiShaderResourceBinding the QRhiBuffer is referenced from). + + \code + if (buf->size() < newSize) { + buf->setSize(newSize); + if (!buf->create()) { error(); } + } + // continue using buf, fill it with new data + \endcode + + When working with uniform buffers, it will sometimes be necessary to + combine data for multiple draw calls into a single buffer for efficiency + reasons. Be aware of the aligment requirements: with some graphics APIs + offsets for a uniform buffer must be aligned to 256 bytes. This applies + both to QRhiShaderResourceBinding and to the dynamic offsets passed to + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()}. Use the + \l{QRhi::ubufAlignment()}{ubufAlignment()} and + \l{QRhi::ubufAligned()}{ubufAligned()} functions to create portable code. + As an example, the following is an outline for issuing multiple (\c N) draw + calls with the same pipeline and geometry, but with a different data in the + uniform buffers exposed at binding point 0. This assumes the buffer is + exposed via + \l{QRhiShaderResourceBinding::uniformBufferWithDynamicOffset()}{uniformBufferWithDynamicOffset()} + which allows passing a QRhiCommandBuffer::DynamicOffset list to + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()}. + + \code + const int N = 2; + const int UB_SIZE = 64 + 4; // assuming a uniform block with { mat4 matrix; float opacity; } + const int ONE_UBUF_SIZE = rhi->ubufAligned(UB_SIZE); + const int TOTAL_UBUF_SIZE = N * ONE_UBUF_SIZE; + QRhiBuffer *ubuf = rhi->newBuffer(QRhiBuffer::Dynamic, QRhiBuffer::UniformBuffer, TOTAL_UBUF_SIZE); + if (!ubuf->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + for (int i = 0; i < N; ++i) { + batch->updateDynamicBuffer(ubuf, i * ONE_UBUF_SIZE, 64, matrix.constData()); + batch->updateDynamicBuffer(ubuf, i * ONE_UBUF_SIZE + 64, 4, &opacity); + } + // ... + // beginPass(), set pipeline, etc., and then: + for (int i = 0; i < N; ++i) { + QRhiCommandBuffer::DynamicOffset dynOfs[] = { { 0, i * ONE_UBUF_SIZE } }; + cb->setShaderResources(srb, 1, dynOfs); + cb->draw(36); + } + \endcode + + \sa QRhiResourceUpdateBatch, QRhi, QRhiCommandBuffer */ /*! @@ -2220,14 +3765,14 @@ QRhi *QRhiResource::rhi() const Flag values to specify how the buffer is going to be used. \value VertexBuffer Vertex buffer. This allows the QRhiBuffer to be used in - \l{setVertexInput()}{QRhiCommandBuffer::setVertexInput()}. + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()}. \value IndexBuffer Index buffer. This allows the QRhiBuffer to be used in - \l{setVertexInput()}{QRhiCommandBuffer::setVertexInput()}. + \l{QRhiCommandBuffer::setVertexInput()}{setVertexInput()}. \value UniformBuffer Uniform buffer (also called constant buffer). This allows the QRhiBuffer to be used in combination with - \l{UniformBuffer}{QRhiShaderResourceBinding::UniformBuffer}. When + \l{QRhiShaderResourceBinding::UniformBuffer}{UniformBuffer}. When \l{QRhi::NonDynamicUniformBuffers}{NonDynamicUniformBuffers} is reported as not supported, this usage can only be combined with the type Dynamic. @@ -2241,21 +3786,8 @@ QRhi *QRhiResource::rhi() const */ /*! - \fn void QRhiBuffer::setSize(int sz) - - Sets the size of the buffer in bytes. The size is normally specified in - QRhi::newBuffer() so this function is only used when the size has to be - changed. As with other setters, the size only takes effect when calling - create(), and for already created buffers this involves releasing the previous - native resource and creating new ones under the hood. - - Backends may choose to allocate buffers bigger than \a sz in order to - fulfill alignment requirements. This is hidden from the applications and - size() will always report the size requested in \a sz. - */ - -/*! \class QRhiBuffer::NativeBuffer + \inmodule QtGui \brief Contains information about the underlying native resources of a buffer. */ @@ -2267,10 +3799,13 @@ QRhi *QRhiResource::rhi() const objects array are pointers to a GLuint. With Vulkan, the native handle is a VkBuffer, so the elements of the array are pointers to a VkBuffer. With Direct3D 11 and Metal the elements are pointers to a ID3D11Buffer or - MTLBuffer pointer, respectively. + MTLBuffer pointer, respectively. With Direct3D 12, the elements are + pointers to a ID3D12Resource. \note Pay attention to the fact that the elements are always pointers to the native buffer handle type, even if the native type itself is a pointer. + (so the elements are \c{VkBuffer *} on Vulkan, even though VkBuffer itself + is a pointer on 64-bit architectures). */ /*! @@ -2306,7 +3841,7 @@ QRhiResource::Type QRhiBuffer::resourceType() const } /*! - \fn bool QRhiBuffer::create() + \fn virtual bool QRhiBuffer::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2317,6 +3852,50 @@ QRhiResource::Type QRhiBuffer::resourceType() const */ /*! + \fn QRhiBuffer::Type QRhiBuffer::type() const + \return the buffer type. + */ + +/*! + \fn void QRhiBuffer::setType(Type t) + Sets the buffer's type to \a t. + */ + +/*! + \fn QRhiBuffer::UsageFlags QRhiBuffer::usage() const + \return the buffer's usage flags. + */ + +/*! + \fn void QRhiBuffer::setUsage(UsageFlags u) + Sets the buffer's usage flags to \a u. + */ + +/*! + \fn quint32 QRhiBuffer::size() const + + \return the buffer's size in bytes. + + This is always the value that was passed to setSize() or QRhi::newBuffer(). + Internally, the native buffers may be bigger if that is required by the + underlying graphics API. + */ + +/*! + \fn void QRhiBuffer::setSize(quint32 sz) + + Sets the size of the buffer in bytes. The size is normally specified in + QRhi::newBuffer() so this function is only used when the size has to be + changed. As with other setters, the size only takes effect when calling + create(), and for already created buffers this involves releasing the previous + native resource and creating new ones under the hood. + + Backends may choose to allocate buffers bigger than \a sz in order to + fulfill alignment requirements. This is hidden from the applications and + size() will always report the size requested in \a sz. + */ + +/*! \return the underlying native resources for this buffer. The returned value will be empty if exposing the underlying native resources is not supported by the backend. @@ -2404,18 +3983,18 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() /*! \class QRhiRenderBuffer - \internal \inmodule QtGui + \since 6.6 \brief Renderbuffer resource. Renderbuffers cannot be sampled or read but have some benefits over textures in some cases: - A DepthStencil renderbuffer may be lazily allocated and be backed by + A \l DepthStencil renderbuffer may be lazily allocated and be backed by transient memory with some APIs. On some platforms this may mean the depth/stencil buffer uses no physical backing at all. - Color renderbuffers are useful since QRhi::MultisampleRenderBuffer may be + \l Color renderbuffers are useful since QRhi::MultisampleRenderBuffer may be supported even when QRhi::MultisampleTexture is not. How the renderbuffer is implemented by a backend is not exposed to the @@ -2429,6 +4008,9 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() QRhi provides automatic sizing behavior to match the color buffers, which means calling setPixelSize() and create() are not necessary for such renderbuffers. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2440,6 +4022,22 @@ void QRhiBuffer::endFullDynamicBufferUpdateForCurrentFrame() */ /*! + \struct QRhiRenderBuffer::NativeRenderBuffer + \inmodule QtGui + \brief Wraps a native renderbuffer object. + */ + +/*! + \variable QRhiRenderBuffer::NativeRenderBuffer::object + \brief 64-bit integer containing the native object handle. + + Used with QRhiRenderBuffer::createFrom(). + + With OpenGL the native handle is a GLuint value. \c object is expected to + be a valid OpenGL renderbuffer object ID. + */ + +/*! \enum QRhiRenderBuffer::Flag Flag values for flags() and setFlags() @@ -2476,7 +4074,7 @@ QRhiResource::Type QRhiRenderBuffer::resourceType() const } /*! - \fn bool QRhiRenderBuffer::create() + \fn virtual bool QRhiRenderBuffer::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2525,16 +4123,121 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) } /*! - \fn QRhiTexture::Format QRhiRenderBuffer::backingFormat() const + \fn QRhiRenderBuffer::Type QRhiRenderBuffer::type() const + \return the renderbuffer type. + */ + +/*! + \fn void QRhiRenderBuffer::setType(Type t) + Sets the type to \a t. + */ + +/*! + \fn QSize QRhiRenderBuffer::pixelSize() const + \return the pixel size. + */ + +/*! + \fn void QRhiRenderBuffer::setPixelSize(const QSize &sz) + Sets the size (in pixels) to \a sz. + */ + +/*! + \fn int QRhiRenderBuffer::sampleCount() const + \return the sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiRenderBuffer::setSampleCount(int s) + Sets the sample count to \a s. + */ + +/*! + \fn QRhiRenderBuffer::Flags QRhiRenderBuffer::flags() const + \return the flags. + */ + +/*! + \fn void QRhiRenderBuffer::setFlags(Flags f) + Sets the flags to \a f. + */ + +/*! + \fn virtual QRhiTexture::Format QRhiRenderBuffer::backingFormat() const = 0 \internal */ /*! \class QRhiTexture - \internal \inmodule QtGui + \since 6.6 \brief Texture resource. + + A QRhiTexture encapsulates a native texture object, such as a \c VkImage or + \c MTLTexture. + + A QRhiTexture instance is always created by calling + \l{QRhi::newTexture()}{the QRhi's newTexture() function}. This creates no + native graphics resources. To do that, call create() after setting the + appropriate options, such as the format and size, although in most cases + these are already set based on the arguments passed to + \l{QRhi::newTexture()}{newTexture()}. + + Setting the \l{QRhiTexture::Flags}{flags} correctly is essential, otherwise + various errors can occur depending on the underlying QRhi backend and + graphics API. For example, when a texture will be rendered into from a + render pass via QRhiTextureRenderTarget, the texture must be created with + the \l RenderTarget flag set. Similarly, when the texture is going to be + \l{QRhiResourceUpdateBatch::readBackTexture()}{read back}, the \l + UsedAsTransferSource flag must be set upfront. Mipmapped textures must have + the MipMapped flag set. And so on. It is not possible to change the flags + once create() has succeeded. To release the existing and create a new + native texture object with the changed settings, call the setters and call + create() again. This then might be a potentially expensive operation. + + \section2 Example usage + + To create a 2D texture with a size of 512x512 pixels and set its contents to all green: + + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, QSize(512, 512)); + if (!texture->create()) { error(); } + QRhiResourceUpdateBatch *batch = rhi->nextResourceUpdateBatch(); + QImage image(512, 512, QImage::Format_RGBA8888); + image.fill(Qt::green); + batch->uploadTexture(texture, image); + // ... + commandBuffer->resourceUpdate(batch); // or, alternatively, pass 'batch' to a beginPass() call + \endcode + + \section2 Common patterns + + A call to create() destroys any existing native resources if create() was + successfully called before. If those native resources are still in use by + an in-flight frame (i.e., there's a chance they are still read by the GPU), + the destroying of those resources is deferred automatically. Thus a very + common and convenient pattern to safely change the size of an already + existing texture is the following. In practice this drops and creates a + whole new native texture resource underneath, so it is not necessarily a + cheap operation, but is more convenient and still faster than the + alternatives, because by not destroying the \c texture object itself, all + references to it stay valid in other data structures (e.g., in any + QShaderResourceBinding the QRhiTexture is referenced from). + + \code + // determine newSize, e.g. based on the swapchain's output size or other factors + if (texture->pixelSize() != newSize) { + texture->setPixelSize(newSize); + if (!texture->create()) { error(); } + } + // continue using texture, fill it with new data + \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiResourceUpdateBatch, QRhi, QRhiTextureRenderTarget */ /*! @@ -2688,7 +4391,8 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) */ /*! - \class QRhiTexture::NativeTexture + \struct QRhiTexture::NativeTexture + \inmodule QtGui \brief Contains information about the underlying native resources of a texture. */ @@ -2697,9 +4401,10 @@ bool QRhiRenderBuffer::createFrom(NativeRenderBuffer src) \brief 64-bit integer containing the native object handle. With OpenGL, the native handle is a GLuint value, so \c object can then be - cast to a GLuint. With Vulkan, the native handle is a VkImage, so \c - object can be cast to a VkImage. With Direct3D 11 and Metal \c - object contains a ID3D11Texture2D or MTLTexture pointer, respectively. + cast to a GLuint. With Vulkan, the native handle is a VkImage, so \c object + can be cast to a VkImage. With Direct3D 11 and Metal \c object contains a + ID3D11Texture2D or MTLTexture pointer, respectively. With Direct3D 12 + \c object contains a ID3D12Resource pointer. */ /*! @@ -2729,7 +4434,7 @@ QRhiResource::Type QRhiTexture::resourceType() const } /*! - \fn bool QRhiTexture::create() + \fn virtual bool QRhiTexture::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -2752,13 +4457,16 @@ QRhiTexture::NativeTexture QRhiTexture::nativeTexture() } /*! - Similar to create() except that no new native textures are created. Instead, - the native texture resources specified by \a src is used. + Similar to create(), except that no new native textures are created. + Instead, the native texture resources specified by \a src is used. This allows importing an existing native texture object (which must belong to the same device or sharing context, depending on the graphics API) from an external graphics engine. + \return true if the specified existing native texture object has been + successfully wrapped as a non-owning QRhiTexture. + \note format(), pixelSize(), sampleCount(), and flags() must still be set correctly. Passing incorrect sizes and other values to QRhi::newTexture() and then following it with a createFrom() expecting that the native texture @@ -2811,10 +4519,196 @@ void QRhiTexture::setNativeLayout(int layout) } /*! + \fn QRhiTexture::Format QRhiTexture::format() const + \return the texture format. + */ + +/*! + \fn void QRhiTexture::setFormat(QRhiTexture::Format fmt) + + Sets the requested texture format to \a fmt. + + \note The value set is only taken into account upon the next call to + create(), i.e. when the underlying graphics resource are (re)created. + Setting a new value is futile otherwise and must be avoided since it can + lead to inconsistent state. + */ + +/*! + \fn QSize QRhiTexture::pixelSize() const + \return the size in pixels. + */ + +/*! + \fn void QRhiTexture::setPixelSize(const QSize &sz) + + Sets the texture size, specified in pixels, to \a sz. + + \note The value set is only taken into account upon the next call to + create(), i.e. when the underlying graphics resource are (re)created. + Setting a new value is futile otherwise and must be avoided since it can + lead to inconsistent state. The same applies to all other setters as well. + */ + +/*! + \fn int QRhiTexture::depth() const + \return the depth for 3D textures. + */ + +/*! + \fn void QRhiTexture::setDepth(int depth) + Sets the \a depth for a 3D texture. + */ + +/*! + \fn int QRhiTexture::arraySize() const + \return the texture array size. + */ + +/*! + \fn void QRhiTexture::setArraySize(int arraySize) + Sets the texture \a arraySize. + */ + +/*! + \fn int QRhiTexture::arrayRangeStart() const + + \return the first array layer when setArrayRange() was called. + + \sa setArrayRange() + */ + +/*! + \fn int QRhiTexture::arrayRangeLength() const + + \return the exposed array range size when setArrayRange() was called. + + \sa setArrayRange() +*/ + +/*! + \fn void QRhiTexture::setArrayRange(int startIndex, int count) + + Normally all array layers are exposed and it is up to the shader to select + the layer via the third coordinate passed to the \c{texture()} GLSL + function when sampling the \c sampler2DArray. When QRhi::TextureArrayRange + is reported as supported, calling setArrayRange() before create() or + createFrom() requests selecting only the specified range, \a count elements + starting from \a startIndex. The shader logic can then be written with this + in mind. + + \sa QRhi::TextureArrayRange + */ + +/*! + \fn Flags QRhiTexture::flags() const + \return the texture flags. + */ + +/*! + \fn void QRhiTexture::setFlags(Flags f) + Sets the texture flags to \a f. + */ + +/*! + \fn int QRhiTexture::sampleCount() const + \return the sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiTexture::setSampleCount(int s) + Sets the sample count to \a s. + */ + +/*! + \struct QRhiTexture::ViewFormat + \inmodule QtGui + \since 6.8 + \brief Specifies the view format for reading or writing from or to the texture. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiTexture::ViewFormat::format + */ + +/*! + \variable QRhiTexture::ViewFormat::srgb + */ + +/*! + \fn QRhiTexture::ViewFormat QRhiTexture::readViewFormat() const + \since 6.8 + \return the view format used when sampling the texture. When not called, the view + format is assumed to be the same as format(). + */ + +/*! + \fn void QRhiTexture::setReadViewFormat(const ViewFormat &fmt) + \since 6.8 + + Sets the shader resource view format (or the format of the view used for + sampling the texture) to \a fmt. By default the same format (and sRGB-ness) + is used as the texture itself, and in most cases this function does not need + to be called. + + This setting is only taken into account when the \l TextureViewFormat + feature is reported as supported. + + \note This functionality is provided to allow "casting" between + non-sRGB and sRGB in order to get the shader reads perform, or not perform, + the implicit sRGB conversions. Other types of casting may or may not be + functional. + */ + +/*! + \fn QRhiTexture::ViewFormat QRhiTexture::writeViewFormat() const + \since 6.8 + \return the view format used when writing to the texture and when using it + with image load/store. When not called, the view format is assumed to be the + same as format(). + */ + +/*! + \fn void QRhiTexture::setWriteViewFormat(const ViewFormat &fmt) + \since 6.8 + + Sets the render target view format to \a fmt. By default the same format + (and sRGB-ness) is used as the texture itself, and in most cases this + function does not need to be called. + + One common use case for providing a write view format is working with + externally provided textures that, outside of our control, use an sRGB + format with 3D APIs such as Vulkan or Direct 3D, but the rendering engine is + already prepared to handle linearization and conversion to sRGB at the end + of its shading pipeline. In this case what is wanted when rendering into + such a texture is a render target view (e.g. VkImageView) that has the same, + but non-sRGB format. (if e.g. from an OpenXR implementation one gets a + VK_FORMAT_R8G8B8A8_SRGB texture, it is likely that rendering into it should + be done using a VK_FORMAT_R8G8B8A8_UNORM view, if that is what the rendering + engine's pipeline requires; in this example one would call this function + with a ViewFormat that has a format of QRhiTexture::RGBA8 and \c srgb set to + \c false). + + This setting is only taken into account when the \l TextureViewFormat + feature is reported as supported. + + \note This functionality is provided to allow "casting" between + non-sRGB and sRGB in order to get the shader write not perform, or perform, + the implicit sRGB conversions. Other types of casting may or may not be + functional. + */ + +/*! \class QRhiSampler - \internal \inmodule QtGui + \since 6.6 \brief Sampler resource. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2871,14 +4765,91 @@ QRhiResource::Type QRhiSampler::resourceType() const } /*! + \fn QRhiSampler::Filter QRhiSampler::magFilter() const + \return the magnification filter mode. + */ + +/*! + \fn void QRhiSampler::setMagFilter(Filter f) + Sets the magnification filter mode to \a f. + */ + +/*! + \fn QRhiSampler::Filter QRhiSampler::minFilter() const + \return the minification filter mode. + */ + +/*! + \fn void QRhiSampler::setMinFilter(Filter f) + Sets the minification filter mode to \a f. + */ + +/*! + \fn QRhiSampler::Filter QRhiSampler::mipmapMode() const + \return the mipmap filter mode. + */ + +/*! + \fn void QRhiSampler::setMipmapMode(Filter f) + + Sets the mipmap filter mode to \a f. + + Leave this set to None when the texture has no mip levels, or when the mip + levels are not to be taken into account. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressU() const + \return the horizontal wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressU(AddressMode mode) + Sets the horizontal wrap \a mode. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressV() const + \return the vertical wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressV(AddressMode mode) + Sets the vertical wrap \a mode. + */ + +/*! + \fn QRhiSampler::AddressMode QRhiSampler::addressW() const + \return the depth wrap mode. + */ + +/*! + \fn void QRhiSampler::setAddressW(AddressMode mode) + Sets the depth wrap \a mode. + */ + +/*! + \fn QRhiSampler::CompareOp QRhiSampler::textureCompareOp() const + \return the texture comparison function. + */ + +/*! + \fn void QRhiSampler::setTextureCompareOp(CompareOp op) + Sets the texture comparison function \a op. + */ + +/*! \class QRhiRenderPassDescriptor - \internal \inmodule QtGui + \since 6.6 \brief Render pass resource. A render pass, if such a concept exists in the underlying graphics API, is a collection of attachments (color, depth, stencil) and describes how those attachments are used. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -2898,7 +4869,7 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const } /*! - \fn bool QRhiRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const + \fn virtual bool QRhiRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const = 0 \return true if the \a other QRhiRenderPassDescriptor is compatible with this one, meaning \c this and \a other can be used interchangebly in @@ -2933,7 +4904,7 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const */ /*! - \fn QRhiRenderPassDescriptor *QRhiRenderPassDescriptor::newCompatibleRenderPassDescriptor() const + \fn virtual QRhiRenderPassDescriptor *QRhiRenderPassDescriptor::newCompatibleRenderPassDescriptor() const = 0 \return a new QRhiRenderPassDescriptor that is \l{isCompatible()}{compatible} with this one. @@ -2954,7 +4925,7 @@ QRhiResource::Type QRhiRenderPassDescriptor::resourceType() const */ /*! - \fn QVector<quint32> QRhiRenderPassDescriptor::serializedFormat() const + \fn virtual QVector<quint32> QRhiRenderPassDescriptor::serializedFormat() const = 0 \return a vector of integers containing an opaque blob describing the data relevant for \l{isCompatible()}{compatibility}. @@ -2985,10 +4956,20 @@ const QRhiNativeHandles *QRhiRenderPassDescriptor::nativeHandles() /*! \class QRhiRenderTarget - \internal \inmodule QtGui + \since 6.6 \brief Represents an onscreen (swapchain) or offscreen (texture) render target. + Applications do not create an instance of this class directly. Rather, it + is the subclass QRhiTextureRenderTarget that is instantiable by clients of + the API via \l{QRhi::newTextureRenderTarget()}{newTextureRenderTarget()}. + The other subclass is QRhiSwapChainRenderTarget, which is the type + QRhiSwapChain returns when calling + \l{QRhiSwapChain::currentFrameRenderTarget()}{currentFrameRenderTarget()}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiSwapChainRenderTarget, QRhiTextureRenderTarget */ @@ -3001,7 +4982,7 @@ QRhiRenderTarget::QRhiRenderTarget(QRhiImplementation *rhi) } /*! - \fn QSize QRhiRenderTarget::pixelSize() const + \fn virtual QSize QRhiRenderTarget::pixelSize() const = 0 \return the size in pixels. @@ -3019,7 +5000,7 @@ QRhiRenderTarget::QRhiRenderTarget(QRhiImplementation *rhi) */ /*! - \fn float QRhiRenderTarget::devicePixelRatio() const + \fn virtual float QRhiRenderTarget::devicePixelRatio() const = 0 \return the device pixel ratio. For QRhiTextureRenderTarget this is always 1. For targets retrieved from a QRhiSwapChain the value reflects the @@ -3028,6 +5009,25 @@ QRhiRenderTarget::QRhiRenderTarget(QRhiImplementation *rhi) */ /*! + \fn virtual int QRhiRenderTarget::sampleCount() const = 0 + + \return the sample count or 1 if multisample antialiasing is not relevant for + this render target. + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiRenderTarget::renderPassDescriptor() const + + \return the associated QRhiRenderPassDescriptor. + */ + +/*! + \fn void QRhiRenderTarget::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + + Sets the QRhiRenderPassDescriptor \a desc for use with this render target. + */ + +/*! \internal */ QRhiSwapChainRenderTarget::QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain_) @@ -3038,14 +5038,17 @@ QRhiSwapChainRenderTarget::QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QR /*! \class QRhiSwapChainRenderTarget - \internal \inmodule QtGui + \since 6.6 \brief Swapchain render target resource. When targeting the color buffers of a swapchain, active render target is a QRhiSwapChainRenderTarget. This is what QRhiSwapChain::currentFrameRenderTarget() returns. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + \sa QRhiSwapChain */ @@ -3065,8 +5068,8 @@ QRhiResource::Type QRhiSwapChainRenderTarget::resourceType() const /*! \class QRhiTextureRenderTarget - \internal \inmodule QtGui + \since 6.6 \brief Texture render target resource. A texture render target allows rendering into one or more textures, @@ -3082,15 +5085,18 @@ QRhiResource::Type QRhiSwapChainRenderTarget::resourceType() const The simplest example of creating a render target with a texture as its single color attachment: - \badcode - texture = rhi->newTexture(QRhiTexture::RGBA8, size, 1, QRhiTexture::RenderTarget); + \code + QRhiTexture *texture = rhi->newTexture(QRhiTexture::RGBA8, size, 1, QRhiTexture::RenderTarget); texture->create(); - rt = rhi->newTextureRenderTarget({ texture }); + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ texture }); rp = rt->newCompatibleRenderPassDescriptor(); rt->setRenderPassDescriptor(rt); rt->create(); // rt can now be used with beginPass() \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -3105,7 +5111,19 @@ QRhiResource::Type QRhiSwapChainRenderTarget::resourceType() const \value PreserveColorContents Indicates that the contents of the color attachments is to be loaded when starting a render pass, instead of clearing. This is potentially more expensive, especially on mobile (tiled) - GPUs, but allows preserving the existing contents between passes. + GPUs, but allows preserving the existing contents between passes. When doing + multisample rendering with a resolve texture set, setting this flag also + requests the multisample color data to be stored (written out) to the + multisample texture or render buffer. (for non-multisample rendering the + color data is always stored, but for MSAA storing the multisample data + decreases efficiency for certain GPU architectures, hence defaulting to not + writing it out) Note however that this is non-portable: in some cases there + is no intermediate multisample texture on the graphics API level, e.g. when + using OpenGL ES's \c{GL_EXT_multisampled_render_to_texture} as it is all + implicit, handled by the OpenGL ES implementation. In that case, + PreserveColorContents will likely have no effect. Therefore, avoid relying + on this flag when using multisample rendering and the color attachment is + using a multisample QRhiTexture (not QRhiRenderBuffer). \value PreserveDepthStencilContents Indicates that the contents of the depth texture is to be loaded when starting a render pass, instead @@ -3113,6 +5131,13 @@ QRhiResource::Type QRhiSwapChainRenderTarget::resourceType() const (QRhiTextureRenderTargetDescription::depthTexture() is set) because depth/stencil renderbuffers may not have any physical backing and data may not be written out in the first place. + + \value DoNotStoreDepthStencilContents Indicates that the contents of the + depth texture does not need to be written out. Relevant only when a + QRhiTexture, not QRhiRenderBuffer, is used as the depth-stencil buffer, + because for QRhiRenderBuffer this is implicit. When a depthResolveTexture is + set, the flag is not relevant, because the behavior is then as if the flag + was set. This enum value is introduced in Qt 6.8. */ /*! @@ -3136,7 +5161,7 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const } /*! - \fn QRhiRenderPassDescriptor *QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor() + \fn virtual QRhiRenderPassDescriptor *QRhiTextureRenderTarget::newCompatibleRenderPassDescriptor() = 0 \return a new QRhiRenderPassDescriptor that is compatible with this render target. @@ -3146,7 +5171,8 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const QRhiGraphicsPipeline::setRenderPassDescriptor(). A render pass descriptor describes the attachments (color, depth/stencil) and the load/store behavior that can be affected by flags(). A QRhiGraphicsPipeline can only - be used in combination with a render target that has the same + be used in combination with a render target that has a + \l{QRhiRenderPassDescriptor::isCompatible()}{compatible} QRhiRenderPassDescriptor set. Two QRhiTextureRenderTarget instances can share the same render pass @@ -3162,7 +5188,7 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const */ /*! - \fn bool QRhiTextureRenderTarget::create() + \fn virtual bool QRhiTextureRenderTarget::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -3186,9 +5212,29 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const */ /*! + \fn QRhiTextureRenderTargetDescription QRhiTextureRenderTarget::description() const + \return the render target description. + */ + +/*! + \fn void QRhiTextureRenderTarget::setDescription(const QRhiTextureRenderTargetDescription &desc) + Sets the render target description \a desc. + */ + +/*! + \fn QRhiTextureRenderTarget::Flags QRhiTextureRenderTarget::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiTextureRenderTarget::setFlags(Flags f) + Sets the flags to \a f. + */ + +/*! \class QRhiShaderResourceBindings - \internal \inmodule QtGui + \since 6.6 \brief Encapsulates resources for making buffer, texture, sampler resources visible to shaders. A QRhiShaderResourceBindings is a collection of QRhiShaderResourceBinding @@ -3209,19 +5255,19 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const QRhiShaderResourceBindings could be created and then passed to QRhiGraphicsPipeline::setShaderResourceBindings(): - \badcode - srb = rhi->newShaderResourceBindings(); + \code + QRhiShaderResourceBindings *srb = rhi->newShaderResourceBindings(); srb->setBindings({ QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage | QRhiShaderResourceBinding::FragmentStage, ubuf), QRhiShaderResourceBinding::sampledTexture(1, QRhiShaderResourceBinding::FragmentStage, texture, sampler) }); srb->create(); - ... - ps = rhi->newGraphicsPipeline(); - ... + // ... + QRhiGraphicsPipeline *ps = rhi->newGraphicsPipeline(); + // ... ps->setShaderResourceBindings(srb); ps->create(); - ... + // ... cb->setGraphicsPipeline(ps); cb->setShaderResources(); // binds srb \endcode @@ -3240,17 +5286,28 @@ QRhiResource::Type QRhiTextureRenderTarget::resourceType() const srb argument. As long as the layouts (so the number of bindings and the binding points) match between two QRhiShaderResourceBindings, they can both be used with the same pipeline, assuming the pipeline was created with one of - them in the first place. + them in the first place. See isLayoutCompatible() for more details. - \badcode - srb2 = rhi->newShaderResourceBindings(); - ... + \code + QRhiShaderResourceBindings *srb2 = rhi->newShaderResourceBindings(); + // ... cb->setGraphicsPipeline(ps); cb->setShaderResources(srb2); // binds srb2 \endcode + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \typedef QRhiShaderResourceBindingSet + \relates QRhi + \since 6.7 + + Synonym for QRhiShaderResourceBindings. +*/ + +/*! \internal */ QRhiShaderResourceBindings::QRhiShaderResourceBindings(QRhiImplementation *rhi) @@ -3313,7 +5370,7 @@ bool QRhiShaderResourceBindings::isLayoutCompatible(const QRhiShaderResourceBind \l{isLayoutCompatible()}{layout compatibility tests}. Given two objects \c srb1 and \c srb2, if the data returned from this - function is identical, then \c{srb1->isLayoutCompatible(srb2), and vice + function is identical, then \c{srb1->isLayoutCompatible(srb2)}, and vice versa hold true as well. \note The returned data is meant to be used for storing in memory and @@ -3338,14 +5395,47 @@ void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) } /*! + \fn void QRhiShaderResourceBindings::setBindings(std::initializer_list<QRhiShaderResourceBinding> list) + Sets the \a list of bindings. + */ + +/*! + \fn template<typename InputIterator> void QRhiShaderResourceBindings::setBindings(InputIterator first, InputIterator last) + Sets the list of bindings from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::cbeginBindings() const + \return a const iterator pointing to the first item in the binding list. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::cendBindings() const + \return a const iterator pointing just after the last item in the binding list. + */ + +/*! + \fn const QRhiShaderResourceBinding *QRhiShaderResourceBindings::bindingAt(qsizetype index) const + \return the binding at the specified \a index. + */ + +/*! + \fn qsizetype QRhiShaderResourceBindings::bindingCount() const + \return the number of bindings. + */ + +/*! \class QRhiShaderResourceBinding - \internal \inmodule QtGui + \since 6.6 \brief Describes the shader resource for a single binding point. A QRhiShaderResourceBinding cannot be constructed directly. Instead, use the static functions such as uniformBuffer() or sampledTexture() to get an instance. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -3402,7 +5492,7 @@ void QRhiImplementation::updateLayoutDesc(QRhiShaderResourceBindings *srb) For example, \c a and \c b below are not equal, but are compatible layout-wise: - \badcode + \code auto a = QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage, buffer); auto b = QRhiShaderResourceBinding::uniformBuffer(0, QRhiShaderResourceBinding::VertexStage, someOtherBuffer, 256); \endcode @@ -4276,18 +6366,44 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) /*! \class QRhiGraphicsPipeline - \internal \inmodule QtGui + \since 6.6 \brief Graphics pipeline state resource. + Represents a graphics pipeline. What exactly this map to in the underlying + native graphics API, varies. Where there is a concept of pipeline objects, + for example with Vulkan, the QRhi backend will create such an object upon + calling create(). Elsewhere, for example with OpenGL, the + QRhiGraphicsPipeline may merely collect the various state, and create()'s + main task is to set up the corresponding shader program, but deferring + looking at any of the requested state to a later point. + + As with all QRhiResource subclasses, the two-phased initialization pattern + applies: setting any values via the setters, for example setDepthTest(), is + only effective after calling create(). Avoid changing any values once the + QRhiGraphicsPipeline has been initialized via create(). To change some + state, set the new value and call create() again. However, that will + effectively release all underlying native resources and create new ones. As + a result, it may be a heavy, expensive operation. Rather, prefer creating + multiple pipelines with the different states, and + \l{QRhiCommandBuffer::setGraphicsPipeline()}{switch between them} when + recording the render pass. + \note Setting the shader stages is mandatory. There must be at least one stage, and there must be a vertex stage. \note Setting the shader resource bindings is mandatory. The referenced QRhiShaderResourceBindings must already have create() called on it by the time create() is called. Associating with a QRhiShaderResourceBindings that - has no bindings is also valid, as long as no shader in any stage expects - any resources. + has no bindings is also valid, as long as no shader in any stage expects any + resources. Using a QRhiShaderResourceBindings object that does not specify + any actual resources (i.e., the buffers, textures, etc. for the binding + points are set to \nullptr) is valid as well, as long as a + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings, that specifies resources for all the bindings, + is going to be set via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} when + recording the render pass. \note Setting the render pass descriptor is mandatory. To obtain a QRhiRenderPassDescriptor that can be passed to setRenderPassDescriptor(), @@ -4300,20 +6416,49 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) render target's color and depth stencil attachments. \note The depth test, depth write, and stencil test are disabled by - default. + default. The face culling mode defaults to no culling. \note stencilReadMask() and stencilWriteMask() apply to both faces. They both default to 0xFF. - */ -/*! - \fn void QRhiGraphicsPipeline::setTargetBlends(const QList<TargetBlend> &blends) + \section2 Example usage + + All settings of a graphics pipeline have defaults which might be suitable + to many applications. Therefore a minimal example of creating a graphics + pipeline could be the following. This assumes that the vertex shader takes + a single \c{vec3 position} input at the input location 0. With the + QRhiShaderResourceBindings and QRhiRenderPassDescriptor objects, plus the + QShader collections for the vertex and fragment stages, a pipeline could be + created like this: + + \code + QRhiShaderResourceBindings *srb; + QRhiRenderPassDescriptor *rpDesc; + QShader vs, fs; + // ... + + QRhiVertexInputLayout inputLayout; + inputLayout.setBindings({ { 3 * sizeof(float) } }); + inputLayout.setAttributes({ { 0, 0, QRhiVertexInputAttribute::Float3, 0 } }); - Sets the blend specification for color attachments. Each element in \a - blends corresponds to a color attachment of the render target. + QRhiGraphicsPipeline *ps = rhi->newGraphicsPipeline(); + ps->setShaderStages({ { QRhiShaderStage::Vertex, vs }, { QRhiShaderStage::Fragment, fs } }); + ps->setVertexInputLayout(inputLayout); + ps->setShaderResourceBindings(srb); + ps->setRenderPassDescriptor(rpDesc); + if (!ps->create()) { error(); } + \endcode + + The above code creates a pipeline object that uses the defaults for many + settings and states. For example, it will use a \l Triangles topology, no + backface culling, blending is disabled but color write is enabled for all + four channels, depth test/write are disabled, stencil operations are + disabled. - By default no blends are set, which is a shortcut to disabling blending and - enabling color write for all four channels. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhiCommandBuffer, QRhi */ /*! @@ -4472,21 +6617,85 @@ QDebug operator<<(QDebug dbg, const QRhiShaderResourceBindings &srb) */ /*! - \class QRhiGraphicsPipeline::TargetBlend - \internal + \struct QRhiGraphicsPipeline::TargetBlend \inmodule QtGui + \since 6.6 \brief Describes the blend state for one color attachment. Defaults to color write enabled, blending disabled. The blend values are set up for pre-multiplied alpha (One, OneMinusSrcAlpha, One, - OneMinusSrcAlpha) by default. + OneMinusSrcAlpha) by default. This means that to get the alpha blending + mode Qt Quick uses, it is enough to set the \c enable flag to true while + leaving other values at their defaults. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! - \class QRhiGraphicsPipeline::StencilOpState - \internal + \variable QRhiGraphicsPipeline::TargetBlend::colorWrite + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::enable + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::srcColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::dstColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::opColor + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::srcAlpha + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::dstAlpha + */ + +/*! + \variable QRhiGraphicsPipeline::TargetBlend::opAlpha + */ + +/*! + \struct QRhiGraphicsPipeline::StencilOpState \inmodule QtGui + \since 6.6 \brief Describes the stencil operation state. + + The default-constructed StencilOpState has the following set: + \list + \li failOp - \l Keep + \li depthFailOp - \l Keep + \li passOp - \l Keep + \li compareOp \l Always + \endlist + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::failOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::depthFailOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::passOp + */ + +/*! + \variable QRhiGraphicsPipeline::StencilOpState::compareOp */ /*! @@ -4506,7 +6715,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const } /*! - \fn bool QRhiGraphicsPipeline::create() + \fn virtual bool QRhiGraphicsPipeline::create() = 0 Creates the corresponding native graphics resources. If there are already resources present due to an earlier create() with no corresponding @@ -4514,23 +6723,132 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const \return \c true when successful, \c false when a graphics operation failed. Regardless of the return value, calling destroy() is always safe. + + \note This may be, depending on the underlying graphics API, an expensive + operation, especially when shaders get compiled/optimized from source or + from an intermediate bytecode format to the GPU's own instruction set. + Where applicable, the QRhi backend automatically sets up the relevant + non-persistent facilities to accelerate this, for example the Vulkan + backend automatically creates a \c VkPipelineCache to improve data reuse + during the lifetime of the application. + + \note Drivers may also employ various persistent (disk-based) caching + strategies for shader and pipeline data, which is hidden to and is outside + of Qt's control. In some cases, depending on the graphics API and the QRhi + backend, there are facilities within QRhi for manually managing such a + cache, allowing the retrieval of a serializable blob that can then be + reloaded in the future runs of the application to ensure faster pipeline + creation times. See QRhi::pipelineCacheData() and + QRhi::setPipelineCacheData() for details. Note also that when working with + a QRhi instance managed by a higher level Qt framework, such as Qt Quick, + it is possible that such disk-based caching is taken care of automatically, + for example QQuickWindow uses a disk-based pipeline cache by default (which + comes in addition to any driver-level caching). + */ + +/*! + \fn QRhiGraphicsPipeline::Flags QRhiGraphicsPipeline::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiGraphicsPipeline::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiGraphicsPipeline::Topology QRhiGraphicsPipeline::topology() const + \return the currently set primitive topology. + */ + +/*! + \fn void QRhiGraphicsPipeline::setTopology(Topology t) + Sets the primitive topology \a t. + */ + +/*! + \fn QRhiGraphicsPipeline::CullMode QRhiGraphicsPipeline::cullMode() const + \return the currently set face culling mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setCullMode(CullMode mode) + Sets the specified face culling \a mode. + */ + +/*! + \fn QRhiGraphicsPipeline::FrontFace QRhiGraphicsPipeline::frontFace() const + \return the currently set front face mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setFrontFace(FrontFace f) + Sets the front face mode \a f. + */ + +/*! + \fn void QRhiGraphicsPipeline::setTargetBlends(std::initializer_list<TargetBlend> list) + + Sets the \a list of render target blend settings. This is a list because + when multiple render targets are used (i.e., a QRhiTextureRenderTarget with + more than one QRhiColorAttachment), there needs to be a TargetBlend + structure per render target (color attachment). + + By default there is one default-constructed TargetBlend set. + + \sa QRhi::MaxColorAttachments + */ + +/*! + \fn template<typename InputIterator> void QRhiGraphicsPipeline::setTargetBlends(InputIterator first, InputIterator last) + Sets the list of render target blend settings from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::cbeginTargetBlends() const + \return a const iterator pointing to the first item in the render target blend setting list. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::cendTargetBlends() const + \return a const iterator pointing just after the last item in the render target blend setting list. + */ + +/*! + \fn const QRhiGraphicsPipeline::TargetBlend *QRhiGraphicsPipeline::targetBlendAt(qsizetype index) const + \return the render target blend setting at the specified \a index. + */ + +/*! + \fn qsizetype QRhiGraphicsPipeline::targetBlendCount() const + \return the number of render target blend settings. + */ + +/*! + \fn bool QRhiGraphicsPipeline::hasDepthTest() const + \return true if depth testing is enabled. */ /*! \fn void QRhiGraphicsPipeline::setDepthTest(bool enable) - Enables or disables depth testing. Both depth test and the writing out of - depth data are disabled by default. + Enables or disables depth testing based on \a enable. Both depth test and + the writing out of depth data are disabled by default. \sa setDepthWrite() */ /*! + \fn bool QRhiGraphicsPipeline::hasDepthWrite() const + \return true if depth write is enabled. + */ + +/*! \fn void QRhiGraphicsPipeline::setDepthWrite(bool enable) - Controls the writing out of depth data into the depth buffer. By default - this is disabled. Depth write is typically enabled together with the depth - test. + Controls the writing out of depth data into the depth buffer based on + \a enable. By default this is disabled. Depth write is typically enabled + together with the depth test. \note Enabling depth write without having depth testing enabled may not lead to the desired result, and should be avoided. @@ -4539,9 +6857,233 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const */ /*! + \fn QRhiGraphicsPipeline::CompareOp QRhiGraphicsPipeline::depthOp() const + \return the depth comparison function. + */ + +/*! + \fn void QRhiGraphicsPipeline::setDepthOp(CompareOp op) + Sets the depth comparison function \a op. + */ + +/*! + \fn bool QRhiGraphicsPipeline::hasStencilTest() const + \return true if stencil testing is enabled. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilTest(bool enable) + Enables or disables stencil tests based on \a enable. + By default this is disabled. + */ + +/*! + \fn QRhiGraphicsPipeline::StencilOpState QRhiGraphicsPipeline::stencilFront() const + \return the current stencil test state for front faces. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilFront(const StencilOpState &state) + Sets the stencil test \a state for front faces. + */ + +/*! + \fn QRhiGraphicsPipeline::StencilOpState QRhiGraphicsPipeline::stencilBack() const + \return the current stencil test state for back faces. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilBack(const StencilOpState &state) + Sets the stencil test \a state for back faces. + */ + +/*! + \fn quint32 QRhiGraphicsPipeline::stencilReadMask() const + \return the currrent stencil read mask. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilReadMask(quint32 mask) + Sets the stencil read \a mask. The default value is 0xFF. + */ + +/*! + \fn quint32 QRhiGraphicsPipeline::stencilWriteMask() const + \return the current stencil write mask. + */ + +/*! + \fn void QRhiGraphicsPipeline::setStencilWriteMask(quint32 mask) + Sets the stencil write \a mask. The default value is 0xFF. + */ + +/*! + \fn int QRhiGraphicsPipeline::sampleCount() const + \return the currently set sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiGraphicsPipeline::setSampleCount(int s) + + Sets the sample count. Typical values for \a s are 1, 4, or 8. The pipeline + must always be compatible with the render target, i.e. the sample counts + must match. + + \sa QRhi::supportedSampleCounts() + */ + +/*! + \fn float QRhiGraphicsPipeline::lineWidth() const + \return the currently set line width. The default is 1.0f. + */ + +/*! + \fn void QRhiGraphicsPipeline::setLineWidth(float width) + + Sets the line \a width. If the QRhi::WideLines feature is reported as + unsupported at runtime, values other than 1.0f are ignored. + */ + +/*! + \fn int QRhiGraphicsPipeline::depthBias() const + \return the currently set depth bias. + */ + +/*! + \fn void QRhiGraphicsPipeline::setDepthBias(int bias) + Sets the depth \a bias. The default value is 0. + */ + +/*! + \fn float QRhiGraphicsPipeline::slopeScaledDepthBias() const + \return the currently set slope scaled depth bias. + */ + +/*! + \fn void QRhiGraphicsPipeline::setSlopeScaledDepthBias(float bias) + Sets the slope scaled depth \a bias. The default value is 0. + */ + +/*! + \fn void QRhiGraphicsPipeline::setShaderStages(std::initializer_list<QRhiShaderStage> list) + Sets the \a list of shader stages. + */ + +/*! + \fn template<typename InputIterator> void QRhiGraphicsPipeline::setShaderStages(InputIterator first, InputIterator last) + Sets the list of shader stages from the iterators \a first and \a last. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::cbeginShaderStages() const + \return a const iterator pointing to the first item in the shader stage list. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::cendShaderStages() const + \return a const iterator pointing just after the last item in the shader stage list. + */ + +/*! + \fn const QRhiShaderStage *QRhiGraphicsPipeline::shaderStageAt(qsizetype index) const + \return the shader stage at the specified \a index. + */ + +/*! + \fn qsizetype QRhiGraphicsPipeline::shaderStageCount() const + \return the number of shader stages in this pipeline. + */ + +/*! + \fn QRhiVertexInputLayout QRhiGraphicsPipeline::vertexInputLayout() const + \return the currently set vertex input layout specification. + */ + +/*! + \fn void QRhiGraphicsPipeline::setVertexInputLayout(const QRhiVertexInputLayout &layout) + Specifies the vertex input \a layout. + */ + +/*! + \fn QRhiShaderResourceBindings *QRhiGraphicsPipeline::shaderResourceBindings() const + \return the currently associated QRhiShaderResourceBindings object. + */ + +/*! + \fn void QRhiGraphicsPipeline::setShaderResourceBindings(QRhiShaderResourceBindings *srb) + + Associates with \a srb describing the resource binding layout and the + resources (QRhiBuffer, QRhiTexture) themselves. The latter is optional, + because only the layout matters during pipeline creation. Therefore, the \a + srb passed in here can leave the actual buffer or texture objects + unspecified (\nullptr) as long as there is another, + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings bound via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} before + recording the draw calls. + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiGraphicsPipeline::renderPassDescriptor() const + \return the currently set QRhiRenderPassDescriptor. + */ + +/*! + \fn void QRhiGraphicsPipeline::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + Associates with the specified QRhiRenderPassDescriptor \a desc. + */ + +/*! + \fn int QRhiGraphicsPipeline::patchControlPointCount() const + \return the currently set patch control point count. + */ + +/*! + \fn void QRhiGraphicsPipeline::setPatchControlPointCount(int count) + + Sets the number of patch control points to \a count. The default value is + 3. This is used only when the topology is set to \l Patches. + */ + +/*! + \fn QRhiGraphicsPipeline::PolygonMode QRhiGraphicsPipeline::polygonMode() const + \return the polygon mode. + */ + +/*! + \fn void QRhiGraphicsPipeline::setPolygonMode(PolygonMode mode) + Sets the polygon \a mode. The default is Fill. + + \sa QRhi::NonFillPolygonMode + */ + +/*! + \fn int QRhiGraphicsPipeline::multiViewCount() const + \return the view count. The default is 0, indicating no multiview rendering. + \since 6.7 + */ + +/*! + \fn void QRhiGraphicsPipeline::setMultiViewCount(int count) + Sets the view \a count for multiview rendering. The default is 0, + indicating no multiview rendering. + \a count must be 2 or larger to trigger multiview rendering. + + Multiview is only available when the \l{QRhi::MultiView}{MultiView feature} + is reported as supported. The render target must be a 2D texture array, and + the color attachment for the render target must have the same \a count set. + + See QRhiColorAttachment::setMultiViewCount() for further details on + multiview rendering. + + \since 6.7 + \sa QRhi::MultiView, QRhiColorAttachment::setMultiViewCount() + */ + +/*! \class QRhiSwapChain - \internal \inmodule QtGui + \since 6.6 \brief Swapchain resource. A swapchain enables presenting rendering results to a surface. A swapchain @@ -4551,7 +7093,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const Below is a typical pattern for creating and managing a swapchain and some associated resources in order to render onto a QWindow: - \badcode + \code void init() { sc = rhi->newSwapChain(); @@ -4597,7 +7139,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const Releasing the swapchain must happen while the QWindow and the underlying native window is fully up and running. Building on the previous example: - \badcode + \code void releaseSwapChain() { if (hasSwapChain) { @@ -4629,7 +7171,7 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const events. QExposeEvent is a loosely specified event that is sent whenever a window gets mapped, obscured, and resized, depending on the platform. - \badcode + \code void Window::exposeEvent(QExposeEvent *) { // initialize and start rendering when the window becomes usable for graphics purposes @@ -4675,6 +7217,9 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const command at the end of a frame. For OpenGL, it is necessary to request the appropriate sample count also via QSurfaceFormat, by calling QSurfaceFormat::setDefaultFormat() before initializing the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -4733,6 +7278,14 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const \enum QRhiSwapChain::Format Describes the swapchain format. The default format is SDR. + This enum is used with + \l{QRhiSwapChain::isFormatSupported()}{isFormatSupported()} to check + upfront if creating the swapchain with the given format is supported by the + platform and the window's associated screen, and with + \l{QRhiSwapChain::setFormat()}{setFormat()} + to set the requested format in the swapchain before calling + \l{QRhiSwapChain::createOrResize()}{createOrResize()} for the first time. + \value SDR 8-bit RGBA or BGRA, depending on the backend and platform. With OpenGL ES in particular, it could happen that the platform provides less than 8 bits (e.g. due to EGL and the QSurfaceFormat choosing a 565 or 444 @@ -4744,10 +7297,14 @@ QRhiResource::Type QRhiGraphicsPipeline::resourceType() const (same as SDR/sRGB) and linear colors. Conversion to the display's native color space (such as, HDR10) is performed by the windowing system. On Windows this is the canonical color space of the system compositor, and is - the recommended format for HDR swapchains in general. + the recommended format for HDR swapchains in general on desktop platforms. \value HDR10 10-bit unsigned int RGB or BGR with 2 bit alpha, high dynamic range, HDR10 (Rec. 2020) color space with an ST2084 PQ transfer function. + + \value HDRExtendedDisplayP3Linear 16-bit float RGBA, high dynamic range, + extended linear Display P3 color space. The primary choice for HDR on + platforms such as iOS and VisionOS. */ /*! @@ -4793,7 +7350,7 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! - \fn QSize QRhiSwapChain::surfacePixelSize() + \fn virtual QSize QRhiSwapChain::surfacePixelSize() = 0 \return The size of the window's associated surface or layer. @@ -4801,13 +7358,13 @@ QRhiResource::Type QRhiSwapChain::resourceType() const QWindow::devicePixelRatio()}. With some graphics APIs and windowing system interfaces (for example, Vulkan) there is a theoretical possibility for a surface to assume a size different from the associated window. To support - these cases, rendering logic must always base size-derived calculations + these cases, \b{rendering logic must always base size-derived calculations (such as, viewports) on the size reported from QRhiSwapChain, and never on - the size queried from QWindow. + the size queried from QWindow}. - \note Can also be called before createOrResize(), if at least window() is - already set) This in combination with currentPixelSize() allows to detect - when a swapchain needs to be resized. However, watch out for the fact that + \note \b{Can also be called before createOrResize(), if at least window() is + already set. This in combination with currentPixelSize() allows to detect + when a swapchain needs to be resized.} However, watch out for the fact that the size of the underlying native object (surface, layer, or similar) is "live", so whenever this function is called, it returns the latest value reported by the underlying implementation, without any atomicity guarantee. @@ -4828,9 +7385,9 @@ QRhiResource::Type QRhiSwapChain::resourceType() const */ /*! - \fn bool QRhiSwapChain::isFormatSuported(Format f) + \fn virtual bool QRhiSwapChain::isFormatSupported(Format f) = 0 - \return true if the given swapchain format is supported. SDR is always + \return true if the given swapchain format \a f is supported. SDR is always supported. \note Can be called independently of createOrResize(), but window() must @@ -4841,20 +7398,45 @@ QRhiResource::Type QRhiSwapChain::resourceType() const time. If the result is true for a HDR format, then creating the swapchain with that format is expected to succeed as long as the window is not moved to another screen in the meantime. + + The main use of this function is to call it before the first + createOrResize() after the window is already set. This allow the QRhi + backends to perform platform or windowing system specific queries to + determine if the window (and the screen it is on) is capable of true HDR + output with the specified format. + + When the format is reported as supported, call setFormat() to set the + requested format and call createOrResize(). Be aware of the consequences + however: successfully requesting a HDR format will involve having to deal + with a different color space, possibly doing white level correction for + non-HDR-aware content, adjusting tonemapping methods, adjusting offscreen + render target settings, etc. + + \sa setFormat() */ /*! - \fn QRhiCommandBuffer *QRhiSwapChain::currentFrameCommandBuffer() + \fn virtual QRhiCommandBuffer *QRhiSwapChain::currentFrameCommandBuffer() = 0 - \return a command buffer on which rendering commands can be recorded. Only - valid within a QRhi::beginFrame() - QRhi::endFrame() block where - beginFrame() was called with this swapchain. + \return a command buffer on which rendering commands and resource updates + can be recorded within a \l{QRhi::beginFrame()}{beginFrame} - + \l{QRhi::endFrame()}{endFrame} block, assuming beginFrame() was called with + this swapchain. - \note the value must not be cached and reused between frames + \note The returned object is valid also after endFrame(), up until the next + beginFrame(), but the returned command buffer should not be used to record + any commands then. Rather, it can be used to query data collected during + the frame (or previous frames), for example by calling + \l{QRhiCommandBuffer::lastCompletedGpuTime()}{lastCompletedGpuTime()}. + + \note The value must not be cached and reused between frames. The caller + should not hold on to the returned object once + \l{QRhi::beginFrame()}{beginFrame()} is called again. Instead, the command + buffer object should be queried again by calling this function. */ /*! - \fn QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget() + \fn virtual QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget() = 0 \return a render target that can used with beginPass() in order to render the swapchain's current backbuffer. Only valid within a @@ -4879,10 +7461,9 @@ QRhiResource::Type QRhiSwapChain::resourceType() const is backed by two color buffers, one for each eye, instead of just one. When stereoscopic rendering is not supported, the return value will be - the default target. For the time being the only backend and 3D API where traditional - stereoscopic rendering is supported is OpenGL (excluding OpenGL ES), in + the default target. It is supported by all hardware backends except for Metal, in combination with \l QSurfaceFormat::StereoBuffers, assuming it is supported - by the graphics and display driver stack at run time. All other backends + by the graphics and display driver stack at run time. Metal and Null backends are going to return the default render target from this overload. \note the value must not be cached and reused between frames @@ -4894,7 +7475,7 @@ QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer tar } /*! - \fn bool QRhiSwapChain::createOrResize() + \fn virtual bool QRhiSwapChain::createOrResize() = 0 Creates the swapchain if not already done and resizes the swapchain buffers to match the current size of the targeted surface. Call this whenever the @@ -4910,18 +7491,121 @@ QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer tar */ /*! + \fn QWindow *QRhiSwapChain::window() const + \return the currently set window. + */ + +/*! + \fn void QRhiSwapChain::setWindow(QWindow *window) + Sets the \a window. + */ + +/*! + \fn QRhiSwapChainProxyData QRhiSwapChain::proxyData() const + \return the currently set proxy data. + */ + +/*! + \fn void QRhiSwapChain::setProxyData(const QRhiSwapChainProxyData &d) + Sets the proxy data \a d. + + \sa QRhi::updateSwapChainProxyData() + */ + +/*! + \fn QRhiSwapChain::Flags QRhiSwapChain::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiSwapChain::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiSwapChain::Format QRhiSwapChain::format() const + \return the currently set format. + */ + +/*! + \fn void QRhiSwapChain::setFormat(Format f) + Sets the format \a f. + + Avoid setting formats that are reported as unsupported from + isFormatSupported(). Note that support for a given format may depend on the + screen the swapchain's associated window is opened on. On some platforms, + such as Windows and macOS, for HDR output to work it is necessary to have + HDR output enabled in the display settings. + + See isFormatSupported(), \l QRhiSwapChainHdrInfo, and \l Format for more + information on high dynamic range output. + */ + +/*! + \fn QRhiRenderBuffer *QRhiSwapChain::depthStencil() const + \return the currently associated renderbuffer for depth-stencil. + */ + +/*! + \fn void QRhiSwapChain::setDepthStencil(QRhiRenderBuffer *ds) + Sets the renderbuffer \a ds for use as a depth-stencil buffer. + */ + +/*! + \fn int QRhiSwapChain::sampleCount() const + \return the currently set sample count. 1 means no multisample antialiasing. + */ + +/*! + \fn void QRhiSwapChain::setSampleCount(int samples) + + Sets the sample count. Common values for \a samples are 1 (no MSAA), 4 (4x + MSAA), or 8 (8x MSAA). + + \sa QRhi::supportedSampleCounts() + */ + +/*! + \fn QRhiRenderPassDescriptor *QRhiSwapChain::renderPassDescriptor() const + \return the currently associated QRhiRenderPassDescriptor object. + */ + +/*! + \fn void QRhiSwapChain::setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) + Associates with the QRhiRenderPassDescriptor \a desc. + */ + +/*! + \fn virtual QRhiRenderPassDescriptor *QRhiSwapChain::newCompatibleRenderPassDescriptor() = 0; + + \return a new QRhiRenderPassDescriptor that is compatible with this swapchain. + + The returned value is used in two ways: it can be passed to + setRenderPassDescriptor() and + QRhiGraphicsPipeline::setRenderPassDescriptor(). A render pass descriptor + describes the attachments (color, depth/stencil) and the load/store + behavior that can be affected by flags(). A QRhiGraphicsPipeline can only + be used in combination with a swapchain that has a + \l{QRhiRenderPassDescriptor::isCompatible()}{compatible} + QRhiRenderPassDescriptor set. + + \sa createOrResize() + */ + +/*! \struct QRhiSwapChainHdrInfo - \internal \inmodule QtGui + \since 6.6 \brief Describes the high dynamic range related information of the swapchain's associated output. - To perform tonemapping, one often needs to know the maximum luminance of - the display the swapchain's window is associated with. While this is often - made user-configurable, it can be highly useful to set defaults based on - the values reported by the display itself, thus providing a decent starting - point. + To perform HDR-compatible tonemapping, where the target range is not [0,1], + one often needs to know the maximum luminance of the display the + swapchain's window is associated with. While this is often made + user-configurable (think brightness, gamma and similar settings in games), + it can be highly useful to set defaults based on the values reported by the + display itself, thus providing a decent starting point. There are some problems however: the information is exposed in different forms on different platforms, whereas with cross-platform graphics APIs @@ -4929,11 +7613,6 @@ QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer tar information is not in the scope of the API (and may rather be retrievable via other platform-specific means, if any). - The struct returned from QRhiSwapChain::hdrInfo() contains either some - hard-coded defaults, indicated by the \c isHardCodedDefaults field, or real - values received from an API such as DXGI (IDXGIOutput6) or Cocoa - (NSScreen). The default is 1000 nits for maximum luminance. - With Metal on macOS/iOS, there is no luminance values exposed in the platform APIs. Instead, the maximum color component value, that would be 1.0 in a non-HDR setup, is provided. The \c limitsType field indicates what @@ -4942,17 +7621,168 @@ QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer tar fit. With an API like Vulkan, where there is no way to get such information, the - values are always the built-in defaults and \c isHardCodedDefaults is - always true. + values are always the built-in defaults. + + Therefore, the struct returned from QRhiSwapChain::hdrInfo() contains + either some hard-coded defaults or real values received from an API such as + DXGI (IDXGIOutput6) or Cocoa (NSScreen). When no platform queries are + available (or needs using platform facilities out of scope for QRhi), the + hard-coded defaults are a maximum luminance of 1000 nits and an SDR white + level of 200. + + The struct also exposes the presumed luminance behavior of the platform and + its compositor, to indicate what a color component value of 1.0 is treated + as in a HDR color buffer. In some cases it will be necessary to perform + color correction of non-HDR content composited with HDR content. To enable + this, the SDR white level is queried from the system on some platforms + (Windows) and exposed here. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. \sa QRhiSwapChain::hdrInfo() */ /*! + \enum QRhiSwapChainHdrInfo::LimitsType + + \value LuminanceInNits Indicates that the \l limits union has its + \c luminanceInNits struct set + + \value ColorComponentValue Indicates that the \l limits union has its + \c colorComponentValue struct set +*/ + +/*! + \enum QRhiSwapChainHdrInfo::LuminanceBehavior + + \value SceneReferred Indicates that the color value of 1.0 is interpreted + as 80 nits. This is the behavior of HDR-enabled windows with the Windows + compositor. See + \l{https://learn.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range}{this + page} for more information on HDR on Windows. + + \value DisplayReferred Indicates that the color value of 1.0 is interpreted + as the value of the SDR white. (which can be e.g. 200 nits, but will vary + depending on screen brightness) This is the behavior of HDR-enabled windows + on Apple platforms. See + \l{https://developer.apple.com/documentation/metal/hdr_content/displaying_hdr_content_in_a_metal_layer}{this + page} for more information on Apple's EDR system. +*/ + +/*! + \variable QRhiSwapChainHdrInfo::limitsType + + With Metal on macOS/iOS, there is no luminance values exposed in the + platform APIs. Instead, the maximum color component value, that would be + 1.0 in a non-HDR setup, is provided. This value indicates what kind of + information is available in \l limits. + + \sa QRhiSwapChain::hdrInfo() +*/ + +/*! + \variable QRhiSwapChainHdrInfo::limits + + Contains the actual values queried from the graphics API or the platform. + The type of data is indicated by \l limitsType. This is therefore a union. + There are currently two options: + + Luminance values in nits: + + \code + struct { + float minLuminance; + float maxLuminance; + } luminanceInNits; + \endcode + + On Windows the minimum and maximum luminance depends on the screen + brightness. While not relevant for desktops, on laptops the screen + brightness may change at any time. Increasing brightness implies decreased + maximum luminance. In addition, the results may also be dependent on the + HDR Content Brightness set in Windows Settings' System/Display/HDR view, + if there is such a setting. + + Note however that the changes made to the laptop screen's brightness or in + the system settings while the application is running are not necessarily + reflected in the returned values, meaning calling hdrInfo() again may still + return the same luminance range as before for the rest of the process' + lifetime. The exact behavior is up to DXGI and Qt has no control over it. + + \note The Windows compositor works in scene-referred mode for HDR content. + A color component value of 1.0 corresponds to a luminance of 80 nits. When + rendering non-HDR content (e.g. 2D UI elements), the correction of the + white level is often necessary. (e.g., outputting the fragment color (1, 1, + 1) will likely lead to showing a shade of white that is too dim on-screen) + See \l sdrWhiteLevel. + + For macOS/iOS, the current maximum and potential maximum color + component values are provided: + + \code + struct { + float maxColorComponentValue; + float maxPotentialColorComponentValue; + } colorComponentValue; + \endcode + + The value may depend on the screen brightness, which on laptops means that + the result may change in the next call to hdrInfo() if the brightness was + changed in the meantime. The maximum screen brightness implies a maximum + color value of 1.0. + + \note Apple's EDR is display-referred. 1.0 corresponds to a luminance level + of SDR white (e.g. 200 nits), the value of which varies based on the screen + brightness and possibly other settings. The exact luminance value for that, + or the maximum luminance of the display, are not exposed to the + applications. + + \note It has been observed that the color component values are not set to + the correct larger-than-1 value right away on startup on some macOS + systems, but the values tend to change during or after the first frame. + + \sa QRhiSwapChain::hdrInfo() +*/ + +/*! + \variable QRhiSwapChainHdrInfo::luminanceBehavior + + Describes the platform's presumed behavior with regards to color values. + + \sa sdrWhiteLevel + */ + +/*! + \variable QRhiSwapChainHdrInfo::sdrWhiteLevel + + On Windows this is the dynamic SDR white level in nits. The value is + dependent on the screen brightness (on laptops), and the SDR or HDR Content + Brightness settings in the Windows settings' System/Display/HDR view. + + To perform white level correction for non-HDR (SDR) content, such as 2D UI + elemenents, multiply the final color with sdrWhiteLevel / 80.0 whenever + \l luminanceBehavior is SceneReferred. (assuming Windows and a linear + extended sRGB (scRGB) color space) + + On other platforms the value is always a pre-defined value, 200. This may + not match the system's actual SDR white level, but the value of this + variable is not relevant in practice when the \l luminanceBehavior is + DisplayReferred, because then the color component value of 1.0 refers to + the SDR white by default. + + \sa luminanceBehavior +*/ + +/*! \return the HDR information for the associated display. - The returned struct is always the default one if createOrResize() has not - been successfully called yet. + Do not assume that this is a cheap operation. Depending on the platform, + this function makes various platform queries which may have a performance + impact. + + \note Can be called before createOrResize() as long as the window is + \l{setWindow()}{set}. \note What happens when moving a window with an initialized swapchain between displays (HDR to HDR with different characteristics, HDR to SDR, @@ -4967,10 +7797,11 @@ QRhiRenderTarget *QRhiSwapChain::currentFrameRenderTarget(StereoTargetBuffer tar QRhiSwapChainHdrInfo QRhiSwapChain::hdrInfo() { QRhiSwapChainHdrInfo info; - info.isHardCodedDefaults = true; info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; info.limits.luminanceInNits.minLuminance = 0.0f; info.limits.luminanceInNits.maxLuminance = 1000.0f; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; + info.sdrWhiteLevel = 200.0f; return info; } @@ -4978,7 +7809,7 @@ QRhiSwapChainHdrInfo QRhiSwapChain::hdrInfo() QDebug operator<<(QDebug dbg, const QRhiSwapChainHdrInfo &info) { QDebugStateSaver saver(dbg); - dbg.nospace() << "QRhiSwapChainHdrInfo(" << (info.isHardCodedDefaults ? "with hard-coded defaults" : "queried from system"); + dbg.nospace() << "QRhiSwapChainHdrInfo("; switch (info.limitsType) { case QRhiSwapChainHdrInfo::LuminanceInNits: dbg.nospace() << " minLuminance=" << info.limits.luminanceInNits.minLuminance @@ -4986,6 +7817,15 @@ QDebug operator<<(QDebug dbg, const QRhiSwapChainHdrInfo &info) break; case QRhiSwapChainHdrInfo::ColorComponentValue: dbg.nospace() << " maxColorComponentValue=" << info.limits.colorComponentValue.maxColorComponentValue; + dbg.nospace() << " maxPotentialColorComponentValue=" << info.limits.colorComponentValue.maxPotentialColorComponentValue; + break; + } + switch (info.luminanceBehavior) { + case QRhiSwapChainHdrInfo::SceneReferred: + dbg.nospace() << " scene-referred, SDR white level=" << info.sdrWhiteLevel; + break; + case QRhiSwapChainHdrInfo::DisplayReferred: + dbg.nospace() << " display-referred"; break; } dbg.nospace() << ')'; @@ -4995,8 +7835,8 @@ QDebug operator<<(QDebug dbg, const QRhiSwapChainHdrInfo &info) /*! \class QRhiComputePipeline - \internal \inmodule QtGui + \since 6.6 \brief Compute pipeline state resource. \note Setting the shader resource bindings is mandatory. The referenced @@ -5004,6 +7844,9 @@ QDebug operator<<(QDebug dbg, const QRhiSwapChainHdrInfo &info) time create() is called. \note Setting the shader is mandatory. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -5033,15 +7876,59 @@ QRhiComputePipeline::QRhiComputePipeline(QRhiImplementation *rhi) } /*! + \fn QRhiComputePipeline::Flags QRhiComputePipeline::flags() const + \return the currently set flags. + */ + +/*! + \fn void QRhiComputePipeline::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QRhiShaderStage QRhiComputePipeline::shaderStage() const + \return the currently set shader. + */ + +/*! + \fn void QRhiComputePipeline::setShaderStage(const QRhiShaderStage &stage) + + Sets the shader to use. \a stage can only refer to the + \l{QRhiShaderStage::Compute}{compute stage}. + */ + +/*! + \fn QRhiShaderResourceBindings *QRhiComputePipeline::shaderResourceBindings() const + \return the currently associated QRhiShaderResourceBindings object. + */ + +/*! + \fn void QRhiComputePipeline::setShaderResourceBindings(QRhiShaderResourceBindings *srb) + + Associates with \a srb describing the resource binding layout and the + resources (QRhiBuffer, QRhiTexture) themselves. The latter is optional. As + with graphics pipelines, the \a srb passed in here can leave the actual + buffer or texture objects unspecified (\nullptr) as long as there is + another, + \l{QRhiShaderResourceBindings::isLayoutCompatible()}{layout-compatible} + QRhiShaderResourceBindings bound via + \l{QRhiCommandBuffer::setShaderResources()}{setShaderResources()} before + recording the dispatch call. + */ + +/*! \class QRhiCommandBuffer - \internal \inmodule QtGui + \since 6.6 \brief Command buffer resource. Not creatable by applications at the moment. The only ways to obtain a valid QRhiCommandBuffer are to get it from the targeted swapchain via QRhiSwapChain::currentFrameCommandBuffer(), or, in case of rendering completely offscreen, initializing one via QRhi::beginOffscreenFrame(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -5102,7 +7989,7 @@ QRhiResource::Type QRhiCommandBuffer::resourceType() const return CommandBuffer; } -static const char *resourceTypeStr(QRhiResource *res) +static const char *resourceTypeStr(const QRhiResource *res) { switch (res->resourceType()) { case QRhiResource::Buffer: @@ -5153,8 +8040,10 @@ QRhiImplementation::~QRhiImplementation() qWarning("QRhi %p going down with %d unreleased resources that own native graphics objects. This is not nice.", q, int(resources.size())); } - for (QRhiResource *res : std::as_const(resources)) { - if (leakCheck) + for (auto it = resources.cbegin(), end = resources.cend(); it != end; ++it) { + QRhiResource *res = it.key(); + const bool ownsNativeResources = it.value(); + if (leakCheck && ownsNativeResources) qWarning(" %s resource %p (%s)", resourceTypeStr(res), res, res->m_objectName.constData()); // Null out the resource's rhi pointer. This is why it makes sense to do null @@ -5364,6 +8253,17 @@ void QRhiImplementation::textureFormatInfo(QRhiTexture::Format format, const QSi *bytesPerPixel = bpc; } +bool QRhiImplementation::isStencilSupportingFormat(QRhiTexture::Format format) const +{ + switch (format) { + case QRhiTexture::D24S8: + return true; + default: + break; + } + return false; +} + bool QRhiImplementation::sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps) { if (ps->cbeginShaderStages() == ps->cendShaderStages()) { @@ -5483,6 +8383,41 @@ bool QRhiImplementation::sanityCheckShaderResourceBindings(QRhiShaderResourceBin return true; } +int QRhiImplementation::effectiveSampleCount(int sampleCount) const +{ + // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. + const int s = qBound(1, sampleCount, 64); + const QList<int> supported = supportedSampleCounts(); + int result = 1; + + // Stay compatible with Qt 5 in that requesting an unsupported sample count + // is not an error (although we still do a categorized debug print about + // this), and rather a supported value, preferably a close one, not just 1, + // is used instead. This is actually deviating from Qt 5 as that performs a + // clamping only and does not handle cases such as when sample count 2 is + // not supported but 4 is. (OpenGL handles things like that gracefully, + // other APIs may not, so improve this by picking the next largest, or in + // absence of that, the largest value; this with the goal to not reduce + // quality by rather picking a larger-than-requested value than a smaller one) + + for (int i = 0, ie = supported.count(); i != ie; ++i) { + // assumes the 'supported' list is sorted + if (supported[i] >= s) { + result = supported[i]; + break; + } + } + + if (result != s) { + if (result == 1 && !supported.isEmpty()) + result = supported.last(); + qCDebug(QRHI_LOG_INFO, "Attempted to set unsupported sample count %d, using %d instead", + sampleCount, result); + } + + return result; +} + /*! \internal */ @@ -5498,15 +8433,31 @@ QRhi::~QRhi() if (!d) return; + runCleanup(); + qDeleteAll(d->pendingDeleteResources); d->pendingDeleteResources.clear(); - runCleanup(); - d->destroy(); delete d; } +void QRhiImplementation::prepareForCreate(QRhi *rhi, QRhi::Implementation impl, QRhi::Flags flags) +{ + q = rhi; + + // Play nice with QSG_INFO since that is still the most commonly used + // way to get graphics info printed from Qt Quick apps, and the Quick + // scenegraph is our primary user. + if (qEnvironmentVariableIsSet("QSG_INFO")) + const_cast<QLoggingCategory &>(QRHI_LOG_INFO()).setEnabled(QtDebugMsg, true); + + debugMarkers = flags.testFlag(QRhi::EnableDebugMarkers); + + implType = impl; + implThread = QThread::currentThread(); +} + /*! \return a new QRhi instance with a backend for the graphics API specified by \a impl with the specified \a flags. @@ -5574,7 +8525,7 @@ QRhi *QRhi::create(Implementation impl, QRhiInitParams *params, Flags flags, QRh break; #endif case Metal: -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) +#if QT_CONFIG(metal) r->d = new QRhiMetal(static_cast<QRhiMetalInitParams *>(params), static_cast<QRhiMetalNativeHandles *>(importDevice)); break; @@ -5584,31 +8535,27 @@ QRhi *QRhi::create(Implementation impl, QRhiInitParams *params, Flags flags, QRh #endif case D3D12: #ifdef Q_OS_WIN +#ifdef QRHI_D3D12_AVAILABLE r->d = new QRhiD3D12(static_cast<QRhiD3D12InitParams *>(params), static_cast<QRhiD3D12NativeHandles *>(importDevice)); break; #else + qWarning("Qt was built without Direct3D 12 support. " + "This is likely due to having ancient SDK headers (such as d3d12.h) in the Qt build environment. " + "Rebuild Qt with an SDK supporting D3D12 features introduced in Windows 10 version 1703, " + "or use an MSVC build as those typically are built with more up-to-date SDKs."); + break; +#endif +#else qWarning("This platform has no Direct3D 12 support"); break; #endif } if (r->d) { - r->d->q = r.get(); - - // Play nice with QSG_INFO since that is still the most commonly used - // way to get graphics info printed from Qt Quick apps, and the Quick - // scenegraph is our primary user. - if (qEnvironmentVariableIsSet("QSG_INFO")) - const_cast<QLoggingCategory &>(QRHI_LOG_INFO()).setEnabled(QtDebugMsg, true); - - r->d->debugMarkers = flags.testFlag(EnableDebugMarkers); - - if (r->d->create(flags)) { - r->d->implType = impl; - r->d->implThread = QThread::currentThread(); + r->d->prepareForCreate(r.get(), impl, flags); + if (r->d->create(flags)) return r.release(); - } } return nullptr; @@ -5637,7 +8584,7 @@ bool QRhi::probe(QRhi::Implementation impl, QRhiInitParams *params) // create() and then drop the result. if (impl == Metal) { -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) +#if QT_CONFIG(metal) ok = QRhiMetal::probe(static_cast<QRhiMetalInitParams *>(params)); #endif } else { @@ -5650,8 +8597,15 @@ bool QRhi::probe(QRhi::Implementation impl, QRhiInitParams *params) /*! \struct QRhiSwapChainProxyData - \internal \inmodule QtGui + \since 6.6 + + \brief Opaque data describing native objects needed to set up a swapchain. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + + \sa QRhi::updateSwapChainProxyData() */ /*! @@ -5681,7 +8635,7 @@ bool QRhi::probe(QRhi::Implementation impl, QRhiInitParams *params) */ QRhiSwapChainProxyData QRhi::updateSwapChainProxyData(QRhi::Implementation impl, QWindow *window) { -#if defined(Q_OS_MACOS) || defined(Q_OS_IOS) +#if QT_CONFIG(metal) if (impl == Metal) return QRhiMetal::updateSwapChainProxyData(window); #else @@ -5733,9 +8687,11 @@ const char *QRhi::backendName() const /*! \enum QRhiDriverInfo::DeviceType - Specifies the graphics device's type, when the information is available. In - practice this is only applicable with Vulkan and Metal. With others the - value will always be UnknownDevice. + Specifies the graphics device's type, when the information is available. + + In practice this is only applicable with Vulkan and Metal. With Direct 3D + 11 and 12, using an adapter with the software flag set leads to the value + \c CpuDevice. Otherwise, and with OpenGL, the value is always UnknownDevice. \value UnknownDevice \value IntegratedDevice @@ -5747,8 +8703,8 @@ const char *QRhi::backendName() const /*! \struct QRhiDriverInfo - \internal \inmodule QtGui + \since 6.6 \brief Describes the physical device, adapter, or graphics API implementation that is used by an initialized QRhi. @@ -5760,8 +8716,35 @@ const char *QRhi::backendName() const \c{GL_VERSION}. The deviceId is always 0 for OpenGL. vendorId is always 0 for OpenGL and Metal. deviceType is always UnknownDevice for OpenGL and Direct 3D. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiDriverInfo::deviceName + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::deviceId + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::vendorId + + \sa QRhi::driverInfo() +*/ + +/*! + \variable QRhiDriverInfo::deviceType + + \sa QRhi::driverInfo(), QRhiDriverInfo::DeviceType +*/ + #ifndef QT_NO_DEBUG_STREAM static inline const char *deviceTypeStr(QRhiDriverInfo::DeviceType type) { @@ -5829,6 +8812,33 @@ void QRhi::addCleanupCallback(const CleanupCallback &callback) } /*! + \overload + + Registers \a callback to be invoked either when the QRhi is destroyed or + when runCleanup() is called. This overload takes an opaque pointer, \a key, + that is used to ensure that a given callback is registered (and so called) + only once. + + \sa removeCleanupCallback() + */ +void QRhi::addCleanupCallback(const void *key, const CleanupCallback &callback) +{ + d->addCleanupCallback(key, callback); +} + +/*! + Deregisters the callback with \a key. If no cleanup callback was registered + with \a key, the function does nothing. Callbacks registered without a key + cannot be removed. + + \sa addCleanupCallback() + */ +void QRhi::removeCleanupCallback(const void *key) +{ + d->removeCleanupCallback(key); +} + +/*! Invokes all registered cleanup functions. The list of cleanup callbacks it then cleared. Normally destroying the QRhi does this automatically, but sometimes it can be useful to trigger cleanup in order to release all @@ -5842,42 +8852,17 @@ void QRhi::runCleanup() f(this); d->cleanupCallbacks.clear(); -} -/*! - Registers a \a callback that is called with an elapsed time calculated from - GPU timestamps asynchronously after a timestamp becomes available at some - point after presenting a frame. - - The callback is called with a float value that is meant to be in - milliseconds and represents the elapsed time on the GPU side for a given - frame. Care must be exercised with the interpretation of the value, as what - it exactly is is not controlled by Qt and depends on the underlying - graphics API and its implementation. In particular, comparing the values - between different graphics APIs is discouraged and may be meaningless. - - The timing values become available asynchronously, sometimes several frames - after the frame has been submitted in endFrame(). There is currently no way - to identify the frame. The callback is invoked whenever the timestamp - queries complete. + for (auto it = d->keyedCleanupCallbacks.cbegin(), end = d->keyedCleanupCallbacks.cend(); it != end; ++it) + it.value()(this); - \note This is only supported when the Timestamp feature is reported as - supported from isFeatureSupported(). Otherwise the \a callback is never - called. - - The \a callback is always called on the thread the QRhi lives and operates - on. While not guaranteed, it is typical that the callback is invoked from - within beginFrame(). - */ -void QRhi::addGpuFrameTimeCallback(const GpuFrameTimeCallback &callback) -{ - d->addGpuFrameTimeCallback(callback); + d->keyedCleanupCallbacks.clear(); } /*! \class QRhiResourceUpdateBatch - \internal \inmodule QtGui + \since 6.6 \brief Records upload and copy type of operations. With QRhi it is no longer possible to perform copy type of operations at @@ -5893,6 +8878,9 @@ void QRhi::addGpuFrameTimeCallback(const GpuFrameTimeCallback &callback) To get an available, empty batch from the pool, call QRhi::nextResourceUpdateBatch(). + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! @@ -5936,29 +8924,26 @@ void QRhiResourceUpdateBatch::release() that is then merged into another when starting to first render pass later on: - \badcode - void init() - { - ... - initialUpdates = rhi->nextResourceUpdateBatch(); - initialUpdates->uploadStaticBuffer(vbuf, vertexData); - initialUpdates->uploadStaticBuffer(ibuf, indexData); - ... - } + \code + void init() + { + initialUpdates = rhi->nextResourceUpdateBatch(); + initialUpdates->uploadStaticBuffer(vbuf, vertexData); + initialUpdates->uploadStaticBuffer(ibuf, indexData); + // ... + } - void render() - { - ... - QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); - if (initialUpdates) { - resUpdates->merge(initialUpdates); - initialUpdates->release(); - initialUpdates = nullptr; + void render() + { + QRhiResourceUpdateBatch *resUpdates = rhi->nextResourceUpdateBatch(); + if (initialUpdates) { + resUpdates->merge(initialUpdates); + initialUpdates->release(); + initialUpdates = nullptr; + } + // resUpdates->updateDynamicBuffer(...); + cb->beginPass(rt, clearCol, clearDs, resUpdates); } - resUpdates->updateDynamicBuffer(...); - ... - cb->beginPass(rt, clearCol, clearDs, resUpdates); - } \endcode */ void QRhiResourceUpdateBatch::merge(QRhiResourceUpdateBatch *other) @@ -6000,8 +8985,8 @@ bool QRhiResourceUpdateBatch::hasOptimalCapacity() const \note QRhi transparently manages double buffering in order to prevent stalling the graphics pipeline. The fact that a QRhiBuffer may have - multiple native underneath can be safely ignored when using the QRhi and - QRhiResourceUpdateBatch. + multiple native buffer objects underneath can be safely ignored when using + the QRhi and QRhiResourceUpdateBatch. */ void QRhiResourceUpdateBatch::updateDynamicBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) { @@ -6035,6 +9020,8 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, quint32 offset } /*! + \overload + Enqueues updating the entire QRhiBuffer \a buf created with the type QRhiBuffer::Immutable or QRhiBuffer::Static. */ @@ -6056,7 +9043,7 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, const void *da A readback is asynchronous. \a result contains a callback that is invoked when the operation has completed. The data is provided in - QRhiBufferReadbackResult::data. Upon successful completion that QByteArray + QRhiReadbackResult::data. Upon successful completion that QByteArray will have a size equal to \a size. On failure the QByteArray will be empty. \note Reading buffers with a usage different than QRhiBuffer::UniformBuffer @@ -6073,7 +9060,7 @@ void QRhiResourceUpdateBatch::uploadStaticBuffer(QRhiBuffer *buf, const void *da \sa readBackTexture(), QRhi::isFeatureSupported(), QRhi::resourceLimit() */ -void QRhiResourceUpdateBatch::readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiBufferReadbackResult *result) +void QRhiResourceUpdateBatch::readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result) { const int idx = d->activeBufferOpCount++; if (idx < d->bufferOps.size()) @@ -6147,10 +9134,10 @@ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, co application. Therefore, \a result provides not just the data but also a callback as operations on the batch are asynchronous by nature: - \badcode - beginFrame(sc); - beginPass - ... + \code + rhi->beginFrame(swapchain); + cb->beginPass(swapchain->currentFrameRenderTarget(), colorClear, dsClear); + // ... QRhiReadbackResult *rbResult = new QRhiReadbackResult; rbResult->completed = [rbResult] { { @@ -6161,11 +9148,11 @@ void QRhiResourceUpdateBatch::copyTexture(QRhiTexture *dst, QRhiTexture *src, co } delete rbResult; }; - u = nextResourceUpdateBatch(); + QRhiResourceUpdateBatch *u = nextResourceUpdateBatch(); QRhiReadbackDescription rb; // no texture -> uses the current backbuffer of sc u->readBackTexture(rb, rbResult); - endPass(u); - endFrame(sc); + cb->endPass(u); + rhi->endFrame(swapchain); \endcode \note The texture must be created with QRhiTexture::UsedAsTransferSource. @@ -6242,11 +9229,43 @@ void QRhiResourceUpdateBatch::generateMips(QRhiTexture *tex) \note Can be called outside beginFrame() - endFrame() as well since a batch instance just collects data on its own, it does not perform any operations. - \warning The maximum number of batches is 64. When this limit is reached, - the function will return null until a batch is returned to the pool. + Due to not being tied to a frame being recorded, the following sequence is + valid for example: + + \code + rhi->beginFrame(swapchain); + QRhiResourceUpdateBatch *u = rhi->nextResourceUpdateBatch(); + u->uploadStaticBuffer(buf, data); + // ... do not commit the batch + rhi->endFrame(); + // u stays valid (assuming buf stays valid as well) + rhi->beginFrame(swapchain); + swapchain->currentFrameCommandBuffer()->resourceUpdate(u); + // ... draw with buf + rhi->endFrame(); + \endcode + + \warning The maximum number of batches per QRhi is 64. When this limit is + reached, the function will return null until a batch is returned to the + pool. */ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() { + // By default we prefer spreading out the utilization of the 64 batches as + // much as possible, meaning we won't pick the first one even if it's free, + // but prefer picking one after the last picked one. Relevant due to how + // QVLA and QRhiBufferData allocations behind the bufferOps are reused; in + // typical Qt Quick scenes this leads to a form of (eventually) seeding all + // the 64 resource batches with buffer operation data allocations which are + // then reused in subsequent frames. This comes at the expense of using + // more memory, but has proven good results when (CPU) profiling typical + // Quick/Quick3D apps. + // + // Prefering memory over performance means that we always pick the first + // free batch, and triggering the aggressive deallocating of all backing + // memory (see trimOpLists) before returning it. + static const bool preferMemoryOverPerformance = qEnvironmentVariableIntValue("QT_RHI_MINIMIZE_POOLS"); + auto nextFreeBatch = [this]() -> QRhiResourceUpdateBatch * { auto isFree = [this](int i) -> QRhiResourceUpdateBatch * { const quint64 mask = 1ULL << quint64(i); @@ -6254,7 +9273,8 @@ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() d->resUpdPoolMap |= mask; QRhiResourceUpdateBatch *u = d->resUpdPool[i]; QRhiResourceUpdateBatchPrivate::get(u)->poolIndex = i; - d->lastResUpdIdx = i; + if (!preferMemoryOverPerformance) + d->lastResUpdIdx = i; return u; } return nullptr; @@ -6283,6 +9303,9 @@ QRhiResourceUpdateBatch *QRhi::nextResourceUpdateBatch() qWarning("Resource update batch pool exhausted (max is 64)"); } + if (preferMemoryOverPerformance && u) + u->d->trimOpLists(); + return u; } @@ -6297,7 +9320,15 @@ void QRhiResourceUpdateBatchPrivate::free() rhi->resUpdPoolMap &= ~mask; poolIndex = -1; + // textureOps is cleared, to not keep the potentially large image pixel + // data alive, but it is expected that the container keeps the list alloc + // at least. Only trimOpList() goes for the more aggressive route with squeeze. textureOps.clear(); + + // bufferOps is not touched, to allow reusing allocations (incl. in the + // elements' QRhiBufferData) as much as possible when this batch is used + // again in the future, which is important for performance, in particular + // with Qt Quick. } void QRhiResourceUpdateBatchPrivate::merge(QRhiResourceUpdateBatchPrivate *other) @@ -6325,13 +9356,21 @@ bool QRhiResourceUpdateBatchPrivate::hasOptimalCapacity() const void QRhiResourceUpdateBatchPrivate::trimOpLists() { - Q_ASSERT(poolIndex == -1); // must not be in use + // Unlike free(), this is expected to aggressively deallocate all memory + // used by both the buffer and texture operation lists. (i.e. using + // squeeze() to only keep the stack prealloc of the QVLAs) + // + // This (e.g. just the destruction of bufferOps elements) may have a + // non-negligible performance impact e.g. with Qt Quick with scenes where + // there are lots of buffer operations per frame. activeBufferOpCount = 0; bufferOps.clear(); + bufferOps.squeeze(); activeTextureOpCount = 0; textureOps.clear(); + textureOps.squeeze(); } /*! @@ -6386,13 +9425,13 @@ void QRhiCommandBuffer::resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates) made on \a rt. Therefore, if \a rt has a QRhiTexture color attachment \c texture, and one needs to make the texture a different size, the following is then valid: - \badcode - rt = rhi->newTextureRenderTarget({ { texture } }); + \code + QRhiTextureRenderTarget *rt = rhi->newTextureRenderTarget({ { texture } }); rt->create(); - ... + // ... texture->setPixelSize(new_size); texture->create(); - cb->beginPass(rt, ...); // this is ok, no explicit rt->create() is required before + cb->beginPass(rt, colorClear, dsClear); // this is ok, no explicit rt->create() is required before \endcode \a flags allow controlling certain advanced functionality. One commonly used @@ -6537,7 +9576,7 @@ void QRhiCommandBuffer::setShaderResources(QRhiShaderResourceBindings *srb, floats for position (so 5 floats per vertex: x, y, r, g, b). A QRhiGraphicsPipeline for this shader can then be created using the input layout: - \badcode + \code QRhiVertexInputLayout inputLayout; inputLayout.setBindings({ { 5 * sizeof(float) } @@ -6550,11 +9589,10 @@ void QRhiCommandBuffer::setShaderResources(QRhiShaderResourceBindings *srb, Here there is one buffer binding (binding number 0), with two inputs referencing it. When recording the pass, once the pipeline is set, the - vertex bindings can be specified simply like the following (using C++11 - initializer syntax), assuming vbuf is the QRhiBuffer with all the - interleaved position+color data: + vertex bindings can be specified simply like the following, assuming vbuf + is the QRhiBuffer with all the interleaved position+color data: - \badcode + \code const QRhiCommandBuffer::VertexInput vbufBinding(vbuf, 0); cb->setVertexInput(0, 1, &vbufBinding); \endcode @@ -6839,9 +9877,9 @@ const QRhiNativeHandles *QRhiCommandBuffer::nativeHandles() called when the pass recording was started with specifying QRhiCommandBuffer::ExternalContent. - With Vulkan or Metal one can query the native command buffer or encoder - objects via nativeHandles() and enqueue commands to them. With OpenGL or - Direct3D 11 the (device) context can be retrieved from + With Vulkan, Metal, or Direct3D 12 one can query the native command buffer + or encoder objects via nativeHandles() and enqueue commands to them. With + OpenGL or Direct3D 11 the (device) context can be retrieved from QRhi::nativeHandles(). However, this must never be done without ensuring the QRhiCommandBuffer's state stays up-to-date. Hence the requirement for wrapping any externally added command recording between beginExternal() and @@ -6889,6 +9927,72 @@ void QRhiCommandBuffer::endExternal() } /*! + \return the last available timestamp, in seconds, when + \l QRhi::EnableTimestamps was enabled when creating the QRhi. The value + indicates the elapsed time on the GPU during the last completed frame. + + \note Do not expect results other than 0 when the QRhi::Timestamps feature + is not reported as supported, or when QRhi::EnableTimestamps was not passed + to QRhi::create(). There are exceptions to this, because with some graphics + APIs (Metal) timings are available without having to perform extra + operations (timestamp queries), but portable applications should always + consciously opt-in to timestamp collection when they know it is needed, and + call this function accordingly. + + Care must be exercised with the interpretation of the value, as its + precision and granularity is often not controlled by Qt, and depends on the + underlying graphics API and its implementation. In particular, comparing + the values between different graphics APIs and hardware is discouraged and + may be meaningless. + + When the frame was recorded with \l{QRhi::beginFrame()}{beginFrame()} and + \l{QRhi::endFrame()}{endFrame()}, i.e., with a swapchain, the timing values + will likely become available asynchronously. The returned value may + therefore be 0 (e.g., for the first 1-2 frames) or the last known value + referring to some previous frame. The value my also + become 0 again under certain conditions, such as when resizing the window. + It can be expected that the most up-to-date available value is retrieved in + beginFrame() and becomes queriable via this function once beginFrame() + returns. + + \note Do not assume that the value refers to the previous + (\c{currently_recorded - 1}) frame. It may refer to \c{currently_recorded - + 2} or \c{currently_recorded - 3} as well. The exact behavior may depend on + the graphics API and its implementation. + + On the other hand, with offscreen frames the returned value is up-to-date + once \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} returns, because + offscreen frames reduce GPU pipelining and wait the the commands to be + complete. + + \note This means that, unlike with swapchain frames, with offscreen frames + the returned value is guaranteed to refer to the frame that has just been + submitted and completed. (assuming this function is called after + endOffscreenFrame() but before the next beginOffscreenFrame()) + + Watch out for the consequences of GPU frequency scaling and GPU clock + changes, depending on the platform. For example, on Windows the returned + timing may vary in a quite wide range between frames with modern graphics + cards, even when submitting frames with a similar, or the same workload. + This is out of scope for Qt to control and solve, generally speaking. + However, the D3D12 backend automatically calls + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-id3d12device-setstablepowerstate}{ID3D12Device::SetStablePowerState()} + whenever the environment variable \c QT_D3D_STABLE_POWER_STATE is set to a + non-zero value. This can greatly stabilize the result. It can also have a + non-insignificant effect on the CPU-side timings measured via QElapsedTimer + for example, especially when offscreen frames are involved. + + \note Do not and never ship applications to production with + \c QT_D3D_STABLE_POWER_STATE set. See the Windows API documentation for details. + + \sa QRhi::Timestamps, QRhi::EnableTimestamps + */ +double QRhiCommandBuffer::lastCompletedGpuTime() +{ + return m_rhi->lastCompletedGpuTime(this); +} + +/*! \return the value (typically an offset) \a v aligned to the uniform buffer alignment given by by ubufAlignment(). */ @@ -7024,7 +10128,8 @@ int QRhi::resourceLimit(ResourceLimit limit) const for the device, context, and similar concepts used by the backend. Cast to QRhiVulkanNativeHandles, QRhiD3D11NativeHandles, - QRhiGles2NativeHandles, QRhiMetalNativeHandles as appropriate. + QRhiD3D12NativeHandles, QRhiGles2NativeHandles, or QRhiMetalNativeHandles + as appropriate. \note No ownership is transferred, neither for the returned pointer nor for any native objects. @@ -7224,12 +10329,70 @@ void QRhi::setPipelineCacheData(const QByteArray &data) /*! \struct QRhiStats - \internal \inmodule QtGui + \since 6.6 \brief Statistics provided from the underlying memory allocator. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiStats::totalPipelineCreationTime + + The total time in milliseconds spent in graphics and compute pipeline + creation, which usually involves shader compilation or cache lookups, and + potentially expensive processing. + + \note The value should not be compared between different backends since the + concept of "pipelines" and what exactly happens under the hood during, for + instance, a call to QRhiGraphicsPipeline::create(), differ greatly between + graphics APIs and their implementations. