Fixes T2T depth copies on Mac Intel where the destination
subresource is a non-zero mip/layer.
Fixed: dawn:1083
Change-Id: If344b46c3fd436d857906850fc0ac5ccb4b93e1d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117592
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
This workaround is needed unless the system is MacOS 13.1+
with architecture at least AMD gcn-4.
Bug: dawn:1389
Change-Id: I3a85be150a0c5ca8b0b74e4b097906f278dd3049
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/118023
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Previously we were checking against a list of skipped messages using
both the pMessageIdName and pMessage from a Vulkan debug info callback.
The spec, however, indicates that the pMessageIdName may be NULL.
This change checks to see if that value is NULL before doing any further
checks, and if so always indicates that the message should not be
skipped. (No other code in Dawn references the pMessageIdName.)
See the appropriate page of the Vulkan spec for details:
https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkDebugUtilsMessengerCallbackDataEXT.html
Change-Id: Idd8bf312db31d6cea8e6ce42a47254f182b1070e
Bug: chromium:1411047
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117594
Kokoro: Kai Ninomiya <kainino@chromium.org>
Kokoro: Brandon Jones <bajones@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
When rotation metadata is added to videos by ffmpeg, it is specified as counter-clockwise. Dawn should follow this format to avoid confusion during integration.
Bug: chromium:1316671
Change-Id: I99ff30bffb1664aafd060d9a5bb1b15845388386
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117912
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Brandon1 Jones <brandon1.jones@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Works around issues on Metal Intel where CopyB2T and WriteTexture
with depth/stencil textures do not work correctly.
Fixes test failures with depth16unorm in the CTS.
Deletes UseTempTextureInStencilTextureToBufferCopy in favor of
the stencil blit. The former supposedly fixes a problem where
the stencil data is not flushed into the real stencil texture
by performing another T2T copy. This only works because the Metal
Intel backend also happens to allocate s8 as d32s8. Copying
the depth aspect as well seems to make the driver remember to
flush the data.
The stencil blit is a better fix for the problem since entirely
avoids getting the driver into a bad state where the stencil data
is not in sync.
Fixed: dawn:1389
Change-Id: If34b1d58996157036c164a5bc329e38b5e53f67a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117910
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
- It doens't know which operator == to use when there is using
EnumClassBitmask that looks like (a & b) == 0. Instead use just
the form (a & b).
- It doesn't do automatic capture of constexpr variables in lambdas so
turn a couple constexpr into regular const.
- It (correctly) warns that if constexpr (std::is_constant_evaluated())
is always true, so remove the constexpr keyword.
Bug: None
Change-Id: If7857abd1c30acb0736557844ff13f32a19d54cf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117888
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Corentin Wallez <cwallez@chromium.org>
This CL deprecates DawnTogglesDeviceDescriptor and replace it with
DawnTogglesDescriptor, which will be also used to provide instance
and adapter toggles with future CLs.
Bug: dawn:1495
Change-Id: I5f5cfdf38dcbcd237727a7f921462597b9301f1f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117174
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Kokoro: Zhaoming Jiang <zhaoming.jiang@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
A bad assert was causing devices that only supported Vulkan 1.0, such as
the Nexus 5X, to crash when querying information about adapter
properties and features. This change removes the need for the ASSERT.
(Note that this does not guarantee WebGPU support on Vulkan 1.0 devices.
For example, the Nexus 5X does not have the fragmentStoresAndAtomics
feature, which is required for WebGPU.)
Bug: dawn:286
Change-Id: I7f884290f29ae7e9607933f6186feac2ce8af540
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117840
Commit-Queue: Brandon Jones <bajones@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
So the next MapAsync() call, doesn't need to record pipeline barrier
command into the pending command buffer, submit it, and wait for it
completion.
Bug: b/265152896
Change-Id: Ied7cbca24a7b1bc81c0a6402179452599aabfe9b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116920
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
The fuzzer produced a shader with this format and hit an UNREACHABLE
in Dawn.
Bug: dawn:1641, chromium:1410048
Change-Id: Ia5ba66a18fc718f7e6dd803f4229ecb12b02753e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117531
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
This CL moves the ast/access files into type/access to resolve the type
dependency on ast
Change-Id: Ic0973c5ad162166736abb07cf9bbed7efa91695b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117605
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Adds a toggle to workaround another issue where Metal fails to set
the depth/stencil attachment correctly for a combined depth stencil
format if just one of the attachments is used. The workaround forces
both attachments to be set, giving the unused one LoadOp::Load and
StoreOp::Store so its contents are preserved.
Bug: dawn:1389
Change-Id: Iacbefcc57b33bf11ca8fcacb03506301646fe59d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117175
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
This CL fix FeatureState for some experimental features to prepare for
refactoring adapter creation with adapter toggles. This CL also fix the
related unittests.
