Fences are no longer part of the WebGPU spec, and have been removed from
Blink.
Bug: dawn:22
Change-Id: I240c4c4107acfaf9facec88a43a38b5ff327c7a6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/50702
Commit-Queue: Brandon Jones <bajones@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Below are the list of all the individual changes, which are a good order
in which to review this CL.
Core changes:
- Change the tracking in the frontend to produce a synchronization
scope per dispatch instead of per compute pass. Some bindgroups might
not be part of any synchronization scopes so we also track all the
referenced resources on the side so they can be checked during
Queue::Submit validation.
- Fix clearing in the GL and Metal backends to use the per-dispatch
synchronization scopes.
- Fix the Vulkan backend to use the per dispatch synchronization scopes
to produce the correct pipeline barriers. This allows the removal of
previous logic that was subtly incorrect for Indirect buffer. This
allows the merging of the Compute and Render DescriptorSetTracker into
a single small helper class.
- D3D12 changes are similar to Vulkan, but the simplification is just a
the suppression of a branch with a lot of code in
BindGroupStateTracker.
Test changes:
- Fixup all the ResourceUsageTracking tests to follow the WebGPU spec
for synchronization scopes (fixing a lot of TODOs).
- Add additional tests checking that Indirect buffers are not allowed
to be used as a writeable storage in the same synchronization scope.
- Add tests for Queue::Submit validation correctly taking into account
resources that are bound but unused in compute passes.
- Add an end2end test for using a buffer as Indirect and Storage at the
same time in a DispatchIndirect, which would previously produce
incorrect barriers in the Vulkan and D3D12 backends.
Other small changes (that I was to lazy to put in a different CL):
- Add the utils::MakePipelineLayout helper function.
- Fix Indirect not being in the list of readonly buffer usages (caught
by a test added in this CL).
Bug: dawn:632
Change-Id: I77263c3535a4ba995faccbf26255da9a2f6ed3b5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49887
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
This change is a preparation for making the compute pass track the
synchronization scope usages per dispatch instead of for the whole pass.
- Split the tracking of render and compute passes usages.
- Rename PassTextureUsage to TextureSubresourceUsage since is it not
per-pass
- Add SyncScopeResourceUsage as a subclass of PassResourceUsage to
start modifying some of the code to work with synchronization scopes
(even if syncscope == pass at the moment).
There are no functional changes.
Bug: dawn:632
Change-Id: Ieeb6d70a44dc1c726f26989eebcd87e63e732785
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/49883
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This is the first patch to implement 3D texture copy. It starts with
implementation for 3D texture copy on D3D12 backend with the simplest
case: copy to the entire 3D texture. And texture's width is aligned
with 256 bytes.
The implementation for 3d texture copy might be inaccurate/incorrect
in some functions for complicated cases. But don't panic. The previous
implementation is also incorrect because many functions assumes that
we are copying to/from 2D textures only. And I will incrementally fix
the incorrect functions via upcoming tests for 3d texture copy.
BUG: dawn:547
Change-Id: I588b09fc8d0f0398e0798573415ba3a6a3f576fc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/45980
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This CL propagates errors from reentrant WriteBuffer calls and makes
procy methods on BufferBase to show that it is safe to call
GetMappedRange and Unmap without handling errors.
Bug: dawn:723
Change-Id: I4ea43adc4844505314bf84e2357b2d928f1d1f8c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/46003
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
An async callback which calls Queue::Submit will cause
reentrance in QueueBase::Tick and CreatePipelineAsyncTracker::Tick,
which invalidates the task queue being used by the original call,
and leads to a crash from an invalid pointer.
The Tick functions should remove the tasks from the queues before
the callbacks are called, so invalidation doesn't cause a crash.
Bug: dawn:729
Change-Id: I0d952d51040a3d1a475767400de3333a8b9b0821
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/45900
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This means that calling wgpu::Object::DoStuff will translate to a call
to dawn_native::ObjectBase::APIDoStuff. This will clarify the
difference between reentrant calls and internal calls in dawn_native.
Avoiding issues in the future.
This CL only changes the code generator to prefix with "API", performs
renames needed to make the code compile, and adds TODOs for things that
should be fixed in follow-up CLs.
Bug: dawn:723
Change-Id: Ie24471fa093adc4179d33d13323429847d076ecb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/45921
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Stephen White <senorblanco@chromium.org>
Still leave deprecated `depth` functional as there are some references in
other clients. Using `depth` and `depthOrArrayLayers` at the same time is
invalid. Add DeprecatedAPITests.
Bug: chromium:1176969
Change-Id: Ia06645e4f3c17588323dd36b11f9f3988b2e3aba
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/44640
Commit-Queue: Shrek Shao <shrekshao@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This is to follow the renames in the upstream WebGPU specification.
Typedef are left in places to make a smooth deprecation period.