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::blockCount + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::allocCount + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::usedBytes + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::unusedBytes + + Statistic reported from the Vulkan or D3D12 memory allocator. + + \sa QRhi::statistics() +*/ + +/*! + \variable QRhiStats::totalUsageBytes + + Valid only with D3D12 currently. Matches IDXGIAdapter3::QueryVideoMemoryInfo(). + + \sa QRhi::statistics() +*/ + #ifndef QT_NO_DEBUG_STREAM QDebug operator<<(QDebug dbg, const QRhiStats &info) { @@ -7489,6 +10652,14 @@ QRhiTexture *QRhi::newTextureArray(QRhiTexture::Format format, minification filter \a minFilter, mipmapping mode \a mipmapMode, and the addressing (wrap) modes \a addressU, \a addressV, and \a addressW. + \note Setting \a mipmapMode to a value other than \c None implies that + images for all relevant mip levels will be provided either via + \l{QRhiResourceUpdateBatch::uploadTexture()}{texture uploads} or by calling + \l{QRhiResourceUpdateBatch::generateMips()}{generateMips()} on the texture + that is used with this sampler. Attempting to use the sampler with a + texture that has no data for all relevant mip levels will lead to rendering + errors, with the exact behavior dependent on the underlying graphics API. + \sa QRhiResource::destroy() */ QRhiSampler *QRhi::newSampler(QRhiSampler::Filter magFilter, @@ -7692,26 +10863,29 @@ int QRhi::currentFrameSlot() const beginOffscreenFrame, endOffscreenFrame, beginFrame, ...) is possible too but it does reduce parallelism so it should be done only infrequently. - Offscreen frames do not let the CPU - potentially - generate another frame + Offscreen frames do not let the CPU potentially generate another frame while the GPU is still processing the previous one. This has the side effect that if readbacks are scheduled, the results are guaranteed to be available once endOffscreenFrame() returns. That is not the case with - frames targeting a swapchain. + frames targeting a swapchain: there the GPU is potentially better utilized, + but working with readback operations needs more care from the application + because endFrame(), unlike endOffscreenFrame(), does not guarantee that the + results from the readback are available at that point. The skeleton of rendering a frame without a swapchain and then reading the frame contents back could look like the following: - \badcode - QRhiReadbackResult rbResult; - QRhiCommandBuffer *cb; - beginOffscreenFrame(&cb); - beginPass - ... - u = nextResourceUpdateBatch(); - u->readBackTexture(rb, &rbResult); - endPass(u); - endOffscreenFrame(); - // image data available in rbResult + \code + QRhiReadbackResult rbResult; + QRhiCommandBuffer *cb; + rhi->beginOffscreenFrame(&cb); + cb->beginPass(rt, colorClear, dsClear); + // ... + u = nextResourceUpdateBatch(); + u->readBackTexture(rb, &rbResult); + cb->endPass(u); + rhi->endOffscreenFrame(); + // image data available in rbResult \endcode \sa endOffscreenFrame(), beginFrame() @@ -7772,6 +10946,9 @@ QRhi::FrameOpResult QRhi::finish() With some backend this list of supported values is fixed in advance, while with some others the (physical) device properties indicate what is supported at run time. + + \sa QRhiRenderBuffer::setSampleCount(), QRhiTexture::setSampleCount(), + QRhiGraphicsPipeline::setSampleCount(), QRhiSwapChain::setSampleCount() */ QList<int> QRhi::supportedSampleCounts() const { diff --git a/src/gui/rhi/qrhi.h b/src/gui/rhi/qrhi.h new file mode 100644 index 0000000000..d20b7e00d1 --- /dev/null +++ b/src/gui/rhi/qrhi.h @@ -0,0 +1,2026 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHI_H +#define QRHI_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qsize.h> +#include <QtCore/qlist.h> +#include <QtCore/qvarlengtharray.h> +#include <QtCore/qthread.h> +#include <QtGui/qmatrix4x4.h> +#include <QtGui/qcolor.h> +#include <QtGui/qimage.h> +#include <functional> +#include <array> + +#include <rhi/qshader.h> + +QT_BEGIN_NAMESPACE + +class QWindow; +class QRhi; +class QRhiImplementation; +class QRhiBuffer; +class QRhiRenderBuffer; +class QRhiTexture; +class QRhiSampler; +class QRhiCommandBuffer; +class QRhiResourceUpdateBatch; +class QRhiResourceUpdateBatchPrivate; +class QRhiSwapChain; + +class Q_GUI_EXPORT QRhiDepthStencilClearValue +{ +public: + QRhiDepthStencilClearValue() = default; + QRhiDepthStencilClearValue(float d, quint32 s); + + float depthClearValue() const { return m_d; } + void setDepthClearValue(float d) { m_d = d; } + + quint32 stencilClearValue() const { return m_s; } + void setStencilClearValue(quint32 s) { m_s = s; } + +private: + float m_d = 1.0f; + quint32 m_s = 0; + + friend bool operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + { + return a.m_d == b.m_d && a.m_s == b.m_s; + } + + friend bool operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiDepthStencilClearValue &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_d); + seed = hash(seed, v.m_s); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiDepthStencilClearValue, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDepthStencilClearValue &); +#endif + +class Q_GUI_EXPORT QRhiViewport +{ +public: + QRhiViewport() = default; + QRhiViewport(float x, float y, float w, float h, float minDepth = 0.0f, float maxDepth = 1.0f); + + std::array<float, 4> viewport() const { return m_rect; } + void setViewport(float x, float y, float w, float h) { + m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + } + + float minDepth() const { return m_minDepth; } + void setMinDepth(float minDepth) { m_minDepth = minDepth; } + + float maxDepth() const { return m_maxDepth; } + void setMaxDepth(float maxDepth) { m_maxDepth = maxDepth; } + +private: + std::array<float, 4> m_rect { { 0.0f, 0.0f, 0.0f, 0.0f } }; + float m_minDepth = 0.0f; + float m_maxDepth = 1.0f; + + friend bool operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept + { + return a.m_rect == b.m_rect + && a.m_minDepth == b.m_minDepth + && a.m_maxDepth == b.m_maxDepth; + } + + friend bool operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiViewport &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_rect[0]); + seed = hash(seed, v.m_rect[1]); + seed = hash(seed, v.m_rect[2]); + seed = hash(seed, v.m_rect[3]); + seed = hash(seed, v.m_minDepth); + seed = hash(seed, v.m_maxDepth); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiViewport, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiViewport &); +#endif + +class Q_GUI_EXPORT QRhiScissor +{ +public: + QRhiScissor() = default; + QRhiScissor(int x, int y, int w, int h); + + std::array<int, 4> scissor() const { return m_rect; } + void setScissor(int x, int y, int w, int h) { + m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + } + +private: + std::array<int, 4> m_rect { { 0, 0, 0, 0 } }; + + friend bool operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept + { + return a.m_rect == b.m_rect; + } + + friend bool operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiScissor &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_rect[0]); + seed = hash(seed, v.m_rect[1]); + seed = hash(seed, v.m_rect[2]); + seed = hash(seed, v.m_rect[3]); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiScissor, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiScissor &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputBinding +{ +public: + enum Classification { + PerVertex, + PerInstance + }; + + QRhiVertexInputBinding() = default; + QRhiVertexInputBinding(quint32 stride, Classification cls = PerVertex, quint32 stepRate = 1); + + quint32 stride() const { return m_stride; } + void setStride(quint32 s) { m_stride = s; } + + Classification classification() const { return m_classification; } + void setClassification(Classification c) { m_classification = c; } + + quint32 instanceStepRate() const { return m_instanceStepRate; } + void setInstanceStepRate(quint32 rate) { m_instanceStepRate = rate; } + +private: + quint32 m_stride = 0; + Classification m_classification = PerVertex; + quint32 m_instanceStepRate = 1; + + friend bool operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + { + return a.m_stride == b.m_stride + && a.m_classification == b.m_classification + && a.m_instanceStepRate == b.m_instanceStepRate; + } + + friend bool operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputBinding &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_stride); + seed = hash(seed, v.m_classification); + seed = hash(seed, v.m_instanceStepRate); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiVertexInputBinding, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputBinding &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputAttribute +{ +public: + enum Format { + Float4, + Float3, + Float2, + Float, + UNormByte4, + UNormByte2, + UNormByte, + UInt4, + UInt3, + UInt2, + UInt, + SInt4, + SInt3, + SInt2, + SInt, + Half4, + Half3, + Half2, + Half, + UShort4, + UShort3, + UShort2, + UShort, + SShort4, + SShort3, + SShort2, + SShort, + }; + + QRhiVertexInputAttribute() = default; + QRhiVertexInputAttribute(int binding, int location, Format format, quint32 offset, int matrixSlice = -1); + + int binding() const { return m_binding; } + void setBinding(int b) { m_binding = b; } + + int location() const { return m_location; } + void setLocation(int loc) { m_location = loc; } + + Format format() const { return m_format; } + void setFormat(Format f) { m_format = f; } + + quint32 offset() const { return m_offset; } + void setOffset(quint32 ofs) { m_offset = ofs; } + + int matrixSlice() const { return m_matrixSlice; } + void setMatrixSlice(int slice) { m_matrixSlice = slice; } + +private: + int m_binding = 0; + int m_location = 0; + Format m_format = Float4; + quint32 m_offset = 0; + int m_matrixSlice = -1; + + friend bool operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + { + return a.m_binding == b.m_binding + && a.m_location == b.m_location + && a.m_format == b.m_format + && a.m_offset == b.m_offset; + // matrixSlice excluded intentionally + } + + friend bool operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputAttribute &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_binding); + seed = hash(seed, v.m_location); + seed = hash(seed, v.m_format); + seed = hash(seed, v.m_offset); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiVertexInputAttribute, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputAttribute &); +#endif + +class Q_GUI_EXPORT QRhiVertexInputLayout +{ +public: + QRhiVertexInputLayout() = default; + + void setBindings(std::initializer_list<QRhiVertexInputBinding> list) { m_bindings = list; } + template<typename InputIterator> + void setBindings(InputIterator first, InputIterator last) + { + m_bindings.clear(); + std::copy(first, last, std::back_inserter(m_bindings)); + } + const QRhiVertexInputBinding *cbeginBindings() const { return m_bindings.cbegin(); } + const QRhiVertexInputBinding *cendBindings() const { return m_bindings.cend(); } + const QRhiVertexInputBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } + qsizetype bindingCount() const { return m_bindings.count(); } + + void setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) { m_attributes = list; } + template<typename InputIterator> + void setAttributes(InputIterator first, InputIterator last) + { + m_attributes.clear(); + std::copy(first, last, std::back_inserter(m_attributes)); + } + const QRhiVertexInputAttribute *cbeginAttributes() const { return m_attributes.cbegin(); } + const QRhiVertexInputAttribute *cendAttributes() const { return m_attributes.cend(); } + const QRhiVertexInputAttribute *attributeAt(qsizetype index) const { return &m_attributes.at(index); } + qsizetype attributeCount() const { return m_attributes.count(); } + +private: + QVarLengthArray<QRhiVertexInputBinding, 8> m_bindings; + QVarLengthArray<QRhiVertexInputAttribute, 8> m_attributes; + + friend bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + { + return a.m_bindings == b.m_bindings && a.m_attributes == b.m_attributes; + } + + friend bool operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiVertexInputLayout &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_bindings); + seed = hash(seed, v.m_attributes); + return seed; + } + + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); +}; + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); +#endif + +class Q_GUI_EXPORT QRhiShaderStage +{ +public: + enum Type { + Vertex, + TessellationControl, + TessellationEvaluation, + Geometry, + Fragment, + Compute + }; + + QRhiShaderStage() = default; + QRhiShaderStage(Type type, const QShader &shader, + QShader::Variant v = QShader::StandardShader); + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + QShader shader() const { return m_shader; } + void setShader(const QShader &s) { m_shader = s; } + + QShader::Variant shaderVariant() const { return m_shaderVariant; } + void setShaderVariant(QShader::Variant v) { m_shaderVariant = v; } + +private: + Type m_type = Vertex; + QShader m_shader; + QShader::Variant m_shaderVariant = QShader::StandardShader; + + friend bool operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + { + return a.m_type == b.m_type + && a.m_shader == b.m_shader + && a.m_shaderVariant == b.m_shaderVariant; + } + + friend bool operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + { + return !(a == b); + } + + friend size_t qHash(const QRhiShaderStage &v, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, v.m_type); + seed = hash(seed, v.m_shader); + seed = hash(seed, v.m_shaderVariant); + return seed; + } +}; + +Q_DECLARE_TYPEINFO(QRhiShaderStage, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderStage &); +#endif + +using QRhiGraphicsShaderStage = QRhiShaderStage; + +class Q_GUI_EXPORT QRhiShaderResourceBinding +{ +public: + enum Type { + UniformBuffer, + SampledTexture, + Texture, + Sampler, + ImageLoad, + ImageStore, + ImageLoadStore, + BufferLoad, + BufferStore, + BufferLoadStore + }; + + enum StageFlag { + VertexStage = 1 << 0, + TessellationControlStage = 1 << 1, + TessellationEvaluationStage = 1 << 2, + GeometryStage = 1 << 3, + FragmentStage = 1 << 4, + ComputeStage = 1 << 5 + }; + Q_DECLARE_FLAGS(StageFlags, StageFlag) + + QRhiShaderResourceBinding() = default; + + bool isLayoutCompatible(const QRhiShaderResourceBinding &other) const; + + static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding uniformBufferWithDynamicOffset(int binding, StageFlags stage, QRhiBuffer *buf, quint32 size); + + static QRhiShaderResourceBinding sampledTexture(int binding, StageFlags stage, QRhiTexture *tex, QRhiSampler *sampler); + + struct TextureAndSampler { + QRhiTexture *tex; + QRhiSampler *sampler; + }; + static QRhiShaderResourceBinding sampledTextures(int binding, StageFlags stage, int count, const TextureAndSampler *texSamplers); + + static QRhiShaderResourceBinding texture(int binding, StageFlags stage, QRhiTexture *tex); + static QRhiShaderResourceBinding textures(int binding, StageFlags stage, int count, QRhiTexture **tex); + static QRhiShaderResourceBinding sampler(int binding, StageFlags stage, QRhiSampler *sampler); + + static QRhiShaderResourceBinding imageLoad(int binding, StageFlags stage, QRhiTexture *tex, int level); + static QRhiShaderResourceBinding imageStore(int binding, StageFlags stage, QRhiTexture *tex, int level); + static QRhiShaderResourceBinding imageLoadStore(int binding, StageFlags stage, QRhiTexture *tex, int level); + + static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf); + static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); + + struct Data + { + int binding; + QRhiShaderResourceBinding::StageFlags stage; + QRhiShaderResourceBinding::Type type; + struct UniformBufferData { + QRhiBuffer *buf; + quint32 offset; + quint32 maybeSize; + bool hasDynamicOffset; + }; + static constexpr int MAX_TEX_SAMPLER_ARRAY_SIZE = 16; + struct TextureAndOrSamplerData { + int count; + TextureAndSampler texSamplers[MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct StorageImageData { + QRhiTexture *tex; + int level; + }; + struct StorageBufferData { + QRhiBuffer *buf; + quint32 offset; + quint32 maybeSize; + }; + union { + UniformBufferData ubuf; + TextureAndOrSamplerData stex; + StorageImageData simage; + StorageBufferData sbuf; + } u; + + int arraySize() const + { + return type == QRhiShaderResourceBinding::SampledTexture || type == QRhiShaderResourceBinding::Texture + ? u.stex.count + : 1; + } + + template<typename Output> + Output serialize(Output dst) const + { + // must write out exactly LAYOUT_DESC_ENTRIES_PER_BINDING elements here + *dst++ = quint32(binding); + *dst++ = quint32(stage); + *dst++ = quint32(type); + *dst++ = quint32(arraySize()); + return dst; + } + }; + + static constexpr int LAYOUT_DESC_ENTRIES_PER_BINDING = 4; + + template<typename Output> + static void serializeLayoutDescription(const QRhiShaderResourceBinding *first, + const QRhiShaderResourceBinding *last, + Output dst) + { + while (first != last) { + dst = first->d.serialize(dst); + ++first; + } + } + +private: + Data d; + friend class QRhiImplementation; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBinding::StageFlags) + +Q_DECLARE_TYPEINFO(QRhiShaderResourceBinding, Q_PRIMITIVE_TYPE); + +Q_GUI_EXPORT bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; +Q_GUI_EXPORT bool operator!=(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; +Q_GUI_EXPORT size_t qHash(const QRhiShaderResourceBinding &b, size_t seed = 0) noexcept; +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBinding &); +#endif + +class Q_GUI_EXPORT QRhiColorAttachment +{ +public: + QRhiColorAttachment() = default; + QRhiColorAttachment(QRhiTexture *texture); + QRhiColorAttachment(QRhiRenderBuffer *renderBuffer); + + QRhiTexture *texture() const { return m_texture; } + void setTexture(QRhiTexture *tex) { m_texture = tex; } + + QRhiRenderBuffer *renderBuffer() const { return m_renderBuffer; } + void setRenderBuffer(QRhiRenderBuffer *rb) { m_renderBuffer = rb; } + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + + QRhiTexture *resolveTexture() const { return m_resolveTexture; } + void setResolveTexture(QRhiTexture *tex) { m_resolveTexture = tex; } + + int resolveLayer() const { return m_resolveLayer; } + void setResolveLayer(int layer) { m_resolveLayer = layer; } + + int resolveLevel() const { return m_resolveLevel; } + void setResolveLevel(int level) { m_resolveLevel = level; } + + int multiViewCount() const { return m_multiViewCount; } + void setMultiViewCount(int count) { m_multiViewCount = count; } + +private: + QRhiTexture *m_texture = nullptr; + QRhiRenderBuffer *m_renderBuffer = nullptr; + int m_layer = 0; + int m_level = 0; + QRhiTexture *m_resolveTexture = nullptr; + int m_resolveLayer = 0; + int m_resolveLevel = 0; + int m_multiViewCount = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiColorAttachment, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureRenderTargetDescription +{ +public: + QRhiTextureRenderTargetDescription() = default; + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment); + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiRenderBuffer *depthStencilBuffer); + QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiTexture *depthTexture); + + void setColorAttachments(std::initializer_list<QRhiColorAttachment> list) { m_colorAttachments = list; } + template<typename InputIterator> + void setColorAttachments(InputIterator first, InputIterator last) + { + m_colorAttachments.clear(); + std::copy(first, last, std::back_inserter(m_colorAttachments)); + } + const QRhiColorAttachment *cbeginColorAttachments() const { return m_colorAttachments.cbegin(); } + const QRhiColorAttachment *cendColorAttachments() const { return m_colorAttachments.cend(); } + const QRhiColorAttachment *colorAttachmentAt(qsizetype index) const { return &m_colorAttachments.at(index); } + qsizetype colorAttachmentCount() const { return m_colorAttachments.count(); } + + QRhiRenderBuffer *depthStencilBuffer() const { return m_depthStencilBuffer; } + void setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) { m_depthStencilBuffer = renderBuffer; } + + QRhiTexture *depthTexture() const { return m_depthTexture; } + void setDepthTexture(QRhiTexture *texture) { m_depthTexture = texture; } + + QRhiTexture *depthResolveTexture() const { return m_depthResolveTexture; } + void setDepthResolveTexture(QRhiTexture *tex) { m_depthResolveTexture = tex; } + +private: + QVarLengthArray<QRhiColorAttachment, 8> m_colorAttachments; + QRhiRenderBuffer *m_depthStencilBuffer = nullptr; + QRhiTexture *m_depthTexture = nullptr; + QRhiTexture *m_depthResolveTexture = nullptr; +}; + +class Q_GUI_EXPORT QRhiTextureSubresourceUploadDescription +{ +public: + QRhiTextureSubresourceUploadDescription() = default; + explicit QRhiTextureSubresourceUploadDescription(const QImage &image); + QRhiTextureSubresourceUploadDescription(const void *data, quint32 size); + explicit QRhiTextureSubresourceUploadDescription(const QByteArray &data); + + QImage image() const { return m_image; } + void setImage(const QImage &image) { m_image = image; } + + QByteArray data() const { return m_data; } + void setData(const QByteArray &data) { m_data = data; } + + quint32 dataStride() const { return m_dataStride; } + void setDataStride(quint32 stride) { m_dataStride = stride; } + + QPoint destinationTopLeft() const { return m_destinationTopLeft; } + void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } + + QSize sourceSize() const { return m_sourceSize; } + void setSourceSize(const QSize &size) { m_sourceSize = size; } + + QPoint sourceTopLeft() const { return m_sourceTopLeft; } + void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } + +private: + QImage m_image; + QByteArray m_data; + quint32 m_dataStride = 0; + QPoint m_destinationTopLeft; + QSize m_sourceSize; + QPoint m_sourceTopLeft; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureSubresourceUploadDescription, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureUploadEntry +{ +public: + QRhiTextureUploadEntry() = default; + QRhiTextureUploadEntry(int layer, int level, const QRhiTextureSubresourceUploadDescription &desc); + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + + QRhiTextureSubresourceUploadDescription description() const { return m_desc; } + void setDescription(const QRhiTextureSubresourceUploadDescription &desc) { m_desc = desc; } + +private: + int m_layer = 0; + int m_level = 0; + QRhiTextureSubresourceUploadDescription m_desc; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureUploadEntry, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiTextureUploadDescription +{ +public: + QRhiTextureUploadDescription() = default; + QRhiTextureUploadDescription(const QRhiTextureUploadEntry &entry); + QRhiTextureUploadDescription(std::initializer_list<QRhiTextureUploadEntry> list); + + void setEntries(std::initializer_list<QRhiTextureUploadEntry> list) { m_entries = list; } + template<typename InputIterator> + void setEntries(InputIterator first, InputIterator last) + { + m_entries.clear(); + std::copy(first, last, std::back_inserter(m_entries)); + } + const QRhiTextureUploadEntry *cbeginEntries() const { return m_entries.cbegin(); } + const QRhiTextureUploadEntry *cendEntries() const { return m_entries.cend(); } + const QRhiTextureUploadEntry *entryAt(qsizetype index) const { return &m_entries.at(index); } + qsizetype entryCount() const { return m_entries.count(); } + +private: + QVarLengthArray<QRhiTextureUploadEntry, 16> m_entries; +}; + +class Q_GUI_EXPORT QRhiTextureCopyDescription +{ +public: + QRhiTextureCopyDescription() = default; + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int sourceLayer() const { return m_sourceLayer; } + void setSourceLayer(int layer) { m_sourceLayer = layer; } + + int sourceLevel() const { return m_sourceLevel; } + void setSourceLevel(int level) { m_sourceLevel = level; } + + QPoint sourceTopLeft() const { return m_sourceTopLeft; } + void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } + + int destinationLayer() const { return m_destinationLayer; } + void setDestinationLayer(int layer) { m_destinationLayer = layer; } + + int destinationLevel() const { return m_destinationLevel; } + void setDestinationLevel(int level) { m_destinationLevel = level; } + + QPoint destinationTopLeft() const { return m_destinationTopLeft; } + void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } + +private: + QSize m_pixelSize; + int m_sourceLayer = 0; + int m_sourceLevel = 0; + QPoint m_sourceTopLeft; + int m_destinationLayer = 0; + int m_destinationLevel = 0; + QPoint m_destinationTopLeft; +}; + +Q_DECLARE_TYPEINFO(QRhiTextureCopyDescription, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QRhiReadbackDescription +{ +public: + QRhiReadbackDescription() = default; + QRhiReadbackDescription(QRhiTexture *texture); + + QRhiTexture *texture() const { return m_texture; } + void setTexture(QRhiTexture *tex) { m_texture = tex; } + + int layer() const { return m_layer; } + void setLayer(int layer) { m_layer = layer; } + + int level() const { return m_level; } + void setLevel(int level) { m_level = level; } + +private: + QRhiTexture *m_texture = nullptr; + int m_layer = 0; + int m_level = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiReadbackDescription, Q_RELOCATABLE_TYPE); + +struct Q_GUI_EXPORT QRhiNativeHandles +{ +}; + +class Q_GUI_EXPORT QRhiResource +{ +public: + enum Type { + Buffer, + Texture, + Sampler, + RenderBuffer, + RenderPassDescriptor, + SwapChainRenderTarget, + TextureRenderTarget, + ShaderResourceBindings, + GraphicsPipeline, + SwapChain, + ComputePipeline, + CommandBuffer + }; + + virtual ~QRhiResource(); + + virtual Type resourceType() const = 0; + + virtual void destroy() = 0; + + void deleteLater(); + + QByteArray name() const; + void setName(const QByteArray &name); + + quint64 globalResourceId() const; + + QRhi *rhi() const; + +protected: + QRhiResource(QRhiImplementation *rhi); + Q_DISABLE_COPY(QRhiResource) + friend class QRhiImplementation; + QRhiImplementation *m_rhi = nullptr; + quint64 m_id; + QByteArray m_objectName; +}; + +class Q_GUI_EXPORT QRhiBuffer : public QRhiResource +{ +public: + enum Type { + Immutable, + Static, + Dynamic + }; + + enum UsageFlag { + VertexBuffer = 1 << 0, + IndexBuffer = 1 << 1, + UniformBuffer = 1 << 2, + StorageBuffer = 1 << 3 + }; + Q_DECLARE_FLAGS(UsageFlags, UsageFlag) + + struct NativeBuffer { + const void *objects[3]; + int slotCount; + }; + + QRhiResource::Type resourceType() const override; + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + UsageFlags usage() const { return m_usage; } + void setUsage(UsageFlags u) { m_usage = u; } + + quint32 size() const { return m_size; } + void setSize(quint32 sz) { m_size = sz; } + + virtual bool create() = 0; + + virtual NativeBuffer nativeBuffer(); + + virtual char *beginFullDynamicBufferUpdateForCurrentFrame(); + virtual void endFullDynamicBufferUpdateForCurrentFrame(); + +protected: + QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, quint32 size_); + Type m_type; + UsageFlags m_usage; + quint32 m_size; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiBuffer::UsageFlags) + +class Q_GUI_EXPORT QRhiTexture : public QRhiResource +{ +public: + enum Flag { + RenderTarget = 1 << 0, + CubeMap = 1 << 2, + MipMapped = 1 << 3, + sRGB = 1 << 4, + UsedAsTransferSource = 1 << 5, + UsedWithGenerateMips = 1 << 6, + UsedWithLoadStore = 1 << 7, + UsedAsCompressedAtlas = 1 << 8, + ExternalOES = 1 << 9, + ThreeDimensional = 1 << 10, + TextureRectangleGL = 1 << 11, + TextureArray = 1 << 12, + OneDimensional = 1 << 13 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Format { + UnknownFormat, + + RGBA8, + BGRA8, + R8, + RG8, + R16, + RG16, + RED_OR_ALPHA8, + + RGBA16F, + RGBA32F, + R16F, + R32F, + + RGB10A2, + + D16, + D24, + D24S8, + D32F, + + BC1, + BC2, + BC3, + BC4, + BC5, + BC6H, + BC7, + + ETC2_RGB8, + ETC2_RGB8A1, + ETC2_RGBA8, + + ASTC_4x4, + ASTC_5x4, + ASTC_5x5, + ASTC_6x5, + ASTC_6x6, + ASTC_8x5, + ASTC_8x6, + ASTC_8x8, + ASTC_10x5, + ASTC_10x6, + ASTC_10x8, + ASTC_10x10, + ASTC_12x10, + ASTC_12x12 + }; + + struct NativeTexture { + quint64 object; + int layout; // or state + }; + + QRhiResource::Type resourceType() const override; + + Format format() const { return m_format; } + void setFormat(Format fmt) { m_format = fmt; } + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int depth() const { return m_depth; } + void setDepth(int depth) { m_depth = depth; } + + int arraySize() const { return m_arraySize; } + void setArraySize(int arraySize) { m_arraySize = arraySize; } + + int arrayRangeStart() const { return m_arrayRangeStart; } + int arrayRangeLength() const { return m_arrayRangeLength; } + void setArrayRange(int startIndex, int count) + { + m_arrayRangeStart = startIndex; + m_arrayRangeLength = count; + } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + struct ViewFormat { + QRhiTexture::Format format; + bool srgb; + }; + ViewFormat readViewFormat() const { return m_readViewFormat; } + void setReadViewFormat(const ViewFormat &fmt) { m_readViewFormat = fmt; } + ViewFormat writeViewFormat() const { return m_writeViewFormat; } + void setWriteViewFormat(const ViewFormat &fmt) { m_writeViewFormat = fmt; } + + virtual bool create() = 0; + virtual NativeTexture nativeTexture(); + virtual bool createFrom(NativeTexture src); + virtual void setNativeLayout(int layout); + +protected: + QRhiTexture(QRhiImplementation *rhi, Format format_, const QSize &pixelSize_, int depth_, + int arraySize_, int sampleCount_, Flags flags_); + Format m_format; + QSize m_pixelSize; + int m_depth; + int m_arraySize; + int m_sampleCount; + Flags m_flags; + int m_arrayRangeStart = -1; + int m_arrayRangeLength = -1; + ViewFormat m_readViewFormat = { UnknownFormat, false }; + ViewFormat m_writeViewFormat = { UnknownFormat, false }; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTexture::Flags) + +class Q_GUI_EXPORT QRhiSampler : public QRhiResource +{ +public: + enum Filter { + None, + Nearest, + Linear + }; + + enum AddressMode { + Repeat, + ClampToEdge, + Mirror, + }; + + enum CompareOp { + Never, + Less, + Equal, + LessOrEqual, + Greater, + NotEqual, + GreaterOrEqual, + Always + }; + + QRhiResource::Type resourceType() const override; + + Filter magFilter() const { return m_magFilter; } + void setMagFilter(Filter f) { m_magFilter = f; } + + Filter minFilter() const { return m_minFilter; } + void setMinFilter(Filter f) { m_minFilter = f; } + + Filter mipmapMode() const { return m_mipmapMode; } + void setMipmapMode(Filter f) { m_mipmapMode = f; } + + AddressMode addressU() const { return m_addressU; } + void setAddressU(AddressMode mode) { m_addressU = mode; } + + AddressMode addressV() const { return m_addressV; } + void setAddressV(AddressMode mode) { m_addressV = mode; } + + AddressMode addressW() const { return m_addressW; } + void setAddressW(AddressMode mode) { m_addressW = mode; } + + CompareOp textureCompareOp() const { return m_compareOp; } + void setTextureCompareOp(CompareOp op) { m_compareOp = op; } + + virtual bool create() = 0; + +protected: + QRhiSampler(QRhiImplementation *rhi, + Filter magFilter_, Filter minFilter_, Filter mipmapMode_, + AddressMode u_, AddressMode v_, AddressMode w_); + Filter m_magFilter; + Filter m_minFilter; + Filter m_mipmapMode; + AddressMode m_addressU; + AddressMode m_addressV; + AddressMode m_addressW; + CompareOp m_compareOp; +}; + +class Q_GUI_EXPORT QRhiRenderBuffer : public QRhiResource +{ +public: + enum Type { + DepthStencil, + Color + }; + + enum Flag { + UsedWithSwapChainOnly = 1 << 0 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + struct NativeRenderBuffer { + quint64 object; + }; + + QRhiResource::Type resourceType() const override; + + Type type() const { return m_type; } + void setType(Type t) { m_type = t; } + + QSize pixelSize() const { return m_pixelSize; } + void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + virtual bool create() = 0; + virtual bool createFrom(NativeRenderBuffer src); + + virtual QRhiTexture::Format backingFormat() const = 0; + +protected: + QRhiRenderBuffer(QRhiImplementation *rhi, Type type_, const QSize &pixelSize_, + int sampleCount_, Flags flags_, QRhiTexture::Format backingFormatHint_); + Type m_type; + QSize m_pixelSize; + int m_sampleCount; + Flags m_flags; + QRhiTexture::Format m_backingFormatHint; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiRenderBuffer::Flags) + +class Q_GUI_EXPORT QRhiRenderPassDescriptor : public QRhiResource +{ +public: + QRhiResource::Type resourceType() const override; + + virtual bool isCompatible(const QRhiRenderPassDescriptor *other) const = 0; + virtual const QRhiNativeHandles *nativeHandles(); + + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const = 0; + + virtual QVector<quint32> serializedFormat() const = 0; + +protected: + QRhiRenderPassDescriptor(QRhiImplementation *rhi); +}; + +class Q_GUI_EXPORT QRhiRenderTarget : public QRhiResource +{ +public: + virtual QSize pixelSize() const = 0; + virtual float devicePixelRatio() const = 0; + virtual int sampleCount() const = 0; + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + +protected: + QRhiRenderTarget(QRhiImplementation *rhi); + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; +}; + +class Q_GUI_EXPORT QRhiSwapChainRenderTarget : public QRhiRenderTarget +{ +public: + QRhiResource::Type resourceType() const override; + QRhiSwapChain *swapChain() const { return m_swapchain; } + +protected: + QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain_); + QRhiSwapChain *m_swapchain; +}; + +class Q_GUI_EXPORT QRhiTextureRenderTarget : public QRhiRenderTarget +{ +public: + enum Flag { + PreserveColorContents = 1 << 0, + PreserveDepthStencilContents = 1 << 1, + DoNotStoreDepthStencilContents = 1 << 2 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QRhiResource::Type resourceType() const override; + + QRhiTextureRenderTargetDescription description() const { return m_desc; } + void setDescription(const QRhiTextureRenderTargetDescription &desc) { m_desc = desc; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; + + virtual bool create() = 0; + +protected: + QRhiTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc_, Flags flags_); + QRhiTextureRenderTargetDescription m_desc; + Flags m_flags; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTextureRenderTarget::Flags) + +class Q_GUI_EXPORT QRhiShaderResourceBindings : public QRhiResource +{ +public: + QRhiResource::Type resourceType() const override; + + void setBindings(std::initializer_list<QRhiShaderResourceBinding> list) { m_bindings = list; } + template<typename InputIterator> + void setBindings(InputIterator first, InputIterator last) + { + m_bindings.clear(); + std::copy(first, last, std::back_inserter(m_bindings)); + } + const QRhiShaderResourceBinding *cbeginBindings() const { return m_bindings.cbegin(); } + const QRhiShaderResourceBinding *cendBindings() const { return m_bindings.cend(); } + const QRhiShaderResourceBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } + qsizetype bindingCount() const { return m_bindings.count(); } + + bool isLayoutCompatible(const QRhiShaderResourceBindings *other) const; + + QVector<quint32> serializedLayoutDescription() const { return m_layoutDesc; } + + virtual bool create() = 0; + + enum UpdateFlag { + BindingsAreSorted = 0x01 + }; + Q_DECLARE_FLAGS(UpdateFlags, UpdateFlag) + + virtual void updateResources(UpdateFlags flags = {}) = 0; + +protected: + static constexpr int BINDING_PREALLOC = 12; + QRhiShaderResourceBindings(QRhiImplementation *rhi); + QVarLengthArray<QRhiShaderResourceBinding, BINDING_PREALLOC> m_bindings; + size_t m_layoutDescHash = 0; + // Intentionally not using QVLA for m_layoutDesc: clients like Qt Quick are much + // better served with an implicitly shared container here, because they will likely + // throw this directly into structs serving as cache keys. + QVector<quint32> m_layoutDesc; + friend class QRhiImplementation; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); +#endif +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBindings::UpdateFlags) + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); +#endif + +// The proper name. Until it gets rolled out universally, have the better name +// as a typedef. Eventually it should be reversed (the old name being a typedef +// to the new one). +using QRhiShaderResourceBindingSet = QRhiShaderResourceBindings; + +class Q_GUI_EXPORT QRhiGraphicsPipeline : public QRhiResource +{ +public: + enum Flag { + UsesBlendConstants = 1 << 0, + UsesStencilRef = 1 << 1, + UsesScissor = 1 << 2, + CompileShadersWithDebugInfo = 1 << 3 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Topology { + Triangles, + TriangleStrip, + TriangleFan, + Lines, + LineStrip, + Points, + Patches + }; + + enum CullMode { + None, + Front, + Back + }; + + enum FrontFace { + CCW, + CW + }; + + enum ColorMaskComponent { + R = 1 << 0, + G = 1 << 1, + B = 1 << 2, + A = 1 << 3 + }; + Q_DECLARE_FLAGS(ColorMask, ColorMaskComponent) + + enum BlendFactor { + Zero, + One, + SrcColor, + OneMinusSrcColor, + DstColor, + OneMinusDstColor, + SrcAlpha, + OneMinusSrcAlpha, + DstAlpha, + OneMinusDstAlpha, + ConstantColor, + OneMinusConstantColor, + ConstantAlpha, + OneMinusConstantAlpha, + SrcAlphaSaturate, + Src1Color, + OneMinusSrc1Color, + Src1Alpha, + OneMinusSrc1Alpha + }; + + enum BlendOp { + Add, + Subtract, + ReverseSubtract, + Min, + Max + }; + + struct TargetBlend { + ColorMask colorWrite = ColorMask(0xF); // R | G | B | A + bool enable = false; + BlendFactor srcColor = One; + BlendFactor dstColor = OneMinusSrcAlpha; + BlendOp opColor = Add; + BlendFactor srcAlpha = One; + BlendFactor dstAlpha = OneMinusSrcAlpha; + BlendOp opAlpha = Add; + }; + + enum CompareOp { + Never, + Less, + Equal, + LessOrEqual, + Greater, + NotEqual, + GreaterOrEqual, + Always + }; + + enum StencilOp { + StencilZero, + Keep, + Replace, + IncrementAndClamp, + DecrementAndClamp, + Invert, + IncrementAndWrap, + DecrementAndWrap + }; + + struct StencilOpState { + StencilOp failOp = Keep; + StencilOp depthFailOp = Keep; + StencilOp passOp = Keep; + CompareOp compareOp = Always; + }; + + enum PolygonMode { + Fill, + Line + }; + + QRhiResource::Type resourceType() const override; + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + Topology topology() const { return m_topology; } + void setTopology(Topology t) { m_topology = t; } + + CullMode cullMode() const { return m_cullMode; } + void setCullMode(CullMode mode) { m_cullMode = mode; } + + FrontFace frontFace() const { return m_frontFace; } + void setFrontFace(FrontFace f) { m_frontFace = f; } + + void setTargetBlends(std::initializer_list<TargetBlend> list) { m_targetBlends = list; } + template<typename InputIterator> + void setTargetBlends(InputIterator first, InputIterator last) + { + m_targetBlends.clear(); + std::copy(first, last, std::back_inserter(m_targetBlends)); + } + const TargetBlend *cbeginTargetBlends() const { return m_targetBlends.cbegin(); } + const TargetBlend *cendTargetBlends() const { return m_targetBlends.cend(); } + const TargetBlend *targetBlendAt(qsizetype index) const { return &m_targetBlends.at(index); } + qsizetype targetBlendCount() const { return m_targetBlends.count(); } + + bool hasDepthTest() const { return m_depthTest; } + void setDepthTest(bool enable) { m_depthTest = enable; } + + bool hasDepthWrite() const { return m_depthWrite; } + void setDepthWrite(bool enable) { m_depthWrite = enable; } + + CompareOp depthOp() const { return m_depthOp; } + void setDepthOp(CompareOp op) { m_depthOp = op; } + + bool hasStencilTest() const { return m_stencilTest; } + void setStencilTest(bool enable) { m_stencilTest = enable; } + + StencilOpState stencilFront() const { return m_stencilFront; } + void setStencilFront(const StencilOpState &state) { m_stencilFront = state; } + + StencilOpState stencilBack() const { return m_stencilBack; } + void setStencilBack(const StencilOpState &state) { m_stencilBack = state; } + + quint32 stencilReadMask() const { return m_stencilReadMask; } + void setStencilReadMask(quint32 mask) { m_stencilReadMask = mask; } + + quint32 stencilWriteMask() const { return m_stencilWriteMask; } + void setStencilWriteMask(quint32 mask) { m_stencilWriteMask = mask; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int s) { m_sampleCount = s; } + + float lineWidth() const { return m_lineWidth; } + void setLineWidth(float width) { m_lineWidth = width; } + + int depthBias() const { return m_depthBias; } + void setDepthBias(int bias) { m_depthBias = bias; } + + float slopeScaledDepthBias() const { return m_slopeScaledDepthBias; } + void setSlopeScaledDepthBias(float bias) { m_slopeScaledDepthBias = bias; } + + void setShaderStages(std::initializer_list<QRhiShaderStage> list) { m_shaderStages = list; } + template<typename InputIterator> + void setShaderStages(InputIterator first, InputIterator last) + { + m_shaderStages.clear(); + std::copy(first, last, std::back_inserter(m_shaderStages)); + } + const QRhiShaderStage *cbeginShaderStages() const { return m_shaderStages.cbegin(); } + const QRhiShaderStage *cendShaderStages() const { return m_shaderStages.cend(); } + const QRhiShaderStage *shaderStageAt(qsizetype index) const { return &m_shaderStages.at(index); } + qsizetype shaderStageCount() const { return m_shaderStages.count(); } + + QRhiVertexInputLayout vertexInputLayout() const { return m_vertexInputLayout; } + void setVertexInputLayout(const QRhiVertexInputLayout &layout) { m_vertexInputLayout = layout; } + + QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } + void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + + int patchControlPointCount() const { return m_patchControlPointCount; } + void setPatchControlPointCount(int count) { m_patchControlPointCount = count; } + + PolygonMode polygonMode() const {return m_polygonMode; } + void setPolygonMode(PolygonMode mode) {m_polygonMode = mode; } + + int multiViewCount() const { return m_multiViewCount; } + void setMultiViewCount(int count) { m_multiViewCount = count; } + + virtual bool create() = 0; + +protected: + QRhiGraphicsPipeline(QRhiImplementation *rhi); + Flags m_flags; + Topology m_topology = Triangles; + CullMode m_cullMode = None; + FrontFace m_frontFace = CCW; + QVarLengthArray<TargetBlend, 8> m_targetBlends; + bool m_depthTest = false; + bool m_depthWrite = false; + CompareOp m_depthOp = Less; + bool m_stencilTest = false; + StencilOpState m_stencilFront; + StencilOpState m_stencilBack; + quint32 m_stencilReadMask = 0xFF; + quint32 m_stencilWriteMask = 0xFF; + int m_sampleCount = 1; + float m_lineWidth = 1.0f; + int m_depthBias = 0; + float m_slopeScaledDepthBias = 0.0f; + int m_patchControlPointCount = 3; + PolygonMode m_polygonMode = Fill; + int m_multiViewCount = 0; + QVarLengthArray<QRhiShaderStage, 4> m_shaderStages; + QRhiVertexInputLayout m_vertexInputLayout; + QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::Flags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::ColorMask) +Q_DECLARE_TYPEINFO(QRhiGraphicsPipeline::TargetBlend, Q_RELOCATABLE_TYPE); + +struct QRhiSwapChainHdrInfo +{ + enum LimitsType { + LuminanceInNits, + ColorComponentValue + }; + + enum LuminanceBehavior { + SceneReferred, + DisplayReferred + }; + + LimitsType limitsType; + union { + struct { + float minLuminance; + float maxLuminance; + } luminanceInNits; + struct { + float maxColorComponentValue; + float maxPotentialColorComponentValue; + } colorComponentValue; + } limits; + LuminanceBehavior luminanceBehavior; + float sdrWhiteLevel; +}; + +Q_DECLARE_TYPEINFO(QRhiSwapChainHdrInfo, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiSwapChainHdrInfo &); +#endif + +struct QRhiSwapChainProxyData +{ + void *reserved[2] = {}; +}; + +class Q_GUI_EXPORT QRhiSwapChain : public QRhiResource +{ +public: + enum Flag { + SurfaceHasPreMulAlpha = 1 << 0, + SurfaceHasNonPreMulAlpha = 1 << 1, + sRGB = 1 << 2, + UsedAsTransferSource = 1 << 3, + NoVSync = 1 << 4, + MinimalBufferCount = 1 << 5 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum Format { + SDR, + HDRExtendedSrgbLinear, + HDR10, + HDRExtendedDisplayP3Linear + }; + + enum StereoTargetBuffer { + LeftBuffer, + RightBuffer + }; + + QRhiResource::Type resourceType() const override; + + QWindow *window() const { return m_window; } + void setWindow(QWindow *window) { m_window = window; } + + QRhiSwapChainProxyData proxyData() const { return m_proxyData; } + void setProxyData(const QRhiSwapChainProxyData &d) { m_proxyData = d; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + Format format() const { return m_format; } + void setFormat(Format f) { m_format = f; } + + QRhiRenderBuffer *depthStencil() const { return m_depthStencil; } + void setDepthStencil(QRhiRenderBuffer *ds) { m_depthStencil = ds; } + + int sampleCount() const { return m_sampleCount; } + void setSampleCount(int samples) { m_sampleCount = samples; } + + QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } + void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + + QSize currentPixelSize() const { return m_currentPixelSize; } + + virtual QRhiCommandBuffer *currentFrameCommandBuffer() = 0; + virtual QRhiRenderTarget *currentFrameRenderTarget() = 0; + virtual QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer); + virtual QSize surfacePixelSize() = 0; + virtual bool isFormatSupported(Format f) = 0; + virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; + virtual bool createOrResize() = 0; + virtual QRhiSwapChainHdrInfo hdrInfo(); + +protected: + QRhiSwapChain(QRhiImplementation *rhi); + QWindow *m_window = nullptr; + Flags m_flags; + Format m_format = SDR; + QRhiRenderBuffer *m_depthStencil = nullptr; + int m_sampleCount = 1; + QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; + QSize m_currentPixelSize; + QRhiSwapChainProxyData m_proxyData; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiSwapChain::Flags) + +class Q_GUI_EXPORT QRhiComputePipeline : public QRhiResource +{ +public: + enum Flag { + CompileShadersWithDebugInfo = 1 << 0 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QRhiResource::Type resourceType() const override; + virtual bool create() = 0; + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + + QRhiShaderStage shaderStage() const { return m_shaderStage; } + void setShaderStage(const QRhiShaderStage &stage) { m_shaderStage = stage; } + + QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } + void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } + +protected: + QRhiComputePipeline(QRhiImplementation *rhi); + Flags m_flags; + QRhiShaderStage m_shaderStage; + QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiComputePipeline::Flags) + +class Q_GUI_EXPORT QRhiCommandBuffer : public QRhiResource +{ +public: + enum IndexFormat { + IndexUInt16, + IndexUInt32 + }; + + enum BeginPassFlag { + ExternalContent = 0x01, + DoNotTrackResourcesForCompute = 0x02 + }; + Q_DECLARE_FLAGS(BeginPassFlags, BeginPassFlag) + + QRhiResource::Type resourceType() const override; + + void resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates); + + void beginPass(QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates = nullptr, + BeginPassFlags flags = {}); + void endPass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); + + void setGraphicsPipeline(QRhiGraphicsPipeline *ps); + using DynamicOffset = QPair<int, quint32>; // binding, offset + void setShaderResources(QRhiShaderResourceBindings *srb = nullptr, + int dynamicOffsetCount = 0, + const DynamicOffset *dynamicOffsets = nullptr); + using VertexInput = QPair<QRhiBuffer *, quint32>; // buffer, offset + void setVertexInput(int startBinding, int bindingCount, const VertexInput *bindings, + QRhiBuffer *indexBuf = nullptr, quint32 indexOffset = 0, + IndexFormat indexFormat = IndexUInt16); + + void setViewport(const QRhiViewport &viewport); + void setScissor(const QRhiScissor &scissor); + void setBlendConstants(const QColor &c); + void setStencilRef(quint32 refValue); + + void draw(quint32 vertexCount, + quint32 instanceCount = 1, + quint32 firstVertex = 0, + quint32 firstInstance = 0); + + void drawIndexed(quint32 indexCount, + quint32 instanceCount = 1, + quint32 firstIndex = 0, + qint32 vertexOffset = 0, + quint32 firstInstance = 0); + + void debugMarkBegin(const QByteArray &name); + void debugMarkEnd(); + void debugMarkMsg(const QByteArray &msg); + + void beginComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr, BeginPassFlags flags = {}); + void endComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); + void setComputePipeline(QRhiComputePipeline *ps); + void dispatch(int x, int y, int z); + + const QRhiNativeHandles *nativeHandles(); + void beginExternal(); + void endExternal(); + + double lastCompletedGpuTime(); + +protected: + QRhiCommandBuffer(QRhiImplementation *rhi); +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiCommandBuffer::BeginPassFlags) + +struct Q_GUI_EXPORT QRhiReadbackResult +{ + std::function<void()> completed = nullptr; + QRhiTexture::Format format; + QSize pixelSize; + QByteArray data; +}; + +class Q_GUI_EXPORT QRhiResourceUpdateBatch +{ +public: + ~QRhiResourceUpdateBatch(); + + void release(); + + void merge(QRhiResourceUpdateBatch *other); + bool hasOptimalCapacity() const; + + void updateDynamicBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); + void uploadStaticBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); + void uploadStaticBuffer(QRhiBuffer *buf, const void *data); + void readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result); + void uploadTexture(QRhiTexture *tex, const QRhiTextureUploadDescription &desc); + void uploadTexture(QRhiTexture *tex, const QImage &image); + void copyTexture(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc = QRhiTextureCopyDescription()); + void readBackTexture(const QRhiReadbackDescription &rb, QRhiReadbackResult *result); + void generateMips(QRhiTexture *tex); + +private: + QRhiResourceUpdateBatch(QRhiImplementation *rhi); + Q_DISABLE_COPY(QRhiResourceUpdateBatch) + QRhiResourceUpdateBatchPrivate *d; + friend class QRhiResourceUpdateBatchPrivate; + friend class QRhi; +}; + +struct Q_GUI_EXPORT QRhiDriverInfo +{ + enum DeviceType { + UnknownDevice, + IntegratedDevice, + DiscreteDevice, + ExternalDevice, + VirtualDevice, + CpuDevice + }; + + QByteArray deviceName; + quint64 deviceId = 0; + quint64 vendorId = 0; + DeviceType deviceType = UnknownDevice; +}; + +Q_DECLARE_TYPEINFO(QRhiDriverInfo, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDriverInfo &); +#endif + +struct Q_GUI_EXPORT QRhiStats +{ + qint64 totalPipelineCreationTime = 0; + // Vulkan or D3D12 memory allocator statistics + quint32 blockCount = 0; + quint32 allocCount = 0; + quint64 usedBytes = 0; + quint64 unusedBytes = 0; + // D3D12 only, from IDXGIAdapter3::QueryVideoMemoryInfo(), incl. all resources + quint64 totalUsageBytes = 0; +}; + +Q_DECLARE_TYPEINFO(QRhiStats, Q_RELOCATABLE_TYPE); + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiStats &); +#endif + +struct Q_GUI_EXPORT QRhiInitParams +{ +}; + +class Q_GUI_EXPORT QRhi +{ +public: + enum Implementation { + Null, + Vulkan, + OpenGLES2, + D3D11, + Metal, + D3D12 + }; + + enum Flag { + EnableDebugMarkers = 1 << 0, + PreferSoftwareRenderer = 1 << 1, + EnablePipelineCacheDataSave = 1 << 2, + EnableTimestamps = 1 << 3, + SuppressSmokeTestWarnings = 1 << 4 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + enum FrameOpResult { + FrameOpSuccess = 0, + FrameOpError, + FrameOpSwapChainOutOfDate, + FrameOpDeviceLost + }; + + enum Feature { + MultisampleTexture = 1, + MultisampleRenderBuffer, + DebugMarkers, + Timestamps, + Instancing, + CustomInstanceStepRate, + PrimitiveRestart, + NonDynamicUniformBuffers, + NonFourAlignedEffectiveIndexBufferOffset, + NPOTTextureRepeat, + RedOrAlpha8IsRed, + ElementIndexUint, + Compute, + WideLines, + VertexShaderPointSize, + BaseVertex, + BaseInstance, + TriangleFanTopology, + ReadBackNonUniformBuffer, + ReadBackNonBaseMipLevel, + TexelFetch, + RenderToNonBaseMipLevel, + IntAttributes, + ScreenSpaceDerivatives, + ReadBackAnyTextureFormat, + PipelineCacheDataLoadSave, + ImageDataStride, + RenderBufferImport, + ThreeDimensionalTextures, + RenderTo3DTextureSlice, + TextureArrays, + Tessellation, + GeometryShader, + TextureArrayRange, + NonFillPolygonMode, + OneDimensionalTextures, + OneDimensionalTextureMipmaps, + HalfAttributes, + RenderToOneDimensionalTexture, + ThreeDimensionalTextureMipmaps, + MultiView, + TextureViewFormat, + ResolveDepthStencil + }; + + enum BeginFrameFlag { + }; + Q_DECLARE_FLAGS(BeginFrameFlags, BeginFrameFlag) + + enum EndFrameFlag { + SkipPresent = 1 << 0 + }; + Q_DECLARE_FLAGS(EndFrameFlags, EndFrameFlag) + + enum ResourceLimit { + TextureSizeMin = 1, + TextureSizeMax, + MaxColorAttachments, + FramesInFlight, + MaxAsyncReadbackFrames, + MaxThreadGroupsPerDimension, + MaxThreadsPerThreadGroup, + MaxThreadGroupX, + MaxThreadGroupY, + MaxThreadGroupZ, + TextureArraySizeMax, + MaxUniformBufferRange, + MaxVertexInputs, + MaxVertexOutputs + }; + + ~QRhi(); + + static QRhi *create(Implementation impl, + QRhiInitParams *params, + Flags flags = {}, + QRhiNativeHandles *importDevice = nullptr); + static bool probe(Implementation impl, QRhiInitParams *params); + + Implementation backend() const; + const char *backendName() const; + static const char *backendName(Implementation impl); + QRhiDriverInfo driverInfo() const; + QThread *thread() const; + + using CleanupCallback = std::function<void(QRhi *)>; + void addCleanupCallback(const CleanupCallback &callback); + void addCleanupCallback(const void *key, const CleanupCallback &callback); + void removeCleanupCallback(const void *key); + void runCleanup(); + + QRhiGraphicsPipeline *newGraphicsPipeline(); + QRhiComputePipeline *newComputePipeline(); + QRhiShaderResourceBindings *newShaderResourceBindings(); + + QRhiBuffer *newBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size); + + QRhiRenderBuffer *newRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount = 1, + QRhiRenderBuffer::Flags flags = {}, + QRhiTexture::Format backingFormatHint = QRhiTexture::UnknownFormat); + + QRhiTexture *newTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiTexture *newTexture(QRhiTexture::Format format, + int width, int height, int depth, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiTexture *newTextureArray(QRhiTexture::Format format, + int arraySize, + const QSize &pixelSize, + int sampleCount = 1, + QRhiTexture::Flags flags = {}); + + QRhiSampler *newSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler::AddressMode addressU, + QRhiSampler::AddressMode addressV, + QRhiSampler::AddressMode addressW = QRhiSampler::Repeat); + + QRhiTextureRenderTarget *newTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags = {}); + + QRhiSwapChain *newSwapChain(); + FrameOpResult beginFrame(QRhiSwapChain *swapChain, BeginFrameFlags flags = {}); + FrameOpResult endFrame(QRhiSwapChain *swapChain, EndFrameFlags flags = {}); + bool isRecordingFrame() const; + int currentFrameSlot() const; + + FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, BeginFrameFlags flags = {}); + FrameOpResult endOffscreenFrame(EndFrameFlags flags = {}); + + QRhi::FrameOpResult finish(); + + QRhiResourceUpdateBatch *nextResourceUpdateBatch(); + + QList<int> supportedSampleCounts() const; + + int ubufAlignment() const; + int ubufAligned(int v) const; + + static int mipLevelsForSize(const QSize &size); + static QSize sizeForMipLevel(int mipLevel, const QSize &baseLevelSize); + + bool isYUpInFramebuffer() const; + bool isYUpInNDC() const; + bool isClipDepthZeroToOne() const; + + QMatrix4x4 clipSpaceCorrMatrix() const; + + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags = {}) const; + bool isFeatureSupported(QRhi::Feature feature) const; + int resourceLimit(ResourceLimit limit) const; + + const QRhiNativeHandles *nativeHandles(); + bool makeThreadLocalNativeContextCurrent(); + + static constexpr int MAX_MIP_LEVELS = 16; // -> max width or height is 65536 + + void releaseCachedResources(); + + bool isDeviceLost() const; + + QByteArray pipelineCacheData(); + void setPipelineCacheData(const QByteArray &data); + + QRhiStats statistics() const; + + static QRhiSwapChainProxyData updateSwapChainProxyData(Implementation impl, QWindow *window); + +protected: + QRhi(); + +private: + Q_DISABLE_COPY(QRhi) + QRhiImplementation *d = nullptr; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::Flags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::BeginFrameFlags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::EndFrameFlags) + +QT_END_NAMESPACE + +#include <rhi/qrhi_platform.h> + +#endif diff --git a/src/gui/rhi/qrhi_p.h b/src/gui/rhi/qrhi_p.h index 561ab2dd9e..b5429372a8 100644 --- a/src/gui/rhi/qrhi_p.h +++ b/src/gui/rhi/qrhi_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHI_H -#define QRHI_H +#ifndef QRHI_P_H +#define QRHI_P_H // // W A R N I N G @@ -15,1961 +15,803 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <QSize> -#include <QMatrix4x4> -#include <QList> -#include <QVarLengthArray> -#include <QThread> -#include <QColor> -#include <QImage> -#include <functional> -#include <array> -#include <private/qshader_p.h> +#include <rhi/qrhi.h> +#include <QBitArray> +#include <QAtomicInt> +#include <QElapsedTimer> +#include <QLoggingCategory> +#include <QtCore/qset.h> +#include <QtCore/qvarlengtharray.h> QT_BEGIN_NAMESPACE -class QWindow; -class QRhi; -class QRhiImplementation; -class QRhiBuffer; -class QRhiRenderBuffer; -class QRhiTexture; -class QRhiSampler; -class QRhiCommandBuffer; -class QRhiResourceUpdateBatch; -class QRhiResourceUpdateBatchPrivate; -class QRhiSwapChain; - -class Q_GUI_EXPORT QRhiDepthStencilClearValue -{ -public: - QRhiDepthStencilClearValue() = default; - QRhiDepthStencilClearValue(float d, quint32 s); +#define QRHI_RES(t, x) static_cast<t *>(x) +#define QRHI_RES_RHI(t) t *rhiD = static_cast<t *>(m_rhi) - float depthClearValue() const { return m_d; } - void setDepthClearValue(float d) { m_d = d; } +Q_DECLARE_LOGGING_CATEGORY(QRHI_LOG_INFO) - quint32 stencilClearValue() const { return m_s; } - void setStencilClearValue(quint32 s) { m_s = s; } +class QRhiImplementation +{ +public: + virtual ~QRhiImplementation(); -private: - float m_d = 1.0f; - quint32 m_s = 0; + virtual bool create(QRhi::Flags flags) = 0; + virtual void destroy() = 0; - friend bool operator==(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + virtual QRhiGraphicsPipeline *createGraphicsPipeline() = 0; + virtual QRhiComputePipeline *createComputePipeline() = 0; + virtual QRhiShaderResourceBindings *createShaderResourceBindings() = 0; + virtual QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) = 0; + virtual QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) = 0; + virtual QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) = 0; + virtual QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) = 0; + + virtual QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) = 0; + + virtual QRhiSwapChain *createSwapChain() = 0; + virtual QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) = 0; + virtual QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) = 0; + virtual QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) = 0; + virtual QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) = 0; + virtual QRhi::FrameOpResult finish() = 0; + + virtual void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + + virtual void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) = 0; + virtual void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + + virtual void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) = 0; + + virtual void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) = 0; + + virtual void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) = 0; + + virtual void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) = 0; + virtual void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) = 0; + virtual void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) = 0; + virtual void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) = 0; + + virtual void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) = 0; + virtual void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) = 0; + + virtual void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) = 0; + virtual void debugMarkEnd(QRhiCommandBuffer *cb) = 0; + virtual void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) = 0; + + virtual void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) = 0; + virtual void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; + virtual void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) = 0; + virtual void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) = 0; + + virtual const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) = 0; + virtual void beginExternal(QRhiCommandBuffer *cb) = 0; + virtual void endExternal(QRhiCommandBuffer *cb) = 0; + virtual double lastCompletedGpuTime(QRhiCommandBuffer *cb) = 0; + + virtual QList<int> supportedSampleCounts() const = 0; + virtual int ubufAlignment() const = 0; + virtual bool isYUpInFramebuffer() const = 0; + virtual bool isYUpInNDC() const = 0; + virtual bool isClipDepthZeroToOne() const = 0; + virtual QMatrix4x4 clipSpaceCorrMatrix() const = 0; + virtual bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const = 0; + virtual bool isFeatureSupported(QRhi::Feature feature) const = 0; + virtual int resourceLimit(QRhi::ResourceLimit limit) const = 0; + virtual const QRhiNativeHandles *nativeHandles() = 0; + virtual QRhiDriverInfo driverInfo() const = 0; + virtual QRhiStats statistics() = 0; + virtual bool makeThreadLocalNativeContextCurrent() = 0; + virtual void releaseCachedResources() = 0; + virtual bool isDeviceLost() const = 0; + + virtual QByteArray pipelineCacheData() = 0; + virtual void setPipelineCacheData(const QByteArray &data) = 0; + + void prepareForCreate(QRhi *rhi, QRhi::Implementation impl, QRhi::Flags flags); + + bool isCompressedFormat(QRhiTexture::Format format) const; + void compressedFormatInfo(QRhiTexture::Format format, const QSize &size, + quint32 *bpl, quint32 *byteSize, + QSize *blockDim) const; + void textureFormatInfo(QRhiTexture::Format format, const QSize &size, + quint32 *bpl, quint32 *byteSize, quint32 *bytesPerPixel) const; + bool isStencilSupportingFormat(QRhiTexture::Format format) const; + + void registerResource(QRhiResource *res, bool ownsNativeResources = true) { - return a.m_d == b.m_d && a.m_s == b.m_s; + // The ownsNativeResources is relevant for the (graphics resource) leak + // check in ~QRhiImplementation; when false, the registration's sole + // purpose is to automatically null out the resource's m_rhi pointer in + // case the rhi goes away first. (which should not happen in + // well-written applications but we try to be graceful) + resources.insert(res, ownsNativeResources); } - friend bool operator!=(const QRhiDepthStencilClearValue &a, const QRhiDepthStencilClearValue &b) noexcept + void unregisterResource(QRhiResource *res) { - return !(a == b); + resources.remove(res); } - friend size_t qHash(const QRhiDepthStencilClearValue &v, size_t seed = 0) noexcept + void addDeleteLater(QRhiResource *res) { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_d); - seed = hash(seed, v.m_s); - return seed; - } -}; - -Q_DECLARE_TYPEINFO(QRhiDepthStencilClearValue, Q_RELOCATABLE_TYPE); - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDepthStencilClearValue &); -#endif - -class Q_GUI_EXPORT QRhiViewport -{ -public: - QRhiViewport() = default; - QRhiViewport(float x, float y, float w, float h, float minDepth = 0.0f, float maxDepth = 1.0f); - - std::array<float, 4> viewport() const { return m_rect; } - void setViewport(float x, float y, float w, float h) { - m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + if (inFrame) + pendingDeleteResources.insert(res); + else + delete res; } - float minDepth() const { return m_minDepth; } - void setMinDepth(float minDepth) { m_minDepth = minDepth; } - - float maxDepth() const { return m_maxDepth; } - void setMaxDepth(float maxDepth) { m_maxDepth = maxDepth; } - -private: - std::array<float, 4> m_rect { { 0.0f, 0.0f, 0.0f, 0.0f } }; - float m_minDepth = 0.0f; - float m_maxDepth = 1.0f; - - friend bool operator==(const QRhiViewport &a, const QRhiViewport &b) noexcept + void addCleanupCallback(const QRhi::CleanupCallback &callback) { - return a.m_rect == b.m_rect - && a.m_minDepth == b.m_minDepth - && a.m_maxDepth == b.m_maxDepth; + cleanupCallbacks.append(callback); } - friend bool operator!=(const QRhiViewport &a, const QRhiViewport &b) noexcept + void addCleanupCallback(const void *key, const QRhi::CleanupCallback &callback) { - return !(a == b); + keyedCleanupCallbacks[key] = callback; } - friend size_t qHash(const QRhiViewport &v, size_t seed = 0) noexcept + void removeCleanupCallback(const void *key) { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_rect[0]); - seed = hash(seed, v.m_rect[1]); - seed = hash(seed, v.m_rect[2]); - seed = hash(seed, v.m_rect[3]); - seed = hash(seed, v.m_minDepth); - seed = hash(seed, v.m_maxDepth); - return seed; + keyedCleanupCallbacks.remove(key); } -}; - -Q_DECLARE_TYPEINFO(QRhiViewport, Q_RELOCATABLE_TYPE); -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiViewport &); -#endif + bool sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps); + bool sanityCheckShaderResourceBindings(QRhiShaderResourceBindings *srb); + void updateLayoutDesc(QRhiShaderResourceBindings *srb); -class Q_GUI_EXPORT QRhiScissor -{ -public: - QRhiScissor() = default; - QRhiScissor(int x, int y, int w, int h); - - std::array<int, 4> scissor() const { return m_rect; } - void setScissor(int x, int y, int w, int h) { - m_rect[0] = x; m_rect[1] = y; m_rect[2] = w; m_rect[3] = h; + quint32 pipelineCacheRhiId() const + { + const quint32 ver = (QT_VERSION_MAJOR << 16) | (QT_VERSION_MINOR << 8) | (QT_VERSION_PATCH); + return (quint32(implType) << 24) | ver; } -private: - std::array<int, 4> m_rect { { 0, 0, 0, 0 } }; - - friend bool operator==(const QRhiScissor &a, const QRhiScissor &b) noexcept + void pipelineCreationStart() { - return a.m_rect == b.m_rect; + pipelineCreationTimer.start(); } - friend bool operator!=(const QRhiScissor &a, const QRhiScissor &b) noexcept + void pipelineCreationEnd() { - return !(a == b); + accumulatedPipelineCreationTime += pipelineCreationTimer.elapsed(); } - friend size_t qHash(const QRhiScissor &v, size_t seed = 0) noexcept + qint64 totalPipelineCreationTime() const { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_rect[0]); - seed = hash(seed, v.m_rect[1]); - seed = hash(seed, v.m_rect[2]); - seed = hash(seed, v.m_rect[3]); - return seed; + return accumulatedPipelineCreationTime; } -}; - -Q_DECLARE_TYPEINFO(QRhiScissor, Q_RELOCATABLE_TYPE); -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiScissor &); -#endif - -class Q_GUI_EXPORT QRhiVertexInputBinding -{ -public: - enum Classification { - PerVertex, - PerInstance - }; - - QRhiVertexInputBinding() = default; - QRhiVertexInputBinding(quint32 stride, Classification cls = PerVertex, quint32 stepRate = 1); - - quint32 stride() const { return m_stride; } - void setStride(quint32 s) { m_stride = s; } - - Classification classification() const { return m_classification; } - void setClassification(Classification c) { m_classification = c; } + QRhiVertexInputAttribute::Format shaderDescVariableFormatToVertexInputFormat(QShaderDescription::VariableType type) const; + quint32 byteSizePerVertexForVertexInputFormat(QRhiVertexInputAttribute::Format format) const; - quint32 instanceStepRate() const { return m_instanceStepRate; } - void setInstanceStepRate(quint32 rate) { m_instanceStepRate = rate; } - -private: - quint32 m_stride = 0; - Classification m_classification = PerVertex; - quint32 m_instanceStepRate = 1; - - friend bool operator==(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + static const QRhiShaderResourceBinding::Data *shaderResourceBindingData(const QRhiShaderResourceBinding &binding) { - return a.m_stride == b.m_stride - && a.m_classification == b.m_classification - && a.m_instanceStepRate == b.m_instanceStepRate; + return &binding.d; } - friend bool operator!=(const QRhiVertexInputBinding &a, const QRhiVertexInputBinding &b) noexcept + static QRhiShaderResourceBinding::Data *shaderResourceBindingData(QRhiShaderResourceBinding &binding) { - return !(a == b); + return &binding.d; } - friend size_t qHash(const QRhiVertexInputBinding &v, size_t seed = 0) noexcept + static bool sortedBindingLessThan(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_stride); - seed = hash(seed, v.m_classification); - seed = hash(seed, v.m_instanceStepRate); - return seed; + return a.d.binding < b.d.binding; } -}; -Q_DECLARE_TYPEINFO(QRhiVertexInputBinding, Q_RELOCATABLE_TYPE); + int effectiveSampleCount(int sampleCount) const; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputBinding &); -#endif + QRhi *q; -class Q_GUI_EXPORT QRhiVertexInputAttribute -{ -public: - enum Format { - Float4, - Float3, - Float2, - Float, - UNormByte4, - UNormByte2, - UNormByte, - UInt4, - UInt3, - UInt2, - UInt, - SInt4, - SInt3, - SInt2, - SInt, - Half4, - Half3, - Half2, - Half - }; + static const int MAX_SHADER_CACHE_ENTRIES = 128; - QRhiVertexInputAttribute() = default; - QRhiVertexInputAttribute(int binding, int location, Format format, quint32 offset, int matrixSlice = -1); + bool debugMarkers = false; + int currentFrameSlot = 0; // for vk, mtl, and similar. unused by gl and d3d11. + bool inFrame = false; - int binding() const { return m_binding; } - void setBinding(int b) { m_binding = b; } +private: + QRhi::Implementation implType; + QThread *implThread; + QVarLengthArray<QRhiResourceUpdateBatch *, 4> resUpdPool; + quint64 resUpdPoolMap = 0; + int lastResUpdIdx = -1; + QHash<QRhiResource *, bool> resources; + QSet<QRhiResource *> pendingDeleteResources; + QVarLengthArray<QRhi::CleanupCallback, 4> cleanupCallbacks; + QHash<const void *, QRhi::CleanupCallback> keyedCleanupCallbacks; + QElapsedTimer pipelineCreationTimer; + qint64 accumulatedPipelineCreationTime = 0; - int location() const { return m_location; } - void setLocation(int loc) { m_location = loc; } + friend class QRhi; + friend class QRhiResourceUpdateBatchPrivate; +}; - Format format() const { return m_format; } - void setFormat(Format f) { m_format = f; } +enum QRhiTargetRectBoundMode +{ + UnBounded, + Bounded +}; - quint32 offset() const { return m_offset; } - void setOffset(quint32 ofs) { m_offset = ofs; } +template<QRhiTargetRectBoundMode boundingMode, typename T, size_t N> +bool qrhi_toTopLeftRenderTargetRect(const QSize &outputSize, const std::array<T, N> &r, + T *x, T *y, T *w, T *h) +{ + // x,y are bottom-left in QRhiScissor and QRhiViewport but top-left in + // Vulkan/Metal/D3D. Our input is an OpenGL-style scissor rect where both + // negative x or y, and partly or completely out of bounds rects are + // allowed. The only thing the input here cannot have is a negative width + // or height. We must handle all other input gracefully, clamping to a zero + // width or height rect in the worst case, and ensuring the resulting rect + // is inside the rendertarget's bounds because some APIs' validation/debug + // layers are allergic to out of bounds scissor rects. - int matrixSlice() const { return m_matrixSlice; } - void setMatrixSlice(int slice) { m_matrixSlice = slice; } + const T outputWidth = outputSize.width(); + const T outputHeight = outputSize.height(); + const T inputWidth = r[2]; + const T inputHeight = r[3]; -private: - int m_binding = 0; - int m_location = 0; - Format m_format = Float4; - quint32 m_offset = 0; - int m_matrixSlice = -1; + if (inputWidth < 0 || inputHeight < 0) + return false; - friend bool operator==(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept - { - return a.m_binding == b.m_binding - && a.m_location == b.m_location - && a.m_format == b.m_format - && a.m_offset == b.m_offset; - // matrixSlice excluded intentionally - } + *x = r[0]; + *y = outputHeight - (r[1] + inputHeight); + *w = inputWidth; + *h = inputHeight; - friend bool operator!=(const QRhiVertexInputAttribute &a, const QRhiVertexInputAttribute &b) noexcept - { - return !(a == b); - } + if (boundingMode == Bounded) { + const T widthOffset = *x < 0 ? -*x : 0; + const T heightOffset = *y < 0 ? -*y : 0; + *w = *x < outputWidth ? qMax<T>(0, inputWidth - widthOffset) : 0; + *h = *y < outputHeight ? qMax<T>(0, inputHeight - heightOffset) : 0; - friend size_t qHash(const QRhiVertexInputAttribute &v, size_t seed = 0) noexcept - { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_binding); - seed = hash(seed, v.m_location); - seed = hash(seed, v.m_format); - seed = hash(seed, v.m_offset); - return seed; - } -}; + if (outputWidth > 0) + *x = qBound<T>(0, *x, outputWidth - 1); + if (outputHeight > 0) + *y = qBound<T>(0, *y, outputHeight - 1); -Q_DECLARE_TYPEINFO(QRhiVertexInputAttribute, Q_RELOCATABLE_TYPE); + if (*x + *w > outputWidth) + *w = qMax<T>(0, outputWidth - *x); + if (*y + *h > outputHeight) + *h = qMax<T>(0, outputHeight - *y); + } + return true; +} -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputAttribute &); -#endif +struct QRhiBufferDataPrivate +{ + Q_DISABLE_COPY_MOVE(QRhiBufferDataPrivate) + QRhiBufferDataPrivate() { } + ~QRhiBufferDataPrivate() { delete[] largeData; } + int ref = 1; + quint32 size = 0; + quint32 largeAlloc = 0; + char *largeData = nullptr; + static constexpr quint32 SMALL_DATA_SIZE = 1024; + char data[SMALL_DATA_SIZE]; +}; -class Q_GUI_EXPORT QRhiVertexInputLayout +// no detach-with-contents, no atomic refcount, no shrink +class QRhiBufferData { public: - QRhiVertexInputLayout() = default; - - void setBindings(std::initializer_list<QRhiVertexInputBinding> list) { m_bindings = list; } - template<typename InputIterator> - void setBindings(InputIterator first, InputIterator last) + QRhiBufferData() = default; + ~QRhiBufferData() { - m_bindings.clear(); - std::copy(first, last, std::back_inserter(m_bindings)); + if (d && !--d->ref) + delete d; } - const QRhiVertexInputBinding *cbeginBindings() const { return m_bindings.cbegin(); } - const QRhiVertexInputBinding *cendBindings() const { return m_bindings.cend(); } - const QRhiVertexInputBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } - qsizetype bindingCount() const { return m_bindings.count(); } - - void setAttributes(std::initializer_list<QRhiVertexInputAttribute> list) { m_attributes = list; } - template<typename InputIterator> - void setAttributes(InputIterator first, InputIterator last) + QRhiBufferData(const QRhiBufferData &other) + : d(other.d) { - m_attributes.clear(); - std::copy(first, last, std::back_inserter(m_attributes)); + if (d) + d->ref += 1; } - const QRhiVertexInputAttribute *cbeginAttributes() const { return m_attributes.cbegin(); } - const QRhiVertexInputAttribute *cendAttributes() const { return m_attributes.cend(); } - const QRhiVertexInputAttribute *attributeAt(qsizetype index) const { return &m_attributes.at(index); } - qsizetype attributeCount() const { return m_attributes.count(); } - -private: - QVarLengthArray<QRhiVertexInputBinding, 8> m_bindings; - QVarLengthArray<QRhiVertexInputAttribute, 8> m_attributes; - - friend bool operator==(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept - { - return a.m_bindings == b.m_bindings && a.m_attributes == b.m_attributes; - } - - friend bool operator!=(const QRhiVertexInputLayout &a, const QRhiVertexInputLayout &b) noexcept - { - return !(a == b); - } - - friend size_t qHash(const QRhiVertexInputLayout &v, size_t seed = 0) noexcept + QRhiBufferData &operator=(const QRhiBufferData &other) { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_bindings); - seed = hash(seed, v.m_attributes); - return seed; - } - - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); -}; - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiVertexInputLayout &); -#endif - -class Q_GUI_EXPORT QRhiShaderStage -{ -public: - enum Type { - Vertex, - TessellationControl, - TessellationEvaluation, - Geometry, - Fragment, - Compute - }; - - QRhiShaderStage() = default; - QRhiShaderStage(Type type, const QShader &shader, - QShader::Variant v = QShader::StandardShader); - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - QShader shader() const { return m_shader; } - void setShader(const QShader &s) { m_shader = s; } - - QShader::Variant shaderVariant() const { return m_shaderVariant; } - void setShaderVariant(QShader::Variant v) { m_shaderVariant = v; } - -private: - Type m_type = Vertex; - QShader m_shader; - QShader::Variant m_shaderVariant = QShader::StandardShader; - - friend bool operator==(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + if (d == other.d) + return *this; + if (other.d) + other.d->ref += 1; + if (d && !--d->ref) + delete d; + d = other.d; + return *this; + } + const char *constData() const { - return a.m_type == b.m_type - && a.m_shader == b.m_shader - && a.m_shaderVariant == b.m_shaderVariant; + return d->size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE ? d->data : d->largeData; } - - friend bool operator!=(const QRhiShaderStage &a, const QRhiShaderStage &b) noexcept + quint32 size() const { - return !(a == b); + return d->size; } - - friend size_t qHash(const QRhiShaderStage &v, size_t seed = 0) noexcept + void assign(const char *s, quint32 size) { - QtPrivate::QHashCombine hash; - seed = hash(seed, v.m_type); - seed = hash(seed, v.m_shader); - seed = hash(seed, v.m_shaderVariant); - return seed; + if (!d) { + d = new QRhiBufferDataPrivate; + } else if (d->ref != 1) { + d->ref -= 1; + d = new QRhiBufferDataPrivate; + } + d->size = size; + if (size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE) { + memcpy(d->data, s, size); + } else { + if (d->largeAlloc < size) { + delete[] d->largeData; + d->largeAlloc = size; + d->largeData = new char[size]; + } + memcpy(d->largeData, s, size); + } } +private: + QRhiBufferDataPrivate *d = nullptr; }; -Q_DECLARE_TYPEINFO(QRhiShaderStage, Q_RELOCATABLE_TYPE); - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderStage &); -#endif +Q_DECLARE_TYPEINFO(QRhiBufferData, Q_RELOCATABLE_TYPE); -using QRhiGraphicsShaderStage = QRhiShaderStage; - -class Q_GUI_EXPORT QRhiShaderResourceBinding +class QRhiResourceUpdateBatchPrivate { public: - enum Type { - UniformBuffer, - SampledTexture, - Texture, - Sampler, - ImageLoad, - ImageStore, - ImageLoadStore, - BufferLoad, - BufferStore, - BufferLoadStore - }; - - enum StageFlag { - VertexStage = 1 << 0, - TessellationControlStage = 1 << 1, - TessellationEvaluationStage = 1 << 2, - GeometryStage = 1 << 3, - FragmentStage = 1 << 4, - ComputeStage = 1 << 5 - }; - Q_DECLARE_FLAGS(StageFlags, StageFlag) - - QRhiShaderResourceBinding() = default; + struct BufferOp { + enum Type { + DynamicUpdate, + StaticUpload, + Read + }; + Type type; + QRhiBuffer *buf; + quint32 offset; + QRhiBufferData data; + quint32 readSize; + QRhiReadbackResult *result; + + static BufferOp dynamicUpdate(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + BufferOp op = {}; + op.type = DynamicUpdate; + op.buf = buf; + op.offset = offset; + const int effectiveSize = size ? size : buf->size(); + op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); + return op; + } - bool isLayoutCompatible(const QRhiShaderResourceBinding &other) const; + static void changeToDynamicUpdate(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + op->type = DynamicUpdate; + op->buf = buf; + op->offset = offset; + const int effectiveSize = size ? size : buf->size(); + op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); + } - static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding uniformBuffer(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); - static QRhiShaderResourceBinding uniformBufferWithDynamicOffset(int binding, StageFlags stage, QRhiBuffer *buf, quint32 size); + static BufferOp staticUpload(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + BufferOp op = {}; + op.type = StaticUpload; + op.buf = buf; + op.offset = offset; + const int effectiveSize = size ? size : buf->size(); + op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); + return op; + } - static QRhiShaderResourceBinding sampledTexture(int binding, StageFlags stage, QRhiTexture *tex, QRhiSampler *sampler); + static void changeToStaticUpload(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) + { + op->type = StaticUpload; + op->buf = buf; + op->offset = offset; + const int effectiveSize = size ? size : buf->size(); + op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); + } - struct TextureAndSampler { - QRhiTexture *tex; - QRhiSampler *sampler; + static BufferOp read(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiReadbackResult *result) + { + BufferOp op = {}; + op.type = Read; + op.buf = buf; + op.offset = offset; + op.readSize = size; + op.result = result; + return op; + } }; - static QRhiShaderResourceBinding sampledTextures(int binding, StageFlags stage, int count, const TextureAndSampler *texSamplers); - - static QRhiShaderResourceBinding texture(int binding, StageFlags stage, QRhiTexture *tex); - static QRhiShaderResourceBinding textures(int binding, StageFlags stage, int count, QRhiTexture **tex); - static QRhiShaderResourceBinding sampler(int binding, StageFlags stage, QRhiSampler *sampler); - - static QRhiShaderResourceBinding imageLoad(int binding, StageFlags stage, QRhiTexture *tex, int level); - static QRhiShaderResourceBinding imageStore(int binding, StageFlags stage, QRhiTexture *tex, int level); - static QRhiShaderResourceBinding imageLoadStore(int binding, StageFlags stage, QRhiTexture *tex, int level); - - static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferLoad(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); - static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); - static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf); - static QRhiShaderResourceBinding bufferLoadStore(int binding, StageFlags stage, QRhiBuffer *buf, quint32 offset, quint32 size); - struct Data - { - int binding; - QRhiShaderResourceBinding::StageFlags stage; - QRhiShaderResourceBinding::Type type; - struct UniformBufferData { - QRhiBuffer *buf; - quint32 offset; - quint32 maybeSize; - bool hasDynamicOffset; - }; - static const int MAX_TEX_SAMPLER_ARRAY_SIZE = 16; - struct TextureAndOrSamplerData { - int count; - TextureAndSampler texSamplers[MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct StorageImageData { - QRhiTexture *tex; - int level; + struct TextureOp { + enum Type { + Upload, + Copy, + Read, + GenMips }; - struct StorageBufferData { - QRhiBuffer *buf; - quint32 offset; - quint32 maybeSize; - }; - union { - UniformBufferData ubuf; - TextureAndOrSamplerData stex; - StorageImageData simage; - StorageBufferData sbuf; - } u; - - int arraySize() const + Type type; + QRhiTexture *dst; + // Specifying multiple uploads for a subresource must be supported. + // In the backend this can then end up, where applicable, as a + // single, batched copy operation with only one set of barriers. + // This helps when doing for example glyph cache fills. + using MipLevelUploadList = std::array<QVector<QRhiTextureSubresourceUploadDescription>, QRhi::MAX_MIP_LEVELS>; + QVarLengthArray<MipLevelUploadList, 6> subresDesc; + QRhiTexture *src; + QRhiTextureCopyDescription desc; + QRhiReadbackDescription rb; + QRhiReadbackResult *result; + + static TextureOp upload(QRhiTexture *tex, const QRhiTextureUploadDescription &desc) { - return type == QRhiShaderResourceBinding::SampledTexture || type == QRhiShaderResourceBinding::Texture - ? u.stex.count - : 1; + TextureOp op = {}; + op.type = Upload; + op.dst = tex; + int maxLayer = -1; + for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) { + if (it->layer() > maxLayer) + maxLayer = it->layer(); + } + op.subresDesc.resize(maxLayer + 1); + for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) + op.subresDesc[it->layer()][it->level()].append(it->description()); + return op; } - template<typename Output> - Output serialize(Output dst) const + static TextureOp copy(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc) { - // must write out exactly LAYOUT_DESC_ENTRIES_PER_BINDING elements here - *dst++ = quint32(binding); - *dst++ = quint32(stage); - *dst++ = quint32(type); - *dst++ = quint32(arraySize()); - return dst; + TextureOp op = {}; + op.type = Copy; + op.dst = dst; + op.src = src; + op.desc = desc; + return op; } - }; - static const int LAYOUT_DESC_ENTRIES_PER_BINDING = 4; - - template<typename Output> - static void serializeLayoutDescription(const QRhiShaderResourceBinding *first, - const QRhiShaderResourceBinding *last, - Output dst) - { - while (first != last) { - dst = first->d.serialize(dst); - ++first; + static TextureOp read(const QRhiReadbackDescription &rb, QRhiReadbackResult *result) + { + TextureOp op = {}; + op.type = Read; + op.rb = rb; + op.result = result; + return op; } - } - -private: - Data d; - friend class QRhiImplementation; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBinding::StageFlags) - -Q_DECLARE_TYPEINFO(QRhiShaderResourceBinding, Q_PRIMITIVE_TYPE); - -Q_GUI_EXPORT bool operator==(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; -Q_GUI_EXPORT bool operator!=(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) noexcept; -Q_GUI_EXPORT size_t qHash(const QRhiShaderResourceBinding &b, size_t seed = 0) noexcept; -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBinding &); -#endif - -class Q_GUI_EXPORT QRhiColorAttachment -{ -public: - QRhiColorAttachment() = default; - QRhiColorAttachment(QRhiTexture *texture); - QRhiColorAttachment(QRhiRenderBuffer *renderBuffer); - - QRhiTexture *texture() const { return m_texture; } - void setTexture(QRhiTexture *tex) { m_texture = tex; } - QRhiRenderBuffer *renderBuffer() const { return m_renderBuffer; } - void setRenderBuffer(QRhiRenderBuffer *rb) { m_renderBuffer = rb; } - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } + static TextureOp genMips(QRhiTexture *tex) + { + TextureOp op = {}; + op.type = GenMips; + op.dst = tex; + return op; + } + }; - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } + int activeBufferOpCount = 0; // this is the real number of used elements in bufferOps, not bufferOps.count() + static const int BUFFER_OPS_STATIC_ALLOC = 1024; + QVarLengthArray<BufferOp, BUFFER_OPS_STATIC_ALLOC> bufferOps; - QRhiTexture *resolveTexture() const { return m_resolveTexture; } - void setResolveTexture(QRhiTexture *tex) { m_resolveTexture = tex; } + int activeTextureOpCount = 0; // this is the real number of used elements in textureOps, not textureOps.count() + static const int TEXTURE_OPS_STATIC_ALLOC = 256; + QVarLengthArray<TextureOp, TEXTURE_OPS_STATIC_ALLOC> textureOps; - int resolveLayer() const { return m_resolveLayer; } - void setResolveLayer(int layer) { m_resolveLayer = layer; } + QRhiResourceUpdateBatch *q = nullptr; + QRhiImplementation *rhi = nullptr; + int poolIndex = -1; - int resolveLevel() const { return m_resolveLevel; } - void setResolveLevel(int level) { m_resolveLevel = level; } + void free(); + void merge(QRhiResourceUpdateBatchPrivate *other); + bool hasOptimalCapacity() const; + void trimOpLists(); -private: - QRhiTexture *m_texture = nullptr; - QRhiRenderBuffer *m_renderBuffer = nullptr; - int m_layer = 0; - int m_level = 0; - QRhiTexture *m_resolveTexture = nullptr; - int m_resolveLayer = 0; - int m_resolveLevel = 0; + static QRhiResourceUpdateBatchPrivate *get(QRhiResourceUpdateBatch *b) { return b->d; } }; -Q_DECLARE_TYPEINFO(QRhiColorAttachment, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QRhiTextureRenderTargetDescription +template<typename T> +struct QRhiBatchedBindings { -public: - QRhiTextureRenderTargetDescription() = default; - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment); - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiRenderBuffer *depthStencilBuffer); - QRhiTextureRenderTargetDescription(const QRhiColorAttachment &colorAttachment, QRhiTexture *depthTexture); - - void setColorAttachments(std::initializer_list<QRhiColorAttachment> list) { m_colorAttachments = list; } - template<typename InputIterator> - void setColorAttachments(InputIterator first, InputIterator last) - { - m_colorAttachments.clear(); - std::copy(first, last, std::back_inserter(m_colorAttachments)); + void feed(int binding, T resource) { // binding must be strictly increasing + if (curBinding == -1 || binding > curBinding + 1) { + finish(); + curBatch.startBinding = binding; + curBatch.resources.clear(); + curBatch.resources.append(resource); + } else { + Q_ASSERT(binding == curBinding + 1); + curBatch.resources.append(resource); + } + curBinding = binding; } - const QRhiColorAttachment *cbeginColorAttachments() const { return m_colorAttachments.cbegin(); } - const QRhiColorAttachment *cendColorAttachments() const { return m_colorAttachments.cend(); } - const QRhiColorAttachment *colorAttachmentAt(qsizetype index) const { return &m_colorAttachments.at(index); } - qsizetype colorAttachmentCount() const { return m_colorAttachments.count(); } - - QRhiRenderBuffer *depthStencilBuffer() const { return m_depthStencilBuffer; } - void setDepthStencilBuffer(QRhiRenderBuffer *renderBuffer) { m_depthStencilBuffer = renderBuffer; } - - QRhiTexture *depthTexture() const { return m_depthTexture; } - void setDepthTexture(QRhiTexture *texture) { m_depthTexture = texture; } - -private: - QVarLengthArray<QRhiColorAttachment, 8> m_colorAttachments; - QRhiRenderBuffer *m_depthStencilBuffer = nullptr; - QRhiTexture *m_depthTexture = nullptr; -}; - -class Q_GUI_EXPORT QRhiTextureSubresourceUploadDescription -{ -public: - QRhiTextureSubresourceUploadDescription() = default; - explicit QRhiTextureSubresourceUploadDescription(const QImage &image); - QRhiTextureSubresourceUploadDescription(const void *data, quint32 size); - explicit QRhiTextureSubresourceUploadDescription(const QByteArray &data); - - QImage image() const { return m_image; } - void setImage(const QImage &image) { m_image = image; } - - QByteArray data() const { return m_data; } - void setData(const QByteArray &data) { m_data = data; } - - quint32 dataStride() const { return m_dataStride; } - void setDataStride(quint32 stride) { m_dataStride = stride; } - - QPoint destinationTopLeft() const { return m_destinationTopLeft; } - void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } - - QSize sourceSize() const { return m_sourceSize; } - void setSourceSize(const QSize &size) { m_sourceSize = size; } - - QPoint sourceTopLeft() const { return m_sourceTopLeft; } - void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } - -private: - QImage m_image; - QByteArray m_data; - quint32 m_dataStride = 0; - QPoint m_destinationTopLeft; - QSize m_sourceSize; - QPoint m_sourceTopLeft; -}; - -Q_DECLARE_TYPEINFO(QRhiTextureSubresourceUploadDescription, Q_RELOCATABLE_TYPE); -class Q_GUI_EXPORT QRhiTextureUploadEntry -{ -public: - QRhiTextureUploadEntry() = default; - QRhiTextureUploadEntry(int layer, int level, const QRhiTextureSubresourceUploadDescription &desc); - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } + bool finish() { + if (!curBatch.resources.isEmpty()) + batches.append(curBatch); + return !batches.isEmpty(); + } - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } + void clear() { + batches.clear(); + curBatch.resources.clear(); + curBinding = -1; + } - QRhiTextureSubresourceUploadDescription description() const { return m_desc; } - void setDescription(const QRhiTextureSubresourceUploadDescription &desc) { m_desc = desc; } + struct Batch { + uint startBinding; + QVarLengthArray<T, 4> resources; -private: - int m_layer = 0; - int m_level = 0; - QRhiTextureSubresourceUploadDescription m_desc; -}; + bool operator==(const Batch &other) const + { + return startBinding == other.startBinding && resources == other.resources; + } -Q_DECLARE_TYPEINFO(QRhiTextureUploadEntry, Q_RELOCATABLE_TYPE); + bool operator!=(const Batch &other) const + { + return !operator==(other); + } + }; -class Q_GUI_EXPORT QRhiTextureUploadDescription -{ -public: - QRhiTextureUploadDescription() = default; - QRhiTextureUploadDescription(const QRhiTextureUploadEntry &entry); - QRhiTextureUploadDescription(std::initializer_list<QRhiTextureUploadEntry> list); + QVarLengthArray<Batch, 4> batches; // sorted by startBinding - void setEntries(std::initializer_list<QRhiTextureUploadEntry> list) { m_entries = list; } - template<typename InputIterator> - void setEntries(InputIterator first, InputIterator last) + bool operator==(const QRhiBatchedBindings<T> &other) const { - m_entries.clear(); - std::copy(first, last, std::back_inserter(m_entries)); + return batches == other.batches; } - const QRhiTextureUploadEntry *cbeginEntries() const { return m_entries.cbegin(); } - const QRhiTextureUploadEntry *cendEntries() const { return m_entries.cend(); } - const QRhiTextureUploadEntry *entryAt(qsizetype index) const { return &m_entries.at(index); } - qsizetype entryCount() const { return m_entries.count(); } - -private: - QVarLengthArray<QRhiTextureUploadEntry, 16> m_entries; -}; - -class Q_GUI_EXPORT QRhiTextureCopyDescription -{ -public: - QRhiTextureCopyDescription() = default; - - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } - - int sourceLayer() const { return m_sourceLayer; } - void setSourceLayer(int layer) { m_sourceLayer = layer; } - - int sourceLevel() const { return m_sourceLevel; } - void setSourceLevel(int level) { m_sourceLevel = level; } - - QPoint sourceTopLeft() const { return m_sourceTopLeft; } - void setSourceTopLeft(const QPoint &p) { m_sourceTopLeft = p; } - - int destinationLayer() const { return m_destinationLayer; } - void setDestinationLayer(int layer) { m_destinationLayer = layer; } - - int destinationLevel() const { return m_destinationLevel; } - void setDestinationLevel(int level) { m_destinationLevel = level; } - - QPoint destinationTopLeft() const { return m_destinationTopLeft; } - void setDestinationTopLeft(const QPoint &p) { m_destinationTopLeft = p; } - -private: - QSize m_pixelSize; - int m_sourceLayer = 0; - int m_sourceLevel = 0; - QPoint m_sourceTopLeft; - int m_destinationLayer = 0; - int m_destinationLevel = 0; - QPoint m_destinationTopLeft; -}; -Q_DECLARE_TYPEINFO(QRhiTextureCopyDescription, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QRhiReadbackDescription -{ -public: - QRhiReadbackDescription() = default; - QRhiReadbackDescription(QRhiTexture *texture); - - QRhiTexture *texture() const { return m_texture; } - void setTexture(QRhiTexture *tex) { m_texture = tex; } - - int layer() const { return m_layer; } - void setLayer(int layer) { m_layer = layer; } - - int level() const { return m_level; } - void setLevel(int level) { m_level = level; } + bool operator!=(const QRhiBatchedBindings<T> &other) const + { + return !operator==(other); + } private: - QRhiTexture *m_texture = nullptr; - int m_layer = 0; - int m_level = 0; + Batch curBatch; + int curBinding = -1; }; -Q_DECLARE_TYPEINFO(QRhiReadbackDescription, Q_RELOCATABLE_TYPE); - -struct Q_GUI_EXPORT QRhiNativeHandles -{ -}; - -class Q_GUI_EXPORT QRhiResource +class QRhiGlobalObjectIdGenerator { public: - enum Type { - Buffer, - Texture, - Sampler, - RenderBuffer, - RenderPassDescriptor, - SwapChainRenderTarget, - TextureRenderTarget, - ShaderResourceBindings, - GraphicsPipeline, - SwapChain, - ComputePipeline, - CommandBuffer - }; - - virtual ~QRhiResource(); - - virtual Type resourceType() const = 0; - - virtual void destroy() = 0; - - void deleteLater(); - - QByteArray name() const; - void setName(const QByteArray &name); - - quint64 globalResourceId() const; - - QRhi *rhi() const; - -protected: - QRhiResource(QRhiImplementation *rhi); - Q_DISABLE_COPY(QRhiResource) - friend class QRhiImplementation; - QRhiImplementation *m_rhi = nullptr; - quint64 m_id; - QByteArray m_objectName; +#ifdef Q_ATOMIC_INT64_IS_SUPPORTED + using Type = quint64; +#else + using Type = quint32; +#endif + static Type newId(); }; -class Q_GUI_EXPORT QRhiBuffer : public QRhiResource +class QRhiPassResourceTracker { public: - enum Type { - Immutable, - Static, - Dynamic - }; - - enum UsageFlag { - VertexBuffer = 1 << 0, - IndexBuffer = 1 << 1, - UniformBuffer = 1 << 2, - StorageBuffer = 1 << 3 - }; - Q_DECLARE_FLAGS(UsageFlags, UsageFlag) + bool isEmpty() const; + void reset(); - struct NativeBuffer { - const void *objects[3]; - int slotCount; + struct UsageState { + int layout; + int access; + int stage; }; - QRhiResource::Type resourceType() const override; - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - UsageFlags usage() const { return m_usage; } - void setUsage(UsageFlags u) { m_usage = u; } - - quint32 size() const { return m_size; } - void setSize(quint32 sz) { m_size = sz; } - - virtual bool create() = 0; - - virtual NativeBuffer nativeBuffer(); - - virtual char *beginFullDynamicBufferUpdateForCurrentFrame(); - virtual void endFullDynamicBufferUpdateForCurrentFrame(); - -protected: - QRhiBuffer(QRhiImplementation *rhi, Type type_, UsageFlags usage_, quint32 size_); - Type m_type; - UsageFlags m_usage; - quint32 m_size; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiBuffer::UsageFlags) - -class Q_GUI_EXPORT QRhiTexture : public QRhiResource -{ -public: - enum Flag { - RenderTarget = 1 << 0, - CubeMap = 1 << 2, - MipMapped = 1 << 3, - sRGB = 1 << 4, - UsedAsTransferSource = 1 << 5, - UsedWithGenerateMips = 1 << 6, - UsedWithLoadStore = 1 << 7, - UsedAsCompressedAtlas = 1 << 8, - ExternalOES = 1 << 9, - ThreeDimensional = 1 << 10, - TextureRectangleGL = 1 << 11, - TextureArray = 1 << 12, - OneDimensional = 1 << 13 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum Format { - UnknownFormat, - - RGBA8, - BGRA8, - R8, - RG8, - R16, - RG16, - RED_OR_ALPHA8, - - RGBA16F, - RGBA32F, - R16F, - R32F, - - RGB10A2, - - D16, - D24, - D24S8, - D32F, - - BC1, - BC2, - BC3, - BC4, - BC5, - BC6H, - BC7, - - ETC2_RGB8, - ETC2_RGB8A1, - ETC2_RGBA8, - - ASTC_4x4, - ASTC_5x4, - ASTC_5x5, - ASTC_6x5, - ASTC_6x6, - ASTC_8x5, - ASTC_8x6, - ASTC_8x8, - ASTC_10x5, - ASTC_10x6, - ASTC_10x8, - ASTC_10x10, - ASTC_12x10, - ASTC_12x12 + enum BufferStage { + BufVertexInputStage, + BufVertexStage, + BufTCStage, + BufTEStage, + BufFragmentStage, + BufComputeStage, + BufGeometryStage }; - struct NativeTexture { - quint64 object; - int layout; // or state + enum BufferAccess { + BufVertexInput, + BufIndexRead, + BufUniformRead, + BufStorageLoad, + BufStorageStore, + BufStorageLoadStore }; - QRhiResource::Type resourceType() const override; - - Format format() const { return m_format; } - void setFormat(Format fmt) { m_format = fmt; } - - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } - - int depth() const { return m_depth; } - void setDepth(int depth) { m_depth = depth; } - - int arraySize() const { return m_arraySize; } - void setArraySize(int arraySize) { m_arraySize = arraySize; } - - int arrayRangeStart() const { return m_arrayRangeStart; } - int arrayRangeLength() const { return m_arrayRangeLength; } - void setArrayRange(int startIndex, int count) - { - m_arrayRangeStart = startIndex; - m_arrayRangeLength = count; - } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } - - virtual bool create() = 0; - virtual NativeTexture nativeTexture(); - virtual bool createFrom(NativeTexture src); - virtual void setNativeLayout(int layout); - -protected: - QRhiTexture(QRhiImplementation *rhi, Format format_, const QSize &pixelSize_, int depth_, - int arraySize_, int sampleCount_, Flags flags_); - Format m_format; - QSize m_pixelSize; - int m_depth; - int m_arraySize; - int m_sampleCount; - Flags m_flags; - int m_arrayRangeStart = -1; - int m_arrayRangeLength = -1; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTexture::Flags) - -class Q_GUI_EXPORT QRhiSampler : public QRhiResource -{ -public: - enum Filter { - None, - Nearest, - Linear - }; + void registerBuffer(QRhiBuffer *buf, int slot, BufferAccess *access, BufferStage *stage, + const UsageState &state); - enum AddressMode { - Repeat, - ClampToEdge, - Mirror, + enum TextureStage { + TexVertexStage, + TexTCStage, + TexTEStage, + TexFragmentStage, + TexColorOutputStage, + TexDepthOutputStage, + TexComputeStage, + TexGeometryStage }; - enum CompareOp { - Never, - Less, - Equal, - LessOrEqual, - Greater, - NotEqual, - GreaterOrEqual, - Always + enum TextureAccess { + TexSample, + TexColorOutput, + TexDepthOutput, + TexStorageLoad, + TexStorageStore, + TexStorageLoadStore }; - QRhiResource::Type resourceType() const override; - - Filter magFilter() const { return m_magFilter; } - void setMagFilter(Filter f) { m_magFilter = f; } - - Filter minFilter() const { return m_minFilter; } - void setMinFilter(Filter f) { m_minFilter = f; } - - Filter mipmapMode() const { return m_mipmapMode; } - void setMipmapMode(Filter f) { m_mipmapMode = f; } - - AddressMode addressU() const { return m_addressU; } - void setAddressU(AddressMode mode) { m_addressU = mode; } - - AddressMode addressV() const { return m_addressV; } - void setAddressV(AddressMode mode) { m_addressV = mode; } - - AddressMode addressW() const { return m_addressW; } - void setAddressW(AddressMode mode) { m_addressW = mode; } - - CompareOp textureCompareOp() const { return m_compareOp; } - void setTextureCompareOp(CompareOp op) { m_compareOp = op; } + void registerTexture(QRhiTexture *tex, TextureAccess *access, TextureStage *stage, + const UsageState &state); - virtual bool create() = 0; - -protected: - QRhiSampler(QRhiImplementation *rhi, - Filter magFilter_, Filter minFilter_, Filter mipmapMode_, - AddressMode u_, AddressMode v_, AddressMode w_); - Filter m_magFilter; - Filter m_minFilter; - Filter m_mipmapMode; - AddressMode m_addressU; - AddressMode m_addressV; - AddressMode m_addressW; - CompareOp m_compareOp; -}; - -class Q_GUI_EXPORT QRhiRenderBuffer : public QRhiResource -{ -public: - enum Type { - DepthStencil, - Color + struct Buffer { + int slot; + BufferAccess access; + BufferStage stage; + UsageState stateAtPassBegin; }; - enum Flag { - UsedWithSwapChainOnly = 1 << 0 - }; - Q_DECLARE_FLAGS(Flags, Flag) + using BufferIterator = QHash<QRhiBuffer *, Buffer>::const_iterator; + BufferIterator cbeginBuffers() const { return m_buffers.cbegin(); } + BufferIterator cendBuffers() const { return m_buffers.cend(); } - struct NativeRenderBuffer { - quint64 object; + struct Texture { + TextureAccess access; + TextureStage stage; + UsageState stateAtPassBegin; }; - QRhiResource::Type resourceType() const override; - - Type type() const { return m_type; } - void setType(Type t) { m_type = t; } - - QSize pixelSize() const { return m_pixelSize; } - void setPixelSize(const QSize &sz) { m_pixelSize = sz; } + using TextureIterator = QHash<QRhiTexture *, Texture>::const_iterator; + TextureIterator cbeginTextures() const { return m_textures.cbegin(); } + TextureIterator cendTextures() const { return m_textures.cend(); } - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } + static BufferStage toPassTrackerBufferStage(QRhiShaderResourceBinding::StageFlags stages); + static TextureStage toPassTrackerTextureStage(QRhiShaderResourceBinding::StageFlags stages); - Flags flags() const { return m_flags; } - void setFlags(Flags h) { m_flags = h; } - - virtual bool create() = 0; - virtual bool createFrom(NativeRenderBuffer src); - - virtual QRhiTexture::Format backingFormat() const = 0; - -protected: - QRhiRenderBuffer(QRhiImplementation *rhi, Type type_, const QSize &pixelSize_, - int sampleCount_, Flags flags_, QRhiTexture::Format backingFormatHint_); - Type m_type; - QSize m_pixelSize; - int m_sampleCount; - Flags m_flags; - QRhiTexture::Format m_backingFormatHint; +private: + QHash<QRhiBuffer *, Buffer> m_buffers; + QHash<QRhiTexture *, Texture> m_textures; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiRenderBuffer::Flags) +Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Buffer, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Texture, Q_RELOCATABLE_TYPE); -class Q_GUI_EXPORT QRhiRenderPassDescriptor : public QRhiResource +template<typename T, int GROW = 1024> +class QRhiBackendCommandList { public: - QRhiResource::Type resourceType() const override; - - virtual bool isCompatible(const QRhiRenderPassDescriptor *other) const = 0; - virtual const QRhiNativeHandles *nativeHandles(); - - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const = 0; - - virtual QVector<quint32> serializedFormat() const = 0; - -protected: - QRhiRenderPassDescriptor(QRhiImplementation *rhi); + QRhiBackendCommandList() = default; + ~QRhiBackendCommandList() { delete[] v; } + inline void reset() { p = 0; } + inline bool isEmpty() const { return p == 0; } + inline T &get() { + if (p == a) { + a += GROW; + T *nv = new T[a]; + if (v) { + memcpy(nv, v, p * sizeof(T)); + delete[] v; + } + v = nv; + } + return v[p++]; + } + inline void unget() { --p; } + inline T *cbegin() const { return v; } + inline T *cend() const { return v + p; } + inline T *begin() { return v; } + inline T *end() { return v + p; } +private: + Q_DISABLE_COPY(QRhiBackendCommandList) + T *v = nullptr; + int a = 0; + int p = 0; }; -class Q_GUI_EXPORT QRhiRenderTarget : public QRhiResource +struct QRhiRenderTargetAttachmentTracker { -public: - virtual QSize pixelSize() const = 0; - virtual float devicePixelRatio() const = 0; - virtual int sampleCount() const = 0; + struct ResId { quint64 id; uint generation; }; + using ResIdList = QVarLengthArray<ResId, 8 * 2 + 1>; // color, resolve, ds - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } + template<typename TexType, typename RenderBufferType> + static void updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst); -protected: - QRhiRenderTarget(QRhiImplementation *rhi); - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; + template<typename TexType, typename RenderBufferType> + static bool isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList); }; -class Q_GUI_EXPORT QRhiSwapChainRenderTarget : public QRhiRenderTarget +inline bool operator==(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) { -public: - QRhiResource::Type resourceType() const override; - QRhiSwapChain *swapChain() const { return m_swapchain; } - -protected: - QRhiSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain_); - QRhiSwapChain *m_swapchain; -}; - -class Q_GUI_EXPORT QRhiTextureRenderTarget : public QRhiRenderTarget -{ -public: - enum Flag { - PreserveColorContents = 1 << 0, - PreserveDepthStencilContents = 1 << 1 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - QRhiResource::Type resourceType() const override; - - QRhiTextureRenderTargetDescription description() const { return m_desc; } - void setDescription(const QRhiTextureRenderTargetDescription &desc) { m_desc = desc; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } + return a.id == b.id && a.generation == b.generation; +} - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; - - virtual bool create() = 0; - -protected: - QRhiTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc_, Flags flags_); - QRhiTextureRenderTargetDescription m_desc; - Flags m_flags; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiTextureRenderTarget::Flags) - -class Q_GUI_EXPORT QRhiShaderResourceBindings : public QRhiResource +inline bool operator!=(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) { -public: - QRhiResource::Type resourceType() const override; - - void setBindings(std::initializer_list<QRhiShaderResourceBinding> list) { m_bindings = list; } - template<typename InputIterator> - void setBindings(InputIterator first, InputIterator last) - { - m_bindings.clear(); - std::copy(first, last, std::back_inserter(m_bindings)); - } - const QRhiShaderResourceBinding *cbeginBindings() const { return m_bindings.cbegin(); } - const QRhiShaderResourceBinding *cendBindings() const { return m_bindings.cend(); } - const QRhiShaderResourceBinding *bindingAt(qsizetype index) const { return &m_bindings.at(index); } - qsizetype bindingCount() const { return m_bindings.count(); } - - bool isLayoutCompatible(const QRhiShaderResourceBindings *other) const; - - QVector<quint32> serializedLayoutDescription() const { return m_layoutDesc; } - - virtual bool create() = 0; - - enum UpdateFlag { - BindingsAreSorted = 0x01 - }; - Q_DECLARE_FLAGS(UpdateFlags, UpdateFlag) - - virtual void updateResources(UpdateFlags flags = {}) = 0; - -protected: - static const int BINDING_PREALLOC = 12; - QRhiShaderResourceBindings(QRhiImplementation *rhi); - QVarLengthArray<QRhiShaderResourceBinding, BINDING_PREALLOC> m_bindings; - size_t m_layoutDescHash = 0; - // Intentionally not using QVLA for m_layoutDesc: clients like Qt Quick are much - // better served with an implicitly shared container here, because they will likely - // throw this directly into structs serving as cache keys. - QVector<quint32> m_layoutDesc; - friend class QRhiImplementation; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); -#endif -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiShaderResourceBindings::UpdateFlags) - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiShaderResourceBindings &); -#endif + return !(a == b); +} -class Q_GUI_EXPORT QRhiGraphicsPipeline : public QRhiResource +template<typename TexType, typename RenderBufferType> +void QRhiRenderTargetAttachmentTracker::updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst) { -public: - enum Flag { - UsesBlendConstants = 1 << 0, - UsesStencilRef = 1 << 1, - UsesScissor = 1 << 2, - CompileShadersWithDebugInfo = 1 << 3 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum Topology { - Triangles, - TriangleStrip, - TriangleFan, - Lines, - LineStrip, - Points, - Patches - }; - - enum CullMode { - None, - Front, - Back - }; - - enum FrontFace { - CCW, - CW - }; - - enum ColorMaskComponent { - R = 1 << 0, - G = 1 << 1, - B = 1 << 2, - A = 1 << 3 - }; - Q_DECLARE_FLAGS(ColorMask, ColorMaskComponent) - - enum BlendFactor { - Zero, - One, - SrcColor, - OneMinusSrcColor, - DstColor, - OneMinusDstColor, - SrcAlpha, - OneMinusSrcAlpha, - DstAlpha, - OneMinusDstAlpha, - ConstantColor, - OneMinusConstantColor, - ConstantAlpha, - OneMinusConstantAlpha, - SrcAlphaSaturate, - Src1Color, - OneMinusSrc1Color, - Src1Alpha, - OneMinusSrc1Alpha - }; - - enum BlendOp { - Add, - Subtract, - ReverseSubtract, - Min, - Max - }; - - struct TargetBlend { - ColorMask colorWrite = ColorMask(0xF); // R | G | B | A - bool enable = false; - BlendFactor srcColor = One; - BlendFactor dstColor = OneMinusSrcAlpha; - BlendOp opColor = Add; - BlendFactor srcAlpha = One; - BlendFactor dstAlpha = OneMinusSrcAlpha; - BlendOp opAlpha = Add; - }; - - enum CompareOp { - Never, - Less, - Equal, - LessOrEqual, - Greater, - NotEqual, - GreaterOrEqual, - Always - }; - - enum StencilOp { - StencilZero, - Keep, - Replace, - IncrementAndClamp, - DecrementAndClamp, - Invert, - IncrementAndWrap, - DecrementAndWrap - }; - - struct StencilOpState { - StencilOp failOp = Keep; - StencilOp depthFailOp = Keep; - StencilOp passOp = Keep; - CompareOp compareOp = Always; - }; - - enum PolygonMode { - Fill, - Line - }; - - QRhiResource::Type resourceType() const override; - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - Topology topology() const { return m_topology; } - void setTopology(Topology t) { m_topology = t; } - - CullMode cullMode() const { return m_cullMode; } - void setCullMode(CullMode mode) { m_cullMode = mode; } - - FrontFace frontFace() const { return m_frontFace; } - void setFrontFace(FrontFace f) { m_frontFace = f; } - - void setTargetBlends(std::initializer_list<TargetBlend> list) { m_targetBlends = list; } - template<typename InputIterator> - void setTargetBlends(InputIterator first, InputIterator last) - { - m_targetBlends.clear(); - std::copy(first, last, std::back_inserter(m_targetBlends)); + const bool hasDepthStencil = desc.depthStencilBuffer() || desc.depthTexture(); + dst->resize(desc.colorAttachmentCount() * 2 + (hasDepthStencil ? 