Bug: dawn:628
Change-Id: Ib6a6be0e5467d3cdd6432834d6f9d5c2000b18be
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/76340
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Loko Kung <lokokung@google.com>
Specifically ensuring that Tint's generator is individually traced so
that we can better see how much time is going into Tint's transforms
vs the backend's compiler.
Change-Id: I9903cdca137d652ee400e94f0570eeeb17779207
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/76260
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
However there are a lot more things to implement for full support:
- Copies from/to buffers.
- Copies from/to other textures.
- WriteTexture
- Lazy initialization (if needed)
- Anything using views and 1D textures in shaders
So they are currently marked as unsafe API.
Bug: dawn:814
Change-Id: I3f1aac87bd5bc27f710d58e525938c1226d093d8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/64542
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Bug: dawn:160
Change-Id: Ifbce6f71fdf43a749c332bd691b63119929e0128
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75640
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
But keep a namespace alias to avoid breaking project that depend on the
previous namespace name while they get updated.
Done with through the following steps:
- git grep -l dawn_native:: | xargs sed -i "" "s/dawn_native::/dawn::native::/g"
- git grep -l "namespace dawn_native" | xargs sed -i "" "s/namespace dawn_native/namespace dawn::native/g"
- git cl format
- Manual fixups in generator/templates (and the addition of
namespace_case in dawn_json_generator.py).
- The addition of the namespace alias in DawnNative.h
Bug: dawn:824
Change-Id: I676cc4e3ced2e0e4bab32a0d66d7eaf9537e3f09
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75982
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
D3D keyed mutex cannot be acquired recursively and therefore external
D3D textures cannot be concurrently read within Dawn (e.g. multiple
imports of the same video frame) or across Dawn and GL (e.g. both WebGPU
and WebGL importing a video frame).
Within Dawn, we can scope keyed mutex acquire/release to command list
submission so that Chromium can always guarantee that Dawn doesn't hold
access to any resources after running Dawn wire commands. We also check
that keyed mutexes aren't recursively acquired in Dawn by keeping track
of the acquire count per resource.
This solves the multiple acquire problem within Dawn and provides a path
for concurrent read access across Dawn and GL once the GL decoders are
changed so that they also don't hold access to resources after switching
contexts.
Bug: chromium:1241533
Change-Id: If88fd4a4f798b972836a134809e4fed8832ec89c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75644
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Sunny Sachanandani <sunnyps@chromium.org>
Auto-Submit: Sunny Sachanandani <sunnyps@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
But keep a namespace alias to avoid breaking project that depend on the
previous namespace name while they get updated.
Some TraceEvent.h macro were using "platform" as an argument name so
that was renamed to "platformObj" to avoid conflicting with ::platform::
Bug: dawn:824
Change-Id: Iaf14853f02b0d9fcf866ce87788f87a3fdf8f364
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75541
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Chromium has switched to using a fixed key (0) for quite some time.
Using a fixed key enables future work needed for concurrent access of
external textures both inside Dawn (e.g. importing a video frame twice)
or across Dawn and GL. We want to try scoping the Acquire/Release calls
to command list submission, but with a non-fixed key concurrent access
for the same resource becomes ambiguous.
Bug: chromium:1241533
Change-Id: Ia8ff473b8c9c731c411a3fd59d69213f2d903e61
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75642
Auto-Submit: Sunny Sachanandani <sunnyps@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This patch adds StoreOp to the computation of key of RenderPassCache,
which was missed in the last CL that fixes a failure of end2end test
on Windows Intel Vulkan drivers.
BUG=dawn:1151
Change-Id: Iec03d70303c89906154db5256fd8691cc018a91a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75860
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This helps Chromium move to using this method and stop using
dawn_native::AdapterType/BackendType.
Bug: dawn:824
Change-Id: I9e16edd271d3406bc45e3b9fd472bd94bbf53b30
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75583
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Brandon Jones <bajones@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
Dawn allows texture-to-texture copy happens between the textures that
formats only have diff on srgb-ness.
CopyTextureForBrowser could align on this rule to achieve copying to
*-srgb dst texture and keep the bytes the same as copying to non-srgb
formats.
This CL add support for *-srgb textures as dst textures and using an
extra gamma decoding step for this.
Bug: dawn:1195
Change-Id: I665dbca473aa84b9d87b7a35c4f90ce1897ade7b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/74580
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This patch adds the missing comparisons on the storeOp when comparing
the key of RenderPassCache as now we support more than one storeOp
(Store and Discard).
With this patch the following end2end test will pass on Windows Intel
Vulkan drivers:
- TextureZeroInitTest.IndependentDepthStencilLoadAfterDiscard
BUG=dawn:1151
TEST=dawn_end2end_tests
Change-Id: I44d6f3e589341bba761503b1a06c388db92d1295
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75482
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This was done with these two commands and a couple manual fixups for
namespaces that had more than one space in the comment in the closing
brace, as well as vulkan_platform.h
git grep -l "namespace .* { namespace " | xargs sed -i "" "s/namespace \(.*\) { namespace /namespace \1::/"
git grep -l "}} // namespace" | xargs sed -i "" "s%}} // namespace%} // namespace%"
Bug: dawn:824
Change-Id: I6f448b820c12fc1004ea5270bf8e1f466b0c0aab
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75400
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
In previous T2T copy, Dawn requires textures have the same formats. But
Vulkan/Metal/D3D12 have ability to copy between "compatible" formats textures.