Bug: dawn:22
Change-Id: I5134b897930c1fa883c49dd80d2665d6684ec022
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/43882
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Brandon Jones <bajones@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
The are replaced by Queue::OnSubmittedWorkDone. Only Queue::CreateFence
is deprecated since all other fence-related calls require a fence to
work.
Also ports a number of uses of fences in test harness to use
Queue::OnSubmittedWorkDone instead.
Bug: chromium:1177476
Change-Id: I479415f72b08158a3484013e00db8facd11e6f33
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/42660
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Brandon Jones <bajones@chromium.org>
This is the replacement for Fence in the single-queue WebGPU world. To
keep this CL focused, it doesn't deprecate the fences yet.
Bug: chromium:1177476
Change-Id: I09d60732ec67bc1deb49f7a9d57699c049475acf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41723
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
In upstream WebGPU, error scopes do not wait for queue operations or
async operations like create*PipelineAsync or mapAsync. This simplifies
the implementation so we don't need to track error scopes by parent
pointers but can instead have a simple stack.
Bug: dawn:22, chromium:1177107
Change-Id: Ic7344cbd96e257cbabc0f414934a5e42a4020a13
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/41980
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This is more in line with what happens in dawn_wire and Blink's WebGPU
implementation. It also allows fixing the Fence-related DeviceLost tests
to destroy the mock fence callback on destruction, which in turns fixes
a crash on dawn_end2end_tests exit on MSVC x64 debug.
Bug: dawn:602
Change-Id: I277e7fa284a573854ed46576602d5f6819db1357
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/38526
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
This CL enable CopyTextureForBrowser to accept options. The first
supported option is flipY, which can be implemented through scale and
offset uniforms.
BUG=dawn:465
Change-Id: Ia90153ee63a50e0e40beb1c13c63764d19a0b809
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/34402
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This makes a nearly one-to-one mapping between the JS and C APIs, which
benefits projects like Blink and Emscripten.
- JavaScript's `undefined` is equivalent to C `WGPU_STRIDE_UNDEFINED`.
- JavaScript's `0` is equivalent to C `0`.
- To implement the API correctly, Blink must special-case an actual
value coming in from JS that is equal to WGPU_STRIDE_UNDEFINED
(0xFFFF'FFFF), and inject an error.
Keeps but deprecates a reasonable approximation of the old behavior.
Bug: dawn:520
Change-Id: Ie9c992ffab82830090d0dfc3120731e89cd9691c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/31140
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Like copyTextureCHROMIUM in Chromium, CopyImageBitmapToTexture can use
internal pipeline to do the GPU uploading. But Dawn doesn't support
internal pipeline now.
This patch adds the first internal pipeline for Dawn and add the API
CopyTextureForBrowser to use internal pipeline to do gpu uploading.
The patch integrates very simple wgsl vertex/fragment shaders to do
simple direct blit to verify the whole system works.
BUG=dawn:465
Change-Id: I8b566af38a10eea00f7426c39e752958ef057abd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30960
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
With some refactoring of the relevant validation code.
Bug: dawn:520
Change-Id: Iedda0f7b1b67c20d3a88f2c4183dcc8eeae2096f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30742
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Changes Device::Tick to return a boolean that denotes whether or not
Tick needs to be called again.
Bug: dawn:119
Change-Id: I9d4c7e291536d676b33fc61d652667c1fbff8c62
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29980
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
In follow up CLs additional will be added to the AspectInfo, like the
supported component types.
Also simplify the logic for GetTexelInfo since all aspects are the first
aspects, except stencil which is always stencil8.
Bug: dawn:517
Change-Id: Iebbcb8a7f8fa2c4b7b06f65d6e4e8917c0a85366
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/30100
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Use QueueBase to track fences in flight and map requests so that they
can be resolved in the order they were added. Before these tasks were
separately tracked in FenceSignalTracker and MapRequestTracker, so tasks
would be resolving out of order.
Bug: dawn:404
Change-Id: I8b58fb72c99f43bc4593f56e08920d48ac506157
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/29441
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
This will prevent mixing it up with other serial types in the future.
Bug: dawn:442
Change-Id: I1cdb35ee01be3c771183003bc357c84d46aa5745
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28922
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
A lot of our switches over enum values use the following pattern:
default:
UNREACHABLE();
return foo;
This is problematic because when adding a new value to one of the WebGPU
enums, there is no compilation error for switches that are missing it.
Currently we're supposed to write code and tests and fix UNREACHABLEs when
we see them.
Instead we should strive to have most switches on enums to be complete
and explicitily tag unreachable values as UNREACHABLE. Some switches
might still want to use default: UNREACHABLE() if only a couple values
need to be handled out of very many.
In this CL we go through all the UNRAECHABLEs and change them if need
be. Also an ErrorQueue class is added to avoid having
QueueBase::SubmitImpl just be UNREACHABLE (and force overriding
instead).