1 : 0)); + int n = 0; + for (auto it = desc.cbeginColorAttachments(), itEnd = desc.cendColorAttachments(); it != itEnd; ++it, ++n) { + const QRhiColorAttachment &colorAtt(*it); + if (colorAtt.texture()) { + TexType *texD = QRHI_RES(TexType, colorAtt.texture()); + (*dst)[n] = { texD->globalResourceId(), texD->generation }; + } else if (colorAtt.renderBuffer()) { + RenderBufferType *rbD = QRHI_RES(RenderBufferType, colorAtt.renderBuffer()); + (*dst)[n] = { rbD->globalResourceId(), rbD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } + ++n; + if (colorAtt.resolveTexture()) { + TexType *texD = QRHI_RES(TexType, colorAtt.resolveTexture()); + (*dst)[n] = { texD->globalResourceId(), texD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } } - const TargetBlend *cbeginTargetBlends() const { return m_targetBlends.cbegin(); } - const TargetBlend *cendTargetBlends() const { return m_targetBlends.cend(); } - const TargetBlend *targetBlendAt(qsizetype index) const { return &m_targetBlends.at(index); } - qsizetype targetBlendCount() const { return m_targetBlends.count(); } - - bool hasDepthTest() const { return m_depthTest; } - void setDepthTest(bool enable) { m_depthTest = enable; } - - bool hasDepthWrite() const { return m_depthWrite; } - void setDepthWrite(bool enable) { m_depthWrite = enable; } - - CompareOp depthOp() const { return m_depthOp; } - void setDepthOp(CompareOp op) { m_depthOp = op; } - - bool hasStencilTest() const { return m_stencilTest; } - void setStencilTest(bool enable) { m_stencilTest = enable; } - - StencilOpState stencilFront() const { return m_stencilFront; } - void setStencilFront(const StencilOpState &state) { m_stencilFront = state; } - - StencilOpState stencilBack() const { return m_stencilBack; } - void setStencilBack(const StencilOpState &state) { m_stencilBack = state; } - - quint32 stencilReadMask() const { return m_stencilReadMask; } - void setStencilReadMask(quint32 mask) { m_stencilReadMask = mask; } - - quint32 stencilWriteMask() const { return m_stencilWriteMask; } - void setStencilWriteMask(quint32 mask) { m_stencilWriteMask = mask; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int s) { m_sampleCount = s; } - - float lineWidth() const { return m_lineWidth; } - void setLineWidth(float width) { m_lineWidth = width; } - - int depthBias() const { return m_depthBias; } - void setDepthBias(int bias) { m_depthBias = bias; } - - float slopeScaledDepthBias() const { return m_slopeScaledDepthBias; } - void setSlopeScaledDepthBias(float bias) { m_slopeScaledDepthBias = bias; } - - void setShaderStages(std::initializer_list<QRhiShaderStage> list) { m_shaderStages = list; } - template<typename InputIterator> - void setShaderStages(InputIterator first, InputIterator last) - { - m_shaderStages.clear(); - std::copy(first, last, std::back_inserter(m_shaderStages)); + if (hasDepthStencil) { + if (desc.depthTexture()) { + TexType *depthTexD = QRHI_RES(TexType, desc.depthTexture()); + (*dst)[n] = { depthTexD->globalResourceId(), depthTexD->generation }; + } else if (desc.depthStencilBuffer()) { + RenderBufferType *depthRbD = QRHI_RES(RenderBufferType, desc.depthStencilBuffer()); + (*dst)[n] = { depthRbD->globalResourceId(), depthRbD->generation }; + } else { + (*dst)[n] = { 0, 0 }; + } } - const QRhiShaderStage *cbeginShaderStages() const { return m_shaderStages.cbegin(); } - const QRhiShaderStage *cendShaderStages() const { return m_shaderStages.cend(); } - const QRhiShaderStage *shaderStageAt(qsizetype index) const { return &m_shaderStages.at(index); } - qsizetype shaderStageCount() const { return m_shaderStages.count(); } - - QRhiVertexInputLayout vertexInputLayout() const { return m_vertexInputLayout; } - void setVertexInputLayout(const QRhiVertexInputLayout &layout) { m_vertexInputLayout = layout; } - - QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } - void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } - - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } - - int patchControlPointCount() const { return m_patchControlPointCount; } - void setPatchControlPointCount(int count) { m_patchControlPointCount = count; } - - PolygonMode polygonMode() const {return m_polygonMode; } - void setPolygonMode(PolygonMode mode) {m_polygonMode = mode; } - - virtual bool create() = 0; - -protected: - QRhiGraphicsPipeline(QRhiImplementation *rhi); - Flags m_flags; - Topology m_topology = Triangles; - CullMode m_cullMode = None; - FrontFace m_frontFace = CCW; - QVarLengthArray<TargetBlend, 8> m_targetBlends; - bool m_depthTest = false; - bool m_depthWrite = false; - CompareOp m_depthOp = Less; - bool m_stencilTest = false; - StencilOpState m_stencilFront; - StencilOpState m_stencilBack; - quint32 m_stencilReadMask = 0xFF; - quint32 m_stencilWriteMask = 0xFF; - int m_sampleCount = 1; - float m_lineWidth = 1.0f; - int m_depthBias = 0; - float m_slopeScaledDepthBias = 0.0f; - int m_patchControlPointCount = 3; - PolygonMode m_polygonMode = Fill; - QVarLengthArray<QRhiShaderStage, 4> m_shaderStages; - QRhiVertexInputLayout m_vertexInputLayout; - QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::Flags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiGraphicsPipeline::ColorMask) -Q_DECLARE_TYPEINFO(QRhiGraphicsPipeline::TargetBlend, Q_RELOCATABLE_TYPE); - -struct QRhiSwapChainHdrInfo -{ - bool isHardCodedDefaults; - enum LimitsType { - LuminanceInNits, - ColorComponentValue - }; - LimitsType limitsType; - union { - struct { - float minLuminance; - float maxLuminance; - } luminanceInNits; - struct { - float maxColorComponentValue; - } colorComponentValue; - } limits; -}; - -Q_DECLARE_TYPEINFO(QRhiSwapChainHdrInfo, Q_RELOCATABLE_TYPE); - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiSwapChainHdrInfo &); -#endif - -struct QRhiSwapChainProxyData -{ - void *reserved[2] = {}; -}; - -class Q_GUI_EXPORT QRhiSwapChain : public QRhiResource -{ -public: - enum Flag { - SurfaceHasPreMulAlpha = 1 << 0, - SurfaceHasNonPreMulAlpha = 1 << 1, - sRGB = 1 << 2, - UsedAsTransferSource = 1 << 3, - NoVSync = 1 << 4, - MinimalBufferCount = 1 << 5 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum Format { - SDR, - HDRExtendedSrgbLinear, - HDR10 - }; - - enum StereoTargetBuffer { - LeftBuffer, - RightBuffer - }; - - QRhiResource::Type resourceType() const override; - - QWindow *window() const { return m_window; } - void setWindow(QWindow *window) { m_window = window; } - - QRhiSwapChainProxyData proxyData() const { return m_proxyData; } - void setProxyData(const QRhiSwapChainProxyData &d) { m_proxyData = d; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - Format format() const { return m_format; } - void setFormat(Format f) { m_format = f; } - - QRhiRenderBuffer *depthStencil() const { return m_depthStencil; } - void setDepthStencil(QRhiRenderBuffer *ds) { m_depthStencil = ds; } - - int sampleCount() const { return m_sampleCount; } - void setSampleCount(int samples) { m_sampleCount = samples; } - - QRhiRenderPassDescriptor *renderPassDescriptor() const { return m_renderPassDesc; } - void setRenderPassDescriptor(QRhiRenderPassDescriptor *desc) { m_renderPassDesc = desc; } - - QSize currentPixelSize() const { return m_currentPixelSize; } - - virtual QRhiCommandBuffer *currentFrameCommandBuffer() = 0; - virtual QRhiRenderTarget *currentFrameRenderTarget() = 0; - virtual QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer); - virtual QSize surfacePixelSize() = 0; - virtual bool isFormatSupported(Format f) = 0; - virtual QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() = 0; - virtual bool createOrResize() = 0; - virtual QRhiSwapChainHdrInfo hdrInfo(); - -protected: - QRhiSwapChain(QRhiImplementation *rhi); - QWindow *m_window = nullptr; - Flags m_flags; - Format m_format = SDR; - QRhiRenderBuffer *m_depthStencil = nullptr; - int m_sampleCount = 1; - QRhiRenderPassDescriptor *m_renderPassDesc = nullptr; - QSize m_currentPixelSize; - QRhiSwapChainProxyData m_proxyData; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiSwapChain::Flags) - -class Q_GUI_EXPORT QRhiComputePipeline : public QRhiResource -{ -public: - enum Flag { - CompileShadersWithDebugInfo = 1 << 0 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - QRhiResource::Type resourceType() const override; - virtual bool create() = 0; - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - - QRhiShaderStage shaderStage() const { return m_shaderStage; } - void setShaderStage(const QRhiShaderStage &stage) { m_shaderStage = stage; } - - QRhiShaderResourceBindings *shaderResourceBindings() const { return m_shaderResourceBindings; } - void setShaderResourceBindings(QRhiShaderResourceBindings *srb) { m_shaderResourceBindings = srb; } - -protected: - QRhiComputePipeline(QRhiImplementation *rhi); - Flags m_flags; - QRhiShaderStage m_shaderStage; - QRhiShaderResourceBindings *m_shaderResourceBindings = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiComputePipeline::Flags) - -class Q_GUI_EXPORT QRhiCommandBuffer : public QRhiResource -{ -public: - enum IndexFormat { - IndexUInt16, - IndexUInt32 - }; - - enum BeginPassFlag { - ExternalContent = 0x01, - DoNotTrackResourcesForCompute = 0x02 - }; - Q_DECLARE_FLAGS(BeginPassFlags, BeginPassFlag) - - QRhiResource::Type resourceType() const override; - - void resourceUpdate(QRhiResourceUpdateBatch *resourceUpdates); - - void beginPass(QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates = nullptr, - BeginPassFlags flags = {}); - void endPass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); - - void setGraphicsPipeline(QRhiGraphicsPipeline *ps); - using DynamicOffset = QPair<int, quint32>; // binding, offset - void setShaderResources(QRhiShaderResourceBindings *srb = nullptr, - int dynamicOffsetCount = 0, - const DynamicOffset *dynamicOffsets = nullptr); - using VertexInput = QPair<QRhiBuffer *, quint32>; // buffer, offset - void setVertexInput(int startBinding, int bindingCount, const VertexInput *bindings, - QRhiBuffer *indexBuf = nullptr, quint32 indexOffset = 0, - IndexFormat indexFormat = IndexUInt16); - - void setViewport(const QRhiViewport &viewport); - void setScissor(const QRhiScissor &scissor); - void setBlendConstants(const QColor &c); - void setStencilRef(quint32 refValue); - - void draw(quint32 vertexCount, - quint32 instanceCount = 1, - quint32 firstVertex = 0, - quint32 firstInstance = 0); - - void drawIndexed(quint32 indexCount, - quint32 instanceCount = 1, - quint32 firstIndex = 0, - qint32 vertexOffset = 0, - quint32 firstInstance = 0); - - void debugMarkBegin(const QByteArray &name); - void debugMarkEnd(); - void debugMarkMsg(const QByteArray &msg); - - void beginComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr, BeginPassFlags flags = {}); - void endComputePass(QRhiResourceUpdateBatch *resourceUpdates = nullptr); - void setComputePipeline(QRhiComputePipeline *ps); - void dispatch(int x, int y, int z); - - const QRhiNativeHandles *nativeHandles(); - void beginExternal(); - void endExternal(); - -protected: - QRhiCommandBuffer(QRhiImplementation *rhi); -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhiCommandBuffer::BeginPassFlags) - -struct Q_GUI_EXPORT QRhiReadbackResult -{ - std::function<void()> completed = nullptr; - QRhiTexture::Format format; - QSize pixelSize; - QByteArray data; -}; - -using QRhiBufferReadbackResult = QRhiReadbackResult; - -class Q_GUI_EXPORT QRhiResourceUpdateBatch -{ -public: - ~QRhiResourceUpdateBatch(); - - void release(); - - void merge(QRhiResourceUpdateBatch *other); - bool hasOptimalCapacity() const; - - void updateDynamicBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); - void uploadStaticBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data); - void uploadStaticBuffer(QRhiBuffer *buf, const void *data); - void readBackBuffer(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiBufferReadbackResult *result); - void uploadTexture(QRhiTexture *tex, const QRhiTextureUploadDescription &desc); - void uploadTexture(QRhiTexture *tex, const QImage &image); - void copyTexture(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc = QRhiTextureCopyDescription()); - void readBackTexture(const QRhiReadbackDescription &rb, QRhiReadbackResult *result); - void generateMips(QRhiTexture *tex); - -private: - QRhiResourceUpdateBatch(QRhiImplementation *rhi); - Q_DISABLE_COPY(QRhiResourceUpdateBatch) - QRhiResourceUpdateBatchPrivate *d; - friend class QRhiResourceUpdateBatchPrivate; - friend class QRhi; -}; - -struct Q_GUI_EXPORT QRhiDriverInfo -{ - enum DeviceType { - UnknownDevice, - IntegratedDevice, - DiscreteDevice, - ExternalDevice, - VirtualDevice, - CpuDevice - }; - - QByteArray deviceName; - quint64 deviceId = 0; - quint64 vendorId = 0; - DeviceType deviceType = UnknownDevice; -}; - -Q_DECLARE_TYPEINFO(QRhiDriverInfo, Q_RELOCATABLE_TYPE); +} -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiDriverInfo &); -#endif - -struct Q_GUI_EXPORT QRhiStats +template<typename TexType, typename RenderBufferType> +bool QRhiRenderTargetAttachmentTracker::isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList) { - qint64 totalPipelineCreationTime = 0; - // Vulkan or D3D12 memory allocator statistics - quint32 blockCount = 0; - quint32 allocCount = 0; - quint64 usedBytes = 0; - quint64 unusedBytes = 0; - // D3D12 only, from IDXGIAdapter3::QueryVideoMemoryInfo(), incl. all resources - quint64 totalUsageBytes = 0; -}; + // Just as setShaderResources() recognizes if an srb's referenced + // resources have been rebuilt (got a create() since the srb's + // create()), we should do the same for the textures and renderbuffers + // referenced from the rendertarget. It is not uncommon that a texture + // or ds buffer gets resized due to following a window size in some + // form, which involves a create() on them. It is then nice if the + // render target auto-rebuilds in beginPass(). -Q_DECLARE_TYPEINFO(QRhiStats, Q_RELOCATABLE_TYPE); + ResIdList resIdList; + updateResIdList<TexType, RenderBufferType>(desc, &resIdList); + return resIdList == currentResIdList; +} -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QRhiStats &); -#endif - -struct Q_GUI_EXPORT QRhiInitParams -{ -}; - -class Q_GUI_EXPORT QRhi +template<typename T> +inline T *qrhi_objectFromProxyData(QRhiSwapChainProxyData *pd, QWindow *window, QRhi::Implementation impl, uint objectIndex) { -public: - enum Implementation { - Null, - Vulkan, - OpenGLES2, - D3D11, - Metal, - D3D12 - }; - - enum Flag { - EnableProfiling = 1 << 0, - EnableDebugMarkers = 1 << 1, - PreferSoftwareRenderer = 1 << 2, - EnablePipelineCacheDataSave = 1 << 3 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - enum FrameOpResult { - FrameOpSuccess = 0, - FrameOpError, - FrameOpSwapChainOutOfDate, - FrameOpDeviceLost - }; - - enum Feature { - MultisampleTexture = 1, - MultisampleRenderBuffer, - DebugMarkers, - Timestamps, - Instancing, - CustomInstanceStepRate, - PrimitiveRestart, - NonDynamicUniformBuffers, - NonFourAlignedEffectiveIndexBufferOffset, - NPOTTextureRepeat, - RedOrAlpha8IsRed, - ElementIndexUint, - Compute, - WideLines, - VertexShaderPointSize, - BaseVertex, - BaseInstance, - TriangleFanTopology, - ReadBackNonUniformBuffer, - ReadBackNonBaseMipLevel, - TexelFetch, - RenderToNonBaseMipLevel, - IntAttributes, - ScreenSpaceDerivatives, - ReadBackAnyTextureFormat, - PipelineCacheDataLoadSave, - ImageDataStride, - RenderBufferImport, - ThreeDimensionalTextures, - RenderTo3DTextureSlice, - TextureArrays, - Tessellation, - GeometryShader, - TextureArrayRange, - NonFillPolygonMode, - OneDimensionalTextures, - OneDimensionalTextureMipmaps, - HalfAttributes, - RenderToOneDimensionalTexture, - ThreeDimensionalTextureMipmaps - }; - - enum BeginFrameFlag { - }; - Q_DECLARE_FLAGS(BeginFrameFlags, BeginFrameFlag) - - enum EndFrameFlag { - SkipPresent = 1 << 0 - }; - Q_DECLARE_FLAGS(EndFrameFlags, EndFrameFlag) - - enum ResourceLimit { - TextureSizeMin = 1, - TextureSizeMax, - MaxColorAttachments, - FramesInFlight, - MaxAsyncReadbackFrames, - MaxThreadGroupsPerDimension, - MaxThreadsPerThreadGroup, - MaxThreadGroupX, - MaxThreadGroupY, - MaxThreadGroupZ, - TextureArraySizeMax, - MaxUniformBufferRange, - MaxVertexInputs, - MaxVertexOutputs - }; - - ~QRhi(); - - static QRhi *create(Implementation impl, - QRhiInitParams *params, - Flags flags = {}, - QRhiNativeHandles *importDevice = nullptr); - static bool probe(Implementation impl, QRhiInitParams *params); - - Implementation backend() const; - const char *backendName() const; - static const char *backendName(Implementation impl); - QRhiDriverInfo driverInfo() const; - QThread *thread() const; - - using CleanupCallback = std::function<void(QRhi *)>; - void addCleanupCallback(const CleanupCallback &callback); - void runCleanup(); - - using GpuFrameTimeCallback = std::function<void(float t)>; - void addGpuFrameTimeCallback(const GpuFrameTimeCallback &callback); - - QRhiGraphicsPipeline *newGraphicsPipeline(); - QRhiComputePipeline *newComputePipeline(); - QRhiShaderResourceBindings *newShaderResourceBindings(); - - QRhiBuffer *newBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size); - - QRhiRenderBuffer *newRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount = 1, - QRhiRenderBuffer::Flags flags = {}, - QRhiTexture::Format backingFormatHint = QRhiTexture::UnknownFormat); - - QRhiTexture *newTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int sampleCount = 1, - QRhiTexture::Flags flags = {}); - - QRhiTexture *newTexture(QRhiTexture::Format format, - int width, int height, int depth, - int sampleCount = 1, - QRhiTexture::Flags flags = {}); - - QRhiTexture *newTextureArray(QRhiTexture::Format format, - int arraySize, - const QSize &pixelSize, - int sampleCount = 1, - QRhiTexture::Flags flags = {}); - - QRhiSampler *newSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler::AddressMode addressU, - QRhiSampler::AddressMode addressV, - QRhiSampler::AddressMode addressW = QRhiSampler::Repeat); - - QRhiTextureRenderTarget *newTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags = {}); - - QRhiSwapChain *newSwapChain(); - FrameOpResult beginFrame(QRhiSwapChain *swapChain, BeginFrameFlags flags = {}); - FrameOpResult endFrame(QRhiSwapChain *swapChain, EndFrameFlags flags = {}); - bool isRecordingFrame() const; - int currentFrameSlot() const; - - FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, BeginFrameFlags flags = {}); - FrameOpResult endOffscreenFrame(EndFrameFlags flags = {}); - - QRhi::FrameOpResult finish(); - - QRhiResourceUpdateBatch *nextResourceUpdateBatch(); - - QList<int> supportedSampleCounts() const; - - int ubufAlignment() const; - int ubufAligned(int v) const; - - static int mipLevelsForSize(const QSize &size); - static QSize sizeForMipLevel(int mipLevel, const QSize &baseLevelSize); - - bool isYUpInFramebuffer() const; - bool isYUpInNDC() const; - bool isClipDepthZeroToOne() const; - - QMatrix4x4 clipSpaceCorrMatrix() const; - - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags = {}) const; - bool isFeatureSupported(QRhi::Feature feature) const; - int resourceLimit(ResourceLimit limit) const; - - const QRhiNativeHandles *nativeHandles(); - bool makeThreadLocalNativeContextCurrent(); - - static const int MAX_MIP_LEVELS = 16; // a width and/or height of 65536 should be enough for everyone - - void releaseCachedResources(); - - bool isDeviceLost() const; - - QByteArray pipelineCacheData(); - void setPipelineCacheData(const QByteArray &data); - - QRhiStats statistics() const; - - static QRhiSwapChainProxyData updateSwapChainProxyData(Implementation impl, QWindow *window); - -protected: - QRhi(); - -private: - Q_DISABLE_COPY(QRhi) - QRhiImplementation *d = nullptr; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::Flags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::BeginFrameFlags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QRhi::EndFrameFlags) + Q_ASSERT(objectIndex < std::size(pd->reserved)); + if (!pd->reserved[objectIndex]) // // was not set, no other choice, do it here, whatever thread this is + *pd = QRhi::updateSwapChainProxyData(impl, window); + return static_cast<T *>(pd->reserved[objectIndex]); +} QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhi_p_p.h b/src/gui/rhi/qrhi_p_p.h deleted file mode 100644 index 4e2b62bda6..0000000000 --- a/src/gui/rhi/qrhi_p_p.h +++ /dev/null @@ -1,819 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHI_P_H -#define QRHI_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhi_p.h" -#include <QBitArray> -#include <QAtomicInt> -#include <QElapsedTimer> -#include <QLoggingCategory> -#include <QtCore/qset.h> -#include <QtCore/qvarlengtharray.h> - -QT_BEGIN_NAMESPACE - -#define QRHI_RES(t, x) static_cast<t *>(x) -#define QRHI_RES_RHI(t) t *rhiD = static_cast<t *>(m_rhi) - -Q_DECLARE_LOGGING_CATEGORY(QRHI_LOG_INFO) - -class QRhiImplementation -{ -public: - virtual ~QRhiImplementation(); - - virtual bool create(QRhi::Flags flags) = 0; - virtual void destroy() = 0; - - virtual QRhiGraphicsPipeline *createGraphicsPipeline() = 0; - virtual QRhiComputePipeline *createComputePipeline() = 0; - virtual QRhiShaderResourceBindings *createShaderResourceBindings() = 0; - virtual QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) = 0; - virtual QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) = 0; - virtual QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) = 0; - virtual QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) = 0; - - virtual QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) = 0; - - virtual QRhiSwapChain *createSwapChain() = 0; - virtual QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) = 0; - virtual QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) = 0; - virtual QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) = 0; - virtual QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) = 0; - virtual QRhi::FrameOpResult finish() = 0; - - virtual void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - - virtual void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) = 0; - virtual void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - - virtual void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) = 0; - - virtual void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) = 0; - - virtual void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) = 0; - - virtual void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) = 0; - virtual void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) = 0; - virtual void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) = 0; - virtual void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) = 0; - - virtual void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) = 0; - virtual void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) = 0; - - virtual void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) = 0; - virtual void debugMarkEnd(QRhiCommandBuffer *cb) = 0; - virtual void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) = 0; - - virtual void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) = 0; - virtual void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) = 0; - virtual void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) = 0; - virtual void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) = 0; - - virtual const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) = 0; - virtual void beginExternal(QRhiCommandBuffer *cb) = 0; - virtual void endExternal(QRhiCommandBuffer *cb) = 0; - - virtual QList<int> supportedSampleCounts() const = 0; - virtual int ubufAlignment() const = 0; - virtual bool isYUpInFramebuffer() const = 0; - virtual bool isYUpInNDC() const = 0; - virtual bool isClipDepthZeroToOne() const = 0; - virtual QMatrix4x4 clipSpaceCorrMatrix() const = 0; - virtual bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const = 0; - virtual bool isFeatureSupported(QRhi::Feature feature) const = 0; - virtual int resourceLimit(QRhi::ResourceLimit limit) const = 0; - virtual const QRhiNativeHandles *nativeHandles() = 0; - virtual QRhiDriverInfo driverInfo() const = 0; - virtual QRhiStats statistics() = 0; - virtual bool makeThreadLocalNativeContextCurrent() = 0; - virtual void releaseCachedResources() = 0; - virtual bool isDeviceLost() const = 0; - - virtual QByteArray pipelineCacheData() = 0; - virtual void setPipelineCacheData(const QByteArray &data) = 0; - - bool isCompressedFormat(QRhiTexture::Format format) const; - void compressedFormatInfo(QRhiTexture::Format format, const QSize &size, - quint32 *bpl, quint32 *byteSize, - QSize *blockDim) const; - void textureFormatInfo(QRhiTexture::Format format, const QSize &size, - quint32 *bpl, quint32 *byteSize, quint32 *bytesPerPixel) const; - - // only really care about resources that own native graphics resources underneath - void registerResource(QRhiResource *res) - { - resources.insert(res); - } - - void unregisterResource(QRhiResource *res) - { - resources.remove(res); - } - - QSet<QRhiResource *> activeResources() const - { - return resources; - } - - void addDeleteLater(QRhiResource *res) - { - if (inFrame) - pendingDeleteResources.insert(res); - else - delete res; - } - - void addCleanupCallback(const QRhi::CleanupCallback &callback) - { - cleanupCallbacks.append(callback); - } - - void addGpuFrameTimeCallback(const QRhi::GpuFrameTimeCallback &callback) - { - gpuFrameTimeCallbacks.append(callback); - } - - bool hasGpuFrameTimeCallback() const - { - return !gpuFrameTimeCallbacks.isEmpty(); - } - - void runGpuFrameTimeCallbacks(float t) - { - for (const QRhi::GpuFrameTimeCallback &f : std::as_const(gpuFrameTimeCallbacks)) - f(t); - } - - bool sanityCheckGraphicsPipeline(QRhiGraphicsPipeline *ps); - bool sanityCheckShaderResourceBindings(QRhiShaderResourceBindings *srb); - void updateLayoutDesc(QRhiShaderResourceBindings *srb); - - quint32 pipelineCacheRhiId() const - { - const quint32 ver = (QT_VERSION_MAJOR << 16) | (QT_VERSION_MINOR << 8) | (QT_VERSION_PATCH); - return (quint32(implType) << 24) | ver; - } - - void pipelineCreationStart() - { - pipelineCreationTimer.start(); - } - - void pipelineCreationEnd() - { - accumulatedPipelineCreationTime += pipelineCreationTimer.elapsed(); - } - - qint64 totalPipelineCreationTime() const - { - return accumulatedPipelineCreationTime; - } - - QRhiVertexInputAttribute::Format shaderDescVariableFormatToVertexInputFormat(QShaderDescription::VariableType type) const; - quint32 byteSizePerVertexForVertexInputFormat(QRhiVertexInputAttribute::Format format) const; - - static const QRhiShaderResourceBinding::Data *shaderResourceBindingData(const QRhiShaderResourceBinding &binding) - { - return &binding.d; - } - - static QRhiShaderResourceBinding::Data *shaderResourceBindingData(QRhiShaderResourceBinding &binding) - { - return &binding.d; - } - - static bool sortedBindingLessThan(const QRhiShaderResourceBinding &a, const QRhiShaderResourceBinding &b) - { - return a.d.binding < b.d.binding; - } - - QRhi *q; - - static const int MAX_SHADER_CACHE_ENTRIES = 128; - - bool debugMarkers = false; - int currentFrameSlot = 0; // for vk, mtl, and similar. unused by gl and d3d11. - bool inFrame = false; - -private: - QRhi::Implementation implType; - QThread *implThread; - QVarLengthArray<QRhiResourceUpdateBatch *, 4> resUpdPool; - quint64 resUpdPoolMap = 0; - int lastResUpdIdx = -1; - QSet<QRhiResource *> resources; - QSet<QRhiResource *> pendingDeleteResources; - QVarLengthArray<QRhi::CleanupCallback, 4> cleanupCallbacks; - QVarLengthArray<QRhi::GpuFrameTimeCallback, 4> gpuFrameTimeCallbacks; - QElapsedTimer pipelineCreationTimer; - qint64 accumulatedPipelineCreationTime = 0; - - friend class QRhi; - friend class QRhiResourceUpdateBatchPrivate; -}; - -enum QRhiTargetRectBoundMode -{ - UnBounded, - Bounded -}; - -template<QRhiTargetRectBoundMode boundingMode, typename T, size_t N> -bool qrhi_toTopLeftRenderTargetRect(const QSize &outputSize, const std::array<T, N> &r, - T *x, T *y, T *w, T *h) -{ - // x,y are bottom-left in QRhiScissor and QRhiViewport but top-left in - // Vulkan/Metal/D3D. Our input is an OpenGL-style scissor rect where both - // negative x or y, and partly or completely out of bounds rects are - // allowed. The only thing the input here cannot have is a negative width - // or height. We must handle all other input gracefully, clamping to a zero - // width or height rect in the worst case, and ensuring the resulting rect - // is inside the rendertarget's bounds because some APIs' validation/debug - // layers are allergic to out of bounds scissor rects. - - const T outputWidth = outputSize.width(); - const T outputHeight = outputSize.height(); - const T inputWidth = r[2]; - const T inputHeight = r[3]; - - if (inputWidth < 0 || inputHeight < 0) - return false; - - *x = r[0]; - *y = outputHeight - (r[1] + inputHeight); - *w = inputWidth; - *h = inputHeight; - - if (boundingMode == Bounded) { - const T widthOffset = *x < 0 ? -*x : 0; - const T heightOffset = *y < 0 ? -*y : 0; - *w = *x < outputWidth ? qMax<T>(0, inputWidth - widthOffset) : 0; - *h = *y < outputHeight ? qMax<T>(0, inputHeight - heightOffset) : 0; - - if (outputWidth > 0) - *x = qBound<T>(0, *x, outputWidth - 1); - if (outputHeight > 0) - *y = qBound<T>(0, *y, outputHeight - 1); - - if (*x + *w > outputWidth) - *w = qMax<T>(0, outputWidth - *x); - if (*y + *h > outputHeight) - *h = qMax<T>(0, outputHeight - *y); - } - return true; -} - -struct QRhiBufferDataPrivate -{ - Q_DISABLE_COPY_MOVE(QRhiBufferDataPrivate) - QRhiBufferDataPrivate() { } - ~QRhiBufferDataPrivate() { delete[] largeData; } - int ref = 1; - quint32 size = 0; - quint32 largeAlloc = 0; - char *largeData = nullptr; - static constexpr quint32 SMALL_DATA_SIZE = 1024; - char data[SMALL_DATA_SIZE]; -}; - -// no detach-with-contents, no atomic refcount, no shrink -class QRhiBufferData -{ -public: - QRhiBufferData() = default; - ~QRhiBufferData() - { - if (d && !--d->ref) - delete d; - } - QRhiBufferData(const QRhiBufferData &other) - : d(other.d) - { - if (d) - d->ref += 1; - } - QRhiBufferData &operator=(const QRhiBufferData &other) - { - if (d == other.d) - return *this; - if (other.d) - other.d->ref += 1; - if (d && !--d->ref) - delete d; - d = other.d; - return *this; - } - const char *constData() const - { - return d->size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE ? d->data : d->largeData; - } - quint32 size() const - { - return d->size; - } - void assign(const char *s, quint32 size) - { - if (!d) { - d = new QRhiBufferDataPrivate; - } else if (d->ref != 1) { - d->ref -= 1; - d = new QRhiBufferDataPrivate; - } - d->size = size; - if (size <= QRhiBufferDataPrivate::SMALL_DATA_SIZE) { - memcpy(d->data, s, size); - } else { - if (d->largeAlloc < size) { - delete[] d->largeData; - d->largeAlloc = size; - d->largeData = new char[size]; - } - memcpy(d->largeData, s, size); - } - } -private: - QRhiBufferDataPrivate *d = nullptr; -}; - -Q_DECLARE_TYPEINFO(QRhiBufferData, Q_RELOCATABLE_TYPE); - -class QRhiResourceUpdateBatchPrivate -{ -public: - struct BufferOp { - enum Type { - DynamicUpdate, - StaticUpload, - Read - }; - Type type; - QRhiBuffer *buf; - quint32 offset; - QRhiBufferData data; - quint32 readSize; - QRhiBufferReadbackResult *result; - - static BufferOp dynamicUpdate(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) - { - BufferOp op = {}; - op.type = DynamicUpdate; - op.buf = buf; - op.offset = offset; - const int effectiveSize = size ? size : buf->size(); - op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); - return op; - } - - static void changeToDynamicUpdate(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) - { - op->type = DynamicUpdate; - op->buf = buf; - op->offset = offset; - const int effectiveSize = size ? size : buf->size(); - op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); - } - - static BufferOp staticUpload(QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) - { - BufferOp op = {}; - op.type = StaticUpload; - op.buf = buf; - op.offset = offset; - const int effectiveSize = size ? size : buf->size(); - op.data.assign(reinterpret_cast<const char *>(data), effectiveSize); - return op; - } - - static void changeToStaticUpload(BufferOp *op, QRhiBuffer *buf, quint32 offset, quint32 size, const void *data) - { - op->type = StaticUpload; - op->buf = buf; - op->offset = offset; - const int effectiveSize = size ? size : buf->size(); - op->data.assign(reinterpret_cast<const char *>(data), effectiveSize); - } - - static BufferOp read(QRhiBuffer *buf, quint32 offset, quint32 size, QRhiBufferReadbackResult *result) - { - BufferOp op = {}; - op.type = Read; - op.buf = buf; - op.offset = offset; - op.readSize = size; - op.result = result; - return op; - } - }; - - struct TextureOp { - enum Type { - Upload, - Copy, - Read, - GenMips - }; - Type type; - QRhiTexture *dst; - // Specifying multiple uploads for a subresource must be supported. - // In the backend this can then end up, where applicable, as a - // single, batched copy operation with only one set of barriers. - // This helps when doing for example glyph cache fills. - using MipLevelUploadList = std::array<QVector<QRhiTextureSubresourceUploadDescription>, QRhi::MAX_MIP_LEVELS>; - QVarLengthArray<MipLevelUploadList, 6> subresDesc; - QRhiTexture *src; - QRhiTextureCopyDescription desc; - QRhiReadbackDescription rb; - QRhiReadbackResult *result; - - static TextureOp upload(QRhiTexture *tex, const QRhiTextureUploadDescription &desc) - { - TextureOp op = {}; - op.type = Upload; - op.dst = tex; - int maxLayer = -1; - for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) { - if (it->layer() > maxLayer) - maxLayer = it->layer(); - } - op.subresDesc.resize(maxLayer + 1); - for (auto it = desc.cbeginEntries(), itEnd = desc.cendEntries(); it != itEnd; ++it) - op.subresDesc[it->layer()][it->level()].append(it->description()); - return op; - } - - static TextureOp copy(QRhiTexture *dst, QRhiTexture *src, const QRhiTextureCopyDescription &desc) - { - TextureOp op = {}; - op.type = Copy; - op.dst = dst; - op.src = src; - op.desc = desc; - return op; - } - - static TextureOp read(const QRhiReadbackDescription &rb, QRhiReadbackResult *result) - { - TextureOp op = {}; - op.type = Read; - op.rb = rb; - op.result = result; - return op; - } - - static TextureOp genMips(QRhiTexture *tex) - { - TextureOp op = {}; - op.type = GenMips; - op.dst = tex; - return op; - } - }; - - int activeBufferOpCount = 0; // this is the real number of used elements in bufferOps, not bufferOps.count() - static const int BUFFER_OPS_STATIC_ALLOC = 1024; - QVarLengthArray<BufferOp, BUFFER_OPS_STATIC_ALLOC> bufferOps; - - int activeTextureOpCount = 0; // this is the real number of used elements in textureOps, not textureOps.count() - static const int TEXTURE_OPS_STATIC_ALLOC = 256; - QVarLengthArray<TextureOp, TEXTURE_OPS_STATIC_ALLOC> textureOps; - - QRhiResourceUpdateBatch *q = nullptr; - QRhiImplementation *rhi = nullptr; - int poolIndex = -1; - - void free(); - void merge(QRhiResourceUpdateBatchPrivate *other); - bool hasOptimalCapacity() const; - void trimOpLists(); - - static QRhiResourceUpdateBatchPrivate *get(QRhiResourceUpdateBatch *b) { return b->d; } -}; - -template<typename T> -struct QRhiBatchedBindings -{ - void feed(int binding, T resource) { // binding must be strictly increasing - if (curBinding == -1 || binding > curBinding + 1) { - finish(); - curBatch.startBinding = binding; - curBatch.resources.clear(); - curBatch.resources.append(resource); - } else { - Q_ASSERT(binding == curBinding + 1); - curBatch.resources.append(resource); - } - curBinding = binding; - } - - bool finish() { - if (!curBatch.resources.isEmpty()) - batches.append(curBatch); - return !batches.isEmpty(); - } - - void clear() { - batches.clear(); - curBatch.resources.clear(); - curBinding = -1; - } - - struct Batch { - uint startBinding; - QVarLengthArray<T, 4> resources; - - bool operator==(const Batch &other) const - { - return startBinding == other.startBinding && resources == other.resources; - } - - bool operator!=(const Batch &other) const - { - return !operator==(other); - } - }; - - QVarLengthArray<Batch, 4> batches; // sorted by startBinding - - bool operator==(const QRhiBatchedBindings<T> &other) const - { - return batches == other.batches; - } - - bool operator!=(const QRhiBatchedBindings<T> &other) const - { - return !operator==(other); - } - -private: - Batch curBatch; - int curBinding = -1; -}; - -class QRhiGlobalObjectIdGenerator -{ -public: -#ifdef Q_ATOMIC_INT64_IS_SUPPORTED - using Type = quint64; -#else - using Type = quint32; -#endif - static Type newId(); -}; - -class QRhiPassResourceTracker -{ -public: - bool isEmpty() const; - void reset(); - - struct UsageState { - int layout; - int access; - int stage; - }; - - enum BufferStage { - BufVertexInputStage, - BufVertexStage, - BufTCStage, - BufTEStage, - BufFragmentStage, - BufComputeStage, - BufGeometryStage - }; - - enum BufferAccess { - BufVertexInput, - BufIndexRead, - BufUniformRead, - BufStorageLoad, - BufStorageStore, - BufStorageLoadStore - }; - - void registerBuffer(QRhiBuffer *buf, int slot, BufferAccess *access, BufferStage *stage, - const UsageState &state); - - enum TextureStage { - TexVertexStage, - TexTCStage, - TexTEStage, - TexFragmentStage, - TexColorOutputStage, - TexDepthOutputStage, - TexComputeStage, - TexGeometryStage - }; - - enum TextureAccess { - TexSample, - TexColorOutput, - TexDepthOutput, - TexStorageLoad, - TexStorageStore, - TexStorageLoadStore - }; - - void registerTexture(QRhiTexture *tex, TextureAccess *access, TextureStage *stage, - const UsageState &state); - - struct Buffer { - int slot; - BufferAccess access; - BufferStage stage; - UsageState stateAtPassBegin; - }; - - using BufferIterator = QHash<QRhiBuffer *, Buffer>::const_iterator; - BufferIterator cbeginBuffers() const { return m_buffers.cbegin(); } - BufferIterator cendBuffers() const { return m_buffers.cend(); } - - struct Texture { - TextureAccess access; - TextureStage stage; - UsageState stateAtPassBegin; - }; - - using TextureIterator = QHash<QRhiTexture *, Texture>::const_iterator; - TextureIterator cbeginTextures() const { return m_textures.cbegin(); } - TextureIterator cendTextures() const { return m_textures.cend(); } - - static BufferStage toPassTrackerBufferStage(QRhiShaderResourceBinding::StageFlags stages); - static TextureStage toPassTrackerTextureStage(QRhiShaderResourceBinding::StageFlags stages); - -private: - QHash<QRhiBuffer *, Buffer> m_buffers; - QHash<QRhiTexture *, Texture> m_textures; -}; - -Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Buffer, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiPassResourceTracker::Texture, Q_RELOCATABLE_TYPE); - -template<typename T, int GROW = 1024> -class QRhiBackendCommandList -{ -public: - QRhiBackendCommandList() = default; - ~QRhiBackendCommandList() { delete[] v; } - inline void reset() { p = 0; } - inline bool isEmpty() const { return p == 0; } - inline T &get() { - if (p == a) { - a += GROW; - T *nv = new T[a]; - if (v) { - memcpy(nv, v, p * sizeof(T)); - delete[] v; - } - v = nv; - } - return v[p++]; - } - inline void unget() { --p; } - inline T *cbegin() const { return v; } - inline T *cend() const { return v + p; } - inline T *begin() { return v; } - inline T *end() { return v + p; } -private: - Q_DISABLE_COPY(QRhiBackendCommandList) - T *v = nullptr; - int a = 0; - int p = 0; -}; - -struct QRhiRenderTargetAttachmentTracker -{ - struct ResId { quint64 id; uint generation; }; - using ResIdList = QVarLengthArray<ResId, 8 * 2 + 1>; // color, resolve, ds - - template<typename TexType, typename RenderBufferType> - static void updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst); - - template<typename TexType, typename RenderBufferType> - static bool isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList); -}; - -inline bool operator==(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) -{ - return a.id == b.id && a.generation == b.generation; -} - -inline bool operator!=(const QRhiRenderTargetAttachmentTracker::ResId &a, const QRhiRenderTargetAttachmentTracker::ResId &b) -{ - return !(a == b); -} - -template<typename TexType, typename RenderBufferType> -void QRhiRenderTargetAttachmentTracker::updateResIdList(const QRhiTextureRenderTargetDescription &desc, ResIdList *dst) -{ - const bool hasDepthStencil = desc.depthStencilBuffer() || desc.depthTexture(); - dst->resize(desc.colorAttachmentCount() * 2 + (hasDepthStencil ? 1 : 0)); - int n = 0; - for (auto it = desc.cbeginColorAttachments(), itEnd = desc.cendColorAttachments(); it != itEnd; ++it, ++n) { - const QRhiColorAttachment &colorAtt(*it); - if (colorAtt.texture()) { - TexType *texD = QRHI_RES(TexType, colorAtt.texture()); - (*dst)[n] = { texD->globalResourceId(), texD->generation }; - } else if (colorAtt.renderBuffer()) { - RenderBufferType *rbD = QRHI_RES(RenderBufferType, colorAtt.renderBuffer()); - (*dst)[n] = { rbD->globalResourceId(), rbD->generation }; - } else { - (*dst)[n] = { 0, 0 }; - } - ++n; - if (colorAtt.resolveTexture()) { - TexType *texD = QRHI_RES(TexType, colorAtt.resolveTexture()); - (*dst)[n] = { texD->globalResourceId(), texD->generation }; - } else { - (*dst)[n] = { 0, 0 }; - } - } - if (hasDepthStencil) { - if (desc.depthTexture()) { - TexType *depthTexD = QRHI_RES(TexType, desc.depthTexture()); - (*dst)[n] = { depthTexD->globalResourceId(), depthTexD->generation }; - } else if (desc.depthStencilBuffer()) { - RenderBufferType *depthRbD = QRHI_RES(RenderBufferType, desc.depthStencilBuffer()); - (*dst)[n] = { depthRbD->globalResourceId(), depthRbD->generation }; - } else { - (*dst)[n] = { 0, 0 }; - } - } -} - -template<typename TexType, typename RenderBufferType> -bool QRhiRenderTargetAttachmentTracker::isUpToDate(const QRhiTextureRenderTargetDescription &desc, const ResIdList ¤tResIdList) -{ - // Just as setShaderResources() recognizes if an srb's referenced - // resources have been rebuilt (got a create() since the srb's - // create()), we should do the same for the textures and renderbuffers - // referenced from the rendertarget. It is not uncommon that a texture - // or ds buffer gets resized due to following a window size in some - // form, which involves a create() on them. It is then nice if the - // render target auto-rebuilds in beginPass(). - - ResIdList resIdList; - updateResIdList<TexType, RenderBufferType>(desc, &resIdList); - return resIdList == currentResIdList; -} - -template<typename T> -inline T *qrhi_objectFromProxyData(QRhiSwapChainProxyData *pd, QWindow *window, QRhi::Implementation impl, uint objectIndex) -{ - Q_ASSERT(objectIndex < std::size(pd->reserved)); - if (!pd->reserved[objectIndex]) // // was not set, no other choice, do it here, whatever thread this is - *pd = QRhi::updateSwapChainProxyData(impl, window); - return static_cast<T *>(pd->reserved[objectIndex]); -} - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhi_platform.h b/src/gui/rhi/qrhi_platform.h new file mode 100644 index 0000000000..e7be522c52 --- /dev/null +++ b/src/gui/rhi/qrhi_platform.h @@ -0,0 +1,175 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHIPLATFORM_H +#define QRHIPLATFORM_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <rhi/qrhi.h> + +#if QT_CONFIG(opengl) +#include <QtGui/qsurfaceformat.h> +#endif + +#if QT_CONFIG(vulkan) +#include <QtGui/qvulkaninstance.h> +#endif + +#if QT_CONFIG(metal) || defined(Q_QDOC) +Q_FORWARD_DECLARE_OBJC_CLASS(MTLDevice); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandQueue); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandBuffer); +Q_FORWARD_DECLARE_OBJC_CLASS(MTLRenderCommandEncoder); +#endif + +QT_BEGIN_NAMESPACE + +struct Q_GUI_EXPORT QRhiNullInitParams : public QRhiInitParams +{ +}; + +struct Q_GUI_EXPORT QRhiNullNativeHandles : public QRhiNativeHandles +{ +}; + +#if QT_CONFIG(opengl) || defined(Q_QDOC) + +class QOpenGLContext; +class QOffscreenSurface; +class QSurface; +class QWindow; + +struct Q_GUI_EXPORT QRhiGles2InitParams : public QRhiInitParams +{ + QRhiGles2InitParams(); + + QSurfaceFormat format; + QSurface *fallbackSurface = nullptr; + QWindow *window = nullptr; + QOpenGLContext *shareContext = nullptr; + + static QOffscreenSurface *newFallbackSurface(const QSurfaceFormat &format = QSurfaceFormat::defaultFormat()); +}; + +struct Q_GUI_EXPORT QRhiGles2NativeHandles : public QRhiNativeHandles +{ + QOpenGLContext *context = nullptr; +}; + +#endif // opengl/qdoc + +#if (QT_CONFIG(vulkan) && __has_include(<vulkan/vulkan.h>)) || defined(Q_QDOC) + +struct Q_GUI_EXPORT QRhiVulkanInitParams : public QRhiInitParams +{ + QVulkanInstance *inst = nullptr; + QWindow *window = nullptr; + QByteArrayList deviceExtensions; + + static QByteArrayList preferredInstanceExtensions(); + static QByteArrayList preferredExtensionsForImportedDevice(); +}; + +struct Q_GUI_EXPORT QRhiVulkanNativeHandles : public QRhiNativeHandles +{ + // to import a physical device (always required) + VkPhysicalDevice physDev = VK_NULL_HANDLE; + // to import a device and queue + VkDevice dev = VK_NULL_HANDLE; + quint32 gfxQueueFamilyIdx = 0; + quint32 gfxQueueIdx = 0; + // and optionally, the mem allocator + void *vmemAllocator = nullptr; + + // only for querying (rhi->nativeHandles()) + VkQueue gfxQueue = VK_NULL_HANDLE; + QVulkanInstance *inst = nullptr; +}; + +struct Q_GUI_EXPORT QRhiVulkanCommandBufferNativeHandles : public QRhiNativeHandles +{ + VkCommandBuffer commandBuffer = VK_NULL_HANDLE; +}; + +struct Q_GUI_EXPORT QRhiVulkanRenderPassNativeHandles : public QRhiNativeHandles +{ + VkRenderPass renderPass = VK_NULL_HANDLE; +}; + +#endif // vulkan/qdoc + +#if defined(Q_OS_WIN) || defined(Q_QDOC) + +// no d3d includes here, to prevent precompiled header mess due to COM, hence the void pointers + +struct Q_GUI_EXPORT QRhiD3D11InitParams : public QRhiInitParams +{ + bool enableDebugLayer = false; +}; + +struct Q_GUI_EXPORT QRhiD3D11NativeHandles : public QRhiNativeHandles +{ + // to import a device and a context + void *dev = nullptr; + void *context = nullptr; + // alternatively, to specify the device feature level and/or the adapter to use + int featureLevel = 0; + quint32 adapterLuidLow = 0; + qint32 adapterLuidHigh = 0; +}; + +struct Q_GUI_EXPORT QRhiD3D12InitParams : public QRhiInitParams +{ + bool enableDebugLayer = false; +}; + +struct Q_GUI_EXPORT QRhiD3D12NativeHandles : public QRhiNativeHandles +{ + // to import a device + void *dev = nullptr; + int minimumFeatureLevel = 0; + // to just specify the adapter to use, set these and leave dev set to null + quint32 adapterLuidLow = 0; + qint32 adapterLuidHigh = 0; + // in addition, can specify the command queue to use + void *commandQueue = nullptr; +}; + +struct Q_GUI_EXPORT QRhiD3D12CommandBufferNativeHandles : public QRhiNativeHandles +{ + void *commandList = nullptr; // ID3D12GraphicsCommandList1 +}; + +#endif // WIN/QDOC + +#if QT_CONFIG(metal) || defined(Q_QDOC) + +struct Q_GUI_EXPORT QRhiMetalInitParams : public QRhiInitParams +{ +}; + +struct Q_GUI_EXPORT QRhiMetalNativeHandles : public QRhiNativeHandles +{ + MTLDevice *dev = nullptr; + MTLCommandQueue *cmdQueue = nullptr; +}; + +struct Q_GUI_EXPORT QRhiMetalCommandBufferNativeHandles : public QRhiNativeHandles +{ + MTLCommandBuffer *commandBuffer = nullptr; + MTLRenderCommandEncoder *encoder = nullptr; +}; + +#endif // MACOS/IOS/QDOC + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qrhid3d11.cpp b/src/gui/rhi/qrhid3d11.cpp index d9f323eb1a..b09baf57b2 100644 --- a/src/gui/rhi/qrhid3d11.cpp +++ b/src/gui/rhi/qrhid3d11.cpp @@ -1,16 +1,14 @@ // Copyright (C) 2019 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhid3d11_p_p.h" -#include "qshader_p.h" +#include "qrhid3d11_p.h" +#include "qshader.h" #include "vs_test_p.h" #include <QWindow> #include <qmath.h> -#include <private/qsystemlibrary_p.h> #include <QtCore/qcryptographichash.h> #include <QtCore/private/qsystemerror_p.h> - -#include <d3dcompiler.h> +#include "qrhid3dhelpers_p.h" QT_BEGIN_NAMESPACE @@ -28,10 +26,13 @@ using namespace Qt::StringLiterals; /*! \class QRhiD3D11InitParams - \internal \inmodule QtGui + \since 6.6 \brief Direct3D 11 specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A D3D11-based QRhi needs no special parameters for initialization. If desired, enableDebugLayer can be set to \c true to enable the Direct3D debug layer. This can be useful during development, but should be avoided @@ -70,16 +71,71 @@ using namespace Qt::StringLiterals; */ /*! + \variable QRhiD3D11InitParams::enableDebugLayer + + When set to true, a debug device is created, assuming the debug layer is + available. The default value is false. +*/ + +/*! \class QRhiD3D11NativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the D3D device and device context used by the QRhi. \note The class uses \c{void *} as the type since including the COM-based \c{d3d11.h} headers is not acceptable here. The actual types are \c{ID3D11Device *} and \c{ID3D11DeviceContext *}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiD3D11NativeHandles::dev + + Points to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nn-d3d11-id3d11device}{ID3D11Device} + or left set to \nullptr if no existing device is to be imported. + + \note When importing a device, both the device and the device context must be set to valid objects. +*/ + +/*! + \variable QRhiD3D11NativeHandles::context + + Points to a \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nn-d3d11-id3d11devicecontext}{ID3D11DeviceContext} + or left set to \nullptr if no existing device context is to be imported. + + \note When importing a device, both the device and the device context must be set to valid objects. +*/ + +/*! + \variable QRhiD3D11NativeHandles::featureLevel + + Specifies the feature level passed to + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nf-d3d11-d3d11createdevice}{D3D11CreateDevice()}. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. When not set, the default rules outlined in the D3D + documentation apply. +*/ + +/*! + \variable QRhiD3D11NativeHandles::adapterLuidLow + + The low part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D11NativeHandles::adapterLuidHigh + + The high part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + // help mingw with its ancient sdk headers #ifndef DXGI_ADAPTER_FLAG_SOFTWARE #define DXGI_ADAPTER_FLAG_SOFTWARE 2 @@ -161,13 +217,14 @@ bool QRhiD3D11::create(QRhi::Flags flags) if (qEnvironmentVariableIntValue("QT_D3D_FLIP_DISCARD")) qWarning("The default swap effect is FLIP_DISCARD, QT_D3D_FLIP_DISCARD is now ignored"); - if (qEnvironmentVariableIntValue("QT_D3D_NO_FLIP")) - qWarning("Non-FLIP swapchains are no longer supported, QT_D3D_NO_FLIP is now ignored"); - - qCDebug(QRHI_LOG_INFO, "FLIP_* swapchain supported = true, ALLOW_TEARING supported = %s", - supportsAllowTearing ? "true" : "false"); + // Support for flip model swapchains is required now (since we are + // targeting Windows 10+), but the option for using the old model is still + // there. (some features are not supported then, however) + useLegacySwapchainModel = qEnvironmentVariableIntValue("QT_D3D_NO_FLIP"); - qCDebug(QRHI_LOG_INFO, "Default swap effect: FLIP_DISCARD"); + qCDebug(QRHI_LOG_INFO, "FLIP_* swapchain supported = true, ALLOW_TEARING supported = %s, use legacy (non-FLIP) model = %s", + supportsAllowTearing ? "true" : "false", + useLegacySwapchainModel ? "true" : "false"); if (!importedDeviceAndContext) { IDXGIAdapter1 *adapter; @@ -216,9 +273,7 @@ bool QRhiD3D11::create(QRhi::Flags flags) if (!activeAdapter && (requestedAdapterIndex < 0 || requestedAdapterIndex == adapterIndex)) { activeAdapter = adapter; adapterLuid = desc.AdapterLuid; - driverInfoStruct.deviceName = name.toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); qCDebug(QRHI_LOG_INFO, " using this adapter"); } else { adapter->Release(); @@ -261,21 +316,46 @@ bool QRhiD3D11::create(QRhi::Flags flags) return false; } + const bool supports11_1 = SUCCEEDED(ctx->QueryInterface(__uuidof(ID3D11DeviceContext1), reinterpret_cast<void **>(&context))); + ctx->Release(); + if (!supports11_1) { + qWarning("ID3D11DeviceContext1 not supported"); + return false; + } + // Test if creating a Shader Model 5.0 vertex shader works; we want to // fail already in create() if that's not the case. ID3D11VertexShader *testShader = nullptr; if (SUCCEEDED(dev->CreateVertexShader(g_testVertexShader, sizeof(g_testVertexShader), nullptr, &testShader))) { testShader->Release(); } else { - qWarning("D3D11 smoke test failed (failed to create vertex shader)"); - ctx->Release(); + static const char *msg = "D3D11 smoke test: Failed to create vertex shader"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); return false; } - const bool supports11_1 = SUCCEEDED(ctx->QueryInterface(__uuidof(ID3D11DeviceContext1), reinterpret_cast<void **>(&context))); - ctx->Release(); - if (!supports11_1) { - qWarning("ID3D11DeviceContext1 not supported"); + D3D11_FEATURE_DATA_D3D11_OPTIONS features = {}; + if (SUCCEEDED(dev->CheckFeatureSupport(D3D11_FEATURE_D3D11_OPTIONS, &features, sizeof(features)))) { + // The D3D _runtime_ may be 11.1, but the underlying _driver_ may + // still not support this D3D_FEATURE_LEVEL_11_1 feature. (e.g. + // because it only does 11_0) + if (!features.ConstantBufferOffsetting) { + static const char *msg = "D3D11 smoke test: Constant buffer offsetting is not supported by the driver"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); + return false; + } + } else { + static const char *msg = "D3D11 smoke test: Failed to query D3D11_FEATURE_D3D11_OPTIONS"; + if (flags.testFlag(QRhi::SuppressSmokeTestWarnings)) + qCDebug(QRHI_LOG_INFO, "%s", msg); + else + qWarning("%s", msg); return false; } } else { @@ -285,12 +365,14 @@ bool QRhiD3D11::create(QRhi::Flags flags) if (SUCCEEDED(dev->QueryInterface(__uuidof(IDXGIDevice), reinterpret_cast<void **>(&dxgiDev)))) { IDXGIAdapter *adapter = nullptr; if (SUCCEEDED(dxgiDev->GetAdapter(&adapter))) { - DXGI_ADAPTER_DESC desc; - adapter->GetDesc(&desc); - adapterLuid = desc.AdapterLuid; - driverInfoStruct.deviceName = QString::fromUtf16(reinterpret_cast<char16_t *>(desc.Description)).toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + IDXGIAdapter1 *adapter1 = nullptr; + if (SUCCEEDED(adapter->QueryInterface(__uuidof(IDXGIAdapter1), reinterpret_cast<void **>(&adapter1)))) { + DXGI_ADAPTER_DESC1 desc; + adapter1->GetDesc1(&desc); + adapterLuid = desc.AdapterLuid; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); + adapter1->Release(); + } adapter->Release(); } dxgiDev->Release(); @@ -326,6 +408,17 @@ void QRhiD3D11::destroy() clearShaderCache(); + if (ofr.tsDisjointQuery) { + ofr.tsDisjointQuery->Release(); + ofr.tsDisjointQuery = nullptr; + } + for (int i = 0; i < 2; ++i) { + if (ofr.tsQueries[i]) { + ofr.tsQueries[i]->Release(); + ofr.tsQueries[i] = nullptr; + } + } + if (annotations) { annotations->Release(); annotations = nullptr; @@ -373,19 +466,13 @@ QList<int> QRhiD3D11::supportedSampleCounts() const return { 1, 2, 4, 8 }; } -DXGI_SAMPLE_DESC QRhiD3D11::effectiveSampleCount(int sampleCount) const +DXGI_SAMPLE_DESC QRhiD3D11::effectiveSampleDesc(int sampleCount) const { DXGI_SAMPLE_DESC desc; desc.Count = 1; desc.Quality = 0; - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - int s = qBound(1, sampleCount, 64); - - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return desc; - } + const int s = effectiveSampleCount(sampleCount); desc.Count = UINT(s); if (s > 1) @@ -534,6 +621,12 @@ bool QRhiD3D11::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::ThreeDimensionalTextureMipmaps: return true; + case QRhi::MultiView: + return false; + case QRhi::TextureViewFormat: + return false; // because we use fully typed formats for textures and relaxed casting is a D3D12 thing + case QRhi::ResolveDepthStencil: + return false; default: Q_UNREACHABLE(); return false; @@ -1204,7 +1297,6 @@ const QRhiNativeHandles *QRhiD3D11::nativeHandles(QRhiCommandBuffer *cb) void QRhiD3D11::beginExternal(QRhiCommandBuffer *cb) { QD3D11CommandBuffer *cbD = QRHI_RES(QD3D11CommandBuffer, cb); - // no timestampSwapChain, in order to avoid timestamp mess executeCommandBuffer(cbD); cbD->resetCommands(); } @@ -1221,6 +1313,25 @@ void QRhiD3D11::endExternal(QRhiCommandBuffer *cb) } } +double QRhiD3D11::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QD3D11CommandBuffer *cbD = QRHI_RES(QD3D11CommandBuffer, cb); + return cbD->lastGpuTime; +} + +static inline QD3D11RenderTargetData *rtData(QRhiRenderTarget *rt) +{ + switch (rt->resourceType()) { + case QRhiResource::SwapChainRenderTarget: + return &QRHI_RES(QD3D11SwapChainRenderTarget, rt)->d; + case QRhiResource::TextureRenderTarget: + return &QRHI_RES(QD3D11TextureRenderTarget, rt)->d; + default: + Q_UNREACHABLE(); + return nullptr; + } +} + QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); @@ -1229,30 +1340,6 @@ QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF contextState.currentSwapChain = swapChainD; const int currentFrameSlot = swapChainD->currentFrameSlot; - if (swapChainD->timestampActive[currentFrameSlot]) { - ID3D11Query *tsDisjoint = swapChainD->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = swapChainD->timestampQuery[tsIdx]; - ID3D11Query *tsEnd = swapChainD->timestampQuery[tsIdx + 1]; - quint64 timestamps[2]; - D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; - bool ok = true; - ok &= context->GetData(tsDisjoint, &dj, sizeof(dj), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - ok &= context->GetData(tsEnd, ×tamps[1], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - // this above is often not ready, not even in frame_where_recorded+2, - // not clear why. so make the whole thing async and do not touch the - // queries until they are finally all available in frame this+2 or - // this+4 or ... - ok &= context->GetData(tsStart, ×tamps[0], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; - if (ok) { - if (!dj.Disjoint && dj.Frequency) { - const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; - runGpuFrameTimeCallbacks(elapsedMs); - } - swapChainD->timestampActive[currentFrameSlot] = false; - } // else leave timestampActive set to true, will retry in a subsequent beginFrame - } - swapChainD->cb.resetState(); swapChainD->rt.d.rtv[0] = swapChainD->sampleDesc.Count > 1 ? @@ -1261,6 +1348,22 @@ QRhi::FrameOpResult QRhiD3D11::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF finishActiveReadbacks(); + if (swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]) { + double elapsedSec = 0; + if (swapChainD->timestamps.tryQueryTimestamps(swapChainD->currentTimestampPairIndex, context, &elapsedSec)) + swapChainD->cb.lastGpuTime = elapsedSec; + } + + ID3D11Query *tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + ID3D11Query *tsDisjoint = swapChainD->timestamps.disjointQuery[swapChainD->currentTimestampPairIndex]; + const bool recordTimestamps = tsStart && tsDisjoint && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + + QD3D11CommandBuffer::Command &cmd(swapChainD->cb.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::BeginFrame; + cmd.args.beginFrame.tsQuery = recordTimestamps ? tsStart : nullptr; + cmd.args.beginFrame.tsDisjointQuery = recordTimestamps ? tsDisjoint : nullptr; + cmd.args.beginFrame.swapchainData = rtData(&swapChainD->rt); + return QRhi::FrameOpSuccess; } @@ -1270,17 +1373,13 @@ QRhi::FrameOpResult QRhiD3D11::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame Q_ASSERT(contextState.currentSwapChain = swapChainD); const int currentFrameSlot = swapChainD->currentFrameSlot; - ID3D11Query *tsDisjoint = swapChainD->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = swapChainD->timestampQuery[tsIdx]; - ID3D11Query *tsEnd = swapChainD->timestampQuery[tsIdx + 1]; - const bool recordTimestamps = tsDisjoint && tsStart && tsEnd && !swapChainD->timestampActive[currentFrameSlot]; + QD3D11CommandBuffer::Command &cmd(swapChainD->cb.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::EndFrame; + cmd.args.endFrame.tsQuery = nullptr; // done later manually, see below + cmd.args.endFrame.tsDisjointQuery = nullptr; // send all commands to the context - if (recordTimestamps) - executeCommandBuffer(&swapChainD->cb, swapChainD); - else - executeCommandBuffer(&swapChainD->cb); + executeCommandBuffer(&swapChainD->cb); if (swapChainD->sampleDesc.Count > 1) { context->ResolveSubresource(swapChainD->backBufferTex, 0, @@ -1288,17 +1387,25 @@ QRhi::FrameOpResult QRhiD3D11::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame swapChainD->colorFormat); } - // this is here because we want to include the time spent on the resolve as well + // this is here because we want to include the time spent on the ResolveSubresource as well + ID3D11Query *tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + ID3D11Query *tsDisjoint = swapChainD->timestamps.disjointQuery[swapChainD->currentTimestampPairIndex]; + const bool recordTimestamps = tsEnd && tsDisjoint && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; if (recordTimestamps) { context->End(tsEnd); context->End(tsDisjoint); - swapChainD->timestampActive[currentFrameSlot] = true; + swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex] = true; + swapChainD->currentTimestampPairIndex = (swapChainD->currentTimestampPairIndex + 1) % QD3D11SwapChainTimestamps::TIMESTAMP_PAIRS; } if (!flags.testFlag(QRhi::SkipPresent)) { UINT presentFlags = 0; if (swapChainD->swapInterval == 0 && (swapChainD->swapChainFlags & DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING)) presentFlags |= DXGI_PRESENT_ALLOW_TEARING; + if (!swapChainD->swapChain) { + qWarning("Failed to present: IDXGISwapChain is unavailable"); + return QRhi::FrameOpError; + } HRESULT hr = swapChainD->swapChain->Present(swapChainD->swapInterval, presentFlags); if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { qWarning("Device loss detected in Present()"); @@ -1333,6 +1440,36 @@ QRhi::FrameOpResult QRhiD3D11::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: ofr.cbWrapper.resetState(); *cb = &ofr.cbWrapper; + if (rhiFlags.testFlag(QRhi::EnableTimestamps)) { + D3D11_QUERY_DESC queryDesc = {}; + if (!ofr.tsDisjointQuery) { + queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; + HRESULT hr = dev->CreateQuery(&queryDesc, &ofr.tsDisjointQuery); + if (FAILED(hr)) { + qWarning("Failed to create timestamp disjoint query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + } + queryDesc.Query = D3D11_QUERY_TIMESTAMP; + for (int i = 0; i < 2; ++i) { + if (!ofr.tsQueries[i]) { + HRESULT hr = dev->CreateQuery(&queryDesc, &ofr.tsQueries[i]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return QRhi::FrameOpError; + } + } + } + } + + QD3D11CommandBuffer::Command &cmd(ofr.cbWrapper.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::BeginFrame; + cmd.args.beginFrame.tsQuery = ofr.tsQueries[0] ? ofr.tsQueries[0] : nullptr; + cmd.args.beginFrame.tsDisjointQuery = ofr.tsDisjointQuery ? ofr.tsDisjointQuery : nullptr; + cmd.args.beginFrame.swapchainData = nullptr; + return QRhi::FrameOpSuccess; } @@ -1341,10 +1478,41 @@ QRhi::FrameOpResult QRhiD3D11::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_UNUSED(flags); ofr.active = false; + QD3D11CommandBuffer::Command &cmd(ofr.cbWrapper.commands.get()); + cmd.cmd = QD3D11CommandBuffer::Command::EndFrame; + cmd.args.endFrame.tsQuery = ofr.tsQueries[1] ? ofr.tsQueries[1] : nullptr; + cmd.args.endFrame.tsDisjointQuery = ofr.tsDisjointQuery ? ofr.tsDisjointQuery : nullptr; + executeCommandBuffer(&ofr.cbWrapper); + context->Flush(); finishActiveReadbacks(); + if (ofr.tsQueries[0]) { + quint64 timestamps[2]; + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; + HRESULT hr; + bool ok = true; + do { + hr = context->GetData(ofr.tsDisjointQuery, &dj, sizeof(dj), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + do { + hr = context->GetData(ofr.tsQueries[1], ×tamps[1], sizeof(quint64), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + do { + hr = context->GetData(ofr.tsQueries[0], ×tamps[0], sizeof(quint64), 0); + } while (hr == S_FALSE); + ok &= hr == S_OK; + if (ok) { + if (!dj.Disjoint && dj.Frequency) { + const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; + ofr.cbWrapper.lastGpuTime = elapsedMs / 1000.0; + } + } + } + return QRhi::FrameOpSuccess; } @@ -1382,9 +1550,9 @@ static inline DXGI_FORMAT toD3DTextureFormat(QRhiTexture::Format format, QRhiTex case QRhiTexture::D16: return DXGI_FORMAT_R16_TYPELESS; case QRhiTexture::D24: - return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + return DXGI_FORMAT_R24G8_TYPELESS; case QRhiTexture::D24S8: - return DXGI_FORMAT_D24_UNORM_S8_UINT; + return DXGI_FORMAT_R24G8_TYPELESS; case QRhiTexture::D32F: return DXGI_FORMAT_R32_TYPELESS; @@ -1485,7 +1653,7 @@ QRhi::FrameOpResult QRhiD3D11::finish() } else { Q_ASSERT(contextState.currentSwapChain); Q_ASSERT(contextState.currentSwapChain->cb.recordingPass == QD3D11CommandBuffer::NoPass); - executeCommandBuffer(&contextState.currentSwapChain->cb); // no timestampSwapChain, in order to avoid timestamp mess + executeCommandBuffer(&contextState.currentSwapChain->cb); contextState.currentSwapChain->cb.resetCommands(); } } @@ -1860,19 +2028,6 @@ void QRhiD3D11::finishActiveReadbacks() f(); } -static inline QD3D11RenderTargetData *rtData(QRhiRenderTarget *rt) -{ - switch (rt->resourceType()) { - case QRhiResource::SwapChainRenderTarget: - return &QRHI_RES(QD3D11SwapChainRenderTarget, rt)->d; - case QRhiResource::TextureRenderTarget: - return &QRHI_RES(QD3D11TextureRenderTarget, rt)->d; - default: - Q_UNREACHABLE(); - return nullptr; - } -} - void QRhiD3D11::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) { Q_ASSERT(QRHI_RES(QD3D11CommandBuffer, cb)->recordingPass == QD3D11CommandBuffer::NoPass); @@ -1991,6 +2146,8 @@ void QRhiD3D11::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resource cmd.args.resolveSubRes.srcSubRes = D3D11CalcSubresource(0, UINT(colorAtt.layer()), 1); cmd.args.resolveSubRes.format = dstTexD->dxgiFormat; } + if (rtTex->m_desc.depthResolveTexture()) + qWarning("Resolving multisample depth-stencil buffers is not supported with D3D"); } cbD->recordingPass = QD3D11CommandBuffer::NoPass; @@ -2578,7 +2735,7 @@ void QRhiD3D11::resetShaderResources() currentShaderMask &= ~StageU##MaskBit; \ } -void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain *timestampSwapChain) +void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD) { quint32 stencilRef = 0; float blendConstants[] = { 1, 1, 1, 1 }; @@ -2591,26 +2748,30 @@ void QRhiD3D11::executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain * }; int currentShaderMask = 0xFF; - if (timestampSwapChain) { - const int currentFrameSlot = timestampSwapChain->currentFrameSlot; - ID3D11Query *tsDisjoint = timestampSwapChain->timestampDisjointQuery[currentFrameSlot]; - const int tsIdx = QD3D11SwapChain::BUFFER_COUNT * currentFrameSlot; - ID3D11Query *tsStart = timestampSwapChain->timestampQuery[tsIdx]; - if (tsDisjoint && tsStart && !timestampSwapChain->timestampActive[currentFrameSlot]) { - // The timestamps seem to include vsync time with Present(1), except - // when running on a non-primary gpu. This is not ideal. So try working - // it around by issuing a semi-fake OMSetRenderTargets early and - // writing the first timestamp only afterwards. - context->Begin(tsDisjoint); - QD3D11RenderTargetData *rtD = rtData(×tampSwapChain->rt); - context->OMSetRenderTargets(UINT(rtD->colorAttCount), rtD->colorAttCount ? rtD->rtv : nullptr, rtD->dsv); - context->End(tsStart); // just record a timestamp, no Begin needed - } - } - for (auto it = cbD->commands.cbegin(), end = cbD->commands.cend(); it != end; ++it) { const QD3D11CommandBuffer::Command &cmd(*it); switch (cmd.cmd) { + case QD3D11CommandBuffer::Command::BeginFrame: + if (cmd.args.beginFrame.tsDisjointQuery) + context->Begin(cmd.args.beginFrame.tsDisjointQuery); + if (cmd.args.beginFrame.tsQuery) { + if (cmd.args.beginFrame.swapchainData) { + // The timestamps seem to include vsync time with Present(1), except + // when running on a non-primary gpu. This is not ideal. So try working + // it around by issuing a semi-fake OMSetRenderTargets early and + // writing the first timestamp only afterwards. + QD3D11RenderTargetData *rtD = cmd.args.beginFrame.swapchainData; + context->OMSetRenderTargets(UINT(rtD->colorAttCount), rtD->colorAttCount ? rtD->rtv : nullptr, rtD->dsv); + } + context->End(cmd.args.beginFrame.tsQuery); // no Begin() for D3D11_QUERY_TIMESTAMP + } + break; + case QD3D11CommandBuffer::Command::EndFrame: + if (cmd.args.endFrame.tsQuery) + context->End(cmd.args.endFrame.tsQuery); + if (cmd.args.endFrame.tsDisjointQuery) + context->End(cmd.args.endFrame.tsDisjointQuery); + break; case QD3D11CommandBuffer::Command::ResetShaderResources: resetShaderResources(); break; @@ -2964,7 +3125,7 @@ bool QD3D11RenderBuffer::create() return false; QRHI_RES_RHI(QRhiD3D11); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); D3D11_TEXTURE2D_DESC desc = {}; desc.Width = UINT(m_pixelSize.width()); @@ -3106,7 +3267,7 @@ static inline DXGI_FORMAT toD3DDepthTextureDSVFormat(QRhiTexture::Format format) case QRhiTexture::Format::D16: return DXGI_FORMAT_D16_UNORM; case QRhiTexture::Format::D24: - return DXGI_FORMAT_R24_UNORM_X8_TYPELESS; + return DXGI_FORMAT_D24_UNORM_S8_UINT; case QRhiTexture::Format::D24S8: return DXGI_FORMAT_D24_UNORM_S8_UINT; case QRhiTexture::Format::D32F: @@ -3135,7 +3296,7 @@ bool QD3D11Texture::prepareCreate(QSize *adjustedSize) QRHI_RES_RHI(QRhiD3D11); dxgiFormat = toD3DTextureFormat(m_format, m_flags); mipLevelCount = uint(hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); if (sampleDesc.Count > 1) { if (isCube) { qWarning("Cubemap texture cannot be multisample"); @@ -3170,12 +3331,10 @@ bool QD3D11Texture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both 1D and 3D"); return false; } - m_depth = qMax(1, m_depth); if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } - m_arraySize = qMax(0, m_arraySize); if (m_arraySize > 0 && !isArray) { qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); return false; @@ -3215,7 +3374,7 @@ bool QD3D11Texture::finishCreate() srvDesc.Texture1DArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture1DArray.FirstArraySlice = 0; - srvDesc.Texture1DArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture1DArray.ArraySize = UINT(qMax(0, m_arraySize)); } } else { srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE1D; @@ -3229,7 +3388,7 @@ bool QD3D11Texture::finishCreate() srvDesc.Texture2DMSArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture2DMSArray.FirstArraySlice = 0; - srvDesc.Texture2DMSArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, m_arraySize)); } } else { srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2DARRAY; @@ -3239,7 +3398,7 @@ bool QD3D11Texture::finishCreate() srvDesc.Texture2DArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture2DArray.FirstArraySlice = 0; - srvDesc.Texture2DArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); } } } else { @@ -3302,7 +3461,7 @@ bool QD3D11Texture::create() D3D11_TEXTURE1D_DESC desc = {}; desc.Width = UINT(size.width()); desc.MipLevels = mipLevelCount; - desc.ArraySize = isArray ? UINT(m_arraySize) : 1; + desc.ArraySize = isArray ? UINT(qMax(0, m_arraySize)) : 1; desc.Format = dxgiFormat; desc.Usage = D3D11_USAGE_DEFAULT; desc.BindFlags = bindFlags; @@ -3322,7 +3481,7 @@ bool QD3D11Texture::create() desc.Width = UINT(size.width()); desc.Height = UINT(size.height()); desc.MipLevels = mipLevelCount; - desc.ArraySize = isCube ? 6 : (isArray ? UINT(m_arraySize) : 1); + desc.ArraySize = isCube ? 6 : (isArray ? UINT(qMax(0, m_arraySize)) : 1); desc.Format = dxgiFormat; desc.SampleDesc = sampleDesc; desc.Usage = D3D11_USAGE_DEFAULT; @@ -3341,7 +3500,7 @@ bool QD3D11Texture::create() D3D11_TEXTURE3D_DESC desc = {}; desc.Width = UINT(size.width()); desc.Height = UINT(size.height()); - desc.Depth = UINT(m_depth); + desc.Depth = UINT(qMax(1, m_depth)); desc.MipLevels = mipLevelCount; desc.Format = dxgiFormat; desc.Usage = D3D11_USAGE_DEFAULT; @@ -3414,7 +3573,7 @@ ID3D11UnorderedAccessView *QD3D11Texture::unorderedAccessViewForLevel(int level) desc.ViewDimension = D3D11_UAV_DIMENSION_TEXTURE2DARRAY; desc.Texture2DArray.MipSlice = UINT(level); desc.Texture2DArray.FirstArraySlice = 0; - desc.Texture2DArray.ArraySize = UINT(m_arraySize); + desc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); } else if (is3D) { desc.ViewDimension = D3D11_UAV_DIMENSION_TEXTURE3D; desc.Texture3D.MipSlice = UINT(level); @@ -3574,7 +3733,9 @@ QD3D11RenderPassDescriptor::~QD3D11RenderPassDescriptor() void QD3D11RenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -3585,7 +3746,10 @@ bool QD3D11RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *ot QRhiRenderPassDescriptor *QD3D11RenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } QVector<quint32> QD3D11RenderPassDescriptor::serializedFormat() const @@ -3667,7 +3831,10 @@ void QD3D11TextureRenderTarget::destroy() QRhiRenderPassDescriptor *QD3D11TextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } bool QD3D11TextureRenderTarget::create() @@ -3763,6 +3930,27 @@ bool QD3D11TextureRenderTarget::create() dsvDesc.Format = toD3DDepthTextureDSVFormat(depthTexD->format()); dsvDesc.ViewDimension = depthTexD->sampleDesc.Count > 1 ? D3D11_DSV_DIMENSION_TEXTURE2DMS : D3D11_DSV_DIMENSION_TEXTURE2D; + if (depthTexD->flags().testFlag(QRhiTexture::TextureArray)) { + if (depthTexD->sampleDesc.Count > 1) { + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DMSARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DMSArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DMSArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DMSArray.FirstArraySlice = 0; + dsvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } else { + dsvDesc.ViewDimension = D3D11_DSV_DIMENSION_TEXTURE2DARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DArray.FirstArraySlice = 0; + dsvDesc.Texture2DArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } + } HRESULT hr = rhiD->dev->CreateDepthStencilView(depthTexD->tex, &dsvDesc, &dsv); if (FAILED(hr)) { qWarning("Failed to create dsv: %s", @@ -3831,6 +4019,10 @@ void QD3D11ShaderResourceBindings::destroy() { sortedBindings.clear(); boundResourceData.clear(); + + QRHI_RES_RHI(QRhiD3D11); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D11ShaderResourceBindings::create() @@ -3862,6 +4054,7 @@ bool QD3D11ShaderResourceBindings::create() } generation += 1; + rhiD->registerResource(this, false); return true; } @@ -4052,6 +4245,22 @@ static inline DXGI_FORMAT toD3DAttributeFormat(QRhiVertexInputAttribute::Format return DXGI_FORMAT_R16G16_FLOAT; case QRhiVertexInputAttribute::Half: return DXGI_FORMAT_R16_FLOAT; + case QRhiVertexInputAttribute::UShort4: + // Note: D3D does not support UShort3. Pass through UShort3 as UShort4. + case QRhiVertexInputAttribute::UShort3: + return DXGI_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort2: + return DXGI_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return DXGI_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + // Note: D3D does not support SShort3. Pass through SShort3 as SShort4. + case QRhiVertexInputAttribute::SShort3: + return DXGI_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort2: + return DXGI_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return DXGI_FORMAT_R16_SINT; default: Q_UNREACHABLE(); return DXGI_FORMAT_R32G32B32A32_FLOAT; @@ -4164,18 +4373,6 @@ static inline D3D11_BLEND_OP toD3DBlendOp(QRhiGraphicsPipeline::BlendOp op) } } -static pD3DCompile resolveD3DCompile() -{ - for (const wchar_t *libraryName : {L"D3DCompiler_47", L"D3DCompiler_43"}) { - QSystemLibrary library(libraryName); - if (library.load()) { - if (auto symbol = library.resolve("D3DCompile")) - return reinterpret_cast<pD3DCompile>(symbol); - } - } - return nullptr; -} - static inline QByteArray sourceHash(const QByteArray &source) { // taken from the GL backend, use the same mechanism to get a key @@ -4241,7 +4438,7 @@ QByteArray QRhiD3D11::compileHlslShaderSource(const QShader &shader, QShader::Va return cacheIt.value(); } - static const pD3DCompile d3dCompile = resolveD3DCompile(); + static const pD3DCompile d3dCompile = QRhiD3D::resolveD3DCompile(); if (d3dCompile == nullptr) { qWarning("Unable to resolve function D3DCompile()"); return QByteArray(); @@ -4291,7 +4488,7 @@ bool QD3D11GraphicsPipeline::create() rastDesc.SlopeScaledDepthBias = m_slopeScaledDepthBias; rastDesc.DepthClipEnable = true; rastDesc.ScissorEnable = m_flags.testFlag(UsesScissor); - rastDesc.MultisampleEnable = rhiD->effectiveSampleCount(m_sampleCount).Count > 1; + rastDesc.MultisampleEnable = rhiD->effectiveSampleDesc(m_sampleCount).Count > 1; HRESULT hr = rhiD->dev->CreateRasterizerState(&rastDesc, &rastState); if (FAILED(hr)) { qWarning("Failed to create rasterizer state: %s", @@ -4609,20 +4806,93 @@ void QD3D11CommandBuffer::destroy() // nothing to do here } +bool QD3D11SwapChainTimestamps::prepare(QRhiD3D11 *rhiD) +{ + // Creates the query objects if not yet done, but otherwise calling this + // function is expected to be a no-op. + + D3D11_QUERY_DESC queryDesc = {}; + for (int i = 0; i < TIMESTAMP_PAIRS; ++i) { + if (!disjointQuery[i]) { + queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; + HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, &disjointQuery[i]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp disjoint query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + queryDesc.Query = D3D11_QUERY_TIMESTAMP; + for (int j = 0; j < 2; ++j) { + const int idx = 2 * i + j; + if (!query[idx]) { + HRESULT hr = rhiD->dev->CreateQuery(&queryDesc, &query[idx]); + if (FAILED(hr)) { + qWarning("Failed to create timestamp query: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + } + } + return true; +} + +void QD3D11SwapChainTimestamps::destroy() +{ + for (int i = 0; i < TIMESTAMP_PAIRS; ++i) { + active[i] = false; + if (disjointQuery[i]) { + disjointQuery[i]->Release(); + disjointQuery[i] = nullptr; + } + for (int j = 0; j < 2; ++j) { + const int idx = TIMESTAMP_PAIRS * i + j; + if (query[idx]) { + query[idx]->Release(); + query[idx] = nullptr; + } + } + } +} + +bool QD3D11SwapChainTimestamps::tryQueryTimestamps(int pairIndex, ID3D11DeviceContext *context, double *elapsedSec) +{ + bool result = false; + if (!active[pairIndex]) + return result; + + ID3D11Query *tsDisjoint = disjointQuery[pairIndex]; + ID3D11Query *tsStart = query[pairIndex * 2]; + ID3D11Query *tsEnd = query[pairIndex * 2 + 1]; + quint64 timestamps[2]; + D3D11_QUERY_DATA_TIMESTAMP_DISJOINT dj; + + bool ok = true; + ok &= context->GetData(tsDisjoint, &dj, sizeof(dj), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + ok &= context->GetData(tsEnd, ×tamps[1], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + ok &= context->GetData(tsStart, ×tamps[0], sizeof(quint64), D3D11_ASYNC_GETDATA_DONOTFLUSH) == S_OK; + + if (ok) { + if (!dj.Disjoint && dj.Frequency) { + const float elapsedMs = (timestamps[1] - timestamps[0]) / float(dj.Frequency) * 1000.0f; + *elapsedSec = elapsedMs / 1000.0; + result = true; + } + active[pairIndex] = false; + } // else leave active set, will retry in a subsequent beginFrame + + return result; +} + QD3D11SwapChain::QD3D11SwapChain(QRhiImplementation *rhi) - : QRhiSwapChain(rhi), - rt(rhi, this), - cb(rhi) + : QRhiSwapChain(rhi), rt(rhi, this), rtRight(rhi, this), cb(rhi) { backBufferTex = nullptr; backBufferRtv = nullptr; for (int i = 0; i < BUFFER_COUNT; ++i) { msaaTex[i] = nullptr; msaaRtv[i] = nullptr; - timestampActive[i] = false; - timestampDisjointQuery[i] = nullptr; - timestampQuery[2 * i] = nullptr; - timestampQuery[2 * i + 1] = nullptr; } } @@ -4637,6 +4907,10 @@ void QD3D11SwapChain::releaseBuffers() backBufferRtv->Release(); backBufferRtv = nullptr; } + if (backBufferRtvRight) { + backBufferRtvRight->Release(); + backBufferRtvRight = nullptr; + } if (backBufferTex) { backBufferTex->Release(); backBufferTex = nullptr; @@ -4660,19 +4934,7 @@ void QD3D11SwapChain::destroy() releaseBuffers(); - for (int i = 0; i < BUFFER_COUNT; ++i) { - if (timestampDisjointQuery[i]) { - timestampDisjointQuery[i]->Release(); - timestampDisjointQuery[i] = nullptr; - } - for (int j = 0; j < 2; ++j) { - const int idx = BUFFER_COUNT * i + j; - if (timestampQuery[idx]) { - timestampQuery[idx]->Release(); - timestampQuery[idx] = nullptr; - } - } - } + timestamps.destroy(); swapChain->Release(); swapChain = nullptr; @@ -4688,8 +4950,12 @@ void QD3D11SwapChain::destroy() } QRHI_RES_RHI(QRhiD3D11); - if (rhiD) + if (rhiD) { rhiD->unregisterResource(this); + // See Deferred Destruction Issues with Flip Presentation Swap Chains in + // https://learn.microsoft.com/en-us/windows/win32/api/d3d11/nf-d3d11-id3d11devicecontext-flush + rhiD->context->Flush(); + } } QRhiCommandBuffer *QD3D11SwapChain::currentFrameCommandBuffer() @@ -4702,48 +4968,15 @@ QRhiRenderTarget *QD3D11SwapChain::currentFrameRenderTarget() return &rt; } -QSize QD3D11SwapChain::surfacePixelSize() +QRhiRenderTarget *QD3D11SwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) { - Q_ASSERT(m_window); - return m_window->size() * m_window->devicePixelRatio(); + return targetBuffer == StereoTargetBuffer::LeftBuffer? &rt: &rtRight; } -static bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result) -{ - bool ok = false; - QRect wr = w->geometry(); - wr = QRect(wr.topLeft() * w->devicePixelRatio(), wr.size() * w->devicePixelRatio()); - const QPoint center = wr.center(); - IDXGIOutput *currentOutput = nullptr; - IDXGIOutput *output = nullptr; - for (UINT i = 0; adapter->EnumOutputs(i, &output) != DXGI_ERROR_NOT_FOUND; ++i) { - DXGI_OUTPUT_DESC desc; - output->GetDesc(&desc); - const RECT r = desc.DesktopCoordinates; - const QRect dr(QPoint(r.left, r.top), QPoint(r.right - 1, r.bottom - 1)); - if (dr.contains(center)) { - currentOutput = output; - break; - } else { - output->Release(); - } - } - if (currentOutput) { - ok = SUCCEEDED(currentOutput->QueryInterface(__uuidof(IDXGIOutput6), reinterpret_cast<void **>(result))); - currentOutput->Release(); - } - return ok; -} - -static bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result) +QSize QD3D11SwapChain::surfacePixelSize() { - bool ok = false; - IDXGIOutput6 *out6 = nullptr; - if (output6ForWindow(w, adapter, &out6)) { - ok = SUCCEEDED(out6->GetDesc1(result)); - out6->Release(); - } - return ok; + Q_ASSERT(m_window); + return m_window->size() * m_window->devicePixelRatio(); } bool QD3D11SwapChain::isFormatSupported(Format f) @@ -4758,8 +4991,10 @@ bool QD3D11SwapChain::isFormatSupported(Format f) QRHI_RES_RHI(QRhiD3D11); DXGI_OUTPUT_DESC1 desc1; - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) - return desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) { + if (desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) + return f == QRhiSwapChain::HDRExtendedSrgbLinear || f == QRhiSwapChain::HDR10; + } return false; } @@ -4767,14 +5002,16 @@ bool QD3D11SwapChain::isFormatSupported(Format f) QRhiSwapChainHdrInfo QD3D11SwapChain::hdrInfo() { QRhiSwapChainHdrInfo info = QRhiSwapChain::hdrInfo(); - if (m_format != QRhiSwapChain::SDR && m_window) { + // Must use m_window, not window, given this may be called before createOrResize(). + if (m_window) { QRHI_RES_RHI(QRhiD3D11); DXGI_OUTPUT_DESC1 hdrOutputDesc; - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { - info.isHardCodedDefaults = false; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; info.limits.luminanceInNits.minLuminance = hdrOutputDesc.MinLuminance; info.limits.luminanceInNits.maxLuminance = hdrOutputDesc.MaxLuminance; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; // 1.0 = 80 nits + info.sdrWhiteLevel = QRhiD3D::sdrWhiteLevelInNits(hdrOutputDesc); } } return info; @@ -4782,7 +5019,10 @@ QRhiSwapChainHdrInfo QD3D11SwapChain::hdrInfo() QRhiRenderPassDescriptor *QD3D11SwapChain::newCompatibleRenderPassDescriptor() { - return new QD3D11RenderPassDescriptor(m_rhi); + QD3D11RenderPassDescriptor *rpD = new QD3D11RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiD3D11); + rhiD->registerResource(rpD, false); + return rpD; } bool QD3D11SwapChain::newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, @@ -4821,26 +5061,19 @@ bool QD3D11SwapChain::newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI return true; } -static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; -static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; - bool QRhiD3D11::ensureDirectCompositionDevice() { if (dcompDevice) return true; qCDebug(QRHI_LOG_INFO, "Creating Direct Composition device (needed for semi-transparent windows)"); - - HRESULT hr = DCompositionCreateDevice(nullptr, __uuidof(IDCompositionDevice), reinterpret_cast<void **>(&dcompDevice)); - if (FAILED(hr)) { - qWarning("Failed to Direct Composition device: %s", - qPrintable(QSystemError::windowsComString(hr))); - return false; - } - - return true; + dcompDevice = QRhiD3D::createDirectCompositionDevice(); + return dcompDevice ? true : false; } +static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; +static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; + bool QD3D11SwapChain::createOrResize() { // Can be called multiple times due to window resizes - that is not the @@ -4848,6 +5081,7 @@ bool QD3D11SwapChain::createOrResize() // resize the buffers then. const bool needsRegistration = !window || window != m_window; + const bool stereo = m_window->format().stereo(); // except if the window actually changes if (window && window != m_window) @@ -4866,9 +5100,9 @@ bool QD3D11SwapChain::createOrResize() QRHI_RES_RHI(QRhiD3D11); if (m_flags.testFlag(SurfaceHasPreMulAlpha) || m_flags.testFlag(SurfaceHasNonPreMulAlpha)) { - if (rhiD->ensureDirectCompositionDevice()) { + if (!rhiD->useLegacySwapchainModel && rhiD->ensureDirectCompositionDevice()) { if (!dcompTarget) { - hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, true, &dcompTarget); + hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, false, &dcompTarget); if (FAILED(hr)) { qWarning("Failed to create Direct Compsition target for the window: %s", qPrintable(QSystemError::windowsComString(hr))); @@ -4899,13 +5133,13 @@ bool QD3D11SwapChain::createOrResize() swapChainFlags |= DXGI_SWAP_CHAIN_FLAG_ALLOW_TEARING; if (!swapChain) { - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount); colorFormat = DEFAULT_FORMAT; srgbAdjustedColorFormat = m_flags.testFlag(sRGB) ? DEFAULT_SRGB_FORMAT : DEFAULT_FORMAT; DXGI_COLOR_SPACE_TYPE hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G22_NONE_P709; // SDR DXGI_OUTPUT_DESC1 hdrOutputDesc; - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { // https://docs.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range if (hdrOutputDesc.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) { switch (m_format) { @@ -4945,8 +5179,9 @@ bool QD3D11SwapChain::createOrResize() desc.BufferUsage = DXGI_USAGE_RENDER_TARGET_OUTPUT; desc.BufferCount = BUFFER_COUNT; desc.Flags = swapChainFlags; - desc.Scaling = DXGI_SCALING_NONE; - desc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_DISCARD; + desc.Scaling = rhiD->useLegacySwapchainModel ? DXGI_SCALING_STRETCH : DXGI_SCALING_NONE; + desc.SwapEffect = rhiD->useLegacySwapchainModel ? DXGI_SWAP_EFFECT_DISCARD : DXGI_SWAP_EFFECT_FLIP_DISCARD; + desc.Stereo = stereo; if (dcompVisual) { // With DirectComposition setting AlphaMode to STRAIGHT fails the @@ -5006,14 +5241,19 @@ bool QD3D11SwapChain::createOrResize() qWarning("Failed to set content for Direct Composition visual: %s", qPrintable(QSystemError::windowsComString(hr))); } + } else { + // disable Alt+Enter; not relevant when using DirectComposition + rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); } } if (FAILED(hr)) { - qWarning("Failed to create D3D11 swapchain: %s", - qPrintable(QSystemError::windowsComString(hr))); + qWarning("Failed to create D3D11 swapchain: %s" + " (Width=%u Height=%u Format=%u SampleCount=%u BufferCount=%u Scaling=%u SwapEffect=%u Stereo=%u)", + qPrintable(QSystemError::windowsComString(hr)), + desc.Width, desc.Height, UINT(desc.Format), desc.SampleDesc.Count, + desc.BufferCount, UINT(desc.Scaling), UINT(desc.SwapEffect), UINT(desc.Stereo)); return false; } - rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); } else { releaseBuffers(); // flip model -> buffer count is the real buffer count, not 1 like with the legacy modes @@ -5060,6 +5300,19 @@ bool QD3D11SwapChain::createOrResize() return false; } + if (stereo) { + // Create a second render target view for the right eye + rtvDesc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.FirstArraySlice = 1; + rtvDesc.Texture2DArray.ArraySize = 1; + hr = rhiD->dev->CreateRenderTargetView(backBufferTex, &rtvDesc, &backBufferRtvRight); + if (FAILED(hr)) { + qWarning("Failed to create rtv for swapchain backbuffer (right eye): %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + } + // Try to reduce stalls by having a dedicated MSAA texture per swapchain buffer. for (int i = 0; i < BUFFER_COUNT; ++i) { if (sampleDesc.Count > 1) { @@ -5098,31 +5351,20 @@ bool QD3D11SwapChain::createOrResize() rtD->d.colorAttCount = 1; rtD->d.dsAttCount = m_depthStencil ? 1 : 0; - if (rhiD->hasGpuFrameTimeCallback()) { - D3D11_QUERY_DESC queryDesc = {}; - for (int i = 0; i < BUFFER_COUNT; ++i) { - if (!timestampDisjointQuery[i]) { - queryDesc.Query = D3D11_QUERY_TIMESTAMP_DISJOINT; - hr = rhiD->dev->CreateQuery(&queryDesc, ×tampDisjointQuery[i]); - if (FAILED(hr)) { - qWarning("Failed to create timestamp disjoint query: %s", - qPrintable(QSystemError::windowsComString(hr))); - break; - } - } - queryDesc.Query = D3D11_QUERY_TIMESTAMP; - for (int j = 0; j < 2; ++j) { - const int idx = BUFFER_COUNT * i + j; // one pair per buffer (frame) - if (!timestampQuery[idx]) { - hr = rhiD->dev->CreateQuery(&queryDesc, ×tampQuery[idx]); - if (FAILED(hr)) { - qWarning("Failed to create timestamp query: %s", - qPrintable(QSystemError::windowsComString(hr))); - break; - } - } - } - } + if (stereo) { + rtD = QRHI_RES(QD3D11SwapChainRenderTarget, &rtRight); + rtD->d.rp = QRHI_RES(QD3D11RenderPassDescriptor, m_renderPassDesc); + rtD->d.pixelSize = pixelSize; + rtD->d.dpr = float(window->devicePixelRatio()); + rtD->d.sampleCount = int(sampleDesc.Count); + rtD->d.colorAttCount = 1; + rtD->d.dsAttCount = m_depthStencil ? 1 : 0; + rtD->d.rtv[0] = backBufferRtvRight; + rtD->d.dsv = ds ? ds->dsv : nullptr; + } + + if (rhiD->rhiFlags.testFlag(QRhi::EnableTimestamps)) { + timestamps.prepare(rhiD); // timestamp queries are optional so we can go on even if they failed } diff --git a/src/gui/rhi/qrhid3d11_p.h b/src/gui/rhi/qrhid3d11_p.h index 2ab3e9fea3..7644748407 100644 --- a/src/gui/rhi/qrhid3d11_p.h +++ b/src/gui/rhi/qrhid3d11_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHID3D11_H -#define QRHID3D11_H +#ifndef QRHID3D11_P_H +#define QRHID3D11_P_H // // W A R N I N G @@ -15,28 +15,852 @@ // We mean it. // -#include <private/qrhi_p.h> +#include "qrhi_p.h" +#include <rhi/qshaderdescription.h> +#include <QWindow> -// no d3d includes here, to prevent precompiled header mess due to COM +#include <d3d11_1.h> +#include <dxgi1_6.h> +#include <dcomp.h> QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiD3D11InitParams : public QRhiInitParams +class QRhiD3D11; + +struct QD3D11Buffer : public QRhiBuffer +{ + QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QD3D11Buffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + ID3D11UnorderedAccessView *unorderedAccessView(quint32 offset); + + ID3D11Buffer *buffer = nullptr; + char *dynBuf = nullptr; + bool hasPendingDynamicUpdates = false; + QHash<quint32, ID3D11UnorderedAccessView *> uavs; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11RenderBuffer : public QRhiRenderBuffer +{ + QD3D11RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QD3D11RenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + ID3D11Texture2D *tex = nullptr; + ID3D11DepthStencilView *dsv = nullptr; + ID3D11RenderTargetView *rtv = nullptr; + DXGI_FORMAT dxgiFormat; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11Texture : public QRhiTexture +{ + QD3D11Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QD3D11Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + ID3D11UnorderedAccessView *unorderedAccessViewForLevel(int level); + ID3D11Resource *textureResource() const + { + if (tex) + return tex; + else if (tex1D) + return tex1D; + return tex3D; + } + + ID3D11Texture2D *tex = nullptr; + ID3D11Texture3D *tex3D = nullptr; + ID3D11Texture1D *tex1D = nullptr; + bool owns = true; + ID3D11ShaderResourceView *srv = nullptr; + DXGI_FORMAT dxgiFormat; + uint mipLevelCount = 0; + DXGI_SAMPLE_DESC sampleDesc; + ID3D11UnorderedAccessView *perLevelViews[QRhi::MAX_MIP_LEVELS]; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11Sampler : public QRhiSampler +{ + QD3D11Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QD3D11Sampler(); + void destroy() override; + bool create() override; + + ID3D11SamplerState *samplerState = nullptr; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QD3D11RenderPassDescriptor(QRhiImplementation *rhi); + ~QD3D11RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QD3D11RenderTargetData +{ + QD3D11RenderTargetData(QRhiImplementation *) + { + for (int i = 0; i < MAX_COLOR_ATTACHMENTS; ++i) + rtv[i] = nullptr; + } + + QD3D11RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + + static const int MAX_COLOR_ATTACHMENTS = 8; + ID3D11RenderTargetView *rtv[MAX_COLOR_ATTACHMENTS]; + ID3D11DepthStencilView *dsv = nullptr; + + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; +}; + +struct QD3D11SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QD3D11SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QD3D11SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QD3D11RenderTargetData d; +}; + +struct QD3D11TextureRenderTarget : public QRhiTextureRenderTarget +{ + QD3D11TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QD3D11TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QD3D11RenderTargetData d; + bool ownsRtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; + ID3D11RenderTargetView *rtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; + bool ownsDsv = false; + ID3D11DepthStencilView *dsv = nullptr; + friend class QRhiD3D11; +}; + +struct QD3D11ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QD3D11ShaderResourceBindings(QRhiImplementation *rhi); + ~QD3D11ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + bool hasDynamicOffset = false; + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + uint generation = 0; + + // Keep track of the generation number of each referenced QRhi* to be able + // to detect that the batched bindings are out of date. + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData; + + struct StageUniformBufferBatches { + bool present = false; + QRhiBatchedBindings<ID3D11Buffer *> ubufs; + QRhiBatchedBindings<UINT> ubuforigbindings; + QRhiBatchedBindings<UINT> ubufoffsets; + QRhiBatchedBindings<UINT> ubufsizes; + void finish() { + present = ubufs.finish(); + ubuforigbindings.finish(); + ubufoffsets.finish(); + ubufsizes.finish(); + } + void clear() { + ubufs.clear(); + ubuforigbindings.clear(); + ubufoffsets.clear(); + ubufsizes.clear(); + } + }; + + struct StageSamplerBatches { + bool present = false; + QRhiBatchedBindings<ID3D11SamplerState *> samplers; + QRhiBatchedBindings<ID3D11ShaderResourceView *> shaderresources; + void finish() { + present = samplers.finish(); + shaderresources.finish(); + } + void clear() { + samplers.clear(); + shaderresources.clear(); + } + }; + + struct StageUavBatches { + bool present = false; + QRhiBatchedBindings<ID3D11UnorderedAccessView *> uavs; + void finish() { + present = uavs.finish(); + } + void clear() { + uavs.clear(); + } + }; + + StageUniformBufferBatches vsUniformBufferBatches; + StageUniformBufferBatches hsUniformBufferBatches; + StageUniformBufferBatches dsUniformBufferBatches; + StageUniformBufferBatches gsUniformBufferBatches; + StageUniformBufferBatches fsUniformBufferBatches; + StageUniformBufferBatches csUniformBufferBatches; + + StageSamplerBatches vsSamplerBatches; + StageSamplerBatches hsSamplerBatches; + StageSamplerBatches dsSamplerBatches; + StageSamplerBatches gsSamplerBatches; + StageSamplerBatches fsSamplerBatches; + StageSamplerBatches csSamplerBatches; + + StageUavBatches csUavBatches; + + friend class QRhiD3D11; +}; + +Q_DECLARE_TYPEINFO(QD3D11ShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); + +struct QD3D11GraphicsPipeline : public QRhiGraphicsPipeline +{ + QD3D11GraphicsPipeline(QRhiImplementation *rhi); + ~QD3D11GraphicsPipeline(); + void destroy() override; + bool create() override; + + ID3D11DepthStencilState *dsState = nullptr; + ID3D11BlendState *blendState = nullptr; + struct { + ID3D11VertexShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } vs; + struct { + ID3D11HullShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } hs; + struct { + ID3D11DomainShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } ds; + struct { + ID3D11GeometryShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } gs; + struct { + ID3D11PixelShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } fs; + ID3D11InputLayout *inputLayout = nullptr; + D3D11_PRIMITIVE_TOPOLOGY d3dTopology = D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; + ID3D11RasterizerState *rastState = nullptr; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11ComputePipeline : public QRhiComputePipeline { - bool enableDebugLayer = false; + QD3D11ComputePipeline(QRhiImplementation *rhi); + ~QD3D11ComputePipeline(); + void destroy() override; + bool create() override; + + struct { + ID3D11ComputeShader *shader = nullptr; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + } cs; + uint generation = 0; + friend class QRhiD3D11; +}; + +struct QD3D11SwapChain; + +struct QD3D11CommandBuffer : public QRhiCommandBuffer +{ + QD3D11CommandBuffer(QRhiImplementation *rhi); + ~QD3D11CommandBuffer(); + void destroy() override; + + // these must be kept at a reasonably low value otherwise sizeof Command explodes + static const int MAX_DYNAMIC_OFFSET_COUNT = 8; + static const int MAX_VERTEX_BUFFER_BINDING_COUNT = 8; + + struct Command { + enum Cmd { + BeginFrame, + EndFrame, + ResetShaderResources, + SetRenderTarget, + Clear, + Viewport, + Scissor, + BindVertexBuffers, + BindIndexBuffer, + BindGraphicsPipeline, + BindShaderResources, + StencilRef, + BlendConstants, + Draw, + DrawIndexed, + UpdateSubRes, + CopySubRes, + ResolveSubRes, + GenMip, + DebugMarkBegin, + DebugMarkEnd, + DebugMarkMsg, + BindComputePipeline, + Dispatch + }; + enum ClearFlag { Color = 1, Depth = 2, Stencil = 4 }; + Cmd cmd; + + // QRhi*/QD3D11* references should be kept at minimum (so no + // QRhiTexture/Buffer/etc. pointers). + union Args { + struct { + ID3D11Query *tsQuery; + ID3D11Query *tsDisjointQuery; + QD3D11RenderTargetData *swapchainData; + } beginFrame; + struct { + ID3D11Query *tsQuery; + ID3D11Query *tsDisjointQuery; + } endFrame; + struct { + QRhiRenderTarget *rt; + } setRenderTarget; + struct { + QRhiRenderTarget *rt; + int mask; + float c[4]; + float d; + quint32 s; + } clear; + struct { + float x, y, w, h; + float d0, d1; + } viewport; + struct { + int x, y, w, h; + } scissor; + struct { + int startSlot; + int slotCount; + ID3D11Buffer *buffers[MAX_VERTEX_BUFFER_BINDING_COUNT]; + UINT offsets[MAX_VERTEX_BUFFER_BINDING_COUNT]; + UINT strides[MAX_VERTEX_BUFFER_BINDING_COUNT]; + } bindVertexBuffers; + struct { + ID3D11Buffer *buffer; + quint32 offset; + DXGI_FORMAT format; + } bindIndexBuffer; + struct { + QD3D11GraphicsPipeline *ps; + } bindGraphicsPipeline; + struct { + QD3D11ShaderResourceBindings *srb; + bool offsetOnlyChange; + int dynamicOffsetCount; + uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offsetInConstants + } bindShaderResources; + struct { + QD3D11GraphicsPipeline *ps; + quint32 ref; + } stencilRef; + struct { + QD3D11GraphicsPipeline *ps; + float c[4]; + } blendConstants; + struct { + QD3D11GraphicsPipeline *ps; + quint32 vertexCount; + quint32 instanceCount; + quint32 firstVertex; + quint32 firstInstance; + } draw; + struct { + QD3D11GraphicsPipeline *ps; + quint32 indexCount; + quint32 instanceCount; + quint32 firstIndex; + qint32 vertexOffset; + quint32 firstInstance; + } drawIndexed; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + bool hasDstBox; + D3D11_BOX dstBox; + const void *src; // must come from retain*() + UINT srcRowPitch; + } updateSubRes; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + UINT dstX; + UINT dstY; + UINT dstZ; + ID3D11Resource *src; + UINT srcSubRes; + bool hasSrcBox; + D3D11_BOX srcBox; + } copySubRes; + struct { + ID3D11Resource *dst; + UINT dstSubRes; + ID3D11Resource *src; + UINT srcSubRes; + DXGI_FORMAT format; + } resolveSubRes; + struct { + ID3D11ShaderResourceView *srv; + } genMip; + struct { + char s[64]; + } debugMark; + struct { + QD3D11ComputePipeline *ps; + } bindComputePipeline; + struct { + UINT x; + UINT y; + UINT z; + } dispatch; + } args; + }; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + QRhiBackendCommandList<Command> commands; + PassType recordingPass; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + ID3D11Buffer *currentIndexBuffer; + quint32 currentIndexOffset; + DXGI_FORMAT currentIndexFormat; + ID3D11Buffer *currentVertexBuffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + quint32 currentVertexOffsets[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + + QVarLengthArray<QByteArray, 4> dataRetainPool; + QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; + QVarLengthArray<QImage, 4> imageRetainPool; + + // relies heavily on implicit sharing (no copies of the actual data will be made) + const uchar *retainData(const QByteArray &data) { + dataRetainPool.append(data); + return reinterpret_cast<const uchar *>(dataRetainPool.last().constData()); + } + const uchar *retainBufferData(const QRhiBufferData &data) { + bufferDataRetainPool.append(data); + return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); + } + const uchar *retainImage(const QImage &image) { + imageRetainPool.append(image); + return imageRetainPool.last().constBits(); + } + void resetCommands() { + commands.reset(); + dataRetainPool.clear(); + bufferDataRetainPool.clear(); + imageRetainPool.clear(); + } + void resetState() { + recordingPass = NoPass; + // do not zero lastGpuTime + currentTarget = nullptr; + resetCommands(); + resetCachedState(); + } + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentIndexBuffer = nullptr; + currentIndexOffset = 0; + currentIndexFormat = DXGI_FORMAT_R16_UINT; + memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); + memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); + } +}; + +struct QD3D11SwapChainTimestamps +{ + static const int TIMESTAMP_PAIRS = 2; + + bool active[TIMESTAMP_PAIRS] = {}; + ID3D11Query *disjointQuery[TIMESTAMP_PAIRS] = {}; + ID3D11Query *query[TIMESTAMP_PAIRS * 2] = {}; + + bool prepare(QRhiD3D11 *rhiD); + void destroy(); + bool tryQueryTimestamps(int idx, ID3D11DeviceContext *context, double *elapsedSec); +}; + +struct QD3D11SwapChain : public QRhiSwapChain +{ + QD3D11SwapChain(QRhiImplementation *rhi); + ~QD3D11SwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + QRhiSwapChainHdrInfo hdrInfo() override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void releaseBuffers(); + bool newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, + ID3D11Texture2D **tex, ID3D11RenderTargetView **rtv) const; + + QWindow *window = nullptr; + QSize pixelSize; + QD3D11SwapChainRenderTarget rt; + QD3D11SwapChainRenderTarget rtRight; + QD3D11CommandBuffer cb; + DXGI_FORMAT colorFormat; + DXGI_FORMAT srgbAdjustedColorFormat; + IDXGISwapChain *swapChain = nullptr; + UINT swapChainFlags = 0; + ID3D11Texture2D *backBufferTex; + ID3D11RenderTargetView *backBufferRtv; + ID3D11RenderTargetView *backBufferRtvRight = nullptr; + static const int BUFFER_COUNT = 2; + ID3D11Texture2D *msaaTex[BUFFER_COUNT]; + ID3D11RenderTargetView *msaaRtv[BUFFER_COUNT]; + DXGI_SAMPLE_DESC sampleDesc; + int currentFrameSlot = 0; + int frameCount = 0; + QD3D11RenderBuffer *ds = nullptr; + UINT swapInterval = 1; + IDCompositionTarget *dcompTarget = nullptr; + IDCompositionVisual *dcompVisual = nullptr; + QD3D11SwapChainTimestamps timestamps; + int currentTimestampPairIndex = 0; }; -struct Q_GUI_EXPORT QRhiD3D11NativeHandles : public QRhiNativeHandles +class QRhiD3D11 : public QRhiImplementation { - // to import a device and a context - void *dev = nullptr; - void *context = nullptr; - // alternatively, to specify the device feature level and/or the adapter to use - int featureLevel = 0; - quint32 adapterLuidLow = 0; - qint32 adapterLuidHigh = 0; +public: + QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void enqueueSubresUpload(QD3D11Texture *texD, QD3D11CommandBuffer *cbD, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, + const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[]); + void executeBufferHostWrites(QD3D11Buffer *bufD); + void bindShaderResources(QD3D11ShaderResourceBindings *srbD, + const uint *dynOfsPairs, int dynOfsPairCount, + bool offsetOnlyChange); + void resetShaderResources(); + void executeCommandBuffer(QD3D11CommandBuffer *cbD); + DXGI_SAMPLE_DESC effectiveSampleDesc(int sampleCount) const; + void finishActiveReadbacks(); + void reportLiveObjects(ID3D11Device *device); + void clearShaderCache(); + QByteArray compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, uint flags, + QString *error, QShaderKey *usedShaderKey); + bool ensureDirectCompositionDevice(); + + QRhi::Flags rhiFlags; + bool debugLayer = false; + bool importedDeviceAndContext = false; + ID3D11Device *dev = nullptr; + ID3D11DeviceContext1 *context = nullptr; + D3D_FEATURE_LEVEL featureLevel = D3D_FEATURE_LEVEL(0); + LUID adapterLuid = {}; + ID3DUserDefinedAnnotation *annotations = nullptr; + IDXGIAdapter1 *activeAdapter = nullptr; + IDXGIFactory1 *dxgiFactory = nullptr; + IDCompositionDevice *dcompDevice = nullptr; + bool supportsAllowTearing = false; + bool useLegacySwapchainModel = false; + bool deviceLost = false; + QRhiD3D11NativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + + struct { + int vsHighestActiveVertexBufferBinding = -1; + bool vsHasIndexBufferBound = false; + int vsHighestActiveSrvBinding = -1; + int hsHighestActiveSrvBinding = -1; + int dsHighestActiveSrvBinding = -1; + int gsHighestActiveSrvBinding = -1; + int fsHighestActiveSrvBinding = -1; + int csHighestActiveSrvBinding = -1; + int csHighestActiveUavBinding = -1; + QD3D11SwapChain *currentSwapChain = nullptr; + } contextState; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } + bool active = false; + QD3D11CommandBuffer cbWrapper; + ID3D11Query *tsQueries[2] = {}; + ID3D11Query *tsDisjointQuery = nullptr; + } ofr; + + struct TextureReadback { + QRhiReadbackDescription desc; + QRhiReadbackResult *result; + ID3D11Texture2D *stagingTex; + quint32 byteSize; + quint32 bpl; + QSize pixelSize; + QRhiTexture::Format format; + }; + QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; + struct BufferReadback { + QRhiReadbackResult *result; + quint32 byteSize; + ID3D11Buffer *stagingBuf; + }; + QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; + + struct Shader { + Shader() = default; + Shader(IUnknown *s, const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) + : s(s), bytecode(bytecode), nativeResourceBindingMap(rbm) { } + IUnknown *s; + QByteArray bytecode; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + }; + QHash<QRhiShaderStage, Shader> m_shaderCache; + + // This is what gets exposed as the "pipeline cache", not that that concept + // applies anyway. Here we are just storing the DX bytecode for a shader so + // we can skip the HLSL->DXBC compilation when the QShader has HLSL source + // code and the same shader source has already been compiled before. + // m_shaderCache seemingly does the same, but this here does not care about + // the ID3D11*Shader, this is just about the bytecode and about allowing + // the data to be serialized to persistent storage and then reloaded in + // future runs of the app, or when creating another QRhi, etc. + struct BytecodeCacheKey { + QByteArray sourceHash; + QByteArray target; + QByteArray entryPoint; + uint compileFlags; + }; + QHash<BytecodeCacheKey, QByteArray> m_bytecodeCache; }; +Q_DECLARE_TYPEINFO(QRhiD3D11::TextureReadback, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiD3D11::BufferReadback, Q_RELOCATABLE_TYPE); + +inline bool operator==(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept +{ + return a.sourceHash == b.sourceHash + && a.target == b.target + && a.entryPoint == b.entryPoint + && a.compileFlags == b.compileFlags; +} + +inline bool operator!=(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept +{ + return !