This makes the Vulkan backend use the BuddyMemoryAllocator to
sub-allocate small resources inside a larger VkDeviceMemory object.
Right now the heuristic to decide to do suballocation is naive and
should be improved.
BUG=dawn:27
Change-Id: Idcc7b6686c086633c85328a7afb91ee84abf7b8c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12662
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This mostly makes the MemoryAllocator not owned by the
BuddyResourceAllocator so that we don't need an extra class for the
dependency injection in the Vulkan backend. (the container for the
BuddyMemoryAllocator can be it's MemoryAllocator at the same time).
Also renames methods of MemoryAllocator to be more explicit.
Also renames the constructor parameter of BuddyMemoryAllocator to be
(subjectively) closer to what the represent.
BUG=dawn:27
Change-Id: I37355ad5b3cded143956f0adc4742fa1b717e9bc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12661
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This removes the duplication of the memory allocators in preparation for
using sub-allocation in the Vulkan backend too.
Also renames ResourceMemory to ResourceHeap and MemoryResourceAllocator
to ResourceMemoryAllocator, and fixes a number of unused includes.
BUG=dawn:27
Change-Id: I1a9e7d41e5efafa5192bda1d89dc06455fa2af40
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12660
Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Updates TextureD3D to use the allocation handle and
defaults to using MSAA heaps.
BUG=dawn:27
Change-Id: I2318bb8d068df86364cb2ebc433f4737e9e121aa
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12580
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
This patch extends the BindGroupTracker in the D3D12 and Vulkan backends to
track bound storage buffers. We insert barriers between dispatches to properly
synchronize writes to storage buffers.
Bug: dawn:236
Change-Id: Iab3f964c345b64755557ab206e05a2ff7b0a3a1f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12301
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
webgpu.h is the "official" header for WebGPU in native and is being
developed in https://github.com/webgpu-native/webgpu-headers
dawn.h and dawncpp.h are changed to become webgpu.h and webgpu_cpp.h
respectively and use the new naming convention. New dawn.h and dawncpp.h
headers are created that just proxy the types, constants and functions
to their WebGPU counterpart.
Almost no naming change is done in Dawn in this commit, which help check
that the proxying headers work correctly. A couple changes were
necessary, in particular for tests of the internal of dawncpp.h, and a
workaround for a standard library bug for std::underlying_type was
removed because it is no longer needed and got in the way.
Finally since some templates were renamed to match the name of the file
they create instead of using the generic "api" name.
BUG=dawn:22
Change-Id: I12ee22d0b02ccb5b8a52ceccabb3e63ce74da007
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12480
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
11on12 has a bug where D3D12 resources used only for keyed
shared mutexes are not released until work is submitted to
the device context and flushed.
The most minimal work we can get away with is issuing a
TiledResourceBarrier.
ID3D11DeviceContext2 is available in Win8.1 and above.
This suffices for a D3D12 backend since both D3D12 and 11on12
first appeared in Windows 10.
Bug:dawn:217
Change-Id: I422eedf0de9c41777ab704f63520e2f2584f2afb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12520
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
Parameter syncInterval of function Preset specifies how to synchronize
presentation of a frame. To turn off vsync in D3D12 backend,
set syncInterval to 0 which represents the presentation occurs
immediately.
BUG=dawn:237
Change-Id: Ic17f00bae5af9fd6bca4130d6e2282f3c34de4e6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12303
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yizhou Jiang <yizhou.jiang@intel.com>
Dawn creates a keyed shared mutex for all wrapped resources and acquires
the mutex before the texture is used in an ExecuteCommandList call.
To coordinate with external clients, backend API has been extended
to allow clients to get and set the acquire key.
Pending and queue command lists have now been merged into one.
A future change will adjust GetPendingCommandContext to return a raw
command context without the need for error handling.
Bug:dawn:217
Change-Id: Ia96c449c305586407153f05ce75a40794b96027e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12220
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
An exported function of libdawn_native wasn't in libdawn_native_sources
instead of libdawn_native which made it not exported on Windows.
BUG=dawn:22
Change-Id: I824f1d4af18f8308b88e1650837cbe62374ace72
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12320
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Added D3D12Error to check HRESULTS and return error messages
with the correct error names.
Remove ASSERT_SUCCESS from D3D12 backend and use
MaybeError and ResultError instead to handle errors.
Bug: dawn:19
Change-Id: Idf2f1987725e7e658bd29a9b13653125ab43c564
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12000
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Do explicit uint8_t -> float conversions (as was previously done for TextureVk).
Add a missing #include.