Bug: dawn:1495
Change-Id: Ibf043ed74c0bfc79c64986f2f96135d92adf3930
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116841
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Auto-Submit: Zhaoming Jiang <zhaoming.jiang@intel.com>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
Browsers won't be exposing the ability to pass SPIR-V shaders, and the
ability to consume and validate them is adding a non-trivial amount to
the browser binary size on platforms like Android. To avoid that
overhead, this change puts those features behind a flag so that browser
usage can easily omit them.
Bug: dawn:286
Change-Id: Idf70683f2c4ccf479b723c00ba6914e27e4f765f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117241
Commit-Queue: Brandon Jones <bajones@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Add the last usage serial in Buffer, it is used for optimizing
MapAsync(), so the callback of MapAsync() is called when the last usage
serial is done instead of using the current pending serial when
the MapAsync() is called.
Bug: b/265151060
Change-Id: Ibc95d4e41d41896f0a49b0fd1068912b46ea14e1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116693
Commit-Queue: Peng Huang <penghuang@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This CL removes StagingBuffer to start to unify implementation
code paths for WriteTexture/Buffer and CopyBufferToTexture/Buffer.
This will help implementing a buffer-to-stencil copy workaround.
Bug: dawn:1389
Change-Id: Ieb23b8d871f14544ef01445a495dc1077274c9f2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117167
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
WebGPU clears [pending_map] before resolving or rejecting promises,
which means that the callback for .then or .catch is allowed to unmap,
remap, or do anything with the buffer assuming previous mapping
operations are entirely done the buffer.
Mimic this in Dawn by finishing all operations related to mapping
before calling callbacks they might trigger.
bug: dawn:1619
Change-Id: I9e5b82789a68f28a496a54c31bf9fe0ffde23ccf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116220
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Takahiro <hogehoge@gachapin.jp>
Reviewed-by: Austin Eng <enga@chromium.org>
Stencil to buffer copies don't capture contents
written in a rendering stage. Copying through an
intermediate texture fixes this problem on Metal Intel.
Add test for the code path that found the issue:
Nonzero-mip stencil copy, discard, then read
Bug: dawn:1389
Change-Id: I63d982df6bace4b5053d3c643b8abda1682490d1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116851
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Currently Queue::SubmitImpl() may cause two queue submissions, one is
because of Tick() call which will submit pending commands if there are.
The other one is after converting frontend commands to backend command
buffer. Queue::SubmitImpl() will submit converted commands. However
usually queue submissions are expensive, so merge those two queue
submissions into one by not calling Tick() before converting frontend
commands, so pending commands and recorded commands can be submitted
together. After that, we call Tick() to resolve callbacks and perform
bookkeeping operations (deallocating memory for passed operations,
etc).
Bug: b/265151060
Change-Id: Ia171771bcc1061dc599a58aa6d213a645696fb75
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116929
Auto-Submit: Peng Huang <penghuang@chromium.org>
Kokoro: Ben Clayton <bclayton@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Peng Huang <penghuang@chromium.org>
Do not transition if the buffer usage is already MapWrite
Change-Id: I16cf68501e19a5b60d6c6f28eaa539df691f14fd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117240
Reviewed-by: Austin Eng <enga@chromium.org>
Auto-Submit: Peng Huang <penghuang@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Peng Huang <penghuang@chromium.org>
This reverts commit afc3f9d4c7d91898f772962cc7105b8d43c99ff8.
Reason for revert: crbug.com/dawn/1615 should be fixed now that
vulkan-deps has rolled into Chromium.
Original change's description:
> Suppress vkUpdateDescriptorSets-None-03047
>
> Bug: dawn:1615
> Change-Id: I40f745fe7f7736c9c5ece34fc9fb114d478c9118
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116926
> Commit-Queue: Austin Eng <enga@chromium.org>
> Reviewed-by: Zhaoming Jiang <zhaoming.jiang@intel.com>
> Auto-Submit: Austin Eng <enga@chromium.org>
> Kokoro: Kokoro <noreply+kokoro@google.com>
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: dawn:1615
Change-Id: I5dbd7ae82a0c4f3b3b88ea70a9faac35507a0418
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/117320
Kokoro: Austin Eng <enga@google.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Auto-Submit: Austin Eng <enga@google.com>
Commit-Queue: Austin Eng <enga@google.com>
Stencil8 textures don't work correctly on some drivers. Workaround
by allocating a combined depth-stencil texture instead.
Suppress newly added test on Windows ANGLE
Bug: dawn:1389, dawn:1637
Change-Id: Iea03e8f3e5e7f663ffc7b344ab007a73836557cc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116846
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
CopyTextureForBrowser() handles flipY by flipping source texture then
applying copy op. But this is not correct. Instead
it should find the copy rect and apply flipY op in copy rect only,
which is the same as WebGPU GPUImageCopyExternalImage.flipY definition
(https://www.w3.org/TR/webgpu/#dom-gpuimagecopyexternalimage-flipy).