(a == b); +} + +inline size_t qHash(const QRhiD3D11::BytecodeCacheKey &k, size_t seed = 0) noexcept +{ + return qHash(k.sourceHash, seed) ^ qHash(k.target) ^ qHash(k.entryPoint) ^ k.compileFlags; +} + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhid3d11_p_p.h b/src/gui/rhi/qrhid3d11_p_p.h deleted file mode 100644 index 8b56ccbf33..0000000000 --- a/src/gui/rhi/qrhid3d11_p_p.h +++ /dev/null @@ -1,833 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHID3D11_P_H -#define QRHID3D11_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhid3d11_p.h" -#include "qrhi_p_p.h" -#include "qshaderdescription_p.h" -#include <QWindow> - -#include <d3d11_1.h> -#include <dxgi1_6.h> -#include <dcomp.h> - -QT_BEGIN_NAMESPACE - -struct QD3D11Buffer : public QRhiBuffer -{ - QD3D11Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QD3D11Buffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - ID3D11UnorderedAccessView *unorderedAccessView(quint32 offset); - - ID3D11Buffer *buffer = nullptr; - char *dynBuf = nullptr; - bool hasPendingDynamicUpdates = false; - QHash<quint32, ID3D11UnorderedAccessView *> uavs; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11RenderBuffer : public QRhiRenderBuffer -{ - QD3D11RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QD3D11RenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - ID3D11Texture2D *tex = nullptr; - ID3D11DepthStencilView *dsv = nullptr; - ID3D11RenderTargetView *rtv = nullptr; - DXGI_FORMAT dxgiFormat; - DXGI_SAMPLE_DESC sampleDesc; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11Texture : public QRhiTexture -{ - QD3D11Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QD3D11Texture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - bool finishCreate(); - ID3D11UnorderedAccessView *unorderedAccessViewForLevel(int level); - ID3D11Resource *textureResource() const - { - if (tex) - return tex; - else if (tex1D) - return tex1D; - return tex3D; - } - - ID3D11Texture2D *tex = nullptr; - ID3D11Texture3D *tex3D = nullptr; - ID3D11Texture1D *tex1D = nullptr; - bool owns = true; - ID3D11ShaderResourceView *srv = nullptr; - DXGI_FORMAT dxgiFormat; - uint mipLevelCount = 0; - DXGI_SAMPLE_DESC sampleDesc; - ID3D11UnorderedAccessView *perLevelViews[QRhi::MAX_MIP_LEVELS]; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11Sampler : public QRhiSampler -{ - QD3D11Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QD3D11Sampler(); - void destroy() override; - bool create() override; - - ID3D11SamplerState *samplerState = nullptr; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11RenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QD3D11RenderPassDescriptor(QRhiImplementation *rhi); - ~QD3D11RenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; -}; - -struct QD3D11RenderTargetData -{ - QD3D11RenderTargetData(QRhiImplementation *) - { - for (int i = 0; i < MAX_COLOR_ATTACHMENTS; ++i) - rtv[i] = nullptr; - } - - QD3D11RenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - - static const int MAX_COLOR_ATTACHMENTS = 8; - ID3D11RenderTargetView *rtv[MAX_COLOR_ATTACHMENTS]; - ID3D11DepthStencilView *dsv = nullptr; - - QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; -}; - -struct QD3D11SwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QD3D11SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QD3D11SwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QD3D11RenderTargetData d; -}; - -struct QD3D11TextureRenderTarget : public QRhiTextureRenderTarget -{ - QD3D11TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QD3D11TextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QD3D11RenderTargetData d; - bool ownsRtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; - ID3D11RenderTargetView *rtv[QD3D11RenderTargetData::MAX_COLOR_ATTACHMENTS]; - bool ownsDsv = false; - ID3D11DepthStencilView *dsv = nullptr; - friend class QRhiD3D11; -}; - -struct QD3D11ShaderResourceBindings : public QRhiShaderResourceBindings -{ - QD3D11ShaderResourceBindings(QRhiImplementation *rhi); - ~QD3D11ShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; - - bool hasDynamicOffset = false; - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - uint generation = 0; - - // Keep track of the generation number of each referenced QRhi* to be able - // to detect that the batched bindings are out of date. - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData; - - struct StageUniformBufferBatches { - bool present = false; - QRhiBatchedBindings<ID3D11Buffer *> ubufs; - QRhiBatchedBindings<UINT> ubuforigbindings; - QRhiBatchedBindings<UINT> ubufoffsets; - QRhiBatchedBindings<UINT> ubufsizes; - void finish() { - present = ubufs.finish(); - ubuforigbindings.finish(); - ubufoffsets.finish(); - ubufsizes.finish(); - } - void clear() { - ubufs.clear(); - ubuforigbindings.clear(); - ubufoffsets.clear(); - ubufsizes.clear(); - } - }; - - struct StageSamplerBatches { - bool present = false; - QRhiBatchedBindings<ID3D11SamplerState *> samplers; - QRhiBatchedBindings<ID3D11ShaderResourceView *> shaderresources; - void finish() { - present = samplers.finish(); - shaderresources.finish(); - } - void clear() { - samplers.clear(); - shaderresources.clear(); - } - }; - - struct StageUavBatches { - bool present = false; - QRhiBatchedBindings<ID3D11UnorderedAccessView *> uavs; - void finish() { - present = uavs.finish(); - } - void clear() { - uavs.clear(); - } - }; - - StageUniformBufferBatches vsUniformBufferBatches; - StageUniformBufferBatches hsUniformBufferBatches; - StageUniformBufferBatches dsUniformBufferBatches; - StageUniformBufferBatches gsUniformBufferBatches; - StageUniformBufferBatches fsUniformBufferBatches; - StageUniformBufferBatches csUniformBufferBatches; - - StageSamplerBatches vsSamplerBatches; - StageSamplerBatches hsSamplerBatches; - StageSamplerBatches dsSamplerBatches; - StageSamplerBatches gsSamplerBatches; - StageSamplerBatches fsSamplerBatches; - StageSamplerBatches csSamplerBatches; - - StageUavBatches csUavBatches; - - friend class QRhiD3D11; -}; - -Q_DECLARE_TYPEINFO(QD3D11ShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); - -struct QD3D11GraphicsPipeline : public QRhiGraphicsPipeline -{ - QD3D11GraphicsPipeline(QRhiImplementation *rhi); - ~QD3D11GraphicsPipeline(); - void destroy() override; - bool create() override; - - ID3D11DepthStencilState *dsState = nullptr; - ID3D11BlendState *blendState = nullptr; - struct { - ID3D11VertexShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } vs; - struct { - ID3D11HullShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } hs; - struct { - ID3D11DomainShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } ds; - struct { - ID3D11GeometryShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } gs; - struct { - ID3D11PixelShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } fs; - ID3D11InputLayout *inputLayout = nullptr; - D3D11_PRIMITIVE_TOPOLOGY d3dTopology = D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST; - ID3D11RasterizerState *rastState = nullptr; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11ComputePipeline : public QRhiComputePipeline -{ - QD3D11ComputePipeline(QRhiImplementation *rhi); - ~QD3D11ComputePipeline(); - void destroy() override; - bool create() override; - - struct { - ID3D11ComputeShader *shader = nullptr; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - } cs; - uint generation = 0; - friend class QRhiD3D11; -}; - -struct QD3D11SwapChain; - -struct QD3D11CommandBuffer : public QRhiCommandBuffer -{ - QD3D11CommandBuffer(QRhiImplementation *rhi); - ~QD3D11CommandBuffer(); - void destroy() override; - - // these must be kept at a reasonably low value otherwise sizeof Command explodes - static const int MAX_DYNAMIC_OFFSET_COUNT = 8; - static const int MAX_VERTEX_BUFFER_BINDING_COUNT = 8; - - struct Command { - enum Cmd { - ResetShaderResources, - SetRenderTarget, - Clear, - Viewport, - Scissor, - BindVertexBuffers, - BindIndexBuffer, - BindGraphicsPipeline, - BindShaderResources, - StencilRef, - BlendConstants, - Draw, - DrawIndexed, - UpdateSubRes, - CopySubRes, - ResolveSubRes, - GenMip, - DebugMarkBegin, - DebugMarkEnd, - DebugMarkMsg, - BindComputePipeline, - Dispatch - }; - enum ClearFlag { Color = 1, Depth = 2, Stencil = 4 }; - Cmd cmd; - - // QRhi*/QD3D11* references should be kept at minimum (so no - // QRhiTexture/Buffer/etc. pointers). - union Args { - struct { - QRhiRenderTarget *rt; - } setRenderTarget; - struct { - QRhiRenderTarget *rt; - int mask; - float c[4]; - float d; - quint32 s; - } clear; - struct { - float x, y, w, h; - float d0, d1; - } viewport; - struct { - int x, y, w, h; - } scissor; - struct { - int startSlot; - int slotCount; - ID3D11Buffer *buffers[MAX_VERTEX_BUFFER_BINDING_COUNT]; - UINT offsets[MAX_VERTEX_BUFFER_BINDING_COUNT]; - UINT strides[MAX_VERTEX_BUFFER_BINDING_COUNT]; - } bindVertexBuffers; - struct { - ID3D11Buffer *buffer; - quint32 offset; - DXGI_FORMAT format; - } bindIndexBuffer; - struct { - QD3D11GraphicsPipeline *ps; - } bindGraphicsPipeline; - struct { - QD3D11ShaderResourceBindings *srb; - bool offsetOnlyChange; - int dynamicOffsetCount; - uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offsetInConstants - } bindShaderResources; - struct { - QD3D11GraphicsPipeline *ps; - quint32 ref; - } stencilRef; - struct { - QD3D11GraphicsPipeline *ps; - float c[4]; - } blendConstants; - struct { - QD3D11GraphicsPipeline *ps; - quint32 vertexCount; - quint32 instanceCount; - quint32 firstVertex; - quint32 firstInstance; - } draw; - struct { - QD3D11GraphicsPipeline *ps; - quint32 indexCount; - quint32 instanceCount; - quint32 firstIndex; - qint32 vertexOffset; - quint32 firstInstance; - } drawIndexed; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - bool hasDstBox; - D3D11_BOX dstBox; - const void *src; // must come from retain*() - UINT srcRowPitch; - } updateSubRes; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - UINT dstX; - UINT dstY; - UINT dstZ; - ID3D11Resource *src; - UINT srcSubRes; - bool hasSrcBox; - D3D11_BOX srcBox; - } copySubRes; - struct { - ID3D11Resource *dst; - UINT dstSubRes; - ID3D11Resource *src; - UINT srcSubRes; - DXGI_FORMAT format; - } resolveSubRes; - struct { - ID3D11ShaderResourceView *srv; - } genMip; - struct { - char s[64]; - } debugMark; - struct { - QD3D11ComputePipeline *ps; - } bindComputePipeline; - struct { - UINT x; - UINT y; - UINT z; - } dispatch; - } args; - }; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - QRhiBackendCommandList<Command> commands; - PassType recordingPass; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - ID3D11Buffer *currentIndexBuffer; - quint32 currentIndexOffset; - DXGI_FORMAT currentIndexFormat; - ID3D11Buffer *currentVertexBuffers[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - quint32 currentVertexOffsets[D3D11_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - - QVarLengthArray<QByteArray, 4> dataRetainPool; - QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; - QVarLengthArray<QImage, 4> imageRetainPool; - - // relies heavily on implicit sharing (no copies of the actual data will be made) - const uchar *retainData(const QByteArray &data) { - dataRetainPool.append(data); - return reinterpret_cast<const uchar *>(dataRetainPool.last().constData()); - } - const uchar *retainBufferData(const QRhiBufferData &data) { - bufferDataRetainPool.append(data); - return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); - } - const uchar *retainImage(const QImage &image) { - imageRetainPool.append(image); - return imageRetainPool.last().constBits(); - } - void resetCommands() { - commands.reset(); - dataRetainPool.clear(); - bufferDataRetainPool.clear(); - imageRetainPool.clear(); - } - void resetState() { - recordingPass = NoPass; - currentTarget = nullptr; - resetCommands(); - resetCachedState(); - } - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - currentIndexBuffer = nullptr; - currentIndexOffset = 0; - currentIndexFormat = DXGI_FORMAT_R16_UINT; - memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); - memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); - } -}; - -struct QD3D11SwapChain : public QRhiSwapChain -{ - QD3D11SwapChain(QRhiImplementation *rhi); - ~QD3D11SwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - QRhiSwapChainHdrInfo hdrInfo() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - void releaseBuffers(); - bool newColorBuffer(const QSize &size, DXGI_FORMAT format, DXGI_SAMPLE_DESC sampleDesc, - ID3D11Texture2D **tex, ID3D11RenderTargetView **rtv) const; - - QWindow *window = nullptr; - QSize pixelSize; - QD3D11SwapChainRenderTarget rt; - QD3D11CommandBuffer cb; - DXGI_FORMAT colorFormat; - DXGI_FORMAT srgbAdjustedColorFormat; - IDXGISwapChain *swapChain = nullptr; - UINT swapChainFlags = 0; - static const int BUFFER_COUNT = 2; - ID3D11Texture2D *backBufferTex; - ID3D11RenderTargetView *backBufferRtv; - ID3D11Texture2D *msaaTex[BUFFER_COUNT]; - ID3D11RenderTargetView *msaaRtv[BUFFER_COUNT]; - DXGI_SAMPLE_DESC sampleDesc; - int currentFrameSlot = 0; - int frameCount = 0; - QD3D11RenderBuffer *ds = nullptr; - bool timestampActive[BUFFER_COUNT]; - ID3D11Query *timestampDisjointQuery[BUFFER_COUNT]; - ID3D11Query *timestampQuery[BUFFER_COUNT * 2]; - UINT swapInterval = 1; - IDCompositionTarget *dcompTarget = nullptr; - IDCompositionVisual *dcompVisual = nullptr; -}; - -class QRhiD3D11 : public QRhiImplementation -{ -public: - QRhiD3D11(QRhiD3D11InitParams *params, QRhiD3D11NativeHandles *importDevice = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void enqueueSubresUpload(QD3D11Texture *texD, QD3D11CommandBuffer *cbD, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void updateShaderResourceBindings(QD3D11ShaderResourceBindings *srbD, - const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[]); - void executeBufferHostWrites(QD3D11Buffer *bufD); - void bindShaderResources(QD3D11ShaderResourceBindings *srbD, - const uint *dynOfsPairs, int dynOfsPairCount, - bool offsetOnlyChange); - void resetShaderResources(); - void executeCommandBuffer(QD3D11CommandBuffer *cbD, QD3D11SwapChain *timestampSwapChain = nullptr); - DXGI_SAMPLE_DESC effectiveSampleCount(int sampleCount) const; - void finishActiveReadbacks(); - void reportLiveObjects(ID3D11Device *device); - void clearShaderCache(); - QByteArray compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, uint flags, - QString *error, QShaderKey *usedShaderKey); - bool ensureDirectCompositionDevice(); - - QRhi::Flags rhiFlags; - bool debugLayer = false; - bool importedDeviceAndContext = false; - ID3D11Device *dev = nullptr; - ID3D11DeviceContext1 *context = nullptr; - D3D_FEATURE_LEVEL featureLevel = D3D_FEATURE_LEVEL(0); - LUID adapterLuid = {}; - ID3DUserDefinedAnnotation *annotations = nullptr; - IDXGIAdapter1 *activeAdapter = nullptr; - IDXGIFactory1 *dxgiFactory = nullptr; - IDCompositionDevice *dcompDevice = nullptr; - bool supportsAllowTearing = false; - bool deviceLost = false; - QRhiD3D11NativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - - struct { - int vsHighestActiveVertexBufferBinding = -1; - bool vsHasIndexBufferBound = false; - int vsHighestActiveSrvBinding = -1; - int hsHighestActiveSrvBinding = -1; - int dsHighestActiveSrvBinding = -1; - int gsHighestActiveSrvBinding = -1; - int fsHighestActiveSrvBinding = -1; - int csHighestActiveSrvBinding = -1; - int csHighestActiveUavBinding = -1; - QD3D11SwapChain *currentSwapChain = nullptr; - } contextState; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } - bool active = false; - QD3D11CommandBuffer cbWrapper; - } ofr; - - struct TextureReadback { - QRhiReadbackDescription desc; - QRhiReadbackResult *result; - ID3D11Texture2D *stagingTex; - quint32 byteSize; - quint32 bpl; - QSize pixelSize; - QRhiTexture::Format format; - }; - QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; - struct BufferReadback { - QRhiBufferReadbackResult *result; - quint32 byteSize; - ID3D11Buffer *stagingBuf; - }; - QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; - - struct Shader { - Shader() = default; - Shader(IUnknown *s, const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) - : s(s), bytecode(bytecode), nativeResourceBindingMap(rbm) { } - IUnknown *s; - QByteArray bytecode; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - }; - QHash<QRhiShaderStage, Shader> m_shaderCache; - - // This is what gets exposed as the "pipeline cache", not that that concept - // applies anyway. Here we are just storing the DX bytecode for a shader so - // we can skip the HLSL->DXBC compilation when the QShader has HLSL source - // code and the same shader source has already been compiled before. - // m_shaderCache seemingly does the same, but this here does not care about - // the ID3D11*Shader, this is just about the bytecode and about allowing - // the data to be serialized to persistent storage and then reloaded in - // future runs of the app, or when creating another QRhi, etc. - struct BytecodeCacheKey { - QByteArray sourceHash; - QByteArray target; - QByteArray entryPoint; - uint compileFlags; - }; - QHash<BytecodeCacheKey, QByteArray> m_bytecodeCache; -}; - -Q_DECLARE_TYPEINFO(QRhiD3D11::TextureReadback, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiD3D11::BufferReadback, Q_RELOCATABLE_TYPE); - -inline bool operator==(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept -{ - return a.sourceHash == b.sourceHash - && a.target == b.target - && a.entryPoint == b.entryPoint - && a.compileFlags == b.compileFlags; -} - -inline bool operator!=(const QRhiD3D11::BytecodeCacheKey &a, const QRhiD3D11::BytecodeCacheKey &b) noexcept -{ - return !(a == b); -} - -inline size_t qHash(const QRhiD3D11::BytecodeCacheKey &k, size_t seed = 0) noexcept -{ - return qHash(k.sourceHash, seed) ^ qHash(k.target) ^ qHash(k.entryPoint) ^ k.compileFlags; -} - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhid3d12.cpp b/src/gui/rhi/qrhid3d12.cpp index 20d01db1fa..0f176c683d 100644 --- a/src/gui/rhi/qrhid3d12.cpp +++ b/src/gui/rhi/qrhid3d12.cpp @@ -1,19 +1,20 @@ -// Copyright (C) 2022 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhid3d12_p_p.h" -#include "qshader_p.h" -#include <QWindow> +#include "qrhid3d12_p.h" #include <qmath.h> -#include <private/qsystemlibrary_p.h> -#include <QtCore/qcryptographichash.h> #include <QtCore/private/qsystemerror_p.h> - -#include <d3dcompiler.h> #include <comdef.h> - +#include "qrhid3dhelpers_p.h" #include "cs_mipmap_p.h" +#if __has_include(<pix.h>) +#include <pix.h> +#define QRHI_D3D12_HAS_OLD_PIX +#endif + +#ifdef __ID3D12Device2_INTERFACE_DEFINED__ + QT_BEGIN_NAMESPACE /* @@ -25,6 +26,9 @@ QT_BEGIN_NAMESPACE \inmodule QtGui \brief Direct3D 12 specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A D3D12-based QRhi needs no special parameters for initialization. If desired, enableDebugLayer can be set to \c true to enable the Direct3D debug layer. This can be useful during development, but should be avoided @@ -58,27 +62,90 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiD3D12InitParams::enableDebugLayer + + When set to true, the debug layer is enabled, if installed and available. + The default value is false. +*/ + +/*! \class QRhiD3D12NativeHandles \inmodule QtGui \brief Holds the D3D12 device used by the QRhi. \note The class uses \c{void *} as the type since including the COM-based - \c{d3d12.h} headers is not acceptable here. The actual type is - \c{ID3D12Device *}. + \c{d3d12.h} headers is not acceptable here. The actual types are + \c{ID3D12Device *} and \c{ID3D12CommandQueue *}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiD3D12NativeHandles::dev + + Points to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nn-d3d12-id3d12device}{ID3D12Device} + or left set to \nullptr if no existing device is to be imported. +*/ + +/*! + \variable QRhiD3D12NativeHandles::minimumFeatureLevel + + Specifies the \b minimum feature level passed to + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-d3d12createdevice}{D3D12CreateDevice()}. + When not set, \c{D3D_FEATURE_LEVEL_11_0} is used. See + \l{https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-levels}{this + page} for details. + + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::adapterLuidLow + + The low part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::adapterLuidHigh + + The high part of the local identifier (LUID) of the DXGI adapter to use. + Relevant only when QRhi creates the device, ignored when importing a device + and device context. +*/ + +/*! + \variable QRhiD3D12NativeHandles::commandQueue + + When set, must point to a + \l{https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nn-d3d12-id3d12commandqueue}{ID3D12CommandQueue}. + It allows to optionally import a command queue as well, in addition to a + device. +*/ + +/*! \class QRhiD3D12CommandBufferNativeHandles \inmodule QtGui - \brief Holds the ID3D12GraphicsCommandList object that is backing a QRhiCommandBuffer. + \brief Holds the ID3D12GraphicsCommandList1 object that is backing a QRhiCommandBuffer. \note The command list object is only guaranteed to be valid, and in recording state, while recording a frame. That is, between a \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} or \l{QRhi::beginOffscreenFrame()}{beginOffscreenFrame()} - \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} pair. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiD3D12CommandBufferNativeHandles::commandList +*/ + // https://learn.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-levels static const D3D_FEATURE_LEVEL MIN_FEATURE_LEVEL = D3D_FEATURE_LEVEL_11_0; @@ -87,8 +154,14 @@ QRhiD3D12::QRhiD3D12(QRhiD3D12InitParams *params, QRhiD3D12NativeHandles *import debugLayer = params->enableDebugLayer; if (importParams) { if (importParams->dev) { - dev = reinterpret_cast<ID3D12Device *>(importParams->dev); - importedDevice = true; + ID3D12Device *d3d12Device = reinterpret_cast<ID3D12Device *>(importParams->dev); + if (SUCCEEDED(d3d12Device->QueryInterface(__uuidof(ID3D12Device2), reinterpret_cast<void **>(&dev)))) { + // get rid of the ref added by QueryInterface + d3d12Device->Release(); + importedDevice = true; + } else { + qWarning("ID3D12Device2 not supported, cannot import device"); + } } if (importParams->commandQueue) { cmdQueue = reinterpret_cast<ID3D12CommandQueue *>(importParams->commandQueue); @@ -134,9 +207,20 @@ bool QRhiD3D12::create(QRhi::Flags flags) factoryFlags |= DXGI_CREATE_FACTORY_DEBUG; HRESULT hr = CreateDXGIFactory2(factoryFlags, __uuidof(IDXGIFactory2), reinterpret_cast<void **>(&dxgiFactory)); if (FAILED(hr)) { - qWarning("CreateDXGIFactory2() failed to create DXGI factory: %s", - qPrintable(QSystemError::windowsComString(hr))); - return false; + // retry without debug, if it was requested (to match D3D11 backend behavior) + if (debugLayer) { + qCDebug(QRHI_LOG_INFO, "Debug layer was requested but is not available. " + "Attempting to create DXGIFactory2 without it."); + factoryFlags &= ~DXGI_CREATE_FACTORY_DEBUG; + hr = CreateDXGIFactory2(factoryFlags, __uuidof(IDXGIFactory2), reinterpret_cast<void **>(&dxgiFactory)); + } + if (SUCCEEDED(hr)) { + debugLayer = false; + } else { + qWarning("CreateDXGIFactory2() failed to create DXGI factory: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } } supportsAllowTearing = false; @@ -204,9 +288,7 @@ bool QRhiD3D12::create(QRhi::Flags flags) if (!activeAdapter && (requestedAdapterIndex < 0 || requestedAdapterIndex == adapterIndex)) { activeAdapter = adapter; adapterLuid = desc.AdapterLuid; - driverInfoStruct.deviceName = name.toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); qCDebug(QRHI_LOG_INFO, " using this adapter"); } else { adapter->Release(); @@ -222,7 +304,7 @@ bool QRhiD3D12::create(QRhi::Flags flags) hr = D3D12CreateDevice(activeAdapter, minimumFeatureLevel, - __uuidof(ID3D12Device), + __uuidof(ID3D12Device2), reinterpret_cast<void **>(&dev)); if (FAILED(hr)) { qWarning("Failed to create D3D12 device: %s", qPrintable(QSystemError::windowsComString(hr))); @@ -236,16 +318,20 @@ bool QRhiD3D12::create(QRhi::Flags flags) for (int adapterIndex = 0; dxgiFactory->EnumAdapters1(UINT(adapterIndex), &adapter) != DXGI_ERROR_NOT_FOUND; ++adapterIndex) { DXGI_ADAPTER_DESC1 desc; adapter->GetDesc1(&desc); - adapter->Release(); if (desc.AdapterLuid.LowPart == adapterLuid.LowPart && desc.AdapterLuid.HighPart == adapterLuid.HighPart) { - driverInfoStruct.deviceName = QString::fromUtf16(reinterpret_cast<char16_t *>(desc.Description)).toUtf8(); - driverInfoStruct.deviceId = desc.DeviceId; - driverInfoStruct.vendorId = desc.VendorId; + activeAdapter = adapter; + QRhiD3D::fillDriverInfo(&driverInfoStruct, desc); break; + } else { + adapter->Release(); } } + if (!activeAdapter) { + qWarning("No adapter"); + return false; + } qCDebug(QRHI_LOG_INFO, "Using imported device %p", dev); } @@ -347,6 +433,9 @@ bool QRhiD3D12::create(QRhi::Flags flags) qWarning("Could not create host-visible staging area"); return false; } + QString decoratedName = QLatin1String("Small staging area buffer/"); + decoratedName += QString::number(i); + smallStagingAreas[i].mem.buffer->SetName(reinterpret_cast<LPCWSTR>(decoratedName.utf16())); } if (!shaderVisibleCbvSrvUavHeap.create(dev, @@ -357,6 +446,53 @@ bool QRhiD3D12::create(QRhi::Flags flags) return false; } + if (flags.testFlag(QRhi::EnableTimestamps)) { + static bool wantsStablePowerState = qEnvironmentVariableIntValue("QT_D3D_STABLE_POWER_STATE"); + // + // https://learn.microsoft.com/en-us/windows/win32/api/d3d12/nf-d3d12-id3d12device-setstablepowerstate + // + // NB! This is a _global_ setting, affecting other processes (and 3D + // APIs such as Vulkan), as long as this application is running. Hence + // making it an env.var. for now. Never enable it in production. But + // extremely useful for the GPU timings with NVIDIA at least; the + // timestamps become stable and smooth, making the number readable and + // actually useful e.g. in Quick 3D's DebugView when this is enabled. + // (otherwise the number's all over the place) + // + // See also + // https://developer.nvidia.com/blog/advanced-api-performance-setstablepowerstate/ + // for possible other approaches. + // + if (wantsStablePowerState) + dev->SetStablePowerState(TRUE); + + hr = cmdQueue->GetTimestampFrequency(×tampTicksPerSecond); + if (FAILED(hr)) { + qWarning("Failed to query timestamp frequency: %s", + qPrintable(QSystemError::windowsComString(hr))); + return false; + } + if (!timestampQueryHeap.create(dev, QD3D12_FRAMES_IN_FLIGHT * 2, D3D12_QUERY_HEAP_TYPE_TIMESTAMP)) { + qWarning("Failed to create timestamp query pool"); + return false; + } + const quint32 readbackBufSize = QD3D12_FRAMES_IN_FLIGHT * 2 * sizeof(quint64); + if (!timestampReadbackArea.create(this, readbackBufSize, D3D12_HEAP_TYPE_READBACK)) { + qWarning("Failed to create timestamp readback buffer"); + return false; + } + timestampReadbackArea.mem.buffer->SetName(L"Timestamp readback buffer"); + memset(timestampReadbackArea.mem.p, 0, readbackBufSize); + } + + caps = {}; + D3D12_FEATURE_DATA_D3D12_OPTIONS3 options3 = {}; + if (SUCCEEDED(dev->CheckFeatureSupport(D3D12_FEATURE_D3D12_OPTIONS3, &options3, sizeof(options3)))) { + caps.multiView = options3.ViewInstancingTier != D3D12_VIEW_INSTANCING_TIER_NOT_SUPPORTED; + // https://microsoft.github.io/DirectX-Specs/d3d/RelaxedCasting.html + caps.textureViewFormat = options3.CastingFullyTypedFormatSupported; + } + deviceLost = false; offscreenActive = false; @@ -385,6 +521,9 @@ void QRhiD3D12::destroy() } } + timestampQueryHeap.destroy(); + timestampReadbackArea.destroy(); + shaderVisibleCbvSrvUavHeap.destroy(); for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) @@ -400,8 +539,10 @@ void QRhiD3D12::destroy() cbvSrvUavPool.destroy(); for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { - cmdAllocators[i]->Release(); - cmdAllocators[i] = nullptr; + if (cmdAllocators[i]) { + cmdAllocators[i]->Release(); + cmdAllocators[i] = nullptr; + } } if (fullFenceEvent) { @@ -514,9 +655,13 @@ bool QRhiD3D12::isFeatureSupported(QRhi::Feature feature) const case QRhi::MultisampleRenderBuffer: return true; case QRhi::DebugMarkers: - return false; // ### +#ifdef QRHI_D3D12_HAS_OLD_PIX + return true; +#else + return false; +#endif case QRhi::Timestamps: - return false; // ### + return true; case QRhi::Instancing: return true; case QRhi::CustomInstanceStepRate: @@ -589,6 +734,14 @@ bool QRhiD3D12::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::ThreeDimensionalTextureMipmaps: return false; // we generate mipmaps ourselves with compute and this is not implemented + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return caps.textureViewFormat; + case QRhi::ResolveDepthStencil: + // there is no Multisample Resolve support for depth/stencil formats + // https://learn.microsoft.com/en-us/windows/win32/direct3ddxgi/hardware-support-for-direct3d-12-1-formats + return false; } return false; } @@ -745,15 +898,18 @@ void QRhiD3D12::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline } cbD->cmdList->IASetPrimitiveTopology(psD->topology); + + if (psD->viewInstanceMask) + cbD->cmdList->SetViewInstanceMask(psD->viewInstanceMask); } } -void QRhiD3D12::visitUniformBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::UniformBufferData &d, - int, - int binding, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) +void QD3D12CommandBuffer::visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int, + int binding, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) { QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, d.buf); quint32 offset = d.offset; @@ -766,29 +922,30 @@ void QRhiD3D12::visitUniformBuffer(QD3D12Stage s, } } } - visitorData.cbufs[s].append({ bufD->handles[currentFrameSlot], offset }); + QRHI_RES_RHI(QRhiD3D12); + visitorData.cbufs[s].append({ bufD->handles[rhiD->currentFrameSlot], offset }); } -void QRhiD3D12::visitTexture(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int) +void QD3D12CommandBuffer::visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int) { QD3D12Texture *texD = QRHI_RES(QD3D12Texture, d.tex); visitorData.srvs[s].append(texD->srv); } -void QRhiD3D12::visitSampler(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int) +void QD3D12CommandBuffer::visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int) { QD3D12Sampler *samplerD = QRHI_RES(QD3D12Sampler, d.sampler); visitorData.samplers[s].append(samplerD->lookupOrCreateShaderVisibleDescriptor()); } -void QRhiD3D12::visitStorageBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageBufferData &d, - QD3D12ShaderResourceVisitor::StorageOp, - int) +void QD3D12CommandBuffer::visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp, + int) { QD3D12Buffer *bufD = QRHI_RES(QD3D12Buffer, d.buf); // SPIRV-Cross generated HLSL uses RWByteAddressBuffer @@ -801,17 +958,17 @@ void QRhiD3D12::visitStorageBuffer(QD3D12Stage s, visitorData.uavs[s].append({ bufD->handles[0], uavDesc }); } -void QRhiD3D12::visitStorageImage(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageImageData &d, - QD3D12ShaderResourceVisitor::StorageOp, - int) +void QD3D12CommandBuffer::visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp, + int) { QD3D12Texture *texD = QRHI_RES(QD3D12Texture, d.tex); const bool isCube = texD->m_flags.testFlag(QRhiTexture::CubeMap); const bool isArray = texD->m_flags.testFlag(QRhiTexture::TextureArray); const bool is3D = texD->m_flags.testFlag(QRhiTexture::ThreeDimensional); D3D12_UNORDERED_ACCESS_VIEW_DESC uavDesc = {}; - uavDesc.Format = texD->dxgiFormat; + uavDesc.Format = texD->rtFormat; if (isCube) { uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY; uavDesc.Texture2DArray.MipSlice = UINT(d.level); @@ -821,7 +978,7 @@ void QRhiD3D12::visitStorageImage(QD3D12Stage s, uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE2DARRAY; uavDesc.Texture2DArray.MipSlice = UINT(d.level); uavDesc.Texture2DArray.FirstArraySlice = 0; - uavDesc.Texture2DArray.ArraySize = UINT(texD->m_arraySize); + uavDesc.Texture2DArray.ArraySize = UINT(qMax(0, texD->m_arraySize)); } else if (is3D) { uavDesc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE3D; uavDesc.Texture3D.MipSlice = UINT(d.level); @@ -850,8 +1007,8 @@ void QRhiD3D12::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind QD3D12ShaderResourceBindings *srbD = QRHI_RES(QD3D12ShaderResourceBindings, srb); - for (int i = 0, ie = srbD->sortedBindings.size(); i != ie; ++i) { - const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->sortedBindings[i]); + for (int i = 0, ie = srbD->m_bindings.size(); i != ie; ++i) { + const QRhiShaderResourceBinding::Data *b = shaderResourceBindingData(srbD->m_bindings[i]); switch (b->type) { case QRhiShaderResourceBinding::UniformBuffer: { @@ -963,14 +1120,15 @@ void QRhiD3D12::setShaderResources(QRhiCommandBuffer *cb, QRhiShaderResourceBind QD3D12ShaderResourceVisitor visitor(srbD, stageData, gfxPsD ? 5 : 1); + QD3D12CommandBuffer::VisitorData &visitorData(cbD->visitorData); visitorData = {}; using namespace std::placeholders; - visitor.uniformBuffer = std::bind(&QRhiD3D12::visitUniformBuffer, this, _1, _2, _3, _4, dynamicOffsetCount, dynamicOffsets); - visitor.texture = std::bind(&QRhiD3D12::visitTexture, this, _1, _2, _3); - visitor.sampler = std::bind(&QRhiD3D12::visitSampler, this, _1, _2, _3); - visitor.storageBuffer = std::bind(&QRhiD3D12::visitStorageBuffer, this, _1, _2, _3, _4); - visitor.storageImage = std::bind(&QRhiD3D12::visitStorageImage, this, _1, _2, _3, _4); + visitor.uniformBuffer = std::bind(&QD3D12CommandBuffer::visitUniformBuffer, cbD, _1, _2, _3, _4, dynamicOffsetCount, dynamicOffsets); + visitor.texture = std::bind(&QD3D12CommandBuffer::visitTexture, cbD, _1, _2, _3); + visitor.sampler = std::bind(&QD3D12CommandBuffer::visitSampler, cbD, _1, _2, _3); + visitor.storageBuffer = std::bind(&QD3D12CommandBuffer::visitStorageBuffer, cbD, _1, _2, _3, _4); + visitor.storageImage = std::bind(&QD3D12CommandBuffer::visitStorageImage, cbD, _1, _2, _3, _4); visitor.visit(); @@ -1261,19 +1419,43 @@ void QRhiD3D12::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, void QRhiD3D12::debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) { - Q_UNUSED(cb); + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXBeginEvent(cbD->cmdList, PIX_COLOR_DEFAULT, reinterpret_cast<LPCWSTR>(QString::fromLatin1(name).utf16())); +#else + Q_UNUSED(cbD); Q_UNUSED(name); +#endif } void QRhiD3D12::debugMarkEnd(QRhiCommandBuffer *cb) { - Q_UNUSED(cb); + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXEndEvent(cbD->cmdList); +#else + Q_UNUSED(cbD); +#endif } void QRhiD3D12::debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) { - Q_UNUSED(cb); + if (!debugMarkers) + return; + + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); +#ifdef QRHI_D3D12_HAS_OLD_PIX + PIXSetMarker(cbD->cmdList, PIX_COLOR_DEFAULT, reinterpret_cast<LPCWSTR>(QString::fromLatin1(msg).utf16())); +#else + Q_UNUSED(cbD); Q_UNUSED(msg); +#endif } const QRhiNativeHandles *QRhiD3D12::nativeHandles(QRhiCommandBuffer *cb) @@ -1300,6 +1482,28 @@ void QRhiD3D12::endExternal(QRhiCommandBuffer *cb) } } +double QRhiD3D12::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QD3D12CommandBuffer *cbD = QRHI_RES(QD3D12CommandBuffer, cb); + return cbD->lastGpuTime; +} + +static void calculateGpuTime(QD3D12CommandBuffer *cbD, + int timestampPairStartIndex, + const quint8 *readbackBufPtr, + quint64 timestampTicksPerSecond) +{ + const size_t byteOffset = timestampPairStartIndex * sizeof(quint64); + const quint64 *p = reinterpret_cast<const quint64 *>(readbackBufPtr + byteOffset); + const quint64 startTime = *p++; + const quint64 endTime = *p; + if (startTime < endTime) { + const quint64 ticks = endTime - startTime; + const double timeSec = ticks / double(timestampTicksPerSecond); + cbD->lastGpuTime = timeSec; + } +} + QRhi::FrameOpResult QRhiD3D12::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); @@ -1321,7 +1525,7 @@ QRhi::FrameOpResult QRhiD3D12::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF // be in flight anymore). With Qt Quick this situation cannot happen anyway // by design (one QRhi per window). for (QD3D12SwapChain *sc : std::as_const(swapchains)) - sc->waitCommandCompletionForFrameSlot(sc->currentFrameSlot); + sc->waitCommandCompletionForFrameSlot(currentFrameSlot); // note: swapChainD->currentFrameSlot, not sc's HRESULT hr = cmdAllocators[currentFrameSlot]->Reset(); if (FAILED(hr)) { @@ -1343,6 +1547,16 @@ QRhi::FrameOpResult QRhiD3D12::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF swapChainD->rtWrapper.d.dsv = swapChainD->ds ? swapChainD->ds->dsv.cpuHandle : D3D12_CPU_DESCRIPTOR_HANDLE { 0 }; + if (swapChainD->stereo) { + swapChainD->rtWrapperRight.d.rtv[0] = swapChainD->sampleDesc.Count > 1 + ? swapChainD->msaaRtvs[swapChainD->currentBackBufferIndex].cpuHandle + : swapChainD->rtvsRight[swapChainD->currentBackBufferIndex].cpuHandle; + + swapChainD->rtWrapperRight.d.dsv = + swapChainD->ds ? swapChainD->ds->dsv.cpuHandle : D3D12_CPU_DESCRIPTOR_HANDLE{ 0 }; + } + + // Time to release things that are marked for currentFrameSlot since due to // the wait above we know that the previous commands on the GPU for this // slot must have finished already. @@ -1360,6 +1574,20 @@ QRhi::FrameOpResult QRhiD3D12::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF finishActiveReadbacks(); // last, in case the readback-completed callback issues rhi calls + if (timestampQueryHeap.isValid() && timestampTicksPerSecond) { + // Read the timestamps for the previous frame for this slot. (the + // ResolveQuery() should have completed by now due to the wait above) + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + calculateGpuTime(cbD, + timestampPairStartIndex, + timestampReadbackArea.mem.p, + timestampTicksPerSecond); + // Write the start timestamp for this frame for this slot. + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex); + } + return QRhi::FrameOpSuccess; } @@ -1384,7 +1612,20 @@ QRhi::FrameOpResult QRhiD3D12::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame barrierGen.addTransitionBarrier(backBufferResourceHandle, D3D12_RESOURCE_STATE_PRESENT); barrierGen.enqueueBufferedTransitionBarriers(cbD); - ID3D12GraphicsCommandList *cmdList = cbD->cmdList; + if (timestampQueryHeap.isValid()) { + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex + 1); + cbD->cmdList->ResolveQueryData(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex, + 2, + timestampReadbackArea.mem.buffer, + timestampPairStartIndex * sizeof(quint64)); + } + + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; HRESULT hr = cmdList->Close(); if (FAILED(hr)) { qWarning("Failed to close command list: %s", @@ -1402,6 +1643,10 @@ QRhi::FrameOpResult QRhiD3D12::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame { presentFlags |= DXGI_PRESENT_ALLOW_TEARING; } + if (!swapChainD->swapChain) { + qWarning("Failed to present, no swapchain"); + return QRhi::FrameOpError; + } HRESULT hr = swapChainD->swapChain->Present(swapChainD->swapInterval, presentFlags); if (hr == DXGI_ERROR_DEVICE_REMOVED || hr == DXGI_ERROR_DEVICE_RESET) { qWarning("Device loss detected in Present()"); @@ -1468,6 +1713,12 @@ QRhi::FrameOpResult QRhiD3D12::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: bindShaderVisibleHeaps(cbD); + if (timestampQueryHeap.isValid() && timestampTicksPerSecond) { + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT); + } + offscreenActive = true; *cb = cbD; @@ -1481,7 +1732,20 @@ QRhi::FrameOpResult QRhiD3D12::endOffscreenFrame(QRhi::EndFrameFlags flags) offscreenActive = false; QD3D12CommandBuffer *cbD = offscreenCb[currentFrameSlot]; - ID3D12GraphicsCommandList *cmdList = cbD->cmdList; + if (timestampQueryHeap.isValid()) { + const int timestampPairStartIndex = currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT; + cbD->cmdList->EndQuery(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex + 1); + cbD->cmdList->ResolveQueryData(timestampQueryHeap.heap, + D3D12_QUERY_TYPE_TIMESTAMP, + timestampPairStartIndex, + 2, + timestampReadbackArea.mem.buffer, + timestampPairStartIndex * sizeof(quint64)); + } + + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; HRESULT hr = cmdList->Close(); if (FAILED(hr)) { qWarning("Failed to close command list: %s", @@ -1501,6 +1765,14 @@ QRhi::FrameOpResult QRhiD3D12::endOffscreenFrame(QRhi::EndFrameFlags flags) // previous) frame is safe since we waited for completion above. finishActiveReadbacks(true); + // the timestamp query results should be available too, given the wait + if (timestampQueryHeap.isValid()) { + calculateGpuTime(cbD, + currentFrameSlot * QD3D12_FRAMES_IN_FLIGHT, + timestampReadbackArea.mem.p, + timestampTicksPerSecond); + } + return QRhi::FrameOpSuccess; } @@ -1522,7 +1794,7 @@ QRhi::FrameOpResult QRhiD3D12::finish() Q_ASSERT(cbD->recordingPass == QD3D12CommandBuffer::NoPass); - ID3D12GraphicsCommandList *cmdList = cbD->cmdList; + ID3D12GraphicsCommandList1 *cmdList = cbD->cmdList; HRESULT hr = cmdList->Close(); if (FAILED(hr)) { qWarning("Failed to close command list: %s", @@ -1707,15 +1979,19 @@ void QRhiD3D12::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resource barrierGen.addTransitionBarrier(dstTexD->handle, D3D12_RESOURCE_STATE_RESOLVE_DEST); barrierGen.enqueueBufferedTransitionBarriers(cbD); - const UINT srcSubresource = calcSubresource(0, UINT(colorAtt.layer()), 1); - const UINT dstSubresource = calcSubresource(UINT(colorAtt.resolveLevel()), - UINT(colorAtt.resolveLayer()), - dstTexD->mipLevelCount); - cbD->cmdList->ResolveSubresource(dstRes->resource, dstSubresource, - srcRes->resource, srcSubresource, - dstTexD->dxgiFormat); + const UINT resolveCount = colorAtt.multiViewCount() >= 2 ? colorAtt.multiViewCount() : 1; + for (UINT resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + const UINT srcSubresource = calcSubresource(0, UINT(colorAtt.layer()) + resolveIdx, 1); + const UINT dstSubresource = calcSubresource(UINT(colorAtt.resolveLevel()), + UINT(colorAtt.resolveLayer()) + resolveIdx, + dstTexD->mipLevelCount); + cbD->cmdList->ResolveSubresource(dstRes->resource, dstSubresource, + srcRes->resource, srcSubresource, + dstTexD->dxgiFormat); + } } - + if (rtTex->m_desc.depthResolveTexture()) + qWarning("Resolving multisample depth-stencil buffers is not supported with D3D"); } cbD->recordingPass = QD3D12CommandBuffer::NoPass; @@ -1962,6 +2238,36 @@ void QD3D12CpuDescriptorPool::release(const QD3D12Descriptor &descriptor, quint3 quint64(descriptor.cpuHandle.ptr)); } +bool QD3D12QueryHeap::create(ID3D12Device *device, + quint32 queryCount, + D3D12_QUERY_HEAP_TYPE heapType) +{ + capacity = queryCount; + + D3D12_QUERY_HEAP_DESC heapDesc = {}; + heapDesc.Type = heapType; + heapDesc.Count = capacity; + + HRESULT hr = device->CreateQueryHeap(&heapDesc, __uuidof(ID3D12QueryHeap), reinterpret_cast<void **>(&heap)); + if (FAILED(hr)) { + qWarning("Failed to create query heap: %s", qPrintable(QSystemError::windowsComString(hr))); + heap = nullptr; + capacity = 0; + return false; + } + + return true; +} + +void QD3D12QueryHeap::destroy() +{ + if (heap) { + heap->Release(); + heap = nullptr; + } + capacity = 0; +} + bool QD3D12StagingArea::create(QRhiD3D12 *rhi, quint32 capacity, D3D12_HEAP_TYPE heapType) { Q_ASSERT(heapType == D3D12_HEAP_TYPE_UPLOAD || heapType == D3D12_HEAP_TYPE_READBACK); @@ -2302,8 +2608,8 @@ static inline QPair<int, int> mapBinding(int binding, const QShader::NativeResou void QD3D12ShaderResourceVisitor::visit() { - for (int bindingIdx = 0, bindingCount = srb->sortedBindings.count(); bindingIdx != bindingCount; ++bindingIdx) { - const QRhiShaderResourceBinding &b(srb->sortedBindings[bindingIdx]); + for (int bindingIdx = 0, bindingCount = srb->m_bindings.count(); bindingIdx != bindingCount; ++bindingIdx) { + const QRhiShaderResourceBinding &b(srb->m_bindings[bindingIdx]); const QRhiShaderResourceBinding::Data *bd = QRhiImplementation::shaderResourceBindingData(b); for (int stageIdx = 0; stageIdx < stageCount; ++stageIdx) { @@ -2454,6 +2760,7 @@ bool QD3D12MipmapGenerator::create(QRhiD3D12 *rhiD) // b0 rootParams[0].ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV; rootParams[0].ShaderVisibility = D3D12_SHADER_VISIBILITY_ALL; + rootParams[0].Descriptor.Flags = D3D12_ROOT_DESCRIPTOR_FLAG_DATA_STATIC; // t0 descriptorRanges[0].RangeType = D3D12_DESCRIPTOR_RANGE_TYPE_SRV; @@ -2800,24 +3107,18 @@ void QRhiD3D12::waitGpu() } } -DXGI_SAMPLE_DESC QRhiD3D12::effectiveSampleCount(int sampleCount, DXGI_FORMAT format) const +DXGI_SAMPLE_DESC QRhiD3D12::effectiveSampleDesc(int sampleCount, DXGI_FORMAT format) const { DXGI_SAMPLE_DESC desc; desc.Count = 1; desc.Quality = 0; - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - int s = qBound(1, sampleCount, 64); - - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return desc; - } + const int s = effectiveSampleCount(sampleCount); if (s > 1) { D3D12_FEATURE_DATA_MULTISAMPLE_QUALITY_LEVELS msaaInfo = {}; msaaInfo.Format = format; - msaaInfo.SampleCount = s; + msaaInfo.SampleCount = UINT(s); if (SUCCEEDED(dev->CheckFeatureSupport(D3D12_FEATURE_MULTISAMPLE_QUALITY_LEVELS, &msaaInfo, sizeof(msaaInfo)))) { if (msaaInfo.NumQualityLevels > 0) { desc.Count = UINT(s); @@ -2831,7 +3132,7 @@ DXGI_SAMPLE_DESC QRhiD3D12::effectiveSampleCount(int sampleCount, DXGI_FORMAT fo return desc; } -bool QRhiD3D12::startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList **cmdList) +bool QRhiD3D12::startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList1 **cmdList) { ID3D12CommandAllocator *cmdAlloc = cmdAllocators[currentFrameSlot]; if (!*cmdList) { @@ -2839,7 +3140,7 @@ bool QRhiD3D12::startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList ** D3D12_COMMAND_LIST_TYPE_DIRECT, cmdAlloc, nullptr, - __uuidof(ID3D12GraphicsCommandList), + __uuidof(ID3D12GraphicsCommandList1), reinterpret_cast<void **>(cmdList)); if (FAILED(hr)) { qWarning("Failed to create command list: %s", qPrintable(QSystemError::windowsComString(hr))); @@ -2997,18 +3298,42 @@ void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpd for (int layer = 0, maxLayer = u.subresDesc.size(); layer < maxLayer; ++layer) { for (int level = 0; level < QRhi::MAX_MIP_LEVELS; ++level) { for (const QRhiTextureSubresourceUploadDescription &subresDesc : std::as_const(u.subresDesc[layer][level])) { - const UINT subresource = calcSubresource(UINT(level), is3D ? 0u : UINT(layer), texD->mipLevelCount); - D3D12_PLACED_SUBRESOURCE_FOOTPRINT layout; - UINT64 totalBytes = 0; - D3D12_RESOURCE_DESC desc = res->desc; - if (is3D) { - desc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; - desc.DepthOrArraySize = 1; + D3D12_SUBRESOURCE_FOOTPRINT footprint = {}; + footprint.Format = res->desc.Format; + footprint.Depth = 1; + quint32 totalBytes = 0; + + const QSize subresSize = subresDesc.sourceSize().isEmpty() ? q->sizeForMipLevel(level, texD->m_pixelSize) + : subresDesc.sourceSize(); + const QPoint srcPos = subresDesc.sourceTopLeft(); + QPoint dstPos = subresDesc.destinationTopLeft(); + + if (!subresDesc.image().isNull()) { + const QImage img = subresDesc.image(); + const int bpl = img.bytesPerLine(); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + totalBytes = footprint.RowPitch * img.height(); + } else if (!subresDesc.data().isEmpty() && isCompressedFormat(texD->m_format)) { + QSize blockDim; + quint32 bpl = 0; + compressedFormatInfo(texD->m_format, subresSize, &bpl, nullptr, &blockDim); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + const int rowCount = aligned(subresSize.height(), blockDim.height()) / blockDim.height(); + totalBytes = footprint.RowPitch * rowCount; + } else if (!subresDesc.data().isEmpty()) { + quint32 bpl = 0; + if (subresDesc.dataStride()) + bpl = subresDesc.dataStride(); + else + textureFormatInfo(texD->m_format, subresSize, &bpl, nullptr, nullptr); + footprint.RowPitch = aligned<UINT>(bpl, D3D12_TEXTURE_DATA_PITCH_ALIGNMENT); + totalBytes = footprint.RowPitch * subresSize.height(); + } else { + qWarning("Invalid texture upload for %p layer=%d mip=%d", texD, layer, level); + continue; } - dev->GetCopyableFootprints(&desc, subresource, 1, 0, - &layout, nullptr, nullptr, &totalBytes); - const quint32 allocSize = QD3D12StagingArea::allocSizeForArray(quint32(totalBytes), 1); + const quint32 allocSize = QD3D12StagingArea::allocSizeForArray(totalBytes, 1); QD3D12StagingArea::Allocation stagingAlloc; if (smallStagingAreas[currentFrameSlot].remainingCapacity() >= allocSize) stagingAlloc = smallStagingAreas[currentFrameSlot].get(allocSize); @@ -3025,32 +3350,29 @@ void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpd } } - const UINT requiredBytesPerLine = layout.Footprint.RowPitch; // multiple of 256 - const QSize subresSize = subresDesc.sourceSize().isEmpty() ? q->sizeForMipLevel(level, texD->m_pixelSize) - : subresDesc.sourceSize(); - const QPoint srcPos = subresDesc.sourceTopLeft(); - QPoint dstPos = subresDesc.destinationTopLeft(); - D3D12_TEXTURE_COPY_LOCATION dst; dst.pResource = res->resource; dst.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; - dst.SubresourceIndex = subresource; + dst.SubresourceIndex = calcSubresource(UINT(level), is3D ? 0u : UINT(layer), texD->mipLevelCount); D3D12_TEXTURE_COPY_LOCATION src; src.pResource = stagingAlloc.buffer; src.Type = D3D12_TEXTURE_COPY_TYPE_PLACED_FOOTPRINT; src.PlacedFootprint.Offset = stagingAlloc.bufferOffset; - src.PlacedFootprint.Footprint = layout.Footprint; D3D12_BOX srcBox; // back, right, bottom are exclusive if (!subresDesc.image().isNull()) { - QImage img = subresDesc.image(); + const QImage img = subresDesc.image(); const int bpc = qMax(1, img.depth() / 8); const int bpl = img.bytesPerLine(); QSize size = subresDesc.sourceSize().isEmpty() ? img.size() : subresDesc.sourceSize(); size.setWidth(qMin(size.width(), img.width() - srcPos.x())); size.setHeight(qMin(size.height(), img.height() - srcPos.y())); + + footprint.Width = size.width(); + footprint.Height = size.height(); + srcBox.left = 0; srcBox.top = 0; srcBox.right = UINT(size.width()); @@ -3061,7 +3383,7 @@ void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpd const uchar *imgPtr = img.constBits(); const quint32 lineBytes = size.width() * bpc; for (int y = 0, h = size.height(); y < h; ++y) { - memcpy(stagingAlloc.p + y * requiredBytesPerLine, + memcpy(stagingAlloc.p + y * footprint.RowPitch, imgPtr + srcPos.x() * bpc + (y + srcPos.y()) * bpl, lineBytes); } @@ -3078,15 +3400,19 @@ void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpd // width and height must be multiples of the block width and height srcBox.right = aligned(subresSize.width(), blockDim.width()); srcBox.bottom = aligned(subresSize.height(), blockDim.height()); + srcBox.front = 0; srcBox.back = 1; - const quint32 copyBytes = qMin(bpl, requiredBytesPerLine); + footprint.Width = aligned(subresSize.width(), blockDim.width()); + footprint.Height = aligned(subresSize.height(), blockDim.height()); + + const quint32 copyBytes = qMin(bpl, footprint.RowPitch); const QByteArray imgData = subresDesc.data(); const char *imgPtr = imgData.constData(); const int rowCount = aligned(subresSize.height(), blockDim.height()) / blockDim.height(); for (int y = 0; y < rowCount; ++y) - memcpy(stagingAlloc.p + y * requiredBytesPerLine, imgPtr + y * bpl, copyBytes); + memcpy(stagingAlloc.p + y * footprint.RowPitch, imgPtr + y * bpl, copyBytes); } else if (!subresDesc.data().isEmpty()) { srcBox.left = 0; srcBox.top = 0; @@ -3095,24 +3421,24 @@ void QRhiD3D12::enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpd srcBox.front = 0; srcBox.back = 1; + footprint.Width = subresSize.width(); + footprint.Height = subresSize.height(); + quint32 bpl = 0; if (subresDesc.dataStride()) bpl = subresDesc.dataStride(); else textureFormatInfo(texD->m_format, subresSize, &bpl, nullptr, nullptr); - const quint32 copyBytes = qMin(bpl, requiredBytesPerLine); + const quint32 copyBytes = qMin(bpl, footprint.RowPitch); const QByteArray data = subresDesc.data(); const char *imgPtr = data.constData(); for (int y = 0, h = subresSize.height(); y < h; ++y) - memcpy(stagingAlloc.p + y * requiredBytesPerLine, imgPtr + y * bpl, copyBytes); - } else { - qWarning("Invalid texture upload for %p layer=%d mip=%d", texD, layer, level); - if (ownStagingArea.has_value()) - ownStagingArea->destroyWithDeferredRelease(&releaseQueue); - continue; + memcpy(stagingAlloc.p + y * footprint.RowPitch, imgPtr + y * bpl, copyBytes); } + src.PlacedFootprint.Footprint = footprint; + cbD->cmdList->CopyTextureRegion(&dst, UINT(dstPos.x()), UINT(dstPos.y()), @@ -3663,7 +3989,7 @@ bool QD3D12RenderBuffer::create() case QRhiRenderBuffer::Color: { dxgiFormat = toD3DTextureFormat(backingFormat(), {}); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount, dxgiFormat); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); D3D12_RESOURCE_DESC resourceDesc = {}; resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; resourceDesc.Width = UINT64(m_pixelSize.width()); @@ -3704,7 +4030,7 @@ bool QD3D12RenderBuffer::create() case QRhiRenderBuffer::DepthStencil: { dxgiFormat = DS_FORMAT; - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount, dxgiFormat); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); D3D12_RESOURCE_DESC resourceDesc = {}; resourceDesc.Dimension = D3D12_RESOURCE_DIMENSION_TEXTURE2D; resourceDesc.Width = UINT64(m_pixelSize.width()); @@ -3857,10 +4183,30 @@ bool QD3D12Texture::prepareCreate(QSize *adjustedSize) const QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); - QRHI_RES_RHI(QRhiD3D12); dxgiFormat = toD3DTextureFormat(m_format, m_flags); + if (isDepth) { + srvFormat = toD3DDepthTextureSRVFormat(m_format); + rtFormat = toD3DDepthTextureDSVFormat(m_format); + } else { + srvFormat = dxgiFormat; + rtFormat = dxgiFormat; + } + if (m_writeViewFormat.format != UnknownFormat) { + if (isDepth) + rtFormat = toD3DDepthTextureDSVFormat(m_writeViewFormat.format); + else + rtFormat = toD3DTextureFormat(m_writeViewFormat.format, m_writeViewFormat.srgb ? sRGB : Flags()); + } + if (m_readViewFormat.format != UnknownFormat) { + if (isDepth) + srvFormat = toD3DDepthTextureSRVFormat(m_readViewFormat.format); + else + srvFormat = toD3DTextureFormat(m_readViewFormat.format, m_readViewFormat.srgb ? sRGB : Flags()); + } + + QRHI_RES_RHI(QRhiD3D12); mipLevelCount = uint(hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1); - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount, dxgiFormat); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, dxgiFormat); if (sampleDesc.Count > 1) { if (isCube) { qWarning("Cubemap texture cannot be multisample"); @@ -3895,12 +4241,10 @@ bool QD3D12Texture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both 1D and 3D"); return false; } - m_depth = qMax(1, m_depth); if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } - m_arraySize = qMax(0, m_arraySize); if (m_arraySize > 0 && !isArray) { qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); return false; @@ -3919,14 +4263,13 @@ bool QD3D12Texture::prepareCreate(QSize *adjustedSize) bool QD3D12Texture::finishCreate() { QRHI_RES_RHI(QRhiD3D12); - const bool isDepth = isDepthTextureFormat(m_format); const bool isCube = m_flags.testFlag(CubeMap); const bool is3D = m_flags.testFlag(ThreeDimensional); const bool isArray = m_flags.testFlag(TextureArray); const bool is1D = m_flags.testFlag(OneDimensional); D3D12_SHADER_RESOURCE_VIEW_DESC srvDesc = {}; - srvDesc.Format = isDepth ? toD3DDepthTextureSRVFormat(m_format) : dxgiFormat; + srvDesc.Format = srvFormat; srvDesc.Shader4ComponentMapping = D3D12_DEFAULT_SHADER_4_COMPONENT_MAPPING; if (isCube) { @@ -3942,7 +4285,7 @@ bool QD3D12Texture::finishCreate() srvDesc.Texture1DArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture1DArray.FirstArraySlice = 0; - srvDesc.Texture1DArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture1DArray.ArraySize = UINT(qMax(0, m_arraySize)); } } else { srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1D; @@ -3956,7 +4299,7 @@ bool QD3D12Texture::finishCreate() srvDesc.Texture2DMSArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture2DMSArray.FirstArraySlice = 0; - srvDesc.Texture2DMSArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, m_arraySize)); } } else { srvDesc.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY; @@ -3966,7 +4309,7 @@ bool QD3D12Texture::finishCreate() srvDesc.Texture2DArray.ArraySize = UINT(m_arrayRangeLength); } else { srvDesc.Texture2DArray.FirstArraySlice = 0; - srvDesc.Texture2DArray.ArraySize = UINT(m_arraySize); + srvDesc.Texture2DArray.ArraySize = UINT(qMax(0, m_arraySize)); } } } else { @@ -4016,7 +4359,7 @@ bool QD3D12Texture::create() bool needsOptimizedClearValueSpecified = false; UINT resourceFlags = 0; - if (m_flags.testFlag(RenderTarget)) { + if (m_flags.testFlag(RenderTarget) || sampleDesc.Count > 1) { if (isDepth) resourceFlags |= D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL; else @@ -4039,7 +4382,10 @@ bool QD3D12Texture::create() : D3D12_RESOURCE_DIMENSION_TEXTURE2D); resourceDesc.Width = UINT64(size.width()); resourceDesc.Height = UINT(size.height()); - resourceDesc.DepthOrArraySize = isCube ? 6 : (isArray ? UINT(m_arraySize) : (is3D ? m_depth : 1)); + resourceDesc.DepthOrArraySize = isCube ? 6 + : (isArray ? UINT(qMax(0, m_arraySize)) + : (is3D ? qMax(1, m_depth) + : 1)); resourceDesc.MipLevels = mipLevelCount; resourceDesc.Format = dxgiFormat; resourceDesc.SampleDesc = sampleDesc; @@ -4135,6 +4481,10 @@ QD3D12Sampler::~QD3D12Sampler() void QD3D12Sampler::destroy() { shaderVisibleDescriptor = {}; + + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); } static inline D3D12_FILTER toD3DFilter(QRhiSampler::Filter minFilter, QRhiSampler::Filter magFilter, QRhiSampler::Filter mipFilter) @@ -4215,6 +4565,9 @@ bool QD3D12Sampler::create() desc.MaxAnisotropy = 1.0f; desc.ComparisonFunc = toD3DTextureComparisonFunc(m_compareOp); desc.MaxLOD = m_mipmapMode == None ? 0.0f : 10000.0f; + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(this, false); return true; } @@ -4275,7 +4628,7 @@ QRhiRenderPassDescriptor *QD3D12TextureRenderTarget::newCompatibleRenderPassDesc QD3D12Texture *texD = QRHI_RES(QD3D12Texture, it->texture()); QD3D12RenderBuffer *rbD = QRHI_RES(QD3D12RenderBuffer, it->renderBuffer()); if (texD) - rpD->colorFormat[rpD->colorAttachmentCount] = texD->dxgiFormat; + rpD->colorFormat[rpD->colorAttachmentCount] = texD->rtFormat; else if (rbD) rpD->colorFormat[rpD->colorAttachmentCount] = rbD->dxgiFormat; rpD->colorAttachmentCount += 1; @@ -4293,6 +4646,8 @@ QRhiRenderPassDescriptor *QD3D12TextureRenderTarget::newCompatibleRenderPassDesc rpD->updateSerializedFormat(); + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); return rpD; } @@ -4321,19 +4676,21 @@ bool QD3D12TextureRenderTarget::create() qWarning("Could not look up texture handle for render target"); return false; } + const bool isMultiView = it->multiViewCount() >= 2; + UINT layerCount = isMultiView ? UINT(it->multiViewCount()) : 1; D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; - rtvDesc.Format = toD3DTextureFormat(texD->format(), texD->flags()); + rtvDesc.Format = texD->rtFormat; if (texD->flags().testFlag(QRhiTexture::CubeMap)) { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); - rtvDesc.Texture2DArray.ArraySize = 1; + rtvDesc.Texture2DArray.ArraySize = layerCount; } else if (texD->flags().testFlag(QRhiTexture::OneDimensional)) { if (texD->flags().testFlag(QRhiTexture::TextureArray)) { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE1DARRAY; rtvDesc.Texture1DArray.MipSlice = UINT(colorAtt.level()); rtvDesc.Texture1DArray.FirstArraySlice = UINT(colorAtt.layer()); - rtvDesc.Texture1DArray.ArraySize = 1; + rtvDesc.Texture1DArray.ArraySize = layerCount; } else { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE1D; rtvDesc.Texture1D.MipSlice = UINT(colorAtt.level()); @@ -4342,18 +4699,18 @@ bool QD3D12TextureRenderTarget::create() if (texD->sampleDesc.Count > 1) { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DMSARRAY; rtvDesc.Texture2DMSArray.FirstArraySlice = UINT(colorAtt.layer()); - rtvDesc.Texture2DMSArray.ArraySize = 1; + rtvDesc.Texture2DMSArray.ArraySize = layerCount; } else { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; rtvDesc.Texture2DArray.MipSlice = UINT(colorAtt.level()); rtvDesc.Texture2DArray.FirstArraySlice = UINT(colorAtt.layer()); - rtvDesc.Texture2DArray.ArraySize = 1; + rtvDesc.Texture2DArray.ArraySize = layerCount; } } else if (texD->flags().testFlag(QRhiTexture::ThreeDimensional)) { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE3D; rtvDesc.Texture3D.MipSlice = UINT(colorAtt.level()); rtvDesc.Texture3D.FirstWSlice = UINT(colorAtt.layer()); - rtvDesc.Texture3D.WSize = 1; + rtvDesc.Texture3D.WSize = layerCount; } else { if (texD->sampleDesc.Count > 1) { rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DMS; @@ -4396,9 +4753,30 @@ bool QD3D12TextureRenderTarget::create() return false; } D3D12_DEPTH_STENCIL_VIEW_DESC dsvDesc = {}; - dsvDesc.Format = toD3DDepthTextureDSVFormat(depthTexD->format()); + dsvDesc.Format = depthTexD->rtFormat; dsvDesc.ViewDimension = depthTexD->sampleDesc.Count > 1 ? D3D12_DSV_DIMENSION_TEXTURE2DMS : D3D12_DSV_DIMENSION_TEXTURE2D; + if (depthTexD->flags().testFlag(QRhiTexture::TextureArray)) { + if (depthTexD->sampleDesc.Count > 1) { + dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DMSARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DMSArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DMSArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DMSArray.FirstArraySlice = 0; + dsvDesc.Texture2DMSArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } else { + dsvDesc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DARRAY; + if (depthTexD->arrayRangeStart() >= 0 && depthTexD->arrayRangeLength() >= 0) { + dsvDesc.Texture2DArray.FirstArraySlice = UINT(depthTexD->arrayRangeStart()); + dsvDesc.Texture2DArray.ArraySize = UINT(depthTexD->arrayRangeLength()); + } else { + dsvDesc.Texture2DArray.FirstArraySlice = 0; + dsvDesc.Texture2DArray.ArraySize = UINT(qMax(0, depthTexD->arraySize())); + } + } + } dsv = rhiD->dsvPool.allocate(1); if (!dsv.isValid()) { qWarning("Failed to allocate DSV for texture render target"); @@ -4465,25 +4843,21 @@ QD3D12ShaderResourceBindings::~QD3D12ShaderResourceBindings() void QD3D12ShaderResourceBindings::destroy() { - sortedBindings.clear(); + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D12ShaderResourceBindings::create() { - if (!sortedBindings.isEmpty()) - destroy(); - QRHI_RES_RHI(QRhiD3D12); if (!rhiD->sanityCheckShaderResourceBindings(this)) return false; rhiD->updateLayoutDesc(this); - std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); - std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); - hasDynamicOffset = false; - for (const QRhiShaderResourceBinding &b : sortedBindings) { + for (const QRhiShaderResourceBinding &b : std::as_const(m_bindings)) { const QRhiShaderResourceBinding::Data *bd = QRhiImplementation::shaderResourceBindingData(b); if (bd->type == QRhiShaderResourceBinding::UniformBuffer && bd->u.ubuf.hasDynamicOffset) { hasDynamicOffset = true; @@ -4500,16 +4874,13 @@ bool QD3D12ShaderResourceBindings::create() // therefore impossible. generation += 1; + rhiD->registerResource(this, false); return true; } void QD3D12ShaderResourceBindings::updateResources(UpdateFlags flags) { - sortedBindings.clear(); - std::copy(m_bindings.cbegin(), m_bindings.cend(), std::back_inserter(sortedBindings)); - if (!flags.testFlag(BindingsAreSorted)) - std::sort(sortedBindings.begin(), sortedBindings.end(), QRhiImplementation::sortedBindingLessThan); - + Q_UNUSED(flags); generation += 1; } @@ -4527,6 +4898,7 @@ void QD3D12ShaderResourceBindings::visitUniformBuffer(QD3D12Stage s, rootParam.ParameterType = D3D12_ROOT_PARAMETER_TYPE_CBV; rootParam.ShaderVisibility = qd3d12_stageToVisibility(s); rootParam.Descriptor.ShaderRegister = shaderRegister; + rootParam.Descriptor.Flags = D3D12_ROOT_DESCRIPTOR_FLAG_DATA_STATIC; visitorData.cbParams[s].append(rootParam); } @@ -4718,21 +5090,14 @@ QD3D12ObjectHandle QD3D12ShaderResourceBindings::createRootSignature(const QD3D1 return QD3D12RootSignature::addToPool(&rhiD->rootSignaturePool, rootSig); } -// For now we mirror exactly what's done in the D3D11 backend, meaning we use -// the old shader compiler (so like fxc, not dxc) to generate shader model 5.0 -// output. Some day this should be moved to the new compiler and DXIL. - -static pD3DCompile resolveD3DCompile() -{ - for (const wchar_t *libraryName : {L"D3DCompiler_47", L"D3DCompiler_43"}) { - QSystemLibrary library(libraryName); - if (library.load()) { - if (auto symbol = library.resolve("D3DCompile")) - return reinterpret_cast<pD3DCompile>(symbol); - } - } - return nullptr; -} +// For shader model < 6.0 we do the same as the D3D11 backend: use the old +// compiler (D3DCompile) to generate DXBC, just as qsb does (when -c is passed) +// by invoking fxc, not dxc. For SM >= 6.0 we have to use the new compiler and +// work with DXIL. And that involves IDxcCompiler and needs the presence of +// dxcompiler.dll and dxil.dll at runtime. Plus there's a chance we have +// ancient SDK headers when not using MSVC. So this is heavily optional, +// meaning support for dxc can be disabled both at build time (no dxcapi.h) and +// at run time (no DLLs). static inline void makeHlslTargetString(char target[7], const char stage[3], int version) { @@ -4747,9 +5112,139 @@ static inline void makeHlslTargetString(char target[7], const char stage[3], int target[6] = '\0'; } +enum class HlslCompileFlag +{ + WithDebugInfo = 0x01 +}; + +static QByteArray legacyCompile(const QShaderCode &hlslSource, const char *target, int flags, QString *error) +{ + static const pD3DCompile d3dCompile = QRhiD3D::resolveD3DCompile(); + if (!d3dCompile) { + qWarning("Unable to resolve function D3DCompile()"); + return QByteArray(); + } + + ID3DBlob *bytecode = nullptr; + ID3DBlob *errors = nullptr; + UINT d3dCompileFlags = 0; + if (flags & int(HlslCompileFlag::WithDebugInfo)) + d3dCompileFlags |= D3DCOMPILE_DEBUG; + + HRESULT hr = d3dCompile(hlslSource.shader().constData(), SIZE_T(hlslSource.shader().size()), + nullptr, nullptr, nullptr, + hlslSource.entryPoint().constData(), target, d3dCompileFlags, 0, &bytecode, &errors); + if (FAILED(hr) || !bytecode) { + qWarning("HLSL shader compilation failed: 0x%x", uint(hr)); + if (errors) { + *error = QString::fromUtf8(static_cast<const char *>(errors->GetBufferPointer()), + int(errors->GetBufferSize())); + errors->Release(); + } + return QByteArray(); + } + + QByteArray result; + result.resize(int(bytecode->GetBufferSize())); + memcpy(result.data(), bytecode->GetBufferPointer(), size_t(result.size())); + bytecode->Release(); + return result; +} + +#ifdef QRHI_D3D12_HAS_DXC + +#ifndef DXC_CP_UTF8 +#define DXC_CP_UTF8 65001 +#endif + +#ifndef DXC_ARG_DEBUG +#define DXC_ARG_DEBUG L"-Zi" +#endif + +static QByteArray dxcCompile(const QShaderCode &hlslSource, const char *target, int flags, QString *error) +{ + static std::pair<IDxcCompiler *, IDxcLibrary *> dxc = QRhiD3D::createDxcCompiler(); + IDxcCompiler *compiler = dxc.first; + if (!compiler) { + qWarning("Unable to instantiate IDxcCompiler. Likely no dxcompiler.dll and dxil.dll present. " + "Use windeployqt or try https://github.com/microsoft/DirectXShaderCompiler/releases"); + return QByteArray(); + } + IDxcLibrary *library = dxc.second; + if (!library) + return QByteArray(); + + IDxcBlobEncoding *sourceBlob = nullptr; + HRESULT hr = library->CreateBlobWithEncodingOnHeapCopy(hlslSource.shader().constData(), + UINT32(hlslSource.shader().size()), + DXC_CP_UTF8, + &sourceBlob); + if (FAILED(hr)) { + qWarning("Failed to create source blob for dxc: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + return QByteArray(); + } + + const QString entryPointStr = QString::fromLatin1(hlslSource.entryPoint()); + const QString targetStr = QString::fromLatin1(target); + + QVarLengthArray<LPCWSTR, 4> argPtrs; + QString debugArg; + if (flags & int(HlslCompileFlag::WithDebugInfo)) { + debugArg = QString::fromUtf16(reinterpret_cast<const char16_t *>(DXC_ARG_DEBUG)); + argPtrs.append(reinterpret_cast<LPCWSTR>(debugArg.utf16())); + } + + IDxcOperationResult *result = nullptr; + hr = compiler->Compile(sourceBlob, + nullptr, + reinterpret_cast<LPCWSTR>(entryPointStr.utf16()), + reinterpret_cast<LPCWSTR>(targetStr.utf16()), + argPtrs.data(), argPtrs.count(), + nullptr, 0, + nullptr, + &result); + sourceBlob->Release(); + if (SUCCEEDED(hr)) + result->GetStatus(&hr); + if (FAILED(hr)) { + qWarning("HLSL shader compilation failed: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + if (result) { + IDxcBlobEncoding *errorsBlob = nullptr; + if (SUCCEEDED(result->GetErrorBuffer(&errorsBlob))) { + if (errorsBlob) { + *error = QString::fromUtf8(static_cast<const char *>(errorsBlob->GetBufferPointer()), + int(errorsBlob->GetBufferSize())); + errorsBlob->Release(); + } + } + } + return QByteArray(); + } + + IDxcBlob *bytecode = nullptr; + if FAILED(result->GetResult(&bytecode)) { + qWarning("No result from IDxcCompiler: 0x%x (%s)", + uint(hr), + qPrintable(QSystemError::windowsComString(hr))); + return QByteArray(); + } + + QByteArray ba; + ba.resize(int(bytecode->GetBufferSize())); + memcpy(ba.data(), bytecode->GetBufferPointer(), size_t(ba.size())); + bytecode->Release(); + return ba; +} + +#endif // QRHI_D3D12_HAS_DXC + static QByteArray compileHlslShaderSource(const QShader &shader, QShader::Variant shaderVariant, - UINT flags, + int flags, QString *error, QShaderKey *usedShaderKey) { @@ -4806,33 +5301,17 @@ static QByteArray compileHlslShaderSource(const QShader &shader, break; } - static const pD3DCompile d3dCompile = resolveD3DCompile(); - if (!d3dCompile) { - qWarning("Unable to resolve function D3DCompile()"); - return QByteArray(); - } - - ID3DBlob *bytecode = nullptr; - ID3DBlob *errors = nullptr; - HRESULT hr = d3dCompile(hlslSource.shader().constData(), SIZE_T(hlslSource.shader().size()), - nullptr, nullptr, nullptr, - hlslSource.entryPoint().constData(), target, flags, 0, &bytecode, &errors); - if (FAILED(hr) || !bytecode) { - qWarning("HLSL shader compilation failed: 0x%x", uint(hr)); - if (errors) { - *error = QString::fromUtf8(static_cast<const char *>(errors->GetBufferPointer()), - int(errors->GetBufferSize())); - errors->Release(); - } - return QByteArray(); + if (key.sourceVersion().version() >= 60) { +#ifdef QRHI_D3D12_HAS_DXC + return dxcCompile(hlslSource, target, flags, error); +#else + qWarning("Attempted to runtime-compile HLSL source code for shader model >= 6.0 " + "but the Qt build has no support for DXC. " + "Rebuild Qt with a recent Windows SDK or switch to an MSVC build."); +#endif } - QByteArray result; - result.resize(int(bytecode->GetBufferSize())); - memcpy(result.data(), bytecode->GetBufferPointer(), size_t(result.size())); - bytecode->Release(); - - return result; + return legacyCompile(hlslSource, target, flags, error); } static inline UINT8 toD3DColorWriteMask(QRhiGraphicsPipeline::ColorMask c) @@ -5067,6 +5546,22 @@ static inline DXGI_FORMAT toD3DAttributeFormat(QRhiVertexInputAttribute::Format return DXGI_FORMAT_R16G16_FLOAT; case QRhiVertexInputAttribute::Half: return DXGI_FORMAT_R16_FLOAT; + case QRhiVertexInputAttribute::UShort4: + // Note: D3D does not support UShort3. Pass through UShort3 as UShort4. + case QRhiVertexInputAttribute::UShort3: + return DXGI_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort2: + return DXGI_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return DXGI_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + // Note: D3D does not support SShort3. Pass through SShort3 as SShort4. + case QRhiVertexInputAttribute::SShort3: + return DXGI_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort2: + return DXGI_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return DXGI_FORMAT_R16_SINT; } Q_UNREACHABLE_RETURN(DXGI_FORMAT_R32G32B32A32_FLOAT); } @@ -5122,16 +5617,16 @@ bool QD3D12GraphicsPipeline::create() } else { QString error; QShaderKey shaderKey; - UINT compileFlags = 0; + int compileFlags = 0; if (m_flags.testFlag(CompileShadersWithDebugInfo)) - compileFlags |= D3DCOMPILE_DEBUG; + compileFlags |= int(HlslCompileFlag::WithDebugInfo); const QByteArray bytecode = compileHlslShaderSource(shaderStage.shader(), shaderStage.shaderVariant(), compileFlags, &error, &shaderKey); if (bytecode.isEmpty()) { - qWarning("HLSL compute shader compilation failed: %s", qPrintable(error)); + qWarning("HLSL graphics shader compilation failed: %s", qPrintable(error)); return false; } @@ -5159,32 +5654,94 @@ bool QD3D12GraphicsPipeline::create() } QD3D12RenderPassDescriptor *rpD = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); - const DXGI_SAMPLE_DESC sampleDesc = rhiD->effectiveSampleCount(m_sampleCount, DXGI_FORMAT(rpD->colorFormat[0])); + const DXGI_SAMPLE_DESC sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, DXGI_FORMAT(rpD->colorFormat[0])); + + struct { + QD3D12PipelineStateSubObject<ID3D12RootSignature *, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_ROOT_SIGNATURE> rootSig; + QD3D12PipelineStateSubObject<D3D12_INPUT_LAYOUT_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_INPUT_LAYOUT> inputLayout; + QD3D12PipelineStateSubObject<D3D12_PRIMITIVE_TOPOLOGY_TYPE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_PRIMITIVE_TOPOLOGY> primitiveTopology; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_VS> VS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_HS> HS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DS> DS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_GS> GS; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_PS> PS; + QD3D12PipelineStateSubObject<D3D12_RASTERIZER_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_RASTERIZER> rasterizerState; + QD3D12PipelineStateSubObject<D3D12_DEPTH_STENCIL_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DEPTH_STENCIL> depthStencilState; + QD3D12PipelineStateSubObject<D3D12_BLEND_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_BLEND> blendState; + QD3D12PipelineStateSubObject<D3D12_RT_FORMAT_ARRAY, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_RENDER_TARGET_FORMATS> rtFormats; + QD3D12PipelineStateSubObject<DXGI_FORMAT, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_DEPTH_STENCIL_FORMAT> dsFormat; + QD3D12PipelineStateSubObject<DXGI_SAMPLE_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_SAMPLE_DESC> sampleDesc; + QD3D12PipelineStateSubObject<UINT, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_SAMPLE_MASK> sampleMask; + QD3D12PipelineStateSubObject<D3D12_VIEW_INSTANCING_DESC, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_VIEW_INSTANCING> viewInstancingDesc; + } stream; + + stream.rootSig.object = rootSig; + + QVarLengthArray<D3D12_INPUT_ELEMENT_DESC, 4> inputDescs; + QByteArrayList matrixSliceSemantics; + if (!shaderBytecode[VS].isEmpty()) { + for (auto it = m_vertexInputLayout.cbeginAttributes(), itEnd = m_vertexInputLayout.cendAttributes(); + it != itEnd; ++it) + { + D3D12_INPUT_ELEMENT_DESC desc = {}; + // The output from SPIRV-Cross uses TEXCOORD<location> as the + // semantic, except for matrices that are unrolled into consecutive + // vec2/3/4s attributes and need TEXCOORD<location>_ as + // SemanticName and row/column index as SemanticIndex. + const int matrixSlice = it->matrixSlice(); + if (matrixSlice < 0) { + desc.SemanticName = "TEXCOORD"; + desc.SemanticIndex = UINT(it->location()); + } else { + QByteArray sem; + sem.resize(16); + qsnprintf(sem.data(), sem.size(), "TEXCOORD%d_", it->location() - matrixSlice); + matrixSliceSemantics.append(sem); + desc.SemanticName = matrixSliceSemantics.last().constData(); + desc.SemanticIndex = UINT(matrixSlice); + } + desc.Format = toD3DAttributeFormat(it->format()); + desc.InputSlot = UINT(it->binding()); + desc.AlignedByteOffset = it->offset(); + const QRhiVertexInputBinding *inputBinding = m_vertexInputLayout.bindingAt(it->binding()); + if (inputBinding->classification() == QRhiVertexInputBinding::PerInstance) { + desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_INSTANCE_DATA; + desc.InstanceDataStepRate = inputBinding->instanceStepRate(); + } else { + desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA; + } + inputDescs.append(desc); + } + } + + stream.inputLayout.object.NumElements = inputDescs.count(); + stream.inputLayout.object.pInputElementDescs = inputDescs.isEmpty() ? nullptr : inputDescs.constData(); + + stream.primitiveTopology.object = toD3DTopologyType(m_topology); + topology = toD3DTopology(m_topology, m_patchControlPointCount); - D3D12_GRAPHICS_PIPELINE_STATE_DESC psoDesc = {}; - psoDesc.pRootSignature = rootSig; for (const QRhiShaderStage &shaderStage : std::as_const(m_shaderStages)) { const int d3dStage = qd3d12_stage(shaderStage.type()); switch (d3dStage) { case VS: - psoDesc.VS.pShaderBytecode = shaderBytecode[d3dStage].constData(); - psoDesc.VS.BytecodeLength = shaderBytecode[d3dStage].size(); + stream.VS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.VS.object.BytecodeLength = shaderBytecode[d3dStage].size(); break; case HS: - psoDesc.HS.pShaderBytecode = shaderBytecode[d3dStage].constData(); - psoDesc.HS.BytecodeLength = shaderBytecode[d3dStage].size(); + stream.HS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.HS.object.BytecodeLength = shaderBytecode[d3dStage].size(); break; case DS: - psoDesc.DS.pShaderBytecode = shaderBytecode[d3dStage].constData(); - psoDesc.DS.BytecodeLength = shaderBytecode[d3dStage].size(); + stream.DS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.DS.object.BytecodeLength = shaderBytecode[d3dStage].size(); break; case GS: - psoDesc.GS.pShaderBytecode = shaderBytecode[d3dStage].constData(); - psoDesc.GS.BytecodeLength = shaderBytecode[d3dStage].size(); + stream.GS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.GS.object.BytecodeLength = shaderBytecode[d3dStage].size(); break; case PS: - psoDesc.PS.pShaderBytecode = shaderBytecode[d3dStage].constData(); - psoDesc.PS.BytecodeLength = shaderBytecode[d3dStage].size(); + stream.PS.object.pShaderBytecode = shaderBytecode[d3dStage].constData(); + stream.PS.object.BytecodeLength = shaderBytecode[d3dStage].size(); break; default: Q_UNREACHABLE(); @@ -5192,7 +5749,32 @@ bool QD3D12GraphicsPipeline::create() } } - psoDesc.BlendState.IndependentBlendEnable = m_targetBlends.count() > 1; + stream.rasterizerState.object.FillMode = toD3DFillMode(m_polygonMode); + stream.rasterizerState.object.CullMode = toD3DCullMode(m_cullMode); + stream.rasterizerState.object.FrontCounterClockwise = m_frontFace == CCW; + stream.rasterizerState.object.DepthBias = m_depthBias; + stream.rasterizerState.object.SlopeScaledDepthBias = m_slopeScaledDepthBias; + stream.rasterizerState.object.DepthClipEnable = TRUE; + stream.rasterizerState.object.MultisampleEnable = sampleDesc.Count > 1; + + stream.depthStencilState.object.DepthEnable = m_depthTest; + stream.depthStencilState.object.DepthWriteMask = m_depthWrite ? D3D12_DEPTH_WRITE_MASK_ALL : D3D12_DEPTH_WRITE_MASK_ZERO; + stream.depthStencilState.object.DepthFunc = toD3DCompareOp(m_depthOp); + stream.depthStencilState.object.StencilEnable = m_stencilTest; + if (m_stencilTest) { + stream.depthStencilState.object.StencilReadMask = UINT8(m_stencilReadMask); + stream.depthStencilState.object.StencilWriteMask = UINT8(m_stencilWriteMask); + stream.depthStencilState.object.FrontFace.StencilFailOp = toD3DStencilOp(m_stencilFront.failOp); + stream.depthStencilState.object.FrontFace.StencilDepthFailOp = toD3DStencilOp(m_stencilFront.depthFailOp); + stream.depthStencilState.object.FrontFace.StencilPassOp = toD3DStencilOp(m_stencilFront.passOp); + stream.depthStencilState.object.FrontFace.StencilFunc = toD3DCompareOp(m_stencilFront.compareOp); + stream.depthStencilState.object.BackFace.StencilFailOp = toD3DStencilOp(m_stencilBack.failOp); + stream.depthStencilState.object.BackFace.StencilDepthFailOp = toD3DStencilOp(m_stencilBack.depthFailOp); + stream.depthStencilState.object.BackFace.StencilPassOp = toD3DStencilOp(m_stencilBack.passOp); + stream.depthStencilState.object.BackFace.StencilFunc = toD3DCompareOp(m_stencilBack.compareOp); + } + + stream.blendState.object.IndependentBlendEnable = m_targetBlends.count() > 1; for (int i = 0, ie = m_targetBlends.count(); i != ie; ++i) { const QRhiGraphicsPipeline::TargetBlend &b(m_targetBlends[i]); D3D12_RENDER_TARGET_BLEND_DESC blend = {}; @@ -5204,95 +5786,40 @@ bool QD3D12GraphicsPipeline::create() blend.DestBlendAlpha = toD3DBlendFactor(b.dstAlpha, false); blend.BlendOpAlpha = toD3DBlendOp(b.opAlpha); blend.RenderTargetWriteMask = toD3DColorWriteMask(b.colorWrite); - psoDesc.BlendState.RenderTarget[i] = blend; + stream.blendState.object.RenderTarget[i] = blend; } if (m_targetBlends.isEmpty()) { D3D12_RENDER_TARGET_BLEND_DESC blend = {}; blend.RenderTargetWriteMask = D3D12_COLOR_WRITE_ENABLE_ALL; - psoDesc.BlendState.RenderTarget[0] = blend; + stream.blendState.object.RenderTarget[0] = blend; } - psoDesc.SampleMask = 0xFFFFFFFF; + stream.rtFormats.object.NumRenderTargets = rpD->colorAttachmentCount; + for (int i = 0; i < rpD->colorAttachmentCount; ++i) + stream.rtFormats.object.RTFormats[i] = DXGI_FORMAT(rpD->colorFormat[i]); - psoDesc.RasterizerState.FillMode = toD3DFillMode(m_polygonMode); - psoDesc.RasterizerState.CullMode = toD3DCullMode(m_cullMode); - psoDesc.RasterizerState.FrontCounterClockwise = m_frontFace == CCW; - psoDesc.RasterizerState.DepthBias = m_depthBias; - psoDesc.RasterizerState.SlopeScaledDepthBias = m_slopeScaledDepthBias; - psoDesc.RasterizerState.DepthClipEnable = TRUE; - psoDesc.RasterizerState.MultisampleEnable = sampleDesc.Count > 1; + stream.dsFormat.object = rpD->hasDepthStencil ? DXGI_FORMAT(rpD->dsFormat) : DXGI_FORMAT_UNKNOWN; - psoDesc.DepthStencilState.DepthEnable = m_depthTest; - psoDesc.DepthStencilState.DepthWriteMask = m_depthWrite ? D3D12_DEPTH_WRITE_MASK_ALL : D3D12_DEPTH_WRITE_MASK_ZERO; - psoDesc.DepthStencilState.DepthFunc = toD3DCompareOp(m_depthOp); - psoDesc.DepthStencilState.StencilEnable = m_stencilTest; - if (m_stencilTest) { - psoDesc.DepthStencilState.StencilReadMask = UINT8(m_stencilReadMask); - psoDesc.DepthStencilState.StencilWriteMask = UINT8(m_stencilWriteMask); - psoDesc.DepthStencilState.FrontFace.StencilFailOp = toD3DStencilOp(m_stencilFront.failOp); - psoDesc.DepthStencilState.FrontFace.StencilDepthFailOp = toD3DStencilOp(m_stencilFront.depthFailOp); - psoDesc.DepthStencilState.FrontFace.StencilPassOp = toD3DStencilOp(m_stencilFront.passOp); - psoDesc.DepthStencilState.FrontFace.StencilFunc = toD3DCompareOp(m_stencilFront.compareOp); - psoDesc.DepthStencilState.BackFace.StencilFailOp = toD3DStencilOp(m_stencilBack.failOp); - psoDesc.DepthStencilState.BackFace.StencilDepthFailOp = toD3DStencilOp(m_stencilBack.depthFailOp); - psoDesc.DepthStencilState.BackFace.StencilPassOp = toD3DStencilOp(m_stencilBack.passOp); - psoDesc.DepthStencilState.BackFace.StencilFunc = toD3DCompareOp(m_stencilBack.compareOp); - } + stream.sampleDesc.object = sampleDesc; - QVarLengthArray<D3D12_INPUT_ELEMENT_DESC, 4> inputDescs; - QByteArrayList matrixSliceSemantics; - if (!shaderBytecode[VS].isEmpty()) { - for (auto it = m_vertexInputLayout.cbeginAttributes(), itEnd = m_vertexInputLayout.cendAttributes(); - it != itEnd; ++it) - { - D3D12_INPUT_ELEMENT_DESC desc = {}; - // The output from SPIRV-Cross uses TEXCOORD<location> as the - // semantic, except for matrices that are unrolled into consecutive - // vec2/3/4s attributes and need TEXCOORD<location>_ as - // SemanticName and row/column index as SemanticIndex. - const int matrixSlice = it->matrixSlice(); - if (matrixSlice < 0) { - desc.SemanticName = "TEXCOORD"; - desc.SemanticIndex = UINT(it->location()); - } else { - QByteArray sem; - sem.resize(16); - qsnprintf(sem.data(), sem.size(), "TEXCOORD%d_", it->location() - matrixSlice); - matrixSliceSemantics.append(sem); - desc.SemanticName = matrixSliceSemantics.last().constData(); - desc.SemanticIndex = UINT(matrixSlice); - } - desc.Format = toD3DAttributeFormat(it->format()); - desc.InputSlot = UINT(it->binding()); - desc.AlignedByteOffset = it->offset(); - const QRhiVertexInputBinding *inputBinding = m_vertexInputLayout.bindingAt(it->binding()); - if (inputBinding->classification() == QRhiVertexInputBinding::PerInstance) { - desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_INSTANCE_DATA; - desc.InstanceDataStepRate = inputBinding->instanceStepRate(); - } else { - desc.InputSlotClass = D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA; - } - inputDescs.append(desc); + stream.sampleMask.object = 0xFFFFFFFF; + + viewInstanceMask = 0; + const bool isMultiView = m_multiViewCount >= 2; + stream.viewInstancingDesc.object.ViewInstanceCount = isMultiView ? m_multiViewCount : 0; + QVarLengthArray<D3D12_VIEW_INSTANCE_LOCATION, 4> viewInstanceLocations; + if (isMultiView) { + for (int i = 0; i < m_multiViewCount; ++i) { + viewInstanceMask |= (1 << i); + viewInstanceLocations.append({ 0, UINT(i) }); } - } - if (!inputDescs.isEmpty()) { - psoDesc.InputLayout.pInputElementDescs = inputDescs.constData(); - psoDesc.InputLayout.NumElements = inputDescs.count(); + stream.viewInstancingDesc.object.pViewInstanceLocations = viewInstanceLocations.constData(); } - psoDesc.PrimitiveTopologyType = toD3DTopologyType(m_topology); - topology = toD3DTopology(m_topology, m_patchControlPointCount); - - psoDesc.NumRenderTargets = rpD->colorAttachmentCount; - for (int i = 0; i < rpD->colorAttachmentCount; ++i) - psoDesc.RTVFormats[i] = DXGI_FORMAT(rpD->colorFormat[i]); - psoDesc.DSVFormat = rpD->hasDepthStencil ? DXGI_FORMAT(rpD->dsFormat) : DXGI_FORMAT_UNKNOWN; - psoDesc.SampleDesc = sampleDesc; + const D3D12_PIPELINE_STATE_STREAM_DESC streamDesc = { sizeof(stream), &stream }; ID3D12PipelineState *pso = nullptr; - HRESULT hr = rhiD->dev->CreateGraphicsPipelineState(&psoDesc, - __uuidof(ID3D12PipelineState), - reinterpret_cast<void **>(&pso)); + HRESULT hr = rhiD->dev->CreatePipelineState(&streamDesc, __uuidof(ID3D12PipelineState), reinterpret_cast<void **>(&pso)); if (FAILED(hr)) { qWarning("Failed to create graphics pipeline state: %s", qPrintable(QSystemError::windowsComString(hr))); @@ -5356,9 +5883,9 @@ bool QD3D12ComputePipeline::create() } else { QString error; QShaderKey shaderKey; - UINT compileFlags = 0; + int compileFlags = 0; if (m_flags.testFlag(CompileShadersWithDebugInfo)) - compileFlags |= D3DCOMPILE_DEBUG; + compileFlags |= int(HlslCompileFlag::WithDebugInfo); const QByteArray bytecode = compileHlslShaderSource(m_shaderStage.shader(), m_shaderStage.shaderVariant(), compileFlags, @@ -5391,14 +5918,16 @@ bool QD3D12ComputePipeline::create() return false; } - D3D12_COMPUTE_PIPELINE_STATE_DESC psoDesc = {}; - psoDesc.pRootSignature = rootSig; - psoDesc.CS.pShaderBytecode = shaderBytecode.constData(); - psoDesc.CS.BytecodeLength = shaderBytecode.size(); + struct { + QD3D12PipelineStateSubObject<ID3D12RootSignature *, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_ROOT_SIGNATURE> rootSig; + QD3D12PipelineStateSubObject<D3D12_SHADER_BYTECODE, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE_CS> CS; + } stream; + stream.rootSig.object = rootSig; + stream.CS.object.pShaderBytecode = shaderBytecode.constData(); + stream.CS.object.BytecodeLength = shaderBytecode.size(); + const D3D12_PIPELINE_STATE_STREAM_DESC streamDesc = { sizeof(stream), &stream }; ID3D12PipelineState *pso = nullptr; - HRESULT hr = rhiD->dev->CreateComputePipelineState(&psoDesc, - __uuidof(ID3D12PipelineState), - reinterpret_cast<void **>(&pso)); + HRESULT hr = rhiD->dev->CreatePipelineState(&streamDesc, __uuidof(ID3D12PipelineState), reinterpret_cast<void **>(&pso)); if (FAILED(hr)) { qWarning("Failed to create compute pipeline state: %s", qPrintable(QSystemError::windowsComString(hr))); @@ -5431,7 +5960,9 @@ QD3D12RenderPassDescriptor::~QD3D12RenderPassDescriptor() void QD3D12RenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiD3D12); + if (rhiD) + rhiD->unregisterResource(this); } bool QD3D12RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -5474,13 +6005,17 @@ void QD3D12RenderPassDescriptor::updateSerializedFormat() QRhiRenderPassDescriptor *QD3D12RenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - QD3D12RenderPassDescriptor *rp = new QD3D12RenderPassDescriptor(m_rhi); - rp->colorAttachmentCount = colorAttachmentCount; - rp->hasDepthStencil = hasDepthStencil; - memcpy(rp->colorFormat, colorFormat, sizeof(colorFormat)); - rp->dsFormat = dsFormat; - rp->updateSerializedFormat(); - return rp; + QD3D12RenderPassDescriptor *rpD = new QD3D12RenderPassDescriptor(m_rhi); + rpD->colorAttachmentCount = colorAttachmentCount; + rpD->hasDepthStencil = hasDepthStencil; + memcpy(rpD->colorFormat, colorFormat, sizeof(colorFormat)); + rpD->dsFormat = dsFormat; + + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); + return rpD; } QVector<quint32> QD3D12RenderPassDescriptor::serializedFormat() const @@ -5544,6 +6079,7 @@ int QD3D12SwapChainRenderTarget::sampleCount() const QD3D12SwapChain::QD3D12SwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), rtWrapper(rhi, this), + rtWrapperRight(rhi, this), cbWrapper(rhi) { } @@ -5600,6 +6136,8 @@ void QD3D12SwapChain::releaseBuffers() for (UINT i = 0; i < BUFFER_COUNT; ++i) { rhiD->resourcePool.remove(colorBuffers[i]); rhiD->rtvPool.release(rtvs[i], 1); + if (stereo) + rhiD->rtvPool.release(rtvsRight[i], 1); if (!msaaBuffers[i].isNull()) rhiD->resourcePool.remove(msaaBuffers[i]); if (msaaRtvs[i].isValid()) @@ -5634,48 +6172,15 @@ QRhiRenderTarget *QD3D12SwapChain::currentFrameRenderTarget() return &rtWrapper; } -QSize QD3D12SwapChain::surfacePixelSize() +QRhiRenderTarget *QD3D12SwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) { - Q_ASSERT(m_window); - return m_window->size() * m_window->devicePixelRatio(); -} - -static bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result) -{ - bool ok = false; - QRect wr = w->geometry(); - wr = QRect(wr.topLeft() * w->devicePixelRatio(), wr.size() * w->devicePixelRatio()); - const QPoint center = wr.center(); - IDXGIOutput *currentOutput = nullptr; - IDXGIOutput *output = nullptr; - for (UINT i = 0; adapter->EnumOutputs(i, &output) != DXGI_ERROR_NOT_FOUND; ++i) { - DXGI_OUTPUT_DESC desc; - output->GetDesc(&desc); - const RECT r = desc.DesktopCoordinates; - const QRect dr(QPoint(r.left, r.top), QPoint(r.right - 1, r.bottom - 1)); - if (dr.contains(center)) { - currentOutput = output; - break; - } else { - output->Release(); - } - } - if (currentOutput) { - ok = SUCCEEDED(currentOutput->QueryInterface(__uuidof(IDXGIOutput6), reinterpret_cast<void **>(result))); - currentOutput->Release(); - } - return ok; + return !stereo || targetBuffer == StereoTargetBuffer::LeftBuffer ? &rtWrapper : &rtWrapperRight; } -static bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result) +QSize QD3D12SwapChain::surfacePixelSize() { - bool ok = false; - IDXGIOutput6 *out6 = nullptr; - if (output6ForWindow(w, adapter, &out6)) { - ok = SUCCEEDED(out6->GetDesc1(result)); - out6->Release(); - } - return ok; + Q_ASSERT(m_window); + return m_window->size() * m_window->devicePixelRatio(); } bool QD3D12SwapChain::isFormatSupported(Format f) @@ -5690,8 +6195,10 @@ bool QD3D12SwapChain::isFormatSupported(Format f) QRHI_RES_RHI(QRhiD3D12); DXGI_OUTPUT_DESC1 desc1; - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) - return desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &desc1)) { + if (desc1.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) + return f == QRhiSwapChain::HDRExtendedSrgbLinear || f == QRhiSwapChain::HDR10; + } return false; } @@ -5699,14 +6206,16 @@ bool QD3D12SwapChain::isFormatSupported(Format f) QRhiSwapChainHdrInfo QD3D12SwapChain::hdrInfo() { QRhiSwapChainHdrInfo info = QRhiSwapChain::hdrInfo(); - if (m_format != QRhiSwapChain::SDR && m_window) { + // Must use m_window, not window, given this may be called before createOrResize(). + if (m_window) { QRHI_RES_RHI(QRhiD3D12); DXGI_OUTPUT_DESC1 hdrOutputDesc; - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { - info.isHardCodedDefaults = false; + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc)) { info.limitsType = QRhiSwapChainHdrInfo::LuminanceInNits; info.limits.luminanceInNits.minLuminance = hdrOutputDesc.MinLuminance; info.limits.luminanceInNits.maxLuminance = hdrOutputDesc.MaxLuminance; + info.luminanceBehavior = QRhiSwapChainHdrInfo::SceneReferred; // 1.0 = 80 nits + info.sdrWhiteLevel = QRhiD3D::sdrWhiteLevelInNits(hdrOutputDesc); } } return info; @@ -5723,28 +6232,25 @@ QRhiRenderPassDescriptor *QD3D12SwapChain::newCompatibleRenderPassDescriptor() rpD->colorFormat[0] = int(srgbAdjustedColorFormat); rpD->dsFormat = QD3D12RenderBuffer::DS_FORMAT; rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiD3D12); + rhiD->registerResource(rpD); return rpD; } -static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; -static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; - bool QRhiD3D12::ensureDirectCompositionDevice() { if (dcompDevice) return true; qCDebug(QRHI_LOG_INFO, "Creating Direct Composition device (needed for semi-transparent windows)"); - - HRESULT hr = DCompositionCreateDevice(nullptr, __uuidof(IDCompositionDevice), reinterpret_cast<void **>(&dcompDevice)); - if (FAILED(hr)) { - qWarning("Failed to Direct Composition device: %s", qPrintable(QSystemError::windowsComString(hr))); - return false; - } - - return true; + dcompDevice = QRhiD3D::createDirectCompositionDevice(); + return dcompDevice ? true : false; } +static const DXGI_FORMAT DEFAULT_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM; +static const DXGI_FORMAT DEFAULT_SRGB_FORMAT = DXGI_FORMAT_R8G8B8A8_UNORM_SRGB; + void QD3D12SwapChain::chooseFormats() { colorFormat = DEFAULT_FORMAT; @@ -5752,7 +6258,7 @@ void QD3D12SwapChain::chooseFormats() hdrColorSpace = DXGI_COLOR_SPACE_RGB_FULL_G22_NONE_P709; // SDR DXGI_OUTPUT_DESC1 hdrOutputDesc; QRHI_RES_RHI(QRhiD3D12); - if (outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { + if (QRhiD3D::outputDesc1ForWindow(m_window, rhiD->activeAdapter, &hdrOutputDesc) && m_format != SDR) { // https://docs.microsoft.com/en-us/windows/win32/direct3darticles/high-dynamic-range if (hdrOutputDesc.ColorSpace == DXGI_COLOR_SPACE_RGB_FULL_G2084_NONE_P2020) { switch (m_format) { @@ -5777,7 +6283,7 @@ void QD3D12SwapChain::chooseFormats() "(or Use HDR is Off in the Display Settings), ignoring HDR format request"); } } - sampleDesc = rhiD->effectiveSampleCount(m_sampleCount, colorFormat); + sampleDesc = rhiD->effectiveSampleDesc(m_sampleCount, colorFormat); } bool QD3D12SwapChain::createOrResize() @@ -5802,13 +6308,14 @@ bool QD3D12SwapChain::createOrResize() HWND hwnd = reinterpret_cast<HWND>(window->winId()); HRESULT hr; QRHI_RES_RHI(QRhiD3D12); + stereo = m_window->format().stereo() && rhiD->dxgiFactory->IsWindowedStereoEnabled(); if (m_flags.testFlag(SurfaceHasPreMulAlpha) || m_flags.testFlag(SurfaceHasNonPreMulAlpha)) { if (rhiD->ensureDirectCompositionDevice()) { if (!dcompTarget) { - hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, true, &dcompTarget); + hr = rhiD->dcompDevice->CreateTargetForHwnd(hwnd, false, &dcompTarget); if (FAILED(hr)) { - qWarning("Failed to create Direct Compsition target for the window: %s", + qWarning("Failed to create Direct Composition target for the window: %s", qPrintable(QSystemError::windowsComString(hr))); } } @@ -5844,6 +6351,7 @@ bool QD3D12SwapChain::createOrResize() desc.Flags = swapChainFlags; desc.Scaling = DXGI_SCALING_NONE; desc.SwapEffect = DXGI_SWAP_EFFECT_FLIP_DISCARD; + desc.Stereo = stereo; if (dcompVisual) { // With DirectComposition setting AlphaMode to STRAIGHT fails the @@ -5898,13 +6406,19 @@ bool QD3D12SwapChain::createOrResize() qWarning("Failed to set content for Direct Composition visual: %s", qPrintable(QSystemError::windowsComString(hr))); } + } else { + // disable Alt+Enter; not relevant when using DirectComposition + rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); } } if (FAILED(hr)) { - qWarning("Failed to create D3D12 swapchain: %s", qPrintable(QSystemError::windowsComString(hr))); + qWarning("Failed to create D3D12 swapchain: %s" + " (Width=%u Height=%u Format=%u SampleCount=%u BufferCount=%u Scaling=%u SwapEffect=%u Stereo=%u)", + qPrintable(QSystemError::windowsComString(hr)), + desc.Width, desc.Height, UINT(desc.Format), desc.SampleDesc.Count, + desc.BufferCount, UINT(desc.Scaling), UINT(desc.SwapEffect), UINT(desc.Stereo)); return false; } - rhiD->dxgiFactory->MakeWindowAssociation(hwnd, DXGI_MWA_NO_WINDOW_CHANGES); for (int i = 0; i < QD3D12_FRAMES_IN_FLIGHT; ++i) { hr = rhiD->dev->CreateFence(0, @@ -5951,6 +6465,16 @@ bool QD3D12SwapChain::createOrResize() rtvDesc.Format = srgbAdjustedColorFormat; rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2D; rhiD->dev->CreateRenderTargetView(colorBuffer, &rtvDesc, rtvs[i].cpuHandle); + + if (stereo) { + rtvsRight[i] = rhiD->rtvPool.allocate(1); + D3D12_RENDER_TARGET_VIEW_DESC rtvDesc = {}; + rtvDesc.Format = srgbAdjustedColorFormat; + rtvDesc.ViewDimension = D3D12_RTV_DIMENSION_TEXTURE2DARRAY; + rtvDesc.Texture2DArray.ArraySize = 1; + rtvDesc.Texture2DArray.FirstArraySlice = 1; + rhiD->dev->CreateRenderTargetView(colorBuffer, &rtvDesc, rtvsRight[i].cpuHandle); + } } if (m_depthStencil && m_depthStencil->sampleCount() != m_sampleCount) { @@ -6023,6 +6547,15 @@ bool QD3D12SwapChain::createOrResize() rtD->d.colorAttCount = 1; rtD->d.dsAttCount = m_depthStencil ? 1 : 0; + rtWrapperRight.setRenderPassDescriptor(m_renderPassDesc); + QD3D12SwapChainRenderTarget *rtDr = QRHI_RES(QD3D12SwapChainRenderTarget, &rtWrapperRight); + rtDr->d.rp = QRHI_RES(QD3D12RenderPassDescriptor, m_renderPassDesc); + rtDr->d.pixelSize = pixelSize; + rtDr->d.dpr = float(window->devicePixelRatio()); + rtDr->d.sampleCount = int(sampleDesc.Count); + rtDr->d.colorAttCount = 1; + rtDr->d.dsAttCount = m_depthStencil ? 1 : 0; + if (needsRegistration) { rhiD->swapchains.insert(this); rhiD->registerResource(this); @@ -6032,3 +6565,5 @@ bool QD3D12SwapChain::createOrResize() } QT_END_NAMESPACE + +#endif // __ID3D12Device2_INTERFACE_DEFINED__ diff --git a/src/gui/rhi/qrhid3d12_p.h b/src/gui/rhi/qrhid3d12_p.h index e49ef57eaf..3f9abbb5ac 100644 --- a/src/gui/rhi/qrhid3d12_p.h +++ b/src/gui/rhi/qrhid3d12_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2022 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHID3D12_H -#define QRHID3D12_H +#ifndef QRHID3D12_P_H +#define QRHID3D12_P_H // // W A R N I N G @@ -15,34 +15,1234 @@ // We mean it. // -#include <private/qrhi_p.h> +#include "qrhi_p.h" +#include <QWindow> +#include <QBitArray> -// no d3d includes here, to prevent precompiled header mess due to COM +#include <optional> +#include <array> + +#include <d3d12.h> +#include <d3d12sdklayers.h> +#include <dxgi1_6.h> +#include <dcomp.h> + +#include "D3D12MemAlloc.h" + +// ID3D12Device2 and ID3D12GraphicsCommandList1 and types and enums introduced +// with those are hard requirements now. These should be declared in any +// moderately recent d3d12.h, but if it is an SDK from before Windows 10 +// version 1703 then these types could be missing. In the absence of other +// options, handle this by skipping all the code and making QRhi::create() fail +// in such builds. +#ifdef __ID3D12Device2_INTERFACE_DEFINED__ +#define QRHI_D3D12_AVAILABLE QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiD3D12InitParams : public QRhiInitParams +static const int QD3D12_FRAMES_IN_FLIGHT = 2; + +class QRhiD3D12; + +struct QD3D12Descriptor +{ + D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = {}; + D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = {}; + + bool isValid() const { return cpuHandle.ptr != 0; } +}; + +struct QD3D12ReleaseQueue; + +struct QD3D12DescriptorHeap +{ + bool isValid() const { return heap && capacity; } + bool create(ID3D12Device *device, + quint32 descriptorCount, + D3D12_DESCRIPTOR_HEAP_TYPE heapType, + D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags); + void createWithExisting(const QD3D12DescriptorHeap &other, + quint32 offsetInDescriptors, + quint32 descriptorCount); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + QD3D12Descriptor get(quint32 count); + QD3D12Descriptor at(quint32 index) const; + quint32 remainingCapacity() const { return capacity - head; } + + QD3D12Descriptor incremented(const QD3D12Descriptor &descriptor, quint32 offsetInDescriptors) const + { + D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = descriptor.cpuHandle; + cpuHandle.ptr += offsetInDescriptors * descriptorByteSize; + D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = descriptor.gpuHandle; + if (gpuHandle.ptr) + gpuHandle.ptr += offsetInDescriptors * descriptorByteSize; + return { cpuHandle, gpuHandle }; + } + + ID3D12DescriptorHeap *heap = nullptr; + quint32 capacity = 0; + QD3D12Descriptor heapStart; + quint32 head = 0; + quint32 descriptorByteSize = 0; + D3D12_DESCRIPTOR_HEAP_TYPE heapType; + D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags; +}; + +struct QD3D12CpuDescriptorPool +{ + bool isValid() const { return !heaps.isEmpty(); } + bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE heapType, const char *debugName = ""); + void destroy(); + + QD3D12Descriptor allocate(quint32 count); + void release(const QD3D12Descriptor &descriptor, quint32 count); + + static const int DESCRIPTORS_PER_HEAP = 256; + + struct HeapWithMap { + QD3D12DescriptorHeap heap; + QBitArray map; + static HeapWithMap init(const QD3D12DescriptorHeap &heap, quint32 descriptorCount) { + HeapWithMap result; + result.heap = heap; + result.map.resize(descriptorCount); + return result; + } + }; + + ID3D12Device *device; + quint32 descriptorByteSize; + QVector<HeapWithMap> heaps; + const char *debugName; +}; + +struct QD3D12QueryHeap +{ + bool isValid() const { return heap && capacity; } + bool create(ID3D12Device *device, + quint32 queryCount, + D3D12_QUERY_HEAP_TYPE heapType); + void destroy(); + + ID3D12QueryHeap *heap = nullptr; + quint32 capacity = 0; +}; + +struct QD3D12StagingArea +{ + static const quint32 ALIGNMENT = D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT; // 512 so good enough both for cb and texdata + + struct Allocation { + quint8 *p = nullptr; + D3D12_GPU_VIRTUAL_ADDRESS gpuAddr = 0; + ID3D12Resource *buffer = nullptr; + quint32 bufferOffset = 0; + bool isValid() const { return p != nullptr; } + }; + + bool isValid() const { return allocation && mem.isValid(); } + bool create(QRhiD3D12 *rhi, quint32 capacity, D3D12_HEAP_TYPE heapType); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + Allocation get(quint32 byteSize); + + quint32 remainingCapacity() const + { + return capacity - head; + } + + static quint32 allocSizeForArray(quint32 size, int count = 1) + { + return count * ((size + ALIGNMENT - 1) & ~(ALIGNMENT - 1)); + } + + Allocation mem; + ID3D12Resource *resource = nullptr; + D3D12MA::Allocation *allocation = nullptr; + quint32 head; + quint32 capacity; +}; + +struct QD3D12ObjectHandle +{ + quint32 index = 0; + quint32 generation = 0; + + // the default, null handle is guaranteed to give ObjectPool::isValid() == false + bool isNull() const { return index == 0 && generation == 0; } +}; + +inline bool operator==(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept +{ + return a.index == b.index && a.generation == b.generation; +} + +inline bool operator!=(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept +{ + return !(a == b); +} + +template<typename T> +struct QD3D12ObjectPool +{ + void create(const char *debugName = "") + { + this->debugName = debugName; + Q_ASSERT(data.isEmpty()); + data.append(Data()); // index 0 is always invalid + } + + void destroy() { + int leakCount = 0; // will nicely destroy everything here, but warn about it if enabled + for (Data &d : data) { + if (d.object.has_value()) { + leakCount += 1; + d.object->releaseResources(); + } + } + data.clear(); +#ifndef QT_NO_DEBUG + // debug builds: just do it always + static bool leakCheck = true; +#else + // release builds: opt-in + static bool leakCheck = qEnvironmentVariableIntValue("QT_RHI_LEAK_CHECK"); +#endif + if (leakCheck) { + if (leakCount > 0) { + qWarning("QD3D12ObjectPool::destroy(): Pool %p '%s' had %d unreleased objects", + this, debugName, leakCount); + } + } + } + + bool isValid(const QD3D12ObjectHandle &handle) const + { + return handle.index > 0 + && handle.index < quint32(data.count()) + && handle.generation > 0 + && handle.generation == data[handle.index].generation + && data[handle.index].object.has_value(); + } + + T lookup(const QD3D12ObjectHandle &handle) const + { + return isValid(handle) ? *data[handle.index].object : T(); + } + + const T *lookupRef(const QD3D12ObjectHandle &handle) const + { + return isValid(handle) ? &*data[handle.index].object : nullptr; + } + + T *lookupRef(const QD3D12ObjectHandle &handle) + { + return isValid(handle) ? &*data[handle.index].object : nullptr; + } + + QD3D12ObjectHandle add(const T &object) + { + Q_ASSERT(!data.isEmpty()); + const quint32 count = quint32(data.count()); + quint32 index = 1; // index 0 is always invalid + for (; index < count; ++index) { + if (!data[index].object.has_value()) + break; + } + if (index < count) { + data[index].object = object; + quint32 &generation = data[index].generation; + generation += 1u; + return { index, generation }; + } else { + data.append({ object, 1 }); + return { count, 1 }; + } + } + + void remove(const QD3D12ObjectHandle &handle) + { + if (T *object = lookupRef(handle)) { + object->releaseResources(); + data[handle.index].object.reset(); + } + } + + const char *debugName; + struct Data { + std::optional<T> object; + quint32 generation = 0; + }; + QVector<Data> data; +}; + +struct QD3D12Resource +{ + ID3D12Resource *resource; + D3D12_RESOURCE_STATES state; + D3D12_RESOURCE_DESC desc; + D3D12MA::Allocation *allocation; + void *cpuMapPtr; + enum { UavUsageRead = 0x01, UavUsageWrite = 0x02 }; + int uavUsage; + bool owns; + + // note that this assumes the allocation (if there is one) and the resource + // are separately releaseable, see D3D12MemAlloc docs + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Resource> *pool, + ID3D12Resource *resource, + D3D12_RESOURCE_STATES state, + D3D12MA::Allocation *allocation = nullptr, + void *cpuMapPtr = nullptr) + { + Q_ASSERT(resource); + return pool->add({ resource, state, resource->GetDesc(), allocation, cpuMapPtr, 0, true }); + } + + // for QRhiTexture::createFrom() where the ID3D12Resource is not owned by us + static QD3D12ObjectHandle addNonOwningToPool(QD3D12ObjectPool<QD3D12Resource> *pool, + ID3D12Resource *resource, + D3D12_RESOURCE_STATES state) + { + Q_ASSERT(resource); + return pool->add({ resource, state, resource->GetDesc(), nullptr, nullptr, 0, false }); + } + + void releaseResources() + { + if (owns) { + // order matters: resource first, then the allocation + resource->Release(); + if (allocation) + allocation->Release(); + } + } +}; + +struct QD3D12Pipeline +{ + enum Type { + Graphics, + Compute + }; + Type type; + ID3D12PipelineState *pso; + + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Pipeline> *pool, + Type type, + ID3D12PipelineState *pso) + { + return pool->add({ type, pso }); + } + + void releaseResources() + { + pso->Release(); + } +}; + +struct QD3D12RootSignature { - bool enableDebugLayer = false; + ID3D12RootSignature *rootSig; + + static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12RootSignature> *pool, + ID3D12RootSignature *rootSig) + { + return pool->add({ rootSig }); + } + + void releaseResources() + { + rootSig->Release(); + } }; -struct Q_GUI_EXPORT QRhiD3D12NativeHandles : public QRhiNativeHandles +struct QD3D12ReleaseQueue { - // to import a device - void *dev = nullptr; - int minimumFeatureLevel = 0; - // to just specify the adapter to use, set these and leave dev set to null - quint32 adapterLuidLow = 0; - qint32 adapterLuidHigh = 0; - // in addition, can specify the command queue to use - void *commandQueue = nullptr; + void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool, + QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool, + QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool) + { + this->resourcePool = resourcePool; + this->pipelinePool = pipelinePool; + this->rootSignaturePool = rootSignaturePool; + } + + void deferredReleaseResource(const QD3D12ObjectHandle &handle); + void deferredReleaseResourceWithViews(const QD3D12ObjectHandle &handle, + QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount); + void deferredReleasePipeline(const QD3D12ObjectHandle &handle); + void deferredReleaseRootSignature(const QD3D12ObjectHandle &handle); + void deferredReleaseCallback(std::function<void(void*)> callback, void *userData); + void deferredReleaseResourceAndAllocation(ID3D12Resource *resource, + D3D12MA::Allocation *allocation); + void deferredReleaseDescriptorHeap(ID3D12DescriptorHeap *heap); + void deferredReleaseViews(QD3D12CpuDescriptorPool *pool, + const QD3D12Descriptor &viewsStart, + int viewCount); + + void activatePendingDeferredReleaseRequests(int frameSlot); + void executeDeferredReleases(int frameSlot, bool forced = false); + void releaseAll(); + + struct DeferredReleaseEntry { + enum Type { + Resource, + Pipeline, + RootSignature, + Callback, + ResourceAndAllocation, + DescriptorHeap, + Views + }; + Type type = Resource; + std::optional<int> frameSlotToBeReleasedIn; + QD3D12ObjectHandle handle; + QD3D12CpuDescriptorPool *poolForViews = nullptr; + QD3D12Descriptor viewsStart; + int viewCount = 0; + std::function<void(void*)> callback = nullptr; + void *callbackUserData = nullptr; + QPair<ID3D12Resource *, D3D12MA::Allocation *> resourceAndAllocation = {}; + ID3D12DescriptorHeap *descriptorHeap = nullptr; + }; + QVector<DeferredReleaseEntry> queue; + QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; + QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool = nullptr; + QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool = nullptr; }; -struct Q_GUI_EXPORT QRhiD3D12CommandBufferNativeHandles : public QRhiNativeHandles +struct QD3D12CommandBuffer; + +struct QD3D12ResourceBarrierGenerator { - void *commandList = nullptr; // ID3D12GraphicsCommandList + static const int PREALLOC = 16; + + void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool) + { + this->resourcePool = resourcePool; + } + + void addTransitionBarrier(const QD3D12ObjectHandle &resourceHandle, D3D12_RESOURCE_STATES stateAfter); + void enqueueBufferedTransitionBarriers(QD3D12CommandBuffer *cbD); + void enqueueSubresourceTransitionBarrier(QD3D12CommandBuffer *cbD, + const QD3D12ObjectHandle &resourceHandle, + UINT subresource, + D3D12_RESOURCE_STATES stateBefore, + D3D12_RESOURCE_STATES stateAfter); + void enqueueUavBarrier(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &resourceHandle); + + struct TransitionResourceBarrier { + QD3D12ObjectHandle resourceHandle; + D3D12_RESOURCE_STATES stateBefore; + D3D12_RESOURCE_STATES stateAfter; + }; + QVarLengthArray<TransitionResourceBarrier, PREALLOC> transitionResourceBarriers; + QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; +}; + +struct QD3D12ShaderBytecodeCache +{ + struct Shader { + Shader() = default; + Shader(const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) + : bytecode(bytecode), nativeResourceBindingMap(rbm) + { } + QByteArray bytecode; + QShader::NativeResourceBindingMap nativeResourceBindingMap; + }; + + QHash<QRhiShaderStage, Shader> data; + + void insertWithCapacityLimit(const QRhiShaderStage &key, const Shader &s); +}; + +struct QD3D12ShaderVisibleDescriptorHeap +{ + bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE type, quint32 perFrameDescriptorCount); + void destroy(); + void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); + + QD3D12DescriptorHeap heap; + QD3D12DescriptorHeap perFrameHeapSlice[QD3D12_FRAMES_IN_FLIGHT]; +}; + +// wrap foreign struct so we can legally supply equality operators and qHash: +struct Q_D3D12_SAMPLER_DESC +{ + D3D12_SAMPLER_DESC desc; + + friend bool operator==(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept + { + return lhs.desc.Filter == rhs.desc.Filter + && lhs.desc.AddressU == rhs.desc.AddressU + && lhs.desc.AddressV == rhs.desc.AddressV + && lhs.desc.AddressW == rhs.desc.AddressW + && lhs.desc.MipLODBias == rhs.desc.MipLODBias + && lhs.desc.MaxAnisotropy == rhs.desc.MaxAnisotropy + && lhs.desc.ComparisonFunc == rhs.desc.ComparisonFunc + // BorderColor is never used, skip it + && lhs.desc.MinLOD == rhs.desc.MinLOD + && lhs.desc.MaxLOD == rhs.desc.MaxLOD; + } + + friend bool operator!