-
Show log
Commit
-
Hash : 130597e1
Author :
Date : 2020-11-23T18:59:02
Vulkan: Use PackedScissor struct to reduce GraphicsPipelineDesc size We are running out of space in GraphicsPipelineDesc. According to gpuinfo, the max reported viewport size is 32768. This CL changes scissor's x/y/w/h from uint32_t to uint16_t, which saves two uint32_t for other usages. Bug: b/173800146 Change-Id: Icf2d8ba8ea8a8c412ecef2059401a8d831c410e4 Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/2557218 Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org> Reviewed-by: Tim Van Patten <timvp@google.com> Reviewed-by: Jamie Madill <jmadill@chromium.org>
-
Files
- BUILD.gn
- BufferVk.cpp
- BufferVk.h
- CommandProcessor.cpp
- CommandProcessor.h
- CompilerVk.cpp
- CompilerVk.h
- ContextVk.cpp
- ContextVk.h
- DebugAnnotatorVk.cpp
- DebugAnnotatorVk.h
- DeviceVk.cpp
- DeviceVk.h
- DisplayVk.cpp
- DisplayVk.h
- DisplayVk_api.h
- FenceNVVk.cpp
- FenceNVVk.h
- FramebufferVk.cpp
- FramebufferVk.h
- GlslangWrapperVk.cpp
- GlslangWrapperVk.h
- ImageVk.cpp
- ImageVk.h
- MemoryObjectVk.cpp
- MemoryObjectVk.h
- OWNERS
- OverlayVk.cpp
- OverlayVk.h
- PersistentCommandPool.cpp
- PersistentCommandPool.h
- ProgramExecutableVk.cpp
- ProgramExecutableVk.h
- ProgramPipelineVk.cpp
- ProgramPipelineVk.h
- ProgramVk.cpp
- ProgramVk.h
- QueryVk.cpp
- QueryVk.h
- README.md
- RenderTargetVk.cpp
- RenderTargetVk.h
- RenderbufferVk.cpp
- RenderbufferVk.h
- RendererVk.cpp
- RendererVk.h
- ResourceVk.cpp
- ResourceVk.h
- SamplerVk.cpp
- SamplerVk.h
- SecondaryCommandBuffer.cpp
- SecondaryCommandBuffer.h
- SemaphoreVk.cpp
- SemaphoreVk.h
- ShaderVk.cpp
- ShaderVk.h
- SurfaceVk.cpp
- SurfaceVk.h
- SyncVk.cpp
- SyncVk.h
- TextureVk.cpp
- TextureVk.h
- TransformFeedbackVk.cpp
- TransformFeedbackVk.h
- UtilsVk.cpp
- UtilsVk.h
- VertexArrayVk.cpp
- VertexArrayVk.h
- android/
- display/
- doc/
- fuchsia/
- gen_vk_format_table.py
- gen_vk_internal_shaders.py
- gen_vk_mandatory_format_support_table.py
- ggp/
- headless/
- mac/
- shaders/
- vk_cache_utils.cpp
- vk_cache_utils.h
- vk_caps_utils.cpp
- vk_caps_utils.h
- vk_format_map.json
- vk_format_table_autogen.cpp
- vk_format_utils.cpp
- vk_format_utils.h
- vk_helpers.cpp
- vk_helpers.h
- vk_internal_shaders_autogen.cpp
- vk_internal_shaders_autogen.gni
- vk_internal_shaders_autogen.h
- vk_mandatory_format_support_data.json
- vk_mandatory_format_support_table_autogen.cpp
- vk_mem_alloc_wrapper.cpp
- vk_mem_alloc_wrapper.h
- vk_utils.cpp
- vk_utils.h
- vk_wrapper.h
- win32/
- xcb/
-
README.md
-
ANGLE: Vulkan Back-end
ANGLE’s Vulkan back-end implementation lives in this folder.
Vulkan is an explicit graphics API. It has a lot in common with other explicit APIs such as Microsoft’s D3D12 and Apple’s Metal. Compared to APIs like OpenGL or D3D11 explicit APIs can offer a number of significant benefits:
- Lower API call CPU overhead.
- A smaller API surface with more direct hardware control.
- Better support for multi-core programming.
- Vulkan in particular has open-source tooling and tests.
Back-end Design
The
RendererVkclass represents anEGLDisplay.RendererVkowns shared global resources like the VkDevice, VkQueue, the Vulkan format tables and internal Vulkan shaders. The ContextVk class implements the back-end of a front-end OpenGL Context. ContextVk processes state changes and handles action commands likeglDrawArraysandglDrawElements.Command recording
The back-end records commands into command buffers via the following
ContextVkAPIs:-
beginNewRenderPass: Writes out (aka flushes) prior pending commands into a primary command buffer, then starts a new render pass. Returns a secondary command buffer inside a render pass instance. -
getOutsideRenderPassCommandBuffer: May flush prior command buffers and close the render pass if necessary, in addition to issuing the appropriate barriers. Returns a secondary command buffer outside a render pass instance. -
getStartedRenderPassCommands: Returns a reference to the currently open render pass’ commands buffer.
The back-end (mostly) records Image and Buffer barriers through additional
CommandBufferAccessAPIs, the result of which is passed togetOutsideRenderPassCommandBuffer. Note that the barriers are not actually recorded untilgetOutsideRenderPassCommandBufferis called:-
onBufferTransferReadandonBufferComputeShaderReadaccumulateVkBufferread barriers. -
onBufferTransferWriteandonBufferComputeShaderWriteaccumulateVkBufferwrite barriers. -
onBuffferSelfCopyis a special case forVkBufferself copies. It behaves the same as write. -
onImageTransferReadandonImageComputerShadeReadaccumulateVkImageread barriers. -
onImageTransferWriteandonImageComputerShadeWriteaccumulateVkImagewrite barriers. -
onImageRenderPassReadandonImageRenderPassWriteaccumulateVkImagebarriers inside a started RenderPass.
After the back-end records commands to the primary buffer and we flush (e.g. on swap) or when we call
ContextVk::finishToSerial, ANGLE submits the primary command buffer to aVkQueue.See the code for more details.
Simple command recording example
In this example we’ll be recording a buffer copy command:
// Ensure that ANGLE sets proper read and write barriers for the Buffers. vk::CommandBufferAccess access; access.onBufferTransferWrite(destBuffer); access.onBufferTransferRead(srcBuffer); // Get a pointer to a secondary command buffer for command recording. vk::CommandBuffer *commandBuffer; ANGLE_TRY(contextVk->getOutsideRenderPassCommandBuffer(access, &commandBuffer)); // Record the copy command into the secondary buffer. We're done! commandBuffer->copyBuffer(srcBuffer->getBuffer(), destBuffer->getBuffer(), copyCount, copies);Additional Reading
More implementation details can be found in the
docdirectory: