This CL add Instance and Adapter toggle stage, and promote
DisallowUnsafeAPIs as an instance toggle, and can be required using
DawnTogglesDescriptor chained in instance descriptor when creating
instance. The instance's toggles state will get inherited to adapters
and devices it create. Related tests are implemented and updated.
Toggles inheritance can be overriden if not forced, so requiring
DisallowUnsafeAPIs when creating device is still available and working
like before.
Note that currently we don't have toggle of adapter stage, and can not
require toggles when creating adapter, until follow up CLs implement it.
Currently the toggles state of a adapter is simply inherited from
instance.
Bug: dawn:1495
Change-Id: I6bf7aa0f950a99451afcc2cab5322c924b7d9520
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122021
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch uses VK_QUEUE_FAMILY_IGNORED on buffer barriers to align
with what we are doing on image barriers.
Bug: dawn:1688
Change-Id: Icabae22270fc9c92a69ce18e0a63db49de9d7481
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122444
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
There are some scenarios where buffers can be used in pending
commands that are not retained by a command buffer.
They must be retained in the set of mappable buffers for eager
transition to prevent a use-after-free violation.
Fixed: chromium:1421170
Change-Id: I452d80b2513a7726a003d44e2a7850292d798bb1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122580
Auto-Submit: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Peng Huang <penghuang@chromium.org>
Currently in the middle of some functions, we execute callbacks
immediately such as inside Buffer::APIMapAsync(), Device::HandleError()
or Queue::Submit().
Firstly, this has risks. The functions might be in a middle of modifying
internal states. By triggering callbacks, users might call API
functions again which could further modify the internal states
unexpectedly or access the states in an inconsistent way.
Secondly, upcoming thread safe API which locks the public functions with
a mutex might encounter deadlock. Because callbacks might cause
re-entrances which would unexpectedly lock the public function again.
This CL attempts to limit number of functions that are allowed to
trigger callbacks. Other functions that want to trigger callbacks will
instead enqueue a request to execute callbacks in the next
Device::APITick() call.
Currently the functions that will be allowed to trigger callbacks are:
- Device::WillDropLastExternalRef()
- Device::APITick()
- Device::APISetLoggingCallback()
- Device::APISetUncapturedErrorCallback()
- Device::APISetDeviceLostCallback()
Bug: dawn:1672
Change-Id: Iabca00f1b6f8f69eb5e966ffaa43dda5ae20fa8b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120940
Commit-Queue: Quyen Le <lehoangquyen@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
- Note the duplication (in LIMITS_MAX_BUFFER_SIZE and LIMITS_OTHER)
caused tiering to be overridden.
- Also updated binding size test to align buffer sizes properly since
the test was failing locally for me.
Fixes: dawn:1683
Change-Id: I8d05f863ea9bf4dc8e620b7803bedb913af9f67b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122260
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Loko Kung <lokokung@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
There's no good reason for this to be public.
Move it into the writers, and expose a 'disable_robustness' option to
turn it off. This can be expanded to hold more fine-grain control in the
future.
Change-Id: I6ea6e54a27b2ae0fbcba5fdf45539063045cc15a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122203
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Ben Clayton <bclayton@google.com>
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This change makes it harder to misuse ConsumedError which can cause
device loss. When it is a known error, instead use HandleError to
bypass the "unlikely" if clause.
Bug: dawn:1336
Change-Id: I3052db343fe4080b257f1c2f9535f743a0e78526
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120384
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Note that we can just log these errors and avoid bubbling them up since
cache creation failure should not be fatal in Vulkan.
Bug: dawn:1336
Change-Id: Ie49d433f9b991508859f4969f2d4bf3b7c9e66d3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122024
Commit-Queue: Loko Kung <lokokung@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
This commit makes depthWriteEnabled and depthCompare required and
makes depthClearValue conditionally required for the spec change
in WebGPU V1.
https://github.com/gpuweb/gpuweb/pull/3849
depthClearValue is required if depthLoadOp is clear and the
attachment has a depth aspect. To simulate it, this commit lets
NAN represent unspecified depthClearValue and lets the default
value of depthClearValue be NAN.
Bug: dawn:1669
Change-Id: I469338e909b1d3c345bc2642ee47daee858909ca
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120620
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Up to 1024 threads in the X and Y dimensions. This is supported on all
Mac, all Windows, and most Vulkan devices.
Bug: dawn:685
Change-Id: I067a492bdfc32d6a186d6875e545238babca8784
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122003
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Shrek Shao <shrekshao@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
- Defaults consume error calls to only allow validation and device loss
errors.
- Allows OOM errors on Buffers, QuerySets, and Textures only.
- Adds initial suite of unit tests (and any necessary updates to mock
framework).
Bug: dawn:1336
Change-Id: I82112ea6c147e894280e605bf8ae0ce00488c9f3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119800
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
The higher tier for this limit is available on all D3D12, all Metal,
and most Vulkan devices.
Bug: dawn:685
Change-Id: Ic2a39ad7908ea178e7aac48b7bb54b262d7039cf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121543
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Shrek Shao <shrekshao@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Only load extension entry points if the extension is supported.
From the eglGetProcAddress manpage:
"A non-NULL return value does not guarantee that an extension function
is actually supported at runtime. The client must also make a
corresponding query, such as glGetString(GL_EXTENSIONS) for OpenGL and
OpenGL ES extensions [...] to determine if a function is supported by
EGL or a specific client API context."
This required moving extension initialization from OpenGLFunctions
into OpenGLFunctionsBase.
Change-Id: Ib4e8360ba455818701990b4476689b651d097ca8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121760
Commit-Queue: Stephen White <senorblanco@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This CL add a toggle-controlled Tint polyfill for reflect on vec2<f32>,
and enable this toggle by default on D3D12 Intel device when using FXC.
This CL works around issue tint:1798.
Bug: tint:1798
Change-Id: If2f4de836eaf5e7374bc2c1ae3fbe06b91a5bbd5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121160
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
The higher tier currently supports all D3D12 devices, all Metal,
and most Vulkan devices.
Bug: dawn:685
Change-Id: I5bcb778b92a073c9c1af943acee193073c0741ff
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121101
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Instead, compile pipelines synchronously.
When backend validation is on with Mac AMD, the driver crashes during
pipeline compilation. This seems to only occur when partition alloc
is enabled.
Bug: dawn:1200
Change-Id: I520087eeae9c356990962d0c4a96eba017735af1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121400
Commit-Queue: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
ExternalTexture has active, expired and destroyed states.
Only active state external texture is valid to submit.
Expired state external texture can be refresh to active but
destroyed external texture cannot be refresh.
Bug: chromium:1412338
Change-Id: Ic7f12d274d27b644f19ec3ef8b46c110610afa2b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120982
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
On Intel Gen9 and Gen9.5 GPUs there is a D3D12 driver bug about
using DstAlpha as the source blend factors in both color and alpha
blending.
Although we cannot add workaround for all such cases, we have
confirmed that when the blend operation is 'add', and the
destination blend factors are ‘zero', we can workaround this driver
bug by replacing 'add' to 'minus'. We cannot add workaround when
destination blend factor is another value.
This patch adds such workaround on the affected platforms.
Bug: dawn:1579
Test: dawn_end2end_tests
Change-Id: If7c82b9d559b876e42fb36e2f539f9ff2f71ea91
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120981
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Validate as the bind group lazy aspect at each dispatch/draw
call.
Use nested loops to iterate through each bind group and binding
to find if any aliasing exists, which has time complexity of
O(N^2) and can be further optimized to use O(NlogN) algorithm.
Bug: dawn:1642
Change-Id: I8c43128cdeea75352c194752fb22258b6a73430e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/118440
Commit-Queue: Shrek Shao <shrekshao@google.com>
Reviewed-by: Loko Kung <lokokung@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
- Plumbs all descriptors through so that the objects act much more like
a real implementation.
- Lots of changes to DestroyObjectTests.cpp to update usages and test
through the APIs when applicable.
- Put in defaults for the mock device and make it as easy to use as
possible.
Change-Id: I85b243a18ec1872aff0172549aec0f599967ea0e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119821
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Loko Kung <lokokung@google.com>
This follows the recent semantic from the WebGPU spec and fixes a couple
failing CTS tests.
Bug: None
Change-Id: I9197ea3fda6c15bd9302f8d60e70ee00f1d37708
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120980
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Query D3D12_FEATURE_DATA_D3D12_OPTIONS13 to determine whether the device
supports copies between textures of different dimensions. If it does
not, or if D3D12_FEATURE_D3D12_OPTIONS13 is not supported, force enable
toggle
D3D12UseTempBufferInTextureToTextureCopyBetweenDifferentDimensions.
Fixed: dawn:1216
Change-Id: I976e7fad291126f8dcee31ce6b681314d38e69e7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120882
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Auto-Submit: Austin Eng <enga@chromium.org>
This CL removes two deprecated fields from the inspector. The deprecated
alias to sem::SamplerTexturePair remains as it will require creating a
structure on the inspector side to replicate.
Change-Id: I6001fbd475260079dd49d50f8e8619241bb0b468
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120880
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
This patch replaces D3D12_BLEND_DEST_ALPHA with D3D12_BLEND_ONE
when the color target formats have no alpha-channel.
Using D3D12_BLEND_ONE is an optimization over
D3D12_BLEND_DEST_ALPHA as it means the GPU hardware doesn't
need to get destination pixel at all. As D3D SPED requires the
default value for missing components in an element format is
"0" for any component except A, which gets "1", using
D3D12_BLEND_DEST_ALPHA takes same effect with
D3D12_BLEND_ONE when the color target formats have no alpha
channel.