=(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept + { + return !(lhs == rhs); + } + + friend size_t qHash(const Q_D3D12_SAMPLER_DESC &key, size_t seed = 0) noexcept + { + QtPrivate::QHashCombine hash; + seed = hash(seed, key.desc.Filter); + seed = hash(seed, key.desc.AddressU); + seed = hash(seed, key.desc.AddressV); + seed = hash(seed, key.desc.AddressW); + seed = hash(seed, key.desc.MipLODBias); + seed = hash(seed, key.desc.MaxAnisotropy); + seed = hash(seed, key.desc.ComparisonFunc); + // BorderColor is never used, skip it + seed = hash(seed, key.desc.MinLOD); + seed = hash(seed, key.desc.MaxLOD); + return seed; + } +}; + +struct QD3D12SamplerManager +{ + const quint32 MAX_SAMPLERS = 512; + + bool create(ID3D12Device *device); + void destroy(); + + QD3D12Descriptor getShaderVisibleDescriptor(const D3D12_SAMPLER_DESC &desc); + + ID3D12Device *device = nullptr; + QD3D12ShaderVisibleDescriptorHeap shaderVisibleSamplerHeap; + QHash<Q_D3D12_SAMPLER_DESC, QD3D12Descriptor> gpuMap; +}; + +enum QD3D12Stage { VS = 0, HS, DS, GS, PS, CS }; + +static inline QD3D12Stage qd3d12_stage(QRhiShaderStage::Type type) +{ + switch (type) { + case QRhiShaderStage::Vertex: + return VS; + case QRhiShaderStage::TessellationControl: + return HS; + case QRhiShaderStage::TessellationEvaluation: + return DS; + case QRhiShaderStage::Geometry: + return GS; + case QRhiShaderStage::Fragment: + return PS; + case QRhiShaderStage::Compute: + return CS; + } + Q_UNREACHABLE_RETURN(VS); +} + +static inline D3D12_SHADER_VISIBILITY qd3d12_stageToVisibility(QD3D12Stage s) +{ + switch (s) { + case VS: + return D3D12_SHADER_VISIBILITY_VERTEX; + case HS: + return D3D12_SHADER_VISIBILITY_HULL; + case DS: + return D3D12_SHADER_VISIBILITY_DOMAIN; + case GS: + return D3D12_SHADER_VISIBILITY_GEOMETRY; + case PS: + return D3D12_SHADER_VISIBILITY_PIXEL; + case CS: + return D3D12_SHADER_VISIBILITY_ALL; + } + Q_UNREACHABLE_RETURN(D3D12_SHADER_VISIBILITY_ALL); +} + +static inline QRhiShaderResourceBinding::StageFlag qd3d12_stageToSrb(QD3D12Stage s) +{ + switch (s) { + case VS: + return QRhiShaderResourceBinding::VertexStage; + case HS: + return QRhiShaderResourceBinding::TessellationControlStage; + case DS: + return QRhiShaderResourceBinding::TessellationEvaluationStage; + case GS: + return QRhiShaderResourceBinding::GeometryStage; + case PS: + return QRhiShaderResourceBinding::FragmentStage; + case CS: + return QRhiShaderResourceBinding::ComputeStage; + } + Q_UNREACHABLE_RETURN(QRhiShaderResourceBinding::VertexStage); +} + +struct QD3D12ShaderStageData +{ + bool valid = false; // to allow simple arrays where unused stages are indicated by !valid + QD3D12Stage stage = VS; + QShader::NativeResourceBindingMap nativeResourceBindingMap; +}; + +struct QD3D12ShaderResourceBindings; + +struct QD3D12ShaderResourceVisitor +{ + enum StorageOp { Load = 0, Store, LoadStore }; + + QD3D12ShaderResourceVisitor(const QD3D12ShaderResourceBindings *srb, + const QD3D12ShaderStageData *stageData, + int stageCount) + : srb(srb), + stageData(stageData), + stageCount(stageCount) + { + } + + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::UniformBufferData &, int, int)> uniformBuffer = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> texture = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> sampler = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageImageData &, StorageOp, int)> storageImage = nullptr; + std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageBufferData &, StorageOp, int)> storageBuffer = nullptr; + + void visit(); + + const QD3D12ShaderResourceBindings *srb; + const QD3D12ShaderStageData *stageData; + int stageCount; +}; + +struct QD3D12Readback +{ + // common + int frameSlot = -1; + QRhiReadbackResult *result = nullptr; + QD3D12StagingArea staging; + quint32 byteSize = 0; + // textures + quint32 bytesPerLine = 0; + QSize pixelSize; + QRhiTexture::Format format = QRhiTexture::UnknownFormat; + quint32 stagingRowPitch = 0; +}; + +struct QD3D12MipmapGenerator +{ + bool create(QRhiD3D12 *rhiD); + void destroy(); + void generate(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &textureHandle); + + QRhiD3D12 *rhiD; + QD3D12ObjectHandle rootSigHandle; + QD3D12ObjectHandle pipelineHandle; +}; + +struct QD3D12MemoryAllocator +{ + bool create(ID3D12Device *device, IDXGIAdapter1 *adapter); + void destroy(); + + HRESULT createResource(D3D12_HEAP_TYPE heapType, + const D3D12_RESOURCE_DESC *resourceDesc, + D3D12_RESOURCE_STATES initialState, + const D3D12_CLEAR_VALUE *optimizedClearValue, + D3D12MA::Allocation **maybeAllocation, + REFIID riidResource, + void **ppvResource); + + void getBudget(D3D12MA::Budget *localBudget, D3D12MA::Budget *nonLocalBudget); + + bool isUsingD3D12MA() const { return allocator != nullptr; } + + ID3D12Device *device = nullptr; + D3D12MA::Allocator *allocator = nullptr; +}; + +struct QD3D12Buffer : public QRhiBuffer +{ + QD3D12Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QD3D12Buffer(); + + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + void executeHostWritesForFrameSlot(int frameSlot); + + QD3D12ObjectHandle handles[QD3D12_FRAMES_IN_FLIGHT] = {}; + struct HostWrite { + quint32 offset; + QRhiBufferData data; + }; + QVarLengthArray<HostWrite, 16> pendingHostWrites[QD3D12_FRAMES_IN_FLIGHT]; + friend class QRhiD3D12; + friend struct QD3D12CommandBuffer; +}; + +struct QD3D12RenderBuffer : public QRhiRenderBuffer +{ + QD3D12RenderBuffer(QRhiImplementation *rhi, + Type type, + const QSize &pixelSize, + int sampleCount, + Flags flags, + QRhiTexture::Format backingFormatHint); + ~QD3D12RenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + static const DXGI_FORMAT DS_FORMAT = DXGI_FORMAT_D24_UNORM_S8_UINT; + + QD3D12ObjectHandle handle; + QD3D12Descriptor rtv; + QD3D12Descriptor dsv; + DXGI_FORMAT dxgiFormat; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12Texture : public QRhiTexture +{ + QD3D12Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QD3D12Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + void setNativeLayout(int layout) override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + + QD3D12ObjectHandle handle; + QD3D12Descriptor srv; + DXGI_FORMAT dxgiFormat; + DXGI_FORMAT srvFormat; + DXGI_FORMAT rtFormat; // RTV/DSV/UAV + uint mipLevelCount; + DXGI_SAMPLE_DESC sampleDesc; + uint generation = 0; + friend class QRhiD3D12; + friend struct QD3D12CommandBuffer; +}; + +struct QD3D12Sampler : public QRhiSampler +{ + QD3D12Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QD3D12Sampler(); + void destroy() override; + bool create() override; + + QD3D12Descriptor lookupOrCreateShaderVisibleDescriptor(); + + D3D12_SAMPLER_DESC desc = {}; + QD3D12Descriptor shaderVisibleDescriptor; +}; + +struct QD3D12RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QD3D12RenderPassDescriptor(QRhiImplementation *rhi); + ~QD3D12RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + + void updateSerializedFormat(); + + static const int MAX_COLOR_ATTACHMENTS = 8; + int colorAttachmentCount = 0; + bool hasDepthStencil = false; + int colorFormat[MAX_COLOR_ATTACHMENTS]; + int dsFormat; + QVector<quint32> serializedFormatData; +}; + +struct QD3D12RenderTargetData +{ + QD3D12RenderTargetData(QRhiImplementation *) { } + + QD3D12RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + static const int MAX_COLOR_ATTACHMENTS = QD3D12RenderPassDescriptor::MAX_COLOR_ATTACHMENTS; + D3D12_CPU_DESCRIPTOR_HANDLE rtv[MAX_COLOR_ATTACHMENTS]; + D3D12_CPU_DESCRIPTOR_HANDLE dsv; +}; + +struct QD3D12SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QD3D12SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QD3D12SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QD3D12RenderTargetData d; +}; + +struct QD3D12TextureRenderTarget : public QRhiTextureRenderTarget +{ + QD3D12TextureRenderTarget(QRhiImplementation *rhi, + const QRhiTextureRenderTargetDescription &desc, + Flags flags); + ~QD3D12TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QD3D12RenderTargetData d; + bool ownsRtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; + QD3D12Descriptor rtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; + bool ownsDsv = false; + QD3D12Descriptor dsv; + friend class QRhiD3D12; +}; + +struct QD3D12ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QD3D12ShaderResourceBindings(QRhiImplementation *rhi); + ~QD3D12ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QD3D12ObjectHandle createRootSignature(const QD3D12ShaderStageData *stageData, int stageCount); + + struct VisitorData { + QVarLengthArray<D3D12_ROOT_PARAMETER1, 2> cbParams[6]; + + D3D12_ROOT_PARAMETER1 srvTables[6] = {}; + QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> srvRanges[6]; + quint32 currentSrvRangeOffset[6] = {}; + + QVarLengthArray<D3D12_ROOT_PARAMETER1, 4> samplerTables[6]; + std::array<D3D12_DESCRIPTOR_RANGE1, 16> samplerRanges[6] = {}; + int samplerRangeHeads[6] = {}; + + D3D12_ROOT_PARAMETER1 uavTables[6] = {}; + QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> uavRanges[6]; + quint32 currentUavRangeOffset[6] = {}; + } visitorData; + + + void visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int shaderRegister, + int binding); + void visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + void visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + + bool hasDynamicOffset = false; + uint generation = 0; + + friend class QRhiD3D12; + friend struct QD3D12ShaderResourceVisitor; +}; + +struct QD3D12GraphicsPipeline : public QRhiGraphicsPipeline +{ + QD3D12GraphicsPipeline(QRhiImplementation *rhi); + ~QD3D12GraphicsPipeline(); + void destroy() override; + bool create() override; + + QD3D12ObjectHandle handle; + QD3D12ObjectHandle rootSigHandle; + std::array<QD3D12ShaderStageData, 5> stageData; + D3D12_PRIMITIVE_TOPOLOGY topology; + UINT viewInstanceMask = 0; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12ComputePipeline : public QRhiComputePipeline +{ + QD3D12ComputePipeline(QRhiImplementation *rhi); + ~QD3D12ComputePipeline(); + void destroy() override; + bool create() override; + + QD3D12ObjectHandle handle; + QD3D12ObjectHandle rootSigHandle; + QD3D12ShaderStageData stageData; + uint generation = 0; + friend class QRhiD3D12; +}; + +struct QD3D12CommandBuffer : public QRhiCommandBuffer +{ + QD3D12CommandBuffer(QRhiImplementation *rhi); + ~QD3D12CommandBuffer(); + void destroy() override; + + const QRhiNativeHandles *nativeHandles(); + + ID3D12GraphicsCommandList1 *cmdList = nullptr; // not owned + QRhiD3D12CommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + void resetState() + { + recordingPass = NoPass; + currentTarget = nullptr; + + resetPerPassState(); + } + + void resetPerPassState() + { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentIndexBuffer = {}; + currentIndexOffset = 0; + currentIndexFormat = DXGI_FORMAT_R16_UINT; + currentVertexBuffers = {}; + currentVertexOffsets = {}; + } + + // per-frame + PassType recordingPass; + QRhiRenderTarget *currentTarget; + + // per-pass + QD3D12GraphicsPipeline *currentGraphicsPipeline; + QD3D12ComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + QD3D12ObjectHandle currentIndexBuffer; + quint32 currentIndexOffset; + DXGI_FORMAT currentIndexFormat; + std::array<QD3D12ObjectHandle, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexBuffers; + std::array<quint32, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexOffsets; + + // global + double lastGpuTime = 0; + + // per-setShaderResources + struct VisitorData { + QVarLengthArray<QPair<QD3D12ObjectHandle, quint32>, 4> cbufs[6]; + QVarLengthArray<QD3D12Descriptor, 8> srvs[6]; + QVarLengthArray<QD3D12Descriptor, 8> samplers[6]; + QVarLengthArray<QPair<QD3D12ObjectHandle, D3D12_UNORDERED_ACCESS_VIEW_DESC>, 4> uavs[6]; + } visitorData; + + void visitUniformBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::UniformBufferData &d, + int shaderRegister, + int binding, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets); + void visitTexture(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitSampler(QD3D12Stage s, + const QRhiShaderResourceBinding::TextureAndSampler &d, + int shaderRegister); + void visitStorageBuffer(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageBufferData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); + void visitStorageImage(QD3D12Stage s, + const QRhiShaderResourceBinding::Data::StorageImageData &d, + QD3D12ShaderResourceVisitor::StorageOp op, + int shaderRegister); +}; + +struct QD3D12SwapChain : public QRhiSwapChain +{ + QD3D12SwapChain(QRhiImplementation *rhi); + ~QD3D12SwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + QRhiSwapChainHdrInfo hdrInfo() override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void releaseBuffers(); + void waitCommandCompletionForFrameSlot(int frameSlot); + void addCommandCompletionSignalForCurrentFrameSlot(); + void chooseFormats(); + + QWindow *window = nullptr; + IDXGISwapChain1 *sourceSwapChain1 = nullptr; + IDXGISwapChain3 *swapChain = nullptr; + QSize pixelSize; + UINT swapInterval = 1; + UINT swapChainFlags = 0; + BOOL stereo = false; + DXGI_FORMAT colorFormat; + DXGI_FORMAT srgbAdjustedColorFormat; + DXGI_COLOR_SPACE_TYPE hdrColorSpace; + IDCompositionTarget *dcompTarget = nullptr; + IDCompositionVisual *dcompVisual = nullptr; + static const UINT BUFFER_COUNT = 3; + QD3D12ObjectHandle colorBuffers[BUFFER_COUNT]; + QD3D12Descriptor rtvs[BUFFER_COUNT]; + QD3D12Descriptor rtvsRight[BUFFER_COUNT]; + DXGI_SAMPLE_DESC sampleDesc; + QD3D12ObjectHandle msaaBuffers[BUFFER_COUNT]; + QD3D12Descriptor msaaRtvs[BUFFER_COUNT]; + QD3D12RenderBuffer *ds = nullptr; + UINT currentBackBufferIndex = 0; + QD3D12SwapChainRenderTarget rtWrapper; + QD3D12SwapChainRenderTarget rtWrapperRight; + QD3D12CommandBuffer cbWrapper; + + struct FrameResources { + ID3D12Fence *fence = nullptr; + HANDLE fenceEvent = nullptr; + UINT64 fenceCounter = 0; + ID3D12GraphicsCommandList1 *cmdList = nullptr; + } frameRes[QD3D12_FRAMES_IN_FLIGHT]; + + int currentFrameSlot = 0; // index in frameRes +}; + +template<typename T, D3D12_PIPELINE_STATE_SUBOBJECT_TYPE Type> +struct alignas(void*) QD3D12PipelineStateSubObject +{ + D3D12_PIPELINE_STATE_SUBOBJECT_TYPE type = Type; + T object = {}; +}; + +class QRhiD3D12 : public QRhiImplementation +{ +public: + // 16MB * QD3D12_FRAMES_IN_FLIGHT; buffer and texture upload staging data that + // gets no space from this will get their own temporary staging areas. + static const quint32 SMALL_STAGING_AREA_BYTES_PER_FRAME = 16 * 1024 * 1024; + + static const quint32 SHADER_VISIBLE_CBV_SRV_UAV_HEAP_PER_FRAME_START_SIZE = 16384; + + QRhiD3D12(QRhiD3D12InitParams *params, QRhiD3D12NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void waitGpu(); + DXGI_SAMPLE_DESC effectiveSampleDesc(int sampleCount, DXGI_FORMAT format) const; + bool ensureDirectCompositionDevice(); + bool startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList1 **cmdList); + void enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); + void finishActiveReadbacks(bool forced = false); + bool ensureShaderVisibleDescriptorHeapCapacity(QD3D12ShaderVisibleDescriptorHeap *h, + D3D12_DESCRIPTOR_HEAP_TYPE type, + int frameSlot, + quint32 neededDescriptorCount, + bool *gotNew); + void bindShaderVisibleHeaps(QD3D12CommandBuffer *cbD); + + bool debugLayer = false; + ID3D12Device2 *dev = nullptr; + D3D_FEATURE_LEVEL minimumFeatureLevel = D3D_FEATURE_LEVEL(0); + LUID adapterLuid = {}; + bool importedDevice = false; + bool importedCommandQueue = false; + QRhi::Flags rhiFlags; + IDXGIFactory2 *dxgiFactory = nullptr; + bool supportsAllowTearing = false; + IDXGIAdapter1 *activeAdapter = nullptr; + QRhiDriverInfo driverInfoStruct; + QRhiD3D12NativeHandles nativeHandlesStruct; + bool deviceLost = false; + ID3D12CommandQueue *cmdQueue = nullptr; + ID3D12Fence *fullFence = nullptr; + HANDLE fullFenceEvent = nullptr; + UINT64 fullFenceCounter = 0; + ID3D12CommandAllocator *cmdAllocators[QD3D12_FRAMES_IN_FLIGHT] = {}; + QD3D12MemoryAllocator vma; + QD3D12CpuDescriptorPool rtvPool; + QD3D12CpuDescriptorPool dsvPool; + QD3D12CpuDescriptorPool cbvSrvUavPool; + QD3D12ObjectPool<QD3D12Resource> resourcePool; + QD3D12ObjectPool<QD3D12Pipeline> pipelinePool; + QD3D12ObjectPool<QD3D12RootSignature> rootSignaturePool; + QD3D12ReleaseQueue releaseQueue; + QD3D12ResourceBarrierGenerator barrierGen; + QD3D12SamplerManager samplerMgr; + QD3D12MipmapGenerator mipmapGen; + QD3D12StagingArea smallStagingAreas[QD3D12_FRAMES_IN_FLIGHT]; + QD3D12ShaderVisibleDescriptorHeap shaderVisibleCbvSrvUavHeap; + UINT64 timestampTicksPerSecond = 0; + QD3D12QueryHeap timestampQueryHeap; + QD3D12StagingArea timestampReadbackArea; + IDCompositionDevice *dcompDevice = nullptr; + QD3D12SwapChain *currentSwapChain = nullptr; + QSet<QD3D12SwapChain *> swapchains; + QD3D12ShaderBytecodeCache shaderBytecodeCache; + QVarLengthArray<QD3D12Readback, 4> activeReadbacks; + bool offscreenActive = false; + QD3D12CommandBuffer *offscreenCb[QD3D12_FRAMES_IN_FLIGHT] = {}; + + struct { + bool multiView = false; + bool textureViewFormat = false; + } caps; }; QT_END_NAMESPACE +#endif // __ID3D12Device2_INTERFACE_DEFINED__ + #endif diff --git a/src/gui/rhi/qrhid3d12_p_p.h b/src/gui/rhi/qrhid3d12_p_p.h deleted file mode 100644 index dec217e771..0000000000 --- a/src/gui/rhi/qrhid3d12_p_p.h +++ /dev/null @@ -1,1194 +0,0 @@ -// Copyright (C) 2022 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHID3D12_P_H -#define QRHID3D12_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhid3d12_p.h" -#include "qrhi_p_p.h" -#include "qshaderdescription_p.h" -#include <QWindow> -#include <QBitArray> - -#include <optional> -#include <array> - -#include <d3d12.h> -#include <d3d12sdklayers.h> -#include <dxgi1_6.h> -#include <dcomp.h> - -#include "D3D12MemAlloc.h" - -QT_BEGIN_NAMESPACE - -static const int QD3D12_FRAMES_IN_FLIGHT = 2; - -class QRhiD3D12; - -struct QD3D12Descriptor -{ - D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = {}; - D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = {}; - - bool isValid() const { return cpuHandle.ptr != 0; } -}; - -struct QD3D12ReleaseQueue; - -struct QD3D12DescriptorHeap -{ - bool isValid() const { return heap && capacity; } - bool create(ID3D12Device *device, - quint32 descriptorCount, - D3D12_DESCRIPTOR_HEAP_TYPE heapType, - D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags); - void createWithExisting(const QD3D12DescriptorHeap &other, - quint32 offsetInDescriptors, - quint32 descriptorCount); - void destroy(); - void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); - - QD3D12Descriptor get(quint32 count); - QD3D12Descriptor at(quint32 index) const; - quint32 remainingCapacity() const { return capacity - head; } - - QD3D12Descriptor incremented(const QD3D12Descriptor &descriptor, quint32 offsetInDescriptors) const - { - D3D12_CPU_DESCRIPTOR_HANDLE cpuHandle = descriptor.cpuHandle; - cpuHandle.ptr += offsetInDescriptors * descriptorByteSize; - D3D12_GPU_DESCRIPTOR_HANDLE gpuHandle = descriptor.gpuHandle; - if (gpuHandle.ptr) - gpuHandle.ptr += offsetInDescriptors * descriptorByteSize; - return { cpuHandle, gpuHandle }; - } - - ID3D12DescriptorHeap *heap = nullptr; - quint32 capacity = 0; - QD3D12Descriptor heapStart; - quint32 head = 0; - quint32 descriptorByteSize = 0; - D3D12_DESCRIPTOR_HEAP_TYPE heapType; - D3D12_DESCRIPTOR_HEAP_FLAGS heapFlags; -}; - -struct QD3D12CpuDescriptorPool -{ - bool isValid() const { return !heaps.isEmpty(); } - bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE heapType, const char *debugName = ""); - void destroy(); - - QD3D12Descriptor allocate(quint32 count); - void release(const QD3D12Descriptor &descriptor, quint32 count); - - static const int DESCRIPTORS_PER_HEAP = 256; - - struct HeapWithMap { - QD3D12DescriptorHeap heap; - QBitArray map; - static HeapWithMap init(const QD3D12DescriptorHeap &heap, quint32 descriptorCount) { - HeapWithMap result; - result.heap = heap; - result.map.resize(descriptorCount); - return result; - } - }; - - ID3D12Device *device; - quint32 descriptorByteSize; - QVector<HeapWithMap> heaps; - const char *debugName; -}; - -struct QD3D12StagingArea -{ - static const quint32 ALIGNMENT = D3D12_TEXTURE_DATA_PLACEMENT_ALIGNMENT; // 512 so good enough both for cb and texdata - - struct Allocation { - quint8 *p = nullptr; - D3D12_GPU_VIRTUAL_ADDRESS gpuAddr = 0; - ID3D12Resource *buffer = nullptr; - quint32 bufferOffset = 0; - bool isValid() const { return p != nullptr; } - }; - - bool isValid() const { return allocation && mem.isValid(); } - bool create(QRhiD3D12 *rhi, quint32 capacity, D3D12_HEAP_TYPE heapType); - void destroy(); - void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); - - Allocation get(quint32 byteSize); - - quint32 remainingCapacity() const - { - return capacity - head; - } - - static quint32 allocSizeForArray(quint32 size, int count = 1) - { - return count * ((size + ALIGNMENT - 1) & ~(ALIGNMENT - 1)); - } - - Allocation mem; - ID3D12Resource *resource = nullptr; - D3D12MA::Allocation *allocation = nullptr; - quint32 head; - quint32 capacity; -}; - -struct QD3D12ObjectHandle -{ - quint32 index = 0; - quint32 generation = 0; - - // the default, null handle is guaranteed to give ObjectPool::isValid() == false - bool isNull() const { return index == 0 && generation == 0; } -}; - -inline bool operator==(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept -{ - return a.index == b.index && a.generation == b.generation; -} - -inline bool operator!=(const QD3D12ObjectHandle &a, const QD3D12ObjectHandle &b) noexcept -{ - return !(a == b); -} - -template<typename T> -struct QD3D12ObjectPool -{ - void create(const char *debugName = "") - { - this->debugName = debugName; - Q_ASSERT(data.isEmpty()); - data.append(Data()); // index 0 is always invalid - } - - void destroy() { - int leakCount = 0; // will nicely destroy everything here, but warn about it if enabled - for (Data &d : data) { - if (d.object.has_value()) { - leakCount += 1; - d.object->releaseResources(); - } - } - data.clear(); -#ifndef QT_NO_DEBUG - // debug builds: just do it always - static bool leakCheck = true; -#else - // release builds: opt-in - static bool leakCheck = qEnvironmentVariableIntValue("QT_RHI_LEAK_CHECK"); -#endif - if (leakCheck) { - if (leakCount > 0) { - qWarning("QD3D12ObjectPool::destroy(): Pool %p '%s' had %d unreleased objects", - this, debugName, leakCount); - } - } - } - - bool isValid(const QD3D12ObjectHandle &handle) const - { - return handle.index > 0 - && handle.index < quint32(data.count()) - && handle.generation > 0 - && handle.generation == data[handle.index].generation - && data[handle.index].object.has_value(); - } - - T lookup(const QD3D12ObjectHandle &handle) const - { - return isValid(handle) ? *data[handle.index].object : T(); - } - - const T *lookupRef(const QD3D12ObjectHandle &handle) const - { - return isValid(handle) ? &*data[handle.index].object : nullptr; - } - - T *lookupRef(const QD3D12ObjectHandle &handle) - { - return isValid(handle) ? &*data[handle.index].object : nullptr; - } - - QD3D12ObjectHandle add(const T &object) - { - Q_ASSERT(!data.isEmpty()); - const quint32 count = quint32(data.count()); - quint32 index = 1; // index 0 is always invalid - for (; index < count; ++index) { - if (!data[index].object.has_value()) - break; - } - if (index < count) { - data[index].object = object; - quint32 &generation = data[index].generation; - generation += 1u; - return { index, generation }; - } else { - data.append({ object, 1 }); - return { count, 1 }; - } - } - - void remove(const QD3D12ObjectHandle &handle) - { - if (T *object = lookupRef(handle)) { - object->releaseResources(); - data[handle.index].object.reset(); - } - } - - const char *debugName; - struct Data { - std::optional<T> object; - quint32 generation = 0; - }; - QVector<Data> data; -}; - -struct QD3D12Resource -{ - ID3D12Resource *resource; - D3D12_RESOURCE_STATES state; - D3D12_RESOURCE_DESC desc; - D3D12MA::Allocation *allocation; - void *cpuMapPtr; - enum { UavUsageRead = 0x01, UavUsageWrite = 0x02 }; - int uavUsage; - bool owns; - - // note that this assumes the allocation (if there is one) and the resource - // are separately releaseable, see D3D12MemAlloc docs - static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Resource> *pool, - ID3D12Resource *resource, - D3D12_RESOURCE_STATES state, - D3D12MA::Allocation *allocation = nullptr, - void *cpuMapPtr = nullptr) - { - Q_ASSERT(resource); - return pool->add({ resource, state, resource->GetDesc(), allocation, cpuMapPtr, 0, true }); - } - - // for QRhiTexture::createFrom() where the ID3D12Resource is not owned by us - static QD3D12ObjectHandle addNonOwningToPool(QD3D12ObjectPool<QD3D12Resource> *pool, - ID3D12Resource *resource, - D3D12_RESOURCE_STATES state) - { - Q_ASSERT(resource); - return pool->add({ resource, state, resource->GetDesc(), nullptr, nullptr, 0, false }); - } - - void releaseResources() - { - if (owns) { - // order matters: resource first, then the allocation - resource->Release(); - if (allocation) - allocation->Release(); - } - } -}; - -struct QD3D12Pipeline -{ - enum Type { - Graphics, - Compute - }; - Type type; - ID3D12PipelineState *pso; - - static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12Pipeline> *pool, - Type type, - ID3D12PipelineState *pso) - { - return pool->add({ type, pso }); - } - - void releaseResources() - { - pso->Release(); - } -}; - -struct QD3D12RootSignature -{ - ID3D12RootSignature *rootSig; - - static QD3D12ObjectHandle addToPool(QD3D12ObjectPool<QD3D12RootSignature> *pool, - ID3D12RootSignature *rootSig) - { - return pool->add({ rootSig }); - } - - void releaseResources() - { - rootSig->Release(); - } -}; - -struct QD3D12ReleaseQueue -{ - void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool, - QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool, - QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool) - { - this->resourcePool = resourcePool; - this->pipelinePool = pipelinePool; - this->rootSignaturePool = rootSignaturePool; - } - - void deferredReleaseResource(const QD3D12ObjectHandle &handle); - void deferredReleaseResourceWithViews(const QD3D12ObjectHandle &handle, - QD3D12CpuDescriptorPool *pool, - const QD3D12Descriptor &viewsStart, - int viewCount); - void deferredReleasePipeline(const QD3D12ObjectHandle &handle); - void deferredReleaseRootSignature(const QD3D12ObjectHandle &handle); - void deferredReleaseCallback(std::function<void(void*)> callback, void *userData); - void deferredReleaseResourceAndAllocation(ID3D12Resource *resource, - D3D12MA::Allocation *allocation); - void deferredReleaseDescriptorHeap(ID3D12DescriptorHeap *heap); - void deferredReleaseViews(QD3D12CpuDescriptorPool *pool, - const QD3D12Descriptor &viewsStart, - int viewCount); - - void activatePendingDeferredReleaseRequests(int frameSlot); - void executeDeferredReleases(int frameSlot, bool forced = false); - void releaseAll(); - - struct DeferredReleaseEntry { - enum Type { - Resource, - Pipeline, - RootSignature, - Callback, - ResourceAndAllocation, - DescriptorHeap, - Views - }; - Type type = Resource; - std::optional<int> frameSlotToBeReleasedIn; - QD3D12ObjectHandle handle; - QD3D12CpuDescriptorPool *poolForViews = nullptr; - QD3D12Descriptor viewsStart; - int viewCount = 0; - std::function<void(void*)> callback = nullptr; - void *callbackUserData = nullptr; - QPair<ID3D12Resource *, D3D12MA::Allocation *> resourceAndAllocation = {}; - ID3D12DescriptorHeap *descriptorHeap = nullptr; - }; - QVector<DeferredReleaseEntry> queue; - QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; - QD3D12ObjectPool<QD3D12Pipeline> *pipelinePool = nullptr; - QD3D12ObjectPool<QD3D12RootSignature> *rootSignaturePool = nullptr; -}; - -struct QD3D12CommandBuffer; - -struct QD3D12ResourceBarrierGenerator -{ - static const int PREALLOC = 16; - - void create(QD3D12ObjectPool<QD3D12Resource> *resourcePool) - { - this->resourcePool = resourcePool; - } - - void addTransitionBarrier(const QD3D12ObjectHandle &resourceHandle, D3D12_RESOURCE_STATES stateAfter); - void enqueueBufferedTransitionBarriers(QD3D12CommandBuffer *cbD); - void enqueueSubresourceTransitionBarrier(QD3D12CommandBuffer *cbD, - const QD3D12ObjectHandle &resourceHandle, - UINT subresource, - D3D12_RESOURCE_STATES stateBefore, - D3D12_RESOURCE_STATES stateAfter); - void enqueueUavBarrier(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &resourceHandle); - - struct TransitionResourceBarrier { - QD3D12ObjectHandle resourceHandle; - D3D12_RESOURCE_STATES stateBefore; - D3D12_RESOURCE_STATES stateAfter; - }; - QVarLengthArray<TransitionResourceBarrier, PREALLOC> transitionResourceBarriers; - QD3D12ObjectPool<QD3D12Resource> *resourcePool = nullptr; -}; - -struct QD3D12ShaderBytecodeCache -{ - struct Shader { - Shader() = default; - Shader(const QByteArray &bytecode, const QShader::NativeResourceBindingMap &rbm) - : bytecode(bytecode), nativeResourceBindingMap(rbm) - { } - QByteArray bytecode; - QShader::NativeResourceBindingMap nativeResourceBindingMap; - }; - - QHash<QRhiShaderStage, Shader> data; - - void insertWithCapacityLimit(const QRhiShaderStage &key, const Shader &s); -}; - -struct QD3D12ShaderVisibleDescriptorHeap -{ - bool create(ID3D12Device *device, D3D12_DESCRIPTOR_HEAP_TYPE type, quint32 perFrameDescriptorCount); - void destroy(); - void destroyWithDeferredRelease(QD3D12ReleaseQueue *releaseQueue); - - QD3D12DescriptorHeap heap; - QD3D12DescriptorHeap perFrameHeapSlice[QD3D12_FRAMES_IN_FLIGHT]; -}; - -// wrap foreign struct so we can legally supply equality operators and qHash: -struct Q_D3D12_SAMPLER_DESC -{ - D3D12_SAMPLER_DESC desc; - - friend bool operator==(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept - { - return lhs.desc.Filter == rhs.desc.Filter - && lhs.desc.AddressU == rhs.desc.AddressU - && lhs.desc.AddressV == rhs.desc.AddressV - && lhs.desc.AddressW == rhs.desc.AddressW - && lhs.desc.MipLODBias == rhs.desc.MipLODBias - && lhs.desc.MaxAnisotropy == rhs.desc.MaxAnisotropy - && lhs.desc.ComparisonFunc == rhs.desc.ComparisonFunc - // BorderColor is never used, skip it - && lhs.desc.MinLOD == rhs.desc.MinLOD - && lhs.desc.MaxLOD == rhs.desc.MaxLOD; - } - - friend bool operator!=(const Q_D3D12_SAMPLER_DESC &lhs, const Q_D3D12_SAMPLER_DESC &rhs) noexcept - { - return !(lhs == rhs); - } - - friend size_t qHash(const Q_D3D12_SAMPLER_DESC &key, size_t seed = 0) noexcept - { - QtPrivate::QHashCombine hash; - seed = hash(seed, key.desc.Filter); - seed = hash(seed, key.desc.AddressU); - seed = hash(seed, key.desc.AddressV); - seed = hash(seed, key.desc.AddressW); - seed = hash(seed, key.desc.MipLODBias); - seed = hash(seed, key.desc.MaxAnisotropy); - seed = hash(seed, key.desc.ComparisonFunc); - // BorderColor is never used, skip it - seed = hash(seed, key.desc.MinLOD); - seed = hash(seed, key.desc.MaxLOD); - return seed; - } -}; - -struct QD3D12SamplerManager -{ - const quint32 MAX_SAMPLERS = 512; - - bool create(ID3D12Device *device); - void destroy(); - - QD3D12Descriptor getShaderVisibleDescriptor(const D3D12_SAMPLER_DESC &desc); - - ID3D12Device *device = nullptr; - QD3D12ShaderVisibleDescriptorHeap shaderVisibleSamplerHeap; - QHash<Q_D3D12_SAMPLER_DESC, QD3D12Descriptor> gpuMap; -}; - -enum QD3D12Stage { VS = 0, HS, DS, GS, PS, CS }; - -static inline QD3D12Stage qd3d12_stage(QRhiShaderStage::Type type) -{ - switch (type) { - case QRhiShaderStage::Vertex: - return VS; - case QRhiShaderStage::TessellationControl: - return HS; - case QRhiShaderStage::TessellationEvaluation: - return DS; - case QRhiShaderStage::Geometry: - return GS; - case QRhiShaderStage::Fragment: - return PS; - case QRhiShaderStage::Compute: - return CS; - } - Q_UNREACHABLE_RETURN(VS); -} - -static inline D3D12_SHADER_VISIBILITY qd3d12_stageToVisibility(QD3D12Stage s) -{ - switch (s) { - case VS: - return D3D12_SHADER_VISIBILITY_VERTEX; - case HS: - return D3D12_SHADER_VISIBILITY_HULL; - case DS: - return D3D12_SHADER_VISIBILITY_DOMAIN; - case GS: - return D3D12_SHADER_VISIBILITY_GEOMETRY; - case PS: - return D3D12_SHADER_VISIBILITY_PIXEL; - case CS: - return D3D12_SHADER_VISIBILITY_ALL; - } - Q_UNREACHABLE_RETURN(D3D12_SHADER_VISIBILITY_ALL); -} - -static inline QRhiShaderResourceBinding::StageFlag qd3d12_stageToSrb(QD3D12Stage s) -{ - switch (s) { - case VS: - return QRhiShaderResourceBinding::VertexStage; - case HS: - return QRhiShaderResourceBinding::TessellationControlStage; - case DS: - return QRhiShaderResourceBinding::TessellationEvaluationStage; - case GS: - return QRhiShaderResourceBinding::GeometryStage; - case PS: - return QRhiShaderResourceBinding::FragmentStage; - case CS: - return QRhiShaderResourceBinding::ComputeStage; - } - Q_UNREACHABLE_RETURN(QRhiShaderResourceBinding::VertexStage); -} - -struct QD3D12ShaderStageData -{ - bool valid = false; // to allow simple arrays where unused stages are indicated by !valid - QD3D12Stage stage = VS; - QShader::NativeResourceBindingMap nativeResourceBindingMap; -}; - -struct QD3D12ShaderResourceBindings; - -struct QD3D12ShaderResourceVisitor -{ - enum StorageOp { Load = 0, Store, LoadStore }; - - QD3D12ShaderResourceVisitor(const QD3D12ShaderResourceBindings *srb, - const QD3D12ShaderStageData *stageData, - int stageCount) - : srb(srb), - stageData(stageData), - stageCount(stageCount) - { - } - - std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::UniformBufferData &, int, int)> uniformBuffer = nullptr; - std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> texture = nullptr; - std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::TextureAndSampler &, int)> sampler = nullptr; - std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageImageData &, StorageOp, int)> storageImage = nullptr; - std::function<void(QD3D12Stage, const QRhiShaderResourceBinding::Data::StorageBufferData &, StorageOp, int)> storageBuffer = nullptr; - - void visit(); - - const QD3D12ShaderResourceBindings *srb; - const QD3D12ShaderStageData *stageData; - int stageCount; -}; - -struct QD3D12Readback -{ - // common - int frameSlot = -1; - QRhiReadbackResult *result = nullptr; - QD3D12StagingArea staging; - quint32 byteSize = 0; - // textures - quint32 bytesPerLine = 0; - QSize pixelSize; - QRhiTexture::Format format = QRhiTexture::UnknownFormat; - quint32 stagingRowPitch = 0; -}; - -struct QD3D12MipmapGenerator -{ - bool create(QRhiD3D12 *rhiD); - void destroy(); - void generate(QD3D12CommandBuffer *cbD, const QD3D12ObjectHandle &textureHandle); - - QRhiD3D12 *rhiD; - QD3D12ObjectHandle rootSigHandle; - QD3D12ObjectHandle pipelineHandle; -}; - -struct QD3D12MemoryAllocator -{ - bool create(ID3D12Device *device, IDXGIAdapter1 *adapter); - void destroy(); - - HRESULT createResource(D3D12_HEAP_TYPE heapType, - const D3D12_RESOURCE_DESC *resourceDesc, - D3D12_RESOURCE_STATES initialState, - const D3D12_CLEAR_VALUE *optimizedClearValue, - D3D12MA::Allocation **maybeAllocation, - REFIID riidResource, - void **ppvResource); - - void getBudget(D3D12MA::Budget *localBudget, D3D12MA::Budget *nonLocalBudget); - - bool isUsingD3D12MA() const { return allocator != nullptr; } - - ID3D12Device *device = nullptr; - D3D12MA::Allocator *allocator = nullptr; -}; - -struct QD3D12Buffer : public QRhiBuffer -{ - QD3D12Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QD3D12Buffer(); - - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - void executeHostWritesForFrameSlot(int frameSlot); - - QD3D12ObjectHandle handles[QD3D12_FRAMES_IN_FLIGHT] = {}; - struct HostWrite { - quint32 offset; - QRhiBufferData data; - }; - QVarLengthArray<HostWrite, 16> pendingHostWrites[QD3D12_FRAMES_IN_FLIGHT]; - friend class QRhiD3D12; -}; - -struct QD3D12RenderBuffer : public QRhiRenderBuffer -{ - QD3D12RenderBuffer(QRhiImplementation *rhi, - Type type, - const QSize &pixelSize, - int sampleCount, - Flags flags, - QRhiTexture::Format backingFormatHint); - ~QD3D12RenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - static const DXGI_FORMAT DS_FORMAT = DXGI_FORMAT_D24_UNORM_S8_UINT; - - QD3D12ObjectHandle handle; - QD3D12Descriptor rtv; - QD3D12Descriptor dsv; - DXGI_FORMAT dxgiFormat; - DXGI_SAMPLE_DESC sampleDesc; - uint generation = 0; - friend class QRhiD3D12; -}; - -struct QD3D12Texture : public QRhiTexture -{ - QD3D12Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QD3D12Texture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - void setNativeLayout(int layout) override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - bool finishCreate(); - - QD3D12ObjectHandle handle; - QD3D12Descriptor srv; - DXGI_FORMAT dxgiFormat; - uint mipLevelCount; - DXGI_SAMPLE_DESC sampleDesc; - uint generation = 0; - friend class QRhiD3D12; -}; - -struct QD3D12Sampler : public QRhiSampler -{ - QD3D12Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QD3D12Sampler(); - void destroy() override; - bool create() override; - - QD3D12Descriptor lookupOrCreateShaderVisibleDescriptor(); - - D3D12_SAMPLER_DESC desc = {}; - QD3D12Descriptor shaderVisibleDescriptor; -}; - -struct QD3D12RenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QD3D12RenderPassDescriptor(QRhiImplementation *rhi); - ~QD3D12RenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; - - void updateSerializedFormat(); - - static const int MAX_COLOR_ATTACHMENTS = 8; - int colorAttachmentCount = 0; - bool hasDepthStencil = false; - int colorFormat[MAX_COLOR_ATTACHMENTS]; - int dsFormat; - QVector<quint32> serializedFormatData; -}; - -struct QD3D12RenderTargetData -{ - QD3D12RenderTargetData(QRhiImplementation *) { } - - QD3D12RenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; - static const int MAX_COLOR_ATTACHMENTS = QD3D12RenderPassDescriptor::MAX_COLOR_ATTACHMENTS; - D3D12_CPU_DESCRIPTOR_HANDLE rtv[MAX_COLOR_ATTACHMENTS]; - D3D12_CPU_DESCRIPTOR_HANDLE dsv; -}; - -struct QD3D12SwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QD3D12SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QD3D12SwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QD3D12RenderTargetData d; -}; - -struct QD3D12TextureRenderTarget : public QRhiTextureRenderTarget -{ - QD3D12TextureRenderTarget(QRhiImplementation *rhi, - const QRhiTextureRenderTargetDescription &desc, - Flags flags); - ~QD3D12TextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QD3D12RenderTargetData d; - bool ownsRtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; - QD3D12Descriptor rtv[QD3D12RenderTargetData::MAX_COLOR_ATTACHMENTS]; - bool ownsDsv = false; - QD3D12Descriptor dsv; - friend class QRhiD3D12; -}; - -struct QD3D12ShaderResourceBindings : public QRhiShaderResourceBindings -{ - QD3D12ShaderResourceBindings(QRhiImplementation *rhi); - ~QD3D12ShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; - - QD3D12ObjectHandle createRootSignature(const QD3D12ShaderStageData *stageData, int stageCount); - - struct VisitorData { - QVarLengthArray<D3D12_ROOT_PARAMETER1, 2> cbParams[6]; - - D3D12_ROOT_PARAMETER1 srvTables[6] = {}; - QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> srvRanges[6]; - quint32 currentSrvRangeOffset[6] = {}; - - QVarLengthArray<D3D12_ROOT_PARAMETER1, 4> samplerTables[6]; - std::array<D3D12_DESCRIPTOR_RANGE1, 16> samplerRanges[6] = {}; - int samplerRangeHeads[6] = {}; - - D3D12_ROOT_PARAMETER1 uavTables[6] = {}; - QVarLengthArray<D3D12_DESCRIPTOR_RANGE1, 4> uavRanges[6]; - quint32 currentUavRangeOffset[6] = {}; - } visitorData; - - - void visitUniformBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::UniformBufferData &d, - int shaderRegister, - int binding); - void visitTexture(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int shaderRegister); - void visitSampler(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int shaderRegister); - void visitStorageBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageBufferData &d, - QD3D12ShaderResourceVisitor::StorageOp op, - int shaderRegister); - void visitStorageImage(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageImageData &d, - QD3D12ShaderResourceVisitor::StorageOp op, - int shaderRegister); - - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - bool hasDynamicOffset = false; - uint generation = 0; -}; - -struct QD3D12GraphicsPipeline : public QRhiGraphicsPipeline -{ - QD3D12GraphicsPipeline(QRhiImplementation *rhi); - ~QD3D12GraphicsPipeline(); - void destroy() override; - bool create() override; - - QD3D12ObjectHandle handle; - QD3D12ObjectHandle rootSigHandle; - std::array<QD3D12ShaderStageData, 5> stageData; - D3D12_PRIMITIVE_TOPOLOGY topology; - uint generation = 0; - friend class QRhiD3D12; -}; - -struct QD3D12ComputePipeline : public QRhiComputePipeline -{ - QD3D12ComputePipeline(QRhiImplementation *rhi); - ~QD3D12ComputePipeline(); - void destroy() override; - bool create() override; - - QD3D12ObjectHandle handle; - QD3D12ObjectHandle rootSigHandle; - QD3D12ShaderStageData stageData; - uint generation = 0; - friend class QRhiD3D12; -}; - -struct QD3D12CommandBuffer : public QRhiCommandBuffer -{ - QD3D12CommandBuffer(QRhiImplementation *rhi); - ~QD3D12CommandBuffer(); - void destroy() override; - - const QRhiNativeHandles *nativeHandles(); - - ID3D12GraphicsCommandList *cmdList = nullptr; // not owned - QRhiD3D12CommandBufferNativeHandles nativeHandlesStruct; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - void resetState() - { - recordingPass = NoPass; - currentTarget = nullptr; - - resetPerPassState(); - } - - void resetPerPassState() - { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - currentIndexBuffer = {}; - currentIndexOffset = 0; - currentIndexFormat = DXGI_FORMAT_R16_UINT; - currentVertexBuffers = {}; - currentVertexOffsets = {}; - } - - PassType recordingPass; - QRhiRenderTarget *currentTarget; - - QD3D12GraphicsPipeline *currentGraphicsPipeline; - QD3D12ComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - QD3D12ObjectHandle currentIndexBuffer; - quint32 currentIndexOffset; - DXGI_FORMAT currentIndexFormat; - std::array<QD3D12ObjectHandle, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexBuffers; - std::array<quint32, D3D12_IA_VERTEX_INPUT_RESOURCE_SLOT_COUNT> currentVertexOffsets; -}; - -struct QD3D12SwapChain : public QRhiSwapChain -{ - QD3D12SwapChain(QRhiImplementation *rhi); - ~QD3D12SwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - QRhiSwapChainHdrInfo hdrInfo() override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - void releaseBuffers(); - void waitCommandCompletionForFrameSlot(int frameSlot); - void addCommandCompletionSignalForCurrentFrameSlot(); - void chooseFormats(); - - QWindow *window = nullptr; - IDXGISwapChain1 *sourceSwapChain1 = nullptr; - IDXGISwapChain3 *swapChain = nullptr; - QSize pixelSize; - UINT swapInterval = 1; - UINT swapChainFlags = 0; - DXGI_FORMAT colorFormat; - DXGI_FORMAT srgbAdjustedColorFormat; - DXGI_COLOR_SPACE_TYPE hdrColorSpace; - IDCompositionTarget *dcompTarget = nullptr; - IDCompositionVisual *dcompVisual = nullptr; - static const UINT BUFFER_COUNT = 3; - QD3D12ObjectHandle colorBuffers[BUFFER_COUNT]; - QD3D12Descriptor rtvs[BUFFER_COUNT]; - DXGI_SAMPLE_DESC sampleDesc; - QD3D12ObjectHandle msaaBuffers[BUFFER_COUNT]; - QD3D12Descriptor msaaRtvs[BUFFER_COUNT]; - QD3D12RenderBuffer *ds = nullptr; - UINT currentBackBufferIndex = 0; - QD3D12SwapChainRenderTarget rtWrapper; - QD3D12CommandBuffer cbWrapper; - - struct FrameResources { - ID3D12Fence *fence = nullptr; - HANDLE fenceEvent = nullptr; - UINT64 fenceCounter = 0; - ID3D12GraphicsCommandList *cmdList = nullptr; - } frameRes[QD3D12_FRAMES_IN_FLIGHT]; - - int currentFrameSlot = 0; // index in frameRes -}; - -class QRhiD3D12 : public QRhiImplementation -{ -public: - // 16MB * QD3D12_FRAMES_IN_FLIGHT; buffer and texture upload staging data that - // gets no space from this will get their own temporary staging areas. - static const quint32 SMALL_STAGING_AREA_BYTES_PER_FRAME = 16 * 1024 * 1024; - - static const quint32 SHADER_VISIBLE_CBV_SRV_UAV_HEAP_PER_FRAME_START_SIZE = 16384; - - QRhiD3D12(QRhiD3D12InitParams *params, QRhiD3D12NativeHandles *importDevice = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void waitGpu(); - DXGI_SAMPLE_DESC effectiveSampleCount(int sampleCount, DXGI_FORMAT format) const; - bool ensureDirectCompositionDevice(); - bool startCommandListForCurrentFrameSlot(ID3D12GraphicsCommandList **cmdList); - void enqueueResourceUpdates(QD3D12CommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); - void finishActiveReadbacks(bool forced = false); - bool ensureShaderVisibleDescriptorHeapCapacity(QD3D12ShaderVisibleDescriptorHeap *h, - D3D12_DESCRIPTOR_HEAP_TYPE type, - int frameSlot, - quint32 neededDescriptorCount, - bool *gotNew); - void bindShaderVisibleHeaps(QD3D12CommandBuffer *cbD); - - bool debugLayer = false; - ID3D12Device *dev = nullptr; - D3D_FEATURE_LEVEL minimumFeatureLevel = D3D_FEATURE_LEVEL(0); - LUID adapterLuid = {}; - bool importedDevice = false; - bool importedCommandQueue = false; - QRhi::Flags rhiFlags; - IDXGIFactory2 *dxgiFactory = nullptr; - bool supportsAllowTearing = false; - IDXGIAdapter1 *activeAdapter = nullptr; - QRhiDriverInfo driverInfoStruct; - QRhiD3D12NativeHandles nativeHandlesStruct; - bool deviceLost = false; - ID3D12CommandQueue *cmdQueue = nullptr; - ID3D12Fence *fullFence = nullptr; - HANDLE fullFenceEvent = nullptr; - UINT64 fullFenceCounter = 0; - ID3D12CommandAllocator *cmdAllocators[QD3D12_FRAMES_IN_FLIGHT] = {}; - QD3D12MemoryAllocator vma; - QD3D12CpuDescriptorPool rtvPool; - QD3D12CpuDescriptorPool dsvPool; - QD3D12CpuDescriptorPool cbvSrvUavPool; - QD3D12ObjectPool<QD3D12Resource> resourcePool; - QD3D12ObjectPool<QD3D12Pipeline> pipelinePool; - QD3D12ObjectPool<QD3D12RootSignature> rootSignaturePool; - QD3D12ReleaseQueue releaseQueue; - QD3D12ResourceBarrierGenerator barrierGen; - QD3D12SamplerManager samplerMgr; - QD3D12MipmapGenerator mipmapGen; - QD3D12StagingArea smallStagingAreas[QD3D12_FRAMES_IN_FLIGHT]; - QD3D12ShaderVisibleDescriptorHeap shaderVisibleCbvSrvUavHeap; - IDCompositionDevice *dcompDevice = nullptr; - QD3D12SwapChain *currentSwapChain = nullptr; - QSet<QD3D12SwapChain *> swapchains; - QD3D12ShaderBytecodeCache shaderBytecodeCache; - QVarLengthArray<QD3D12Readback, 4> activeReadbacks; - bool offscreenActive = false; - QD3D12CommandBuffer *offscreenCb[QD3D12_FRAMES_IN_FLIGHT] = {}; - - struct VisitorData { - QVarLengthArray<QPair<QD3D12ObjectHandle, quint32>, 4> cbufs[6]; - QVarLengthArray<QD3D12Descriptor, 8> srvs[6]; - QVarLengthArray<QD3D12Descriptor, 8> samplers[6]; - QVarLengthArray<QPair<QD3D12ObjectHandle, D3D12_UNORDERED_ACCESS_VIEW_DESC>, 4> uavs[6]; - } visitorData; - - void visitUniformBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::UniformBufferData &d, - int shaderRegister, - int binding, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets); - void visitTexture(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int shaderRegister); - void visitSampler(QD3D12Stage s, - const QRhiShaderResourceBinding::TextureAndSampler &d, - int shaderRegister); - void visitStorageBuffer(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageBufferData &d, - QD3D12ShaderResourceVisitor::StorageOp op, - int shaderRegister); - void visitStorageImage(QD3D12Stage s, - const QRhiShaderResourceBinding::Data::StorageImageData &d, - QD3D12ShaderResourceVisitor::StorageOp op, - int shaderRegister); -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhid3dhelpers.cpp b/src/gui/rhi/qrhid3dhelpers.cpp new file mode 100644 index 0000000000..216c358cbe --- /dev/null +++ b/src/gui/rhi/qrhid3dhelpers.cpp @@ -0,0 +1,172 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#include "qrhid3dhelpers_p.h" +#include <QtCore/private/qsystemlibrary_p.h> +#include <QtCore/private/qsystemerror_p.h> + +QT_BEGIN_NAMESPACE + +namespace QRhiD3D { + +bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result) +{ + bool ok = false; + QRect wr = w->geometry(); + wr = QRect(wr.topLeft() * w->devicePixelRatio(), wr.size() * w->devicePixelRatio()); + const QPoint center = wr.center(); + IDXGIOutput *currentOutput = nullptr; + IDXGIOutput *output = nullptr; + for (UINT i = 0; adapter->EnumOutputs(i, &output) != DXGI_ERROR_NOT_FOUND; ++i) { + DXGI_OUTPUT_DESC desc; + output->GetDesc(&desc); + const RECT r = desc.DesktopCoordinates; + const QRect dr(QPoint(r.left, r.top), QPoint(r.right - 1, r.bottom - 1)); + if (dr.contains(center)) { + currentOutput = output; + break; + } else { + output->Release(); + } + } + if (currentOutput) { + ok = SUCCEEDED(currentOutput->QueryInterface(__uuidof(IDXGIOutput6), reinterpret_cast<void **>(result))); + currentOutput->Release(); + } + return ok; +} + +bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result) +{ + bool ok = false; + IDXGIOutput6 *out6 = nullptr; + if (output6ForWindow(w, adapter, &out6)) { + ok = SUCCEEDED(out6->GetDesc1(result)); + out6->Release(); + } + return ok; +} + +float sdrWhiteLevelInNits(const DXGI_OUTPUT_DESC1 &outputDesc) +{ + QVector<DISPLAYCONFIG_PATH_INFO> pathInfos; + uint32_t pathInfoCount, modeInfoCount; + LONG result; + do { + if (GetDisplayConfigBufferSizes(QDC_ONLY_ACTIVE_PATHS, &pathInfoCount, &modeInfoCount) == ERROR_SUCCESS) { + pathInfos.resize(pathInfoCount); + QVector<DISPLAYCONFIG_MODE_INFO> modeInfos(modeInfoCount); + result = QueryDisplayConfig(QDC_ONLY_ACTIVE_PATHS, &pathInfoCount, pathInfos.data(), &modeInfoCount, modeInfos.data(), nullptr); + } else { + return 200.0f; + } + } while (result == ERROR_INSUFFICIENT_BUFFER); + + MONITORINFOEX monitorInfo = {}; + monitorInfo.cbSize = sizeof(monitorInfo); + GetMonitorInfo(outputDesc.Monitor, &monitorInfo); + + for (const DISPLAYCONFIG_PATH_INFO &info : pathInfos) { + DISPLAYCONFIG_SOURCE_DEVICE_NAME deviceName = {}; + deviceName.header.type = DISPLAYCONFIG_DEVICE_INFO_GET_SOURCE_NAME; + deviceName.header.size = sizeof(deviceName); + deviceName.header.adapterId = info.sourceInfo.adapterId; + deviceName.header.id = info.sourceInfo.id; + if (DisplayConfigGetDeviceInfo(&deviceName.header) == ERROR_SUCCESS) { + if (!wcscmp(monitorInfo.szDevice, deviceName.viewGdiDeviceName)) { + DISPLAYCONFIG_SDR_WHITE_LEVEL whiteLevel = {}; + whiteLevel.header.type = DISPLAYCONFIG_DEVICE_INFO_GET_SDR_WHITE_LEVEL; + whiteLevel.header.size = sizeof(DISPLAYCONFIG_SDR_WHITE_LEVEL); + whiteLevel.header.adapterId = info.targetInfo.adapterId; + whiteLevel.header.id = info.targetInfo.id; + if (DisplayConfigGetDeviceInfo(&whiteLevel.header) == ERROR_SUCCESS) + return whiteLevel.SDRWhiteLevel * 80 / 1000.0f; + } + } + } + + return 200.0f; +} + +pD3DCompile resolveD3DCompile() +{ + for (const wchar_t *libraryName : {L"D3DCompiler_47", L"D3DCompiler_43"}) { + QSystemLibrary library(libraryName); + if (library.load()) { + if (auto symbol = library.resolve("D3DCompile")) + return reinterpret_cast<pD3DCompile>(symbol); + } else { + qWarning("Failed to load D3DCompiler_47/43.dll"); + } + } + return nullptr; +} + +IDCompositionDevice *createDirectCompositionDevice() +{ + QSystemLibrary dcomplib(QStringLiteral("dcomp")); + typedef HRESULT (__stdcall *DCompositionCreateDeviceFuncPtr)( + _In_opt_ IDXGIDevice *dxgiDevice, + _In_ REFIID iid, + _Outptr_ void **dcompositionDevice); + DCompositionCreateDeviceFuncPtr func = reinterpret_cast<DCompositionCreateDeviceFuncPtr>( + dcomplib.resolve("DCompositionCreateDevice")); + if (!func) { + qWarning("Unable to resolve DCompositionCreateDevice, perhaps dcomp.dll is missing?"); + return nullptr; + } + IDCompositionDevice *device = nullptr; + HRESULT hr = func(nullptr, __uuidof(IDCompositionDevice), reinterpret_cast<void **>(&device)); + if (FAILED(hr)) { + qWarning("Failed to create Direct Composition device: %s", + qPrintable(QSystemError::windowsComString(hr))); + return nullptr; + } + return device; +} + +#ifdef QRHI_D3D12_HAS_DXC +std::pair<IDxcCompiler *, IDxcLibrary *> createDxcCompiler() +{ + QSystemLibrary dxclib(QStringLiteral("dxcompiler")); + // this will not be in the system library location, hence onlySystemDirectory==false + if (!dxclib.load(false)) { + qWarning("Failed to load dxcompiler.dll"); + return {}; + } + DxcCreateInstanceProc func = reinterpret_cast<DxcCreateInstanceProc>(dxclib.resolve("DxcCreateInstance")); + if (!func) { + qWarning("Unable to resolve DxcCreateInstance"); + return {}; + } + IDxcCompiler *compiler = nullptr; + HRESULT hr = func(CLSID_DxcCompiler, __uuidof(IDxcCompiler), reinterpret_cast<void**>(&compiler)); + if (FAILED(hr)) { + qWarning("Failed to create dxc compiler instance: %s", + qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + IDxcLibrary *library = nullptr; + hr = func(CLSID_DxcLibrary, __uuidof(IDxcLibrary), reinterpret_cast<void**>(&library)); + if (FAILED(hr)) { + qWarning("Failed to create dxc library instance: %s", + qPrintable(QSystemError::windowsComString(hr))); + return {}; + } + return { compiler, library }; +} +#endif + +void fillDriverInfo(QRhiDriverInfo *info, const DXGI_ADAPTER_DESC1 &desc) +{ + const QString name = QString::fromUtf16(reinterpret_cast<const char16_t *>(desc.Description)); + info->deviceName = name.toUtf8(); + info->deviceId = desc.DeviceId; + info->vendorId = desc.VendorId; + info->deviceType = (desc.Flags & DXGI_ADAPTER_FLAG_SOFTWARE) ? QRhiDriverInfo::CpuDevice + : QRhiDriverInfo::UnknownDevice; +} + +} // namespace + +QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhid3dhelpers_p.h b/src/gui/rhi/qrhid3dhelpers_p.h new file mode 100644 index 0000000000..f31cdc8d11 --- /dev/null +++ b/src/gui/rhi/qrhid3dhelpers_p.h @@ -0,0 +1,53 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QRHID3DHELPERS_P_H +#define QRHID3DHELPERS_P_H + +// +// W A R N I N G +// ------------- +// +// This file is not part of the Qt API. It exists purely as an +// implementation detail. This header file may change from version to +// version without notice, or even be removed. +// +// We mean it. +// + +#include <rhi/qrhi.h> + +#include <QtGui/qwindow.h> + +#include <dxgi1_6.h> +#include <dcomp.h> +#include <d3dcompiler.h> + +#if __has_include(<dxcapi.h>) +#include <dxcapi.h> +#define QRHI_D3D12_HAS_DXC +#endif + +QT_BEGIN_NAMESPACE + +namespace QRhiD3D { + +bool output6ForWindow(QWindow *w, IDXGIAdapter1 *adapter, IDXGIOutput6 **result); +bool outputDesc1ForWindow(QWindow *w, IDXGIAdapter1 *adapter, DXGI_OUTPUT_DESC1 *result); +float sdrWhiteLevelInNits(const DXGI_OUTPUT_DESC1 &outputDesc); + +pD3DCompile resolveD3DCompile(); + +IDCompositionDevice *createDirectCompositionDevice(); + +#ifdef QRHI_D3D12_HAS_DXC +std::pair<IDxcCompiler *, IDxcLibrary *> createDxcCompiler(); +#endif + +void fillDriverInfo(QRhiDriverInfo *info, const DXGI_ADAPTER_DESC1 &desc); + +} // namespace + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qrhigles2.cpp b/src/gui/rhi/qrhigles2.cpp index a9bb8473a2..7e886a5d00 100644 --- a/src/gui/rhi/qrhigles2.cpp +++ b/src/gui/rhi/qrhigles2.cpp @@ -1,12 +1,13 @@ -// Copyright (C) 2021 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhigles2_p_p.h" +#include "qrhigles2_p.h" #include <QOffscreenSurface> #include <QOpenGLContext> #include <QtCore/qmap.h> #include <QtGui/private/qopenglextensions_p.h> #include <QtGui/private/qopenglprogrambinarycache_p.h> +#include <QtGui/private/qwindow_p.h> #include <qpa/qplatformopenglcontext.h> #include <qmath.h> @@ -27,10 +28,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiGles2InitParams - \internal \inmodule QtGui + \since 6.6 \brief OpenGL specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + An OpenGL-based QRhi needs an already created QSurface that can be used in combination with QOpenGLContext. Most commonly, this is a QOffscreenSurface in practice. Additionally, while optional, it is recommended that the QWindow @@ -104,12 +108,50 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiGles2InitParams::format + + The QSurfaceFormat, initialized to QSurfaceFormat::defaultFormat() by default. +*/ + +/*! + \variable QRhiGles2InitParams::fallbackSurface + + A QSurface compatible with \l format. Typically a QOffscreenSurface. + Providing this is mandatory. Be aware of the threading implications: a + QOffscreenSurface, like QWindow, must only ever be created and destroyed on + the main (gui) thread, even if the QRhi is created and operates on another + thread. +*/ + +/*! + \variable QRhiGles2InitParams::window + + Optional, but setting it is recommended when targeting a QWindow with the + QRhi. +*/ + +/*! + \variable QRhiGles2InitParams::shareContext + + Optional, the QOpenGLContext to share resource with. QRhi creates its own + context, and setting this member to a valid QOpenGLContext leads to calling + \l{QOpenGLContext::setShareContext()}{setShareContext()} with it. +*/ + +/*! \class QRhiGles2NativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the OpenGL context used by the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiGles2NativeHandles::context +*/ + #ifndef GL_BGRA #define GL_BGRA 0x80E1 #endif @@ -174,6 +216,10 @@ QT_BEGIN_NAMESPACE #define GL_DEPTH_COMPONENT32F 0x8CAC #endif +#ifndef GL_UNSIGNED_INT_24_8 +#define GL_UNSIGNED_INT_24_8 0x84FA +#endif + #ifndef GL_STENCIL_INDEX #define GL_STENCIL_INDEX 0x1901 #endif @@ -199,7 +245,7 @@ QT_BEGIN_NAMESPACE #endif #ifndef GL_FRAMEBUFFER_SRGB -#define GL_FRAMEBUFFER_SRGB 0x8DB9 +#define GL_FRAMEBUFFER_SRGB 0x8DB9 #endif #ifndef GL_READ_FRAMEBUFFER @@ -314,6 +360,10 @@ QT_BEGIN_NAMESPACE #define GL_TEXTURE_2D_MULTISAMPLE 0x9100 #endif +#ifndef GL_TEXTURE_2D_MULTISAMPLE_ARRAY +#define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102 +#endif + #ifndef GL_TEXTURE_EXTERNAL_OES #define GL_TEXTURE_EXTERNAL_OES 0x8D65 #endif @@ -443,15 +493,39 @@ QT_BEGIN_NAMESPACE #endif #ifndef GL_TEXTURE_1D -# define GL_TEXTURE_1D 0x0DE0 +# define GL_TEXTURE_1D 0x0DE0 #endif #ifndef GL_TEXTURE_1D_ARRAY -# define GL_TEXTURE_1D_ARRAY 0x8C18 +# define GL_TEXTURE_1D_ARRAY 0x8C18 #endif #ifndef GL_HALF_FLOAT -#define GL_HALF_FLOAT 0x140B +#define GL_HALF_FLOAT 0x140B +#endif + +#ifndef GL_MAX_VERTEX_OUTPUT_COMPONENTS +#define GL_MAX_VERTEX_OUTPUT_COMPONENTS 0x9122 +#endif + +#ifndef GL_TIMESTAMP +#define GL_TIMESTAMP 0x8E28 +#endif + +#ifndef GL_QUERY_RESULT +#define GL_QUERY_RESULT 0x8866 +#endif + +#ifndef GL_QUERY_RESULT_AVAILABLE +#define GL_QUERY_RESULT_AVAILABLE 0x8867 +#endif + +#ifndef GL_BUFFER +#define GL_BUFFER 0x82E0 +#endif + +#ifndef GL_PROGRAM +#define GL_PROGRAM 0x82E2 #endif /*! @@ -658,6 +732,8 @@ bool QRhiGles2::create(QRhi::Flags flags) return false; f = static_cast<QOpenGLExtensions *>(ctx->extraFunctions()); + const QSurfaceFormat actualFormat = ctx->format(); + caps.gles = actualFormat.renderableType() == QSurfaceFormat::OpenGLES; if (!caps.gles) { glPolygonMode = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum)>( @@ -708,8 +784,6 @@ bool QRhiGles2::create(QRhi::Flags flags) if (version) driverInfoStruct.deviceName += QByteArray(version); - const QSurfaceFormat actualFormat = ctx->format(); - caps.ctxMajor = actualFormat.majorVersion(); caps.ctxMinor = actualFormat.minorVersion(); @@ -769,8 +843,6 @@ bool QRhiGles2::create(QRhi::Flags flags) f->glGetIntegerv(GL_MAX_TEXTURE_SIZE, &caps.maxTextureSize); - caps.gles = actualFormat.renderableType() == QSurfaceFormat::OpenGLES; - if (!caps.gles || caps.ctxMajor >= 3) { // non-ES or ES 3.0+ f->glGetIntegerv(GL_MAX_DRAW_BUFFERS, &caps.maxDrawBuffers); @@ -782,8 +854,8 @@ bool QRhiGles2::create(QRhi::Flags flags) caps.maxDrawBuffers = 1; caps.hasDrawBuffersFunc = false; // This does not mean MSAA is not supported, just that we cannot query - // the supported sample counts. - caps.maxSamples = 1; + // the supported sample counts. Assume that 4x is always supported. + caps.maxSamples = 4; } caps.msaaRenderBuffer = f->hasOpenGLExtension(QOpenGLExtensions::FramebufferMultisample) @@ -820,7 +892,13 @@ bool QRhiGles2::create(QRhi::Flags flags) #else caps.needsDepthStencilCombinedAttach = false; #endif - caps.srgbCapableDefaultFramebuffer = f->hasOpenGLExtension(QOpenGLExtensions::SRGBFrameBuffer); + + // QOpenGLExtensions::SRGBFrameBuffer is not useful here. We need to know if + // controlling the sRGB-on-shader-write state is supported, not that if the + // default framebuffer is sRGB-capable. And there are two different + // extensions for desktop and ES. + caps.srgbWriteControl = ctx->hasExtension("GL_EXT_framebuffer_sRGB") || ctx->hasExtension("GL_EXT_sRGB_write_control"); + caps.coreProfile = actualFormat.profile() == QSurfaceFormat::CoreProfile; if (caps.gles) @@ -941,7 +1019,7 @@ bool QRhiGles2::create(QRhi::Flags flags) f->glGetIntegerv(GL_MAX_VARYING_VECTORS, &caps.maxVertexOutputs); } else if (caps.ctxMajor >= 3) { GLint components = 0; - f->glGetIntegerv(GL_MAX_VARYING_COMPONENTS, &components); + f->glGetIntegerv(caps.coreProfile ? GL_MAX_VERTEX_OUTPUT_COMPONENTS : GL_MAX_VARYING_COMPONENTS, &components); caps.maxVertexOutputs = components / 4; } else { // OpenGL before 3.0 only has this, and not the same as @@ -954,7 +1032,8 @@ bool QRhiGles2::create(QRhi::Flags flags) if (!caps.gles) { f->glEnable(GL_VERTEX_PROGRAM_POINT_SIZE); - f->glEnable(GL_POINT_SPRITE); + if (!caps.coreProfile) + f->glEnable(GL_POINT_SPRITE); } // else (with gles) these are always on // Match D3D and others when it comes to seamless cubemap filtering. @@ -965,6 +1044,60 @@ bool QRhiGles2::create(QRhi::Flags flags) caps.halfAttributes = f->hasOpenGLExtension(QOpenGLExtensions::HalfFloatVertex); + // We always require GL_OVR_multiview2 for symmetry with other backends. + caps.multiView = f->hasOpenGLExtension(QOpenGLExtensions::MultiView) + && f->hasOpenGLExtension(QOpenGLExtensions::MultiViewExtended); + if (caps.multiView) { + glFramebufferTextureMultiviewOVR = + reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLuint, GLint, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTextureMultiviewOVR"))); + } + + // Only do timestamp queries on OpenGL 3.3+. + caps.timestamps = !caps.gles && (caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 3)); + if (caps.timestamps) { + glQueryCounter = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLuint, GLenum)>( + ctx->getProcAddress(QByteArrayLiteral("glQueryCounter"))); + glGetQueryObjectui64v = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLuint, GLenum, quint64 *)>( + ctx->getProcAddress(QByteArrayLiteral("glGetQueryObjectui64v"))); + if (!glQueryCounter || !glGetQueryObjectui64v) + caps.timestamps = false; + } + + // glObjectLabel is available on OpenGL ES 3.2+ and OpenGL 4.3+ + if (caps.gles) + caps.objectLabel = caps.ctxMajor > 3 || (caps.ctxMajor == 3 && caps.ctxMinor >= 2); + else + caps.objectLabel = caps.ctxMajor > 4 || (caps.ctxMajor == 4 && caps.ctxMinor >= 3); + if (caps.objectLabel) { + glObjectLabel = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLuint, GLsizei, const GLchar *)>( + ctx->getProcAddress(QByteArrayLiteral("glObjectLabel"))); + } + + if (caps.gles) { + // This is the third way to get multisample rendering with GLES. (1. is + // multisample render buffer -> resolve to texture; 2. is multisample + // texture with GLES 3.1; 3. is this, avoiding the explicit multisample + // buffer and should be more efficient with tiled architectures. + // Interesting also because 2. does not seem to work in practice on + // devices such as the Quest 3) + caps.glesMultisampleRenderToTexture = ctx->hasExtension("GL_EXT_multisampled_render_to_texture"); + if (caps.glesMultisampleRenderToTexture) { + glFramebufferTexture2DMultisampleEXT = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLenum, GLuint, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTexture2DMultisampleEXT"))); + } + caps.glesMultiviewMultisampleRenderToTexture = ctx->hasExtension("GL_OVR_multiview_multisampled_render_to_texture"); + if (caps.glesMultiviewMultisampleRenderToTexture) { + glFramebufferTextureMultisampleMultiviewOVR = reinterpret_cast<void(QOPENGLF_APIENTRYP)(GLenum, GLenum, GLuint, GLint, GLsizei, GLint, GLsizei)>( + ctx->getProcAddress(QByteArrayLiteral("glFramebufferTextureMultisampleMultiviewOVR"))); + } + } else { + caps.glesMultisampleRenderToTexture = false; + caps.glesMultiviewMultisampleRenderToTexture = false; + } + + caps.unpackRowLength = !caps.gles || caps.ctxMajor >= 3; + nativeHandlesStruct.context = ctx; contextLost = false; @@ -980,6 +1113,11 @@ void QRhiGles2::destroy() ensureContext(); executeDeferredReleases(); + if (ofr.tsQueries[0]) { + f->glDeleteQueries(2, ofr.tsQueries); + ofr.tsQueries[0] = ofr.tsQueries[1] = 0; + } + if (vao) { f->glDeleteVertexArrays(1, &vao); vao = 0; @@ -1017,6 +1155,7 @@ void QRhiGles2::executeDeferredReleases() break; case QRhiGles2::DeferredReleaseEntry::TextureRenderTarget: f->glDeleteFramebuffers(1, &e.textureRenderTarget.framebuffer); + f->glDeleteTextures(1, &e.textureRenderTarget.nonMsaaThrowawayDepthTexture); break; default: Q_UNREACHABLE(); @@ -1036,17 +1175,6 @@ QList<int> QRhiGles2::supportedSampleCounts() const return supportedSampleCountList; } -int QRhiGles2::effectiveSampleCount(int sampleCount) const -{ - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - const int s = qBound(1, sampleCount, 64); - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return 1; - } - return s; -} - QRhiSwapChain *QRhiGles2::createSwapChain() { return new QGles2SwapChain(this); @@ -1171,13 +1299,13 @@ static inline void toGlTextureFormat(QRhiTexture::Format format, const QRhiGles2 *glintformat = GL_DEPTH_COMPONENT24; *glsizedintformat = *glintformat; *glformat = GL_DEPTH_COMPONENT; - *gltype = GL_UNSIGNED_SHORT; + *gltype = GL_UNSIGNED_INT; break; case QRhiTexture::D24S8: *glintformat = GL_DEPTH24_STENCIL8; *glsizedintformat = *glintformat; *glformat = GL_DEPTH_STENCIL; - *gltype = GL_UNSIGNED_SHORT; + *gltype = GL_UNSIGNED_INT_24_8; break; case QRhiTexture::D32F: *glintformat = GL_DEPTH_COMPONENT32F; @@ -1254,7 +1382,7 @@ bool QRhiGles2::isFeatureSupported(QRhi::Feature feature) const case QRhi::DebugMarkers: return false; case QRhi::Timestamps: - return false; + return caps.timestamps; case QRhi::Instancing: return caps.instancing; case QRhi::CustomInstanceStepRate: @@ -1300,7 +1428,7 @@ bool QRhiGles2::isFeatureSupported(QRhi::Feature feature) const case QRhi::PipelineCacheDataLoadSave: return caps.programBinary; case QRhi::ImageDataStride: - return !caps.gles || caps.ctxMajor >= 3; + return caps.unpackRowLength; case QRhi::RenderBufferImport: return true; case QRhi::ThreeDimensionalTextures: @@ -1327,6 +1455,12 @@ bool QRhiGles2::isFeatureSupported(QRhi::Feature feature) const return caps.texture1D; case QRhi::ThreeDimensionalTextureMipmaps: return caps.texture3D; + case QRhi::MultiView: + return caps.multiView && caps.maxTextureArraySize > 0; + case QRhi::TextureViewFormat: + return false; + case QRhi::ResolveDepthStencil: + return true; default: Q_UNREACHABLE_RETURN(false); } @@ -1901,10 +2035,14 @@ const QRhiNativeHandles *QRhiGles2::nativeHandles(QRhiCommandBuffer *cb) return nullptr; } -static inline void addBoundaryCommand(QGles2CommandBuffer *cbD, QGles2CommandBuffer::Command::Cmd type) +static inline void addBoundaryCommand(QGles2CommandBuffer *cbD, QGles2CommandBuffer::Command::Cmd type, GLuint tsQuery = 0) { QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = type; + if (type == QGles2CommandBuffer::Command::BeginFrame) + cmd.args.beginFrame.timestampQuery = tsQuery; + else if (type == QGles2CommandBuffer::Command::EndFrame) + cmd.args.endFrame.timestampQuery = tsQuery; } void QRhiGles2::beginExternal(QRhiCommandBuffer *cb) @@ -1964,6 +2102,12 @@ void QRhiGles2::endExternal(QRhiCommandBuffer *cb) enqueueBindFramebuffer(cbD->currentTarget, cbD); } +double QRhiGles2::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QGles2CommandBuffer *cbD = QRHI_RES(QGles2CommandBuffer, cb); + return cbD->lastGpuTime; +} + QRhi::FrameOpResult QRhiGles2::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags) { QGles2SwapChain *swapChainD = QRHI_RES(QGles2SwapChain, swapChain); @@ -1977,7 +2121,17 @@ QRhi::FrameOpResult QRhiGles2::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF executeDeferredReleases(); swapChainD->cb.resetState(); - addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::BeginFrame); + if (swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]) { + double elapsedSec = 0; + if (swapChainD->timestamps.tryQueryTimestamps(swapChainD->currentTimestampPairIndex, this, &elapsedSec)) + swapChainD->cb.lastGpuTime = elapsedSec; + } + + GLuint tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + GLuint tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + const bool recordTimestamps = tsStart && tsEnd && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + + addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::BeginFrame, recordTimestamps ? tsStart : 0); return QRhi::FrameOpSuccess; } @@ -1987,7 +2141,15 @@ QRhi::FrameOpResult QRhiGles2::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame QGles2SwapChain *swapChainD = QRHI_RES(QGles2SwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); - addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::EndFrame); + GLuint tsStart = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2]; + GLuint tsEnd = swapChainD->timestamps.query[swapChainD->currentTimestampPairIndex * 2 + 1]; + const bool recordTimestamps = tsStart && tsEnd && !swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex]; + if (recordTimestamps) { + swapChainD->timestamps.active[swapChainD->currentTimestampPairIndex] = true; + swapChainD->currentTimestampPairIndex = (swapChainD->currentTimestampPairIndex + 1) % QGles2SwapChainTimestamps::TIMESTAMP_PAIRS; + } + + addBoundaryCommand(&swapChainD->cb, QGles2CommandBuffer::Command::EndFrame, recordTimestamps ? tsEnd : 0); if (!ensureContext(swapChainD->surface)) return contextLost ? QRhi::FrameOpDeviceLost : QRhi::FrameOpError; @@ -2019,7 +2181,12 @@ QRhi::FrameOpResult QRhiGles2::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: executeDeferredReleases(); ofr.cbWrapper.resetState(); - addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::BeginFrame); + if (rhiFlags.testFlag(QRhi::EnableTimestamps) && caps.timestamps) { + if (!ofr.tsQueries[0]) + f->glGenQueries(2, ofr.tsQueries); + } + + addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::BeginFrame, ofr.tsQueries[0]); *cb = &ofr.cbWrapper; return QRhi::FrameOpSuccess; @@ -2031,7 +2198,7 @@ QRhi::FrameOpResult QRhiGles2::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_ASSERT(ofr.active); ofr.active = false; - addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::EndFrame); + addBoundaryCommand(&ofr.cbWrapper, QGles2CommandBuffer::Command::EndFrame, ofr.tsQueries[1]); if (!ensureContext()) return contextLost ? QRhi::FrameOpDeviceLost : QRhi::FrameOpError; @@ -2044,6 +2211,16 @@ QRhi::FrameOpResult QRhiGles2::endOffscreenFrame(QRhi::EndFrameFlags flags) // another, sharing context. f->glFlush(); + if (ofr.tsQueries[0]) { + quint64 timestamps[2]; + glGetQueryObjectui64v(ofr.tsQueries[1], GL_QUERY_RESULT, ×tamps[1]); + glGetQueryObjectui64v(ofr.tsQueries[0], GL_QUERY_RESULT, ×tamps[0]); + if (timestamps[1] >= timestamps[0]) { + const quint64 nanoseconds = timestamps[1] - timestamps[0]; + ofr.cbWrapper.lastGpuTime = nanoseconds / 1000000000.0; // seconds + } + } + return QRhi::FrameOpSuccess; } @@ -2163,17 +2340,15 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb const GLenum effectiveTarget = faceTargetBase + (isCubeMap ? uint(layer) : 0u); const QPoint dp = subresDesc.destinationTopLeft(); const QByteArray rawData = subresDesc.data(); - if (!subresDesc.image().isNull()) { - QImage img = subresDesc.image(); - QSize size = img.size(); + + auto setCmdByNotCompressedData = [&](const void* data, QSize size, quint32 dataStride) + { + quint32 bytesPerLine = 0; + quint32 bytesPerPixel = 0; + textureFormatInfo(texD->m_format, size, &bytesPerLine, nullptr, &bytesPerPixel); + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::SubImage; - if (!subresDesc.sourceSize().isEmpty() || !subresDesc.sourceTopLeft().isNull()) { - const QPoint sp = subresDesc.sourceTopLeft(); - if (!subresDesc.sourceSize().isEmpty()) - size = subresDesc.sourceSize(); - img = img.copy(sp.x(), sp.y(), size.width(), size.height()); - } cmd.args.subImage.target = texD->target; cmd.args.subImage.texture = texD->texture; cmd.args.subImage.faceTarget = effectiveTarget; @@ -2185,10 +2360,38 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb cmd.args.subImage.h = size.height(); cmd.args.subImage.glformat = texD->glformat; cmd.args.subImage.gltype = texD->gltype; - cmd.args.subImage.rowStartAlign = 4; - cmd.args.subImage.rowLength = 0; - cmd.args.subImage.data = cbD->retainImage(img); + + if (dataStride == 0) + dataStride = bytesPerLine; + + cmd.args.subImage.rowStartAlign = (dataStride & 3) ? 1 : 4; + cmd.args.subImage.rowLength = caps.unpackRowLength ? (bytesPerPixel ? dataStride / bytesPerPixel : 0) : 0; + + cmd.args.subImage.data = data; + }; + + if (!subresDesc.image().isNull()) { + QImage img = subresDesc.image(); + QSize size = img.size(); + if (!subresDesc.sourceSize().isEmpty() || !subresDesc.sourceTopLeft().isNull()) { + const QPoint sp = subresDesc.sourceTopLeft(); + if (!subresDesc.sourceSize().isEmpty()) + size = subresDesc.sourceSize(); + + if (caps.unpackRowLength) { + cbD->retainImage(img); + // create a non-owning wrapper for the subimage + const uchar *data = img.constBits() + sp.y() * img.bytesPerLine() + sp.x() * (qMax(1, img.depth() / 8)); + img = QImage(data, size.width(), size.height(), img.bytesPerLine(), img.format()); + } else { + img = img.copy(sp.x(), sp.y(), size.width(), size.height()); + } + } + + setCmdByNotCompressedData(cbD->retainImage(img), size, img.bytesPerLine()); } else if (!rawData.isEmpty() && isCompressed) { + const int depth = qMax(1, texD->m_depth); + const int arraySize = qMax(0, texD->m_arraySize); if ((texD->flags().testFlag(QRhiTexture::UsedAsCompressedAtlas) || is3D || isArray) && !texD->zeroInitialized) { @@ -2200,9 +2403,9 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb quint32 byteSize = 0; compressedFormatInfo(texD->m_format, texD->m_pixelSize, nullptr, &byteSize, nullptr); if (is3D) - byteSize *= texD->m_depth; + byteSize *= depth; if (isArray) - byteSize *= texD->m_arraySize; + byteSize *= arraySize; QByteArray zeroBuf(byteSize, 0); QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::CompressedImage; @@ -2212,9 +2415,8 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb cmd.args.compressedImage.level = level; cmd.args.compressedImage.glintformat = texD->glintformat; cmd.args.compressedImage.w = texD->m_pixelSize.width(); - cmd.args.compressedImage.h = - is1D && isArray ? texD->m_arraySize : texD->m_pixelSize.height(); - cmd.args.compressedImage.depth = is3D ? texD->m_depth : (isArray ? texD->m_arraySize : 0); + cmd.args.compressedImage.h = is1D && isArray ? arraySize : texD->m_pixelSize.height(); + cmd.args.compressedImage.depth = is3D ? depth : (isArray ? arraySize : 0); cmd.args.compressedImage.size = byteSize; cmd.args.compressedImage.data = cbD->retainData(zeroBuf); texD->zeroInitialized = true; @@ -2246,39 +2448,16 @@ void QRhiGles2::enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cb cmd.args.compressedImage.level = level; cmd.args.compressedImage.glintformat = texD->glintformat; cmd.args.compressedImage.w = size.width(); - cmd.args.compressedImage.h = is1D && isArray ? texD->m_arraySize : size.height(); - cmd.args.compressedImage.depth = is3D ? texD->m_depth : (isArray ? texD->m_arraySize : 0); + cmd.args.compressedImage.h = is1D && isArray ? arraySize : size.height(); + cmd.args.compressedImage.depth = is3D ? depth : (isArray ? arraySize : 0); cmd.args.compressedImage.size = rawData.size(); cmd.args.compressedImage.data = cbD->retainData(rawData); } } else if (!rawData.isEmpty()) { const QSize size = subresDesc.sourceSize().isEmpty() ? q->sizeForMipLevel(level, texD->m_pixelSize) : subresDesc.sourceSize(); - quint32 bytesPerLine = 0; - quint32 bytesPerPixel = 0; - textureFormatInfo(texD->m_format, size, &bytesPerLine, nullptr, &bytesPerPixel); - QGles2CommandBuffer::Command &cmd(cbD->commands.get()); - cmd.cmd = QGles2CommandBuffer::Command::SubImage; - cmd.args.subImage.target = texD->target; - cmd.args.subImage.texture = texD->texture; - cmd.args.subImage.faceTarget = effectiveTarget; - cmd.args.subImage.level = level; - cmd.args.subImage.dx = dp.x(); - cmd.args.subImage.dy = is1D && isArray ? layer : dp.y(); - cmd.args.subImage.dz = is3D || isArray ? layer : 0; - cmd.args.subImage.w = size.width(); - cmd.args.subImage.h = size.height(); - cmd.args.subImage.glformat = texD->glformat; - cmd.args.subImage.gltype = texD->gltype; - // Default unpack alignment (row start alignment - // requirement) is 4. QImage guarantees 4 byte aligned - // row starts, but our raw data here does not. - cmd.args.subImage.rowStartAlign = (bytesPerLine & 3) ? 1 : 4; - if (subresDesc.dataStride() && bytesPerPixel) - cmd.args.subImage.rowLength = subresDesc.dataStride() / bytesPerPixel; - else - cmd.args.subImage.rowLength = 0; - cmd.args.subImage.data = cbD->retainData(rawData); + + setCmdByNotCompressedData(cbD->retainData(rawData), size, subresDesc.dataStride()); } else { qWarning("Invalid texture upload for %p layer=%d mip=%d", texD, layer, level); } @@ -2797,6 +2976,8 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) const QGles2CommandBuffer::Command &cmd(*it); switch (cmd.cmd) { case QGles2CommandBuffer::Command::BeginFrame: + if (cmd.args.beginFrame.timestampQuery) + glQueryCounter(cmd.args.beginFrame.timestampQuery, GL_TIMESTAMP); if (caps.coreProfile) { if (!vao) f->glGenVertexArrays(1, &vao); @@ -2823,6 +3004,8 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) #endif if (vao) f->glBindVertexArray(0); + if (cmd.args.endFrame.timestampQuery) + glQueryCounter(cmd.args.endFrame.timestampQuery, GL_TIMESTAMP); break; case QGles2CommandBuffer::Command::ResetFrame: if (vao) @@ -2967,6 +3150,38 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) type = GL_HALF_FLOAT; size = 1; break; + case QRhiVertexInputAttribute::UShort4: + type = GL_UNSIGNED_SHORT; + size = 4; + break; + case QRhiVertexInputAttribute::UShort3: + type = GL_UNSIGNED_SHORT; + size = 3; + break; + case QRhiVertexInputAttribute::UShort2: + type = GL_UNSIGNED_SHORT; + size = 2; + break; + case QRhiVertexInputAttribute::UShort: + type = GL_UNSIGNED_SHORT; + size = 1; + break; + case QRhiVertexInputAttribute::SShort4: + type = GL_SHORT; + size = 4; + break; + case QRhiVertexInputAttribute::SShort3: + type = GL_SHORT; + size = 3; + break; + case QRhiVertexInputAttribute::SShort2: + type = GL_SHORT; + size = 2; + break; + case QRhiVertexInputAttribute::SShort: + type = GL_SHORT; + size = 1; + break; default: break; } @@ -3109,7 +3324,7 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) } if (caps.hasDrawBuffersFunc) f->glDrawBuffers(bufs.count(), bufs.constData()); - if (caps.srgbCapableDefaultFramebuffer) { + if (caps.srgbWriteControl) { if (cmd.args.bindFramebuffer.srgb) f->glEnable(GL_FRAMEBUFFER_SRGB); else @@ -3141,7 +3356,7 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::GetBufferSubData: { - QRhiBufferReadbackResult *result = cmd.args.getBufferSubData.result; + QRhiReadbackResult *result = cmd.args.getBufferSubData.result; bindVertexIndexBufferWithStateReset(&state, f, cmd.args.getBufferSubData.target, cmd.args.getBufferSubData.buffer); if (caps.gles) { if (caps.properMapBuffer) { @@ -3264,6 +3479,14 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) result->data.resize(w * h * 8); f->glReadPixels(0, 0, w, h, GL_RGBA, GL_HALF_FLOAT, result->data.data()); break; + case QRhiTexture::R16F: + result->data.resize(w * h * 2); + f->glReadPixels(0, 0, w, h, GL_RED, GL_HALF_FLOAT, result->data.data()); + break; + case QRhiTexture::R32F: + result->data.resize(w * h * 4); + f->glReadPixels(0, 0, w, h, GL_RED, GL_FLOAT, result->data.data()); + break; case QRhiTexture::RGBA32F: result->data.resize(w * h * 16); f->glReadPixels(0, 0, w, h, GL_RGBA, GL_FLOAT, result->data.data()); @@ -3362,23 +3585,128 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) break; case QGles2CommandBuffer::Command::BlitFromRenderbuffer: { + // Altering the scissor state, so reset the stored state, although + // not strictly required as long as blit is done in endPass() only. + cbD->graphicsPassState.reset(); + f->glDisable(GL_SCISSOR_TEST); GLuint fbo[2]; f->glGenFramebuffers(2, fbo); f->glBindFramebuffer(GL_READ_FRAMEBUFFER, fbo[0]); - f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, - GL_RENDERBUFFER, cmd.args.blitFromRb.renderbuffer); + const bool ds = cmd.args.blitFromRenderbuffer.isDepthStencil; + if (ds) { + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + } else { + f->glFramebufferRenderbuffer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + GL_RENDERBUFFER, cmd.args.blitFromRenderbuffer.renderbuffer); + } f->glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo[1]); - if (cmd.args.blitFromRb.target == GL_TEXTURE_3D || cmd.args.blitFromRb.target == GL_TEXTURE_2D_ARRAY) { - f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, - cmd.args.blitFromRb.texture, cmd.args.blitFromRb.dstLevel, cmd.args.blitFromRb.dstLayer); + if (cmd.args.blitFromRenderbuffer.target == GL_TEXTURE_3D || cmd.args.blitFromRenderbuffer.target == GL_TEXTURE_2D_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + } else { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromRenderbuffer.dstTexture, + cmd.args.blitFromRenderbuffer.dstLevel, + cmd.args.blitFromRenderbuffer.dstLayer); + } } else { - f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromRb.target, - cmd.args.blitFromRb.texture, cmd.args.blitFromRb.dstLevel); + if (ds) { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + } else { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromRenderbuffer.target, + cmd.args.blitFromRenderbuffer.dstTexture, cmd.args.blitFromRenderbuffer.dstLevel); + } } - f->glBlitFramebuffer(0, 0, cmd.args.blitFromRb.w, cmd.args.blitFromRb.h, - 0, 0, cmd.args.blitFromRb.w, cmd.args.blitFromRb.h, - GL_COLOR_BUFFER_BIT, - GL_LINEAR); + f->glBlitFramebuffer(0, 0, cmd.args.blitFromRenderbuffer.w, cmd.args.blitFromRenderbuffer.h, + 0, 0, cmd.args.blitFromRenderbuffer.w, cmd.args.blitFromRenderbuffer.h, + ds ? GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT : GL_COLOR_BUFFER_BIT, + GL_NEAREST); // Qt 5 used Nearest when resolving samples, stick to that + f->glBindFramebuffer(GL_FRAMEBUFFER, ctx->defaultFramebufferObject()); + f->glDeleteFramebuffers(2, fbo); + } + break; + case QGles2CommandBuffer::Command::BlitFromTexture: + { + // Altering the scissor state, so reset the stored state, although + // not strictly required as long as blit is done in endPass() only. + cbD->graphicsPassState.reset(); + f->glDisable(GL_SCISSOR_TEST); + GLuint fbo[2]; + f->glGenFramebuffers(2, fbo); + f->glBindFramebuffer(GL_READ_FRAMEBUFFER, fbo[0]); + const bool ds = cmd.args.blitFromTexture.isDepthStencil; + if (cmd.args.blitFromTexture.srcTarget == GL_TEXTURE_2D_MULTISAMPLE_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + } else { + f->glFramebufferTextureLayer(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromTexture.srcTexture, + cmd.args.blitFromTexture.srcLevel, + cmd.args.blitFromTexture.srcLayer); + } + } else { + if (ds) { + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + } else { + f->glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromTexture.srcTarget, + cmd.args.blitFromTexture.srcTexture, cmd.args.blitFromTexture.srcLevel); + } + } + f->glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo[1]); + if (cmd.args.blitFromTexture.dstTarget == GL_TEXTURE_3D || cmd.args.blitFromTexture.dstTarget == GL_TEXTURE_2D_ARRAY) { + if (ds) { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + } else { + f->glFramebufferTextureLayer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, + cmd.args.blitFromTexture.dstTexture, + cmd.args.blitFromTexture.dstLevel, + cmd.args.blitFromTexture.dstLayer); + } + } else { + if (ds) { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + } else { + f->glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, cmd.args.blitFromTexture.dstTarget, + cmd.args.blitFromTexture.dstTexture, cmd.args.blitFromTexture.dstLevel); + } + } + f->glBlitFramebuffer(0, 0, cmd.args.blitFromTexture.w, cmd.args.blitFromTexture.h, + 0, 0, cmd.args.blitFromTexture.w, cmd.args.blitFromTexture.h, + ds ? GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT : GL_COLOR_BUFFER_BIT, + GL_NEAREST); // Qt 5 used Nearest when resolving samples, stick to that f->glBindFramebuffer(GL_FRAMEBUFFER, ctx->defaultFramebufferObject()); f->glDeleteFramebuffers(2, fbo); } @@ -3427,6 +3755,13 @@ void QRhiGles2::executeCommandBuffer(QRhiCommandBuffer *cb) if (caps.compute) f->glMemoryBarrier(cmd.args.barrier.barriers); break; + case QGles2CommandBuffer::Command::InvalidateFramebuffer: + if (caps.gles && caps.ctxMajor >= 3) { + f->glInvalidateFramebuffer(GL_DRAW_FRAMEBUFFER, + cmd.args.invalidateFramebuffer.attCount, + cmd.args.invalidateFramebuffer.att); + } + break; default: break; } @@ -3682,7 +4017,7 @@ void QRhiGles2::bindCombinedSampler(QGles2CommandBuffer *cbD, QGles2Texture *tex f->glTexParameteri(texD->target, GL_TEXTURE_MAG_FILTER, GLint(samplerD->d.glmagfilter)); f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_S, GLint(samplerD->d.glwraps)); f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_T, GLint(samplerD->d.glwrapt)); - if (caps.texture3D) + if (caps.texture3D && texD->target == GL_TEXTURE_3D) f->glTexParameteri(texD->target, GL_TEXTURE_WRAP_R, GLint(samplerD->d.glwrapr)); if (caps.textureCompareMode) { if (samplerD->d.gltexcomparefunc != GL_NEVER) { @@ -4225,32 +4560,129 @@ void QRhiGles2::endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resource if (cbD->currentTarget->resourceType() == QRhiResource::TextureRenderTarget) { QGles2TextureRenderTarget *rtTex = QRHI_RES(QGles2TextureRenderTarget, cbD->currentTarget); - if (rtTex->m_desc.cbeginColorAttachments() != rtTex->m_desc.cendColorAttachments()) { - // handle only 1 color attachment and only (msaa) renderbuffer - const QRhiColorAttachment &colorAtt(*rtTex->m_desc.cbeginColorAttachments()); - if (colorAtt.resolveTexture()) { - Q_ASSERT(colorAtt.renderBuffer()); + for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); + it != itEnd; ++it) + { + const QRhiColorAttachment &colorAtt(*it); + if (!colorAtt.resolveTexture()) + continue; + + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + const QSize size = resolveTexD->pixelSize(); + if (colorAtt.renderBuffer()) { QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, colorAtt.renderBuffer()); - const QSize size = colorAtt.resolveTexture()->pixelSize(); if (rbD->pixelSize() != size) { qWarning("Resolve source (%dx%d) and target (%dx%d) size does not match", rbD->pixelSize().width(), rbD->pixelSize().height(), size.width(), size.height()); } QGles2CommandBuffer::Command &cmd(cbD->commands.get()); cmd.cmd = QGles2CommandBuffer::Command::BlitFromRenderbuffer; - cmd.args.blitFromRb.renderbuffer = rbD->renderbuffer; - cmd.args.blitFromRb.w = size.width(); - cmd.args.blitFromRb.h = size.height(); - QGles2Texture *colorTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); - if (colorTexD->m_flags.testFlag(QRhiTexture::CubeMap)) - cmd.args.blitFromRb.target = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(colorAtt.resolveLayer()); + cmd.args.blitFromRenderbuffer.renderbuffer = rbD->renderbuffer; + cmd.args.blitFromRenderbuffer.w = size.width(); + cmd.args.blitFromRenderbuffer.h = size.height(); + if (resolveTexD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromRenderbuffer.target = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(colorAtt.resolveLayer()); else - cmd.args.blitFromRb.target = colorTexD->target; - cmd.args.blitFromRb.texture = colorTexD->texture; - cmd.args.blitFromRb.dstLevel = colorAtt.resolveLevel(); - const bool hasZ = colorTexD->m_flags.testFlag(QRhiTexture::ThreeDimensional) - || colorTexD->m_flags.testFlag(QRhiTexture::TextureArray); - cmd.args.blitFromRb.dstLayer = hasZ ? colorAtt.resolveLayer() : 0; + cmd.args.blitFromRenderbuffer.target = resolveTexD->target; + cmd.args.blitFromRenderbuffer.dstTexture = resolveTexD->texture; + cmd.args.blitFromRenderbuffer.dstLevel = colorAtt.resolveLevel(); + const bool hasZ = resolveTexD->m_flags.testFlag(QRhiTexture::ThreeDimensional) + || resolveTexD->m_flags.testFlag(QRhiTexture::TextureArray); + cmd.args.blitFromRenderbuffer.dstLayer = hasZ ? colorAtt.resolveLayer() : 0; + cmd.args.blitFromRenderbuffer.isDepthStencil = false; + } else if (caps.glesMultisampleRenderToTexture) { + // Nothing to do, resolving into colorAtt.resolveTexture() is automatic, + // colorAtt.texture() is in fact not used for anything. + } else { + Q_ASSERT(colorAtt.texture()); + QGles2Texture *texD = QRHI_RES(QGles2Texture, colorAtt.texture()); + if (texD->pixelSize() != size) { + qWarning("Resolve source (%dx%d) and target (%dx%d) size does not match", + texD->pixelSize().width(), texD->pixelSize().height(), size.width(), size.height()); + } + const int resolveCount = colorAtt.multiViewCount() >= 2 ? colorAtt.multiViewCount() : 1; + for (int resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + const int srcLayer = colorAtt.layer() + resolveIdx; + const int dstLayer = colorAtt.resolveLayer() + resolveIdx; + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromTexture; + if (texD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromTexture.srcTarget = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(srcLayer); + else + cmd.args.blitFromTexture.srcTarget = texD->target; + cmd.args.blitFromTexture.srcTexture = texD->texture; + cmd.args.blitFromTexture.srcLevel = colorAtt.level(); + cmd.args.blitFromTexture.srcLayer = 0; + if (texD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || texD->m_flags.testFlag(QRhiTexture::TextureArray)) + cmd.args.blitFromTexture.srcLayer = srcLayer; + cmd.args.blitFromTexture.w = size.width(); + cmd.args.blitFromTexture.h = size.height(); + if (resolveTexD->m_flags.testFlag(QRhiTexture::CubeMap)) + cmd.args.blitFromTexture.dstTarget = GL_TEXTURE_CUBE_MAP_POSITIVE_X + uint(dstLayer); + else + cmd.args.blitFromTexture.dstTarget = resolveTexD->target; + cmd.args.blitFromTexture.dstTexture = resolveTexD->texture; + cmd.args.blitFromTexture.dstLevel = colorAtt.resolveLevel(); + cmd.args.blitFromTexture.dstLayer = 0; + if (resolveTexD->m_flags.testFlag(QRhiTexture::ThreeDimensional) || resolveTexD->m_flags.testFlag(QRhiTexture::TextureArray)) + cmd.args.blitFromTexture.dstLayer = dstLayer; + cmd.args.blitFromTexture.isDepthStencil = false; + } + } + } + + if (rtTex->m_desc.depthResolveTexture()) { + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, rtTex->m_desc.depthResolveTexture()); + const QSize size = depthResolveTexD->pixelSize(); + if (rtTex->m_desc.depthStencilBuffer()) { + QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, rtTex->m_desc.depthStencilBuffer()); + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromRenderbuffer; + cmd.args.blitFromRenderbuffer.renderbuffer = rbD->renderbuffer; + cmd.args.blitFromRenderbuffer.w = size.width(); + cmd.args.blitFromRenderbuffer.h = size.height(); + cmd.args.blitFromRenderbuffer.target = depthResolveTexD->target; + cmd.args.blitFromRenderbuffer.dstTexture = depthResolveTexD->texture; + cmd.args.blitFromRenderbuffer.dstLevel = 0; + cmd.args.blitFromRenderbuffer.dstLayer = 0; + cmd.args.blitFromRenderbuffer.isDepthStencil = true; + } else if (caps.glesMultisampleRenderToTexture) { + // Nothing to do, resolving into depthResolveTexture() is automatic. + } else { + QGles2Texture *depthTexD = QRHI_RES(QGles2Texture, rtTex->m_desc.depthTexture()); + const int resolveCount = depthTexD->arraySize() >= 2 ? depthTexD->arraySize() : 1; + for (int resolveIdx = 0; resolveIdx < resolveCount; ++resolveIdx) { + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::BlitFromTexture; + cmd.args.blitFromTexture.srcTarget = depthTexD->target; + cmd.args.blitFromTexture.srcTexture = depthTexD->texture; + cmd.args.blitFromTexture.srcLevel = 0; + cmd.args.blitFromTexture.srcLayer = resolveIdx; + cmd.args.blitFromTexture.w = size.width(); + cmd.args.blitFromTexture.h = size.height(); + cmd.args.blitFromTexture.dstTarget = depthResolveTexD->target; + cmd.args.blitFromTexture.dstTexture = depthResolveTexD->texture; + cmd.args.blitFromTexture.dstLevel = 0; + cmd.args.blitFromTexture.dstLayer = resolveIdx; + cmd.args.blitFromTexture.isDepthStencil = true; + } + } + } + + const bool mayDiscardDepthStencil = + (rtTex->m_desc.depthStencilBuffer() + || (rtTex->m_desc.depthTexture() && rtTex->m_flags.testFlag(QRhiTextureRenderTarget::DoNotStoreDepthStencilContents))) + && !rtTex->m_desc.depthResolveTexture(); + if (mayDiscardDepthStencil) { + QGles2CommandBuffer::Command &cmd(cbD->commands.get()); + cmd.cmd = QGles2CommandBuffer::Command::InvalidateFramebuffer; + if (caps.needsDepthStencilCombinedAttach) { + cmd.args.invalidateFramebuffer.attCount = 1; + cmd.args.invalidateFramebuffer.att[0] = GL_DEPTH_STENCIL_ATTACHMENT; + } else { + cmd.args.invalidateFramebuffer.attCount = 2; + cmd.args.invalidateFramebuffer.att[0] = GL_DEPTH_ATTACHMENT; + cmd.args.invalidateFramebuffer.att[1] = GL_STENCIL_ATTACHMENT; } } } @@ -4885,6 +5317,9 @@ bool QGles2Buffer::create() rhiD->f->glBindBuffer(targetForDataOps, buffer); rhiD->f->glBufferData(targetForDataOps, nonZeroSize, nullptr, m_type == Dynamic ? GL_DYNAMIC_DRAW : GL_STATIC_DRAW); + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_BUFFER, buffer, -1, m_objectName.constData()); + usageState.access = AccessNone; rhiD->registerResource(this); @@ -5039,6 +5474,9 @@ bool QGles2RenderBuffer::create() break; } + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_RENDERBUFFER, renderbuffer, -1, m_objectName.constData()); + owns = true; generation += 1; rhiD->registerResource(this); @@ -5155,12 +5593,10 @@ bool QGles2Texture::prepareCreate(QSize *adjustedSize) return false; } - m_depth = qMax(1, m_depth); if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } - m_arraySize = qMax(0, m_arraySize); if (m_arraySize > 0 && !isArray) { qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); return false; @@ -5171,7 +5607,7 @@ bool QGles2Texture::prepareCreate(QSize *adjustedSize) } target = isCube ? GL_TEXTURE_CUBE_MAP - : m_sampleCount > 1 ? GL_TEXTURE_2D_MULTISAMPLE + : m_sampleCount > 1 ? (isArray ? GL_TEXTURE_2D_MULTISAMPLE_ARRAY : GL_TEXTURE_2D_MULTISAMPLE) : (is3D ? GL_TEXTURE_3D : (is1D ? (isArray ? GL_TEXTURE_1D_ARRAY : GL_TEXTURE_1D) : (isArray ? GL_TEXTURE_2D_ARRAY : GL_TEXTURE_2D))); @@ -5235,13 +5671,13 @@ bool QGles2Texture::create() const QSize mipSize = rhiD->q->sizeForMipLevel(level, size); if (isArray) rhiD->f->glTexImage2D(target, level, GLint(glintformat), mipSize.width(), - m_arraySize, 0, glformat, gltype, nullptr); + qMax(0, m_arraySize), 0, glformat, gltype, nullptr); else rhiD->glTexImage1D(target, level, GLint(glintformat), mipSize.width(), 0, glformat, gltype, nullptr); } } else if (is3D || isArray) { - const int layerCount = is3D ? m_depth : m_arraySize; + const int layerCount = is3D ? qMax(1, m_depth) : qMax(0, m_arraySize); if (hasMipMaps) { for (int level = 0; level != mipLevelCount; ++level) { const QSize mipSize = rhiD->q->sizeForMipLevel(level, size); @@ -5263,8 +5699,16 @@ bool QGles2Texture::create() } } } else { - rhiD->f->glTexImage2D(target, 0, GLint(glintformat), size.width(), size.height(), - 0, glformat, gltype, nullptr); + // 2D texture. For multisample textures the GLES 3.1 + // glStorage2DMultisample must be used for portability. + if (m_sampleCount > 1 && rhiD->caps.multisampledTexture) { + // internal format must be sized + rhiD->f->glTexStorage2DMultisample(target, m_sampleCount, glsizedintformat, + size.width(), size.height(), GL_TRUE); + } else { + rhiD->f->glTexImage2D(target, 0, GLint(glintformat), size.width(), size.height(), + 0, glformat, gltype, nullptr); + } } } else { // Must be specified with immutable storage functions otherwise @@ -5273,10 +5717,14 @@ bool QGles2Texture::create() if (is1D && !isArray) rhiD->glTexStorage1D(target, mipLevelCount, glsizedintformat, size.width()); else if (!is1D && (is3D || isArray)) - rhiD->f->glTexStorage3D(target, mipLevelCount, glsizedintformat, size.width(), size.height(), is3D ? m_depth : m_arraySize); + rhiD->f->glTexStorage3D(target, mipLevelCount, glsizedintformat, size.width(), size.height(), + is3D ? qMax(1, m_depth) : qMax(0, m_arraySize)); + else if (m_sampleCount > 1) + rhiD->f->glTexStorage2DMultisample(target, m_sampleCount, glsizedintformat, + size.width(), size.height(), GL_TRUE); else rhiD->f->glTexStorage2D(target, mipLevelCount, glsizedintformat, size.width(), - is1D ? m_arraySize : size.height()); + is1D ? qMax(0, m_arraySize) : size.height()); } specified = true; } else { @@ -5286,6 +5734,9 @@ bool QGles2Texture::create() specified = false; } + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_TEXTURE, texture, -1, m_objectName.constData()); + owns = true; generation += 1; @@ -5332,7 +5783,9 @@ QGles2Sampler::~QGles2Sampler() void QGles2Sampler::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2Sampler::create() @@ -5345,6 +5798,8 @@ bool QGles2Sampler::create() d.gltexcomparefunc = toGlTextureCompareFunc(m_compareOp); generation += 1; + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(this, false); return true; } @@ -5361,7 +5816,9 @@ QGles2RenderPassDescriptor::~QGles2RenderPassDescriptor() void QGles2RenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -5372,7 +5829,10 @@ bool QGles2RenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *ot QRhiRenderPassDescriptor *QGles2RenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; } QVector<quint32> QGles2RenderPassDescriptor::serializedFormat() const @@ -5433,8 +5893,10 @@ void QGles2TextureRenderTarget::destroy() e.type = QRhiGles2::DeferredReleaseEntry::TextureRenderTarget; e.textureRenderTarget.framebuffer = framebuffer; + e.textureRenderTarget.nonMsaaThrowawayDepthTexture = nonMsaaThrowawayDepthTexture; framebuffer = 0; + nonMsaaThrowawayDepthTexture = 0; QRHI_RES_RHI(QRhiGles2); if (rhiD) { @@ -5445,7 +5907,10 @@ void QGles2TextureRenderTarget::destroy() QRhiRenderPassDescriptor *QGles2TextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; } bool QGles2TextureRenderTarget::create() @@ -5478,6 +5943,7 @@ bool QGles2TextureRenderTarget::create() d.colorAttCount = 0; int attIndex = 0; + int multiViewCount = 0; for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it, ++attIndex) { d.colorAttCount += 1; const QRhiColorAttachment &colorAtt(*it); @@ -5488,20 +5954,56 @@ bool QGles2TextureRenderTarget::create() QGles2Texture *texD = QRHI_RES(QGles2Texture, texture); Q_ASSERT(texD->texture && texD->specified); if (texD->flags().testFlag(QRhiTexture::ThreeDimensional) || texD->flags().testFlag(QRhiTexture::TextureArray)) { - rhiD->f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), texD->texture, - colorAtt.level(), colorAtt.layer()); + if (colorAtt.multiViewCount() < 2) { + rhiD->f->glFramebufferTextureLayer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), texD->texture, + colorAtt.level(), colorAtt.layer()); + } else { + multiViewCount = colorAtt.multiViewCount(); + if (texD->sampleCount() > 1 && rhiD->caps.glesMultiviewMultisampleRenderToTexture && colorAtt.resolveTexture()) { + // Special path for GLES and GL_OVR_multiview_multisampled_render_to_texture: + // ignore the color attachment's (multisample) texture + // array and give the resolve texture array to GL. (no + // explicit resolving is needed by us later on) + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_COLOR_ATTACHMENT0 + uint(attIndex), + resolveTexD->texture, + colorAtt.resolveLevel(), + texD->sampleCount(), + colorAtt.resolveLayer(), + multiViewCount); + } else { + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, + GL_COLOR_ATTACHMENT0 + uint(attIndex), + texD->texture, + colorAtt.level(), + colorAtt.layer(), + multiViewCount); + } + } } else if (texD->flags().testFlag(QRhiTexture::OneDimensional)) { rhiD->glFramebufferTexture1D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), texD->target + uint(colorAtt.layer()), texD->texture, colorAtt.level()); } else { - const GLenum faceTargetBase = texD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : texD->target; - rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.layer()), - texD->texture, colorAtt.level()); + if (texD->sampleCount() > 1 && rhiD->caps.glesMultisampleRenderToTexture && colorAtt.resolveTexture()) { + // Special path for GLES and GL_EXT_multisampled_render_to_texture: + // ignore the color attachment's (multisample) texture and + // give the resolve texture to GL. (no explicit resolving is + // needed by us later on) + QGles2Texture *resolveTexD = QRHI_RES(QGles2Texture, colorAtt.resolveTexture()); + const GLenum faceTargetBase = resolveTexD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : resolveTexD->target; + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.resolveLayer()), + resolveTexD->texture, colorAtt.level(), texD->sampleCount()); + } else { + const GLenum faceTargetBase = texD->flags().testFlag(QRhiTexture::CubeMap) ? GL_TEXTURE_CUBE_MAP_POSITIVE_X : texD->target; + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0 + uint(attIndex), faceTargetBase + uint(colorAtt.layer()), + texD->texture, colorAtt.level()); + } } if (attIndex == 0) { d.pixelSize = rhiD->q->sizeForMipLevel(colorAtt.level(), texD->pixelSize()); - d.sampleCount = 1; + d.sampleCount = texD->sampleCount(); } } else if (renderBuffer) { QGles2RenderBuffer *rbD = QRHI_RES(QGles2RenderBuffer, renderBuffer); @@ -5522,12 +6024,14 @@ bool QGles2TextureRenderTarget::create() } else { rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, depthRbD->renderbuffer); - if (depthRbD->stencilRenderbuffer) + if (depthRbD->stencilRenderbuffer) { rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, depthRbD->stencilRenderbuffer); - else // packed + } else { + // packed depth-stencil rhiD->f->glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_RENDERBUFFER, depthRbD->renderbuffer); + } } if (d.colorAttCount == 0) { d.pixelSize = depthRbD->pixelSize(); @@ -5535,11 +6039,105 @@ bool QGles2TextureRenderTarget::create() } } else { QGles2Texture *depthTexD = QRHI_RES(QGles2Texture, m_desc.depthTexture()); - rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->target, - depthTexD->texture, 0); + if (multiViewCount < 2) { + if (depthTexD->sampleCount() > 1 && rhiD->caps.glesMultisampleRenderToTexture && m_desc.depthResolveTexture()) { + // Special path for GLES and + // GL_EXT_multisampled_render_to_texture, for depth-stencil. + // Relevant only when depthResolveTexture is set. + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, m_desc.depthResolveTexture()); + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthResolveTexD->target, + depthResolveTexD->texture, 0, depthTexD->sampleCount()); + if (rhiD->isStencilSupportingFormat(depthResolveTexD->format())) { + rhiD->glFramebufferTexture2DMultisampleEXT(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthResolveTexD->target, + depthResolveTexD->texture, 0, depthTexD->sampleCount()); + } + } else { + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->target, + depthTexD->texture, 0); + if (rhiD->isStencilSupportingFormat(depthTexD->format())) { + rhiD->f->glFramebufferTexture2D(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthTexD->target, + depthTexD->texture, 0); + } + } + } else { + if (depthTexD->sampleCount() > 1 && rhiD->caps.glesMultiviewMultisampleRenderToTexture) { + // And so it turns out + // https://registry.khronos.org/OpenGL/extensions/OVR/OVR_multiview.txt + // does not work with multisample 2D texture arrays. (at least + // that's what Issue 30 in the extension spec seems to imply) + // + // There is https://registry.khronos.org/OpenGL/extensions/EXT/EXT_multiview_texture_multisample.txt + // that seems to resolve that, but that does not seem to + // work (or not available) on GLES devices such as the Quest 3. + // + // So instead, on GLES we can use the + // multisample-multiview-auto-resolving version (which in + // turn is not supported on desktop GL e.g. by NVIDIA), too + // bad we have a multisample depth texture array here as + // every other API out there requires that. So, in absence + // of a depthResolveTexture, create a temporary one ignoring + // what the user has already created. + // + if (!m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture()) { + qWarning("Attempted to create a multiview+multisample QRhiTextureRenderTarget, but DoNotStoreDepthStencilContents was not set." + " This path has no choice but to behave as if DoNotStoreDepthStencilContents was set, because QRhi is forced to create" + " a throwaway non-multisample depth texture here. Set the flag to silence this warning, or set a depthResolveTexture."); + } + if (m_desc.depthResolveTexture()) { + QGles2Texture *depthResolveTexD = QRHI_RES(QGles2Texture, m_desc.depthResolveTexture()); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_DEPTH_ATTACHMENT, + depthResolveTexD->texture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + if (rhiD->isStencilSupportingFormat(depthResolveTexD->format())) { + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_STENCIL_ATTACHMENT, + depthResolveTexD->texture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + } + } else { + if (!nonMsaaThrowawayDepthTexture) { + rhiD->f->glGenTextures(1, &nonMsaaThrowawayDepthTexture); + rhiD->f->glBindTexture(GL_TEXTURE_2D_ARRAY, nonMsaaThrowawayDepthTexture); + rhiD->f->glTexStorage3D(GL_TEXTURE_2D_ARRAY, 1, GL_DEPTH24_STENCIL8, + depthTexD->pixelSize().width(), depthTexD->pixelSize().height(), multiViewCount); + } + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_DEPTH_ATTACHMENT, + nonMsaaThrowawayDepthTexture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + rhiD->glFramebufferTextureMultisampleMultiviewOVR(GL_FRAMEBUFFER, + GL_STENCIL_ATTACHMENT, + nonMsaaThrowawayDepthTexture, + 0, + depthTexD->sampleCount(), + 0, + multiViewCount); + } + } else { + // The depth texture here must be an array with at least + // multiViewCount elements, and the format should be D24 or D32F + // for depth only, or D24S8 for depth and stencil. + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, depthTexD->texture, + 0, 0, multiViewCount); + if (rhiD->isStencilSupportingFormat(depthTexD->format())) { + rhiD->glFramebufferTextureMultiviewOVR(GL_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, depthTexD->texture, + 0, 0, multiViewCount); + } + } + } if (d.colorAttCount == 0) { d.pixelSize = depthTexD->pixelSize(); - d.sampleCount = 1; + d.sampleCount = depthTexD->sampleCount(); } } d.dsAttCount = 1; @@ -5556,6 +6154,9 @@ bool QGles2TextureRenderTarget::create() return false; } + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_FRAMEBUFFER, framebuffer, -1, m_objectName.constData()); + QRhiRenderTargetAttachmentTracker::updateResIdList<QGles2Texture, QGles2RenderBuffer>(m_desc, &d.currentResIdList); rhiD->registerResource(this); @@ -5592,7 +6193,9 @@ QGles2ShaderResourceBindings::~QGles2ShaderResourceBindings() void QGles2ShaderResourceBindings::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiGles2); + if (rhiD) + rhiD->unregisterResource(this); } bool QGles2ShaderResourceBindings::create() @@ -5615,6 +6218,7 @@ bool QGles2ShaderResourceBindings::create() rhiD->updateLayoutDesc(this); generation += 1; + rhiD->registerResource(this, false); return true; } @@ -5799,6 +6403,9 @@ bool QGles2GraphicsPipeline::create() currentSrb = nullptr; currentSrbGeneration = 0; + if (rhiD->glObjectLabel) + rhiD->glObjectLabel(GL_PROGRAM, program, -1, m_objectName.constData()); + rhiD->pipelineCreationEnd(); generation += 1; rhiD->registerResource(this); @@ -5969,7 +6576,12 @@ QRhiRenderTarget *QGles2SwapChain::currentFrameRenderTarget(StereoTargetBuffer t QSize QGles2SwapChain::surfacePixelSize() { Q_ASSERT(m_window); - return m_window->size() * m_window->devicePixelRatio(); + if (QPlatformWindow *platformWindow = m_window->handle()) + // Prefer using QPlatformWindow geometry and DPR in order to avoid + // errors due to rounded QWindow geometry. + return platformWindow->geometry().size() * platformWindow->devicePixelRatio(); + else + return m_window->size() * m_window->devicePixelRatio(); } bool QGles2SwapChain::isFormatSupported(Format f) @@ -5979,7 +6591,10 @@ bool QGles2SwapChain::isFormatSupported(Format f) QRhiRenderPassDescriptor *QGles2SwapChain::newCompatibleRenderPassDescriptor() { - return new QGles2RenderPassDescriptor(m_rhi); + QGles2RenderPassDescriptor *rpD = new QGles2RenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiGles2); + rhiD->registerResource(rpD, false); + return rpD; } void QGles2SwapChain::initSwapChainRenderTarget(QGles2SwapChainRenderTarget *rt) @@ -6023,15 +6638,59 @@ bool QGles2SwapChain::createOrResize() frameCount = 0; + QRHI_RES_RHI(QRhiGles2); + if (rhiD->rhiFlags.testFlag(QRhi::EnableTimestamps) && rhiD->caps.timestamps) + timestamps.prepare(rhiD); + // The only reason to register this fairly fake gl swapchain // object with no native resources underneath is to be able to // implement a safe destroy(). - if (needsRegistration) { - QRHI_RES_RHI(QRhiGles2); - rhiD->registerResource(this); - } + if (needsRegistration) + rhiD->registerResource(this, false); return true; } +void QGles2SwapChainTimestamps::prepare(QRhiGles2 *rhiD) +{ + if (!query[0]) + rhiD->f->glGenQueries(TIMESTAMP_PAIRS * 2, query); +} + +void QGles2SwapChainTimestamps::destroy(QRhiGles2 *rhiD) +{ + rhiD->f->glDeleteQueries(TIMESTAMP_PAIRS * 2, query); + memset(active, 0, sizeof(active)); + memset(query, 0, sizeof(query)); +} + +bool QGles2SwapChainTimestamps::tryQueryTimestamps(int pairIndex, QRhiGles2 *rhiD, double *elapsedSec) +{ + if (!active[pairIndex]) + return false; + + GLuint tsStart = query[pairIndex * 2]; + GLuint tsEnd = query[pairIndex * 2 + 1]; + + GLuint ready = GL_FALSE; + rhiD->f->glGetQueryObjectuiv(tsEnd, GL_QUERY_RESULT_AVAILABLE, &ready); + + if (!ready) + return false; + + bool result = false; + quint64 timestamps[2]; + rhiD->glGetQueryObjectui64v(tsStart, GL_QUERY_RESULT, ×tamps[0]); + rhiD->glGetQueryObjectui64v(tsEnd, GL_QUERY_RESULT, ×tamps[1]); + + if (timestamps[1] >= timestamps[0]) { + const quint64 nanoseconds = timestamps[1] - timestamps[0]; + *elapsedSec = nanoseconds / 1000000000.0; + result = true; + } + + active[pairIndex] = false; + return result; +} + QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhigles2_p.h b/src/gui/rhi/qrhigles2_p.h index e3f4fbe7d7..4139579864 100644 --- a/src/gui/rhi/qrhigles2_p.h +++ b/src/gui/rhi/qrhigles2_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHIGLES2_H -#define QRHIGLES2_H +#ifndef QRHIGLES2_P_H +#define QRHIGLES2_P_H // // W A R N I N G @@ -15,33 +15,1125 @@ // We mean it. // -#include <private/qrhi_p.h> -#include <QtGui/qsurfaceformat.h> +#include "qrhi_p.h" +#include <rhi/qshaderdescription.h> +#include <qopengl.h> +#include <QByteArray> +#include <QWindow> +#include <QPointer> +#include <QtCore/private/qduplicatetracker_p.h> +#include <optional> QT_BEGIN_NAMESPACE -class QOpenGLContext; -class QOffscreenSurface; -class QSurface; -class QWindow; +class QOpenGLExtensions; +class QRhiGles2; -struct Q_GUI_EXPORT QRhiGles2InitParams : public QRhiInitParams +struct QGles2Buffer : public QRhiBuffer { - QRhiGles2InitParams(); + QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QGles2Buffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; - QSurfaceFormat format; - QSurface *fallbackSurface = nullptr; - QWindow *window = nullptr; - QOpenGLContext *shareContext = nullptr; + quint32 nonZeroSize = 0; + GLuint buffer = 0; + GLenum targetForDataOps; + QByteArray data; + enum Access { + AccessNone, + AccessVertex, + AccessIndex, + AccessUniform, + AccessStorageRead, + AccessStorageWrite, + AccessStorageReadWrite, + AccessUpdate + }; + struct UsageState { + Access access; + }; + UsageState usageState; + friend class QRhiGles2; +}; + +struct QGles2RenderBuffer : public QRhiRenderBuffer +{ + QGles2RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QGles2RenderBuffer(); + void destroy() override; + bool create() override; + bool createFrom(NativeRenderBuffer src) override; + QRhiTexture::Format backingFormat() const override; + + GLuint renderbuffer = 0; + GLuint stencilRenderbuffer = 0; // when packed depth-stencil not supported + int samples; + bool owns = true; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2SamplerData +{ + GLenum glminfilter = 0; + GLenum glmagfilter = 0; + GLenum glwraps = 0; + GLenum glwrapt = 0; + GLenum glwrapr = 0; + GLenum gltexcomparefunc = 0; +}; + +inline bool operator==(const QGles2SamplerData &a, const QGles2SamplerData &b) +{ + return a.glminfilter == b.glminfilter + && a.glmagfilter == b.glmagfilter + && a.glwraps == b.glwraps + && a.glwrapt == b.glwrapt + && a.glwrapr == b.glwrapr + && a.gltexcomparefunc == b.gltexcomparefunc; +} + +inline bool operator!