Change-Id: If7616ec00e2ea2cf55b9b0cc966b0436a5a21d9d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12280
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Stephen White <senorblanco@chromium.org>
This commit add a toggle to turn off vsync in Dawn.
When turn off vsync, choose vulkan present mode VK_PRESENT_MODE_IMMEDIATE_KHR
if the mode is available on the system, but if the mode isn't supported,
choose default mode VK_PRESENT_MODE_FIFO_KHR.
BUG=dawn:237
Change-Id: If400262b67cc8051422745e3bed737431183c0b6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12100
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yizhou Jiang <yizhou.jiang@intel.com>
QueueWaitIdle can fail but if it fails, the error codes show that we
can't do anything about it so we just ignore the result.
BufferGetMemoryRequirements should always return memory types bits that
allow Dawn to select a good memory type of a resource.
BUG=dawn:19
Change-Id: I52beebf9f2be2bfb997ca9c1bf306618a73c9c4b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12183
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This patch adds an ASSERT expression on the object parameter of
DeviceBase::ValidateObject() so that Dawn won't crash in the debug
build when a nullptr object is being used unexpectedly.
BUG=dawn:209
Change-Id: I52ad57b58eb59edf04afc9963e260436d93e673e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12140
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This commit makes the following changes:
- Unused fences are reset when they will next be used so there is a
single place where error handling is needed, either for resetting
an existing fence, or for creating a new fence.
- All accesses to VkCommandBuffer are moved to using the
CommandRecordingContext and GetPendingCommandBuffer is removed.
- Instead of tracking both a current (VkCommandBuffer + Pool) and
a command recording context that contains the same VkCommandBuffer,
the RecordingContext now holds the pool too, as well as a tag to
know if it was ever queried (meaning it contains commands).
- mRecordingContext is now always valid, such that
GetRecordingContext() doesn't need to return an error.
BUG=dawn:19
Change-Id: I853d4ecdc6905b66e842688f39d863e362f59b66
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12022
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This change refactors D3D12 backend to have CommandRecordingContext
CommandRecordingContext allows us to, in a future, add additional
data to the context such as textures that need to be acquired and
released before command lists are executed.
The Device's pending command list and the command list which resides
in the Queue object were converted to use CommandRecordingContext.
Bug=dawn:234
Change-Id: Ic13a229fc1f15895ef71117ce638c942de224743
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11940
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
According to https://github.com/gpuweb/gpuweb/issues/405, None is a
valid value for GPUBindGroupLayoutBinding visibility to be passed in.
Bug: dawn:22
Change-Id: I7b30b7ab8ed6824718573fa25fad5d509846db55
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11980
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: François Beaufort <beaufort.francois@gmail.com>
Following WebGPU spec change at
https://github.com/gpuweb/gpuweb/pull/468, this CL changes all
occurrences of setVertexBuffers to setVertexBuffer.
Bug: dawn:22
Change-Id: I48b551a89dc0934dfa61e661e9546a2b7eafd2fc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12020
Commit-Queue: François Beaufort <beaufort.francois@gmail.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Allows buffer/texture direct access to underlying resource rather than indirectly with a opaque memory type.
BUG=dawn:27
Change-Id: I2eb69f4e30c96c431dbc96094d671be1e0a29869
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11800
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
This also introduces another combinator to ConsumeError for
ResultOrError.
BUG=dawn:19
Change-Id: Ic204313436f5e919473d604efd049fe3d3c27a66
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11862
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This makes all the regular WebGPU object creation handle errors properly
in the Vulkan backend instead of ASSERTing no Vulkan error is raised.
Static Create functions are added to all these types so that the details
of how the initialization is done is private, and it isn't possible to
construct an object but forget to initialize it.
BUG=dawn:19
Change-Id: I362b2d66b74dd7799ffbf69d732bc58caa97950b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11861
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This is necessary because @available(macOS 10.N, *) is taken on all iOS
versions but there is no way to say the branch is just not taken on any
iOS version. The proper way to deal with this is to add additional #if
guards to just not compile the code on iOS / macOS.
BUG=dawn:225
Change-Id: I76ec01f933364e9b47b5dda1198359f2ab4d1188
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11900
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
There was still some logic associated with push constants in Dawn. This
removes it to remove some code and reclaim one buffer in Metal. Related
code was found via `git grep "push constant"`.
BUG=
Change-Id: I17de9d47872483875b6fa292f8259ef1fc4ecaf3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11904
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Closing a command list can fail. We need to handle the error
gracefully instead of ignoring it.