In addition, replacing D3D12_BLEND_DEST_ALPHA with
D3D12_BLEND_ONE also serves as a workaround against an Intel
driver issue about alpha blending.
Bug: dawn:1579
Change-Id: If79e4c8007b68dd35e142ba9cf8a4921e611890a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120120
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This CL moves the type::Access builtin to the builtin:: tree.
Change-Id: I3276d364f7b597671612a23c8823f0afd1914d81
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/120363
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Dawn provides external semaphore service in static way by compiling
implementation files based on OS type.
A better way is to select external semaphore in runtime.
This CL adds ServiceImplementation interface and implements it for
different semaphore type. Semaphore service delegate work to
implementation in runtime.
This CL also removes "DAWN_USE_SYNC_FD" macro.
Bug=dawn:1593
Change-Id: Idc3f7b644430aad76198ef66dc5ba13e8cfc5a4d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119321
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This was wrong.
DAWN_PLATFORM_IS_X86 is defined for both 32-bit and 64-bit
X86 builds.
Bug: dawn:1254
Change-Id: I308ac10749edcf76643539eb2801d019dd5439f9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119900
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch adds the support of the optional feature "bgra8unorm-storage"
on D3D12 when the platform allows using DXGI_FORMAT_B8G8R8A8_UNORM as
UAVs.
According to D3D12 documents, enabling the flag
D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS when creating textures requires
"either the texture format must support unordered access capabilities at
the current feature level. Or, when the format is a typeless format, a
format within the same typeless group must support unordered access
capabilities at the current feature level".
Bug: dawn:1641
Test: dawn_end2end_tests
Change-Id: If16271b5da52423e73ad4f7ba258af83dbf66dd2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119345
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
This patch implements the optional feature "bgra8unorm-storage" on
Vulkan. As SPIR-V doesn't support 'bgra8' as a valid image format,
we have to create an image view with RGBA8Unorm format on the
BGRA8Unorm texture when we want to use it as a storage texture.
Bug: dawn:1641
Test: dawn_end2end_tests
Change-Id: I4aeea96ae872fe4e6367c535afb6ab896b952453
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/118021
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
These textures still have some outstanding issues on both Intel and AMD,
where the wrong subresource can be read from.
Zero-initialization workarounds have been enabled to ensure that even
when the wrong subresource is read, uninitialized data is not read.
Bug: dawn:704, dawn:791, dawn:838
Change-Id: Ib118eff0c5ed5e7812fa70e88ece5317c1af13b0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/116849
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Adds a basic adapter blocklist and adds two cases to it.
By default, the blocklist is disabled until Chromium can enable it
explicitly - perhaps based on --enable-unsafe-webgpu or some other
flag. It will be switched to default to true in the future.
about://gpu would disable the blocklist so all the adapters are
visible there, but WebGPU would enable it. Trusted users of Dawn
that are aware of potential bugs may also disable it.
One downside is that about://gpu won't surface directly that an
adapter is on the system, but blocklisted for WebGPU. Something
like that can be added in the future, if necessary.
In the future, this should probably be merged with Chromium's
software_rendering_list.json, but that list doesn't support
multi-adapter systems well (for blocklisting just one adapter),
and it doesn't understand all the information we need for the
current blocklist.
Bug: dawn:1254, dawn:1196, tint:1753
Change-Id: I992bcd10dd5d3f5b23319fc4ec699b06bb1117da
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119061
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Still a missing condition to enable timestamp period calculation at
device initialization on Metal.
Refactor GPUTimestampCalibrationTests.cpp to check timestamp query
correctness on both D3D12 and Metal backends.
Bug: dawn:1193
Change-Id: I69feeaea0df309e15c008647d76b11899dcdc727
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119320
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch adds the support of the optional feature "bgra8unorm-storage"
and enable it on Metal.
Bug: dawn:1641
Test: dawn_end2end_tests
Change-Id: Id58cefd8735f46b8d1807376ebcfada10df2890e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/111380
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
To avoid uninitialized reads of depth stencil data, where the Metal
driver incorrectly binds/loads the wrong depth stencil subresource,
always keep all depth stencil subresources initialized. This means
that textures are initialized on creation, and StoreOp::Discard is
never used - Store is used instead.
Texture initialized state is still set as-if the Discard occured, so
Dawn will try to zero-initialize the subresource if it is read from.
In many cases, this will work correctly, and the application will
read back 0, as expected. In some cases, Metal will bind the wrong
subresource, and the previous contents will be read. This is wrong,
but at least it is not uninitialized data.
Bug: dawn:838
Change-Id: I3cc87073d52de60283e3b683bbee7809db803018
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/119344
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Shrek Shao <shrekshao@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