This CL fixed the issue and updated related end2end tests.
Bug: dawn:1635
Change-Id: I3ea7c9de44fb45224bc438486e0e92385446bfb1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116924
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This patch counts the line position, offset and size of the compilation
message in UTF-16 and saves them to WGPUCompilationMessage to align the
latest WebGPU SPEC.
Bug: dawn:1357
Change-Id: If8f4026bd5b4a64a078e100762b6d1f61da50053
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/115640
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
OpenGLES doesn't support 1D textures, so use 2D textures of width x 1
as a workaround (requires dependent Tint change).
Bug: dawn:1301
Change-Id: I99dbccfae497ee86d6f9b9e1ca1608049971016d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114820
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Unsure whether this error is correct, there's an issue open on VVL to
figure that out. Suppress until that's resolved.
Bug: dawn:1627
Change-Id: I9663c1434f6eda48385ff66751432d22b2d8a3d3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116544
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Auto-Submit: Kai Ninomiya <kainino@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
This should help unblock the roll of vulkan-deps into chromium, which is
currently failing dawn_end2end_tests and angle_end2end_tests.
Fixed: dawn:1626
Change-Id: I5c0313e93d0be01c794ce9c1ffe73f16c0dfc8df
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116138
Auto-Submit: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Start using Tint's ClampFragDepth transform in the Vulkan backend when
needed in order to correctly clamp @builtin(frag_depth) on Vulkan. Do
this by always reserving 8 bytes of push constant space to contain the
f32 min and max values from the last viewport command.
Reenables relevant CTS tests that were suppressed on Vulkan.
Bug: dawn:1125, dawn:1576, dawn:1616
Change-Id: I38f4f6c3c51c99b5e591a780fea9859537529534
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/105642
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
The toggle D3D12ForceClearCopyableDepthStencilTextureOnCreation can be
disabled when the toggle D3D12Allocate2DTexturewithCopyDstAsCommittedResource
is enabled because with D3D12Allocate2DTexturewithCopyDstAsCommittedResource
all the depth stencil textures (can only be 2D textures) with CopyDst usage
have already been created with CreateCommittedResource() instead of
CreatePlacedResource(), thus the driver issue about creating depth stencil
texture on a dirty heap with CreatePlacedResource() won't be triggered.
Bug: dawn:1487, chromium:1237175
Change-Id: I872d4d95e6e05e1bcf9489b31a72e61f957de3e5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116129
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This blocks the import into google3.
Change-Id: Ieb5db932553b5f9c8e44e1c334e497004c6bb778
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116361
Commit-Queue: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Removes the ExternalTextureVisibleRect type in favor of the existing
visibleOrigin/visibleRect parameters.
Bug: dawn:1622
Change-Id: Ifa661392b5541543c1445ce3e1e8e5e9db881be4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116124
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Brandon1 Jones <brandon1.jones@intel.com>
Do more math on the CPU to avoid per-fragment ALU operations.
Use a mat3x2 instead of mat2x3 to avoid padding.
Fixed: dawn:1614
Change-Id: Ib0e0f7d44ed9aa16eaca712f6553214fad141feb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116060
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Brandon1 Jones <brandon1.jones@intel.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Scream and shout if we ICE in a debug build.
Fixed: tint:597
Change-Id: Ie349d314a2ddf518f54df81469e4e14aa16cacda
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114540
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
The metal validation layer complains when copying from a 3D texture to
a buffer if the stride for the destination buffer is larger then 2048
bytes.
Bug: dawn:1430
Change-Id: I6ba4508d71610c35dfb0fab7d2bebc91d37504e3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/113426
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Gregg Tavares <gman@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Brandon Jones <bajones@chromium.org>
Previously this toggle was implemented only for the Metal backend, but
a need for it was identified on Android as well. This change moves the
implementation to the backend-agnostic command encoder recording so that
it works for all backends. Fundamentally it's still doing the same
thing, however: Swapping resolve targets that point at a non-zero mip
level or layer with a temporary texture and then performing a copy once
the render pass has ended.
Bug: dawn:1569
Change-Id: I292860cc74f653b2880e727d2ef3a7dfa3f10b91
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/106040
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
2D array texture may corrupt on some Intel devices, making out-of-bound
texture access and memory information leak. It's a critical security
issue. Intel driver team suggested the 24K extra memory approach in
order to mitigate the security issue before.
However, the texture corruption issue (and even the correctness issue)
can be worked around via allocating a few extra layers. And patches
have already been merged in Dawn, with a lot tests for verification. The
24K extra memory for each texture is actually incorrect and unnecessary.
So this patch removes relevant code in Dawn.
This patch mainly reverts some code of this patch below:
https://dawn-review.googlesource.com/c/dawn/+/96220
Bug: dawn:1507
Change-Id: Ic3239115ad4c74bdee928577ccbb20f1e35d13c3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114641
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