As a fallout from adding MaybeError to ExecuteCommandList, need to
also add MaybeError to TickImpl, and OnBeforePresent.
Bug:dawn:19
Change-Id: I13685f3dd731f4ab49cbff4ce4edfa960d630464
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11841
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Rafael Cintron <rafael.cintron@microsoft.com>
If the swapchain creates its own queue, there can be races to determine
whether the submit of the present happens first. The didn't show up
in our samples but was an issue in real apps. Passing the queue to the
swapchain makes it simple to keep operations well ordered.
BUG=dawn:225
Change-Id: I9cc8e6e8140ad4a816373cffc9bda74ee826a41b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11640
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
HandleError assumes that the error is not NoError. InjectError is
an untrusted wire command and needs to validate that NoError is not
passed.
Bug: chromium:1010703, chromium:1004368
Change-Id: Id48a877ded327a6e3a54fedb2be2c54eeca5cd3c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11780
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This patch implements ASSERT_SUCCESS(hr) as a macro instead of a
function so that when hr fails Dawn can print out the line number of the
expression that causes the failure.
For example, previously we will always get the following error message
because ASSERT_SUCCESS is a function:
Assertion failure at ../../src/dawn_native/d3d12/DeviceD3D12.cpp:43
(ASSERT_SUCCESS): (((HRESULT)(hr)) >= 0)
Now we can get more details about where the failure occurs because now
ASSERT_SUCCESS is a macro:
Assertion failure at ../../src/dawn_native/d3d12/DeviceD3D12.cpp:59
(Initialize): (((HRESULT)(mD3d12Device->CreateCommandQueue(&queueDesc,
__uuidof(**(&mFence)), IID_PPV_ARGS_Helper(&mFence)))) >= 0)
BUG=dawn:178
Change-Id: I435ee2f418658bca276f439fcabfabbfecbff998
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11700
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This fixes a problem where client-generated errors weren't properly
captured in error scopes.
Bug: chromium:1004368
Change-Id: Ic0f6e5bc5e281c676ea7154dd217cfc8dd51be5c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11642
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This patch fixes an undefined behaviour on Metal and Vulkan when there
is a color state whose corresponding fragment output is not declared in
the fragment shader.
According to Vulkan SPEC (Chapter 14.3), the input values to blending or
color attachment writes are undefined for components which do not
correspond to a fragment shader output. Vulkan validation layer follows
the SPEC that it only allows the shader to not produce a matching output
if the writemask is 0, or it will report a warning when the application
is against this rule.
When no fragment output matches the color state in a render pipeline,
the output differs on different Metal devices. On some Metal devices the
fragment output will be (0, 0, 0, 0) even if it is not declared in the
shader, while on others there will be no fragment outputs and the content
in the color attachments is not changed.
This patch fixes this issue by setting the color write mask to 0 to
prevent the undefined values being written into the color attachments.
With this patch, the following end2end tests will not report warnings
any more when we enable the Vulkan validation layer:
ObjectCachingTest.RenderPipelineDeduplicationOnLayout/Vulkan
ObjectCachingTest.RenderPipelineDeduplicationOnVertexModule/Vulkan
ObjectCachingTest.RenderPipelineDeduplicationOnFragmentModule/Vulkan
BUG=dawn:209
TEST=dawn_end2end_tests
Change-Id: I5613daa1b9a45349ea1459fbdfe4a12d6149f0f7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11581
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch adds the validation on the compatibility between the format
of the color states and the fragment shader output when we create a
render pipeline state object as is required in Vulkan (Vulkan SPEC
Chapter 14.3 "Fragment Output Interface"):
"if the type of the values written by the fragment shader do not match
the format of the corresponding color attachment, the resulting values
are undefined for those components".
BUG=dawn:202
TEST=dawn_unittests
Change-Id: I3a72baa11999bd07c69050c42b094720ef4708b2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11461
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
When storeOp is clear, texture subresource is set as not initialized
Bug: dawn:145
Change-Id: I364d7239a7ebdb9d5a28a4af559f3212be7ef15a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11560
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This patch adds the validation on the fragment shader output location to
ensure it will never exceed kMaxColorAttachments.
BUG=dawn:202
TEST=dawn_unittests
Change-Id: I4ac4463fd3dfb3c2e9ffecb370f9d9d59393c26d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/11580
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Corentin Wallez
Bryan Bernhart
Austin Eng
Rafael Cintron
Yizhou Jiang
Natasha Lee
Stephen White
Jiawei Shao
François Beaufort
Mathieu-Andre Chiasson
Jinho Bang