Bug: dawn:527
Change-Id: I33dfb4703104912cc5f001f9faf907a61324de68
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/28501
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Depth/stencil copies of partial subresources is disallowed in
WebGPU because this is a D3D12 restriction. This restriction
need also to be enforced on B2T, T2B and WriteTexture.
This CL also fixes the subresource whole size calucation to use
the mip level. Previously, the 0th level size was always used.
This CL updates the validation to be correct and adds tests.
The DepthStencilCopy tests are factored into smaller helpers to
reduce code duplication.
Bug: dawn:439
Change-Id: I45d4836f6be1707c5171bddef875e535e935f7f4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26660
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
More code is now shared across backends.
Bug: dawn:483
Change-Id: I7ca1b8cbc2f12e408c94fbe5bca9fd29e47e0004
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27021
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
When using a dynamic uploader we didn't align the offset
that the allocated memory might have already had.
That fixes WriteTexture, WriteBuffer, ClearTexture and
on D3D12 ClearBuffer.
Bug: dawn:512
Change-Id: I64c7511ad6b0d3d6a28a494e1324a10ad4d38091
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/27020
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Command buffers hold references to all encoded objects. Freeing them
eagerly significantly reduces the amount memory held before the JS GC
clears the command buffers.
Bug: dawn:262, dawn:372
Change-Id: I68dfa973f980fba8d94611ed1de3c593bdb91a63
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26562
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Commit-Queue: Austin Eng <enga@chromium.org>
That was missing. There's a new validation test for that and two
tests for submitting copy commands with destroyed buffer/texture
(I think only a mapped buffer was covered so far).
Also fixing some error state tests.
Bug: dawn:483
Change-Id: I691f34722e96866a06465b4b3b0cae9d31c08a84
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26161
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Chromium always sends arrayLayer instead of origin.depth since
the migration hasn't been finished yet. This wasn't caught
in Dawn's testing since we are using origin.depth everywhere.
Bug: dawn:483
Change-Id: I13b1ccfb016eea01a3291ca439457db09966e9a3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/26160
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Tomek Ponitka <tommek@google.com>
Changed the order of validation blocks in ValidateLinearTextureData.
It doesn't match the order in spec now, but for the algorithm to
compute required bytes in copy we need the conditions which come
after the ones relating to it. Also switched the order of
ValidateLinearTextureData and ValidateTextureCopyRange for similar
reasons.
Bug: dawn:483
Change-Id: If00ae769d170ea12494258721916ec55d79e2880
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/25041
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Added implementation of writeTexture in Metal. It's using a
staging buffer instead of writing directly from the CPU to
the texture, because Dawn uses the private storage mode for
most of the Metal textures.
Bug: dawn:483
Change-Id: I6b85ee8bbe343881337bdb203a122dc1f1523177
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24581
Commit-Queue: Tomek Ponitka <tommek@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Added Queue::WriteTexture with validation but no actual
implementation. Tests were mostly taken from validation tests
for copying buffer to texture. Validation tests for CopyB2T
and WriteTexture do not cover 2d-array textures yet.
Bug: dawn:483
Change-Id: I9027eb615c02fe2265cde912f6ba17a235b94728
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/24440
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Tomek Ponitka <tommek@google.com>
Zero-sized copies are invalid in a couple backends, and in follow up
CLs CreateBufferMapped will be change to handle zero-sized buffers
correctly.
Bug: chromium:1069076
Change-Id: Ieef62a13182bbe1e939a3847980c91339e42aa8f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/22460
Reviewed-by: Stephen White <senorblanco@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Handle DeviceLostCallback once DeviceLost error occurs.
Disallow any other commands or actions on device to happen after device
has been lost.
Bug: dawn:68
Change-Id: Icbbbadf278cae5e6213050d00439118789c863dc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/12801
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This patch factors resource usage tracking so it is separate from
command validation, allowing the bulk of command validation to be
completely skipped.
In DrawCallPerfRun/Vulkan_DynamicPipeline_DynamicBindGroup, disabling
validation cuts roughly 74 nanoseconds (20%) of CPU time, per draw.
In DrawCallPerfRun/Vulkan_DynamicBindGroup, disabling validation
cuts roughly 35 nanoseconds (17%) of CPU time, per draw.
In DrawCallPerfRun/Vulkan_MultipleBindGroups, disabling validation
cuts roughly 45 nanoseconds (14%) of CPU time, per draw.
Bug: dawn:271
Change-Id: I517b85840ba18c6a554b83f34a1d0aef1a8c56a6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/13520
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Brandon Jones
Brandon Jones
Corentin Wallez
Yunchao He
Aleksi Sapon
shrekshao
Austin Eng
Stephen White
Yan, Shaobo
Kai Ninomiya
Natasha Lee
Tomek Ponitka
Enrico Galli
Hao Li