=(const QGles2SamplerData &a, const QGles2SamplerData &b) +{ + return !(a == b); +} + +struct QGles2Texture : public QRhiTexture +{ + QGles2Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QGles2Texture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + + GLuint texture = 0; + bool owns = true; + GLenum target; + GLenum glintformat; + GLenum glsizedintformat; + GLenum glformat; + GLenum gltype; + QGles2SamplerData samplerState; + bool specified = false; + bool zeroInitialized = false; + int mipLevelCount = 0; + + enum Access { + AccessNone, + AccessSample, + AccessFramebuffer, + AccessStorageRead, + AccessStorageWrite, + AccessStorageReadWrite, + AccessUpdate, + AccessRead + }; + struct UsageState { + Access access; + }; + UsageState usageState; + + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2Sampler : public QRhiSampler +{ + QGles2Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QGles2Sampler(); + void destroy() override; + bool create() override; + + QGles2SamplerData d; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2RenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QGles2RenderPassDescriptor(QRhiImplementation *rhi); + ~QGles2RenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QGles2RenderTargetData +{ + QGles2RenderTargetData(QRhiImplementation *) { } + + bool isValid() const { return rp != nullptr; } + + QGles2RenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + bool srgbUpdateAndBlend = false; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + std::optional<QRhiSwapChain::StereoTargetBuffer> stereoTarget; +}; + +struct QGles2SwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QGles2SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QGles2SwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; - static QOffscreenSurface *newFallbackSurface(const QSurfaceFormat &format = QSurfaceFormat::defaultFormat()); + QGles2RenderTargetData d; }; -struct Q_GUI_EXPORT QRhiGles2NativeHandles : public QRhiNativeHandles +struct QGles2TextureRenderTarget : public QRhiTextureRenderTarget { - QOpenGLContext *context = nullptr; + QGles2TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QGles2TextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QGles2RenderTargetData d; + GLuint framebuffer = 0; + GLuint nonMsaaThrowawayDepthTexture = 0; + friend class QRhiGles2; +}; + +struct QGles2ShaderResourceBindings : public QRhiShaderResourceBindings +{ + QGles2ShaderResourceBindings(QRhiImplementation *rhi); + ~QGles2ShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + bool hasDynamicOffset = false; + uint generation = 0; + friend class QRhiGles2; }; +struct QGles2UniformDescription +{ + QShaderDescription::VariableType type; + int glslLocation; + int binding; + quint32 offset; + quint32 size; + int arrayDim; +}; + +Q_DECLARE_TYPEINFO(QGles2UniformDescription, Q_RELOCATABLE_TYPE); + +struct QGles2SamplerDescription +{ + int glslLocation; + int combinedBinding; + int tbinding; + int sbinding; +}; + +Q_DECLARE_TYPEINFO(QGles2SamplerDescription, Q_RELOCATABLE_TYPE); + +using QGles2UniformDescriptionVector = QVarLengthArray<QGles2UniformDescription, 8>; +using QGles2SamplerDescriptionVector = QVarLengthArray<QGles2SamplerDescription, 4>; + +struct QGles2UniformState +{ + static constexpr int MAX_TRACKED_LOCATION = 1023; + int componentCount; + float v[4]; +}; + +struct QGles2GraphicsPipeline : public QRhiGraphicsPipeline +{ + QGles2GraphicsPipeline(QRhiImplementation *rhi); + ~QGles2GraphicsPipeline(); + void destroy() override; + bool create() override; + + GLuint program = 0; + GLenum drawMode = GL_TRIANGLES; + QGles2UniformDescriptionVector uniforms; + QGles2SamplerDescriptionVector samplers; + QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; + QRhiShaderResourceBindings *currentSrb = nullptr; + uint currentSrbGeneration = 0; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2ComputePipeline : public QRhiComputePipeline +{ + QGles2ComputePipeline(QRhiImplementation *rhi); + ~QGles2ComputePipeline(); + void destroy() override; + bool create() override; + + GLuint program = 0; + QGles2UniformDescriptionVector uniforms; + QGles2SamplerDescriptionVector samplers; + QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; + QRhiShaderResourceBindings *currentSrb = nullptr; + uint currentSrbGeneration = 0; + uint generation = 0; + friend class QRhiGles2; +}; + +struct QGles2CommandBuffer : public QRhiCommandBuffer +{ + QGles2CommandBuffer(QRhiImplementation *rhi); + ~QGles2CommandBuffer(); + void destroy() override; + + // keep at a reasonably low value otherwise sizeof Command explodes + static const int MAX_DYNAMIC_OFFSET_COUNT = 8; + + struct Command { + enum Cmd { + BeginFrame, + EndFrame, + ResetFrame, + Viewport, + Scissor, + BlendConstants, + StencilRef, + BindVertexBuffer, + BindIndexBuffer, + Draw, + DrawIndexed, + BindGraphicsPipeline, + BindShaderResources, + BindFramebuffer, + Clear, + BufferSubData, + GetBufferSubData, + CopyTex, + ReadPixels, + SubImage, + CompressedImage, + CompressedSubImage, + BlitFromRenderbuffer, + BlitFromTexture, + GenMip, + BindComputePipeline, + Dispatch, + BarriersForPass, + Barrier, + InvalidateFramebuffer + }; + Cmd cmd; + + // QRhi*/QGles2* references should be kept at minimum (so no + // QRhiTexture/Buffer/etc. pointers). + union Args { + struct { + GLuint timestampQuery; + } beginFrame; + struct { + GLuint timestampQuery; + } endFrame; + struct { + float x, y, w, h; + float d0, d1; + } viewport; + struct { + int x, y, w, h; + } scissor; + struct { + float r, g, b, a; + } blendConstants; + struct { + quint32 ref; + QRhiGraphicsPipeline *ps; + } stencilRef; + struct { + QRhiGraphicsPipeline *ps; + GLuint buffer; + quint32 offset; + int binding; + } bindVertexBuffer; + struct { + GLuint buffer; + quint32 offset; + GLenum type; + } bindIndexBuffer; + struct { + QRhiGraphicsPipeline *ps; + quint32 vertexCount; + quint32 firstVertex; + quint32 instanceCount; + quint32 baseInstance; + } draw; + struct { + QRhiGraphicsPipeline *ps; + quint32 indexCount; + quint32 firstIndex; + quint32 instanceCount; + quint32 baseInstance; + qint32 baseVertex; + } drawIndexed; + struct { + QRhiGraphicsPipeline *ps; + } bindGraphicsPipeline; + struct { + QRhiGraphicsPipeline *maybeGraphicsPs; + QRhiComputePipeline *maybeComputePs; + QRhiShaderResourceBindings *srb; + int dynamicOffsetCount; + uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offset + } bindShaderResources; + struct { + GLbitfield mask; + float c[4]; + float d; + quint32 s; + } clear; + struct { + GLuint fbo; + bool srgb; + int colorAttCount; + bool stereo; + QRhiSwapChain::StereoTargetBuffer stereoTarget; + } bindFramebuffer; + struct { + GLenum target; + GLuint buffer; + int offset; + int size; + const void *data; // must come from retainData() + } bufferSubData; + struct { + QRhiReadbackResult *result; + GLenum target; + GLuint buffer; + int offset; + int size; + } getBufferSubData; + struct { + GLenum srcTarget; + GLenum srcFaceTarget; + GLuint srcTexture; + int srcLevel; + int srcX; + int srcY; + int srcZ; + GLenum dstTarget; + GLuint dstTexture; + GLenum dstFaceTarget; + int dstLevel; + int dstX; + int dstY; + int dstZ; + int w; + int h; + } copyTex; + struct { + QRhiReadbackResult *result; + GLuint texture; + int w; + int h; + QRhiTexture::Format format; + GLenum readTarget; + int level; + int slice3D; + } readPixels; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + int dx; + int dy; + int dz; + int w; + int h; + GLenum glformat; + GLenum gltype; + int rowStartAlign; + int rowLength; + const void *data; // must come from retainImage() + } subImage; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + GLenum glintformat; + int w; + int h; + int depth; + int size; + const void *data; // must come from retainData() + } compressedImage; + struct { + GLenum target; + GLuint texture; + GLenum faceTarget; + int level; + int dx; + int dy; + int dz; + int w; + int h; + GLenum glintformat; + int size; + const void *data; // must come from retainData() + } compressedSubImage; + struct { + GLuint renderbuffer; + int w; + int h; + GLenum target; + GLuint dstTexture; + int dstLevel; + int dstLayer; + bool isDepthStencil; + } blitFromRenderbuffer; + struct { + GLenum srcTarget; + GLuint srcTexture; + int srcLevel; + int srcLayer; + int w; + int h; + GLenum dstTarget; + GLuint dstTexture; + int dstLevel; + int dstLayer; + bool isDepthStencil; + } blitFromTexture; + struct { + GLenum target; + GLuint texture; + } genMip; + struct { + QRhiComputePipeline *ps; + } bindComputePipeline; + struct { + GLuint x; + GLuint y; + GLuint z; + } dispatch; + struct { + int trackerIndex; + } barriersForPass; + struct { + GLbitfield barriers; + } barrier; + struct { + int attCount; + GLenum att[3]; + } invalidateFramebuffer; + } args; + }; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + QRhiBackendCommandList<Command> commands; + QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; + int currentPassResTrackerIndex; + + PassType recordingPass; + bool passNeedsResourceTracking; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + + struct GraphicsPassState { + bool valid = false; + bool scissor; + bool cullFace; + GLenum cullMode; + GLenum frontFace; + bool blendEnabled; + struct ColorMask { bool r, g, b, a; } colorMask; + struct Blend { + GLenum srcColor; + GLenum dstColor; + GLenum srcAlpha; + GLenum dstAlpha; + GLenum opColor; + GLenum opAlpha; + } blend; + bool depthTest; + bool depthWrite; + GLenum depthFunc; + bool stencilTest; + GLuint stencilReadMask; + GLuint stencilWriteMask; + struct StencilFace { + GLenum func; + GLenum failOp; + GLenum zfailOp; + GLenum zpassOp; + } stencil[2]; // front, back + bool polyOffsetFill; + float polyOffsetFactor; + float polyOffsetUnits; + float lineWidth; + int cpCount; + GLenum polygonMode; + void reset() { valid = false; } + struct { + // not part of QRhiGraphicsPipeline but used by setGraphicsPipeline() + GLint stencilRef = 0; + } dynamic; + } graphicsPassState; + + struct ComputePassState { + enum Access { + Read = 0x01, + Write = 0x02 + }; + QHash<QRhiResource *, QPair<int, bool> > writtenResources; + void reset() { + writtenResources.clear(); + } + } computePassState; + + struct TextureUnitState { + void *ps; + uint psGeneration; + uint texture; + } textureUnitState[16]; + + QVarLengthArray<QByteArray, 4> dataRetainPool; + QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; + QVarLengthArray<QImage, 4> imageRetainPool; + + // relies heavily on implicit sharing (no copies of the actual data will be made) + const void *retainData(const QByteArray &data) { + dataRetainPool.append(data); + return dataRetainPool.last().constData(); + } + const uchar *retainBufferData(const QRhiBufferData &data) { + bufferDataRetainPool.append(data); + return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); + } + const void *retainImage(const QImage &image) { + imageRetainPool.append(image); + return imageRetainPool.last().constBits(); + } + void resetCommands() { + commands.reset(); + dataRetainPool.clear(); + bufferDataRetainPool.clear(); + imageRetainPool.clear(); + + passResTrackers.clear(); + currentPassResTrackerIndex = -1; + } + void resetState() { + recordingPass = NoPass; + passNeedsResourceTracking = true; + // do not zero lastGpuTime + currentTarget = nullptr; + resetCommands(); + resetCachedState(); + } + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + graphicsPassState.reset(); + computePassState.reset(); + memset(textureUnitState, 0, sizeof(textureUnitState)); + } +}; + +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, + const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) +{ + return a.func == b.func + && a.failOp == b.failOp + && a.zfailOp == b.zfailOp + && a.zpassOp == b.zpassOp; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, + const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) +{ + return !(a == b); +} + +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, + const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) +{ + return a.r == b.r && a.g == b.g && a.b == b.b && a.a == b.a; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, + const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) +{ + return !(a == b); +} + +inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::Blend &a, + const QGles2CommandBuffer::GraphicsPassState::Blend &b) +{ + return a.srcColor == b.srcColor + && a.dstColor == b.dstColor + && a.srcAlpha == b.srcAlpha + && a.dstAlpha == b.dstAlpha + && a.opColor == b.opColor + && a.opAlpha == b.opAlpha; +} + +inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::Blend &a, + const QGles2CommandBuffer::GraphicsPassState::Blend &b) +{ + return !(a == b); +} + +struct QGles2SwapChainTimestamps +{ + static const int TIMESTAMP_PAIRS = 2; + + bool active[TIMESTAMP_PAIRS] = {}; + GLuint query[TIMESTAMP_PAIRS * 2] = {}; + + void prepare(QRhiGles2 *rhiD); + void destroy(QRhiGles2 *rhiD); + bool tryQueryTimestamps(int pairIndex, QRhiGles2 *rhiD, double *elapsedSec); +}; + +struct QGles2SwapChain : public QRhiSwapChain +{ + QGles2SwapChain(QRhiImplementation *rhi); + ~QGles2SwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + void initSwapChainRenderTarget(QGles2SwapChainRenderTarget *rt); + + QSurface *surface = nullptr; + QSize pixelSize; + QGles2SwapChainRenderTarget rt; + QGles2SwapChainRenderTarget rtLeft; + QGles2SwapChainRenderTarget rtRight; + QGles2CommandBuffer cb; + int frameCount = 0; + QGles2SwapChainTimestamps timestamps; + int currentTimestampPairIndex = 0; +}; + +class QRhiGles2 : public QRhiImplementation +{ +public: + QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *importDevice = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + bool ensureContext(QSurface *surface = nullptr) const; + QSurface *evaluateFallbackSurface() const; + void executeDeferredReleases(); + void trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *bufD, QGles2Buffer::Access access); + void trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Texture::Access access); + void enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cbD, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, + QGles2Buffer *bufD, + QRhiPassResourceTracker::BufferAccess access, + QRhiPassResourceTracker::BufferStage stage); + void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, + QGles2Texture *texD, + QRhiPassResourceTracker::TextureAccess access, + QRhiPassResourceTracker::TextureStage stage); + void executeCommandBuffer(QRhiCommandBuffer *cb); + void executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2GraphicsPipeline *psD); + void bindCombinedSampler(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Sampler *samplerD, + void *ps, uint psGeneration, int glslLocation, + int *texUnit, bool *activeTexUnitAltered); + void bindShaderResources(QGles2CommandBuffer *cbD, + QRhiGraphicsPipeline *maybeGraphicsPs, QRhiComputePipeline *maybeComputePs, + QRhiShaderResourceBindings *srb, + const uint *dynOfsPairs, int dynOfsCount); + QGles2RenderTargetData *enqueueBindFramebuffer(QRhiRenderTarget *rt, QGles2CommandBuffer *cbD, + bool *wantsColorClear = nullptr, bool *wantsDsClear = nullptr); + void enqueueBarriersForPass(QGles2CommandBuffer *cbD); + QByteArray shaderSource(const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); + bool compileShader(GLuint program, const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); + bool linkProgram(GLuint program); + void registerUniformIfActive(const QShaderDescription::BlockVariable &var, + const QByteArray &namePrefix, int binding, int baseOffset, + GLuint program, + QDuplicateTracker<int, 256> *activeUniformLocations, + QGles2UniformDescriptionVector *dst); + void gatherUniforms(GLuint program, const QShaderDescription::UniformBlock &ub, + QDuplicateTracker<int, 256> *activeUniformLocations, QGles2UniformDescriptionVector *dst); + void gatherSamplers(GLuint program, const QShaderDescription::InOutVariable &v, + QGles2SamplerDescriptionVector *dst); + void gatherGeneratedSamplers(GLuint program, + const QShader::SeparateToCombinedImageSamplerMapping &mapping, + QGles2SamplerDescriptionVector *dst); + void sanityCheckVertexFragmentInterface(const QShaderDescription &vsDesc, const QShaderDescription &fsDesc); + bool isProgramBinaryDiskCacheEnabled() const; + + enum ProgramCacheResult { + ProgramCacheHit, + ProgramCacheMiss, + ProgramCacheError + }; + ProgramCacheResult tryLoadFromDiskOrPipelineCache(const QRhiShaderStage *stages, + int stageCount, + GLuint program, + const QVector<QShaderDescription::InOutVariable> &inputVars, + QByteArray *cacheKey); + void trySaveToDiskCache(GLuint program, const QByteArray &cacheKey); + void trySaveToPipelineCache(GLuint program, const QByteArray &cacheKey, bool force = false); + + QRhi::Flags rhiFlags; + QOpenGLContext *ctx = nullptr; + bool importedContext = false; + QSurfaceFormat requestedFormat; + QSurface *fallbackSurface = nullptr; + QPointer<QWindow> maybeWindow = nullptr; + QOpenGLContext *maybeShareContext = nullptr; + mutable bool needsMakeCurrentDueToSwap = false; + QOpenGLExtensions *f = nullptr; + void (QOPENGLF_APIENTRYP glPolygonMode) (GLenum, GLenum) = nullptr; + void(QOPENGLF_APIENTRYP glTexImage1D)(GLenum, GLint, GLint, GLsizei, GLint, GLenum, GLenum, + const void *) = nullptr; + void(QOPENGLF_APIENTRYP glTexStorage1D)(GLenum, GLint, GLenum, GLsizei) = nullptr; + void(QOPENGLF_APIENTRYP glTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, GLenum, + const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glCopyTexSubImage1D)(GLenum, GLint, GLint, GLint, GLint, + GLsizei) = nullptr; + void(QOPENGLF_APIENTRYP glCompressedTexImage1D)(GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, + const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glCompressedTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, + GLsizei, const GLvoid *) = nullptr; + void(QOPENGLF_APIENTRYP glFramebufferTexture1D)(GLenum, GLenum, GLenum, GLuint, + GLint) = nullptr; + void(QOPENGLF_APIENTRYP glFramebufferTextureMultiviewOVR)(GLenum, GLenum, GLuint, GLint, + GLint, GLsizei) = nullptr; + void (QOPENGLF_APIENTRYP glQueryCounter)(GLuint, GLenum) = nullptr; + void (QOPENGLF_APIENTRYP glGetQueryObjectui64v)(GLuint, GLenum, quint64 *) = nullptr; + void (QOPENGLF_APIENTRYP glObjectLabel)(GLenum, GLuint, GLsizei, const GLchar *) = nullptr; + void (QOPENGLF_APIENTRYP glFramebufferTexture2DMultisampleEXT)(GLenum, GLenum, GLenum, GLuint, GLint, GLsizei) = nullptr; + void (QOPENGLF_APIENTRYP glFramebufferTextureMultisampleMultiviewOVR)(GLenum, GLenum, GLuint, GLint, GLsizei, GLint, GLsizei) = nullptr; + uint vao = 0; + struct Caps { + Caps() + : ctxMajor(2), + ctxMinor(0), + maxTextureSize(2048), + maxDrawBuffers(4), + maxSamples(16), + maxTextureArraySize(0), + maxThreadGroupsPerDimension(0), + maxThreadsPerThreadGroup(0), + maxThreadGroupsX(0), + maxThreadGroupsY(0), + maxThreadGroupsZ(0), + maxUniformVectors(4096), + maxVertexInputs(8), + maxVertexOutputs(8), + msaaRenderBuffer(false), + multisampledTexture(false), + npotTextureFull(true), + gles(false), + fixedIndexPrimitiveRestart(false), + bgraExternalFormat(false), + bgraInternalFormat(false), + r8Format(false), + r16Format(false), + floatFormats(false), + rgb10Formats(false), + depthTexture(false), + packedDepthStencil(false), + needsDepthStencilCombinedAttach(false), + srgbWriteControl(false), + coreProfile(false), + uniformBuffers(false), + elementIndexUint(false), + depth24(false), + rgba8Format(false), + instancing(false), + baseVertex(false), + compute(false), + textureCompareMode(false), + properMapBuffer(false), + nonBaseLevelFramebufferTexture(false), + texelFetch(false), + intAttributes(true), + screenSpaceDerivatives(false), + programBinary(false), + texture3D(false), + tessellation(false), + geometryShader(false), + texture1D(false), + hasDrawBuffersFunc(false), + halfAttributes(false), + multiView(false), + timestamps(false), + objectLabel(false), + glesMultisampleRenderToTexture(false), + glesMultiviewMultisampleRenderToTexture(false), + unpackRowLength(false) + { } + int ctxMajor; + int ctxMinor; + int maxTextureSize; + int maxDrawBuffers; + int maxSamples; + int maxTextureArraySize; + int maxThreadGroupsPerDimension; + int maxThreadsPerThreadGroup; + int maxThreadGroupsX; + int maxThreadGroupsY; + int maxThreadGroupsZ; + int maxUniformVectors; + int maxVertexInputs; + int maxVertexOutputs; + // Multisample fb and blit are supported (GLES 3.0 or OpenGL 3.x). Not + // the same as multisample textures! + uint msaaRenderBuffer : 1; + uint multisampledTexture : 1; + uint npotTextureFull : 1; + uint gles : 1; + uint fixedIndexPrimitiveRestart : 1; + uint bgraExternalFormat : 1; + uint bgraInternalFormat : 1; + uint r8Format : 1; + uint r16Format : 1; + uint floatFormats : 1; + uint rgb10Formats : 1; + uint depthTexture : 1; + uint packedDepthStencil : 1; + uint needsDepthStencilCombinedAttach : 1; + uint srgbWriteControl : 1; + uint coreProfile : 1; + uint uniformBuffers : 1; + uint elementIndexUint : 1; + uint depth24 : 1; + uint rgba8Format : 1; + uint instancing : 1; + uint baseVertex : 1; + uint compute : 1; + uint textureCompareMode : 1; + uint properMapBuffer : 1; + uint nonBaseLevelFramebufferTexture : 1; + uint texelFetch : 1; + uint intAttributes : 1; + uint screenSpaceDerivatives : 1; + uint programBinary : 1; + uint texture3D : 1; + uint tessellation : 1; + uint geometryShader : 1; + uint texture1D : 1; + uint hasDrawBuffersFunc : 1; + uint halfAttributes : 1; + uint multiView : 1; + uint timestamps : 1; + uint objectLabel : 1; + uint glesMultisampleRenderToTexture : 1; + uint glesMultiviewMultisampleRenderToTexture : 1; + uint unpackRowLength : 1; + } caps; + QGles2SwapChain *currentSwapChain = nullptr; + QSet<GLint> supportedCompressedFormats; + mutable QList<int> supportedSampleCountList; + QRhiGles2NativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + mutable bool contextLost = false; + + struct DeferredReleaseEntry { + enum Type { + Buffer, + Pipeline, + Texture, + RenderBuffer, + TextureRenderTarget + }; + Type type; + union { + struct { + GLuint buffer; + } buffer; + struct { + GLuint program; + } pipeline; + struct { + GLuint texture; + } texture; + struct { + GLuint renderbuffer; + GLuint renderbuffer2; + } renderbuffer; + struct { + GLuint framebuffer; + GLuint nonMsaaThrowawayDepthTexture; + } textureRenderTarget; + }; + }; + QList<DeferredReleaseEntry> releaseQueue; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } + bool active = false; + QGles2CommandBuffer cbWrapper; + GLuint tsQueries[2] = {}; + } ofr; + + QHash<QRhiShaderStage, uint> m_shaderCache; + + struct PipelineCacheData { + quint32 format; + QByteArray data; + }; + QHash<QByteArray, PipelineCacheData> m_pipelineCache; +}; + +Q_DECLARE_TYPEINFO(QRhiGles2::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhigles2_p_p.h b/src/gui/rhi/qrhigles2_p_p.h deleted file mode 100644 index c3d85c1d09..0000000000 --- a/src/gui/rhi/qrhigles2_p_p.h +++ /dev/null @@ -1,1076 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHIGLES2_P_H -#define QRHIGLES2_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhigles2_p.h" -#include "qrhi_p_p.h" -#include "qshaderdescription_p.h" -#include <qopengl.h> -#include <QByteArray> -#include <QWindow> -#include <QPointer> -#include <QtCore/private/qduplicatetracker_p.h> -#include <optional> - -QT_BEGIN_NAMESPACE - -class QOpenGLExtensions; - -struct QGles2Buffer : public QRhiBuffer -{ - QGles2Buffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QGles2Buffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - quint32 nonZeroSize = 0; - GLuint buffer = 0; - GLenum targetForDataOps; - QByteArray data; - enum Access { - AccessNone, - AccessVertex, - AccessIndex, - AccessUniform, - AccessStorageRead, - AccessStorageWrite, - AccessStorageReadWrite, - AccessUpdate - }; - struct UsageState { - Access access; - }; - UsageState usageState; - friend class QRhiGles2; -}; - -struct QGles2RenderBuffer : public QRhiRenderBuffer -{ - QGles2RenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QGles2RenderBuffer(); - void destroy() override; - bool create() override; - bool createFrom(NativeRenderBuffer src) override; - QRhiTexture::Format backingFormat() const override; - - GLuint renderbuffer = 0; - GLuint stencilRenderbuffer = 0; // when packed depth-stencil not supported - int samples; - bool owns = true; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2SamplerData -{ - GLenum glminfilter = 0; - GLenum glmagfilter = 0; - GLenum glwraps = 0; - GLenum glwrapt = 0; - GLenum glwrapr = 0; - GLenum gltexcomparefunc = 0; -}; - -inline bool operator==(const QGles2SamplerData &a, const QGles2SamplerData &b) -{ - return a.glminfilter == b.glminfilter - && a.glmagfilter == b.glmagfilter - && a.glwraps == b.glwraps - && a.glwrapt == b.glwrapt - && a.glwrapr == b.glwrapr - && a.gltexcomparefunc == b.gltexcomparefunc; -} - -inline bool operator!=(const QGles2SamplerData &a, const QGles2SamplerData &b) -{ - return !(a == b); -} - -struct QGles2Texture : public QRhiTexture -{ - QGles2Texture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QGles2Texture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - - GLuint texture = 0; - bool owns = true; - GLenum target; - GLenum glintformat; - GLenum glsizedintformat; - GLenum glformat; - GLenum gltype; - QGles2SamplerData samplerState; - bool specified = false; - bool zeroInitialized = false; - int mipLevelCount = 0; - - enum Access { - AccessNone, - AccessSample, - AccessFramebuffer, - AccessStorageRead, - AccessStorageWrite, - AccessStorageReadWrite, - AccessUpdate, - AccessRead - }; - struct UsageState { - Access access; - }; - UsageState usageState; - - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2Sampler : public QRhiSampler -{ - QGles2Sampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QGles2Sampler(); - void destroy() override; - bool create() override; - - QGles2SamplerData d; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2RenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QGles2RenderPassDescriptor(QRhiImplementation *rhi); - ~QGles2RenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; -}; - -struct QGles2RenderTargetData -{ - QGles2RenderTargetData(QRhiImplementation *) { } - - bool isValid() const { return rp != nullptr; } - - QGles2RenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - bool srgbUpdateAndBlend = false; - QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; - std::optional<QRhiSwapChain::StereoTargetBuffer> stereoTarget; -}; - -struct QGles2SwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QGles2SwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QGles2SwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QGles2RenderTargetData d; -}; - -struct QGles2TextureRenderTarget : public QRhiTextureRenderTarget -{ - QGles2TextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QGles2TextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QGles2RenderTargetData d; - GLuint framebuffer = 0; - friend class QRhiGles2; -}; - -struct QGles2ShaderResourceBindings : public QRhiShaderResourceBindings -{ - QGles2ShaderResourceBindings(QRhiImplementation *rhi); - ~QGles2ShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; - - bool hasDynamicOffset = false; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2UniformDescription -{ - QShaderDescription::VariableType type; - int glslLocation; - int binding; - quint32 offset; - quint32 size; - int arrayDim; -}; - -Q_DECLARE_TYPEINFO(QGles2UniformDescription, Q_RELOCATABLE_TYPE); - -struct QGles2SamplerDescription -{ - int glslLocation; - int combinedBinding; - int tbinding; - int sbinding; -}; - -Q_DECLARE_TYPEINFO(QGles2SamplerDescription, Q_RELOCATABLE_TYPE); - -using QGles2UniformDescriptionVector = QVarLengthArray<QGles2UniformDescription, 8>; -using QGles2SamplerDescriptionVector = QVarLengthArray<QGles2SamplerDescription, 4>; - -struct QGles2UniformState -{ - static constexpr int MAX_TRACKED_LOCATION = 1023; - int componentCount; - float v[4]; -}; - -struct QGles2GraphicsPipeline : public QRhiGraphicsPipeline -{ - QGles2GraphicsPipeline(QRhiImplementation *rhi); - ~QGles2GraphicsPipeline(); - void destroy() override; - bool create() override; - - GLuint program = 0; - GLenum drawMode = GL_TRIANGLES; - QGles2UniformDescriptionVector uniforms; - QGles2SamplerDescriptionVector samplers; - QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; - QRhiShaderResourceBindings *currentSrb = nullptr; - uint currentSrbGeneration = 0; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2ComputePipeline : public QRhiComputePipeline -{ - QGles2ComputePipeline(QRhiImplementation *rhi); - ~QGles2ComputePipeline(); - void destroy() override; - bool create() override; - - GLuint program = 0; - QGles2UniformDescriptionVector uniforms; - QGles2SamplerDescriptionVector samplers; - QGles2UniformState uniformState[QGles2UniformState::MAX_TRACKED_LOCATION + 1]; - QRhiShaderResourceBindings *currentSrb = nullptr; - uint currentSrbGeneration = 0; - uint generation = 0; - friend class QRhiGles2; -}; - -struct QGles2CommandBuffer : public QRhiCommandBuffer -{ - QGles2CommandBuffer(QRhiImplementation *rhi); - ~QGles2CommandBuffer(); - void destroy() override; - - // keep at a reasonably low value otherwise sizeof Command explodes - static const int MAX_DYNAMIC_OFFSET_COUNT = 8; - - struct Command { - enum Cmd { - BeginFrame, - EndFrame, - ResetFrame, - Viewport, - Scissor, - BlendConstants, - StencilRef, - BindVertexBuffer, - BindIndexBuffer, - Draw, - DrawIndexed, - BindGraphicsPipeline, - BindShaderResources, - BindFramebuffer, - Clear, - BufferSubData, - GetBufferSubData, - CopyTex, - ReadPixels, - SubImage, - CompressedImage, - CompressedSubImage, - BlitFromRenderbuffer, - GenMip, - BindComputePipeline, - Dispatch, - BarriersForPass, - Barrier - }; - Cmd cmd; - - // QRhi*/QGles2* references should be kept at minimum (so no - // QRhiTexture/Buffer/etc. pointers). - union Args { - struct { - float x, y, w, h; - float d0, d1; - } viewport; - struct { - int x, y, w, h; - } scissor; - struct { - float r, g, b, a; - } blendConstants; - struct { - quint32 ref; - QRhiGraphicsPipeline *ps; - } stencilRef; - struct { - QRhiGraphicsPipeline *ps; - GLuint buffer; - quint32 offset; - int binding; - } bindVertexBuffer; - struct { - GLuint buffer; - quint32 offset; - GLenum type; - } bindIndexBuffer; - struct { - QRhiGraphicsPipeline *ps; - quint32 vertexCount; - quint32 firstVertex; - quint32 instanceCount; - quint32 baseInstance; - } draw; - struct { - QRhiGraphicsPipeline *ps; - quint32 indexCount; - quint32 firstIndex; - quint32 instanceCount; - quint32 baseInstance; - qint32 baseVertex; - } drawIndexed; - struct { - QRhiGraphicsPipeline *ps; - } bindGraphicsPipeline; - struct { - QRhiGraphicsPipeline *maybeGraphicsPs; - QRhiComputePipeline *maybeComputePs; - QRhiShaderResourceBindings *srb; - int dynamicOffsetCount; - uint dynamicOffsetPairs[MAX_DYNAMIC_OFFSET_COUNT * 2]; // binding, offset - } bindShaderResources; - struct { - GLbitfield mask; - float c[4]; - float d; - quint32 s; - } clear; - struct { - GLuint fbo; - bool srgb; - int colorAttCount; - bool stereo; - QRhiSwapChain::StereoTargetBuffer stereoTarget; - } bindFramebuffer; - struct { - GLenum target; - GLuint buffer; - int offset; - int size; - const void *data; // must come from retainData() - } bufferSubData; - struct { - QRhiBufferReadbackResult *result; - GLenum target; - GLuint buffer; - int offset; - int size; - } getBufferSubData; - struct { - GLenum srcTarget; - GLenum srcFaceTarget; - GLuint srcTexture; - int srcLevel; - int srcX; - int srcY; - int srcZ; - GLenum dstTarget; - GLuint dstTexture; - GLenum dstFaceTarget; - int dstLevel; - int dstX; - int dstY; - int dstZ; - int w; - int h; - } copyTex; - struct { - QRhiReadbackResult *result; - GLuint texture; - int w; - int h; - QRhiTexture::Format format; - GLenum readTarget; - int level; - int slice3D; - } readPixels; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - int dx; - int dy; - int dz; - int w; - int h; - GLenum glformat; - GLenum gltype; - int rowStartAlign; - int rowLength; - const void *data; // must come from retainImage() - } subImage; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - GLenum glintformat; - int w; - int h; - int depth; - int size; - const void *data; // must come from retainData() - } compressedImage; - struct { - GLenum target; - GLuint texture; - GLenum faceTarget; - int level; - int dx; - int dy; - int dz; - int w; - int h; - GLenum glintformat; - int size; - const void *data; // must come from retainData() - } compressedSubImage; - struct { - GLuint renderbuffer; - int w; - int h; - GLenum target; - GLuint texture; - int dstLevel; - int dstLayer; - } blitFromRb; - struct { - GLenum target; - GLuint texture; - } genMip; - struct { - QRhiComputePipeline *ps; - } bindComputePipeline; - struct { - GLuint x; - GLuint y; - GLuint z; - } dispatch; - struct { - int trackerIndex; - } barriersForPass; - struct { - GLbitfield barriers; - } barrier; - } args; - }; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - QRhiBackendCommandList<Command> commands; - QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; - int currentPassResTrackerIndex; - - PassType recordingPass; - bool passNeedsResourceTracking; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - - struct GraphicsPassState { - bool valid = false; - bool scissor; - bool cullFace; - GLenum cullMode; - GLenum frontFace; - bool blendEnabled; - struct ColorMask { bool r, g, b, a; } colorMask; - struct Blend { - GLenum srcColor; - GLenum dstColor; - GLenum srcAlpha; - GLenum dstAlpha; - GLenum opColor; - GLenum opAlpha; - } blend; - bool depthTest; - bool depthWrite; - GLenum depthFunc; - bool stencilTest; - GLuint stencilReadMask; - GLuint stencilWriteMask; - struct StencilFace { - GLenum func; - GLenum failOp; - GLenum zfailOp; - GLenum zpassOp; - } stencil[2]; // front, back - bool polyOffsetFill; - float polyOffsetFactor; - float polyOffsetUnits; - float lineWidth; - int cpCount; - GLenum polygonMode; - void reset() { valid = false; } - struct { - // not part of QRhiGraphicsPipeline but used by setGraphicsPipeline() - GLint stencilRef = 0; - } dynamic; - } graphicsPassState; - - struct ComputePassState { - enum Access { - Read = 0x01, - Write = 0x02 - }; - QHash<QRhiResource *, QPair<int, bool> > writtenResources; - void reset() { - writtenResources.clear(); - } - } computePassState; - - struct TextureUnitState { - void *ps; - uint psGeneration; - uint texture; - } textureUnitState[16]; - - QVarLengthArray<QByteArray, 4> dataRetainPool; - QVarLengthArray<QRhiBufferData, 4> bufferDataRetainPool; - QVarLengthArray<QImage, 4> imageRetainPool; - - // relies heavily on implicit sharing (no copies of the actual data will be made) - const void *retainData(const QByteArray &data) { - dataRetainPool.append(data); - return dataRetainPool.last().constData(); - } - const uchar *retainBufferData(const QRhiBufferData &data) { - bufferDataRetainPool.append(data); - return reinterpret_cast<const uchar *>(bufferDataRetainPool.last().constData()); - } - const void *retainImage(const QImage &image) { - imageRetainPool.append(image); - return imageRetainPool.last().constBits(); - } - void resetCommands() { - commands.reset(); - dataRetainPool.clear(); - bufferDataRetainPool.clear(); - imageRetainPool.clear(); - - passResTrackers.clear(); - currentPassResTrackerIndex = -1; - } - void resetState() { - recordingPass = NoPass; - passNeedsResourceTracking = true; - currentTarget = nullptr; - resetCommands(); - resetCachedState(); - } - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - graphicsPassState.reset(); - computePassState.reset(); - memset(textureUnitState, 0, sizeof(textureUnitState)); - } -}; - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, - const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) -{ - return a.func == b.func - && a.failOp == b.failOp - && a.zfailOp == b.zfailOp - && a.zpassOp == b.zpassOp; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::StencilFace &a, - const QGles2CommandBuffer::GraphicsPassState::StencilFace &b) -{ - return !(a == b); -} - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, - const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) -{ - return a.r == b.r && a.g == b.g && a.b == b.b && a.a == b.a; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::ColorMask &a, - const QGles2CommandBuffer::GraphicsPassState::ColorMask &b) -{ - return !(a == b); -} - -inline bool operator==(const QGles2CommandBuffer::GraphicsPassState::Blend &a, - const QGles2CommandBuffer::GraphicsPassState::Blend &b) -{ - return a.srcColor == b.srcColor - && a.dstColor == b.dstColor - && a.srcAlpha == b.srcAlpha - && a.dstAlpha == b.dstAlpha - && a.opColor == b.opColor - && a.opAlpha == b.opAlpha; -} - -inline bool operator!=(const QGles2CommandBuffer::GraphicsPassState::Blend &a, - const QGles2CommandBuffer::GraphicsPassState::Blend &b) -{ - return !(a == b); -} - -struct QGles2SwapChain : public QRhiSwapChain -{ - QGles2SwapChain(QRhiImplementation *rhi); - ~QGles2SwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; - - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - void initSwapChainRenderTarget(QGles2SwapChainRenderTarget *rt); - - QSurface *surface = nullptr; - QSize pixelSize; - QGles2SwapChainRenderTarget rt; - QGles2SwapChainRenderTarget rtLeft; - QGles2SwapChainRenderTarget rtRight; - QGles2CommandBuffer cb; - int frameCount = 0; -}; - -class QRhiGles2 : public QRhiImplementation -{ -public: - QRhiGles2(QRhiGles2InitParams *params, QRhiGles2NativeHandles *importDevice = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - bool ensureContext(QSurface *surface = nullptr) const; - QSurface *evaluateFallbackSurface() const; - void executeDeferredReleases(); - void trackedBufferBarrier(QGles2CommandBuffer *cbD, QGles2Buffer *bufD, QGles2Buffer::Access access); - void trackedImageBarrier(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Texture::Access access); - void enqueueSubresUpload(QGles2Texture *texD, QGles2CommandBuffer *cbD, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, - QGles2Buffer *bufD, - QRhiPassResourceTracker::BufferAccess access, - QRhiPassResourceTracker::BufferStage stage); - void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, - QGles2Texture *texD, - QRhiPassResourceTracker::TextureAccess access, - QRhiPassResourceTracker::TextureStage stage); - void executeCommandBuffer(QRhiCommandBuffer *cb); - void executeBindGraphicsPipeline(QGles2CommandBuffer *cbD, QGles2GraphicsPipeline *psD); - void bindCombinedSampler(QGles2CommandBuffer *cbD, QGles2Texture *texD, QGles2Sampler *samplerD, - void *ps, uint psGeneration, int glslLocation, - int *texUnit, bool *activeTexUnitAltered); - void bindShaderResources(QGles2CommandBuffer *cbD, - QRhiGraphicsPipeline *maybeGraphicsPs, QRhiComputePipeline *maybeComputePs, - QRhiShaderResourceBindings *srb, - const uint *dynOfsPairs, int dynOfsCount); - QGles2RenderTargetData *enqueueBindFramebuffer(QRhiRenderTarget *rt, QGles2CommandBuffer *cbD, - bool *wantsColorClear = nullptr, bool *wantsDsClear = nullptr); - void enqueueBarriersForPass(QGles2CommandBuffer *cbD); - int effectiveSampleCount(int sampleCount) const; - QByteArray shaderSource(const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); - bool compileShader(GLuint program, const QRhiShaderStage &shaderStage, QShaderVersion *shaderVersion); - bool linkProgram(GLuint program); - void registerUniformIfActive(const QShaderDescription::BlockVariable &var, - const QByteArray &namePrefix, int binding, int baseOffset, - GLuint program, - QDuplicateTracker<int, 256> *activeUniformLocations, - QGles2UniformDescriptionVector *dst); - void gatherUniforms(GLuint program, const QShaderDescription::UniformBlock &ub, - QDuplicateTracker<int, 256> *activeUniformLocations, QGles2UniformDescriptionVector *dst); - void gatherSamplers(GLuint program, const QShaderDescription::InOutVariable &v, - QGles2SamplerDescriptionVector *dst); - void gatherGeneratedSamplers(GLuint program, - const QShader::SeparateToCombinedImageSamplerMapping &mapping, - QGles2SamplerDescriptionVector *dst); - void sanityCheckVertexFragmentInterface(const QShaderDescription &vsDesc, const QShaderDescription &fsDesc); - bool isProgramBinaryDiskCacheEnabled() const; - - enum ProgramCacheResult { - ProgramCacheHit, - ProgramCacheMiss, - ProgramCacheError - }; - ProgramCacheResult tryLoadFromDiskOrPipelineCache(const QRhiShaderStage *stages, - int stageCount, - GLuint program, - const QVector<QShaderDescription::InOutVariable> &inputVars, - QByteArray *cacheKey); - void trySaveToDiskCache(GLuint program, const QByteArray &cacheKey); - void trySaveToPipelineCache(GLuint program, const QByteArray &cacheKey, bool force = false); - - QRhi::Flags rhiFlags; - QOpenGLContext *ctx = nullptr; - bool importedContext = false; - QSurfaceFormat requestedFormat; - QSurface *fallbackSurface = nullptr; - QPointer<QWindow> maybeWindow = nullptr; - QOpenGLContext *maybeShareContext = nullptr; - mutable bool needsMakeCurrentDueToSwap = false; - QOpenGLExtensions *f = nullptr; - void (QOPENGLF_APIENTRYP glPolygonMode) (GLenum, GLenum) = nullptr; - void(QOPENGLF_APIENTRYP glTexImage1D)(GLenum, GLint, GLint, GLsizei, GLint, GLenum, GLenum, - const void *) = nullptr; - void(QOPENGLF_APIENTRYP glTexStorage1D)(GLenum, GLint, GLenum, GLsizei) = nullptr; - void(QOPENGLF_APIENTRYP glTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, GLenum, - const GLvoid *) = nullptr; - void(QOPENGLF_APIENTRYP glCopyTexSubImage1D)(GLenum, GLint, GLint, GLint, GLint, - GLsizei) = nullptr; - void(QOPENGLF_APIENTRYP glCompressedTexImage1D)(GLenum, GLint, GLenum, GLsizei, GLint, GLsizei, - const GLvoid *) = nullptr; - void(QOPENGLF_APIENTRYP glCompressedTexSubImage1D)(GLenum, GLint, GLint, GLsizei, GLenum, - GLsizei, const GLvoid *) = nullptr; - void(QOPENGLF_APIENTRYP glFramebufferTexture1D)(GLenum, GLenum, GLenum, GLuint, - GLint) = nullptr; - - uint vao = 0; - struct Caps { - Caps() - : ctxMajor(2), - ctxMinor(0), - maxTextureSize(2048), - maxDrawBuffers(4), - maxSamples(16), - maxTextureArraySize(0), - maxThreadGroupsPerDimension(0), - maxThreadsPerThreadGroup(0), - maxThreadGroupsX(0), - maxThreadGroupsY(0), - maxThreadGroupsZ(0), - maxUniformVectors(4096), - maxVertexInputs(8), - maxVertexOutputs(8), - msaaRenderBuffer(false), - multisampledTexture(false), - npotTextureFull(true), - gles(false), - fixedIndexPrimitiveRestart(false), - bgraExternalFormat(false), - bgraInternalFormat(false), - r8Format(false), - r16Format(false), - floatFormats(false), - rgb10Formats(false), - depthTexture(false), - packedDepthStencil(false), - needsDepthStencilCombinedAttach(false), - srgbCapableDefaultFramebuffer(false), - coreProfile(false), - uniformBuffers(false), - elementIndexUint(false), - depth24(false), - rgba8Format(false), - instancing(false), - baseVertex(false), - compute(false), - textureCompareMode(false), - properMapBuffer(false), - nonBaseLevelFramebufferTexture(false), - texelFetch(false), - intAttributes(true), - screenSpaceDerivatives(false), - programBinary(false), - texture3D(false), - tessellation(false), - geometryShader(false), - texture1D(false), - hasDrawBuffersFunc(false), - halfAttributes(false) - { } - int ctxMajor; - int ctxMinor; - int maxTextureSize; - int maxDrawBuffers; - int maxSamples; - int maxTextureArraySize; - int maxThreadGroupsPerDimension; - int maxThreadsPerThreadGroup; - int maxThreadGroupsX; - int maxThreadGroupsY; - int maxThreadGroupsZ; - int maxUniformVectors; - int maxVertexInputs; - int maxVertexOutputs; - // Multisample fb and blit are supported (GLES 3.0 or OpenGL 3.x). Not - // the same as multisample textures! - uint msaaRenderBuffer : 1; - uint multisampledTexture : 1; - uint npotTextureFull : 1; - uint gles : 1; - uint fixedIndexPrimitiveRestart : 1; - uint bgraExternalFormat : 1; - uint bgraInternalFormat : 1; - uint r8Format : 1; - uint r16Format : 1; - uint floatFormats : 1; - uint rgb10Formats : 1; - uint depthTexture : 1; - uint packedDepthStencil : 1; - uint needsDepthStencilCombinedAttach : 1; - uint srgbCapableDefaultFramebuffer : 1; - uint coreProfile : 1; - uint uniformBuffers : 1; - uint elementIndexUint : 1; - uint depth24 : 1; - uint rgba8Format : 1; - uint instancing : 1; - uint baseVertex : 1; - uint compute : 1; - uint textureCompareMode : 1; - uint properMapBuffer : 1; - uint nonBaseLevelFramebufferTexture : 1; - uint texelFetch : 1; - uint intAttributes : 1; - uint screenSpaceDerivatives : 1; - uint programBinary : 1; - uint texture3D : 1; - uint tessellation : 1; - uint geometryShader : 1; - uint texture1D : 1; - uint hasDrawBuffersFunc : 1; - uint halfAttributes : 1; - } caps; - QGles2SwapChain *currentSwapChain = nullptr; - QSet<GLint> supportedCompressedFormats; - mutable QList<int> supportedSampleCountList; - QRhiGles2NativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - mutable bool contextLost = false; - - struct DeferredReleaseEntry { - enum Type { - Buffer, - Pipeline, - Texture, - RenderBuffer, - TextureRenderTarget - }; - Type type; - union { - struct { - GLuint buffer; - } buffer; - struct { - GLuint program; - } pipeline; - struct { - GLuint texture; - } texture; - struct { - GLuint renderbuffer; - GLuint renderbuffer2; - } renderbuffer; - struct { - GLuint framebuffer; - } textureRenderTarget; - }; - }; - QList<DeferredReleaseEntry> releaseQueue; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } - bool active = false; - QGles2CommandBuffer cbWrapper; - } ofr; - - QHash<QRhiShaderStage, uint> m_shaderCache; - - struct PipelineCacheData { - quint32 format; - QByteArray data; - }; - QHash<QByteArray, PipelineCacheData> m_pipelineCache; -}; - -Q_DECLARE_TYPEINFO(QRhiGles2::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhimetal.mm b/src/gui/rhi/qrhimetal.mm index 196016d9ca..887d40c7a5 100644 --- a/src/gui/rhi/qrhimetal.mm +++ b/src/gui/rhi/qrhimetal.mm @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhimetal_p_p.h" -#include <QtGui/private/qshader_p_p.h> +#include "qrhimetal_p.h" +#include "qshader_p.h" #include <QGuiApplication> #include <QWindow> #include <QUrl> @@ -13,6 +13,7 @@ #include <QOperatingSystemVersion> #include <QtCore/private/qcore_mac_p.h> +#include <QtGui/private/qmetallayer_p.h> #ifdef Q_OS_MACOS #include <AppKit/AppKit.h> @@ -20,8 +21,9 @@ #include <UIKit/UIKit.h> #endif +#include <QuartzCore/CATransaction.h> + #include <Metal/Metal.h> -#include <QuartzCore/CAMetalLayer.h> QT_BEGIN_NAMESPACE @@ -40,19 +42,28 @@ QT_BEGIN_NAMESPACE #error ARC not supported #endif -// Note: we expect everything here pass the Metal API validation when running -// in Debug mode in XCode (or with METAL_DEVICE_WRAPPER_TYPE=1). An exception -// is the nextDrawable Called Early blah blah warning, which is plain and -// simply false. This may not be present with newer XCode. There may also be -// warnings about threading (e.g. about accessing view.layer), those are -// expected for now. +// Even though the macOS 13 MTLBinaryArchive problem (QTBUG-106703) seems +// to be solved in later 13.x releases, we have reports from old Intel hardware +// and older macOS versions where this causes problems (QTBUG-114338). +// Thus we no longer do OS version based differentiation, but rather have a +// single toggle that is currently on, and so QRhi::(set)pipelineCache() +// does nothing with Metal. +#define QRHI_METAL_DISABLE_BINARY_ARCHIVE + +// We should be able to operate with command buffers that do not automatically +// retain/release the resources used by them. (since we have logic that mirrors +// other backends such as the Vulkan one anyway) +#define QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES /*! \class QRhiMetalInitParams \inmodule QtRhi - \internal + \since 6.6 \brief Metal specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Metal-based QRhi needs no special parameters for initialization. \badcode @@ -85,14 +96,30 @@ QT_BEGIN_NAMESPACE /*! \class QRhiMetalNativeHandles \inmodule QtRhi - \internal + \since 6.6 \brief Holds the Metal device used by the QRhi. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiMetalNativeHandles::dev + + Set to a valid MTLDevice to import an existing device. +*/ + +/*! + \variable QRhiMetalNativeHandles::cmdQueue + + Set to a valid MTLCommandQueue when importing an existing command queue. + When \nullptr, QRhi will create a new command queue. +*/ + +/*! \class QRhiMetalCommandBufferNativeHandles \inmodule QtRhi - \internal + \since 6.6 \brief Holds the MTLCommandBuffer and MTLRenderCommandEncoder objects that are backing a QRhiCommandBuffer. \note The command buffer object is only guaranteed to be valid while @@ -104,8 +131,19 @@ QT_BEGIN_NAMESPACE \note The command encoder is only valid while recording a pass, that is, between \l{QRhiCommandBuffer::beginPass()} - \l{QRhiCommandBuffer::endPass()}. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiMetalCommandBufferNativeHandles::commandBuffer +*/ + +/*! + \variable QRhiMetalCommandBufferNativeHandles::encoder +*/ + struct QMetalShader { id<MTLLibrary> lib = nil; @@ -133,8 +171,8 @@ struct QRhiMetalData id<MTLDevice> dev = nil; id<MTLCommandQueue> cmdQueue = nil; API_AVAILABLE(macosx(11.0), ios(14.0)) id<MTLBinaryArchive> binArch = nil; - bool binArchWasEmpty = false; + id<MTLCommandBuffer> newCommandBuffer(); MTLRenderPassDescriptor *createDefaultRenderPass(bool hasDepthStencil, const QColor &colorClearValue, const QRhiDepthStencilClearValue &depthStencilClearValue, @@ -194,6 +232,7 @@ struct QRhiMetalData struct OffscreenFrame { OffscreenFrame(QRhiImplementation *rhi) : cbWrapper(rhi) { } bool active = false; + double lastGpuTime = 0; QMetalCommandBuffer cbWrapper; } ofr; @@ -211,7 +250,7 @@ struct QRhiMetalData struct BufferReadback { int activeFrameSlot = -1; - QRhiBufferReadbackResult *result; + QRhiReadbackResult *result; quint32 offset; quint32 readSize; id<MTLBuffer> buf; @@ -296,6 +335,7 @@ struct QMetalShaderResourceBindingsData { struct QMetalCommandBufferData { id<MTLCommandBuffer> cb; + double lastGpuTime = 0; id<MTLRenderCommandEncoder> currentRenderPassEncoder; id<MTLComputeCommandEncoder> currentComputePassEncoder; id<MTLComputeCommandEncoder> tessellationComputeEncoder; @@ -330,8 +370,11 @@ struct QMetalRenderTargetData struct { ColorAtt colorAtt[QMetalRenderPassDescriptor::MAX_COLOR_ATTACHMENTS]; id<MTLTexture> dsTex = nil; + id<MTLTexture> dsResolveTex = nil; bool hasStencil = false; bool depthNeedsStore = false; + bool preserveColor = false; + bool preserveDs = false; } fb; QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; @@ -350,6 +393,15 @@ struct QMetalGraphicsPipelineData float slopeScaledDepthBias; QMetalShader vs; QMetalShader fs; + struct ExtraBufferManager { + enum class WorkBufType { + DeviceLocal, + HostVisible + }; + QMetalBuffer *acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type = WorkBufType::DeviceLocal); + QVector<QMetalBuffer *> deviceLocalWorkBuffers; + QVector<QMetalBuffer *> hostVisibleWorkBuffers; + } extraBufMgr; struct Tessellation { QMetalGraphicsPipelineData *q = nullptr; bool enabled = false; @@ -383,13 +435,6 @@ struct QMetalGraphicsPipelineData id<MTLComputePipelineState> vsCompPipeline(QRhiMetal *rhiD, QShader::Variant vertexCompVariant); id<MTLComputePipelineState> tescCompPipeline(QRhiMetal *rhiD); id<MTLRenderPipelineState> teseFragRenderPipeline(QRhiMetal *rhiD, QMetalGraphicsPipeline *pipeline); - enum class WorkBufType { - DeviceLocal, - HostVisible - }; - QMetalBuffer *acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type = WorkBufType::DeviceLocal); - QVector<QMetalBuffer *> deviceLocalWorkBuffers; - QVector<QMetalBuffer *> hostVisibleWorkBuffers; } tess; void setupVertexInputDescriptor(MTLVertexDescriptor *desc); void setupStageInputDescriptor(MTLStageInputOutputDescriptor *desc); @@ -413,15 +458,11 @@ struct QMetalSwapChainData CAMetalLayer *layer = nullptr; id<CAMetalDrawable> curDrawable = nil; dispatch_semaphore_t sem[QMTL_FRAMES_IN_FLIGHT]; + double lastGpuTime[QMTL_FRAMES_IN_FLIGHT]; MTLRenderPassDescriptor *rp = nullptr; id<MTLTexture> msaaTex[QMTL_FRAMES_IN_FLIGHT]; QRhiTexture::Format rhiColorFormat; MTLPixelFormat colorFormat; -#ifdef Q_OS_MACOS - bool liveResizeObserverSet = false; - QMacNotificationObserver liveResizeStartObserver; - QMacNotificationObserver liveResizeEndObserver; -#endif }; QRhiMetal::QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice) @@ -432,7 +473,7 @@ QRhiMetal::QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *import importedDevice = importDevice != nullptr; if (importedDevice) { - if (d->dev) { + if (importDevice->dev) { d->dev = (id<MTLDevice>) importDevice->dev; importedCmdQueue = importDevice->cmdQueue != nullptr; if (importedCmdQueue) @@ -466,8 +507,24 @@ bool QRhiMetal::probe(QRhiMetalInitParams *params) return false; } +id<MTLCommandBuffer> QRhiMetalData::newCommandBuffer() +{ +#ifdef QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES + // Do not let the command buffer mess with the refcount of objects. We do + // have a proper render loop and will manage lifetimes similarly to other + // backends (Vulkan). + return [cmdQueue commandBufferWithUnretainedReferences]; +#else + return [cmdQueue commandBuffer]; +#endif +} + bool QRhiMetalData::setupBinaryArchive(NSURL *sourceFileUrl) { +#ifdef QRHI_METAL_DISABLE_BINARY_ARCHIVE + return false; +#endif + if (@available(macOS 11.0, iOS 14.0, *)) { [binArch release]; MTLBinaryArchiveDescriptor *binArchDesc = [MTLBinaryArchiveDescriptor new]; @@ -480,7 +537,6 @@ bool QRhiMetalData::setupBinaryArchive(NSURL *sourceFileUrl) qWarning("newBinaryArchiveWithDescriptor failed: %s", qPrintable(msg)); return false; } - binArchWasEmpty = sourceFileUrl == nil; return true; } return false; @@ -508,9 +564,7 @@ bool QRhiMetal::create(QRhi::Flags flags) // suitable as deviceId because it does not seem stable on macOS and can // apparently change when the system is rebooted. -#ifdef Q_OS_IOS - driverInfoStruct.deviceType = QRhiDriverInfo::IntegratedDevice; -#else +#ifdef Q_OS_MACOS if (@available(macOS 10.15, *)) { const MTLDeviceLocation deviceLocation = [d->dev location]; switch (deviceLocation) { @@ -527,6 +581,8 @@ bool QRhiMetal::create(QRhi::Flags flags) break; } } +#else + driverInfoStruct.deviceType = QRhiDriverInfo::IntegratedDevice; #endif const QOperatingSystemVersion ver = QOperatingSystemVersion::current(); @@ -552,6 +608,7 @@ bool QRhiMetal::create(QRhi::Flags flags) if (@available(macOS 10.15, *)) caps.isAppleGPU = [d->dev supportsFamily:MTLGPUFamilyApple7]; caps.maxThreadGroupSize = 1024; + caps.multiView = true; #elif defined(Q_OS_TVOS) if ([d->dev supportsFeatureSet: MTLFeatureSet(30003)]) // MTLFeatureSet_tvOS_GPUFamily2_v1 caps.maxTextureSize = 16384; @@ -578,8 +635,10 @@ bool QRhiMetal::create(QRhi::Flags flags) } caps.isAppleGPU = true; if (@available(iOS 13, *)) { - if ([d->dev supportsFamily:MTLGPUFamilyApple4]) + if ([d->dev supportsFamily: MTLGPUFamilyApple4]) caps.maxThreadGroupSize = 1024; + if ([d->dev supportsFamily: MTLGPUFamilyApple5]) + caps.multiView = true; } #endif @@ -629,17 +688,6 @@ QVector<int> QRhiMetal::supportedSampleCounts() const return caps.supportedSampleCounts; } -int QRhiMetal::effectiveSampleCount(int sampleCount) const -{ - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - const int s = qBound(1, sampleCount, 64); - if (!supportedSampleCounts().contains(s)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return 1; - } - return s; -} - QRhiSwapChain *QRhiMetal::createSwapChain() { return new QMetalSwapChain(this); @@ -727,7 +775,7 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const case QRhi::DebugMarkers: return true; case QRhi::Timestamps: - return false; + return true; case QRhi::Instancing: return true; case QRhi::CustomInstanceStepRate: @@ -805,6 +853,12 @@ bool QRhiMetal::isFeatureSupported(QRhi::Feature feature) const return false; case QRhi::ThreeDimensionalTextureMipmaps: return true; + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return false; + case QRhi::ResolveDepthStencil: + return true; default: Q_UNREACHABLE(); return false; @@ -1441,11 +1495,11 @@ void QRhiMetal::setGraphicsPipeline(QRhiCommandBuffer *cb, QRhiGraphicsPipeline psD->makeActiveForCurrentRenderPassEncoder(cbD); } else { // mark work buffers that can now be safely reused as reusable - for (QMetalBuffer *workBuf : psD->d->tess.deviceLocalWorkBuffers) { + for (QMetalBuffer *workBuf : psD->d->extraBufMgr.deviceLocalWorkBuffers) { if (workBuf && workBuf->lastActiveFrameSlot == currentFrameSlot) workBuf->lastActiveFrameSlot = -1; } - for (QMetalBuffer *workBuf : psD->d->tess.hostVisibleWorkBuffers) { + for (QMetalBuffer *workBuf : psD->d->extraBufMgr.hostVisibleWorkBuffers) { if (workBuf && workBuf->lastActiveFrameSlot == currentFrameSlot) workBuf->lastActiveFrameSlot = -1; } @@ -1937,6 +1991,7 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) const quint32 vertexOrIndexCount = indexed ? args.drawIndexed.indexCount : args.draw.vertexCount; QMetalGraphicsPipelineData::Tessellation &tess(graphicsPipeline->d->tess); + QMetalGraphicsPipelineData::ExtraBufferManager &extraBufMgr(graphicsPipeline->d->extraBufMgr); const quint32 patchCount = tess.patchCountForDrawCall(vertexOrIndexCount, instanceCount); QMetalBuffer *vertOutBuf = nullptr; QMetalBuffer *tescOutBuf = nullptr; @@ -1970,7 +2025,7 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) if (outputBufferBinding >= 0) { const quint32 workBufSize = tess.vsCompOutputBufferSize(vertexOrIndexCount, instanceCount); - vertOutBuf = tess.acquireWorkBuffer(this, workBufSize); + vertOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); if (!vertOutBuf) return; [computeEncoder setBuffer: vertOutBuf->d->buf[0] offset: 0 atIndex: outputBufferBinding]; @@ -2018,7 +2073,7 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) if (outputBufferBinding >= 0) { const quint32 workBufSize = tess.tescCompOutputBufferSize(patchCount); - tescOutBuf = tess.acquireWorkBuffer(this, workBufSize); + tescOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); if (!tescOutBuf) return; [computeEncoder setBuffer: tescOutBuf->d->buf[0] offset: 0 atIndex: outputBufferBinding]; @@ -2026,14 +2081,14 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) if (patchOutputBufferBinding >= 0) { const quint32 workBufSize = tess.tescCompPatchOutputBufferSize(patchCount); - tescPatchOutBuf = tess.acquireWorkBuffer(this, workBufSize); + tescPatchOutBuf = extraBufMgr.acquireWorkBuffer(this, workBufSize); if (!tescPatchOutBuf) return; [computeEncoder setBuffer: tescPatchOutBuf->d->buf[0] offset: 0 atIndex: patchOutputBufferBinding]; } if (tessFactorBufferBinding >= 0) { - tescFactorBuf = tess.acquireWorkBuffer(this, patchCount * sizeof(MTLQuadTessellationFactorsHalf)); + tescFactorBuf = extraBufMgr.acquireWorkBuffer(this, patchCount * sizeof(MTLQuadTessellationFactorsHalf)); [computeEncoder setBuffer: tescFactorBuf->d->buf[0] offset: 0 atIndex: tessFactorBufferBinding]; } @@ -2042,7 +2097,7 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) quint32 inControlPointCount; quint32 patchCount; } params; - tescParamsBuf = tess.acquireWorkBuffer(this, sizeof(params), QMetalGraphicsPipelineData::Tessellation::WorkBufType::HostVisible); + tescParamsBuf = extraBufMgr.acquireWorkBuffer(this, sizeof(params), QMetalGraphicsPipelineData::ExtraBufferManager::WorkBufType::HostVisible); if (!tescParamsBuf) return; params.inControlPointCount = tess.inControlPointCount; @@ -2111,6 +2166,39 @@ void QRhiMetal::tessellatedDraw(const TessDrawArgs &args) } } +void QRhiMetal::adjustForMultiViewDraw(quint32 *instanceCount, QRhiCommandBuffer *cb) +{ + QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); + const int multiViewCount = cbD->currentGraphicsPipeline->m_multiViewCount; + if (multiViewCount <= 1) + return; + + const QMap<int, int> &ebb(cbD->currentGraphicsPipeline->d->vs.nativeShaderInfo.extraBufferBindings); + const int viewMaskBufBinding = ebb.value(QShaderPrivate::MslMultiViewMaskBufferBinding, -1); + if (viewMaskBufBinding == -1) { + qWarning("No extra buffer for multiview in the vertex shader; was it built with --view-count specified?"); + return; + } + struct { + quint32 viewOffset; + quint32 viewCount; + } multiViewInfo; + multiViewInfo.viewOffset = 0; + multiViewInfo.viewCount = quint32(multiViewCount); + QMetalBuffer *buf = cbD->currentGraphicsPipeline->d->extraBufMgr.acquireWorkBuffer(this, sizeof(multiViewInfo), + QMetalGraphicsPipelineData::ExtraBufferManager::WorkBufType::HostVisible); + if (buf) { + id<MTLBuffer> mtlbuf = buf->d->buf[0]; + char *p = reinterpret_cast<char *>([mtlbuf contents]); + memcpy(p, &multiViewInfo, sizeof(multiViewInfo)); + [cbD->d->currentRenderPassEncoder setVertexBuffer: mtlbuf offset: 0 atIndex: viewMaskBufBinding]; + // The instance count is adjusted for layered rendering. The vertex shader is expected to contain something like: + // uint gl_ViewIndex = spvViewMask[0] + (gl_InstanceIndex - gl_BaseInstance) % spvViewMask[1]; + // where spvViewMask is the buffer with multiViewInfo passed in above. + *instanceCount *= multiViewCount; + } +} + void QRhiMetal::draw(QRhiCommandBuffer *cb, quint32 vertexCount, quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) { @@ -2129,6 +2217,8 @@ void QRhiMetal::draw(QRhiCommandBuffer *cb, quint32 vertexCount, return; } + adjustForMultiViewDraw(&instanceCount, cb); + if (caps.baseVertexAndInstance) { [cbD->d->currentRenderPassEncoder drawPrimitives: cbD->currentGraphicsPipeline->d->primitiveType vertexStart: firstVertex vertexCount: vertexCount instanceCount: instanceCount baseInstance: firstInstance]; @@ -2167,6 +2257,8 @@ void QRhiMetal::drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, return; } + adjustForMultiViewDraw(&instanceCount, cb); + if (caps.baseVertexAndInstance) { [cbD->d->currentRenderPassEncoder drawIndexedPrimitives: cbD->currentGraphicsPipeline->d->primitiveType indexCount: indexCount @@ -2237,6 +2329,12 @@ void QRhiMetal::endExternal(QRhiCommandBuffer *cb) cbD->resetPerPassCachedState(); } +double QRhiMetal::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); + return cbD->d->lastGpuTime; +} + QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); @@ -2263,10 +2361,7 @@ QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF [d->captureScope beginScope]; - // Do not let the command buffer mess with the refcount of objects. We do - // have a proper render loop and will manage lifetimes similarly to other - // backends (Vulkan). - swapChainD->cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + swapChainD->cbWrapper.d->cb = d->newCommandBuffer(); QMetalRenderTargetData::ColorAtt colorAtt; if (swapChainD->samples > 1) { @@ -2278,6 +2373,7 @@ QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF swapChainD->rtWrapper.d->fb.colorAtt[0] = colorAtt; swapChainD->rtWrapper.d->fb.dsTex = swapChainD->ds ? swapChainD->ds->d->tex : nil; + swapChainD->rtWrapper.d->fb.dsResolveTex = nil; swapChainD->rtWrapper.d->fb.hasStencil = swapChainD->ds ? true : false; swapChainD->rtWrapper.d->fb.depthNeedsStore = false; @@ -2285,7 +2381,8 @@ QRhi::FrameOpResult QRhiMetal::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginF swapChainD->ds->lastActiveFrameSlot = currentFrameSlot; executeDeferredReleases(); - swapChainD->cbWrapper.resetState(); + swapChainD->cbWrapper.resetState(swapChainD->d->lastGpuTime[currentFrameSlot]); + swapChainD->d->lastGpuTime[currentFrameSlot] = 0; finishActiveReadbacks(); return QRhi::FrameOpSuccess; @@ -2296,30 +2393,89 @@ QRhi::FrameOpResult QRhiMetal::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame QMetalSwapChain *swapChainD = QRHI_RES(QMetalSwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); + // Keep strong reference to command buffer + id<MTLCommandBuffer> commandBuffer = swapChainD->cbWrapper.d->cb; + __block int thisFrameSlot = currentFrameSlot; - [swapChainD->cbWrapper.d->cb addCompletedHandler: ^(id<MTLCommandBuffer>) { + [commandBuffer addCompletedHandler: ^(id<MTLCommandBuffer> cb) { + swapChainD->d->lastGpuTime[thisFrameSlot] += cb.GPUEndTime - cb.GPUStartTime; dispatch_semaphore_signal(swapChainD->d->sem[thisFrameSlot]); }]; - const bool needsPresent = !flags.testFlag(QRhi::SkipPresent); - const bool presentsWithTransaction = swapChainD->d->layer.presentsWithTransaction; - if (!presentsWithTransaction && needsPresent) { - // beginFrame-endFrame without a render pass inbetween means there is no drawable. - if (id<CAMetalDrawable> drawable = swapChainD->d->curDrawable) - [swapChainD->cbWrapper.d->cb presentDrawable: drawable]; - } - - [swapChainD->cbWrapper.d->cb commit]; +#ifdef QRHI_METAL_COMMAND_BUFFERS_WITH_UNRETAINED_REFERENCES + // When Metal API validation diagnostics is enabled in Xcode the texture is + // released before the command buffer is done with it. Manually keep it alive + // to work around this. + id<MTLTexture> drawableTexture = [swapChainD->d->curDrawable.texture retain]; + [commandBuffer addCompletedHandler:^(id<MTLCommandBuffer>) { + [drawableTexture release]; + }]; +#endif - if (presentsWithTransaction && needsPresent) { - // beginFrame-endFrame without a render pass inbetween means there is no drawable. + if (flags.testFlag(QRhi::SkipPresent)) { + // Just need to commit, that's it + [commandBuffer commit]; + } else { if (id<CAMetalDrawable> drawable = swapChainD->d->curDrawable) { - // The layer has presentsWithTransaction set to true to avoid flicker on resizing, - // so here it is important to follow what the Metal docs say when it comes to the - // issuing the present. - [swapChainD->cbWrapper.d->cb waitUntilScheduled]; - [drawable present]; + // Got something to present + if (swapChainD->d->layer.presentsWithTransaction) { + [commandBuffer commit]; + // Keep strong reference to Metal layer + auto *metalLayer = swapChainD->d->layer; + auto presentWithTransaction = ^{ + [commandBuffer waitUntilScheduled]; + // If the layer has been resized while we waited to be scheduled we bail out, + // as the drawable is no longer valid for the layer, and we'll get a follow-up + // display with the right size. We know we are on the main thread here, which + // means we can access the layer directly. We also know that the layer is valid, + // since the block keeps a strong reference to it, compared to the QRhiSwapChain + // that can go away under our feet by the time we're scheduled. + const auto surfaceSize = QSizeF::fromCGSize(metalLayer.bounds.size) * metalLayer.contentsScale; + const auto textureSize = QSizeF(drawable.texture.width, drawable.texture.height); + if (textureSize == surfaceSize) { + [drawable present]; + } else { + qCDebug(QRHI_LOG_INFO) << "Skipping" << drawable << "due to texture size" + << textureSize << "not matching surface size" << surfaceSize; + } + }; + + if (NSThread.currentThread == NSThread.mainThread) { + presentWithTransaction(); + } else { + auto *qtMetalLayer = qt_objc_cast<QMetalLayer*>(swapChainD->d->layer); + Q_ASSERT(qtMetalLayer); + // Let the main thread present the drawable from displayLayer + qtMetalLayer.mainThreadPresentation = presentWithTransaction; + } + } else { + // Keep strong reference to Metal layer so it's valid in the block + auto *qtMetalLayer = qt_objc_cast<QMetalLayer*>(swapChainD->d->layer); + [commandBuffer addScheduledHandler:^(id<MTLCommandBuffer>) { + if (qtMetalLayer) { + // The schedule handler comes in on the com.Metal.CompletionQueueDispatch + // thread, which means we might be racing against a display cycle on the + // main thread. If the displayLayer is already in progress, we don't want + // to step on its toes. + if (qtMetalLayer.displayLock.tryLockForRead()) { + [drawable present]; + qtMetalLayer.displayLock.unlock(); + } else { + qCDebug(QRHI_LOG_INFO) << "Skipping" << drawable + << "due to" << qtMetalLayer << "needing display"; + } + } else { + [drawable present]; + } + }]; + [commandBuffer commit]; + } + } else { + // Still need to commit, even if we don't have a drawable + [commandBuffer commit]; } + + swapChainD->currentFrameSlot = (swapChainD->currentFrameSlot + 1) % QMTL_FRAMES_IN_FLIGHT; } // Must not hold on to the drawable, regardless of needsPresent @@ -2328,9 +2484,6 @@ QRhi::FrameOpResult QRhiMetal::endFrame(QRhiSwapChain *swapChain, QRhi::EndFrame [d->captureScope endScope]; - if (needsPresent) - swapChainD->currentFrameSlot = (swapChainD->currentFrameSlot + 1) % QMTL_FRAMES_IN_FLIGHT; - swapChainD->frameCount += 1; currentSwapChain = nullptr; return QRhi::FrameOpSuccess; @@ -2347,10 +2500,11 @@ QRhi::FrameOpResult QRhiMetal::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi: d->ofr.active = true; *cb = &d->ofr.cbWrapper; - d->ofr.cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + d->ofr.cbWrapper.d->cb = d->newCommandBuffer(); executeDeferredReleases(); - d->ofr.cbWrapper.resetState(); + d->ofr.cbWrapper.resetState(d->ofr.lastGpuTime); + d->ofr.lastGpuTime = 0; finishActiveReadbacks(); return QRhi::FrameOpSuccess; @@ -2362,10 +2516,13 @@ QRhi::FrameOpResult QRhiMetal::endOffscreenFrame(QRhi::EndFrameFlags flags) Q_ASSERT(d->ofr.active); d->ofr.active = false; - [d->ofr.cbWrapper.d->cb commit]; + id<MTLCommandBuffer> cb = d->ofr.cbWrapper.d->cb; + [cb commit]; // offscreen frames wait for completion, unlike swapchain ones - [d->ofr.cbWrapper.d->cb waitUntilCompleted]; + [cb waitUntilCompleted]; + + d->ofr.lastGpuTime += cb.GPUEndTime - cb.GPUStartTime; finishActiveReadbacks(true); @@ -2406,10 +2563,13 @@ QRhi::FrameOpResult QRhiMetal::finish() } if (inFrame) { - if (d->ofr.active) - d->ofr.cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; - else - swapChainD->cbWrapper.d->cb = [d->cmdQueue commandBufferWithUnretainedReferences]; + if (d->ofr.active) { + d->ofr.lastGpuTime += cb.GPUEndTime - cb.GPUStartTime; + d->ofr.cbWrapper.d->cb = d->newCommandBuffer(); + } else { + swapChainD->d->lastGpuTime[currentFrameSlot] += cb.GPUEndTime - cb.GPUStartTime; + swapChainD->cbWrapper.d->cb = d->newCommandBuffer(); + } } executeDeferredReleases(true); @@ -2471,7 +2631,6 @@ void QRhiMetal::enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEnc int w = img.width(); int h = img.height(); int bpl = img.bytesPerLine(); - int srcOffset = 0; if (!subresDesc.sourceSize().isEmpty() || !subresDesc.sourceTopLeft().isNull()) { const int sx = subresDesc.sourceTopLeft().x(); @@ -2480,10 +2639,12 @@ void QRhiMetal::enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEnc w = subresDesc.sourceSize().width(); h = subresDesc.sourceSize().height(); } - if (img.depth() == 32) { - memcpy(reinterpret_cast<char *>(mp) + *curOfs, img.constBits(), size_t(fullImageSizeBytes)); - srcOffset = sy * bpl + sx * 4; - // bpl remains set to the original image's row stride + if (w == img.width()) { + const int bpc = qMax(1, img.depth() / 8); + Q_ASSERT(h * img.bytesPerLine() <= fullImageSizeBytes); + memcpy(reinterpret_cast<char *>(mp) + *curOfs, + img.constBits() + sy * img.bytesPerLine() + sx * bpc, + h * img.bytesPerLine()); } else { img = img.copy(sx, sy, w, h); bpl = img.bytesPerLine(); @@ -2495,7 +2656,7 @@ void QRhiMetal::enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEnc } [blitEnc copyFromBuffer: texD->d->stagingBuf[currentFrameSlot] - sourceOffset: NSUInteger(*curOfs + srcOffset) + sourceOffset: NSUInteger(*curOfs) sourceBytesPerRow: NSUInteger(bpl) sourceBytesPerImage: 0 sourceSize: MTLSizeMake(NSUInteger(w), NSUInteger(h), 1) @@ -2587,6 +2748,15 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate QMetalCommandBuffer *cbD = QRHI_RES(QMetalCommandBuffer, cb); QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); + id<MTLBlitCommandEncoder> blitEnc = nil; + auto ensureBlit = [&blitEnc, cbD, this]() { + if (!blitEnc) { + blitEnc = [cbD->d->cb blitCommandEncoder]; + if (debugMarkers) + [blitEnc pushDebugGroup: @"Texture upload/copy"]; + } + }; + for (int opIdx = 0; opIdx < ud->activeBufferOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::BufferOp &u(ud->bufferOps[opIdx]); if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::DynamicUpdate) { @@ -2625,19 +2795,17 @@ void QRhiMetal::enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdate readback.readSize = u.readSize; readback.result = u.result; d->activeBufferReadbacks.append(readback); +#ifdef Q_OS_MACOS + if (bufD->d->managed) { + // On non-Apple Silicon, manually synchronize memory from GPU to CPU + ensureBlit(); + [blitEnc synchronizeResource:readback.buf]; + } +#endif } } } - id<MTLBlitCommandEncoder> blitEnc = nil; - auto ensureBlit = [&blitEnc, cbD, this] { - if (!blitEnc) { - blitEnc = [cbD->d->cb blitCommandEncoder]; - if (debugMarkers) - [blitEnc pushDebugGroup: @"Texture upload/copy"]; - } - }; - for (int opIdx = 0; opIdx < ud->activeTextureOpCount; ++opIdx) { const QRhiResourceUpdateBatchPrivate::TextureOp &u(ud->textureOps[opIdx]); if (u.type == QRhiResourceUpdateBatchPrivate::TextureOp::Upload) { @@ -2848,28 +3016,39 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, if (!QRhiRenderTargetAttachmentTracker::isUpToDate<QMetalTexture, QMetalRenderBuffer>(rtTex->description(), rtD->currentResIdList)) rtTex->create(); cbD->d->currentPassRpDesc = d->createDefaultRenderPass(rtD->dsAttCount, colorClearValue, depthStencilClearValue, rtD->colorAttCount); - if (rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents)) { + if (rtD->fb.preserveColor) { for (uint i = 0; i < uint(rtD->colorAttCount); ++i) cbD->d->currentPassRpDesc.colorAttachments[i].loadAction = MTLLoadActionLoad; } - if (rtD->dsAttCount && rtTex->m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents)) { + if (rtD->dsAttCount && rtD->fb.preserveDs) { cbD->d->currentPassRpDesc.depthAttachment.loadAction = MTLLoadActionLoad; cbD->d->currentPassRpDesc.stencilAttachment.loadAction = MTLLoadActionLoad; } + int colorAttCount = 0; for (auto it = rtTex->m_desc.cbeginColorAttachments(), itEnd = rtTex->m_desc.cendColorAttachments(); it != itEnd; ++it) { - if (it->texture()) + colorAttCount += 1; + if (it->texture()) { QRHI_RES(QMetalTexture, it->texture())->lastActiveFrameSlot = currentFrameSlot; - else if (it->renderBuffer()) + if (it->multiViewCount() >= 2) + cbD->d->currentPassRpDesc.renderTargetArrayLength = NSUInteger(it->multiViewCount()); + } else if (it->renderBuffer()) { QRHI_RES(QMetalRenderBuffer, it->renderBuffer())->lastActiveFrameSlot = currentFrameSlot; + } if (it->resolveTexture()) QRHI_RES(QMetalTexture, it->resolveTexture())->lastActiveFrameSlot = currentFrameSlot; } if (rtTex->m_desc.depthStencilBuffer()) QRHI_RES(QMetalRenderBuffer, rtTex->m_desc.depthStencilBuffer())->lastActiveFrameSlot = currentFrameSlot; - if (rtTex->m_desc.depthTexture()) - QRHI_RES(QMetalTexture, rtTex->m_desc.depthTexture())->lastActiveFrameSlot = currentFrameSlot; + if (rtTex->m_desc.depthTexture()) { + QMetalTexture *depthTexture = QRHI_RES(QMetalTexture, rtTex->m_desc.depthTexture()); + depthTexture->lastActiveFrameSlot = currentFrameSlot; + if (colorAttCount == 0 && depthTexture->arraySize() >= 2) + cbD->d->currentPassRpDesc.renderTargetArrayLength = NSUInteger(depthTexture->arraySize()); + } + if (rtTex->m_desc.depthResolveTexture()) + QRHI_RES(QMetalTexture, rtTex->m_desc.depthResolveTexture())->lastActiveFrameSlot = currentFrameSlot; } break; default: @@ -2883,7 +3062,8 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, cbD->d->currentPassRpDesc.colorAttachments[i].depthPlane = NSUInteger(rtD->fb.colorAtt[i].slice); cbD->d->currentPassRpDesc.colorAttachments[i].level = NSUInteger(rtD->fb.colorAtt[i].level); if (rtD->fb.colorAtt[i].resolveTex) { - cbD->d->currentPassRpDesc.colorAttachments[i].storeAction = MTLStoreActionMultisampleResolve; + cbD->d->currentPassRpDesc.colorAttachments[i].storeAction = rtD->fb.preserveColor ? MTLStoreActionStoreAndMultisampleResolve + : MTLStoreActionMultisampleResolve; cbD->d->currentPassRpDesc.colorAttachments[i].resolveTexture = rtD->fb.colorAtt[i].resolveTex; cbD->d->currentPassRpDesc.colorAttachments[i].resolveSlice = NSUInteger(rtD->fb.colorAtt[i].resolveLayer); cbD->d->currentPassRpDesc.colorAttachments[i].resolveLevel = NSUInteger(rtD->fb.colorAtt[i].resolveLevel); @@ -2896,6 +3076,15 @@ void QRhiMetal::beginPass(QRhiCommandBuffer *cb, cbD->d->currentPassRpDesc.stencilAttachment.texture = rtD->fb.hasStencil ? rtD->fb.dsTex : nil; if (rtD->fb.depthNeedsStore) // Depth/Stencil is set to DontCare by default, override if needed cbD->d->currentPassRpDesc.depthAttachment.storeAction = MTLStoreActionStore; + if (rtD->fb.dsResolveTex) { + cbD->d->currentPassRpDesc.depthAttachment.storeAction = rtD->fb.depthNeedsStore ? MTLStoreActionStoreAndMultisampleResolve + : MTLStoreActionMultisampleResolve; + cbD->d->currentPassRpDesc.depthAttachment.resolveTexture = rtD->fb.dsResolveTex; + if (rtD->fb.hasStencil) { + cbD->d->currentPassRpDesc.stencilAttachment.resolveTexture = rtD->fb.dsResolveTex; + cbD->d->currentPassRpDesc.stencilAttachment.storeAction = cbD->d->currentPassRpDesc.depthAttachment.storeAction; + } + } } cbD->d->currentRenderPassEncoder = [cbD->d->cb renderCommandEncoderWithDescriptor: cbD->d->currentPassRpDesc]; @@ -3666,12 +3855,10 @@ bool QMetalTexture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both 1D and cube"); return false; } - m_depth = qMax(1, m_depth); if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } - m_arraySize = qMax(0, m_arraySize); if (m_arraySize > 0 && !isArray) { qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); return false; @@ -3707,11 +3894,11 @@ bool QMetalTexture::create() desc.textureType = isArray ? MTLTextureType1DArray : MTLTextureType1D; } else if (isArray) { #ifdef Q_OS_IOS - if (samples > 1) { - // would be available on iOS 14.0+ but cannot test for that with a 13 SDK - qWarning("Multisample 2D texture array is not supported on iOS"); + if (@available(iOS 14, *)) { + desc.textureType = samples > 1 ? MTLTextureType2DMultisampleArray : MTLTextureType2DArray; + } else { + desc.textureType = MTLTextureType2DArray; } - desc.textureType = MTLTextureType2DArray; #else desc.textureType = samples > 1 ? MTLTextureType2DMultisampleArray : MTLTextureType2DArray; #endif @@ -3721,12 +3908,12 @@ bool QMetalTexture::create() desc.pixelFormat = d->format; desc.width = NSUInteger(size.width()); desc.height = NSUInteger(size.height()); - desc.depth = is3D ? m_depth : 1; + desc.depth = is3D ? qMax(1, m_depth) : 1; desc.mipmapLevelCount = NSUInteger(mipLevelCount); if (samples > 1) desc.sampleCount = NSUInteger(samples); if (isArray) - desc.arrayLength = NSUInteger(m_arraySize); + desc.arrayLength = NSUInteger(qMax(0, m_arraySize)); desc.resourceOptions = MTLResourceStorageModePrivate; desc.storageMode = MTLStorageModePrivate; desc.usage = MTLTextureUsageShaderRead; @@ -3785,7 +3972,8 @@ id<MTLTexture> QMetalTextureData::viewForLevel(int level) const bool isCube = q->m_flags.testFlag(QRhiTexture::CubeMap); const bool isArray = q->m_flags.testFlag(QRhiTexture::TextureArray); id<MTLTexture> view = [tex newTextureViewWithPixelFormat: format textureType: type - levels: NSMakeRange(NSUInteger(level), 1) slices: NSMakeRange(0, isCube ? 6 : (isArray ? q->m_arraySize : 1))]; + levels: NSMakeRange(NSUInteger(level), 1) + slices: NSMakeRange(0, isCube ? 6 : (isArray ? qMax(0, q->m_arraySize) : 1))]; perLevelViews[level] = view; return view; @@ -3930,7 +4118,9 @@ QMetalRenderPassDescriptor::~QMetalRenderPassDescriptor() void QMetalRenderPassDescriptor::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } bool QMetalRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -3973,13 +4163,17 @@ void QMetalRenderPassDescriptor::updateSerializedFormat() QRhiRenderPassDescriptor *QMetalRenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - QMetalRenderPassDescriptor *rp = new QMetalRenderPassDescriptor(m_rhi); - rp->colorAttachmentCount = colorAttachmentCount; - rp->hasDepthStencil = hasDepthStencil; - memcpy(rp->colorFormat, colorFormat, sizeof(colorFormat)); - rp->dsFormat = dsFormat; - rp->updateSerializedFormat(); - return rp; + QMetalRenderPassDescriptor *rpD = new QMetalRenderPassDescriptor(m_rhi); + rpD->colorAttachmentCount = colorAttachmentCount; + rpD->hasDepthStencil = hasDepthStencil; + memcpy(rpD->colorFormat, colorFormat, sizeof(colorFormat)); + rpD->dsFormat = dsFormat; + + rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiMetal); + rhiD->registerResource(rpD, false); + return rpD; } QVector<quint32> QMetalRenderPassDescriptor::serializedFormat() const @@ -4035,7 +4229,9 @@ QMetalTextureRenderTarget::~QMetalTextureRenderTarget() void QMetalTextureRenderTarget::destroy() { - // nothing to do here + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } QRhiRenderPassDescriptor *QMetalTextureRenderTarget::newCompatibleRenderPassDescriptor() @@ -4058,6 +4254,9 @@ QRhiRenderPassDescriptor *QMetalTextureRenderTarget::newCompatibleRenderPassDesc rpD->dsFormat = int(QRHI_RES(QMetalRenderBuffer, m_desc.depthStencilBuffer())->d->format); rpD->updateSerializedFormat(); + + QRHI_RES_RHI(QRhiMetal); + rhiD->registerResource(rpD, false); return rpD; } @@ -4108,8 +4307,9 @@ bool QMetalTextureRenderTarget::create() if (m_desc.depthTexture()) { QMetalTexture *depthTexD = QRHI_RES(QMetalTexture, m_desc.depthTexture()); d->fb.dsTex = depthTexD->d->tex; - d->fb.hasStencil = false; - d->fb.depthNeedsStore = true; + d->fb.hasStencil = rhiD->isStencilSupportingFormat(depthTexD->format()); + d->fb.depthNeedsStore = !m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture(); + d->fb.preserveDs = m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents); if (d->colorAttCount == 0) { d->pixelSize = depthTexD->pixelSize(); d->sampleCount = depthTexD->samples; @@ -4119,18 +4319,27 @@ bool QMetalTextureRenderTarget::create() d->fb.dsTex = depthRbD->d->tex; d->fb.hasStencil = true; d->fb.depthNeedsStore = false; + d->fb.preserveDs = false; if (d->colorAttCount == 0) { d->pixelSize = depthRbD->pixelSize(); d->sampleCount = depthRbD->samples; } } + if (m_desc.depthResolveTexture()) { + QMetalTexture *depthResolveTexD = QRHI_RES(QMetalTexture, m_desc.depthResolveTexture()); + d->fb.dsResolveTex = depthResolveTexD->d->tex; + } d->dsAttCount = 1; } else { d->dsAttCount = 0; } + if (d->colorAttCount > 0) + d->fb.preserveColor = m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents); + QRhiRenderTargetAttachmentTracker::updateResIdList<QMetalTexture, QMetalRenderBuffer>(m_desc, &d->currentResIdList); + rhiD->registerResource(this, false); return true; } @@ -4166,6 +4375,10 @@ void QMetalShaderResourceBindings::destroy() { sortedBindings.clear(); maxBinding = -1; + + QRHI_RES_RHI(QRhiMetal); + if (rhiD) + rhiD->unregisterResource(this); } bool QMetalShaderResourceBindings::create() @@ -4192,6 +4405,7 @@ bool QMetalShaderResourceBindings::create() memset(&bd, 0, sizeof(BoundResourceData)); generation += 1; + rhiD->registerResource(this, false); return true; } @@ -4234,10 +4448,10 @@ void QMetalGraphicsPipeline::destroy() d->tess.compTesc.destroy(); d->tess.vertTese.destroy(); - qDeleteAll(d->tess.deviceLocalWorkBuffers); - d->tess.deviceLocalWorkBuffers.clear(); - qDeleteAll(d->tess.hostVisibleWorkBuffers); - d->tess.hostVisibleWorkBuffers.clear(); + qDeleteAll(d->extraBufMgr.deviceLocalWorkBuffers); + d->extraBufMgr.deviceLocalWorkBuffers.clear(); + qDeleteAll(d->extraBufMgr.hostVisibleWorkBuffers); + d->extraBufMgr.hostVisibleWorkBuffers.clear(); delete d->bufferSizeBuffer; d->bufferSizeBuffer = nullptr; @@ -4309,6 +4523,22 @@ static inline MTLVertexFormat toMetalAttributeFormat(QRhiVertexInputAttribute::F return MTLVertexFormatHalf2; case QRhiVertexInputAttribute::Half: return MTLVertexFormatHalf; + case QRhiVertexInputAttribute::UShort4: + return MTLVertexFormatUShort4; + case QRhiVertexInputAttribute::UShort3: + return MTLVertexFormatUShort3; + case QRhiVertexInputAttribute::UShort2: + return MTLVertexFormatUShort2; + case QRhiVertexInputAttribute::UShort: + return MTLVertexFormatUShort; + case QRhiVertexInputAttribute::SShort4: + return MTLVertexFormatShort4; + case QRhiVertexInputAttribute::SShort3: + return MTLVertexFormatShort3; + case QRhiVertexInputAttribute::SShort2: + return MTLVertexFormatShort2; + case QRhiVertexInputAttribute::SShort: + return MTLVertexFormatShort; default: Q_UNREACHABLE(); return MTLVertexFormatFloat4; @@ -4464,6 +4694,24 @@ static inline MTLPrimitiveType toMetalPrimitiveType(QRhiGraphicsPipeline::Topolo } } +static inline MTLPrimitiveTopologyClass toMetalPrimitiveTopologyClass(QRhiGraphicsPipeline::Topology t) +{ + switch (t) { + case QRhiGraphicsPipeline::Triangles: + case QRhiGraphicsPipeline::TriangleStrip: + case QRhiGraphicsPipeline::TriangleFan: + return MTLPrimitiveTopologyClassTriangle; + case QRhiGraphicsPipeline::Lines: + case QRhiGraphicsPipeline::LineStrip: + return MTLPrimitiveTopologyClassLine; + case QRhiGraphicsPipeline::Points: + return MTLPrimitiveTopologyClassPoint; + default: + Q_UNREACHABLE(); + return MTLPrimitiveTopologyClassTriangle; + } +} + static inline MTLCullMode toMetalCullMode(QRhiGraphicsPipeline::CullMode c) { switch (c) { @@ -4651,7 +4899,7 @@ void QMetalGraphicsPipeline::setupAttachmentsInMetalRenderPassDescriptor(void *m } QRHI_RES_RHI(QRhiMetal); - rpDesc.sampleCount = NSUInteger(rhiD->effectiveSampleCount(m_sampleCount)); + rpDesc.rasterSampleCount = NSUInteger(rhiD->effectiveSampleCount(m_sampleCount)); } void QMetalGraphicsPipeline::setupMetalDepthStencilDescriptor(void *metalDsDesc) @@ -4704,6 +4952,7 @@ void QMetalGraphicsPipelineData::setupVertexInputDescriptor(MTLVertexDescriptor desc.attributes[loc].bufferIndex = NSUInteger(firstVertexBinding + it->binding()); } int bindingIndex = 0; + const NSUInteger viewCount = qMax<NSUInteger>(1, q->multiViewCount()); for (auto it = vertexInputLayout.cbeginBindings(), itEnd = vertexInputLayout.cendBindings(); it != itEnd; ++it, ++bindingIndex) { @@ -4712,6 +4961,8 @@ void QMetalGraphicsPipelineData::setupVertexInputDescriptor(MTLVertexDescriptor it->classification() == QRhiVertexInputBinding::PerInstance ? MTLVertexStepFunctionPerInstance : MTLVertexStepFunctionPerVertex; desc.layouts[layoutIdx].stepRate = NSUInteger(it->instanceStepRate()); + if (desc.layouts[layoutIdx].stepFunction == MTLVertexStepFunctionPerInstance) + desc.layouts[layoutIdx].stepRate *= viewCount; desc.layouts[layoutIdx].stride = it->stride(); } } @@ -4760,33 +5011,10 @@ void QRhiMetalData::trySeedingRenderPipelineFromBinaryArchive(MTLRenderPipelineD } } -static bool canAddToBinaryArchive(QRhiMetalData *d) -{ - if (@available(macOS 11.0, iOS 14.0, *)) { - if (!d->binArch) - return false; - - // ### QTBUG-106703, QTBUG-108216, revisit after 13.0 - if (!d->binArchWasEmpty && d->q->osMajor >= 13) { - static bool logPrinted = false; - if (!logPrinted) { - logPrinted = true; - qCDebug(QRHI_LOG_INFO, "Skipping adding more pipelines to MTLBinaryArchive on this OS version (%d.%d) due to known issues.", - d->q->osMajor, d->q->osMinor); - } - return false; - } - - return true; - } else { - return false; - } -} - void QRhiMetalData::addRenderPipelineToBinaryArchive(MTLRenderPipelineDescriptor *rpDesc) { if (@available(macOS 11.0, iOS 14.0, *)) { - if (canAddToBinaryArchive(this)) { + if (binArch) { NSError *err = nil; if (![binArch addRenderPipelineFunctionsWithDescriptor: rpDesc error: &err]) { const QString msg = QString::fromNSString(err.localizedDescription); @@ -4888,6 +5116,9 @@ bool QMetalGraphicsPipeline::createVertexFragmentPipeline() QMetalRenderPassDescriptor *rpD = QRHI_RES(QMetalRenderPassDescriptor, m_renderPassDesc); setupAttachmentsInMetalRenderPassDescriptor(rpDesc, rpD); + if (m_multiViewCount >= 2) + rpDesc.inputPrimitiveTopology = toMetalPrimitiveTopologyClass(m_topology); + rhiD->d->trySeedingRenderPipelineFromBinaryArchive(rpDesc); if (rhiD->rhiFlags.testFlag(QRhi::EnablePipelineCacheDataSave)) @@ -5410,7 +5641,7 @@ id<MTLRenderPipelineState> QMetalGraphicsPipelineData::Tessellation::teseFragRen return ps; } -QMetalBuffer *QMetalGraphicsPipelineData::Tessellation::acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type) +QMetalBuffer *QMetalGraphicsPipelineData::ExtraBufferManager::acquireWorkBuffer(QRhiMetal *rhiD, quint32 size, WorkBufType type) { QVector<QMetalBuffer *> *workBuffers = type == WorkBufType::DeviceLocal ? &deviceLocalWorkBuffers : &hostVisibleWorkBuffers; @@ -5476,6 +5707,9 @@ bool QMetalGraphicsPipeline::createTessellationPipelines(const QShader &tessVert return false; } + if (m_multiViewCount >= 2) + qWarning("Multiview is not supported with tessellation"); + // Now the vertex shader is a compute shader. // It should have three dedicated *VertexAsComputeShader variants. // What the requested variant was (Standard or Batchable) plays no role here. @@ -5766,7 +6000,7 @@ void QRhiMetalData::trySeedingComputePipelineFromBinaryArchive(MTLComputePipelin void QRhiMetalData::addComputePipelineToBinaryArchive(MTLComputePipelineDescriptor *cpDesc) { if (@available(macOS 11.0, iOS 14.0, *)) { - if (canAddToBinaryArchive(this)) { + if (binArch) { NSError *err = nil; if (![binArch addComputePipelineFunctionsWithDescriptor: cpDesc error: &err]) { const QString msg = QString::fromNSString(err.localizedDescription); @@ -5897,8 +6131,9 @@ const QRhiNativeHandles *QMetalCommandBuffer::nativeHandles() return &nativeHandlesStruct; } -void QMetalCommandBuffer::resetState() +void QMetalCommandBuffer::resetState(double lastGpuTime) { + d->lastGpuTime = lastGpuTime; d->currentRenderPassEncoder = nil; d->currentComputePassEncoder = nil; d->tessellationComputeEncoder = nil; @@ -5975,13 +6210,8 @@ void QMetalSwapChain::destroy() d->msaaTex[i] = nil; } -#ifdef Q_OS_MACOS - d->liveResizeStartObserver.remove(); - d->liveResizeEndObserver.remove(); - d->liveResizeObserverSet = false; -#endif - d->layer = nullptr; + m_proxyData = {}; [d->curDrawable release]; d->curDrawable = nil; @@ -6045,14 +6275,24 @@ QSize QMetalSwapChain::surfacePixelSize() bool QMetalSwapChain::isFormatSupported(Format f) { -#ifdef Q_OS_MACOS - return f == SDR || f == HDRExtendedSrgbLinear; -#endif + if (f == HDRExtendedSrgbLinear) { + if (@available(macOS 10.11, iOS 16.0, *)) + return hdrInfo().limits.colorComponentValue.maxPotentialColorComponentValue > 1.0f; + else + return false; + } else if (f == HDRExtendedDisplayP3Linear) { + if (@available(macOS 11.0, iOS 14.0, *)) + return hdrInfo().limits.colorComponentValue.maxPotentialColorComponentValue > 1.0f; + else + return false; + } return f == SDR; } QRhiRenderPassDescriptor *QMetalSwapChain::newCompatibleRenderPassDescriptor() { + QRHI_RES_RHI(QRhiMetal); + chooseFormats(); // ensure colorFormat and similar are filled out QMetalRenderPassDescriptor *rpD = new QMetalRenderPassDescriptor(m_rhi); @@ -6063,7 +6303,6 @@ QRhiRenderPassDescriptor *QMetalSwapChain::newCompatibleRenderPassDescriptor() #ifdef Q_OS_MACOS // m_depthStencil may not be built yet so cannot rely on computed fields in it - QRHI_RES_RHI(QRhiMetal); rpD->dsFormat = rhiD->d->dev.depth24Stencil8PixelFormatSupported ? MTLPixelFormatDepth24Unorm_Stencil8 : MTLPixelFormatDepth32Float_Stencil8; #else @@ -6071,6 +6310,8 @@ QRhiRenderPassDescriptor *QMetalSwapChain::newCompatibleRenderPassDescriptor() #endif rpD->updateSerializedFormat(); + + rhiD->registerResource(rpD, false); return rpD; } @@ -6079,7 +6320,7 @@ void QMetalSwapChain::chooseFormats() QRHI_RES_RHI(QRhiMetal); samples = rhiD->effectiveSampleCount(m_sampleCount); // pick a format that is allowed for CAMetalLayer.pixelFormat - if (m_format == HDRExtendedSrgbLinear) { + if (m_format == HDRExtendedSrgbLinear || m_format == HDRExtendedDisplayP3Linear) { d->colorFormat = MTLPixelFormatRGBA16Float; d->rhiColorFormat = QRhiTexture::RGBA16F; return; @@ -6126,13 +6367,18 @@ bool QMetalSwapChain::createOrResize() chooseFormats(); if (d->colorFormat != d->layer.pixelFormat) d->layer.pixelFormat = d->colorFormat; -#ifdef Q_OS_MACOS - // Can't enable this on iOS until wantsExtendedDynamicRangeContent is available + if (m_format == HDRExtendedSrgbLinear) { - d->layer.colorspace = CGColorSpaceCreateWithName(kCGColorSpaceExtendedLinearSRGB); - d->layer.wantsExtendedDynamicRangeContent = YES; + if (@available(macOS 10.11, iOS 16.0, *)) { + d->layer.colorspace = CGColorSpaceCreateWithName(kCGColorSpaceExtendedLinearSRGB); + d->layer.wantsExtendedDynamicRangeContent = YES; + } + } else if (m_format == HDRExtendedDisplayP3Linear) { + if (@available(macOS 11.0, iOS 16.0, *)) { + d->layer.colorspace = CGColorSpaceCreateWithName(kCGColorSpaceExtendedLinearDisplayP3); + d->layer.wantsExtendedDynamicRangeContent = YES; + } } -#endif if (m_flags.testFlag(UsedAsTransferSource)) d->layer.framebufferOnly = NO; @@ -6170,38 +6416,11 @@ bool QMetalSwapChain::createOrResize() [d->layer setDevice: rhiD->d->dev]; -#ifdef Q_OS_MACOS - // Can only use presentsWithTransaction (to get smooth resizing) when - // presenting from the main (gui) thread. We predict that based on the - // thread this function is called on since if the QRhiSwapChain is - // initialied on a given thread then that's almost certainly the thread on - // which the QRhi renders and presents. - const bool canUsePresentsWithTransaction = NSThread.isMainThread; - - // Have an env.var. just in case it turns out presentsWithTransaction is - // not desired in some specific case. - static bool allowPresentsWithTransaction = !qEnvironmentVariableIntValue("QT_MTL_NO_TRANSACTION"); - - if (allowPresentsWithTransaction && canUsePresentsWithTransaction && !d->liveResizeObserverSet) { - d->liveResizeObserverSet = true; - NSView *view = reinterpret_cast<NSView *>(window->winId()); - NSWindow *window = view.window; - if (window) { - qCDebug(QRHI_LOG_INFO, "will set presentsWithTransaction during live resize"); - d->liveResizeStartObserver = QMacNotificationObserver(window, NSWindowWillStartLiveResizeNotification, [this] { - d->layer.presentsWithTransaction = true; - }); - d->liveResizeEndObserver = QMacNotificationObserver(window, NSWindowDidEndLiveResizeNotification, [this] { - d->layer.presentsWithTransaction = false; - }); - } - } -#endif - [d->curDrawable release]; d->curDrawable = nil; for (int i = 0; i < QMTL_FRAMES_IN_FLIGHT; ++i) { + d->lastGpuTime[i] = 0; if (!d->sem[i]) d->sem[i] = dispatch_semaphore_create(QMTL_FRAMES_IN_FLIGHT - 1); } @@ -6264,21 +6483,28 @@ QRhiSwapChainHdrInfo QMetalSwapChain::hdrInfo() { QRhiSwapChainHdrInfo info; info.limitsType = QRhiSwapChainHdrInfo::ColorComponentValue; - if (m_format == SDR) { - info.limits.colorComponentValue.maxColorComponentValue = 1; - return info; - } + info.limits.colorComponentValue.maxColorComponentValue = 1; + info.limits.colorComponentValue.maxPotentialColorComponentValue = 1; + info.luminanceBehavior = QRhiSwapChainHdrInfo::DisplayReferred; // 1.0 = SDR white + info.sdrWhiteLevel = 200; // typical value, but dummy (don't know the real one); won't matter due to being display-referred -#ifdef Q_OS_MACOS - info.isHardCodedDefaults = false; - NSView *view = reinterpret_cast<NSView *>(window->winId()); - info.limits.colorComponentValue.maxColorComponentValue = view.window.screen.maximumExtendedDynamicRangeColorComponentValue; -#else - // ### Fixme: Maybe retrieve the brightness from the screen and if we're not at full brightness we might be able to do more. - // For now, assume 2, in line with iPhone 12 specs that claim 625 nits max brightness and 1200 nits max HDR brightness. - info.isHardCodedDefaults = true; - info.limits.colorComponentValue.maxColorComponentValue = 2; + if (m_window) { + // Must use m_window, not window, given this may be called before createOrResize(). +#if defined(Q_OS_MACOS) + NSView *view = reinterpret_cast<NSView *>(m_window->winId()); + NSScreen *screen = view.window.screen; + info.limits.colorComponentValue.maxColorComponentValue = screen.maximumExtendedDynamicRangeColorComponentValue; + info.limits.colorComponentValue.maxPotentialColorComponentValue = screen.maximumPotentialExtendedDynamicRangeColorComponentValue; +#elif defined(Q_OS_IOS) + if (@available(iOS 16.0, *)) { + UIView *view = reinterpret_cast<UIView *>(m_window->winId()); + UIScreen *screen = view.window.windowScene.screen; + info.limits.colorComponentValue.maxColorComponentValue = view.window.windowScene.screen.currentEDRHeadroom; + info.limits.colorComponentValue.maxPotentialColorComponentValue = screen.potentialEDRHeadroom; + } #endif + } + return info; } diff --git a/src/gui/rhi/qrhimetal_p.h b/src/gui/rhi/qrhimetal_p.h index 016bf749e2..f539148b2c 100644 --- a/src/gui/rhi/qrhimetal_p.h +++ b/src/gui/rhi/qrhimetal_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHIMETAL_H -#define QRHIMETAL_H +#ifndef QRHIMETAL_P_H +#define QRHIMETAL_P_H // // W A R N I N G @@ -15,29 +15,494 @@ // We mean it. // -#include <private/qrhi_p.h> - -Q_FORWARD_DECLARE_OBJC_CLASS(MTLDevice); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandQueue); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLCommandBuffer); -Q_FORWARD_DECLARE_OBJC_CLASS(MTLRenderCommandEncoder); +#include "qrhi_p.h" +#include <QWindow> QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiMetalInitParams : public QRhiInitParams +static const int QMTL_FRAMES_IN_FLIGHT = 2; + +// have to hide the ObjC stuff, this header cannot contain MTL* at all +struct QMetalBufferData; + +struct QMetalBuffer : public QRhiBuffer +{ + QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QMetalBuffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + QMetalBufferData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; + + static constexpr int WorkBufPoolUsage = 1 << 8; + static_assert(WorkBufPoolUsage > QRhiBuffer::StorageBuffer); +}; + +struct QMetalRenderBufferData; + +struct QMetalRenderBuffer : public QRhiRenderBuffer +{ + QMetalRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QMetalRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + QMetalRenderBufferData *d; + int samples = 1; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalTextureData; + +struct QMetalTexture : public QRhiTexture +{ + QMetalTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QMetalTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + + QMetalTextureData *d; + int mipLevelCount = 0; + int samples = 1; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; + friend struct QMetalTextureData; +}; + +struct QMetalSamplerData; + +struct QMetalSampler : public QRhiSampler +{ + QMetalSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QMetalSampler(); + void destroy() override; + bool create() override; + + QMetalSamplerData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; + friend struct QMetalShaderResourceBindings; +}; + +struct QMetalRenderPassDescriptor : public QRhiRenderPassDescriptor { + QMetalRenderPassDescriptor(QRhiImplementation *rhi); + ~QMetalRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + + void updateSerializedFormat(); + + // there is no MTLRenderPassDescriptor here as one will be created for each pass in beginPass() + + // but the things needed for the render pipeline descriptor have to be provided + static const int MAX_COLOR_ATTACHMENTS = 8; + int colorAttachmentCount = 0; + bool hasDepthStencil = false; + int colorFormat[MAX_COLOR_ATTACHMENTS]; + int dsFormat; + QVector<quint32> serializedFormatData; +}; + +struct QMetalRenderTargetData; + +struct QMetalSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QMetalSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QMetalSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QMetalRenderTargetData *d; +}; + +struct QMetalTextureRenderTarget : public QRhiTextureRenderTarget +{ + QMetalTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QMetalTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QMetalRenderTargetData *d; + friend class QRhiMetal; }; -struct Q_GUI_EXPORT QRhiMetalNativeHandles : public QRhiNativeHandles +struct QMetalShaderResourceBindings : public QRhiShaderResourceBindings { - MTLDevice *dev = nullptr; - MTLCommandQueue *cmdQueue = nullptr; + QMetalShaderResourceBindings(QRhiImplementation *rhi); + ~QMetalShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + int maxBinding = -1; + + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData; + + uint generation = 0; + friend class QRhiMetal; }; -struct Q_GUI_EXPORT QRhiMetalCommandBufferNativeHandles : public QRhiNativeHandles +struct QMetalGraphicsPipelineData; +struct QMetalCommandBuffer; + +struct QMetalGraphicsPipeline : public QRhiGraphicsPipeline { - MTLCommandBuffer *commandBuffer = nullptr; - MTLRenderCommandEncoder *encoder = nullptr; + QMetalGraphicsPipeline(QRhiImplementation *rhi); + ~QMetalGraphicsPipeline(); + void destroy() override; + bool create() override; + + void makeActiveForCurrentRenderPassEncoder(QMetalCommandBuffer *cbD); + void setupAttachmentsInMetalRenderPassDescriptor(void *metalRpDesc, QMetalRenderPassDescriptor *rpD); + void setupMetalDepthStencilDescriptor(void *metalDsDesc); + void mapStates(); + bool createVertexFragmentPipeline(); + bool createTessellationPipelines(const QShader &tessVert, const QShader &tesc, const QShader &tese, const QShader &tessFrag); + + QMetalGraphicsPipelineData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalComputePipelineData; + +struct QMetalComputePipeline : public QRhiComputePipeline +{ + QMetalComputePipeline(QRhiImplementation *rhi); + ~QMetalComputePipeline(); + void destroy() override; + bool create() override; + + QMetalComputePipelineData *d; + uint generation = 0; + int lastActiveFrameSlot = -1; + friend class QRhiMetal; +}; + +struct QMetalCommandBufferData; +struct QMetalSwapChain; + +struct QMetalCommandBuffer : public QRhiCommandBuffer +{ + QMetalCommandBuffer(QRhiImplementation *rhi); + ~QMetalCommandBuffer(); + void destroy() override; + + QMetalCommandBufferData *d = nullptr; + QRhiMetalCommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + // per-pass (render or compute command encoder) persistent state + PassType recordingPass; + QRhiRenderTarget *currentTarget; + + // per-pass (render or compute command encoder) volatile (cached) state + QMetalGraphicsPipeline *currentGraphicsPipeline; + QMetalComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QMetalShaderResourceBindings *currentGraphicsSrb; + QMetalShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + int currentResSlot; + QMetalBuffer *currentIndexBuffer; + quint32 currentIndexOffset; + QRhiCommandBuffer::IndexFormat currentIndexFormat; + int currentCullMode; + int currentTriangleFillMode; + int currentFrontFaceWinding; + QPair<float, float> currentDepthBiasValues; + + const QRhiNativeHandles *nativeHandles(); + void resetState(double lastGpuTime = 0); + void resetPerPassState(); + void resetPerPassCachedState(); +}; + +struct QMetalSwapChainData; + +struct QMetalSwapChain : public QRhiSwapChain +{ + QMetalSwapChain(QRhiImplementation *rhi); + ~QMetalSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + + bool createOrResize() override; + + virtual QRhiSwapChainHdrInfo hdrInfo() override; + + void chooseFormats(); + void waitUntilCompleted(int slot); + + QWindow *window = nullptr; + QSize pixelSize; + int currentFrameSlot = 0; // 0..QMTL_FRAMES_IN_FLIGHT-1 + int frameCount = 0; + int samples = 1; + QMetalSwapChainRenderTarget rtWrapper; + QMetalCommandBuffer cbWrapper; + QMetalRenderBuffer *ds = nullptr; + QMetalSwapChainData *d = nullptr; +}; + +struct QRhiMetalData; + +class QRhiMetal : public QRhiImplementation +{ +public: + QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice = nullptr); + ~QRhiMetal(); + + static bool probe(QRhiMetalInitParams *params); + static QRhiSwapChainProxyData updateSwapChainProxyData(QWindow *window); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void executeDeferredReleases(bool forced = false); + void finishActiveReadbacks(bool forced = false); + qsizetype subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; + void enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEncPtr, + int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc, + qsizetype *curOfs); + void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); + void executeBufferHostWritesForSlot(QMetalBuffer *bufD, int slot); + void executeBufferHostWritesForCurrentFrame(QMetalBuffer *bufD); + static const int SUPPORTED_STAGES = 5; + void enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD, + QMetalCommandBuffer *cbD, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets, + bool offsetOnlyChange, + const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[SUPPORTED_STAGES]); + struct TessDrawArgs { + QMetalCommandBuffer *cbD; + enum { + NonIndexed, + U16Indexed, + U32Indexed + } type; + struct NonIndexedArgs { + quint32 vertexCount; + quint32 instanceCount; + quint32 firstVertex; + quint32 firstInstance; + }; + struct IndexedArgs { + quint32 indexCount; + quint32 instanceCount; + quint32 firstIndex; + qint32 vertexOffset; + quint32 firstInstance; + void *indexBuffer; + }; + union { + NonIndexedArgs draw; + IndexedArgs drawIndexed; + }; + }; + void tessellatedDraw(const TessDrawArgs &args); + void adjustForMultiViewDraw(quint32 *instanceCount, QRhiCommandBuffer *cb); + + QRhi::Flags rhiFlags; + bool importedDevice = false; + bool importedCmdQueue = false; + QMetalSwapChain *currentSwapChain = nullptr; + QSet<QMetalSwapChain *> swapchains; + QRhiMetalNativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + quint32 osMajor = 0; + quint32 osMinor = 0; + + struct { + int maxTextureSize = 4096; + bool baseVertexAndInstance = true; + QVector<int> supportedSampleCounts; + bool isAppleGPU = false; + int maxThreadGroupSize = 512; + bool multiView = false; + } caps; + + QRhiMetalData *d = nullptr; }; QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhimetal_p_p.h b/src/gui/rhi/qrhimetal_p_p.h deleted file mode 100644 index c76a29b0a8..0000000000 --- a/src/gui/rhi/qrhimetal_p_p.h +++ /dev/null @@ -1,509 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHIMETAL_P_H -#define QRHIMETAL_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhimetal_p.h" -#include "qrhi_p_p.h" -#include <QWindow> - -QT_BEGIN_NAMESPACE - -static const int QMTL_FRAMES_IN_FLIGHT = 2; - -// have to hide the ObjC stuff, this header cannot contain MTL* at all -struct QMetalBufferData; - -struct QMetalBuffer : public QRhiBuffer -{ - QMetalBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QMetalBuffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - QMetalBufferData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; - - static constexpr int WorkBufPoolUsage = 1 << 8; - static_assert(WorkBufPoolUsage > QRhiBuffer::StorageBuffer); -}; - -struct QMetalRenderBufferData; - -struct QMetalRenderBuffer : public QRhiRenderBuffer -{ - QMetalRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QMetalRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - QMetalRenderBufferData *d; - int samples = 1; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalTextureData; - -struct QMetalTexture : public QRhiTexture -{ - QMetalTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QMetalTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - - QMetalTextureData *d; - int mipLevelCount = 0; - int samples = 1; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; - friend struct QMetalTextureData; -}; - -struct QMetalSamplerData; - -struct QMetalSampler : public QRhiSampler -{ - QMetalSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QMetalSampler(); - void destroy() override; - bool create() override; - - QMetalSamplerData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; - friend struct QMetalShaderResourceBindings; -}; - -struct QMetalRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QMetalRenderPassDescriptor(QRhiImplementation *rhi); - ~QMetalRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; - - void updateSerializedFormat(); - - // there is no MTLRenderPassDescriptor here as one will be created for each pass in beginPass() - - // but the things needed for the render pipeline descriptor have to be provided - static const int MAX_COLOR_ATTACHMENTS = 8; - int colorAttachmentCount = 0; - bool hasDepthStencil = false; - int colorFormat[MAX_COLOR_ATTACHMENTS]; - int dsFormat; - QVector<quint32> serializedFormatData; -}; - -struct QMetalRenderTargetData; - -struct QMetalSwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QMetalSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QMetalSwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QMetalRenderTargetData *d; -}; - -struct QMetalTextureRenderTarget : public QRhiTextureRenderTarget -{ - QMetalTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QMetalTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QMetalRenderTargetData *d; - friend class QRhiMetal; -}; - -struct QMetalShaderResourceBindings : public QRhiShaderResourceBindings -{ - QMetalShaderResourceBindings(QRhiImplementation *rhi); - ~QMetalShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; - - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - int maxBinding = -1; - - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData; - - uint generation = 0; - friend class QRhiMetal; -}; - -struct QMetalGraphicsPipelineData; -struct QMetalCommandBuffer; - -struct QMetalGraphicsPipeline : public QRhiGraphicsPipeline -{ - QMetalGraphicsPipeline(QRhiImplementation *rhi); - ~QMetalGraphicsPipeline(); - void destroy() override; - bool create() override; - - void makeActiveForCurrentRenderPassEncoder(QMetalCommandBuffer *cbD); - void setupAttachmentsInMetalRenderPassDescriptor(void *metalRpDesc, QMetalRenderPassDescriptor *rpD); - void setupMetalDepthStencilDescriptor(void *metalDsDesc); - void mapStates(); - bool createVertexFragmentPipeline(); - bool createTessellationPipelines(const QShader &tessVert, const QShader &tesc, const QShader &tese, const QShader &tessFrag); - - QMetalGraphicsPipelineData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalComputePipelineData; - -struct QMetalComputePipeline : public QRhiComputePipeline -{ - QMetalComputePipeline(QRhiImplementation *rhi); - ~QMetalComputePipeline(); - void destroy() override; - bool create() override; - - QMetalComputePipelineData *d; - uint generation = 0; - int lastActiveFrameSlot = -1; - friend class QRhiMetal; -}; - -struct QMetalCommandBufferData; -struct QMetalSwapChain; - -struct QMetalCommandBuffer : public QRhiCommandBuffer -{ - QMetalCommandBuffer(QRhiImplementation *rhi); - ~QMetalCommandBuffer(); - void destroy() override; - - QMetalCommandBufferData *d = nullptr; - QRhiMetalCommandBufferNativeHandles nativeHandlesStruct; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - // per-pass (render or compute command encoder) persistent state - PassType recordingPass; - QRhiRenderTarget *currentTarget; - - // per-pass (render or compute command encoder) volatile (cached) state - QMetalGraphicsPipeline *currentGraphicsPipeline; - QMetalComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QMetalShaderResourceBindings *currentGraphicsSrb; - QMetalShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - int currentResSlot; - QMetalBuffer *currentIndexBuffer; - quint32 currentIndexOffset; - QRhiCommandBuffer::IndexFormat currentIndexFormat; - int currentCullMode; - int currentTriangleFillMode; - int currentFrontFaceWinding; - QPair<float, float> currentDepthBiasValues; - - const QRhiNativeHandles *nativeHandles(); - void resetState(); - void resetPerPassState(); - void resetPerPassCachedState(); -}; - -struct QMetalSwapChainData; - -struct QMetalSwapChain : public QRhiSwapChain -{ - QMetalSwapChain(QRhiImplementation *rhi); - ~QMetalSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - - bool createOrResize() override; - - virtual QRhiSwapChainHdrInfo hdrInfo() override; - - void chooseFormats(); - void waitUntilCompleted(int slot); - - QWindow *window = nullptr; - QSize pixelSize; - int currentFrameSlot = 0; // 0..QMTL_FRAMES_IN_FLIGHT-1 - int frameCount = 0; - int samples = 1; - QMetalSwapChainRenderTarget rtWrapper; - QMetalCommandBuffer cbWrapper; - QMetalRenderBuffer *ds = nullptr; - QMetalSwapChainData *d = nullptr; -}; - -struct QRhiMetalData; - -class QRhiMetal : public QRhiImplementation -{ -public: - QRhiMetal(QRhiMetalInitParams *params, QRhiMetalNativeHandles *importDevice = nullptr); - ~QRhiMetal(); - - static bool probe(QRhiMetalInitParams *params); - static QRhiSwapChainProxyData updateSwapChainProxyData(QWindow *window); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void executeDeferredReleases(bool forced = false); - void finishActiveReadbacks(bool forced = false); - qsizetype subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; - void enqueueSubresUpload(QMetalTexture *texD, void *mp, void *blitEncPtr, - int layer, int level, const QRhiTextureSubresourceUploadDescription &subresDesc, - qsizetype *curOfs); - void enqueueResourceUpdates(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates); - void executeBufferHostWritesForSlot(QMetalBuffer *bufD, int slot); - void executeBufferHostWritesForCurrentFrame(QMetalBuffer *bufD); - static const int SUPPORTED_STAGES = 5; - void enqueueShaderResourceBindings(QMetalShaderResourceBindings *srbD, - QMetalCommandBuffer *cbD, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets, - bool offsetOnlyChange, - const QShader::NativeResourceBindingMap *nativeResourceBindingMaps[SUPPORTED_STAGES]); - int effectiveSampleCount(int sampleCount) const; - struct TessDrawArgs { - QMetalCommandBuffer *cbD; - enum { - NonIndexed, - U16Indexed, - U32Indexed - } type; - struct NonIndexedArgs { - quint32 vertexCount; - quint32 instanceCount; - quint32 firstVertex; - quint32 firstInstance; - }; - struct IndexedArgs { - quint32 indexCount; - quint32 instanceCount; - quint32 firstIndex; - qint32 vertexOffset; - quint32 firstInstance; - void *indexBuffer; - }; - union { - NonIndexedArgs draw; - IndexedArgs drawIndexed; - }; - }; - void tessellatedDraw(const TessDrawArgs &args); - - QRhi::Flags rhiFlags; - bool importedDevice = false; - bool importedCmdQueue = false; - QMetalSwapChain *currentSwapChain = nullptr; - QSet<QMetalSwapChain *> swapchains; - QRhiMetalNativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - quint32 osMajor = 0; - quint32 osMinor = 0; - - struct { - int maxTextureSize = 4096; - bool baseVertexAndInstance = true; - QVector<int> supportedSampleCounts; - bool isAppleGPU = false; - int maxThreadGroupSize = 512; - } caps; - - QRhiMetalData *d = nullptr; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhinull.cpp b/src/gui/rhi/qrhinull.cpp index ec1b5aa64a..566b922c1b 100644 --- a/src/gui/rhi/qrhinull.cpp +++ b/src/gui/rhi/qrhinull.cpp @@ -1,7 +1,7 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhinull_p_p.h" +#include "qrhinull_p.h" #include <qmath.h> #include <QPainter> @@ -9,10 +9,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiNullInitParams - \internal \inmodule QtGui + \since 6.6 \brief Null backend specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Null QRhi needs no special parameters for initialization. \badcode @@ -28,9 +31,12 @@ QT_BEGIN_NAMESPACE /*! \class QRhiNullNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Empty. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ QRhiNull::QRhiNull(QRhiNullInitParams *params) @@ -344,6 +350,12 @@ void QRhiNull::endExternal(QRhiCommandBuffer *cb) Q_UNUSED(cb); } +double QRhiNull::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + Q_UNUSED(cb); + return 0; +} + QRhi::FrameOpResult QRhiNull::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) { Q_UNUSED(flags); @@ -456,7 +468,7 @@ void QRhiNull::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *re QNullBuffer *bufD = QRHI_RES(QNullBuffer, u.buf); memcpy(bufD->data + u.offset, u.data.constData(), size_t(u.data.size())); } else if (u.type == QRhiResourceUpdateBatchPrivate::BufferOp::Read) { - QRhiBufferReadbackResult *result = u.result; + QRhiReadbackResult *result = u.result; result->data.resize(u.readSize); QNullBuffer *bufD = QRHI_RES(QNullBuffer, u.buf); memcpy(result->data.data(), bufD->data + u.offset, size_t(u.readSize)); @@ -664,10 +676,11 @@ bool QNullTexture::create() const bool is1D = m_flags.testFlags(OneDimensional); QSize size = is1D ? QSize(qMax(1, m_pixelSize.width()), 1) : (m_pixelSize.isEmpty() ? QSize(1, 1) : m_pixelSize); - m_depth = qMax(1, m_depth); const int mipLevelCount = hasMipMaps ? rhiD->q->mipLevelsForSize(size) : 1; - m_arraySize = qMax(0, m_arraySize); - const int layerCount = is3D ? m_depth : (isCube ? 6 : (isArray ? m_arraySize : 1)); + const int layerCount = is3D ? qMax(1, m_depth) + : (isCube ? 6 + : (isArray ? qMax(0, m_arraySize) + : 1)); if (m_format == RGBA8) { image.resize(layerCount); @@ -716,10 +729,15 @@ QNullSampler::~QNullSampler() void QNullSampler::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullSampler::create() { + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); return true; } @@ -735,6 +753,9 @@ QNullRenderPassDescriptor::~QNullRenderPassDescriptor() void QNullRenderPassDescriptor::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other) const @@ -745,7 +766,10 @@ bool QNullRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *oth QRhiRenderPassDescriptor *QNullRenderPassDescriptor::newCompatibleRenderPassDescriptor() const { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; } QVector<quint32> QNullRenderPassDescriptor::serializedFormat() const @@ -798,11 +822,17 @@ QNullTextureRenderTarget::~QNullTextureRenderTarget() void QNullTextureRenderTarget::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } QRhiRenderPassDescriptor *QNullTextureRenderTarget::newCompatibleRenderPassDescriptor() { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; } bool QNullTextureRenderTarget::create() @@ -820,6 +850,7 @@ bool QNullTextureRenderTarget::create() d.pixelSize = m_desc.depthTexture()->pixelSize(); } QRhiRenderTargetAttachmentTracker::updateResIdList<QNullTexture, QNullRenderBuffer>(m_desc, &d.currentResIdList); + rhiD->registerResource(this); return true; } @@ -853,6 +884,9 @@ QNullShaderResourceBindings::~QNullShaderResourceBindings() void QNullShaderResourceBindings::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullShaderResourceBindings::create() @@ -863,6 +897,7 @@ bool QNullShaderResourceBindings::create() rhiD->updateLayoutDesc(this); + rhiD->registerResource(this, false); return true; } @@ -883,6 +918,9 @@ QNullGraphicsPipeline::~QNullGraphicsPipeline() void QNullGraphicsPipeline::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullGraphicsPipeline::create() @@ -891,6 +929,7 @@ bool QNullGraphicsPipeline::create() if (!rhiD->sanityCheckGraphicsPipeline(this)) return false; + rhiD->registerResource(this); return true; } @@ -906,10 +945,15 @@ QNullComputePipeline::~QNullComputePipeline() void QNullComputePipeline::destroy() { + QRHI_RES_RHI(QRhiNull); + if (rhiD) + rhiD->unregisterResource(this); } bool QNullComputePipeline::create() { + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(this); return true; } @@ -969,7 +1013,10 @@ bool QNullSwapChain::isFormatSupported(Format f) QRhiRenderPassDescriptor *QNullSwapChain::newCompatibleRenderPassDescriptor() { - return new QNullRenderPassDescriptor(m_rhi); + QNullRenderPassDescriptor *rpD = new QNullRenderPassDescriptor(m_rhi); + QRHI_RES_RHI(QRhiNull); + rhiD->registerResource(rpD, false); + return rpD; } bool QNullSwapChain::createOrResize() diff --git a/src/gui/rhi/qrhinull_p.h b/src/gui/rhi/qrhinull_p.h index 061b6eba4a..fc266b4f38 100644 --- a/src/gui/rhi/qrhinull_p.h +++ b/src/gui/rhi/qrhinull_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHINULL_H -#define QRHINULL_H +#ifndef QRHINULL_P_H +#define QRHINULL_P_H // // W A R N I N G @@ -15,16 +15,279 @@ // We mean it. // -#include <private/qrhi_p.h> +#include "qrhi_p.h" QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiNullInitParams : public QRhiInitParams +struct QNullBuffer : public QRhiBuffer { + QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QNullBuffer(); + void destroy() override; + bool create() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + + char *data = nullptr; +}; + +struct QNullRenderBuffer : public QRhiRenderBuffer +{ + QNullRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint); + ~QNullRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + bool valid = false; + uint generation = 0; +}; + +struct QNullTexture : public QRhiTexture +{ + QNullTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QNullTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + + bool valid = false; + QVarLengthArray<std::array<QImage, QRhi::MAX_MIP_LEVELS>, 6> image; + uint generation = 0; +}; + +struct QNullSampler : public QRhiSampler +{ + QNullSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QNullSampler(); + void destroy() override; + bool create() override; +}; + +struct QNullRenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QNullRenderPassDescriptor(QRhiImplementation *rhi); + ~QNullRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; +}; + +struct QNullRenderTargetData +{ + QNullRenderTargetData(QRhiImplementation *) { } + + QNullRenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; +}; + +struct QNullSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QNullSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QNullSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QNullRenderTargetData d; +}; + +struct QNullTextureRenderTarget : public QRhiTextureRenderTarget +{ + QNullTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QNullTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QNullRenderTargetData d; +}; + +struct QNullShaderResourceBindings : public QRhiShaderResourceBindings +{ + QNullShaderResourceBindings(QRhiImplementation *rhi); + ~QNullShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; +}; + +struct QNullGraphicsPipeline : public QRhiGraphicsPipeline +{ + QNullGraphicsPipeline(QRhiImplementation *rhi); + ~QNullGraphicsPipeline(); + void destroy() override; + bool create() override; +}; + +struct QNullComputePipeline : public QRhiComputePipeline +{ + QNullComputePipeline(QRhiImplementation *rhi); + ~QNullComputePipeline(); + void destroy() override; + bool create() override; }; -struct Q_GUI_EXPORT QRhiNullNativeHandles : public QRhiNativeHandles +struct QNullCommandBuffer : public QRhiCommandBuffer { + QNullCommandBuffer(QRhiImplementation *rhi); + ~QNullCommandBuffer(); + void destroy() override; +}; + +struct QNullSwapChain : public QRhiSwapChain +{ + QNullSwapChain(QRhiImplementation *rhi); + ~QNullSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + QWindow *window = nullptr; + QNullSwapChainRenderTarget rt; + QNullCommandBuffer cb; + int frameCount = 0; +}; + +class QRhiNull : public QRhiImplementation +{ +public: + QRhiNull(QRhiNullInitParams *params); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + void simulateTextureUpload(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + void simulateTextureCopy(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + void simulateTextureGenMips(const QRhiResourceUpdateBatchPrivate::TextureOp &u); + + QRhiNullNativeHandles nativeHandlesStruct; + QRhiSwapChain *currentSwapChain = nullptr; + QNullCommandBuffer offscreenCommandBuffer; }; QT_END_NAMESPACE diff --git a/src/gui/rhi/qrhinull_p_p.h b/src/gui/rhi/qrhinull_p_p.h deleted file mode 100644 index 2ace5a36b6..0000000000 --- a/src/gui/rhi/qrhinull_p_p.h +++ /dev/null @@ -1,295 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHINULL_P_H -#define QRHINULL_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhinull_p.h" -#include "qrhi_p_p.h" - -QT_BEGIN_NAMESPACE - -struct QNullBuffer : public QRhiBuffer -{ - QNullBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QNullBuffer(); - void destroy() override; - bool create() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - - char *data = nullptr; -}; - -struct QNullRenderBuffer : public QRhiRenderBuffer -{ - QNullRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint); - ~QNullRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - bool valid = false; - uint generation = 0; -}; - -struct QNullTexture : public QRhiTexture -{ - QNullTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QNullTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - - bool valid = false; - QVarLengthArray<std::array<QImage, QRhi::MAX_MIP_LEVELS>, 6> image; - uint generation = 0; -}; - -struct QNullSampler : public QRhiSampler -{ - QNullSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QNullSampler(); - void destroy() override; - bool create() override; -}; - -struct QNullRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QNullRenderPassDescriptor(QRhiImplementation *rhi); - ~QNullRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; -}; - -struct QNullRenderTargetData -{ - QNullRenderTargetData(QRhiImplementation *) { } - - QNullRenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; -}; - -struct QNullSwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QNullSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QNullSwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QNullRenderTargetData d; -}; - -struct QNullTextureRenderTarget : public QRhiTextureRenderTarget -{ - QNullTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QNullTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QNullRenderTargetData d; -}; - -struct QNullShaderResourceBindings : public QRhiShaderResourceBindings -{ - QNullShaderResourceBindings(QRhiImplementation *rhi); - ~QNullShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; -}; - -struct QNullGraphicsPipeline : public QRhiGraphicsPipeline -{ - QNullGraphicsPipeline(QRhiImplementation *rhi); - ~QNullGraphicsPipeline(); - void destroy() override; - bool create() override; -}; - -struct QNullComputePipeline : public QRhiComputePipeline -{ - QNullComputePipeline(QRhiImplementation *rhi); - ~QNullComputePipeline(); - void destroy() override; - bool create() override; -}; - -struct QNullCommandBuffer : public QRhiCommandBuffer -{ - QNullCommandBuffer(QRhiImplementation *rhi); - ~QNullCommandBuffer(); - void destroy() override; -}; - -struct QNullSwapChain : public QRhiSwapChain -{ - QNullSwapChain(QRhiImplementation *rhi); - ~QNullSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - QWindow *window = nullptr; - QNullSwapChainRenderTarget rt; - QNullCommandBuffer cb; - int frameCount = 0; -}; - -class QRhiNull : public QRhiImplementation -{ -public: - QRhiNull(QRhiNullInitParams *params); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - void simulateTextureUpload(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - void simulateTextureCopy(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - void simulateTextureGenMips(const QRhiResourceUpdateBatchPrivate::TextureOp &u); - - QRhiNullNativeHandles nativeHandlesStruct; - QRhiSwapChain *currentSwapChain = nullptr; - QNullCommandBuffer offscreenCommandBuffer; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhivulkan.cpp b/src/gui/rhi/qrhivulkan.cpp index b2f72dce49..3dd3c57bd4 100644 --- a/src/gui/rhi/qrhivulkan.cpp +++ b/src/gui/rhi/qrhivulkan.cpp @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qrhivulkan_p_p.h" -#include "qrhivulkanext_p.h" +#include "qrhivulkan_p.h" +#include <qpa/qplatformvulkaninstance.h> #define VMA_IMPLEMENTATION #define VMA_DYNAMIC_VULKAN_FUNCTIONS 1 @@ -59,10 +59,13 @@ QT_BEGIN_NAMESPACE /*! \class QRhiVulkanInitParams - \internal \inmodule QtGui + \since 6.6 \brief Vulkan specific initialization parameters. + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. + A Vulkan-based QRhi needs at minimum a valid QVulkanInstance. It is up to the user to ensure this is available and initialized. This is typically done in main() similarly to the following: @@ -165,18 +168,87 @@ QT_BEGIN_NAMESPACE */ /*! + \variable QRhiVulkanInitParams::inst + + The QVulkanInstance that has already been successfully + \l{QVulkanInstance::create()}{created}, required. +*/ + +/*! + \variable QRhiVulkanInitParams::window + + Optional, but recommended when targeting a QWindow. +*/ + +/*! + \variable QRhiVulkanInitParams::deviceExtensions + + Optional, empty by default. The list of Vulkan device extensions to enable. + Unsupported extensions are ignored gracefully. +*/ + +/*! \class QRhiVulkanNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Collects device, queue, and other Vulkan objects that are used by the QRhi. \note Ownership of the Vulkan objects is never transferred. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiVulkanNativeHandles::physDev + + When different from \nullptr, specifies the Vulkan physical device to use. +*/ + +/*! + \variable QRhiVulkanNativeHandles::dev + + When wanting to import not just a physical device, but also use an already + existing VkDevice, set this and the graphics queue index and family index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueueFamilyIdx + + Graphics queue family index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueueIdx + + Graphics queue index. +*/ + +/*! + \variable QRhiVulkanNativeHandles::vmemAllocator + + Relevant only when importing an existing memory allocator object, + leave it set to \nullptr otherwise. +*/ + +/*! + \variable QRhiVulkanNativeHandles::gfxQueue + + Output only, not used by QRhi::create(), only set by the + QRhi::nativeHandles() accessor. The graphics VkQueue used by the QRhi. +*/ + +/*! + \variable QRhiVulkanNativeHandles::inst + + Output only, not used by QRhi::create(), only set by the + QRhi::nativeHandles() accessor. The QVulkanInstance used by the QRhi. +*/ + +/*! \class QRhiVulkanCommandBufferNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the Vulkan command buffer object that is backing a QRhiCommandBuffer. \note The Vulkan command buffer object is only guaranteed to be valid, and @@ -184,15 +256,33 @@ QT_BEGIN_NAMESPACE \l{QRhi::beginFrame()}{beginFrame()} - \l{QRhi::endFrame()}{endFrame()} or \l{QRhi::beginOffscreenFrame()}{beginOffscreenFrame()} - \l{QRhi::endOffscreenFrame()}{endOffscreenFrame()} pair. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ /*! + \variable QRhiVulkanCommandBufferNativeHandles::commandBuffer + + The VkCommandBuffer object. +*/ + +/*! \class QRhiVulkanRenderPassNativeHandles - \internal \inmodule QtGui + \since 6.6 \brief Holds the Vulkan render pass object backing a QRhiRenderPassDescriptor. + + \note This is a RHI API with limited compatibility guarantees, see \l QRhi + for details. */ +/*! + \variable QRhiVulkanRenderPassNativeHandles::renderPass + + The VkRenderPass object. +*/ + template <class Int> inline Int aligned(Int v, Int byteAlign) { @@ -221,6 +311,13 @@ static inline VmaAllocator toVmaAllocator(QVkAllocator a) return reinterpret_cast<VmaAllocator>(a); } +/*! + \return the list of instance extensions that are expected to be enabled on + the QVulkanInstance that is used for the Vulkan-based QRhi. + + The returned list can be safely passed to QVulkanInstance::setExtensions() + as-is, because unsupported extensions are filtered out automatically. + */ QByteArrayList QRhiVulkanInitParams::preferredInstanceExtensions() { return { @@ -228,11 +325,18 @@ QByteArrayList QRhiVulkanInitParams::preferredInstanceExtensions() }; } +/*! + \return the list of device extensions that are expected to be enabled on the + \c VkDevice when creating a Vulkan-based QRhi with an externally created + \c VkDevice object. + */ QByteArrayList QRhiVulkanInitParams::preferredExtensionsForImportedDevice() { return { QByteArrayLiteral("VK_KHR_swapchain"), - QByteArrayLiteral("VK_EXT_vertex_attribute_divisor") + QByteArrayLiteral("VK_EXT_vertex_attribute_divisor"), + QByteArrayLiteral("VK_KHR_create_renderpass2"), + QByteArrayLiteral("VK_KHR_depth_stencil_resolve") }; } @@ -326,6 +430,8 @@ bool QRhiVulkan::create(QRhi::Flags flags) for (const char *ext : inst->extensions()) qCDebug(QRHI_LOG_INFO, " %s", ext); } + + caps = {}; caps.debugUtils = inst->extensions().contains(QByteArrayLiteral("VK_EXT_debug_utils")); QList<VkQueueFamilyProperties> queueFamilyProps; @@ -427,7 +533,65 @@ bool QRhiVulkan::create(QRhi::Flags flags) driverInfoStruct.vendorId = physDevProperties.vendorID; driverInfoStruct.deviceType = toRhiDeviceType(physDevProperties.deviceType); - f->vkGetPhysicalDeviceFeatures(physDev, &physDevFeatures); + bool featuresQueried = false; +#ifdef VK_VERSION_1_1 + VkPhysicalDeviceFeatures2 physDevFeaturesChainable = {}; + physDevFeaturesChainable.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; +#endif + + // Vulkan >=1.2 headers at build time, >=1.2 implementation at run time +#ifdef VK_VERSION_1_2 + if (!featuresQueried) { + // Vulkan11Features, Vulkan12Features, etc. are only in Vulkan 1.2 and newer. + if (caps.apiVersion >= QVersionNumber(1, 2)) { + physDevFeatures11IfApi12OrNewer = {}; + physDevFeatures11IfApi12OrNewer.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES; + physDevFeatures12 = {}; + physDevFeatures12.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES; +#ifdef VK_VERSION_1_3 + physDevFeatures13 = {}; + physDevFeatures13.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_3_FEATURES; +#endif + physDevFeaturesChainable.pNext = &physDevFeatures11IfApi12OrNewer; + physDevFeatures11IfApi12OrNewer.pNext = &physDevFeatures12; +#ifdef VK_VERSION_1_3 + if (caps.apiVersion >= QVersionNumber(1, 3)) + physDevFeatures12.pNext = &physDevFeatures13; +#endif + f->vkGetPhysicalDeviceFeatures2(physDev, &physDevFeaturesChainable); + memcpy(&physDevFeatures, &physDevFeaturesChainable.features, sizeof(VkPhysicalDeviceFeatures)); + featuresQueried = true; + } + } +#endif // VK_VERSION_1_2 + + // Vulkan >=1.1 headers at build time, 1.1 implementation at run time +#ifdef VK_VERSION_1_1 + if (!featuresQueried) { + // Vulkan versioning nightmares: if the runtime API version is 1.1, + // there is no Vulkan11Features (introduced in 1.2+, the headers might + // have the types and structs, but the Vulkan implementation version at + // run time is what matters). But there are individual feature structs. + // For multiview, it is important to get this right since at the time of + // writing Quest 3 Android is a Vulkan 1.1 implementation at run time on + // the headset. + if (caps.apiVersion == QVersionNumber(1, 1)) { + multiviewFeaturesIfApi11 = {}; + multiviewFeaturesIfApi11.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES; + physDevFeaturesChainable.pNext = &multiviewFeaturesIfApi11; + f->vkGetPhysicalDeviceFeatures2(physDev, &physDevFeaturesChainable); + memcpy(&physDevFeatures, &physDevFeaturesChainable.features, sizeof(VkPhysicalDeviceFeatures)); + featuresQueried = true; + } + } +#endif + + if (!featuresQueried) { + // If the API version at run time is 1.0 (or we are building with + // ancient 1.0 headers), then do the Vulkan 1.0 query. + f->vkGetPhysicalDeviceFeatures(physDev, &physDevFeatures); + featuresQueried = true; + } // Choose queue and create device, unless the device was specified in importParams. if (!importedDevice) { @@ -499,13 +663,28 @@ bool QRhiVulkan::create(QRhi::Flags flags) } } - caps.vertexAttribDivisor = false; +#ifdef VK_EXT_vertex_attribute_divisor if (devExts.contains(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME)) { if (hasPhysDevProp2) { requestedDevExts.append(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME); caps.vertexAttribDivisor = true; } } +#endif + +#ifdef VK_KHR_create_renderpass2 + if (devExts.contains(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME)) { + requestedDevExts.append(VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME); + caps.renderPass2KHR = true; + } +#endif + +#ifdef VK_KHR_depth_stencil_resolve + if (devExts.contains(VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME)) { + requestedDevExts.append(VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME); + caps.depthStencilResolveKHR = true; + } +#endif for (const QByteArray &ext : requestedDeviceExtensions) { if (!ext.isEmpty() && !requestedDevExts.contains(ext)) { @@ -558,42 +737,27 @@ bool QRhiVulkan::create(QRhi::Flags flags) // tessellationShader, geometryShader // textureCompressionETC2, textureCompressionASTC_LDR, textureCompressionBC -#ifdef VK_VERSION_1_2 // Vulkan11Features is only in Vulkan 1.2 - VkPhysicalDeviceFeatures2 physDevFeatures2 = {}; - physDevFeatures2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - - VkPhysicalDeviceVulkan11Features features11 = {}; - features11.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES; - VkPhysicalDeviceVulkan12Features features12 = {}; - features12.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES; -#ifdef VK_VERSION_1_3 - VkPhysicalDeviceVulkan13Features features13 = {}; - features13.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_3_FEATURES; +#ifdef VK_VERSION_1_1 + physDevFeaturesChainable.features.robustBufferAccess = VK_FALSE; #endif - - if (caps.apiVersion >= QVersionNumber(1, 2)) { - physDevFeatures2.pNext = &features11; - features11.pNext = &features12; #ifdef VK_VERSION_1_3 - if (caps.apiVersion >= QVersionNumber(1, 3)) - features12.pNext = &features13; + physDevFeatures13.robustImageAccess = VK_FALSE; #endif - f->vkGetPhysicalDeviceFeatures2(physDev, &physDevFeatures2); - physDevFeatures2.features.robustBufferAccess = VK_FALSE; -#ifdef VK_VERSION_1_3 - features13.robustImageAccess = VK_FALSE; +#ifdef VK_VERSION_1_1 + if (caps.apiVersion >= QVersionNumber(1, 1)) { + // For a >=1.2 implementation at run time, this will enable all + // (1.0-1.3) features reported as supported, except the ones we turn + // off explicitly above. For a 1.1 implementation at run time, this + // only enables the 1.0 and multiview features reported as + // supported. We will not be bothering with the Vulkan 1.1 + // individual feature struct nonsense. + devInfo.pNext = &physDevFeaturesChainable; + } else #endif - - devInfo.pNext = &physDevFeatures2; - } -#endif // VK_VERSION_1_2 - - VkPhysicalDeviceFeatures features; - if (!devInfo.pNext) { - memcpy(&features, &physDevFeatures, sizeof(features)); - features.robustBufferAccess = VK_FALSE; - devInfo.pEnabledFeatures = &features; + { + physDevFeatures.robustBufferAccess = VK_FALSE; + devInfo.pEnabledFeatures = &physDevFeatures; } VkResult err = f->vkCreateDevice(physDev, &devInfo, nullptr, &dev); @@ -603,6 +767,13 @@ bool QRhiVulkan::create(QRhi::Flags flags) } } else { qCDebug(QRHI_LOG_INFO, "Using imported device %p", dev); + + // Here we have no way to tell if the extensions got enabled or not. + // Pretend it's all there and supported. If getProcAddress fails, we'll + // handle that gracefully. + caps.vertexAttribDivisor = true; + caps.renderPass2KHR = true; + caps.depthStencilResolveKHR = true; } vkGetPhysicalDeviceSurfaceCapabilitiesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR>( @@ -611,13 +782,6 @@ bool QRhiVulkan::create(QRhi::Flags flags) inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfaceFormatsKHR")); vkGetPhysicalDeviceSurfacePresentModesKHR = reinterpret_cast<PFN_vkGetPhysicalDeviceSurfacePresentModesKHR>( inst->getInstanceProcAddr("vkGetPhysicalDeviceSurfacePresentModesKHR")); - if (!vkGetPhysicalDeviceSurfaceCapabilitiesKHR - || !vkGetPhysicalDeviceSurfaceFormatsKHR - || !vkGetPhysicalDeviceSurfacePresentModesKHR) - { - qWarning("Physical device surface queries not available"); - return false; - } df = inst->deviceFunctions(dev); @@ -657,6 +821,28 @@ bool QRhiVulkan::create(QRhi::Flags flags) caps.nonFillPolygonMode = physDevFeatures.fillModeNonSolid; +#ifdef VK_VERSION_1_2 + if (caps.apiVersion >= QVersionNumber(1, 2)) + caps.multiView = physDevFeatures11IfApi12OrNewer.multiview; +#endif + +#ifdef VK_VERSION_1_1 + if (caps.apiVersion == QVersionNumber(1, 1)) + caps.multiView = multiviewFeaturesIfApi11.multiview; +#endif + + // With Vulkan 1.2 renderpass2 and depth_stencil_resolve are core, but we + // have to support the case of 1.1 + extensions, in particular for the Quest + // 3 (Android, Vulkan 1.1 at the time of writing). Therefore, always rely on + // the KHR extension for now. +#ifdef VK_KHR_create_renderpass2 + if (caps.renderPass2KHR) { + vkCreateRenderPass2KHR = reinterpret_cast<PFN_vkCreateRenderPass2KHR>(f->vkGetDeviceProcAddr(dev, "vkCreateRenderPass2KHR")); + if (!vkCreateRenderPass2KHR) // handle it gracefully, the caps flag may be incorrect when using an imported VkDevice + caps.renderPass2KHR = false; + } +#endif + if (!importedAllocator) { VmaVulkanFunctions funcs = {}; funcs.vkGetInstanceProcAddr = wrap_vkGetInstanceProcAddr; @@ -670,12 +856,9 @@ bool QRhiVulkan::create(QRhi::Flags flags) allocatorInfo.device = dev; allocatorInfo.pVulkanFunctions = &funcs; allocatorInfo.instance = inst->vkInstance(); - const QVersionNumber apiVer = inst->apiVersion(); - if (!apiVer.isNull()) { - allocatorInfo.vulkanApiVersion = VK_MAKE_VERSION(apiVer.majorVersion(), - apiVer.minorVersion(), - apiVer.microVersion()); - } + allocatorInfo.vulkanApiVersion = VK_MAKE_VERSION(caps.apiVersion.majorVersion(), + caps.apiVersion.minorVersion(), + caps.apiVersion.microVersion()); VmaAllocator vmaallocator; VkResult err = vmaCreateAllocator(&allocatorInfo, &vmaallocator); if (err != VK_SUCCESS) { @@ -723,6 +906,7 @@ bool QRhiVulkan::create(QRhi::Flags flags) nativeHandlesStruct.gfxQueueIdx = gfxQueueIdx; nativeHandlesStruct.gfxQueue = gfxQueue; nativeHandlesStruct.vmemAllocator = allocator; + nativeHandlesStruct.inst = inst; return true; } @@ -1183,6 +1367,8 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD rpD->colorRefs.append({ 0, VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }); rpD->hasDepthStencil = hasDepthStencil; + rpD->hasDepthStencilResolve = false; + rpD->multiViewCount = 0; if (hasDepthStencil) { // clear on load + no store + lazy alloc + transient image should play @@ -1252,28 +1438,190 @@ bool QRhiVulkan::createDefaultRenderPass(QVkRenderPassDescriptor *rpD, bool hasD return true; } +struct MultiViewRenderPassSetupHelper +{ + bool prepare(VkRenderPassCreateInfo *rpInfo, int multiViewCount, bool multiViewCap) + { + if (multiViewCount < 2) + return true; + if (!multiViewCap) { + qWarning("Cannot create multiview render pass without support for the Vulkan 1.1 multiview feature"); + return false; + } +#ifdef VK_VERSION_1_1 + uint32_t allViewsMask = 0; + for (uint32_t i = 0; i < uint32_t(multiViewCount); ++i) + allViewsMask |= (1 << i); + multiViewMask = allViewsMask; + multiViewCorrelationMask = allViewsMask; + multiViewInfo.sType = VK_STRUCTURE_TYPE_RENDER_PASS_MULTIVIEW_CREATE_INFO; + multiViewInfo.subpassCount = 1; + multiViewInfo.pViewMasks = &multiViewMask; + multiViewInfo.correlationMaskCount = 1; + multiViewInfo.pCorrelationMasks = &multiViewCorrelationMask; + rpInfo->pNext = &multiViewInfo; +#endif + return true; + } + +#ifdef VK_VERSION_1_1 + VkRenderPassMultiviewCreateInfo multiViewInfo = {}; + uint32_t multiViewMask = 0; + uint32_t multiViewCorrelationMask = 0; +#endif +}; + +#ifdef VK_KHR_create_renderpass2 +// Effectively converts a VkRenderPassCreateInfo into a VkRenderPassCreateInfo2, +// adding depth-stencil resolve support. Assumes a single subpass and no subpass +// dependencies. +struct RenderPass2SetupHelper +{ + bool prepare(VkRenderPassCreateInfo2 *rpInfo2, const VkRenderPassCreateInfo *rpInfo, const QVkRenderPassDescriptor *rpD, int multiViewCount) { + *rpInfo2 = {}; + + viewMask = 0; + if (multiViewCount >= 2) { + for (uint32_t i = 0; i < uint32_t(multiViewCount); ++i) + viewMask |= (1 << i); + } + + attDescs2.resize(rpInfo->attachmentCount); + for (qsizetype i = 0; i < attDescs2.count(); ++i) { + VkAttachmentDescription2KHR &att2(attDescs2[i]); + const VkAttachmentDescription &att(rpInfo->pAttachments[i]); + att2 = {}; + att2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2; + att2.flags = att.flags; + att2.format = att.format; + att2.samples = att.samples; + att2.loadOp = att.loadOp; + att2.storeOp = att.storeOp; + att2.stencilLoadOp = att.stencilLoadOp; + att2.stencilStoreOp = att.stencilStoreOp; + att2.initialLayout = att.initialLayout; + att2.finalLayout = att.finalLayout; + } + + attRefs2.clear(); + subpass2 = {}; + subpass2.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2_KHR; + const VkSubpassDescription &subpassDesc(rpInfo->pSubpasses[0]); + subpass2.flags = subpassDesc.flags; + subpass2.pipelineBindPoint = subpassDesc.pipelineBindPoint; + if (multiViewCount >= 2) + subpass2.viewMask = viewMask; + + // color attachment refs + qsizetype startIndex = attRefs2.count(); + for (uint32_t j = 0; j < subpassDesc.colorAttachmentCount; ++j) { + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(subpassDesc.pColorAttachments[j]); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + } + subpass2.colorAttachmentCount = subpassDesc.colorAttachmentCount; + subpass2.pColorAttachments = attRefs2.constData() + startIndex; + + // color resolve refs + if (subpassDesc.pResolveAttachments) { + startIndex = attRefs2.count(); + for (uint32_t j = 0; j < subpassDesc.colorAttachmentCount; ++j) { + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(subpassDesc.pResolveAttachments[j]); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + } + subpass2.pResolveAttachments = attRefs2.constData() + startIndex; + } + + // depth-stencil ref + if (subpassDesc.pDepthStencilAttachment) { + startIndex = attRefs2.count(); + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + const VkAttachmentReference &attref(*subpassDesc.pDepthStencilAttachment); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = attref.attachment; + attref2.layout = attref.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + subpass2.pDepthStencilAttachment = attRefs2.constData() + startIndex; + } + + // depth-stencil resolve ref +#ifdef VK_KHR_depth_stencil_resolve + dsResolveDesc = {}; + if (rpD->hasDepthStencilResolve) { + startIndex = attRefs2.count(); + attRefs2.append({}); + VkAttachmentReference2KHR &attref2(attRefs2.last()); + attref2.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2_KHR; + attref2.attachment = rpD->dsResolveRef.attachment; + attref2.layout = rpD->dsResolveRef.layout; + attref2.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + dsResolveDesc.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE_KHR; + dsResolveDesc.depthResolveMode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT; + dsResolveDesc.stencilResolveMode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT; + dsResolveDesc.pDepthStencilResolveAttachment = attRefs2.constData() + startIndex; + subpass2.pNext = &dsResolveDesc; + } +#endif + + rpInfo2->sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2_KHR; + rpInfo2->pNext = nullptr; // the 1.1 VkRenderPassMultiviewCreateInfo is part of the '2' structs + rpInfo2->flags = rpInfo->flags; + rpInfo2->attachmentCount = rpInfo->attachmentCount; + rpInfo2->pAttachments = attDescs2.constData(); + rpInfo2->subpassCount = 1; + rpInfo2->pSubpasses = &subpass2; + if (multiViewCount >= 2) { + rpInfo2->correlatedViewMaskCount = 1; + rpInfo2->pCorrelatedViewMasks = &viewMask; + } + return true; + } + + QVarLengthArray<VkAttachmentDescription2KHR, 8> attDescs2; + QVarLengthArray<VkAttachmentReference2KHR, 8> attRefs2; + VkSubpassDescription2KHR subpass2; +#ifdef VK_KHR_depth_stencil_resolve + VkSubpassDescriptionDepthStencilResolveKHR dsResolveDesc; +#endif + uint32_t viewMask; +}; +#endif // VK_KHR_create_renderpass2 + bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, - const QRhiColorAttachment *firstColorAttachment, - const QRhiColorAttachment *lastColorAttachment, + const QRhiColorAttachment *colorAttachmentsBegin, + const QRhiColorAttachment *colorAttachmentsEnd, bool preserveColor, bool preserveDs, + bool storeDs, QRhiRenderBuffer *depthStencilBuffer, - QRhiTexture *depthTexture) + QRhiTexture *depthTexture, + QRhiTexture *depthResolveTexture) { - // attachment list layout is color (0-8), ds (0-1), resolve (0-8) + // attachment list layout is color (0-8), ds (0-1), resolve (0-8), ds resolve (0-1) - for (auto it = firstColorAttachment; it != lastColorAttachment; ++it) { + int multiViewCount = 0; + for (auto it = colorAttachmentsBegin; it != colorAttachmentsEnd; ++it) { QVkTexture *texD = QRHI_RES(QVkTexture, it->texture()); QVkRenderBuffer *rbD = QRHI_RES(QVkRenderBuffer, it->renderBuffer()); Q_ASSERT(texD || rbD); - const VkFormat vkformat = texD ? texD->vkformat : rbD->vkformat; + const VkFormat vkformat = texD ? texD->viewFormat : rbD->vkformat; const VkSampleCountFlagBits samples = texD ? texD->samples : rbD->samples; VkAttachmentDescription attDesc = {}; attDesc.format = vkformat; attDesc.samples = samples; attDesc.loadOp = preserveColor ? VK_ATTACHMENT_LOAD_OP_LOAD : VK_ATTACHMENT_LOAD_OP_CLEAR; - attDesc.storeOp = it->resolveTexture() ? VK_ATTACHMENT_STORE_OP_DONT_CARE : VK_ATTACHMENT_STORE_OP_STORE; + attDesc.storeOp = (it->resolveTexture() && !preserveColor) ? VK_ATTACHMENT_STORE_OP_DONT_CARE : VK_ATTACHMENT_STORE_OP_STORE; attDesc.stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; attDesc.stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; // this has to interact correctly with activateTextureRenderTarget(), hence leaving in COLOR_ATT @@ -1283,16 +1631,27 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, const VkAttachmentReference ref = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL }; rpD->colorRefs.append(ref); + + if (it->multiViewCount() >= 2) { + if (multiViewCount > 0 && multiViewCount != it->multiViewCount()) + qWarning("Inconsistent multiViewCount in color attachment set"); + else + multiViewCount = it->multiViewCount(); + } else if (multiViewCount > 0) { + qWarning("Mixing non-multiview color attachments within a multiview render pass"); + } } + Q_ASSERT(multiViewCount == 0 || multiViewCount >= 2); + rpD->multiViewCount = uint32_t(multiViewCount); rpD->hasDepthStencil = depthStencilBuffer || depthTexture; if (rpD->hasDepthStencil) { - const VkFormat dsFormat = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->vkformat + const VkFormat dsFormat = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->viewFormat : QRHI_RES(QVkRenderBuffer, depthStencilBuffer)->vkformat; const VkSampleCountFlagBits samples = depthTexture ? QRHI_RES(QVkTexture, depthTexture)->samples : QRHI_RES(QVkRenderBuffer, depthStencilBuffer)->samples; const VkAttachmentLoadOp loadOp = preserveDs ? VK_ATTACHMENT_LOAD_OP_LOAD : VK_ATTACHMENT_LOAD_OP_CLEAR; - const VkAttachmentStoreOp storeOp = depthTexture ? VK_ATTACHMENT_STORE_OP_STORE : VK_ATTACHMENT_STORE_OP_DONT_CARE; + const VkAttachmentStoreOp storeOp = storeDs ? VK_ATTACHMENT_STORE_OP_STORE : VK_ATTACHMENT_STORE_OP_DONT_CARE; VkAttachmentDescription attDesc = {}; attDesc.format = dsFormat; attDesc.samples = samples; @@ -1300,13 +1659,17 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, attDesc.storeOp = storeOp; attDesc.stencilLoadOp = loadOp; attDesc.stencilStoreOp = storeOp; - attDesc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + attDesc.initialLayout = preserveDs ? VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL : VK_IMAGE_LAYOUT_UNDEFINED; attDesc.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; rpD->attDescs.append(attDesc); + if (depthTexture && depthTexture->arraySize() >= 2 && colorAttachmentsBegin == colorAttachmentsEnd) { + multiViewCount = depthTexture->arraySize(); + rpD->multiViewCount = multiViewCount; + } } rpD->dsRef = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL }; - for (auto it = firstColorAttachment; it != lastColorAttachment; ++it) { + for (auto it = colorAttachmentsBegin; it != colorAttachmentsEnd; ++it) { if (it->resolveTexture()) { QVkTexture *rtexD = QRHI_RES(QVkTexture, it->resolveTexture()); const VkFormat dstFormat = rtexD->vkformat; @@ -1325,7 +1688,7 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, } VkAttachmentDescription attDesc = {}; - attDesc.format = dstFormat; + attDesc.format = rtexD->viewFormat; attDesc.samples = VK_SAMPLE_COUNT_1_BIT; attDesc.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; // ignored attDesc.storeOp = VK_ATTACHMENT_STORE_OP_STORE; @@ -1344,6 +1707,31 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, } Q_ASSERT(rpD->colorRefs.size() == rpD->resolveRefs.size()); + rpD->hasDepthStencilResolve = rpD->hasDepthStencil && depthResolveTexture; + if (rpD->hasDepthStencilResolve) { + QVkTexture *rtexD = QRHI_RES(QVkTexture, depthResolveTexture); + if (rtexD->samples > VK_SAMPLE_COUNT_1_BIT) + qWarning("Resolving into a multisample depth texture is not supported"); + + QVkTexture *texD = QRHI_RES(QVkTexture, depthResolveTexture); + if (texD->vkformat != rtexD->vkformat) { + qWarning("Multisample resolve between different depth-stencil formats (%d and %d) is not supported.", + int(texD->vkformat), int(rtexD->vkformat)); + } + + VkAttachmentDescription attDesc = {}; + attDesc.format = rtexD->viewFormat; + attDesc.samples = VK_SAMPLE_COUNT_1_BIT; + attDesc.loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; // ignored + attDesc.storeOp = VK_ATTACHMENT_STORE_OP_STORE; + attDesc.stencilLoadOp = attDesc.loadOp; + attDesc.stencilStoreOp = attDesc.storeOp; + attDesc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED; + attDesc.finalLayout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + rpD->attDescs.append(attDesc); + } + rpD->dsResolveRef = { uint32_t(rpD->attDescs.size() - 1), VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL }; + // rpD->subpassDeps stays empty: don't yet know the correct initial/final // access and stage stuff for the implicit deps at this point, so leave it // to the resource tracking and activateTextureRenderTarget() to generate @@ -1353,10 +1741,35 @@ bool QRhiVulkan::createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, VkSubpassDescription subpassDesc; fillRenderPassCreateInfo(&rpInfo, &subpassDesc, rpD); - VkResult err = df->vkCreateRenderPass(dev, &rpInfo, nullptr, &rpD->rp); - if (err != VK_SUCCESS) { - qWarning("Failed to create renderpass: %d", err); + MultiViewRenderPassSetupHelper multiViewHelper; + if (!multiViewHelper.prepare(&rpInfo, multiViewCount, caps.multiView)) return false; + +#ifdef VK_KHR_create_renderpass2 + if (rpD->hasDepthStencilResolve && caps.renderPass2KHR) { + // Use the KHR extension, not the 1.2 core API, in order to support Vulkan 1.1. + VkRenderPassCreateInfo2KHR rpInfo2; + RenderPass2SetupHelper rp2Helper; + if (!rp2Helper.prepare(&rpInfo2, &rpInfo, rpD, multiViewCount)) + return false; + + VkResult err = vkCreateRenderPass2KHR(dev, &rpInfo2, nullptr, &rpD->rp); + if (err != VK_SUCCESS) { + qWarning("Failed to create renderpass (using VkRenderPassCreateInfo2KHR): %d", err); + return false; + } + } else +#endif + { + if (rpD->hasDepthStencilResolve) { + qWarning("Resolving multisample depth-stencil buffers is not supported without " + "VK_KHR_depth_stencil_resolve and VK_KHR_create_renderpass2"); + } + VkResult err = df->vkCreateRenderPass(dev, &rpInfo, nullptr, &rpD->rp); + if (err != VK_SUCCESS) { + qWarning("Failed to create renderpass: %d", err); + return false; + } } return true; @@ -1443,9 +1856,16 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) if (swapChainD->supportsReadback && swapChainD->m_flags.testFlag(QRhiSwapChain::UsedAsTransferSource)) usage |= VK_IMAGE_USAGE_TRANSFER_SRC_BIT; + const bool stereo = bool(swapChainD->m_window) && (swapChainD->m_window->format().stereo()) + && surfaceCaps.maxImageArrayLayers > 1; + swapChainD->stereo = stereo; + VkPresentModeKHR presentMode = VK_PRESENT_MODE_FIFO_KHR; if (swapChainD->m_flags.testFlag(QRhiSwapChain::NoVSync)) { - if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_MAILBOX_KHR)) + // Stereo has a weird bug, when using VK_PRESENT_MODE_MAILBOX_KHR, + // black screen is shown, but there is no validation error. + // Detected on Windows, with NVidia RTX A series (at least 4000 and 6000) driver 535.98 + if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_MAILBOX_KHR) && !stereo) presentMode = VK_PRESENT_MODE_MAILBOX_KHR; else if (swapChainD->supportedPresentationModes.contains(VK_PRESENT_MODE_IMMEDIATE_KHR)) presentMode = VK_PRESENT_MODE_IMMEDIATE_KHR; @@ -1469,7 +1889,7 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) swapChainInfo.imageFormat = swapChainD->colorFormat; swapChainInfo.imageColorSpace = swapChainD->colorSpace; swapChainInfo.imageExtent = VkExtent2D { uint32_t(swapChainD->pixelSize.width()), uint32_t(swapChainD->pixelSize.height()) }; - swapChainInfo.imageArrayLayers = 1; + swapChainInfo.imageArrayLayers = stereo ? 2u : 1u; swapChainInfo.imageUsage = usage; swapChainInfo.imageSharingMode = VK_SHARING_MODE_EXCLUSIVE; swapChainInfo.preTransform = preTransform; @@ -1531,7 +1951,9 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) fenceInfo.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; fenceInfo.flags = VK_FENCE_CREATE_SIGNALED_BIT; - swapChainD->imageRes.resize(swapChainD->bufferCount); + // Double up for stereo + swapChainD->imageRes.resize(swapChainD->bufferCount * (stereo ? 2u : 1u)); + for (int i = 0; i < swapChainD->bufferCount; ++i) { QVkSwapChain::ImageResources &image(swapChainD->imageRes[i]); image.image = swapChainImages[i]; @@ -1559,6 +1981,36 @@ bool QRhiVulkan::recreateSwapChain(QRhiSwapChain *swapChain) image.lastUse = QVkSwapChain::ImageResources::ScImageUseNone; } + if (stereo) { + for (int i = 0; i < swapChainD->bufferCount; ++i) { + QVkSwapChain::ImageResources &image(swapChainD->imageRes[i + swapChainD->bufferCount]); + image.image = swapChainImages[i]; + if (swapChainD->samples > VK_SAMPLE_COUNT_1_BIT) { + image.msaaImage = msaaImages[i]; + image.msaaImageView = msaaViews[i]; + } + + VkImageViewCreateInfo imgViewInfo = {}; + imgViewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + imgViewInfo.image = swapChainImages[i]; + imgViewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; + imgViewInfo.format = swapChainD->colorFormat; + imgViewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + imgViewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + imgViewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + imgViewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + imgViewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT; + imgViewInfo.subresourceRange.baseArrayLayer = 1; + imgViewInfo.subresourceRange.levelCount = imgViewInfo.subresourceRange.layerCount = 1; + err = df->vkCreateImageView(dev, &imgViewInfo, nullptr, &image.imageView); + if (err != VK_SUCCESS) { + qWarning("Failed to create swapchain image view %d: %d", i, err); + return false; + } + + image.lastUse = QVkSwapChain::ImageResources::ScImageUseNone; + } + } swapChainD->currentImageIndex = 0; @@ -1626,7 +2078,7 @@ void QRhiVulkan::releaseSwapChainResources(QRhiSwapChain *swapChain) } } - for (int i = 0; i < swapChainD->bufferCount; ++i) { + for (int i = 0; i < swapChainD->bufferCount * (swapChainD->stereo ? 2 : 1); ++i) { QVkSwapChain::ImageResources &image(swapChainD->imageRes[i]); if (image.fb) { df->vkDestroyFramebuffer(dev, image.fb, nullptr); @@ -1671,12 +2123,32 @@ void QRhiVulkan::ensureCommandPoolForNewFrame() df->vkResetCommandPool(dev, cmdPool[currentFrameSlot], flags); } +double QRhiVulkan::elapsedSecondsFromTimestamp(quint64 timestamp[2], bool *ok) +{ + quint64 mask = 0; + for (quint64 i = 0; i < timestampValidBits; i += 8) + mask |= 0xFFULL << i; + const quint64 ts0 = timestamp[0] & mask; + const quint64 ts1 = timestamp[1] & mask; + const float nsecsPerTick = physDevProperties.limits.timestampPeriod; + if (!qFuzzyIsNull(nsecsPerTick)) { + const float elapsedMs = float(ts1 - ts0) * nsecsPerTick / 1000000.0f; + const double elapsedSec = elapsedMs / 1000.0; + *ok = true; + return elapsedSec; + } + *ok = false; + return 0; +} + QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags) { QVkSwapChain *swapChainD = QRHI_RES(QVkSwapChain, swapChain); const int frameResIndex = swapChainD->bufferCount > 1 ? swapChainD->currentFrameSlot : 0; QVkSwapChain::FrameResources &frame(swapChainD->frameRes[frameResIndex]); + inst->handle()->beginFrame(swapChainD->window); + if (!frame.imageAcquired) { // Wait if we are too far ahead, i.e. the thread gets throttled based on the presentation rate // (note that we are using FIFO mode -> vsync) @@ -1720,30 +2192,6 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin // mess up A's in-flight commands (as they are not in flight anymore). waitCommandCompletion(frameResIndex); - // Now is the time to read the timestamps for the previous frame for this slot. - if (frame.timestampQueryIndex >= 0) { - quint64 timestamp[2] = { 0, 0 }; - VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(frame.timestampQueryIndex), 2, - 2 * sizeof(quint64), timestamp, sizeof(quint64), - VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); - timestampQueryPoolMap.clearBit(frame.timestampQueryIndex / 2); - frame.timestampQueryIndex = -1; - if (err == VK_SUCCESS) { - quint64 mask = 0; - for (quint64 i = 0; i < timestampValidBits; i += 8) - mask |= 0xFFULL << i; - const quint64 ts0 = timestamp[0] & mask; - const quint64 ts1 = timestamp[1] & mask; - const float nsecsPerTick = physDevProperties.limits.timestampPeriod; - if (!qFuzzyIsNull(nsecsPerTick)) { - const float elapsedMs = float(ts1 - ts0) * nsecsPerTick / 1000000.0f; - runGpuFrameTimeCallbacks(elapsedMs); - } - } else { - qWarning("Failed to query timestamp: %d", err); - } - } - currentFrameSlot = int(swapChainD->currentFrameSlot); currentSwapChain = swapChainD; if (swapChainD->ds) @@ -1757,9 +2205,40 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin if (cbres != QRhi::FrameOpSuccess) return cbres; - // when profiling is enabled, pick a free query (pair) from the pool - int timestampQueryIdx = -1; - if (hasGpuFrameTimeCallback() && swapChainD->bufferCount > 1) { // no timestamps if not having at least 2 frames in flight + swapChainD->cbWrapper.cb = frame.cmdBuf; + + QVkSwapChain::ImageResources &image(swapChainD->imageRes[swapChainD->currentImageIndex]); + swapChainD->rtWrapper.d.fb = image.fb; + + if (swapChainD->stereo) { + QVkSwapChain::ImageResources &image( + swapChainD->imageRes[swapChainD->currentImageIndex + swapChainD->bufferCount]); + swapChainD->rtWrapperRight.d.fb = image.fb; + } + + prepareNewFrame(&swapChainD->cbWrapper); + + // Read the timestamps for the previous frame for this slot. + if (frame.timestampQueryIndex >= 0) { + quint64 timestamp[2] = { 0, 0 }; + VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(frame.timestampQueryIndex), 2, + 2 * sizeof(quint64), timestamp, sizeof(quint64), + VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); + timestampQueryPoolMap.clearBit(frame.timestampQueryIndex / 2); + frame.timestampQueryIndex = -1; + if (err == VK_SUCCESS) { + bool ok = false; + const double elapsedSec = elapsedSecondsFromTimestamp(timestamp, &ok); + if (ok) + swapChainD->cbWrapper.lastGpuTime = elapsedSec; + } else { + qWarning("Failed to query timestamp: %d", err); + } + } + + // No timestamps if the client did not opt in, or when not having at least 2 frames in flight. + if (rhiFlags.testFlag(QRhi::EnableTimestamps) && swapChainD->bufferCount > 1) { + int timestampQueryIdx = -1; for (int i = 0; i < timestampQueryPoolMap.size(); ++i) { if (!timestampQueryPoolMap.testBit(i)) { timestampQueryPoolMap.setBit(i); @@ -1767,21 +2246,14 @@ QRhi::FrameOpResult QRhiVulkan::beginFrame(QRhiSwapChain *swapChain, QRhi::Begin break; } } + if (timestampQueryIdx >= 0) { + df->vkCmdResetQueryPool(frame.cmdBuf, timestampQueryPool, uint32_t(timestampQueryIdx), 2); + // record timestamp at the start of the command buffer + df->vkCmdWriteTimestamp(frame.cmdBuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, + timestampQueryPool, uint32_t(timestampQueryIdx)); + frame.timestampQueryIndex = timestampQueryIdx; + } } - if (timestampQueryIdx >= 0) { - df->vkCmdResetQueryPool(frame.cmdBuf, timestampQueryPool, uint32_t(timestampQueryIdx), 2); - // record timestamp at the start of the command buffer - df->vkCmdWriteTimestamp(frame.cmdBuf, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, - timestampQueryPool, uint32_t(timestampQueryIdx)); - frame.timestampQueryIndex = timestampQueryIdx; - } - - swapChainD->cbWrapper.cb = frame.cmdBuf; - - QVkSwapChain::ImageResources &image(swapChainD->imageRes[swapChainD->currentImageIndex]); - swapChainD->rtWrapper.d.fb = image.fb; - - prepareNewFrame(&swapChainD->cbWrapper); return QRhi::FrameOpSuccess; } @@ -1791,6 +2263,10 @@ QRhi::FrameOpResult QRhiVulkan::endFrame(QRhiSwapChain *swapChain, QRhi::EndFram QVkSwapChain *swapChainD = QRHI_RES(QVkSwapChain, swapChain); Q_ASSERT(currentSwapChain == swapChainD); + auto cleanup = qScopeGuard([this, swapChainD] { + inst->handle()->endFrame(swapChainD->window); + }); + recordPrimaryCommandBuffer(&swapChainD->cbWrapper); int frameResIndex = swapChainD->bufferCount > 1 ? swapChainD->currentFrameSlot : 0; @@ -2031,6 +2507,24 @@ QRhi::FrameOpResult QRhiVulkan::beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi prepareNewFrame(cbWrapper); ofr.active = true; + if (rhiFlags.testFlag(QRhi::EnableTimestamps)) { + int timestampQueryIdx = -1; + for (int i = 0; i < timestampQueryPoolMap.size(); ++i) { + if (!timestampQueryPoolMap.testBit(i)) { + timestampQueryPoolMap.setBit(i); + timestampQueryIdx = i * 2; + break; + } + } + if (timestampQueryIdx >= 0) { + df->vkCmdResetQueryPool(cbWrapper->cb, timestampQueryPool, uint32_t(timestampQueryIdx), 2); + // record timestamp at the start of the command buffer + df->vkCmdWriteTimestamp(cbWrapper->cb, VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT, + timestampQueryPool, uint32_t(timestampQueryIdx)); + ofr.timestampQueryIndex = timestampQueryIdx; + } + } + *cb = cbWrapper; return QRhi::FrameOpSuccess; } @@ -2044,6 +2538,12 @@ QRhi::FrameOpResult QRhiVulkan::endOffscreenFrame(QRhi::EndFrameFlags flags) QVkCommandBuffer *cbWrapper(ofr.cbWrapper[currentFrameSlot]); recordPrimaryCommandBuffer(cbWrapper); + // record another timestamp, when enabled + if (ofr.timestampQueryIndex >= 0) { + df->vkCmdWriteTimestamp(cbWrapper->cb, VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT, + timestampQueryPool, uint32_t(ofr.timestampQueryIndex + 1)); + } + if (!ofr.cmdFence) { VkFenceCreateInfo fenceInfo = {}; fenceInfo.sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO; @@ -2066,6 +2566,24 @@ QRhi::FrameOpResult QRhiVulkan::endOffscreenFrame(QRhi::EndFrameFlags flags) // previous) frame is safe since we waited for completion above. finishActiveReadbacks(true); + // Read the timestamps, if we wrote them. + if (ofr.timestampQueryIndex >= 0) { + quint64 timestamp[2] = { 0, 0 }; + VkResult err = df->vkGetQueryPoolResults(dev, timestampQueryPool, uint32_t(ofr.timestampQueryIndex), 2, + 2 * sizeof(quint64), timestamp, sizeof(quint64), + VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT); + timestampQueryPoolMap.clearBit(ofr.timestampQueryIndex / 2); + ofr.timestampQueryIndex = -1; + if (err == VK_SUCCESS) { + bool ok = false; + const double elapsedSec = elapsedSecondsFromTimestamp(timestamp, &ok); + if (ok) + cbWrapper->lastGpuTime = elapsedSec; + } else { + qWarning("Failed to query timestamp: %d", err); + } + } + return QRhi::FrameOpSuccess; } @@ -2183,6 +2701,13 @@ void QRhiVulkan::activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRe QRhiPassResourceTracker::TexDepthOutputStage); depthTexD->lastActiveFrameSlot = currentFrameSlot; } + if (rtD->m_desc.depthResolveTexture()) { + QVkTexture *depthResolveTexD = QRHI_RES(QVkTexture, rtD->m_desc.depthResolveTexture()); + trackedRegisterTexture(&passResTracker, depthResolveTexD, + QRhiPassResourceTracker::TexDepthOutput, + QRhiPassResourceTracker::TexDepthOutputStage); + depthResolveTexD->lastActiveFrameSlot = currentFrameSlot; + } } void QRhiVulkan::resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) @@ -2324,6 +2849,11 @@ void QRhiVulkan::beginPass(QRhiCommandBuffer *cb, float(colorClearValue.alphaF()) } }; cvs.append(cv); } + for (int i = 0; i < rtD->dsResolveAttCount; ++i) { + VkClearValue cv; + cv.depthStencil = { depthStencilClearValue.depthClearValue(), depthStencilClearValue.stencilClearValue() }; + cvs.append(cv); + } rpBeginInfo.clearValueCount = uint32_t(cvs.size()); QVkCommandBuffer::Command &cmd(cbD->commands.get()); @@ -2717,7 +3247,7 @@ void QRhiVulkan::updateShaderResourceBindings(QRhiShaderResourceBindings *srb, i case QRhiShaderResourceBinding::ImageLoadStore: { QVkTexture *texD = QRHI_RES(QVkTexture, b->u.simage.tex); - VkImageView view = texD->imageViewForLevel(b->u.simage.level); + VkImageView view = texD->perLevelImageViewForLoadStore(b->u.simage.level); if (view) { writeInfo.descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_IMAGE; bd.simage.id = texD->m_id; @@ -2971,12 +3501,12 @@ void QRhiVulkan::prepareUploadSubres(QVkTexture *texD, int layer, int level, const int sy = subresDesc.sourceTopLeft().y(); if (!subresDesc.sourceSize().isEmpty()) size = subresDesc.sourceSize(); - if (image.depth() == 32) { - // The staging buffer will get the full image - // regardless, just adjust the vk - // buffer-to-image copy start offset. - copyInfo.bufferOffset += VkDeviceSize(sy * image.bytesPerLine() + sx * 4); - // bufferRowLength remains set to the original image's width + + if (size.width() == image.width()) { + // No need to make a QImage copy here, can copy from the source + // QImage into staging directly. + src = image.constBits() + sy * image.bytesPerLine() + sx * bpc; + copySizeBytes = size.height() * image.bytesPerLine(); } else { image = image.copy(sx, sy, size.width(), size.height()); src = image.constBits(); @@ -3039,6 +3569,12 @@ void QRhiVulkan::prepareUploadSubres(QVkTexture *texD, int layer, int level, } } +void QRhiVulkan::printExtraErrorInfo(VkResult err) +{ + if (err == VK_ERROR_OUT_OF_DEVICE_MEMORY) + qWarning() << "Out of device memory, current allocator statistics are" << statistics(); +} + void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates) { QRhiResourceUpdateBatchPrivate *ud = QRhiResourceUpdateBatchPrivate::get(resourceUpdates); @@ -3076,6 +3612,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat bufD->stagingAllocations[currentFrameSlot] = allocation; } else { qWarning("Failed to create staging buffer of size %u: %d", bufD->m_size, err); + printExtraErrorInfo(err); continue; } } @@ -3164,6 +3701,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat readback.stagingAlloc = allocation; } else { qWarning("Failed to create readback buffer of size %u: %d", readback.byteSize, err); + printExtraErrorInfo(err); continue; } @@ -3213,6 +3751,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat &utexD->stagingBuffers[currentFrameSlot], &allocation, nullptr); if (err != VK_SUCCESS) { qWarning("Failed to create image staging buffer of size %d: %d", int(stagingSize), err); + printExtraErrorInfo(err); continue; } utexD->stagingAllocations[currentFrameSlot] = allocation; @@ -3369,6 +3908,7 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat readback.stagingAlloc = allocation; } else { qWarning("Failed to create readback buffer of size %u: %d", readback.byteSize, err); + printExtraErrorInfo(err); continue; } @@ -3434,10 +3974,10 @@ void QRhiVulkan::enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdat if (!origStage) origStage = VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT; - for (int layer = 0; layer < (isCube ? 6 : (isArray ? utexD->m_arraySize : 1)); ++layer) { + for (int layer = 0; layer < (isCube ? 6 : (isArray ? qMax(0, utexD->m_arraySize) : 1)); ++layer) { int w = utexD->m_pixelSize.width(); int h = utexD->m_pixelSize.height(); - int depth = is3D ? utexD->m_depth : 1; + int depth = is3D ? qMax(1, utexD->m_depth) : 1; for (int level = 1; level < int(utexD->mipLevelCount); ++level) { if (level == 1) { subresourceBarrier(cbD, utexD->image, @@ -3616,6 +4156,8 @@ void QRhiVulkan::executeDeferredReleases(bool forced) df->vkDestroyImageView(dev, e.textureRenderTarget.rtv[att], nullptr); df->vkDestroyImageView(dev, e.textureRenderTarget.resrtv[att], nullptr); } + df->vkDestroyImageView(dev, e.textureRenderTarget.dsv, nullptr); + df->vkDestroyImageView(dev, e.textureRenderTarget.resdsv, nullptr); break; case QRhiVulkan::DeferredReleaseEntry::RenderPass: df->vkDestroyRenderPass(dev, e.renderPass.rp, nullptr); @@ -3725,18 +4267,12 @@ QList<int> QRhiVulkan::supportedSampleCounts() const return result; } -VkSampleCountFlagBits QRhiVulkan::effectiveSampleCount(int sampleCount) +VkSampleCountFlagBits QRhiVulkan::effectiveSampleCountBits(int sampleCount) { - // Stay compatible with QSurfaceFormat and friends where samples == 0 means the same as 1. - sampleCount = qBound(1, sampleCount, 64); - - if (!supportedSampleCounts().contains(sampleCount)) { - qWarning("Attempted to set unsupported sample count %d", sampleCount); - return VK_SAMPLE_COUNT_1_BIT; - } + const int s = effectiveSampleCount(sampleCount); for (const auto &qvk_sampleCount : qvk_sampleCounts) { - if (qvk_sampleCount.count == sampleCount) + if (qvk_sampleCount.count == s) return qvk_sampleCount.mask; } @@ -4119,6 +4655,14 @@ void QRhiVulkan::recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhi QRhiSwapChain *QRhiVulkan::createSwapChain() { + if (!vkGetPhysicalDeviceSurfaceCapabilitiesKHR + || !vkGetPhysicalDeviceSurfaceFormatsKHR + || !vkGetPhysicalDeviceSurfacePresentModesKHR) + { + qWarning("Physical device surface queries not available"); + return nullptr; + } + return new QVkSwapChain(this); } @@ -4271,6 +4815,12 @@ bool QRhiVulkan::isFeatureSupported(QRhi::Feature feature) const return true; case QRhi::ThreeDimensionalTextureMipmaps: return true; + case QRhi::MultiView: + return caps.multiView; + case QRhi::TextureViewFormat: + return true; + case QRhi::ResolveDepthStencil: + return caps.renderPass2KHR && caps.depthStencilResolveKHR; default: Q_UNREACHABLE_RETURN(false); } @@ -4403,6 +4953,7 @@ QByteArray QRhiVulkan::pipelineCacheData() header.deviceId = physDevProperties.deviceID; header.dataSize = quint32(dataSize); header.uuidSize = VK_UUID_SIZE; + header.reserved = 0; memcpy(data.data(), &header, headerSize); memcpy(data.data() + headerSize, physDevProperties.pipelineCacheUUID, VK_UUID_SIZE); @@ -5153,6 +5704,12 @@ void QRhiVulkan::endExternal(QRhiCommandBuffer *cb) cbD->resetCachedState(); } +double QRhiVulkan::lastCompletedGpuTime(QRhiCommandBuffer *cb) +{ + QVkCommandBuffer *cbD = QRHI_RES(QVkCommandBuffer, cb); + return cbD->lastGpuTime; +} + void QRhiVulkan::setObjectName(uint64_t object, VkObjectType type, const QByteArray &name, int slot) { #ifdef VK_EXT_debug_utils @@ -5293,6 +5850,22 @@ static inline VkFormat toVkAttributeFormat(QRhiVertexInputAttribute::Format form return VK_FORMAT_R16G16_SFLOAT; case QRhiVertexInputAttribute::Half: return VK_FORMAT_R16_SFLOAT; + case QRhiVertexInputAttribute::UShort4: + return VK_FORMAT_R16G16B16A16_UINT; + case QRhiVertexInputAttribute::UShort3: + return VK_FORMAT_R16G16B16_UINT; + case QRhiVertexInputAttribute::UShort2: + return VK_FORMAT_R16G16_UINT; + case QRhiVertexInputAttribute::UShort: + return VK_FORMAT_R16_UINT; + case QRhiVertexInputAttribute::SShort4: + return VK_FORMAT_R16G16B16A16_SINT; + case QRhiVertexInputAttribute::SShort3: + return VK_FORMAT_R16G16B16_SINT; + case QRhiVertexInputAttribute::SShort2: + return VK_FORMAT_R16G16_SINT; + case QRhiVertexInputAttribute::SShort: + return VK_FORMAT_R16_SINT; default: Q_UNREACHABLE_RETURN(VK_FORMAT_R32G32B32A32_SFLOAT); } @@ -5662,7 +6235,8 @@ bool QVkBuffer::create() } if (err != VK_SUCCESS) { - qWarning("Failed to create buffer: %d", err); + qWarning("Failed to create buffer of size %u: %d", nonZeroSize, err); + rhiD->printExtraErrorInfo(err); return false; } @@ -5770,7 +6344,7 @@ bool QVkRenderBuffer::create() return false; QRHI_RES_RHI(QRhiVulkan); - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); switch (m_type) { case QRhiRenderBuffer::Color: @@ -5888,6 +6462,15 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) QRHI_RES_RHI(QRhiVulkan); vkformat = toVkTextureFormat(m_format, m_flags); + if (m_writeViewFormat.format != UnknownFormat) + viewFormat = toVkTextureFormat(m_writeViewFormat.format, m_writeViewFormat.srgb ? sRGB : Flags()); + else + viewFormat = vkformat; + if (m_readViewFormat.format != UnknownFormat) + viewFormatForSampling = toVkTextureFormat(m_readViewFormat.format, m_readViewFormat.srgb ? sRGB : Flags()); + else + viewFormatForSampling = vkformat; + VkFormatProperties props; rhiD->f->vkGetPhysicalDeviceFormatProperties(rhiD->physDev, vkformat, &props); const bool canSampleOptimal = (props.optimalTilingFeatures & VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT); @@ -5911,7 +6494,7 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) qWarning("Too many mip levels (%d, max is %d), truncating mip chain", mipLevelCount, maxLevels); mipLevelCount = maxLevels; } - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); if (samples > VK_SAMPLE_COUNT_1_BIT) { if (isCube) { qWarning("Cubemap texture cannot be multisample"); @@ -5942,12 +6525,10 @@ bool QVkTexture::prepareCreate(QSize *adjustedSize) qWarning("Texture cannot be both 1D and 3D"); return false; } - m_depth = qMax(1, m_depth); if (m_depth > 1 && !is3D) { qWarning("Texture cannot have a depth of %d when it is not 3D", m_depth); return false; } - m_arraySize = qMax(0, m_arraySize); if (m_arraySize > 0 && !isArray) { qWarning("Texture cannot have an array size of %d when it is not an array", m_arraySize); return false; @@ -5985,7 +6566,7 @@ bool QVkTexture::finishCreate() : (is3D ? VK_IMAGE_VIEW_TYPE_3D : (is1D ? (isArray ? VK_IMAGE_VIEW_TYPE_1D_ARRAY : VK_IMAGE_VIEW_TYPE_1D) : (isArray ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D))); - viewInfo.format = vkformat; + viewInfo.format = viewFormatForSampling; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -5996,7 +6577,7 @@ bool QVkTexture::finishCreate() viewInfo.subresourceRange.baseArrayLayer = uint32_t(m_arrayRangeStart); viewInfo.subresourceRange.layerCount = uint32_t(m_arrayRangeLength); } else { - viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? m_arraySize : 1); + viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); } VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &imageView); @@ -6050,9 +6631,9 @@ bool QVkTexture::create() imageInfo.format = vkformat; imageInfo.extent.width = uint32_t(size.width()); imageInfo.extent.height = uint32_t(size.height()); - imageInfo.extent.depth = is3D ? m_depth : 1; + imageInfo.extent.depth = is3D ? qMax(1, m_depth) : 1; imageInfo.mipLevels = mipLevelCount; - imageInfo.arrayLayers = isCube ? 6 : (isArray ? m_arraySize : 1); + imageInfo.arrayLayers = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); imageInfo.samples = samples; imageInfo.tiling = VK_IMAGE_TILING_OPTIMAL; imageInfo.initialLayout = VK_IMAGE_LAYOUT_PREINITIALIZED; @@ -6077,7 +6658,14 @@ bool QVkTexture::create() VmaAllocation allocation; VkResult err = vmaCreateImage(toVmaAllocator(rhiD->allocator), &imageInfo, &allocInfo, &image, &allocation, nullptr); if (err != VK_SUCCESS) { - qWarning("Failed to create image: %d", err); + qWarning("Failed to create image (with VkImageCreateInfo %ux%u depth %u vkformat 0x%X mips %u layers %u vksamples 0x%X): %d", + imageInfo.extent.width, imageInfo.extent.height, imageInfo.extent.depth, + int(imageInfo.format), + imageInfo.mipLevels, + imageInfo.arrayLayers, + int(imageInfo.samples), + err); + rhiD->printExtraErrorInfo(err); return false; } imageAlloc = allocation; @@ -6124,7 +6712,7 @@ void QVkTexture::setNativeLayout(int layout) usageState.layout = VkImageLayout(layout); } -VkImageView QVkTexture::imageViewForLevel(int level) +VkImageView QVkTexture::perLevelImageViewForLoadStore(int level) { Q_ASSERT(level >= 0 && level < int(mipLevelCount)); if (perLevelImageViews[level] != VK_NULL_HANDLE) @@ -6144,7 +6732,7 @@ VkImageView QVkTexture::imageViewForLevel(int level) : (is3D ? VK_IMAGE_VIEW_TYPE_3D : (is1D ? (isArray ? VK_IMAGE_VIEW_TYPE_1D_ARRAY : VK_IMAGE_VIEW_TYPE_1D) : (isArray ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D))); - viewInfo.format = vkformat; + viewInfo.format = viewFormat; // this is writeViewFormat, regardless of Load, Store, or LoadStore; intentional viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6153,7 +6741,7 @@ VkImageView QVkTexture::imageViewForLevel(int level) viewInfo.subresourceRange.baseMipLevel = uint32_t(level); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = 0; - viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? m_arraySize : 1); + viewInfo.subresourceRange.layerCount = isCube ? 6 : (isArray ? qMax(0, m_arraySize) : 1); VkImageView v = VK_NULL_HANDLE; QRHI_RES_RHI(QRhiVulkan); @@ -6231,7 +6819,7 @@ bool QVkSampler::create() QVkRenderPassDescriptor::QVkRenderPassDescriptor(QRhiImplementation *rhi) : QRhiRenderPassDescriptor(rhi) { - serializedFormatData.reserve(32); + serializedFormatData.reserve(64); } QVkRenderPassDescriptor::~QVkRenderPassDescriptor() @@ -6294,6 +6882,10 @@ bool QVkRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other return false; if (hasDepthStencil != o->hasDepthStencil) return false; + if (hasDepthStencilResolve != o->hasDepthStencilResolve) + return false; + if (multiViewCount != o->multiViewCount) + return false; for (int i = 0, ie = colorRefs.size(); i != ie; ++i) { const uint32_t attIdx = colorRefs[i].attachment; @@ -6319,6 +6911,14 @@ bool QVkRenderPassDescriptor::isCompatible(const QRhiRenderPassDescriptor *other return false; } + if (hasDepthStencilResolve) { + const uint32_t attIdx = dsResolveRef.attachment; + if (attIdx != o->dsResolveRef.attachment) + return false; + if (attIdx != VK_ATTACHMENT_UNUSED && !attachmentDescriptionEquals(attDescs[attIdx], o->attDescs[attIdx])) + return false; + } + // subpassDeps is not included return true; @@ -6333,6 +6933,8 @@ void QVkRenderPassDescriptor::updateSerializedFormat() *p++ = colorRefs.size(); *p++ = resolveRefs.size(); *p++ = hasDepthStencil; + *p++ = hasDepthStencilResolve; + *p++ = multiViewCount; auto serializeAttachmentData = [this, &p](uint32_t attIdx) { const bool used = attIdx != VK_ATTACHMENT_UNUSED; @@ -6364,6 +6966,12 @@ void QVkRenderPassDescriptor::updateSerializedFormat() *p++ = attIdx; serializeAttachmentData(attIdx); } + + if (hasDepthStencilResolve) { + const uint32_t attIdx = dsResolveRef.attachment; + *p++ = attIdx; + serializeAttachmentData(attIdx); + } } QRhiRenderPassDescriptor *QVkRenderPassDescriptor::newCompatibleRenderPassDescriptor() const @@ -6376,13 +6984,22 @@ QRhiRenderPassDescriptor *QVkRenderPassDescriptor::newCompatibleRenderPassDescri rpD->resolveRefs = resolveRefs; rpD->subpassDeps = subpassDeps; rpD->hasDepthStencil = hasDepthStencil; + rpD->hasDepthStencilResolve = hasDepthStencilResolve; + rpD->multiViewCount = multiViewCount; rpD->dsRef = dsRef; + rpD->dsResolveRef = dsResolveRef; VkRenderPassCreateInfo rpInfo; VkSubpassDescription subpassDesc; fillRenderPassCreateInfo(&rpInfo, &subpassDesc, rpD); QRHI_RES_RHI(QRhiVulkan); + MultiViewRenderPassSetupHelper multiViewHelper; + if (!multiViewHelper.prepare(&rpInfo, multiViewCount, rhiD->caps.multiView)) { + delete rpD; + return nullptr; + } + VkResult err = rhiD->df->vkCreateRenderPass(rhiD->dev, &rpInfo, nullptr, &rpD->rp); if (err != VK_SUCCESS) { qWarning("Failed to create renderpass: %d", err); @@ -6471,6 +7088,11 @@ void QVkTextureRenderTarget::destroy() resrtv[att] = VK_NULL_HANDLE; } + e.textureRenderTarget.dsv = dsv; + dsv = VK_NULL_HANDLE; + e.textureRenderTarget.resdsv = resdsv; + resdsv = VK_NULL_HANDLE; + QRHI_RES_RHI(QRhiVulkan); if (rhiD) { rhiD->releaseQueue.append(e); @@ -6489,8 +7111,10 @@ QRhiRenderPassDescriptor *QVkTextureRenderTarget::newCompatibleRenderPassDescrip m_desc.cendColorAttachments(), m_flags.testFlag(QRhiTextureRenderTarget::PreserveColorContents), m_flags.testFlag(QRhiTextureRenderTarget::PreserveDepthStencilContents), + m_desc.depthTexture() && !m_flags.testFlag(DoNotStoreDepthStencilContents) && !m_desc.depthResolveTexture(), m_desc.depthStencilBuffer(), - m_desc.depthTexture())) + m_desc.depthTexture(), + m_desc.depthResolveTexture())) { delete rp; return nullptr; @@ -6513,6 +7137,7 @@ bool QVkTextureRenderTarget::create() QRHI_RES_RHI(QRhiVulkan); QVarLengthArray<VkImageView, 8> views; + d.multiViewCount = 0; d.colorAttCount = 0; int attIndex = 0; @@ -6523,13 +7148,17 @@ bool QVkTextureRenderTarget::create() Q_ASSERT(texD || rbD); if (texD) { Q_ASSERT(texD->flags().testFlag(QRhiTexture::RenderTarget)); + const bool is1D = texD->flags().testFlag(QRhiTexture::OneDimensional); + const bool isMultiView = it->multiViewCount() >= 2; + if (isMultiView && d.multiViewCount == 0) + d.multiViewCount = it->multiViewCount(); VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = texD->image; - viewInfo.viewType = texD->flags().testFlag(QRhiTexture::OneDimensional) - ? VK_IMAGE_VIEW_TYPE_1D - : VK_IMAGE_VIEW_TYPE_2D; - viewInfo.format = texD->vkformat; + viewInfo.viewType = is1D ? VK_IMAGE_VIEW_TYPE_1D + : (isMultiView ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D); + viewInfo.format = texD->viewFormat; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6538,7 +7167,7 @@ bool QVkTextureRenderTarget::create() viewInfo.subresourceRange.baseMipLevel = uint32_t(it->level()); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = uint32_t(it->layer()); - viewInfo.subresourceRange.layerCount = 1; + viewInfo.subresourceRange.layerCount = uint32_t(isMultiView ? it->multiViewCount() : 1); VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &rtv[attIndex]); if (err != VK_SUCCESS) { qWarning("Failed to create render target image view: %d", err); @@ -6563,7 +7192,25 @@ bool QVkTextureRenderTarget::create() if (hasDepthStencil) { if (m_desc.depthTexture()) { QVkTexture *depthTexD = QRHI_RES(QVkTexture, m_desc.depthTexture()); - views.append(depthTexD->imageView); + // need a dedicated view just because viewFormat may differ from vkformat + VkImageViewCreateInfo viewInfo = {}; + viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + viewInfo.image = depthTexD->image; + viewInfo.viewType = d.multiViewCount > 1 ? VK_IMAGE_VIEW_TYPE_2D_ARRAY : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = depthTexD->viewFormat; + viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + viewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + viewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + viewInfo.subresourceRange.levelCount = 1; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); + VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &dsv); + if (err != VK_SUCCESS) { + qWarning("Failed to create depth-stencil image view for rt: %d", err); + return false; + } + views.append(dsv); if (d.colorAttCount == 0) { d.pixelSize = depthTexD->pixelSize(); d.sampleCount = depthTexD->samples; @@ -6583,6 +7230,7 @@ bool QVkTextureRenderTarget::create() d.resolveAttCount = 0; attIndex = 0; + Q_ASSERT(d.multiViewCount == 0 || d.multiViewCount >= 2); for (auto it = m_desc.cbeginColorAttachments(), itEnd = m_desc.cendColorAttachments(); it != itEnd; ++it, ++attIndex) { if (it->resolveTexture()) { QVkTexture *resTexD = QRHI_RES(QVkTexture, it->resolveTexture()); @@ -6592,8 +7240,9 @@ bool QVkTextureRenderTarget::create() VkImageViewCreateInfo viewInfo = {}; viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; viewInfo.image = resTexD->image; - viewInfo.viewType = VK_IMAGE_VIEW_TYPE_2D; - viewInfo.format = resTexD->vkformat; + viewInfo.viewType = d.multiViewCount ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = resTexD->viewFormat; viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; @@ -6602,7 +7251,7 @@ bool QVkTextureRenderTarget::create() viewInfo.subresourceRange.baseMipLevel = uint32_t(it->resolveLevel()); viewInfo.subresourceRange.levelCount = 1; viewInfo.subresourceRange.baseArrayLayer = uint32_t(it->resolveLayer()); - viewInfo.subresourceRange.layerCount = 1; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &resrtv[attIndex]); if (err != VK_SUCCESS) { qWarning("Failed to create render target resolve image view: %d", err); @@ -6612,6 +7261,36 @@ bool QVkTextureRenderTarget::create() } } + if (m_desc.depthResolveTexture()) { + QVkTexture *resTexD = QRHI_RES(QVkTexture, m_desc.depthResolveTexture()); + Q_ASSERT(resTexD->flags().testFlag(QRhiTexture::RenderTarget)); + + VkImageViewCreateInfo viewInfo = {}; + viewInfo.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO; + viewInfo.image = resTexD->image; + viewInfo.viewType = d.multiViewCount ? VK_IMAGE_VIEW_TYPE_2D_ARRAY + : VK_IMAGE_VIEW_TYPE_2D; + viewInfo.format = resTexD->viewFormat; + viewInfo.components.r = VK_COMPONENT_SWIZZLE_R; + viewInfo.components.g = VK_COMPONENT_SWIZZLE_G; + viewInfo.components.b = VK_COMPONENT_SWIZZLE_B; + viewInfo.components.a = VK_COMPONENT_SWIZZLE_A; + viewInfo.subresourceRange.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT; + viewInfo.subresourceRange.baseMipLevel = 0; + viewInfo.subresourceRange.levelCount = 1; + viewInfo.subresourceRange.baseArrayLayer = 0; + viewInfo.subresourceRange.layerCount = qMax<uint32_t>(1, d.multiViewCount); + VkResult err = rhiD->df->vkCreateImageView(rhiD->dev, &viewInfo, nullptr, &resdsv); + if (err != VK_SUCCESS) { + qWarning("Failed to create render target depth resolve image view: %d", err); + return false; + } + views.append(resdsv); + d.dsResolveAttCount = 1; + } else { + d.dsResolveAttCount = 0; + } + if (!m_renderPassDesc) qWarning("QVkTextureRenderTarget: No renderpass descriptor set. See newCompatibleRenderPassDescriptor() and setRenderPassDescriptor()."); @@ -6621,7 +7300,7 @@ bool QVkTextureRenderTarget::create() VkFramebufferCreateInfo fbInfo = {}; fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; fbInfo.renderPass = d.rp->rp; - fbInfo.attachmentCount = uint32_t(d.colorAttCount + d.dsAttCount + d.resolveAttCount); + fbInfo.attachmentCount = uint32_t(d.colorAttCount + d.dsAttCount + d.resolveAttCount + d.dsResolveAttCount); fbInfo.pAttachments = views.constData(); fbInfo.width = uint32_t(d.pixelSize.width()); fbInfo.height = uint32_t(d.pixelSize.height()); @@ -6880,7 +7559,9 @@ bool QVkGraphicsPipeline::create() pipelineInfo.pStages = shaderStageCreateInfos.constData(); QVarLengthArray<VkVertexInputBindingDescription, 4> vertexBindings; +#ifdef VK_EXT_vertex_attribute_divisor QVarLengthArray<VkVertexInputBindingDivisorDescriptionEXT> nonOneStepRates; +#endif int bindingIndex = 0; for (auto it = m_vertexInputLayout.cbeginBindings(), itEnd = m_vertexInputLayout.cendBindings(); it != itEnd; ++it, ++bindingIndex) @@ -6892,9 +7573,12 @@ bool QVkGraphicsPipeline::create() ? VK_VERTEX_INPUT_RATE_VERTEX : VK_VERTEX_INPUT_RATE_INSTANCE }; if (it->classification() == QRhiVertexInputBinding::PerInstance && it->instanceStepRate() != 1) { +#ifdef VK_EXT_vertex_attribute_divisor if (rhiD->caps.vertexAttribDivisor) { nonOneStepRates.append({ uint32_t(bindingIndex), it->instanceStepRate() }); - } else { + } else +#endif + { qWarning("QRhiVulkan: Instance step rates other than 1 not supported without " "VK_EXT_vertex_attribute_divisor on the device and " "VK_KHR_get_physical_device_properties2 on the instance"); @@ -6920,13 +7604,15 @@ bool QVkGraphicsPipeline::create() vertexInputInfo.pVertexBindingDescriptions = vertexBindings.constData(); vertexInputInfo.vertexAttributeDescriptionCount = uint32_t(vertexAttributes.size()); vertexInputInfo.pVertexAttributeDescriptions = vertexAttributes.constData(); +#ifdef VK_EXT_vertex_attribute_divisor VkPipelineVertexInputDivisorStateCreateInfoEXT divisorInfo = {}; if (!nonOneStepRates.isEmpty()) { - divisorInfo.sType = VkStructureType(1000190001); // VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT + divisorInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_DIVISOR_STATE_CREATE_INFO_EXT; divisorInfo.vertexBindingDivisorCount = uint32_t(nonOneStepRates.size()); divisorInfo.pVertexBindingDivisors = nonOneStepRates.constData(); vertexInputInfo.pNext = &divisorInfo; } +#endif pipelineInfo.pVertexInputState = &vertexInputInfo; QVarLengthArray<VkDynamicState, 8> dynEnable; @@ -6997,7 +7683,7 @@ bool QVkGraphicsPipeline::create() VkPipelineMultisampleStateCreateInfo msInfo = {}; msInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO; - msInfo.rasterizationSamples = rhiD->effectiveSampleCount(m_sampleCount); + msInfo.rasterizationSamples = rhiD->effectiveSampleCountBits(m_sampleCount); pipelineInfo.pMultisampleState = &msInfo; VkPipelineDepthStencilStateCreateInfo dsInfo = {}; @@ -7196,6 +7882,7 @@ const QRhiNativeHandles *QVkCommandBuffer::nativeHandles() QVkSwapChain::QVkSwapChain(QRhiImplementation *rhi) : QRhiSwapChain(rhi), rtWrapper(rhi, this), + rtWrapperRight(rhi, this), cbWrapper(rhi) { } @@ -7238,6 +7925,11 @@ QRhiRenderTarget *QVkSwapChain::currentFrameRenderTarget() return &rtWrapper; } +QRhiRenderTarget *QVkSwapChain::currentFrameRenderTarget(StereoTargetBuffer targetBuffer) +{ + return !stereo || targetBuffer == StereoTargetBuffer::LeftBuffer ? &rtWrapper : &rtWrapperRight; +} + QSize QVkSwapChain::surfacePixelSize() { if (!ensureSurface()) @@ -7265,6 +7957,9 @@ static inline bool hdrFormatMatchesVkSurfaceFormat(QRhiSwapChain::Format f, cons case QRhiSwapChain::HDR10: return (s.format == VK_FORMAT_A2B10G10R10_UNORM_PACK32 || s.format == VK_FORMAT_A2R10G10B10_UNORM_PACK32) && s.colorSpace == VK_COLOR_SPACE_HDR10_ST2084_EXT; + case QRhiSwapChain::HDRExtendedDisplayP3Linear: + return s.format == VK_FORMAT_R16G16B16A16_SFLOAT + && s.colorSpace == VK_COLOR_SPACE_DISPLAY_P3_LINEAR_EXT; default: break; } @@ -7374,11 +8069,9 @@ bool QVkSwapChain::ensureSurface() const bool srgbRequested = m_flags.testFlag(sRGB); for (int i = 0; i < int(formatCount); ++i) { if (formats[i].format != VK_FORMAT_UNDEFINED) { - bool ok = false; - if (m_format == SDR) - ok = srgbRequested == isSrgbFormat(formats[i].format); - else - ok = hdrFormatMatchesVkSurfaceFormat(m_format, formats[i]); + bool ok = srgbRequested == isSrgbFormat(formats[i].format); + if (m_format != SDR) + ok &= hdrFormatMatchesVkSurfaceFormat(m_format, formats[i]); if (ok) { colorFormat = formats[i].format; colorSpace = formats[i].colorSpace; @@ -7387,7 +8080,7 @@ bool QVkSwapChain::ensureSurface() } } - samples = rhiD->effectiveSampleCount(m_sampleCount); + samples = rhiD->effectiveSampleCountBits(m_sampleCount); quint32 presModeCount = 0; rhiD->vkGetPhysicalDeviceSurfacePresentModesKHR(rhiD->physDev, surface, &presModeCount, nullptr); @@ -7458,6 +8151,7 @@ bool QVkSwapChain::createOrResize() rtWrapper.d.dsAttCount = 0; ds = nullptr; } + rtWrapper.d.dsResolveAttCount = 0; if (samples > VK_SAMPLE_COUNT_1_BIT) rtWrapper.d.resolveAttCount = 1; else @@ -7474,7 +8168,7 @@ bool QVkSwapChain::createOrResize() VkFramebufferCreateInfo fbInfo = {}; fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; fbInfo.renderPass = rtWrapper.d.rp->rp; - fbInfo.attachmentCount = uint32_t(rtWrapper.d.colorAttCount + rtWrapper.d.dsAttCount + rtWrapper.d.resolveAttCount); + fbInfo.attachmentCount = uint32_t(rtWrapper.d.colorAttCount + rtWrapper.d.dsAttCount + rtWrapper.d.resolveAttCount + rtWrapper.d.dsResolveAttCount); fbInfo.pAttachments = views; fbInfo.width = uint32_t(pixelSize.width()); fbInfo.height = uint32_t(pixelSize.height()); @@ -7487,6 +8181,56 @@ bool QVkSwapChain::createOrResize() } } + if (stereo) { + rtWrapperRight.setRenderPassDescriptor( + m_renderPassDesc); // for the public getter in QRhiRenderTarget + rtWrapperRight.d.rp = QRHI_RES(QVkRenderPassDescriptor, m_renderPassDesc); + Q_ASSERT(rtWrapperRight.d.rp && rtWrapperRight.d.rp->rp); + + rtWrapperRight.d.pixelSize = pixelSize; + rtWrapperRight.d.dpr = float(window->devicePixelRatio()); + rtWrapperRight.d.sampleCount = samples; + rtWrapperRight.d.colorAttCount = 1; + if (m_depthStencil) { + rtWrapperRight.d.dsAttCount = 1; + ds = QRHI_RES(QVkRenderBuffer, m_depthStencil); + } else { + rtWrapperRight.d.dsAttCount = 0; + ds = nullptr; + } + rtWrapperRight.d.dsResolveAttCount = 0; + if (samples > VK_SAMPLE_COUNT_1_BIT) + rtWrapperRight.d.resolveAttCount = 1; + else + rtWrapperRight.d.resolveAttCount = 0; + + for (int i = 0; i < bufferCount; ++i) { + QVkSwapChain::ImageResources &image(imageRes[i + bufferCount]); + VkImageView views[3] = { + // color, ds, resolve + samples > VK_SAMPLE_COUNT_1_BIT ? image.msaaImageView : image.imageView, + ds ? ds->imageView : VK_NULL_HANDLE, + samples > VK_SAMPLE_COUNT_1_BIT ? image.imageView : VK_NULL_HANDLE + }; + + VkFramebufferCreateInfo fbInfo = {}; + fbInfo.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO; + fbInfo.renderPass = rtWrapperRight.d.rp->rp; + fbInfo.attachmentCount = uint32_t(rtWrapperRight.d.colorAttCount + rtWrapperRight.d.dsAttCount + + rtWrapperRight.d.resolveAttCount + rtWrapperRight.d.dsResolveAttCount); + fbInfo.pAttachments = views; + fbInfo.width = uint32_t(pixelSize.width()); + fbInfo.height = uint32_t(pixelSize.height()); + fbInfo.layers = 1; + + VkResult err = rhiD->df->vkCreateFramebuffer(rhiD->dev, &fbInfo, nullptr, &image.fb); + if (err != VK_SUCCESS) { + qWarning("Failed to create framebuffer: %d", err); + return false; + } + } + } + frameCount = 0; if (needsRegistration) diff --git a/src/gui/rhi/qrhivulkan_p.h b/src/gui/rhi/qrhivulkan_p.h index f592cf7e54..f23d8550f0 100644 --- a/src/gui/rhi/qrhivulkan_p.h +++ b/src/gui/rhi/qrhivulkan_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QRHIVULKAN_H -#define QRHIVULKAN_H +#ifndef QRHIVULKAN_P_H +#define QRHIVULKAN_P_H // // W A R N I N G @@ -15,44 +15,1025 @@ // We mean it. // -#include <private/qrhi_p.h> -#include <QtGui/qvulkaninstance.h> // this is where vulkan.h gets pulled in +#include "qrhi_p.h" QT_BEGIN_NAMESPACE -struct Q_GUI_EXPORT QRhiVulkanInitParams : public QRhiInitParams +class QVulkanFunctions; +class QVulkanDeviceFunctions; + +static const int QVK_FRAMES_IN_FLIGHT = 2; + +static const int QVK_DESC_SETS_PER_POOL = 128; +static const int QVK_UNIFORM_BUFFERS_PER_POOL = 256; +static const int QVK_COMBINED_IMAGE_SAMPLERS_PER_POOL = 256; +static const int QVK_STORAGE_BUFFERS_PER_POOL = 128; +static const int QVK_STORAGE_IMAGES_PER_POOL = 128; + +static const int QVK_MAX_ACTIVE_TIMESTAMP_PAIRS = 16; + +// no vk_mem_alloc.h available here, void* is good enough +typedef void * QVkAlloc; +typedef void * QVkAllocator; + +struct QVkBuffer : public QRhiBuffer { - QVulkanInstance *inst = nullptr; + QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); + ~QVkBuffer(); + void destroy() override; + bool create() override; + QRhiBuffer::NativeBuffer nativeBuffer() override; + char *beginFullDynamicBufferUpdateForCurrentFrame() override; + void endFullDynamicBufferUpdateForCurrentFrame() override; + + VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; + struct DynamicUpdate { + quint32 offset; + QRhiBufferData data; + }; + QVarLengthArray<DynamicUpdate, 16> pendingDynamicUpdates[QVK_FRAMES_IN_FLIGHT]; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + struct UsageState { + VkAccessFlags access = 0; + VkPipelineStageFlags stage = 0; + }; + UsageState usageState[QVK_FRAMES_IN_FLIGHT]; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +Q_DECLARE_TYPEINFO(QVkBuffer::DynamicUpdate, Q_RELOCATABLE_TYPE); + +struct QVkTexture; + +struct QVkRenderBuffer : public QRhiRenderBuffer +{ + QVkRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, + int sampleCount, Flags flags, + QRhiTexture::Format backingFormatHint); + ~QVkRenderBuffer(); + void destroy() override; + bool create() override; + QRhiTexture::Format backingFormat() const override; + + VkDeviceMemory memory = VK_NULL_HANDLE; + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + VkSampleCountFlagBits samples; + QVkTexture *backingTexture = nullptr; + VkFormat vkformat; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkTexture : public QRhiTexture +{ + QVkTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, + int arraySize, int sampleCount, Flags flags); + ~QVkTexture(); + void destroy() override; + bool create() override; + bool createFrom(NativeTexture src) override; + NativeTexture nativeTexture() override; + void setNativeLayout(int layout) override; + + bool prepareCreate(QSize *adjustedSize = nullptr); + bool finishCreate(); + VkImageView perLevelImageViewForLoadStore(int level); + + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + QVkAlloc imageAlloc = nullptr; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + VkImageView perLevelImageViews[QRhi::MAX_MIP_LEVELS]; + bool owns = true; + struct UsageState { + // no tracking of subresource layouts (some operations can keep + // subresources in different layouts for some time, but that does not + // need to be kept track of) + VkImageLayout layout; + VkAccessFlags access; + VkPipelineStageFlags stage; + }; + UsageState usageState; + VkFormat vkformat; + uint mipLevelCount = 0; + VkSampleCountFlagBits samples; + VkFormat viewFormat; + VkFormat viewFormatForSampling; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkSampler : public QRhiSampler +{ + QVkSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, + AddressMode u, AddressMode v, AddressMode w); + ~QVkSampler(); + void destroy() override; + bool create() override; + + VkSampler sampler = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkRenderPassDescriptor : public QRhiRenderPassDescriptor +{ + QVkRenderPassDescriptor(QRhiImplementation *rhi); + ~QVkRenderPassDescriptor(); + void destroy() override; + bool isCompatible(const QRhiRenderPassDescriptor *other) const override; + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; + QVector<quint32> serializedFormat() const override; + const QRhiNativeHandles *nativeHandles() override; + + void updateSerializedFormat(); + + VkRenderPass rp = VK_NULL_HANDLE; + bool ownsRp = false; + QVarLengthArray<VkAttachmentDescription, 8> attDescs; + QVarLengthArray<VkAttachmentReference, 8> colorRefs; + QVarLengthArray<VkAttachmentReference, 8> resolveRefs; + QVarLengthArray<VkSubpassDependency, 2> subpassDeps; + bool hasDepthStencil = false; + bool hasDepthStencilResolve = false; + uint32_t multiViewCount = 0; + VkAttachmentReference dsRef; + VkAttachmentReference dsResolveRef; + QVector<quint32> serializedFormatData; + QRhiVulkanRenderPassNativeHandles nativeHandlesStruct; + int lastActiveFrameSlot = -1; +}; + +struct QVkRenderTargetData +{ + VkFramebuffer fb = VK_NULL_HANDLE; + QVkRenderPassDescriptor *rp = nullptr; + QSize pixelSize; + float dpr = 1; + int sampleCount = 1; + int colorAttCount = 0; + int dsAttCount = 0; + int resolveAttCount = 0; + int dsResolveAttCount = 0; + int multiViewCount = 0; + QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; + static const int MAX_COLOR_ATTACHMENTS = 8; +}; + +struct QVkSwapChainRenderTarget : public QRhiSwapChainRenderTarget +{ + QVkSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); + ~QVkSwapChainRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QVkRenderTargetData d; +}; + +struct QVkTextureRenderTarget : public QRhiTextureRenderTarget +{ + QVkTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); + ~QVkTextureRenderTarget(); + void destroy() override; + + QSize pixelSize() const override; + float devicePixelRatio() const override; + int sampleCount() const override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool create() override; + + QVkRenderTargetData d; + VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView dsv = VK_NULL_HANDLE; + VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView resdsv = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + friend class QRhiVulkan; +}; + +struct QVkShaderResourceBindings : public QRhiShaderResourceBindings +{ + QVkShaderResourceBindings(QRhiImplementation *rhi); + ~QVkShaderResourceBindings(); + void destroy() override; + bool create() override; + void updateResources(UpdateFlags flags) override; + + QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; + bool hasSlottedResource = false; + bool hasDynamicOffset = false; + int poolIndex = -1; + VkDescriptorSetLayout layout = VK_NULL_HANDLE; + VkDescriptorSet descSets[QVK_FRAMES_IN_FLIGHT]; // multiple sets to support dynamic buffers + int lastActiveFrameSlot = -1; + uint generation = 0; + + // Keep track of the generation number of each referenced QRhi* to be able + // to detect that the underlying descriptor set became out of date and they + // need to be written again with the up-to-date VkBuffer etc. objects. + struct BoundUniformBufferData { + quint64 id; + uint generation; + }; + struct BoundSampledTextureData { + int count; + struct { + quint64 texId; + uint texGeneration; + quint64 samplerId; + uint samplerGeneration; + } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; + }; + struct BoundStorageImageData { + quint64 id; + uint generation; + }; + struct BoundStorageBufferData { + quint64 id; + uint generation; + }; + struct BoundResourceData { + union { + BoundUniformBufferData ubuf; + BoundSampledTextureData stex; + BoundStorageImageData simage; + BoundStorageBufferData sbuf; + }; + }; + QVarLengthArray<BoundResourceData, 8> boundResourceData[QVK_FRAMES_IN_FLIGHT]; + + friend class QRhiVulkan; +}; + +Q_DECLARE_TYPEINFO(QVkShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); + +struct QVkGraphicsPipeline : public QRhiGraphicsPipeline +{ + QVkGraphicsPipeline(QRhiImplementation *rhi); + ~QVkGraphicsPipeline(); + void destroy() override; + bool create() override; + + VkPipelineLayout layout = VK_NULL_HANDLE; + VkPipeline pipeline = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkComputePipeline : public QRhiComputePipeline +{ + QVkComputePipeline(QRhiImplementation *rhi); + ~QVkComputePipeline(); + void destroy() override; + bool create() override; + + VkPipelineLayout layout = VK_NULL_HANDLE; + VkPipeline pipeline = VK_NULL_HANDLE; + int lastActiveFrameSlot = -1; + uint generation = 0; + friend class QRhiVulkan; +}; + +struct QVkCommandBuffer : public QRhiCommandBuffer +{ + QVkCommandBuffer(QRhiImplementation *rhi); + ~QVkCommandBuffer(); + void destroy() override; + + const QRhiNativeHandles *nativeHandles(); + + VkCommandBuffer cb = VK_NULL_HANDLE; // primary + QRhiVulkanCommandBufferNativeHandles nativeHandlesStruct; + + enum PassType { + NoPass, + RenderPass, + ComputePass + }; + + void resetState() { + recordingPass = NoPass; + passUsesSecondaryCb = false; + lastGpuTime = 0; + currentTarget = nullptr; + activeSecondaryCbStack.clear(); + resetCommands(); + resetCachedState(); + } + + void resetCachedState() { + currentGraphicsPipeline = nullptr; + currentComputePipeline = nullptr; + currentPipelineGeneration = 0; + currentGraphicsSrb = nullptr; + currentComputeSrb = nullptr; + currentSrbGeneration = 0; + currentDescSetSlot = -1; + currentIndexBuffer = VK_NULL_HANDLE; + currentIndexOffset = 0; + currentIndexFormat = VK_INDEX_TYPE_UINT16; + memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); + memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); + inExternal = false; + } + + PassType recordingPass; + bool passUsesSecondaryCb; + double lastGpuTime = 0; + QRhiRenderTarget *currentTarget; + QRhiGraphicsPipeline *currentGraphicsPipeline; + QRhiComputePipeline *currentComputePipeline; + uint currentPipelineGeneration; + QRhiShaderResourceBindings *currentGraphicsSrb; + QRhiShaderResourceBindings *currentComputeSrb; + uint currentSrbGeneration; + int currentDescSetSlot; + VkBuffer currentIndexBuffer; + quint32 currentIndexOffset; + VkIndexType currentIndexFormat; + static const int VERTEX_INPUT_RESOURCE_SLOT_COUNT = 32; + VkBuffer currentVertexBuffers[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + quint32 currentVertexOffsets[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; + QVarLengthArray<VkCommandBuffer, 4> activeSecondaryCbStack; + bool inExternal; + + struct { + QHash<QRhiResource *, QPair<VkAccessFlags, bool> > writtenResources; + void reset() { + writtenResources.clear(); + } + } computePassState; + + struct Command { + enum Cmd { + CopyBuffer, + CopyBufferToImage, + CopyImage, + CopyImageToBuffer, + ImageBarrier, + BufferBarrier, + BlitImage, + BeginRenderPass, + EndRenderPass, + BindPipeline, + BindDescriptorSet, + BindVertexBuffer, + BindIndexBuffer, + SetViewport, + SetScissor, + SetBlendConstants, + SetStencilRef, + Draw, + DrawIndexed, + DebugMarkerBegin, + DebugMarkerEnd, + DebugMarkerInsert, + TransitionPassResources, + Dispatch, + ExecuteSecondary + }; + Cmd cmd; + + union Args { + struct { + VkBuffer src; + VkBuffer dst; + VkBufferCopy desc; + } copyBuffer; + struct { + VkBuffer src; + VkImage dst; + VkImageLayout dstLayout; + int count; + int bufferImageCopyIndex; + } copyBufferToImage; + struct { + VkImage src; + VkImageLayout srcLayout; + VkImage dst; + VkImageLayout dstLayout; + VkImageCopy desc; + } copyImage; + struct { + VkImage src; + VkImageLayout srcLayout; + VkBuffer dst; + VkBufferImageCopy desc; + } copyImageToBuffer; + struct { + VkPipelineStageFlags srcStageMask; + VkPipelineStageFlags dstStageMask; + int count; + int index; + } imageBarrier; + struct { + VkPipelineStageFlags srcStageMask; + VkPipelineStageFlags dstStageMask; + int count; + int index; + } bufferBarrier; + struct { + VkImage src; + VkImageLayout srcLayout; + VkImage dst; + VkImageLayout dstLayout; + VkFilter filter; + VkImageBlit desc; + } blitImage; + struct { + VkRenderPassBeginInfo desc; + int clearValueIndex; + bool useSecondaryCb; + } beginRenderPass; + struct { + } endRenderPass; + struct { + VkPipelineBindPoint bindPoint; + VkPipeline pipeline; + } bindPipeline; + struct { + VkPipelineBindPoint bindPoint; + VkPipelineLayout pipelineLayout; + VkDescriptorSet descSet; + int dynamicOffsetCount; + int dynamicOffsetIndex; + } bindDescriptorSet; + struct { + int startBinding; + int count; + int vertexBufferIndex; + int vertexBufferOffsetIndex; + } bindVertexBuffer; + struct { + VkBuffer buf; + VkDeviceSize ofs; + VkIndexType type; + } bindIndexBuffer; + struct { + VkViewport viewport; + } setViewport; + struct { + VkRect2D scissor; + } setScissor; + struct { + float c[4]; + } setBlendConstants; + struct { + uint32_t ref; + } setStencilRef; + struct { + uint32_t vertexCount; + uint32_t instanceCount; + uint32_t firstVertex; + uint32_t firstInstance; + } draw; + struct { + uint32_t indexCount; + uint32_t instanceCount; + uint32_t firstIndex; + int32_t vertexOffset; + uint32_t firstInstance; + } drawIndexed; + struct { +#ifdef VK_EXT_debug_utils + VkDebugUtilsLabelEXT label; + int labelNameIndex; +#endif + } debugMarkerBegin; + struct { + } debugMarkerEnd; + struct { +#ifdef VK_EXT_debug_utils + VkDebugUtilsLabelEXT label; + int labelNameIndex; +#endif + } debugMarkerInsert; + struct { + int trackerIndex; + } transitionResources; + struct { + int x, y, z; + } dispatch; + struct { + VkCommandBuffer cb; + } executeSecondary; + } args; + }; + + QRhiBackendCommandList<Command> commands; + QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; + int currentPassResTrackerIndex; + + void resetCommands() { + commands.reset(); + resetPools(); + + passResTrackers.clear(); + currentPassResTrackerIndex = -1; + } + + void resetPools() { + pools.clearValue.clear(); + pools.bufferImageCopy.clear(); + pools.dynamicOffset.clear(); + pools.vertexBuffer.clear(); + pools.vertexBufferOffset.clear(); + pools.debugMarkerData.clear(); + pools.imageBarrier.clear(); + pools.bufferBarrier.clear(); + } + + struct { + QVarLengthArray<VkClearValue, 4> clearValue; + QVarLengthArray<VkBufferImageCopy, 16> bufferImageCopy; + QVarLengthArray<uint32_t, 4> dynamicOffset; + QVarLengthArray<VkBuffer, 4> vertexBuffer; + QVarLengthArray<VkDeviceSize, 4> vertexBufferOffset; + QVarLengthArray<QByteArray, 4> debugMarkerData; + QVarLengthArray<VkImageMemoryBarrier, 8> imageBarrier; + QVarLengthArray<VkBufferMemoryBarrier, 8> bufferBarrier; + } pools; + + friend class QRhiVulkan; +}; + +struct QVkSwapChain : public QRhiSwapChain +{ + QVkSwapChain(QRhiImplementation *rhi); + ~QVkSwapChain(); + void destroy() override; + + QRhiCommandBuffer *currentFrameCommandBuffer() override; + QRhiRenderTarget *currentFrameRenderTarget() override; + QRhiRenderTarget *currentFrameRenderTarget(StereoTargetBuffer targetBuffer) override; + + QSize surfacePixelSize() override; + bool isFormatSupported(Format f) override; + + QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; + bool createOrResize() override; + + bool ensureSurface(); + + static const quint32 EXPECTED_MAX_BUFFER_COUNT = 4; + QWindow *window = nullptr; - QByteArrayList deviceExtensions; + QSize pixelSize; + bool supportsReadback = false; + bool stereo = false; + VkSwapchainKHR sc = VK_NULL_HANDLE; + int bufferCount = 0; + VkSurfaceKHR surface = VK_NULL_HANDLE; + VkSurfaceKHR lastConnectedSurface = VK_NULL_HANDLE; + VkFormat colorFormat = VK_FORMAT_B8G8R8A8_UNORM; + VkColorSpaceKHR colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; + QVkRenderBuffer *ds = nullptr; + VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT; + QVarLengthArray<VkPresentModeKHR, 8> supportedPresentationModes; + VkDeviceMemory msaaImageMem = VK_NULL_HANDLE; + QVkSwapChainRenderTarget rtWrapper; + QVkSwapChainRenderTarget rtWrapperRight; + QVkCommandBuffer cbWrapper; + + struct ImageResources { + VkImage image = VK_NULL_HANDLE; + VkImageView imageView = VK_NULL_HANDLE; + VkFramebuffer fb = VK_NULL_HANDLE; + VkImage msaaImage = VK_NULL_HANDLE; + VkImageView msaaImageView = VK_NULL_HANDLE; + enum LastUse { + ScImageUseNone, + ScImageUseRender, + ScImageUseTransferSource + }; + LastUse lastUse = ScImageUseNone; + }; + QVarLengthArray<ImageResources, EXPECTED_MAX_BUFFER_COUNT> imageRes; + + struct FrameResources { + VkFence imageFence = VK_NULL_HANDLE; + bool imageFenceWaitable = false; + VkSemaphore imageSem = VK_NULL_HANDLE; + VkSemaphore drawSem = VK_NULL_HANDLE; + bool imageAcquired = false; + bool imageSemWaitable = false; + VkFence cmdFence = VK_NULL_HANDLE; + bool cmdFenceWaitable = false; + VkCommandBuffer cmdBuf = VK_NULL_HANDLE; // primary + int timestampQueryIndex = -1; + } frameRes[QVK_FRAMES_IN_FLIGHT]; - static QByteArrayList preferredInstanceExtensions(); - static QByteArrayList preferredExtensionsForImportedDevice(); + quint32 currentImageIndex = 0; // index in imageRes + quint32 currentFrameSlot = 0; // index in frameRes + int frameCount = 0; + + friend class QRhiVulkan; }; -struct Q_GUI_EXPORT QRhiVulkanNativeHandles : public QRhiNativeHandles +class QRhiVulkan : public QRhiImplementation { - // to import a physical device (always required) +public: + QRhiVulkan(QRhiVulkanInitParams *params, QRhiVulkanNativeHandles *importParams = nullptr); + + bool create(QRhi::Flags flags) override; + void destroy() override; + + QRhiGraphicsPipeline *createGraphicsPipeline() override; + QRhiComputePipeline *createComputePipeline() override; + QRhiShaderResourceBindings *createShaderResourceBindings() override; + QRhiBuffer *createBuffer(QRhiBuffer::Type type, + QRhiBuffer::UsageFlags usage, + quint32 size) override; + QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, + const QSize &pixelSize, + int sampleCount, + QRhiRenderBuffer::Flags flags, + QRhiTexture::Format backingFormatHint) override; + QRhiTexture *createTexture(QRhiTexture::Format format, + const QSize &pixelSize, + int depth, + int arraySize, + int sampleCount, + QRhiTexture::Flags flags) override; + QRhiSampler *createSampler(QRhiSampler::Filter magFilter, + QRhiSampler::Filter minFilter, + QRhiSampler::Filter mipmapMode, + QRhiSampler:: AddressMode u, + QRhiSampler::AddressMode v, + QRhiSampler::AddressMode w) override; + + QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, + QRhiTextureRenderTarget::Flags flags) override; + + QRhiSwapChain *createSwapChain() override; + QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; + QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; + QRhi::FrameOpResult finish() override; + + void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void beginPass(QRhiCommandBuffer *cb, + QRhiRenderTarget *rt, + const QColor &colorClearValue, + const QRhiDepthStencilClearValue &depthStencilClearValue, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + + void setGraphicsPipeline(QRhiCommandBuffer *cb, + QRhiGraphicsPipeline *ps) override; + + void setShaderResources(QRhiCommandBuffer *cb, + QRhiShaderResourceBindings *srb, + int dynamicOffsetCount, + const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; + + void setVertexInput(QRhiCommandBuffer *cb, + int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, + QRhiBuffer *indexBuf, quint32 indexOffset, + QRhiCommandBuffer::IndexFormat indexFormat) override; + + void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; + void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; + void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; + void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; + + void draw(QRhiCommandBuffer *cb, quint32 vertexCount, + quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; + + void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, + quint32 instanceCount, quint32 firstIndex, + qint32 vertexOffset, quint32 firstInstance) override; + + void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; + void debugMarkEnd(QRhiCommandBuffer *cb) override; + void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; + + void beginComputePass(QRhiCommandBuffer *cb, + QRhiResourceUpdateBatch *resourceUpdates, + QRhiCommandBuffer::BeginPassFlags flags) override; + void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; + void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; + void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; + + const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; + void beginExternal(QRhiCommandBuffer *cb) override; + void endExternal(QRhiCommandBuffer *cb) override; + double lastCompletedGpuTime(QRhiCommandBuffer *cb) override; + + QList<int> supportedSampleCounts() const override; + int ubufAlignment() const override; + bool isYUpInFramebuffer() const override; + bool isYUpInNDC() const override; + bool isClipDepthZeroToOne() const override; + QMatrix4x4 clipSpaceCorrMatrix() const override; + bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; + bool isFeatureSupported(QRhi::Feature feature) const override; + int resourceLimit(QRhi::ResourceLimit limit) const override; + const QRhiNativeHandles *nativeHandles() override; + QRhiDriverInfo driverInfo() const override; + QRhiStats statistics() override; + bool makeThreadLocalNativeContextCurrent() override; + void releaseCachedResources() override; + bool isDeviceLost() const override; + + QByteArray pipelineCacheData() override; + void setPipelineCacheData(const QByteArray &data) override; + + VkResult createDescriptorPool(VkDescriptorPool *pool); + bool allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, VkDescriptorSet *result, int *resultPoolIndex); + uint32_t chooseTransientImageMemType(VkImage img, uint32_t startIndex); + bool createTransientImage(VkFormat format, const QSize &pixelSize, VkImageUsageFlags usage, + VkImageAspectFlags aspectMask, VkSampleCountFlagBits samples, + VkDeviceMemory *mem, VkImage *images, VkImageView *views, int count); + + bool recreateSwapChain(QRhiSwapChain *swapChain); + void releaseSwapChainResources(QRhiSwapChain *swapChain); + + VkFormat optimalDepthStencilFormat(); + VkSampleCountFlagBits effectiveSampleCountBits(int sampleCount); + bool createDefaultRenderPass(QVkRenderPassDescriptor *rpD, + bool hasDepthStencil, + VkSampleCountFlagBits samples, + VkFormat colorFormat); + bool createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, + const QRhiColorAttachment *colorAttachmentsBegin, + const QRhiColorAttachment *colorAttachmentsEnd, + bool preserveColor, + bool preserveDs, + bool storeDs, + QRhiRenderBuffer *depthStencilBuffer, + QRhiTexture *depthTexture, + QRhiTexture *depthResolveTexture); + bool ensurePipelineCache(const void *initialData = nullptr, size_t initialDataSize = 0); + VkShaderModule createShader(const QByteArray &spirv); + + void prepareNewFrame(QRhiCommandBuffer *cb); + VkCommandBuffer startSecondaryCommandBuffer(QVkRenderTargetData *rtD = nullptr); + void endAndEnqueueSecondaryCommandBuffer(VkCommandBuffer cb, QVkCommandBuffer *cbD); + QRhi::FrameOpResult startPrimaryCommandBuffer(VkCommandBuffer *cb); + QRhi::FrameOpResult endAndSubmitPrimaryCommandBuffer(VkCommandBuffer cb, VkFence cmdFence, + VkSemaphore *waitSem, VkSemaphore *signalSem); + void waitCommandCompletion(int frameSlot); + VkDeviceSize subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; + using BufferImageCopyList = QVarLengthArray<VkBufferImageCopy, 16>; + void prepareUploadSubres(QVkTexture *texD, int layer, int level, + const QRhiTextureSubresourceUploadDescription &subresDesc, + size_t *curOfs, void *mp, + BufferImageCopyList *copyInfos); + void enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); + void executeBufferHostWritesForSlot(QVkBuffer *bufD, int slot); + void enqueueTransitionPassResources(QVkCommandBuffer *cbD); + void recordPrimaryCommandBuffer(QVkCommandBuffer *cbD); + void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, + QVkBuffer *bufD, + int slot, + QRhiPassResourceTracker::BufferAccess access, + QRhiPassResourceTracker::BufferStage stage); + void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, + QVkTexture *texD, + QRhiPassResourceTracker::TextureAccess access, + QRhiPassResourceTracker::TextureStage stage); + void recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhiPassResourceTracker &tracker); + void activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRenderTarget *rtD); + void executeDeferredReleases(bool forced = false); + void finishActiveReadbacks(bool forced = false); + + void setObjectName(uint64_t object, VkObjectType type, const QByteArray &name, int slot = -1); + void trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, int slot, + VkAccessFlags access, VkPipelineStageFlags stage); + void trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, + VkImageLayout layout, VkAccessFlags access, VkPipelineStageFlags stage); + void depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuffer *rbD); + void subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, + VkImageLayout oldLayout, VkImageLayout newLayout, + VkAccessFlags srcAccess, VkAccessFlags dstAccess, + VkPipelineStageFlags srcStage, VkPipelineStageFlags dstStage, + int startLayer, int layerCount, + int startLevel, int levelCount); + void updateShaderResourceBindings(QRhiShaderResourceBindings *srb, int descSetIdx = -1); + void ensureCommandPoolForNewFrame(); + double elapsedSecondsFromTimestamp(quint64 timestamp[2], bool *ok); + void printExtraErrorInfo(VkResult err); + + QVulkanInstance *inst = nullptr; + QWindow *maybeWindow = nullptr; + QByteArrayList requestedDeviceExtensions; + bool importedDevice = false; VkPhysicalDevice physDev = VK_NULL_HANDLE; - // to import a device and queue VkDevice dev = VK_NULL_HANDLE; + VkCommandPool cmdPool[QVK_FRAMES_IN_FLIGHT] = {}; quint32 gfxQueueFamilyIdx = 0; quint32 gfxQueueIdx = 0; VkQueue gfxQueue = VK_NULL_HANDLE; - // and optionally, the mem allocator - void *vmemAllocator = nullptr; -}; + quint32 timestampValidBits = 0; + bool importedAllocator = false; + QVkAllocator allocator = nullptr; + QVulkanFunctions *f = nullptr; + QVulkanDeviceFunctions *df = nullptr; + QRhi::Flags rhiFlags; + VkPhysicalDeviceFeatures physDevFeatures; +#ifdef VK_VERSION_1_1 + VkPhysicalDeviceMultiviewFeatures multiviewFeaturesIfApi11; +#endif +#ifdef VK_VERSION_1_2 + VkPhysicalDeviceVulkan11Features physDevFeatures11IfApi12OrNewer; + VkPhysicalDeviceVulkan12Features physDevFeatures12; +#endif +#ifdef VK_VERSION_1_3 + VkPhysicalDeviceVulkan13Features physDevFeatures13; +#endif + VkPhysicalDeviceProperties physDevProperties; + VkDeviceSize ubufAlign; + VkDeviceSize texbufAlign; + bool deviceLost = false; + bool releaseCachedResourcesCalledBeforeFrameStart = false; -struct Q_GUI_EXPORT QRhiVulkanCommandBufferNativeHandles : public QRhiNativeHandles -{ - VkCommandBuffer commandBuffer = VK_NULL_HANDLE; -}; +#ifdef VK_EXT_debug_utils + PFN_vkSetDebugUtilsObjectNameEXT vkSetDebugUtilsObjectNameEXT = nullptr; + PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT = nullptr; + PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT = nullptr; + PFN_vkCmdInsertDebugUtilsLabelEXT vkCmdInsertDebugUtilsLabelEXT = nullptr; +#endif -struct Q_GUI_EXPORT QRhiVulkanRenderPassNativeHandles : public QRhiNativeHandles -{ - VkRenderPass renderPass = VK_NULL_HANDLE; + PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR = nullptr; + PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR; + PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR; + PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR; + PFN_vkQueuePresentKHR vkQueuePresentKHR; + PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR; + PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR; + PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR; + +#ifdef VK_KHR_create_renderpass2 + PFN_vkCreateRenderPass2KHR vkCreateRenderPass2KHR = nullptr; +#endif + + struct { + bool compute = false; + bool wideLines = false; + bool debugUtils = false; + bool vertexAttribDivisor = false; + bool texture3DSliceAs2D = false; + bool tessellation = false; + bool geometryShader = false; + bool nonFillPolygonMode = false; + bool multiView = false; + bool renderPass2KHR = false; + bool depthStencilResolveKHR = false; + QVersionNumber apiVersion; + } caps; + + VkPipelineCache pipelineCache = VK_NULL_HANDLE; + struct DescriptorPoolData { + DescriptorPoolData() { } + DescriptorPoolData(VkDescriptorPool pool_) + : pool(pool_) + { } + VkDescriptorPool pool = VK_NULL_HANDLE; + int refCount = 0; + int allocedDescSets = 0; + }; + QVarLengthArray<DescriptorPoolData, 8> descriptorPools; + QVarLengthArray<VkCommandBuffer, 4> freeSecondaryCbs[QVK_FRAMES_IN_FLIGHT]; + + VkQueryPool timestampQueryPool = VK_NULL_HANDLE; + QBitArray timestampQueryPoolMap; + + VkFormat optimalDsFormat = VK_FORMAT_UNDEFINED; + QMatrix4x4 clipCorrectMatrix; + + QVkSwapChain *currentSwapChain = nullptr; + QSet<QVkSwapChain *> swapchains; + QRhiVulkanNativeHandles nativeHandlesStruct; + QRhiDriverInfo driverInfoStruct; + + struct OffscreenFrame { + OffscreenFrame(QRhiImplementation *rhi) + { + for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) + cbWrapper[i] = new QVkCommandBuffer(rhi); + } + ~OffscreenFrame() + { + for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) + delete cbWrapper[i]; + } + bool active = false; + QVkCommandBuffer *cbWrapper[QVK_FRAMES_IN_FLIGHT]; + VkFence cmdFence = VK_NULL_HANDLE; + int timestampQueryIndex = -1; + } ofr; + + struct TextureReadback { + int activeFrameSlot = -1; + QRhiReadbackDescription desc; + QRhiReadbackResult *result; + VkBuffer stagingBuf; + QVkAlloc stagingAlloc; + quint32 byteSize; + QSize pixelSize; + QRhiTexture::Format format; + }; + QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; + struct BufferReadback { + int activeFrameSlot = -1; + QRhiReadbackResult *result; + quint32 byteSize; + VkBuffer stagingBuf; + QVkAlloc stagingAlloc; + }; + QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; + + struct DeferredReleaseEntry { + enum Type { + Pipeline, + ShaderResourceBindings, + Buffer, + RenderBuffer, + Texture, + Sampler, + TextureRenderTarget, + RenderPass, + StagingBuffer, + SecondaryCommandBuffer + }; + Type type; + int lastActiveFrameSlot; // -1 if not used otherwise 0..FRAMES_IN_FLIGHT-1 + union { + struct { + VkPipeline pipeline; + VkPipelineLayout layout; + } pipelineState; + struct { + int poolIndex; + VkDescriptorSetLayout layout; + } shaderResourceBindings; + struct { + VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + } buffer; + struct { + VkDeviceMemory memory; + VkImage image; + VkImageView imageView; + } renderBuffer; + struct { + VkImage image; + VkImageView imageView; + QVkAlloc allocation; + VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; + QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; + VkImageView extraImageViews[QRhi::MAX_MIP_LEVELS]; + } texture; + struct { + VkSampler sampler; + } sampler; + struct { + VkFramebuffer fb; + VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; + VkImageView dsv; + VkImageView resdsv; + } textureRenderTarget; + struct { + VkRenderPass rp; + } renderPass; + struct { + VkBuffer stagingBuffer; + QVkAlloc stagingAllocation; + } stagingBuffer; + struct { + VkCommandBuffer cb; + } secondaryCommandBuffer; + }; + }; + QList<DeferredReleaseEntry> releaseQueue; }; +Q_DECLARE_TYPEINFO(QRhiVulkan::DescriptorPoolData, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::TextureReadback, Q_RELOCATABLE_TYPE); +Q_DECLARE_TYPEINFO(QRhiVulkan::BufferReadback, Q_RELOCATABLE_TYPE); + QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qrhivulkan_p_p.h b/src/gui/rhi/qrhivulkan_p_p.h deleted file mode 100644 index 22214723ae..0000000000 --- a/src/gui/rhi/qrhivulkan_p_p.h +++ /dev/null @@ -1,1001 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHIVULKAN_P_H -#define QRHIVULKAN_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhivulkan_p.h" -#include "qrhi_p_p.h" - -QT_BEGIN_NAMESPACE - -class QVulkanFunctions; -class QVulkanDeviceFunctions; - -static const int QVK_FRAMES_IN_FLIGHT = 2; - -static const int QVK_DESC_SETS_PER_POOL = 128; -static const int QVK_UNIFORM_BUFFERS_PER_POOL = 256; -static const int QVK_COMBINED_IMAGE_SAMPLERS_PER_POOL = 256; -static const int QVK_STORAGE_BUFFERS_PER_POOL = 128; -static const int QVK_STORAGE_IMAGES_PER_POOL = 128; - -static const int QVK_MAX_ACTIVE_TIMESTAMP_PAIRS = 16; - -// no vk_mem_alloc.h available here, void* is good enough -typedef void * QVkAlloc; -typedef void * QVkAllocator; - -struct QVkBuffer : public QRhiBuffer -{ - QVkBuffer(QRhiImplementation *rhi, Type type, UsageFlags usage, quint32 size); - ~QVkBuffer(); - void destroy() override; - bool create() override; - QRhiBuffer::NativeBuffer nativeBuffer() override; - char *beginFullDynamicBufferUpdateForCurrentFrame() override; - void endFullDynamicBufferUpdateForCurrentFrame() override; - - VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; - struct DynamicUpdate { - quint32 offset; - QRhiBufferData data; - }; - QVarLengthArray<DynamicUpdate, 16> pendingDynamicUpdates[QVK_FRAMES_IN_FLIGHT]; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - struct UsageState { - VkAccessFlags access = 0; - VkPipelineStageFlags stage = 0; - }; - UsageState usageState[QVK_FRAMES_IN_FLIGHT]; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -Q_DECLARE_TYPEINFO(QVkBuffer::DynamicUpdate, Q_RELOCATABLE_TYPE); - -struct QVkTexture; - -struct QVkRenderBuffer : public QRhiRenderBuffer -{ - QVkRenderBuffer(QRhiImplementation *rhi, Type type, const QSize &pixelSize, - int sampleCount, Flags flags, - QRhiTexture::Format backingFormatHint); - ~QVkRenderBuffer(); - void destroy() override; - bool create() override; - QRhiTexture::Format backingFormat() const override; - - VkDeviceMemory memory = VK_NULL_HANDLE; - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - VkSampleCountFlagBits samples; - QVkTexture *backingTexture = nullptr; - VkFormat vkformat; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkTexture : public QRhiTexture -{ - QVkTexture(QRhiImplementation *rhi, Format format, const QSize &pixelSize, int depth, - int arraySize, int sampleCount, Flags flags); - ~QVkTexture(); - void destroy() override; - bool create() override; - bool createFrom(NativeTexture src) override; - NativeTexture nativeTexture() override; - void setNativeLayout(int layout) override; - - bool prepareCreate(QSize *adjustedSize = nullptr); - bool finishCreate(); - VkImageView imageViewForLevel(int level); - - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - QVkAlloc imageAlloc = nullptr; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - VkImageView perLevelImageViews[QRhi::MAX_MIP_LEVELS]; - bool owns = true; - struct UsageState { - // no tracking of subresource layouts (some operations can keep - // subresources in different layouts for some time, but that does not - // need to be kept track of) - VkImageLayout layout; - VkAccessFlags access; - VkPipelineStageFlags stage; - }; - UsageState usageState; - VkFormat vkformat; - uint mipLevelCount = 0; - VkSampleCountFlagBits samples; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkSampler : public QRhiSampler -{ - QVkSampler(QRhiImplementation *rhi, Filter magFilter, Filter minFilter, Filter mipmapMode, - AddressMode u, AddressMode v, AddressMode w); - ~QVkSampler(); - void destroy() override; - bool create() override; - - VkSampler sampler = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkRenderPassDescriptor : public QRhiRenderPassDescriptor -{ - QVkRenderPassDescriptor(QRhiImplementation *rhi); - ~QVkRenderPassDescriptor(); - void destroy() override; - bool isCompatible(const QRhiRenderPassDescriptor *other) const override; - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() const override; - QVector<quint32> serializedFormat() const override; - const QRhiNativeHandles *nativeHandles() override; - - void updateSerializedFormat(); - - VkRenderPass rp = VK_NULL_HANDLE; - bool ownsRp = false; - QVarLengthArray<VkAttachmentDescription, 8> attDescs; - QVarLengthArray<VkAttachmentReference, 8> colorRefs; - QVarLengthArray<VkAttachmentReference, 8> resolveRefs; - QVarLengthArray<VkSubpassDependency, 2> subpassDeps; - bool hasDepthStencil = false; - VkAttachmentReference dsRef; - QVector<quint32> serializedFormatData; - QRhiVulkanRenderPassNativeHandles nativeHandlesStruct; - int lastActiveFrameSlot = -1; -}; - -struct QVkRenderTargetData -{ - VkFramebuffer fb = VK_NULL_HANDLE; - QVkRenderPassDescriptor *rp = nullptr; - QSize pixelSize; - float dpr = 1; - int sampleCount = 1; - int colorAttCount = 0; - int dsAttCount = 0; - int resolveAttCount = 0; - QRhiRenderTargetAttachmentTracker::ResIdList currentResIdList; - static const int MAX_COLOR_ATTACHMENTS = 8; -}; - -struct QVkSwapChainRenderTarget : public QRhiSwapChainRenderTarget -{ - QVkSwapChainRenderTarget(QRhiImplementation *rhi, QRhiSwapChain *swapchain); - ~QVkSwapChainRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QVkRenderTargetData d; -}; - -struct QVkTextureRenderTarget : public QRhiTextureRenderTarget -{ - QVkTextureRenderTarget(QRhiImplementation *rhi, const QRhiTextureRenderTargetDescription &desc, Flags flags); - ~QVkTextureRenderTarget(); - void destroy() override; - - QSize pixelSize() const override; - float devicePixelRatio() const override; - int sampleCount() const override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool create() override; - - QVkRenderTargetData d; - VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - int lastActiveFrameSlot = -1; - friend class QRhiVulkan; -}; - -struct QVkShaderResourceBindings : public QRhiShaderResourceBindings -{ - QVkShaderResourceBindings(QRhiImplementation *rhi); - ~QVkShaderResourceBindings(); - void destroy() override; - bool create() override; - void updateResources(UpdateFlags flags) override; - - QVarLengthArray<QRhiShaderResourceBinding, 8> sortedBindings; - bool hasSlottedResource = false; - bool hasDynamicOffset = false; - int poolIndex = -1; - VkDescriptorSetLayout layout = VK_NULL_HANDLE; - VkDescriptorSet descSets[QVK_FRAMES_IN_FLIGHT]; // multiple sets to support dynamic buffers - int lastActiveFrameSlot = -1; - uint generation = 0; - - // Keep track of the generation number of each referenced QRhi* to be able - // to detect that the underlying descriptor set became out of date and they - // need to be written again with the up-to-date VkBuffer etc. objects. - struct BoundUniformBufferData { - quint64 id; - uint generation; - }; - struct BoundSampledTextureData { - int count; - struct { - quint64 texId; - uint texGeneration; - quint64 samplerId; - uint samplerGeneration; - } d[QRhiShaderResourceBinding::Data::MAX_TEX_SAMPLER_ARRAY_SIZE]; - }; - struct BoundStorageImageData { - quint64 id; - uint generation; - }; - struct BoundStorageBufferData { - quint64 id; - uint generation; - }; - struct BoundResourceData { - union { - BoundUniformBufferData ubuf; - BoundSampledTextureData stex; - BoundStorageImageData simage; - BoundStorageBufferData sbuf; - }; - }; - QVarLengthArray<BoundResourceData, 8> boundResourceData[QVK_FRAMES_IN_FLIGHT]; - - friend class QRhiVulkan; -}; - -Q_DECLARE_TYPEINFO(QVkShaderResourceBindings::BoundResourceData, Q_RELOCATABLE_TYPE); - -struct QVkGraphicsPipeline : public QRhiGraphicsPipeline -{ - QVkGraphicsPipeline(QRhiImplementation *rhi); - ~QVkGraphicsPipeline(); - void destroy() override; - bool create() override; - - VkPipelineLayout layout = VK_NULL_HANDLE; - VkPipeline pipeline = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkComputePipeline : public QRhiComputePipeline -{ - QVkComputePipeline(QRhiImplementation *rhi); - ~QVkComputePipeline(); - void destroy() override; - bool create() override; - - VkPipelineLayout layout = VK_NULL_HANDLE; - VkPipeline pipeline = VK_NULL_HANDLE; - int lastActiveFrameSlot = -1; - uint generation = 0; - friend class QRhiVulkan; -}; - -struct QVkCommandBuffer : public QRhiCommandBuffer -{ - QVkCommandBuffer(QRhiImplementation *rhi); - ~QVkCommandBuffer(); - void destroy() override; - - const QRhiNativeHandles *nativeHandles(); - - VkCommandBuffer cb = VK_NULL_HANDLE; // primary - QRhiVulkanCommandBufferNativeHandles nativeHandlesStruct; - - enum PassType { - NoPass, - RenderPass, - ComputePass - }; - - void resetState() { - recordingPass = NoPass; - passUsesSecondaryCb = false; - currentTarget = nullptr; - activeSecondaryCbStack.clear(); - resetCommands(); - resetCachedState(); - } - - void resetCachedState() { - currentGraphicsPipeline = nullptr; - currentComputePipeline = nullptr; - currentPipelineGeneration = 0; - currentGraphicsSrb = nullptr; - currentComputeSrb = nullptr; - currentSrbGeneration = 0; - currentDescSetSlot = -1; - currentIndexBuffer = VK_NULL_HANDLE; - currentIndexOffset = 0; - currentIndexFormat = VK_INDEX_TYPE_UINT16; - memset(currentVertexBuffers, 0, sizeof(currentVertexBuffers)); - memset(currentVertexOffsets, 0, sizeof(currentVertexOffsets)); - inExternal = false; - } - - PassType recordingPass; - bool passUsesSecondaryCb; - QRhiRenderTarget *currentTarget; - QRhiGraphicsPipeline *currentGraphicsPipeline; - QRhiComputePipeline *currentComputePipeline; - uint currentPipelineGeneration; - QRhiShaderResourceBindings *currentGraphicsSrb; - QRhiShaderResourceBindings *currentComputeSrb; - uint currentSrbGeneration; - int currentDescSetSlot; - VkBuffer currentIndexBuffer; - quint32 currentIndexOffset; - VkIndexType currentIndexFormat; - static const int VERTEX_INPUT_RESOURCE_SLOT_COUNT = 32; - VkBuffer currentVertexBuffers[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - quint32 currentVertexOffsets[VERTEX_INPUT_RESOURCE_SLOT_COUNT]; - QVarLengthArray<VkCommandBuffer, 4> activeSecondaryCbStack; - bool inExternal; - - struct { - QHash<QRhiResource *, QPair<VkAccessFlags, bool> > writtenResources; - void reset() { - writtenResources.clear(); - } - } computePassState; - - struct Command { - enum Cmd { - CopyBuffer, - CopyBufferToImage, - CopyImage, - CopyImageToBuffer, - ImageBarrier, - BufferBarrier, - BlitImage, - BeginRenderPass, - EndRenderPass, - BindPipeline, - BindDescriptorSet, - BindVertexBuffer, - BindIndexBuffer, - SetViewport, - SetScissor, - SetBlendConstants, - SetStencilRef, - Draw, - DrawIndexed, - DebugMarkerBegin, - DebugMarkerEnd, - DebugMarkerInsert, - TransitionPassResources, - Dispatch, - ExecuteSecondary - }; - Cmd cmd; - - union Args { - struct { - VkBuffer src; - VkBuffer dst; - VkBufferCopy desc; - } copyBuffer; - struct { - VkBuffer src; - VkImage dst; - VkImageLayout dstLayout; - int count; - int bufferImageCopyIndex; - } copyBufferToImage; - struct { - VkImage src; - VkImageLayout srcLayout; - VkImage dst; - VkImageLayout dstLayout; - VkImageCopy desc; - } copyImage; - struct { - VkImage src; - VkImageLayout srcLayout; - VkBuffer dst; - VkBufferImageCopy desc; - } copyImageToBuffer; - struct { - VkPipelineStageFlags srcStageMask; - VkPipelineStageFlags dstStageMask; - int count; - int index; - } imageBarrier; - struct { - VkPipelineStageFlags srcStageMask; - VkPipelineStageFlags dstStageMask; - int count; - int index; - } bufferBarrier; - struct { - VkImage src; - VkImageLayout srcLayout; - VkImage dst; - VkImageLayout dstLayout; - VkFilter filter; - VkImageBlit desc; - } blitImage; - struct { - VkRenderPassBeginInfo desc; - int clearValueIndex; - bool useSecondaryCb; - } beginRenderPass; - struct { - } endRenderPass; - struct { - VkPipelineBindPoint bindPoint; - VkPipeline pipeline; - } bindPipeline; - struct { - VkPipelineBindPoint bindPoint; - VkPipelineLayout pipelineLayout; - VkDescriptorSet descSet; - int dynamicOffsetCount; - int dynamicOffsetIndex; - } bindDescriptorSet; - struct { - int startBinding; - int count; - int vertexBufferIndex; - int vertexBufferOffsetIndex; - } bindVertexBuffer; - struct { - VkBuffer buf; - VkDeviceSize ofs; - VkIndexType type; - } bindIndexBuffer; - struct { - VkViewport viewport; - } setViewport; - struct { - VkRect2D scissor; - } setScissor; - struct { - float c[4]; - } setBlendConstants; - struct { - uint32_t ref; - } setStencilRef; - struct { - uint32_t vertexCount; - uint32_t instanceCount; - uint32_t firstVertex; - uint32_t firstInstance; - } draw; - struct { - uint32_t indexCount; - uint32_t instanceCount; - uint32_t firstIndex; - int32_t vertexOffset; - uint32_t firstInstance; - } drawIndexed; - struct { -#ifdef VK_EXT_debug_utils - VkDebugUtilsLabelEXT label; - int labelNameIndex; -#endif - } debugMarkerBegin; - struct { - } debugMarkerEnd; - struct { -#ifdef VK_EXT_debug_utils - VkDebugUtilsLabelEXT label; - int labelNameIndex; -#endif - } debugMarkerInsert; - struct { - int trackerIndex; - } transitionResources; - struct { - int x, y, z; - } dispatch; - struct { - VkCommandBuffer cb; - } executeSecondary; - } args; - }; - - QRhiBackendCommandList<Command> commands; - QVarLengthArray<QRhiPassResourceTracker, 8> passResTrackers; - int currentPassResTrackerIndex; - - void resetCommands() { - commands.reset(); - resetPools(); - - passResTrackers.clear(); - currentPassResTrackerIndex = -1; - } - - void resetPools() { - pools.clearValue.clear(); - pools.bufferImageCopy.clear(); - pools.dynamicOffset.clear(); - pools.vertexBuffer.clear(); - pools.vertexBufferOffset.clear(); - pools.debugMarkerData.clear(); - pools.imageBarrier.clear(); - pools.bufferBarrier.clear(); - } - - struct { - QVarLengthArray<VkClearValue, 4> clearValue; - QVarLengthArray<VkBufferImageCopy, 16> bufferImageCopy; - QVarLengthArray<uint32_t, 4> dynamicOffset; - QVarLengthArray<VkBuffer, 4> vertexBuffer; - QVarLengthArray<VkDeviceSize, 4> vertexBufferOffset; - QVarLengthArray<QByteArray, 4> debugMarkerData; - QVarLengthArray<VkImageMemoryBarrier, 8> imageBarrier; - QVarLengthArray<VkBufferMemoryBarrier, 8> bufferBarrier; - } pools; - - friend class QRhiVulkan; -}; - -struct QVkSwapChain : public QRhiSwapChain -{ - QVkSwapChain(QRhiImplementation *rhi); - ~QVkSwapChain(); - void destroy() override; - - QRhiCommandBuffer *currentFrameCommandBuffer() override; - QRhiRenderTarget *currentFrameRenderTarget() override; - - QSize surfacePixelSize() override; - bool isFormatSupported(Format f) override; - - QRhiRenderPassDescriptor *newCompatibleRenderPassDescriptor() override; - bool createOrResize() override; - - bool ensureSurface(); - - static const quint32 EXPECTED_MAX_BUFFER_COUNT = 4; - - QWindow *window = nullptr; - QSize pixelSize; - bool supportsReadback = false; - VkSwapchainKHR sc = VK_NULL_HANDLE; - int bufferCount = 0; - VkSurfaceKHR surface = VK_NULL_HANDLE; - VkSurfaceKHR lastConnectedSurface = VK_NULL_HANDLE; - VkFormat colorFormat = VK_FORMAT_B8G8R8A8_UNORM; - VkColorSpaceKHR colorSpace = VK_COLOR_SPACE_SRGB_NONLINEAR_KHR; - QVkRenderBuffer *ds = nullptr; - VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT; - QVarLengthArray<VkPresentModeKHR, 8> supportedPresentationModes; - VkDeviceMemory msaaImageMem = VK_NULL_HANDLE; - QVkSwapChainRenderTarget rtWrapper; - QVkCommandBuffer cbWrapper; - - struct ImageResources { - VkImage image = VK_NULL_HANDLE; - VkImageView imageView = VK_NULL_HANDLE; - VkFramebuffer fb = VK_NULL_HANDLE; - VkImage msaaImage = VK_NULL_HANDLE; - VkImageView msaaImageView = VK_NULL_HANDLE; - enum LastUse { - ScImageUseNone, - ScImageUseRender, - ScImageUseTransferSource - }; - LastUse lastUse = ScImageUseNone; - }; - QVarLengthArray<ImageResources, EXPECTED_MAX_BUFFER_COUNT> imageRes; - - struct FrameResources { - VkFence imageFence = VK_NULL_HANDLE; - bool imageFenceWaitable = false; - VkSemaphore imageSem = VK_NULL_HANDLE; - VkSemaphore drawSem = VK_NULL_HANDLE; - bool imageAcquired = false; - bool imageSemWaitable = false; - VkFence cmdFence = VK_NULL_HANDLE; - bool cmdFenceWaitable = false; - VkCommandBuffer cmdBuf = VK_NULL_HANDLE; // primary - int timestampQueryIndex = -1; - } frameRes[QVK_FRAMES_IN_FLIGHT]; - - quint32 currentImageIndex = 0; // index in imageRes - quint32 currentFrameSlot = 0; // index in frameRes - int frameCount = 0; - - friend class QRhiVulkan; -}; - -class QRhiVulkan : public QRhiImplementation -{ -public: - QRhiVulkan(QRhiVulkanInitParams *params, QRhiVulkanNativeHandles *importParams = nullptr); - - bool create(QRhi::Flags flags) override; - void destroy() override; - - QRhiGraphicsPipeline *createGraphicsPipeline() override; - QRhiComputePipeline *createComputePipeline() override; - QRhiShaderResourceBindings *createShaderResourceBindings() override; - QRhiBuffer *createBuffer(QRhiBuffer::Type type, - QRhiBuffer::UsageFlags usage, - quint32 size) override; - QRhiRenderBuffer *createRenderBuffer(QRhiRenderBuffer::Type type, - const QSize &pixelSize, - int sampleCount, - QRhiRenderBuffer::Flags flags, - QRhiTexture::Format backingFormatHint) override; - QRhiTexture *createTexture(QRhiTexture::Format format, - const QSize &pixelSize, - int depth, - int arraySize, - int sampleCount, - QRhiTexture::Flags flags) override; - QRhiSampler *createSampler(QRhiSampler::Filter magFilter, - QRhiSampler::Filter minFilter, - QRhiSampler::Filter mipmapMode, - QRhiSampler:: AddressMode u, - QRhiSampler::AddressMode v, - QRhiSampler::AddressMode w) override; - - QRhiTextureRenderTarget *createTextureRenderTarget(const QRhiTextureRenderTargetDescription &desc, - QRhiTextureRenderTarget::Flags flags) override; - - QRhiSwapChain *createSwapChain() override; - QRhi::FrameOpResult beginFrame(QRhiSwapChain *swapChain, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endFrame(QRhiSwapChain *swapChain, QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult beginOffscreenFrame(QRhiCommandBuffer **cb, QRhi::BeginFrameFlags flags) override; - QRhi::FrameOpResult endOffscreenFrame(QRhi::EndFrameFlags flags) override; - QRhi::FrameOpResult finish() override; - - void resourceUpdate(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void beginPass(QRhiCommandBuffer *cb, - QRhiRenderTarget *rt, - const QColor &colorClearValue, - const QRhiDepthStencilClearValue &depthStencilClearValue, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endPass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - - void setGraphicsPipeline(QRhiCommandBuffer *cb, - QRhiGraphicsPipeline *ps) override; - - void setShaderResources(QRhiCommandBuffer *cb, - QRhiShaderResourceBindings *srb, - int dynamicOffsetCount, - const QRhiCommandBuffer::DynamicOffset *dynamicOffsets) override; - - void setVertexInput(QRhiCommandBuffer *cb, - int startBinding, int bindingCount, const QRhiCommandBuffer::VertexInput *bindings, - QRhiBuffer *indexBuf, quint32 indexOffset, - QRhiCommandBuffer::IndexFormat indexFormat) override; - - void setViewport(QRhiCommandBuffer *cb, const QRhiViewport &viewport) override; - void setScissor(QRhiCommandBuffer *cb, const QRhiScissor &scissor) override; - void setBlendConstants(QRhiCommandBuffer *cb, const QColor &c) override; - void setStencilRef(QRhiCommandBuffer *cb, quint32 refValue) override; - - void draw(QRhiCommandBuffer *cb, quint32 vertexCount, - quint32 instanceCount, quint32 firstVertex, quint32 firstInstance) override; - - void drawIndexed(QRhiCommandBuffer *cb, quint32 indexCount, - quint32 instanceCount, quint32 firstIndex, - qint32 vertexOffset, quint32 firstInstance) override; - - void debugMarkBegin(QRhiCommandBuffer *cb, const QByteArray &name) override; - void debugMarkEnd(QRhiCommandBuffer *cb) override; - void debugMarkMsg(QRhiCommandBuffer *cb, const QByteArray &msg) override; - - void beginComputePass(QRhiCommandBuffer *cb, - QRhiResourceUpdateBatch *resourceUpdates, - QRhiCommandBuffer::BeginPassFlags flags) override; - void endComputePass(QRhiCommandBuffer *cb, QRhiResourceUpdateBatch *resourceUpdates) override; - void setComputePipeline(QRhiCommandBuffer *cb, QRhiComputePipeline *ps) override; - void dispatch(QRhiCommandBuffer *cb, int x, int y, int z) override; - - const QRhiNativeHandles *nativeHandles(QRhiCommandBuffer *cb) override; - void beginExternal(QRhiCommandBuffer *cb) override; - void endExternal(QRhiCommandBuffer *cb) override; - - QList<int> supportedSampleCounts() const override; - int ubufAlignment() const override; - bool isYUpInFramebuffer() const override; - bool isYUpInNDC() const override; - bool isClipDepthZeroToOne() const override; - QMatrix4x4 clipSpaceCorrMatrix() const override; - bool isTextureFormatSupported(QRhiTexture::Format format, QRhiTexture::Flags flags) const override; - bool isFeatureSupported(QRhi::Feature feature) const override; - int resourceLimit(QRhi::ResourceLimit limit) const override; - const QRhiNativeHandles *nativeHandles() override; - QRhiDriverInfo driverInfo() const override; - QRhiStats statistics() override; - bool makeThreadLocalNativeContextCurrent() override; - void releaseCachedResources() override; - bool isDeviceLost() const override; - - QByteArray pipelineCacheData() override; - void setPipelineCacheData(const QByteArray &data) override; - - VkResult createDescriptorPool(VkDescriptorPool *pool); - bool allocateDescriptorSet(VkDescriptorSetAllocateInfo *allocInfo, VkDescriptorSet *result, int *resultPoolIndex); - uint32_t chooseTransientImageMemType(VkImage img, uint32_t startIndex); - bool createTransientImage(VkFormat format, const QSize &pixelSize, VkImageUsageFlags usage, - VkImageAspectFlags aspectMask, VkSampleCountFlagBits samples, - VkDeviceMemory *mem, VkImage *images, VkImageView *views, int count); - - bool recreateSwapChain(QRhiSwapChain *swapChain); - void releaseSwapChainResources(QRhiSwapChain *swapChain); - - VkFormat optimalDepthStencilFormat(); - VkSampleCountFlagBits effectiveSampleCount(int sampleCount); - bool createDefaultRenderPass(QVkRenderPassDescriptor *rpD, - bool hasDepthStencil, - VkSampleCountFlagBits samples, - VkFormat colorFormat); - bool createOffscreenRenderPass(QVkRenderPassDescriptor *rpD, - const QRhiColorAttachment *firstColorAttachment, - const QRhiColorAttachment *lastColorAttachment, - bool preserveColor, - bool preserveDs, - QRhiRenderBuffer *depthStencilBuffer, - QRhiTexture *depthTexture); - bool ensurePipelineCache(const void *initialData = nullptr, size_t initialDataSize = 0); - VkShaderModule createShader(const QByteArray &spirv); - - void prepareNewFrame(QRhiCommandBuffer *cb); - VkCommandBuffer startSecondaryCommandBuffer(QVkRenderTargetData *rtD = nullptr); - void endAndEnqueueSecondaryCommandBuffer(VkCommandBuffer cb, QVkCommandBuffer *cbD); - QRhi::FrameOpResult startPrimaryCommandBuffer(VkCommandBuffer *cb); - QRhi::FrameOpResult endAndSubmitPrimaryCommandBuffer(VkCommandBuffer cb, VkFence cmdFence, - VkSemaphore *waitSem, VkSemaphore *signalSem); - void waitCommandCompletion(int frameSlot); - VkDeviceSize subresUploadByteSize(const QRhiTextureSubresourceUploadDescription &subresDesc) const; - using BufferImageCopyList = QVarLengthArray<VkBufferImageCopy, 16>; - void prepareUploadSubres(QVkTexture *texD, int layer, int level, - const QRhiTextureSubresourceUploadDescription &subresDesc, - size_t *curOfs, void *mp, - BufferImageCopyList *copyInfos); - void enqueueResourceUpdates(QVkCommandBuffer *cbD, QRhiResourceUpdateBatch *resourceUpdates); - void executeBufferHostWritesForSlot(QVkBuffer *bufD, int slot); - void enqueueTransitionPassResources(QVkCommandBuffer *cbD); - void recordPrimaryCommandBuffer(QVkCommandBuffer *cbD); - void trackedRegisterBuffer(QRhiPassResourceTracker *passResTracker, - QVkBuffer *bufD, - int slot, - QRhiPassResourceTracker::BufferAccess access, - QRhiPassResourceTracker::BufferStage stage); - void trackedRegisterTexture(QRhiPassResourceTracker *passResTracker, - QVkTexture *texD, - QRhiPassResourceTracker::TextureAccess access, - QRhiPassResourceTracker::TextureStage stage); - void recordTransitionPassResources(QVkCommandBuffer *cbD, const QRhiPassResourceTracker &tracker); - void activateTextureRenderTarget(QVkCommandBuffer *cbD, QVkTextureRenderTarget *rtD); - void executeDeferredReleases(bool forced = false); - void finishActiveReadbacks(bool forced = false); - - void setObjectName(uint64_t object, VkObjectType type, const QByteArray &name, int slot = -1); - void trackedBufferBarrier(QVkCommandBuffer *cbD, QVkBuffer *bufD, int slot, - VkAccessFlags access, VkPipelineStageFlags stage); - void trackedImageBarrier(QVkCommandBuffer *cbD, QVkTexture *texD, - VkImageLayout layout, VkAccessFlags access, VkPipelineStageFlags stage); - void depthStencilExplicitBarrier(QVkCommandBuffer *cbD, QVkRenderBuffer *rbD); - void subresourceBarrier(QVkCommandBuffer *cbD, VkImage image, - VkImageLayout oldLayout, VkImageLayout newLayout, - VkAccessFlags srcAccess, VkAccessFlags dstAccess, - VkPipelineStageFlags srcStage, VkPipelineStageFlags dstStage, - int startLayer, int layerCount, - int startLevel, int levelCount); - void updateShaderResourceBindings(QRhiShaderResourceBindings *srb, int descSetIdx = -1); - void ensureCommandPoolForNewFrame(); - - QVulkanInstance *inst = nullptr; - QWindow *maybeWindow = nullptr; - QByteArrayList requestedDeviceExtensions; - bool importedDevice = false; - VkPhysicalDevice physDev = VK_NULL_HANDLE; - VkDevice dev = VK_NULL_HANDLE; - VkCommandPool cmdPool[QVK_FRAMES_IN_FLIGHT] = {}; - quint32 gfxQueueFamilyIdx = 0; - quint32 gfxQueueIdx = 0; - VkQueue gfxQueue = VK_NULL_HANDLE; - quint32 timestampValidBits = 0; - bool importedAllocator = false; - QVkAllocator allocator = nullptr; - QVulkanFunctions *f = nullptr; - QVulkanDeviceFunctions *df = nullptr; - QRhi::Flags rhiFlags; - VkPhysicalDeviceFeatures physDevFeatures; - VkPhysicalDeviceProperties physDevProperties; - VkDeviceSize ubufAlign; - VkDeviceSize texbufAlign; - bool deviceLost = false; - bool releaseCachedResourcesCalledBeforeFrameStart = false; - -#ifdef VK_EXT_debug_utils - PFN_vkSetDebugUtilsObjectNameEXT vkSetDebugUtilsObjectNameEXT = nullptr; - PFN_vkCmdBeginDebugUtilsLabelEXT vkCmdBeginDebugUtilsLabelEXT = nullptr; - PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT = nullptr; - PFN_vkCmdInsertDebugUtilsLabelEXT vkCmdInsertDebugUtilsLabelEXT = nullptr; -#endif - - PFN_vkCreateSwapchainKHR vkCreateSwapchainKHR = nullptr; - PFN_vkDestroySwapchainKHR vkDestroySwapchainKHR; - PFN_vkGetSwapchainImagesKHR vkGetSwapchainImagesKHR; - PFN_vkAcquireNextImageKHR vkAcquireNextImageKHR; - PFN_vkQueuePresentKHR vkQueuePresentKHR; - PFN_vkGetPhysicalDeviceSurfaceCapabilitiesKHR vkGetPhysicalDeviceSurfaceCapabilitiesKHR; - PFN_vkGetPhysicalDeviceSurfaceFormatsKHR vkGetPhysicalDeviceSurfaceFormatsKHR; - PFN_vkGetPhysicalDeviceSurfacePresentModesKHR vkGetPhysicalDeviceSurfacePresentModesKHR; - - struct { - bool compute = false; - bool wideLines = false; - bool debugUtils = false; - bool vertexAttribDivisor = false; - bool texture3DSliceAs2D = false; - bool tessellation = false; - bool geometryShader = false; - bool nonFillPolygonMode = false; - QVersionNumber apiVersion; - } caps; - - VkPipelineCache pipelineCache = VK_NULL_HANDLE; - struct DescriptorPoolData { - DescriptorPoolData() { } - DescriptorPoolData(VkDescriptorPool pool_) - : pool(pool_) - { } - VkDescriptorPool pool = VK_NULL_HANDLE; - int refCount = 0; - int allocedDescSets = 0; - }; - QVarLengthArray<DescriptorPoolData, 8> descriptorPools; - QVarLengthArray<VkCommandBuffer, 4> freeSecondaryCbs[QVK_FRAMES_IN_FLIGHT]; - - VkQueryPool timestampQueryPool = VK_NULL_HANDLE; - QBitArray timestampQueryPoolMap; - - VkFormat optimalDsFormat = VK_FORMAT_UNDEFINED; - QMatrix4x4 clipCorrectMatrix; - - QVkSwapChain *currentSwapChain = nullptr; - QSet<QVkSwapChain *> swapchains; - QRhiVulkanNativeHandles nativeHandlesStruct; - QRhiDriverInfo driverInfoStruct; - - struct OffscreenFrame { - OffscreenFrame(QRhiImplementation *rhi) - { - for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) - cbWrapper[i] = new QVkCommandBuffer(rhi); - } - ~OffscreenFrame() - { - for (int i = 0; i < QVK_FRAMES_IN_FLIGHT; ++i) - delete cbWrapper[i]; - } - bool active = false; - QVkCommandBuffer *cbWrapper[QVK_FRAMES_IN_FLIGHT]; - VkFence cmdFence = VK_NULL_HANDLE; - } ofr; - - struct TextureReadback { - int activeFrameSlot = -1; - QRhiReadbackDescription desc; - QRhiReadbackResult *result; - VkBuffer stagingBuf; - QVkAlloc stagingAlloc; - quint32 byteSize; - QSize pixelSize; - QRhiTexture::Format format; - }; - QVarLengthArray<TextureReadback, 2> activeTextureReadbacks; - struct BufferReadback { - int activeFrameSlot = -1; - QRhiBufferReadbackResult *result; - quint32 byteSize; - VkBuffer stagingBuf; - QVkAlloc stagingAlloc; - }; - QVarLengthArray<BufferReadback, 2> activeBufferReadbacks; - - struct DeferredReleaseEntry { - enum Type { - Pipeline, - ShaderResourceBindings, - Buffer, - RenderBuffer, - Texture, - Sampler, - TextureRenderTarget, - RenderPass, - StagingBuffer, - SecondaryCommandBuffer - }; - Type type; - int lastActiveFrameSlot; // -1 if not used otherwise 0..FRAMES_IN_FLIGHT-1 - union { - struct { - VkPipeline pipeline; - VkPipelineLayout layout; - } pipelineState; - struct { - int poolIndex; - VkDescriptorSetLayout layout; - } shaderResourceBindings; - struct { - VkBuffer buffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc allocations[QVK_FRAMES_IN_FLIGHT]; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - } buffer; - struct { - VkDeviceMemory memory; - VkImage image; - VkImageView imageView; - } renderBuffer; - struct { - VkImage image; - VkImageView imageView; - QVkAlloc allocation; - VkBuffer stagingBuffers[QVK_FRAMES_IN_FLIGHT]; - QVkAlloc stagingAllocations[QVK_FRAMES_IN_FLIGHT]; - VkImageView extraImageViews[QRhi::MAX_MIP_LEVELS]; - } texture; - struct { - VkSampler sampler; - } sampler; - struct { - VkFramebuffer fb; - VkImageView rtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - VkImageView resrtv[QVkRenderTargetData::MAX_COLOR_ATTACHMENTS]; - } textureRenderTarget; - struct { - VkRenderPass rp; - } renderPass; - struct { - VkBuffer stagingBuffer; - QVkAlloc stagingAllocation; - } stagingBuffer; - struct { - VkCommandBuffer cb; - } secondaryCommandBuffer; - }; - }; - QList<DeferredReleaseEntry> releaseQueue; -}; - -Q_DECLARE_TYPEINFO(QRhiVulkan::DescriptorPoolData, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::DeferredReleaseEntry, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::TextureReadback, Q_RELOCATABLE_TYPE); -Q_DECLARE_TYPEINFO(QRhiVulkan::BufferReadback, Q_RELOCATABLE_TYPE); - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qrhivulkanext_p.h b/src/gui/rhi/qrhivulkanext_p.h deleted file mode 100644 index 02b346948b..0000000000 --- a/src/gui/rhi/qrhivulkanext_p.h +++ /dev/null @@ -1,48 +0,0 @@ -// Copyright (C) 2018 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QRHIVULKANEXT_P_H -#define QRHIVULKANEXT_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists purely as an -// implementation detail. This header file may change from version to -// version without notice, or even be removed. -// -// We mean it. -// - -#include "qrhivulkan_p.h" - -QT_BEGIN_NAMESPACE - -#ifndef VK_EXT_vertex_attribute_divisor -#define VK_EXT_vertex_attribute_divisor 1 -#define VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_SPEC_VERSION 2 -#define VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME "VK_EXT_vertex_attribute_divisor" - -typedef struct VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT { - VkStructureType sType; - void* pNext; - uint32_t maxVertexAttribDivisor; -} VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT; - -typedef struct VkVertexInputBindingDivisorDescriptionEXT { - uint32_t binding; - uint32_t divisor; -} VkVertexInputBindingDivisorDescriptionEXT; - -typedef struct VkPipelineVertexInputDivisorStateCreateInfoEXT { - VkStructureType sType; - const void* pNext; - uint32_t vertexBindingDivisorCount; - const VkVertexInputBindingDivisorDescriptionEXT* pVertexBindingDivisors; -} VkPipelineVertexInputDivisorStateCreateInfoEXT; -#endif // VK_EXT_vertex_attribute_divisor - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qshader.cpp b/src/gui/rhi/qshader.cpp index 52cb503521..d5fb53e7e6 100644 --- a/src/gui/rhi/qshader.cpp +++ b/src/gui/rhi/qshader.cpp @@ -1,7 +1,7 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qshader_p_p.h" +#include "qshader_p.h" #include <QDataStream> #include <QBuffer> @@ -9,8 +9,9 @@ QT_BEGIN_NAMESPACE /*! \class QShader - \internal + \ingroup painting-3D \inmodule QtGui + \since 6.6 \brief Contains multiple versions of a shader translated to multiple shading languages, together with reflection metadata. @@ -21,6 +22,16 @@ QT_BEGIN_NAMESPACE as, Vulkan, Metal, Direct3D, and OpenGL, take QShader as their input whenever a shader needs to be specified. + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qshader.h>}. + A QShader instance is empty and thus invalid by default. To get a useful instance, the two typical methods are: @@ -68,8 +79,9 @@ QT_BEGIN_NAMESPACE can be returned or passed by value. Detach happens implicitly when calling a setter. - For reference, QRhi expects that a QShader suitable for all its - backends contains at least the following: + For reference, a typical, portable QRhi expects that a QShader suitable for + all its backends contains at least the following. (this excludes support + for core profile OpenGL contexts, add GLSL 150 or newer for that) \list @@ -77,11 +89,11 @@ QT_BEGIN_NAMESPACE \li GLSL/ES 100 source code suitable for OpenGL ES 2.0 or newer - \li GLSL 120 source code suitable for OpenGL 2.1 + \li GLSL 120 source code suitable for OpenGL 2.1 or newer - \li HLSL Shader Model 5.0 source code or the corresponding DXBC bytecode suitable for Direct3D 11 + \li HLSL Shader Model 5.0 source code or the corresponding DXBC bytecode suitable for Direct3D 11/12 - \li Metal Shading Language 1.2 source code or the corresponding bytecode suitable for Metal + \li Metal Shading Language 1.2 source code or the corresponding bytecode suitable for Metal 1.2 or newer \endlist @@ -102,8 +114,8 @@ QT_BEGIN_NAMESPACE /*! \class QShaderVersion - \internal \inmodule QtGui + \since 6.6 \brief Specifies the shading language version. @@ -128,6 +140,9 @@ QT_BEGIN_NAMESPACE A default constructed QShaderVersion contains a version of 100 and no flags set. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! @@ -140,13 +155,16 @@ QT_BEGIN_NAMESPACE /*! \class QShaderKey - \internal \inmodule QtGui + \since 6.6 \brief Specifies the shading language, the version with flags, and the variant. A default constructed QShaderKey has source set to SpirvShader and sourceVersion set to 100. sourceVariant defaults to StandardShader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! @@ -196,14 +214,39 @@ QT_BEGIN_NAMESPACE */ /*! + \enum QShader::SerializedFormatVersion + Describes the desired output format when serializing the QShader. + + The default value for the \c version argument of serialized() is \c Latest. + This is sufficient in the vast majority of cases. Specifying another value + is needed only when the intention is to generate serialized data that can + be loaded by earlier Qt versions. For example, the \c qsb tool uses these + enum values when the \c{--qsbversion} command-line argument is given. + + \note Targeting earlier versions will make certain features disfunctional + with the generated asset. This is not an issue when using the asset with + the specified, older Qt version, given that that Qt version does not have + the newer features in newer Qt versions that rely on additional data + generated in the QShader and the serialized data stream, but may become a + problem if the generated asset is then used with a newer Qt version. + + \value Latest The current Qt version + \value Qt_6_5 Qt 6.5 + \value Qt_6_4 Qt 6.4 + */ + +/*! \class QShaderCode - \internal \inmodule QtGui + \since 6.6 \brief Contains source or binary code for a shader and additional metadata. When shader() is empty after retrieving a QShaderCode instance from QShader, it indicates no shader code was found for the requested key. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! @@ -226,7 +269,7 @@ void QShader::detach() } /*! - \internal + Constructs a copy of \a other. */ QShader::QShader(const QShader &other) : d(other.d) @@ -236,7 +279,7 @@ QShader::QShader(const QShader &other) } /*! - \internal + Assigns \a other to this object. */ QShader &QShader::operator=(const QShader &other) { @@ -256,6 +299,28 @@ QShader &QShader::operator=(const QShader &other) } /*! + \fn QShader::QShader(QShader &&other) noexcept + \since 6.7 + + Move-constructs a new QShader from \a other. + + \note The moved-from object \a other is placed in a + partially-formed state, in which the only valid operations are + destruction and assignment of a new value. +*/ + +/*! + \fn QShader &QShader::operator=(QShader &&other) + \since 6.7 + + Move-assigns \a other to this QShader instance. + + \note The moved-from object \a other is placed in a + partially-formed state, in which the only valid operations are + destruction and assignment of a new value. +*/ + +/*! Destructor. */ QShader::~QShader() @@ -265,6 +330,14 @@ QShader::~QShader() } /*! + \fn void QShader::swap(QShader &other) + \since 6.7 + + Swaps shader \a other with this shader. This operation is very fast and + never fails. +*/ + +/*! \return true if the QShader contains at least one shader version. */ bool QShader::isValid() const @@ -366,7 +439,11 @@ static void writeShaderKey(QDataStream *ds, const QShaderKey &k) QShader, suitable for writing to files or other I/O devices. By default the latest serialization format is used. Use \a version - parameter to serialize for a compatibility Qt version. + parameter to serialize for a compatibility Qt version. Only when it is + known that the generated data stream must be made compatible with an older + Qt version at the expense of making it incompatible with features + introduced since that Qt version, should another value (for example, + \l{SerializedFormatVersion}{Qt_6_5} for Qt 6.5) be used. \sa fromSerialized() */ @@ -455,6 +532,9 @@ static void readShaderKey(QDataStream *ds, QShaderKey *k) /*! Creates a new QShader instance from the given \a data. + If \a data cannot be deserialized successfully, the result is a default + constructed QShader for which isValid() returns \c false. + \sa serialized() */ QShader QShader::fromSerialized(const QByteArray &data) @@ -577,16 +657,79 @@ QShader QShader::fromSerialized(const QByteArray &data) return bs; } +/*! + \fn QShaderVersion::QShaderVersion() = default + */ + +/*! + Constructs a new QShaderVersion with version \a v and flags \a f. + */ QShaderVersion::QShaderVersion(int v, Flags f) : m_version(v), m_flags(f) { } +/*! + \fn int QShaderVersion::version() const + \return the version. + */ + +/*! + \fn void QShaderVersion::setVersion(int v) + Sets the shading language version to \a v. + */ + +/*! + \fn QShaderVersion::Flags QShaderVersion::flags() const + \return the flags. + */ + +/*! + \fn void QShaderVersion::setFlags(Flags f) + Sets the flags \a f. + */ + +/*! + \fn QShaderCode::QShaderCode() = default + */ + +/*! + Constructs a new QShaderCode with the specified shader source \a code and + \a entry point name. + */ QShaderCode::QShaderCode(const QByteArray &code, const QByteArray &entry) : m_shader(code), m_entryPoint(entry) { } +/*! + \fn QByteArray QShaderCode::shader() const + \return the shader source or bytecode. + */ + +/*! + \fn void QShaderCode::setShader(const QByteArray &code) + Sets the shader source or byte \a code. + */ + +/*! + \fn QByteArray QShaderCode::entryPoint() const + \return the entry point name. + */ + +/*! + \fn void QShaderCode::setEntryPoint(const QByteArray &entry) + Sets the \a entry point name. + */ + +/*! + \fn QShaderKey::QShaderKey() = default + */ + +/*! + Constructs a new QShaderKey with shader type \a s, version \a sver, and + variant \a svar. + */ QShaderKey::QShaderKey(QShader::Source s, const QShaderVersion &sver, QShader::Variant svar) @@ -597,6 +740,36 @@ QShaderKey::QShaderKey(QShader::Source s, } /*! + \fn QShader::Source QShaderKey::source() const + \return the shader type. + */ + +/*! + \fn void QShaderKey::setSource(QShader::Source s) + Sets the shader type \a s. + */ + +/*! + \fn QShaderVersion QShaderKey::sourceVersion() const + \return the shading language version. + */ + +/*! + \fn void QShaderKey::setSourceVersion(const QShaderVersion &sver) + Sets the shading language version \a sver. + */ + +/*! + \fn QShader::Variant QShaderKey::sourceVariant() const + \return the type of the variant to use. + */ + +/*! + \fn void QShaderKey::setSourceVariant(QShader::Variant svar) + Sets the type of variant to use to \a svar. + */ + +/*! Returns \c true if the two QShader objects \a lhs and \a rhs are equal, meaning they are for the same stage with matching sets of shader source or binary code. @@ -614,10 +787,9 @@ bool operator==(const QShader &lhs, const QShader &rhs) noexcept } /*! - \internal \fn bool operator!=(const QShader &lhs, const QShader &rhs) - Returns \c false if the values in the two QShader objects \a a and \a b + Returns \c false if the values in the two QShader objects \a lhs and \a rhs are equal; otherwise returns \c true. \relates QShader @@ -635,7 +807,7 @@ size_t qHash(const QShader &s, size_t seed) noexcept seed = hash(seed, s.stage()); if (!s.d->shaders.isEmpty()) { seed = hash(seed, s.d->shaders.firstKey()); - seed = hash(seed, s.d->shaders.first()); + seed = hash(seed, std::as_const(s.d->shaders).first()); } } return seed; @@ -660,6 +832,8 @@ size_t qHash(const QShaderVersion &s, size_t seed) noexcept #endif /*! + \return true if \a lhs is smaller than \a rhs. + Establishes a sorting order between the two QShaderVersion \a lhs and \a rhs. \relates QShaderVersion @@ -676,11 +850,10 @@ bool operator<(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept } /*! - \internal \fn bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) - Returns \c false if the values in the two QShaderVersion objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderVersion objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderVersion */ @@ -697,6 +870,8 @@ bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept } /*! + \return true if \a lhs is smaller than \a rhs. + Establishes a sorting order between the two keys \a lhs and \a rhs. \relates QShaderKey @@ -719,11 +894,10 @@ bool operator<(const QShaderKey &lhs, const QShaderKey &rhs) noexcept } /*! - \internal \fn bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) - Returns \c false if the values in the two QShaderKey objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderKey objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderKey */ @@ -753,11 +927,10 @@ bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept } /*! - \internal \fn bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) - Returns \c false if the values in the two QShaderCode objects \a a - and \a b are equal; otherwise returns \c true. + Returns \c false if the values in the two QShaderCode objects \a lhs + and \a rhs are equal; otherwise returns \c true. \relates QShaderCode */ @@ -895,6 +1068,8 @@ void QShader::removeResourceBindingMap(const QShaderKey &key) /*! \struct QShader::SeparateToCombinedImageSamplerMapping + \inmodule QtGui + \brief Mapping metadata for sampler uniforms. Describes a mapping from a traditional combined image sampler uniform to binding points for a separate texture and sampler. @@ -904,9 +1079,24 @@ void QShader::removeResourceBindingMap(const QShaderKey &key) contains a \c sampler2D (or sampler3D, etc.) uniform with the name of \c{_54} which corresponds to two separate resource bindings (\c 1 and \c 2) in the original shader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! + \variable QShader::SeparateToCombinedImageSamplerMapping::combinedSamplerName +*/ + +/*! + \variable QShader::SeparateToCombinedImageSamplerMapping::textureBinding +*/ + +/*! + \variable QShader::SeparateToCombinedImageSamplerMapping::samplerBinding +*/ + +/*! \return the combined image sampler mapping list for \a key, or an empty list if there is no data available for \a key, for example because such a mapping is not applicable for the shading language. @@ -953,6 +1143,8 @@ void QShader::removeSeparateToCombinedImageSamplerMappingList(const QShaderKey & /*! \struct QShader::NativeShaderInfo + \inmodule QtGui + \brief Additional metadata about the native shader code. Describes information about the native shader code, if applicable. This becomes relevant with certain shader languages for certain shader stages, @@ -969,9 +1161,20 @@ void QShader::removeSeparateToCombinedImageSamplerMappingList(const QShaderKey & not be present at all if per-patch output variables were not used. The fact that the shader code relies on such a buffer present can be indicated by the data in this struct. + + \note This is a RHI API with limited compatibility guarantees, see \l QShader + for details. */ /*! + \variable QShader::NativeShaderInfo::flags +*/ + +/*! + \variable QShader::NativeShaderInfo::extraBufferBindings +*/ + +/*! \return the native shader info struct for \a key, or an empty object if there is no data available for \a key, for example because such a mapping is not applicable for the shading language or the shader stage. diff --git a/src/gui/rhi/qshader.h b/src/gui/rhi/qshader.h new file mode 100644 index 0000000000..2465081366 --- /dev/null +++ b/src/gui/rhi/qshader.h @@ -0,0 +1,241 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QSHADER_H +#define QSHADER_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qhash.h> +#include <QtCore/qmap.h> +#include <rhi/qshaderdescription.h> + +QT_BEGIN_NAMESPACE + +struct QShaderPrivate; +class QShaderKey; + +#ifdef Q_OS_INTEGRITY + class QShaderVersion; + size_t qHash(const QShaderVersion &, size_t = 0) noexcept; +#endif + +class Q_GUI_EXPORT QShaderVersion +{ +public: + enum Flag { + GlslEs = 0x01 + }; + Q_DECLARE_FLAGS(Flags, Flag) + + QShaderVersion() = default; + QShaderVersion(int v, Flags f = Flags()); + + int version() const { return m_version; } + void setVersion(int v) { m_version = v; } + + Flags flags() const { return m_flags; } + void setFlags(Flags f) { m_flags = f; } + +private: + int m_version = 100; + Flags m_flags; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderVersion::Flags) +Q_DECLARE_TYPEINFO(QShaderVersion, Q_RELOCATABLE_TYPE); + +class QShaderCode; +Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t = 0) noexcept; + +class Q_GUI_EXPORT QShaderCode +{ +public: + QShaderCode() = default; + QShaderCode(const QByteArray &code, const QByteArray &entry = QByteArray()); + + QByteArray shader() const { return m_shader; } + void setShader(const QByteArray &code) { m_shader = code; } + + QByteArray entryPoint() const { return m_entryPoint; } + void setEntryPoint(const QByteArray &entry) { m_entryPoint = entry; } + +private: + friend Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t) noexcept; + + QByteArray m_shader; + QByteArray m_entryPoint; +}; + +Q_DECLARE_TYPEINFO(QShaderCode, Q_RELOCATABLE_TYPE); + +class Q_GUI_EXPORT QShader +{ +public: + enum Stage { + VertexStage = 0, + TessellationControlStage, + TessellationEvaluationStage, + GeometryStage, + FragmentStage, + ComputeStage + }; + + enum Source { + SpirvShader = 0, + GlslShader, + HlslShader, + DxbcShader, // fxc + MslShader, + DxilShader, // dxc + MetalLibShader, // xcrun metal + xcrun metallib + WgslShader + }; + + enum Variant { + StandardShader = 0, + BatchableVertexShader, + UInt16IndexedVertexAsComputeShader, + UInt32IndexedVertexAsComputeShader, + NonIndexedVertexAsComputeShader + }; + + enum class SerializedFormatVersion { + Latest = 0, + Qt_6_5, + Qt_6_4 + }; + + QShader(); + QShader(const QShader &other); + QShader &operator=(const QShader &other); + QShader(QShader &&other) noexcept : d(std::exchange(other.d, nullptr)) {} + QT_MOVE_ASSIGNMENT_OPERATOR_IMPL_VIA_PURE_SWAP(QShader) + ~QShader(); + + void swap(QShader &other) noexcept { qt_ptr_swap(d, other.d); } + void detach(); + + bool isValid() const; + + Stage stage() const; + void setStage(Stage stage); + + QShaderDescription description() const; + void setDescription(const QShaderDescription &desc); + + QList<QShaderKey> availableShaders() const; + QShaderCode shader(const QShaderKey &key) const; + void setShader(const QShaderKey &key, const QShaderCode &shader); + void removeShader(const QShaderKey &key); + + QByteArray serialized(SerializedFormatVersion version = SerializedFormatVersion::Latest) const; + static QShader fromSerialized(const QByteArray &data); + + using NativeResourceBindingMap = QMap<int, QPair<int, int> >; // binding -> native_binding[, native_binding] + NativeResourceBindingMap nativeResourceBindingMap(const QShaderKey &key) const; + void setResourceBindingMap(const QShaderKey &key, const NativeResourceBindingMap &map); + void removeResourceBindingMap(const QShaderKey &key); + + struct SeparateToCombinedImageSamplerMapping { + QByteArray combinedSamplerName; + int textureBinding; + int samplerBinding; + }; + using SeparateToCombinedImageSamplerMappingList = QList<SeparateToCombinedImageSamplerMapping>; + SeparateToCombinedImageSamplerMappingList separateToCombinedImageSamplerMappingList(const QShaderKey &key) const; + void setSeparateToCombinedImageSamplerMappingList(const QShaderKey &key, + const SeparateToCombinedImageSamplerMappingList &list); + void removeSeparateToCombinedImageSamplerMappingList(const QShaderKey &key); + + struct NativeShaderInfo { + int flags = 0; + QMap<int, int> extraBufferBindings; + }; + NativeShaderInfo nativeShaderInfo(const QShaderKey &key) const; + void setNativeShaderInfo(const QShaderKey &key, const NativeShaderInfo &info); + void removeNativeShaderInfo(const QShaderKey &key); + +private: + QShaderPrivate *d; + friend struct QShaderPrivate; + friend Q_GUI_EXPORT bool operator==(const QShader &, const QShader &) noexcept; + friend Q_GUI_EXPORT size_t qHash(const QShader &, size_t) noexcept; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); +#endif +}; + +class Q_GUI_EXPORT QShaderKey +{ +public: + QShaderKey() = default; + QShaderKey(QShader::Source s, + const QShaderVersion &sver, + QShader::Variant svar = QShader::StandardShader); + + QShader::Source source() const { return m_source; } + void setSource(QShader::Source s) { m_source = s; } + + QShaderVersion sourceVersion() const { return m_sourceVersion; } + void setSourceVersion(const QShaderVersion &sver) { m_sourceVersion = sver; } + + QShader::Variant sourceVariant() const { return m_sourceVariant; } + void setSourceVariant(QShader::Variant svar) { m_sourceVariant = svar; } + +private: + QShader::Source m_source = QShader::SpirvShader; + QShaderVersion m_sourceVersion; + QShader::Variant m_sourceVariant = QShader::StandardShader; +}; + +Q_DECLARE_TYPEINFO(QShaderKey, Q_RELOCATABLE_TYPE); + +Q_GUI_EXPORT bool operator==(const QShader &lhs, const QShader &rhs) noexcept; +Q_GUI_EXPORT size_t qHash(const QShader &s, size_t seed = 0) noexcept; + +inline bool operator!=(const QShader &lhs, const QShader &rhs) noexcept +{ + return !(lhs == rhs); +} + +Q_GUI_EXPORT bool operator==(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; +Q_GUI_EXPORT bool operator<(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; +Q_GUI_EXPORT bool operator<(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept; + +inline bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) noexcept +{ + return !(lhs == rhs); +} + +Q_GUI_EXPORT size_t qHash(const QShaderKey &k, size_t seed = 0) noexcept; + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); +Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderKey &k); +Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderVersion &v); +#endif + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qshader_p.h b/src/gui/rhi/qshader_p.h index f85b075a83..f77bcb1259 100644 --- a/src/gui/rhi/qshader_p.h +++ b/src/gui/rhi/qshader_p.h @@ -1,4 +1,4 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only #ifndef QSHADER_P_H @@ -15,225 +15,77 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <QtCore/qhash.h> -#include <QtCore/qmap.h> -#include <private/qshaderdescription_p.h> +#include <rhi/qshader.h> +#include <QtCore/QAtomicInt> +#include <QtCore/QMap> +#include <QtCore/QDebug> QT_BEGIN_NAMESPACE -struct QShaderPrivate; -class QShaderKey; - -#ifdef Q_OS_INTEGRITY - class QShaderVersion; - size_t qHash(const QShaderVersion &, size_t = 0) noexcept; -#endif - -class Q_GUI_EXPORT QShaderVersion -{ -public: - enum Flag { - GlslEs = 0x01 - }; - Q_DECLARE_FLAGS(Flags, Flag) - - QShaderVersion() = default; - QShaderVersion(int v, Flags f = Flags()); - - int version() const { return m_version; } - void setVersion(int v) { m_version = v; } - - Flags flags() const { return m_flags; } - void setFlags(Flags f) { m_flags = f; } - -private: - int m_version = 100; - Flags m_flags; -}; - -Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderVersion::Flags) -Q_DECLARE_TYPEINFO(QShaderVersion, Q_RELOCATABLE_TYPE); - -class QShaderCode; -Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t = 0) noexcept; - -class Q_GUI_EXPORT QShaderCode -{ -public: - QShaderCode() = default; - QShaderCode(const QByteArray &code, const QByteArray &entry = QByteArray()); - - QByteArray shader() const { return m_shader; } - void setShader(const QByteArray &code) { m_shader = code; } - - QByteArray entryPoint() const { return m_entryPoint; } - void setEntryPoint(const QByteArray &entry) { m_entryPoint = entry; } - -private: - friend Q_GUI_EXPORT size_t qHash(const QShaderCode &, size_t) noexcept; - - QByteArray m_shader; - QByteArray m_entryPoint; -}; - -Q_DECLARE_TYPEINFO(QShaderCode, Q_RELOCATABLE_TYPE); - -class Q_GUI_EXPORT QShader +struct Q_GUI_EXPORT QShaderPrivate { -public: - enum Stage { - VertexStage = 0, - TessellationControlStage, - TessellationEvaluationStage, - GeometryStage, - FragmentStage, - ComputeStage - }; - - enum Source { - SpirvShader = 0, - GlslShader, - HlslShader, - DxbcShader, // fxc - MslShader, - DxilShader, // dxc - MetalLibShader, // xcrun metal + xcrun metallib - WgslShader - }; - - enum Variant { - StandardShader = 0, - BatchableVertexShader, - UInt16IndexedVertexAsComputeShader, - UInt32IndexedVertexAsComputeShader, - NonIndexedVertexAsComputeShader - }; - - enum class SerializedFormatVersion { - Latest = 0, - Qt_6_5, - Qt_6_4 - }; - - QShader(); - QShader(const QShader &other); - QShader &operator=(const QShader &other); - ~QShader(); - void detach(); - - bool isValid() const; - - Stage stage() const; - void setStage(Stage stage); - - QShaderDescription description() const; - void setDescription(const QShaderDescription &desc); - - QList<QShaderKey> availableShaders() const; - QShaderCode shader(const QShaderKey &key) const; - void setShader(const QShaderKey &key, const QShaderCode &shader); - void removeShader(const QShaderKey &key); - - QByteArray serialized(SerializedFormatVersion version = SerializedFormatVersion::Latest) const; - static QShader fromSerialized(const QByteArray &data); - - using NativeResourceBindingMap = QMap<int, QPair<int, int> >; // binding -> native_binding[, native_binding] - NativeResourceBindingMap nativeResourceBindingMap(const QShaderKey &key) const; - void setResourceBindingMap(const QShaderKey &key, const NativeResourceBindingMap &map); - void removeResourceBindingMap(const QShaderKey &key); - - struct SeparateToCombinedImageSamplerMapping { - QByteArray combinedSamplerName; - int textureBinding; - int samplerBinding; - }; - using SeparateToCombinedImageSamplerMappingList = QList<SeparateToCombinedImageSamplerMapping>; - SeparateToCombinedImageSamplerMappingList separateToCombinedImageSamplerMappingList(const QShaderKey &key) const; - void setSeparateToCombinedImageSamplerMappingList(const QShaderKey &key, - const SeparateToCombinedImageSamplerMappingList &list); - void removeSeparateToCombinedImageSamplerMappingList(const QShaderKey &key); - - struct NativeShaderInfo { - int flags = 0; - QMap<int, int> extraBufferBindings; + static const int QSB_VERSION = 9; + static const int QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS = 8; + static const int QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO = 7; + static const int QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO = 6; + static const int QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS = 5; + static const int QSB_VERSION_WITHOUT_VAR_ARRAYDIMS = 4; + static const int QSB_VERSION_WITH_CBOR = 3; + static const int QSB_VERSION_WITH_BINARY_JSON = 2; + static const int QSB_VERSION_WITHOUT_BINDINGS = 1; + + enum MslNativeShaderInfoExtraBufferBindings { + MslTessVertIndicesBufferBinding = 0, + MslTessVertTescOutputBufferBinding, + MslTessTescTessLevelBufferBinding, + MslTessTescPatchOutputBufferBinding, + MslTessTescParamsBufferBinding, + MslTessTescInputBufferBinding, + MslBufferSizeBufferBinding, + MslMultiViewMaskBufferBinding }; - NativeShaderInfo nativeShaderInfo(const QShaderKey &key) const; - void setNativeShaderInfo(const QShaderKey &key, const NativeShaderInfo &info); - void removeNativeShaderInfo(const QShaderKey &key); - -private: - QShaderPrivate *d; - friend struct QShaderPrivate; - friend Q_GUI_EXPORT bool operator==(const QShader &, const QShader &) noexcept; - friend Q_GUI_EXPORT size_t qHash(const QShader &, size_t) noexcept; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); -#endif -}; - -class Q_GUI_EXPORT QShaderKey -{ -public: - QShaderKey() = default; - QShaderKey(QShader::Source s, - const QShaderVersion &sver, - QShader::Variant svar = QShader::StandardShader); - - QShader::Source source() const { return m_source; } - void setSource(QShader::Source s) { m_source = s; } - - QShaderVersion sourceVersion() const { return m_sourceVersion; } - void setSourceVersion(const QShaderVersion &sver) { m_sourceVersion = sver; } - QShader::Variant sourceVariant() const { return m_sourceVariant; } - void setSourceVariant(QShader::Variant svar) { m_sourceVariant = svar; } - -private: - QShader::Source m_source = QShader::SpirvShader; - QShaderVersion m_sourceVersion; - QShader::Variant m_sourceVariant = QShader::StandardShader; + QShaderPrivate() + : ref(1) + { + } + + QShaderPrivate(const QShaderPrivate &other) + : ref(1), + qsbVersion(other.qsbVersion), + stage(other.stage), + desc(other.desc), + shaders(other.shaders), + bindings(other.bindings), + combinedImageMap(other.combinedImageMap), + nativeShaderInfoMap(other.nativeShaderInfoMap) + { + } + + static QShaderPrivate *get(QShader *s) { return s->d; } + static const QShaderPrivate *get(const QShader *s) { return s->d; } + static int qtQsbVersion(QShader::SerializedFormatVersion qtVersion) { + switch (qtVersion) { + case QShader::SerializedFormatVersion::Qt_6_4: + return (QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS + 1); + case QShader::SerializedFormatVersion::Qt_6_5: + return (QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO + 1); + default: + return QShaderPrivate::QSB_VERSION; + } + } + + QAtomicInt ref; + int qsbVersion = QSB_VERSION; + QShader::Stage stage = QShader::VertexStage; + QShaderDescription desc; + // QMap not QHash because we need to be able to iterate based on sorted keys + QMap<QShaderKey, QShaderCode> shaders; + QMap<QShaderKey, QShader::NativeResourceBindingMap> bindings; + QMap<QShaderKey, QShader::SeparateToCombinedImageSamplerMappingList> combinedImageMap; + QMap<QShaderKey, QShader::NativeShaderInfo> nativeShaderInfoMap; }; -Q_DECLARE_TYPEINFO(QShaderKey, Q_RELOCATABLE_TYPE); - -Q_GUI_EXPORT bool operator==(const QShader &lhs, const QShader &rhs) noexcept; -Q_GUI_EXPORT size_t qHash(const QShader &s, size_t seed = 0) noexcept; - -inline bool operator!=(const QShader &lhs, const QShader &rhs) noexcept -{ - return !(lhs == rhs); -} - -Q_GUI_EXPORT bool operator==(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; -Q_GUI_EXPORT bool operator<(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; -Q_GUI_EXPORT bool operator<(const QShaderKey &lhs, const QShaderKey &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderCode &lhs, const QShaderCode &rhs) noexcept; - -inline bool operator!=(const QShaderVersion &lhs, const QShaderVersion &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderKey &lhs, const QShaderKey &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderCode &lhs, const QShaderCode &rhs) noexcept -{ - return !(lhs == rhs); -} - -Q_GUI_EXPORT size_t qHash(const QShaderKey &k, size_t seed = 0) noexcept; - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShader &); -Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderKey &k); -Q_GUI_EXPORT QDebug operator<<(QDebug dbg, const QShaderVersion &v); -#endif - QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qshader_p_p.h b/src/gui/rhi/qshader_p_p.h deleted file mode 100644 index 9cc9207f69..0000000000 --- a/src/gui/rhi/qshader_p_p.h +++ /dev/null @@ -1,90 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QSHADER_P_P_H -#define QSHADER_P_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists for the convenience -// of a number of Qt sources files. This header file may change from -// version to version without notice, or even be removed. -// -// We mean it. -// - -#include "qshader_p.h" -#include <QtCore/QAtomicInt> -#include <QtCore/QMap> -#include <QtCore/QDebug> - -QT_BEGIN_NAMESPACE - -struct Q_GUI_EXPORT QShaderPrivate -{ - static const int QSB_VERSION = 9; - static const int QSB_VERSION_WITHOUT_INPUT_OUTPUT_INTERFACE_BLOCKS = 8; - static const int QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO = 7; - static const int QSB_VERSION_WITHOUT_NATIVE_SHADER_INFO = 6; - static const int QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS = 5; - static const int QSB_VERSION_WITHOUT_VAR_ARRAYDIMS = 4; - static const int QSB_VERSION_WITH_CBOR = 3; - static const int QSB_VERSION_WITH_BINARY_JSON = 2; - static const int QSB_VERSION_WITHOUT_BINDINGS = 1; - - enum MslNativeShaderInfoExtraBufferBindings { - MslTessVertIndicesBufferBinding = 0, - MslTessVertTescOutputBufferBinding, - MslTessTescTessLevelBufferBinding, - MslTessTescPatchOutputBufferBinding, - MslTessTescParamsBufferBinding, - MslTessTescInputBufferBinding, - MslBufferSizeBufferBinding - }; - - QShaderPrivate() - : ref(1) - { - } - - QShaderPrivate(const QShaderPrivate &other) - : ref(1), - qsbVersion(other.qsbVersion), - stage(other.stage), - desc(other.desc), - shaders(other.shaders), - bindings(other.bindings), - combinedImageMap(other.combinedImageMap), - nativeShaderInfoMap(other.nativeShaderInfoMap) - { - } - - static QShaderPrivate *get(QShader *s) { return s->d; } - static const QShaderPrivate *get(const QShader *s) { return s->d; } - static int qtQsbVersion(QShader::SerializedFormatVersion qtVersion) { - switch (qtVersion) { - case QShader::SerializedFormatVersion::Qt_6_4: - return (QShaderPrivate::QSB_VERSION_WITHOUT_SEPARATE_IMAGES_AND_SAMPLERS + 1); - case QShader::SerializedFormatVersion::Qt_6_5: - return (QShaderPrivate::QSB_VERSION_WITHOUT_EXTENDED_STORAGE_BUFFER_INFO + 1); - default: - return QShaderPrivate::QSB_VERSION; - } - } - - QAtomicInt ref; - int qsbVersion = QSB_VERSION; - QShader::Stage stage = QShader::VertexStage; - QShaderDescription desc; - // QMap not QHash because we need to be able to iterate based on sorted keys - QMap<QShaderKey, QShaderCode> shaders; - QMap<QShaderKey, QShader::NativeResourceBindingMap> bindings; - QMap<QShaderKey, QShader::SeparateToCombinedImageSamplerMappingList> combinedImageMap; - QMap<QShaderKey, QShader::NativeShaderInfo> nativeShaderInfoMap; -}; - -QT_END_NAMESPACE - -#endif diff --git a/src/gui/rhi/qshaderdescription.cpp b/src/gui/rhi/qshaderdescription.cpp index 44fcd936fd..f64daf02ef 100644 --- a/src/gui/rhi/qshaderdescription.cpp +++ b/src/gui/rhi/qshaderdescription.cpp @@ -1,8 +1,8 @@ -// Copyright (C) 2019 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#include "qshaderdescription_p_p.h" -#include "qshader_p_p.h" +#include "qshaderdescription_p.h" +#include "qshader_p.h" #include <QDebug> #include <QDataStream> #include <QJsonObject> @@ -12,11 +12,22 @@ QT_BEGIN_NAMESPACE /*! \class QShaderDescription - \internal + \ingroup painting-3D \inmodule QtGui + \since 6.6 \brief Describes the interface of a shader. + \warning The QRhi family of classes in the Qt Gui module, including QShader + and QShaderDescription, offer limited compatibility guarantees. There are + no source or binary compatibility guarantees for these classes, meaning the + API is only guaranteed to work with the Qt version the application was + developed against. Source incompatible changes are however aimed to be kept + at a minimum and will only be made in minor releases (6.7, 6.8, and so on). + To use these classes in an application, link to + \c{Qt::GuiPrivate} (if using CMake), and include the headers with the \c + rhi prefix, for example \c{#include <rhi/qshaderdescription.h>}. + A shader typically has a set of inputs and outputs. A vertex shader for example has a number of input variables and may use one or more uniform buffers to access data (e.g. a modelview matrix) provided by the @@ -51,8 +62,6 @@ QT_BEGIN_NAMESPACE float opacity; } ubuf; - out gl_PerVertex { vec4 gl_Position; }; - void main() { v_color = color; @@ -200,28 +209,179 @@ QT_BEGIN_NAMESPACE \value ImageRect \value ImageBuffer \value Struct + \value Half + \value Half2 + \value Half3 + \value Half4 */ /*! - \class QShaderDescription::InOutVariable - \internal + \enum QShaderDescription::ImageFormat + Image format. + + \value ImageFormatUnknown + \value ImageFormatRgba32f + \value ImageFormatRgba16f + \value ImageFormatR32f + \value ImageFormatRgba8 + \value ImageFormatRgba8Snorm + \value ImageFormatRg32f + \value ImageFormatRg16f + \value ImageFormatR11fG11fB10f + \value ImageFormatR16f + \value ImageFormatRgba16 + \value ImageFormatRgb10A2 + \value ImageFormatRg16 + \value ImageFormatRg8 + \value ImageFormatR16 + \value ImageFormatR8 + \value ImageFormatRgba16Snorm + \value ImageFormatRg16Snorm + \value ImageFormatRg8Snorm + \value ImageFormatR16Snorm + \value ImageFormatR8Snorm + \value ImageFormatRgba32i + \value ImageFormatRgba16i + \value ImageFormatRgba8i + \value ImageFormatR32i + \value ImageFormatRg32i + \value ImageFormatRg16i + \value ImageFormatRg8i + \value ImageFormatR16i + \value ImageFormatR8i + \value ImageFormatRgba32ui + \value ImageFormatRgba16ui + \value ImageFormatRgba8ui + \value ImageFormatR32ui + \value ImageFormatRgb10a2ui + \value ImageFormatRg32ui + \value ImageFormatRg16ui + \value ImageFormatRg8ui + \value ImageFormatR16ui + \value ImageFormatR8ui + */ + +/*! + \enum QShaderDescription::ImageFlag + Image flags. + + \value ReadOnlyImage + \value WriteOnlyImage + */ + +/*! + \enum QShaderDescription::QualifierFlag + Qualifier flags. + + \value QualifierReadOnly + \value QualifierWriteOnly + \value QualifierCoherent + \value QualifierVolatile + \value QualifierRestrict + */ + +/*! + \struct QShaderDescription::InOutVariable \inmodule QtGui + \since 6.6 \brief Describes an input or output variable in the shader. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::BlockVariable - \internal + \variable QShaderDescription::InOutVariable::name + */ + +/*! + \variable QShaderDescription::InOutVariable::type + */ + +/*! + \variable QShaderDescription::InOutVariable::location + */ + +/*! + \variable QShaderDescription::InOutVariable::binding + */ + +/*! + \variable QShaderDescription::InOutVariable::descriptorSet + */ + +/*! + \variable QShaderDescription::InOutVariable::imageFormat + */ + +/*! + \variable QShaderDescription::InOutVariable::imageFlags + */ + +/*! + \variable QShaderDescription::InOutVariable::arrayDims + */ + +/*! + \variable QShaderDescription::InOutVariable::perPatch + */ + +/*! + \variable QShaderDescription::InOutVariable::structMembers + */ + +/*! + \struct QShaderDescription::BlockVariable \inmodule QtGui + \since 6.6 \brief Describes a member of a uniform or push constant block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::UniformBlock - \internal + \variable QShaderDescription::BlockVariable::name + */ + +/*! + \variable QShaderDescription::BlockVariable::type + */ + +/*! + \variable QShaderDescription::BlockVariable::offset + */ + +/*! + \variable QShaderDescription::BlockVariable::size + */ + +/*! + \variable QShaderDescription::BlockVariable::arrayDims + */ + +/*! + \variable QShaderDescription::BlockVariable::arrayStride + */ + +/*! + \variable QShaderDescription::BlockVariable::matrixStride + */ + +/*! + \variable QShaderDescription::BlockVariable::matrixIsRowMajor + */ + +/*! + \variable QShaderDescription::BlockVariable::structMembers + */ + +/*! + \struct QShaderDescription::UniformBlock \inmodule QtGui + \since 6.6 \brief Describes a uniform block. @@ -229,22 +389,157 @@ QT_BEGIN_NAMESPACE (like GLSL 120 or GLSL/ES 100), uniform blocks are replaced with ordinary uniforms in a struct. The name of the struct, and so the prefix for the uniforms generated from the block members, is given by structName. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::PushConstantBlock - \internal + \variable QShaderDescription::UniformBlock::blockName + */ + +/*! + \variable QShaderDescription::UniformBlock::structName + */ + +/*! + \variable QShaderDescription::UniformBlock::size + */ + +/*! + \variable QShaderDescription::UniformBlock::binding + */ + +/*! + \variable QShaderDescription::UniformBlock::descriptorSet + */ + +/*! + \variable QShaderDescription::UniformBlock::members + */ + +/*! + \struct QShaderDescription::PushConstantBlock \inmodule QtGui + \since 6.6 \brief Describes a push constant block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. */ /*! - \class QShaderDescription::StorageBlock - \internal + \variable QShaderDescription::PushConstantBlock::name + */ + +/*! + \variable QShaderDescription::PushConstantBlock::size + */ + +/*! + \variable QShaderDescription::PushConstantBlock::members + */ + +/*! + \struct QShaderDescription::StorageBlock \inmodule QtGui + \since 6.6 \brief Describes a shader storage block. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. + */ + +/*! + \variable QShaderDescription::StorageBlock::blockName + */ + +/*! + \variable QShaderDescription::StorageBlock::instanceName + */ + +/*! + \variable QShaderDescription::StorageBlock::knownSize + */ + +/*! + \variable QShaderDescription::StorageBlock::binding + */ + +/*! + \variable QShaderDescription::StorageBlock::descriptorSet + */ + +/*! + \variable QShaderDescription::StorageBlock::members + */ + +/*! + \variable QShaderDescription::StorageBlock::runtimeArrayStride + */ + +/*! + \variable QShaderDescription::StorageBlock::qualifierFlags + */ + +/*! + \struct QShaderDescription::BuiltinVariable + \inmodule QtGui + \since 6.6 + + \brief Describes a built-in variable. + + \note This is a RHI API with limited compatibility guarantees, see \l QShaderDescription + for details. + */ + +/*! + \variable QShaderDescription::BuiltinVariable::type + */ + +/*! + \variable QShaderDescription::BuiltinVariable::varType + */ + +/*! + \variable QShaderDescription::BuiltinVariable::arrayDims + */ + +/*! + \enum QShaderDescription::BuiltinType + Built-in variable type. + + \value PositionBuiltin + \value PointSizeBuiltin + \value ClipDistanceBuiltin + \value CullDistanceBuiltin + \value VertexIdBuiltin + \value InstanceIdBuiltin + \value PrimitiveIdBuiltin + \value InvocationIdBuiltin + \value LayerBuiltin + \value ViewportIndexBuiltin + \value TessLevelOuterBuiltin + \value TessLevelInnerBuiltin + \value TessCoordBuiltin + \value PatchVerticesBuiltin + \value FragCoordBuiltin + \value PointCoordBuiltin + \value FrontFacingBuiltin + \value SampleIdBuiltin + \value SamplePositionBuiltin + \value SampleMaskBuiltin + \value FragDepthBuiltin + \value NumWorkGroupsBuiltin + \value WorkgroupSizeBuiltin + \value WorkgroupIdBuiltin + \value LocalInvocationIdBuiltin + \value GlobalInvocationIdBuiltin + \value LocalInvocationIndexBuiltin + \value VertexIndexBuiltin + \value InstanceIndexBuiltin */ /*! @@ -267,7 +562,7 @@ void QShaderDescription::detach() } /*! - \internal + Constructs a copy of \a other. */ QShaderDescription::QShaderDescription(const QShaderDescription &other) : d(other.d) @@ -276,7 +571,7 @@ QShaderDescription::QShaderDescription(const QShaderDescription &other) } /*! - \internal + Assigns \a other to this object. */ QShaderDescription &QShaderDescription::operator=(const QShaderDescription &other) { @@ -574,7 +869,7 @@ uint QShaderDescription::tessellationOutputVertexCount() const \value UnknownTessellationMode \value TrianglesTessellationMode \value QuadTessellationMode - \value IsolinesTessellationMode + \value IsolineTessellationMode */ /*! diff --git a/src/gui/rhi/qshaderdescription.h b/src/gui/rhi/qshaderdescription.h new file mode 100644 index 0000000000..02492a1598 --- /dev/null +++ b/src/gui/rhi/qshaderdescription.h @@ -0,0 +1,386 @@ +// Copyright (C) 2023 The Qt Company Ltd. +// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only + +#ifndef QSHADERDESCRIPTION_H +#define QSHADERDESCRIPTION_H + +// +// W A R N I N G +// ------------- +// +// This file is part of the RHI API, with limited compatibility guarantees. +// Usage of this API may make your code source and binary incompatible with +// future versions of Qt. +// + +#include <QtGui/qtguiglobal.h> +#include <QtCore/qstring.h> +#include <QtCore/qlist.h> +#include <array> + +QT_BEGIN_NAMESPACE + +struct QShaderDescriptionPrivate; +class QDataStream; + +class Q_GUI_EXPORT QShaderDescription +{ +public: + QShaderDescription(); + QShaderDescription(const QShaderDescription &other); + QShaderDescription &operator=(const QShaderDescription &other); + ~QShaderDescription(); + void detach(); + + bool isValid() const; + + void serialize(QDataStream *stream, int version) const; + QByteArray toJson() const; + + static QShaderDescription deserialize(QDataStream *stream, int version); + + enum VariableType { + Unknown = 0, + + // do not reorder + Float, + Vec2, + Vec3, + Vec4, + Mat2, + Mat2x3, + Mat2x4, + Mat3, + Mat3x2, + Mat3x4, + Mat4, + Mat4x2, + Mat4x3, + + Int, + Int2, + Int3, + Int4, + + Uint, + Uint2, + Uint3, + Uint4, + + Bool, + Bool2, + Bool3, + Bool4, + + Double, + Double2, + Double3, + Double4, + DMat2, + DMat2x3, + DMat2x4, + DMat3, + DMat3x2, + DMat3x4, + DMat4, + DMat4x2, + DMat4x3, + + Sampler1D, + Sampler2D, + Sampler2DMS, + Sampler3D, + SamplerCube, + Sampler1DArray, + Sampler2DArray, + Sampler2DMSArray, + Sampler3DArray, + SamplerCubeArray, + SamplerRect, + SamplerBuffer, + SamplerExternalOES, + Sampler, + + Image1D, + Image2D, + Image2DMS, + Image3D, + ImageCube, + Image1DArray, + Image2DArray, + Image2DMSArray, + Image3DArray, + ImageCubeArray, + ImageRect, + ImageBuffer, + + Struct, + + Half, + Half2, + Half3, + Half4 + }; + + enum ImageFormat { + // must match SPIR-V's ImageFormat + ImageFormatUnknown = 0, + ImageFormatRgba32f = 1, + ImageFormatRgba16f = 2, + ImageFormatR32f = 3, + ImageFormatRgba8 = 4, + ImageFormatRgba8Snorm = 5, + ImageFormatRg32f = 6, + ImageFormatRg16f = 7, + ImageFormatR11fG11fB10f = 8, + ImageFormatR16f = 9, + ImageFormatRgba16 = 10, + ImageFormatRgb10A2 = 11, + ImageFormatRg16 = 12, + ImageFormatRg8 = 13, + ImageFormatR16 = 14, + ImageFormatR8 = 15, + ImageFormatRgba16Snorm = 16, + ImageFormatRg16Snorm = 17, + ImageFormatRg8Snorm = 18, + ImageFormatR16Snorm = 19, + ImageFormatR8Snorm = 20, + ImageFormatRgba32i = 21, + ImageFormatRgba16i = 22, + ImageFormatRgba8i = 23, + ImageFormatR32i = 24, + ImageFormatRg32i = 25, + ImageFormatRg16i = 26, + ImageFormatRg8i = 27, + ImageFormatR16i = 28, + ImageFormatR8i = 29, + ImageFormatRgba32ui = 30, + ImageFormatRgba16ui = 31, + ImageFormatRgba8ui = 32, + ImageFormatR32ui = 33, + ImageFormatRgb10a2ui = 34, + ImageFormatRg32ui = 35, + ImageFormatRg16ui = 36, + ImageFormatRg8ui = 37, + ImageFormatR16ui = 38, + ImageFormatR8ui = 39 + }; + + enum ImageFlag { + ReadOnlyImage = 1 << 0, + WriteOnlyImage = 1 << 1 + }; + Q_DECLARE_FLAGS(ImageFlags, ImageFlag) + + enum QualifierFlag { + QualifierReadOnly = 1 << 0, + QualifierWriteOnly = 1 << 1, + QualifierCoherent = 1 << 2, + QualifierVolatile = 1 << 3, + QualifierRestrict = 1 << 4, + }; + Q_DECLARE_FLAGS(QualifierFlags, QualifierFlag) + + // Optional data (like decorations) usually default to an otherwise invalid value (-1 or 0). This is intentional. + + struct BlockVariable { + QByteArray name; + VariableType type = Unknown; + int offset = 0; + int size = 0; + QList<int> arrayDims; + int arrayStride = 0; + int matrixStride = 0; + bool matrixIsRowMajor = false; + QList<BlockVariable> structMembers; + }; + + struct InOutVariable { + QByteArray name; + VariableType type = Unknown; + int location = -1; + int binding = -1; + int descriptorSet = -1; + ImageFormat imageFormat = ImageFormatUnknown; + ImageFlags imageFlags; + QList<int> arrayDims; + bool perPatch = false; + QList<BlockVariable> structMembers; + }; + + struct UniformBlock { + QByteArray blockName; + QByteArray structName; // instanceName + int size = 0; + int binding = -1; + int descriptorSet = -1; + QList<BlockVariable> members; + }; + + struct PushConstantBlock { + QByteArray name; + int size = 0; + QList<BlockVariable> members; + }; + + struct StorageBlock { + QByteArray blockName; + QByteArray instanceName; + int knownSize = 0; + int binding = -1; + int descriptorSet = -1; + QList<BlockVariable> members; + int runtimeArrayStride = 0; + QualifierFlags qualifierFlags; + }; + + QList<InOutVariable> inputVariables() const; + QList<InOutVariable> outputVariables() const; + QList<UniformBlock> uniformBlocks() const; + QList<PushConstantBlock> pushConstantBlocks() const; + QList<StorageBlock> storageBlocks() const; + QList<InOutVariable> combinedImageSamplers() const; + QList<InOutVariable> separateImages() const; + QList<InOutVariable> separateSamplers() const; + QList<InOutVariable> storageImages() const; + + enum BuiltinType { + // must match SpvBuiltIn + PositionBuiltin = 0, + PointSizeBuiltin = 1, + ClipDistanceBuiltin = 3, + CullDistanceBuiltin = 4, + VertexIdBuiltin = 5, + InstanceIdBuiltin = 6, + PrimitiveIdBuiltin = 7, + InvocationIdBuiltin = 8, + LayerBuiltin = 9, + ViewportIndexBuiltin = 10, + TessLevelOuterBuiltin = 11, + TessLevelInnerBuiltin = 12, + TessCoordBuiltin = 13, + PatchVerticesBuiltin = 14, + FragCoordBuiltin = 15, + PointCoordBuiltin = 16, + FrontFacingBuiltin = 17, + SampleIdBuiltin = 18, + SamplePositionBuiltin = 19, + SampleMaskBuiltin = 20, + FragDepthBuiltin = 22, + NumWorkGroupsBuiltin = 24, + WorkgroupSizeBuiltin = 25, + WorkgroupIdBuiltin = 26, + LocalInvocationIdBuiltin = 27, + GlobalInvocationIdBuiltin = 28, + LocalInvocationIndexBuiltin = 29, + VertexIndexBuiltin = 42, + InstanceIndexBuiltin = 43 + }; + + struct BuiltinVariable { + BuiltinType type; + VariableType varType; + QList<int> arrayDims; + }; + + QList<BuiltinVariable> inputBuiltinVariables() const; + QList<BuiltinVariable> outputBuiltinVariables() const; + + std::array<uint, 3> computeShaderLocalSize() const; + + uint tessellationOutputVertexCount() const; + + enum TessellationMode { + UnknownTessellationMode, + TrianglesTessellationMode, + QuadTessellationMode, + IsolineTessellationMode + }; + + TessellationMode tessellationMode() const; + + enum TessellationWindingOrder { + UnknownTessellationWindingOrder, + CwTessellationWindingOrder, + CcwTessellationWindingOrder + }; + + TessellationWindingOrder tessellationWindingOrder() const; + + enum TessellationPartitioning { + UnknownTessellationPartitioning, + EqualTessellationPartitioning, + FractionalEvenTessellationPartitioning, + FractionalOddTessellationPartitioning + }; + + TessellationPartitioning tessellationPartitioning() const; + +private: + QShaderDescriptionPrivate *d; + friend struct QShaderDescriptionPrivate; +#ifndef QT_NO_DEBUG_STREAM + friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); +#endif + friend Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; +}; + +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::ImageFlags) +Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::QualifierFlags) + +#ifndef QT_NO_DEBUG_STREAM +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::InOutVariable &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BlockVariable &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::UniformBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::PushConstantBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::StorageBlock &); +Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BuiltinVariable &); +#endif + +Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept; +Q_GUI_EXPORT bool operator==(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept; + +inline bool operator!=(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept +{ + return !(lhs == rhs); +} + +inline bool operator!=(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept +{ + return !(lhs == rhs); +} + +QT_END_NAMESPACE + +#endif diff --git a/src/gui/rhi/qshaderdescription_p.h b/src/gui/rhi/qshaderdescription_p.h index 990cb12003..df694bbf40 100644 --- a/src/gui/rhi/qshaderdescription_p.h +++ b/src/gui/rhi/qshaderdescription_p.h @@ -1,8 +1,8 @@ -// Copyright (C) 2021 The Qt Company Ltd. +// Copyright (C) 2023 The Qt Company Ltd. // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only -#ifndef QSHADERDESCRIPTION_H -#define QSHADERDESCRIPTION_H +#ifndef QSHADERDESCRIPTION_P_H +#define QSHADERDESCRIPTION_P_H // // W A R N I N G @@ -15,376 +15,67 @@ // We mean it. // -#include <QtGui/qtguiglobal.h> -#include <QtCore/QString> +#include <rhi/qshaderdescription.h> #include <QtCore/QList> -#include <QtCore/private/qglobal_p.h> -#include <array> +#include <QtCore/QAtomicInt> +#include <QtCore/QJsonDocument> QT_BEGIN_NAMESPACE -struct QShaderDescriptionPrivate; -class QDataStream; - -class Q_GUI_EXPORT QShaderDescription +struct Q_GUI_EXPORT QShaderDescriptionPrivate { -public: - QShaderDescription(); - QShaderDescription(const QShaderDescription &other); - QShaderDescription &operator=(const QShaderDescription &other); - ~QShaderDescription(); - void detach(); - - bool isValid() const; - - void serialize(QDataStream *stream, int version) const; - QByteArray toJson() const; - - static QShaderDescription deserialize(QDataStream *stream, int version); - - enum VariableType { - Unknown = 0, - - // do not reorder - Float, - Vec2, - Vec3, - Vec4, - Mat2, - Mat2x3, - Mat2x4, - Mat3, - Mat3x2, - Mat3x4, - Mat4, - Mat4x2, - Mat4x3, - - Int, - Int2, - Int3, - Int4, - - Uint, - Uint2, - Uint3, - Uint4, - - Bool, - Bool2, - Bool3, - Bool4, - - Double, - Double2, - Double3, - Double4, - DMat2, - DMat2x3, - DMat2x4, - DMat3, - DMat3x2, - DMat3x4, - DMat4, - DMat4x2, - DMat4x3, - - Sampler1D, - Sampler2D, - Sampler2DMS, - Sampler3D, - SamplerCube, - Sampler1DArray, - Sampler2DArray, - Sampler2DMSArray, - Sampler3DArray, - SamplerCubeArray, - SamplerRect, - SamplerBuffer, - SamplerExternalOES, - Sampler, - - Image1D, - Image2D, - Image2DMS, - Image3D, - ImageCube, - Image1DArray, - Image2DArray, - Image2DMSArray, - Image3DArray, - ImageCubeArray, - ImageRect, - ImageBuffer, - - Struct, - - Half, - Half2, - Half3, - Half4 - }; - - enum ImageFormat { - // must match SPIR-V's ImageFormat - ImageFormatUnknown = 0, - ImageFormatRgba32f = 1, - ImageFormatRgba16f = 2, - ImageFormatR32f = 3, - ImageFormatRgba8 = 4, - ImageFormatRgba8Snorm = 5, - ImageFormatRg32f = 6, - ImageFormatRg16f = 7, - ImageFormatR11fG11fB10f = 8, - ImageFormatR16f = 9, - ImageFormatRgba16 = 10, - ImageFormatRgb10A2 = 11, - ImageFormatRg16 = 12, - ImageFormatRg8 = 13, - ImageFormatR16 = 14, - ImageFormatR8 = 15, - ImageFormatRgba16Snorm = 16, - ImageFormatRg16Snorm = 17, - ImageFormatRg8Snorm = 18, - ImageFormatR16Snorm = 19, - ImageFormatR8Snorm = 20, - ImageFormatRgba32i = 21, - ImageFormatRgba16i = 22, - ImageFormatRgba8i = 23, - ImageFormatR32i = 24, - ImageFormatRg32i = 25, - ImageFormatRg16i = 26, - ImageFormatRg8i = 27, - ImageFormatR16i = 28, - ImageFormatR8i = 29, - ImageFormatRgba32ui = 30, - ImageFormatRgba16ui = 31, - ImageFormatRgba8ui = 32, - ImageFormatR32ui = 33, - ImageFormatRgb10a2ui = 34, - ImageFormatRg32ui = 35, - ImageFormatRg16ui = 36, - ImageFormatRg8ui = 37, - ImageFormatR16ui = 38, - ImageFormatR8ui = 39 - }; - - enum ImageFlag { - ReadOnlyImage = 1 << 0, - WriteOnlyImage = 1 << 1 - }; - Q_DECLARE_FLAGS(ImageFlags, ImageFlag) - - enum QualifierFlag { - QualifierReadOnly = 1 << 0, - QualifierWriteOnly = 1 << 1, - QualifierCoherent = 1 << 2, - QualifierVolatile = 1 << 3, - QualifierRestrict = 1 << 4, - }; - Q_DECLARE_FLAGS(QualifierFlags, QualifierFlag) - - // Optional data (like decorations) usually default to an otherwise invalid value (-1 or 0). This is intentional. - - struct BlockVariable + QShaderDescriptionPrivate() + : ref(1) { - QByteArray name; - VariableType type = Unknown; - int offset = 0; - int size = 0; - QList<int> arrayDims; - int arrayStride = 0; - int matrixStride = 0; - bool matrixIsRowMajor = false; - QList<BlockVariable> structMembers; - }; - - struct InOutVariable { - QByteArray name; - VariableType type = Unknown; - int location = -1; - int binding = -1; - int descriptorSet = -1; - ImageFormat imageFormat = ImageFormatUnknown; - ImageFlags imageFlags; - QList<int> arrayDims; - bool perPatch = false; - QList<BlockVariable> structMembers; - }; - - struct UniformBlock { - QByteArray blockName; - QByteArray structName; // instanceName - int size = 0; - int binding = -1; - int descriptorSet = -1; - QList<BlockVariable> members; - }; - - struct PushConstantBlock { - QByteArray name; - int size = 0; - QList<BlockVariable> members; - }; - - struct StorageBlock { - QByteArray blockName; - QByteArray instanceName; - int knownSize = 0; - int binding = -1; - int descriptorSet = -1; - QList<BlockVariable> members; - int runtimeArrayStride = 0; - QualifierFlags qualifierFlags; - }; - - QList<InOutVariable> inputVariables() const; - QList<InOutVariable> outputVariables() const; - QList<UniformBlock> uniformBlocks() const; - QList<PushConstantBlock> pushConstantBlocks() const; - QList<StorageBlock> storageBlocks() const; - QList<InOutVariable> combinedImageSamplers() const; - QList<InOutVariable> separateImages() const; - QList<InOutVariable> separateSamplers() const; - QList<InOutVariable> storageImages() const; - - enum BuiltinType { - // must match SpvBuiltIn - PositionBuiltin = 0, - PointSizeBuiltin = 1, - ClipDistanceBuiltin = 3, - CullDistanceBuiltin = 4, - VertexIdBuiltin = 5, - InstanceIdBuiltin = 6, - PrimitiveIdBuiltin = 7, - InvocationIdBuiltin = 8, - LayerBuiltin = 9, - ViewportIndexBuiltin = 10, - TessLevelOuterBuiltin = 11, - TessLevelInnerBuiltin = 12, - TessCoordBuiltin = 13, - PatchVerticesBuiltin = 14, - FragCoordBuiltin = 15, - PointCoordBuiltin = 16, - FrontFacingBuiltin = 17, - SampleIdBuiltin = 18, - SamplePositionBuiltin = 19, - SampleMaskBuiltin = 20, - FragDepthBuiltin = 22, - NumWorkGroupsBuiltin = 24, - WorkgroupSizeBuiltin = 25, - WorkgroupIdBuiltin = 26, - LocalInvocationIdBuiltin = 27, - GlobalInvocationIdBuiltin = 28, - LocalInvocationIndexBuiltin = 29, - VertexIndexBuiltin = 42, - InstanceIndexBuiltin = 43 - }; - - struct BuiltinVariable { - BuiltinType type; - VariableType varType; - QList<int> arrayDims; - }; - - QList<BuiltinVariable> inputBuiltinVariables() const; - QList<BuiltinVariable> outputBuiltinVariables() const; - - std::array<uint, 3> computeShaderLocalSize() const; - - uint tessellationOutputVertexCount() const; - - enum TessellationMode { - UnknownTessellationMode, - TrianglesTessellationMode, - QuadTessellationMode, - IsolineTessellationMode - }; - - TessellationMode tessellationMode() const; - - enum TessellationWindingOrder { - UnknownTessellationWindingOrder, - CwTessellationWindingOrder, - CcwTessellationWindingOrder - }; - - TessellationWindingOrder tessellationWindingOrder() const; - - enum TessellationPartitioning { - UnknownTessellationPartitioning, - EqualTessellationPartitioning, - FractionalEvenTessellationPartitioning, - FractionalOddTessellationPartitioning - }; - - TessellationPartitioning tessellationPartitioning() const; - -private: - QShaderDescriptionPrivate *d; - friend struct QShaderDescriptionPrivate; -#ifndef QT_NO_DEBUG_STREAM - friend Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); -#endif - friend Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; + } + + QShaderDescriptionPrivate(const QShaderDescriptionPrivate &other) + : ref(1), + inVars(other.inVars), + outVars(other.outVars), + uniformBlocks(other.uniformBlocks), + pushConstantBlocks(other.pushConstantBlocks), + storageBlocks(other.storageBlocks), + combinedImageSamplers(other.combinedImageSamplers), + separateImages(other.separateImages), + separateSamplers(other.separateSamplers), + storageImages(other.storageImages), + inBuiltins(other.inBuiltins), + outBuiltins(other.outBuiltins), + localSize(other.localSize), + tessOutVertCount(other.tessOutVertCount), + tessMode(other.tessMode), + tessWind(other.tessWind), + tessPart(other.tessPart) + { + } + + static QShaderDescriptionPrivate *get(QShaderDescription *desc) { return desc->d; } + static const QShaderDescriptionPrivate *get(const QShaderDescription *desc) { return desc->d; } + + QJsonDocument makeDoc(); + void writeToStream(QDataStream *stream, int version); + void loadFromStream(QDataStream *stream, int version); + + QAtomicInt ref; + QList<QShaderDescription::InOutVariable> inVars; + QList<QShaderDescription::InOutVariable> outVars; + QList<QShaderDescription::UniformBlock> uniformBlocks; + QList<QShaderDescription::PushConstantBlock> pushConstantBlocks; + QList<QShaderDescription::StorageBlock> storageBlocks; + QList<QShaderDescription::InOutVariable> combinedImageSamplers; + QList<QShaderDescription::InOutVariable> separateImages; + QList<QShaderDescription::InOutVariable> separateSamplers; + QList<QShaderDescription::InOutVariable> storageImages; + QList<QShaderDescription::BuiltinVariable> inBuiltins; + QList<QShaderDescription::BuiltinVariable> outBuiltins; + std::array<uint, 3> localSize = {}; + uint tessOutVertCount = 0; + QShaderDescription::TessellationMode tessMode = QShaderDescription::UnknownTessellationMode; + QShaderDescription::TessellationWindingOrder tessWind = QShaderDescription::UnknownTessellationWindingOrder; + QShaderDescription::TessellationPartitioning tessPart = QShaderDescription::UnknownTessellationPartitioning; }; -Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::ImageFlags) -Q_DECLARE_OPERATORS_FOR_FLAGS(QShaderDescription::QualifierFlags) - -#ifndef QT_NO_DEBUG_STREAM -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::InOutVariable &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BlockVariable &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::UniformBlock &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::PushConstantBlock &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::StorageBlock &); -Q_GUI_EXPORT QDebug operator<<(QDebug, const QShaderDescription::BuiltinVariable &); -#endif - -Q_GUI_EXPORT bool operator==(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept; -Q_GUI_EXPORT bool operator==(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept; - -inline bool operator!=(const QShaderDescription &lhs, const QShaderDescription &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::InOutVariable &lhs, const QShaderDescription::InOutVariable &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::BlockVariable &lhs, const QShaderDescription::BlockVariable &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::UniformBlock &lhs, const QShaderDescription::UniformBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::PushConstantBlock &lhs, const QShaderDescription::PushConstantBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::StorageBlock &lhs, const QShaderDescription::StorageBlock &rhs) noexcept -{ - return !(lhs == rhs); -} - -inline bool operator!=(const QShaderDescription::BuiltinVariable &lhs, const QShaderDescription::BuiltinVariable &rhs) noexcept -{ - return !(lhs == rhs); -} - QT_END_NAMESPACE #endif diff --git a/src/gui/rhi/qshaderdescription_p_p.h b/src/gui/rhi/qshaderdescription_p_p.h deleted file mode 100644 index c22ddda8ad..0000000000 --- a/src/gui/rhi/qshaderdescription_p_p.h +++ /dev/null @@ -1,81 +0,0 @@ -// Copyright (C) 2019 The Qt Company Ltd. -// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only - -#ifndef QSHADERDESCRIPTION_P_H -#define QSHADERDESCRIPTION_P_H - -// -// W A R N I N G -// ------------- -// -// This file is not part of the Qt API. It exists for the convenience -// of a number of Qt sources files. This header file may change from -// version to version without notice, or even be removed. -// -// We mean it. -// - -#include "qshaderdescription_p.h" -#include <QtCore/QList> -#include <QtCore/QAtomicInt> -#include <QtCore/QJsonDocument> - -QT_BEGIN_NAMESPACE - -struct Q_GUI_EXPORT QShaderDescriptionPrivate -{ - QShaderDescriptionPrivate() - : ref(1) - { - } - - QShaderDescriptionPrivate(const QShaderDescriptionPrivate &other) - : ref(1), - inVars(other.inVars), - outVars(other.outVars), - uniformBlocks(other.uniformBlocks), - pushConstantBlocks(other.pushConstantBlocks), - storageBlocks(other.storageBlocks), - combinedImageSamplers(other.combinedImageSamplers), - separateImages(other.separateImages), - separateSamplers(other.separateSamplers), - storageImages(other.storageImages), - inBuiltins(other.inBuiltins), - outBuiltins(other.outBuiltins), - localSize(other.localSize), - tessOutVertCount(other.tessOutVertCount), - tessMode(other.tessMode), - tessWind(other.tessWind), - tessPart(other.tessPart) - { - } - - static QShaderDescriptionPrivate *get(QShaderDescription *desc) { return desc->d; } - static const QShaderDescriptionPrivate *get(const QShaderDescription *desc) { return desc->d; } - - QJsonDocument makeDoc(); - void writeToStream(QDataStream *stream, int version); - void loadFromStream(QDataStream *stream, int version); - - QAtomicInt ref; - QList<QShaderDescription::InOutVariable> inVars; - QList<QShaderDescription::InOutVariable> outVars; - QList<QShaderDescription::UniformBlock> uniformBlocks; - QList<QShaderDescription::PushConstantBlock> pushConstantBlocks; - QList<QShaderDescription::StorageBlock> storageBlocks; - QList<QShaderDescription::InOutVariable> combinedImageSamplers; - QList<QShaderDescription::InOutVariable> separateImages; - QList<QShaderDescription::InOutVariable> separateSamplers; - QList<QShaderDescription::InOutVariable> storageImages; - QList<QShaderDescription::BuiltinVariable> inBuiltins; - QList<QShaderDescription::BuiltinVariable> outBuiltins; - std::array<uint, 3> localSize = {}; - uint tessOutVertCount = 0; - QShaderDescription::TessellationMode tessMode = QShaderDescription::UnknownTessellationMode; - QShaderDescription::TessellationWindingOrder tessWind = QShaderDescription::UnknownTessellationWindingOrder; - QShaderDescription::TessellationPartitioning tessPart = QShaderDescription::UnknownTessellationPartitioning; -}; - -QT_END_NAMESPACE - -#endif |