Metal has the most restrict rules without setting interpreter flags when creating
textures. It defines "compatible" texture formats to the formats that only have
difference on srgb-ness.
This CL follow Metal's rule and release the validations for T2T copies. It supports
T2T copy between "compatible" texture format textures.
Bug: dawn:1204
Change-Id: I50bf04ea15e8026530b3a5bdb5725f56aa192d85
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/74301
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
If an error message was reported from Tint that had a end line/position
that occurred before the start line/position then the range would have
a negative length, causing the unsigned length to underflow into a very
large value. Also, specifically, if the start line was non-zero but the
end line was zero (indicating no line, and the value ranges are
constructed with by default) the for loop that adds the line offsets
would underflow and cause it to read off the end of the line list.
Clamping the end of the range to always be less than or equal to the
start of the range avoids both of these problems.
Bug: dawn:1245
Change-Id: I780a1f5acc228297cbbea86f33679d00e9153b4c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/75260
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Brandon Jones <bajones@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
- Add format implementation on D3D12, Metal and Vulkan
- Add more formats in depth/stencil copy, sampling and load op tests and
refactor them to test with parameters.
BUG=dawn:690
Change-Id: I829d1eea3ce35ffb39417ea23fb8afba6d542769
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/73180
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
Move the manually parts to src/dawn_native/webgpu_absl_format.cpp/h.
Rename the template webgpu_absl_format.cpp/h to api_absl_format.cpp.h .
BUG=dawn:1201, dawn:563
Change-Id: Ibbeea43227f4fcf7f1d6b1d0bc3927226e79e6c3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/74300
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Junwei Fu <junwei.fu@intel.com>
This CL implements RequestDevice and also has changes for
Dawn to internally use wgpu::FeatureName enums, instead of
strings. Some of the string handling is kept for now to
support the deprecated creation path. GetFeatureInfo is added
to the instance to get a name and description of the feature,
for reporting in about://gpu.
Dawn device toggles are now passed in an extension struct off
of the device descriptor. This is only supported in dawn_native,
and not dawn_wire, for now, since dawn_wire doesn't have a way
to serialize lists of null-terminated const char*.
To enable the client to check whether the toggle descriptor is
supported, a `dawn-native` feature is added which is supported
all the time with dawn_native, but not supported with dawn_wire.
Feature `dawn-native` also enables a synchronous version of
CreateDevice for convenience.
Bug: dawn:160
Change-Id: Ifc195e7ea808c6c319021528ef4b36bd65583bff
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72020
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
This is a reland of 1a57ee9d4e
Original change's description:
> Bump maxUniformBindingSize to 2^16 from 2^14
>
> This was upped in https://github.com/gpuweb/gpuweb/pull/2159
>
> Bug: dawn:1173
> Change-Id: I98d46b9b5a339280d21ddcb2ab039ce7ebce4ec3
> Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71608
> Commit-Queue: Brandon Jones <bajones@chromium.org>
> Auto-Submit: Corentin Wallez <cwallez@chromium.org>
> Reviewed-by: Brandon Jones <bajones@chromium.org>
Bug: dawn:1173
Change-Id: Id86c6432e792a44bb231ae7778df1d38f4084f5b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/74302
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
This change add a couple end2end tests when readonly depth/stencil
attahcment is enabled in render pass and the pipeline doesn't
sample from the depth/stencil attachment. The pipeline only do
depth/stencil test.
This situation may not be useful in real world applications. But
it is possible that developers do it in this way. And it impacts
the implementation on some backend like Vulkan.
Bug: dawn:485
Change-Id: I8a81330659295cd4a2f00591b6ead719538babd9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71441
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
For unknown reasons formatting of a error context message fails in
RenderEncoderBase::APISetBindGroup fails if the bind group is printed
as a string in the message. This despite the exact same code working as
intended in ComputePassEncoder::APISetBindGroup. Replacing it with a
static "[BindGroup]" to allow the rest of the message to format
correctly until the reason for the failure can be determined.
Bug: dawn:1190
Change-Id: I31105acff4d08849a34e406c6b8e85fab28f17b5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/73280
Auto-Submit: Brandon Jones <bajones@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
CopyTextureForBrowserOptions deprecated AlphaOp after supporting
color space conversion. AlphaMode for src and dst is the replacement.
Bug: dawn:1140
Change-Id: Id507bd7525d74be8a12d212b92cc22f0c7bc94b7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/73141
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This has been removed from WGSL and is now deprecated in Tint.
Bug: tint:1324
Change-Id: Ic187ce3c5ce0723db6f3ca6483e9f5e73ce27acc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72880
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Corentin Wallez <cwallez@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Struct member defined in dawn.json could have a 'length' attribute
defines the length of the member if it is an array.
In previous, the 'length' could only be 'strlen' or other variable
which has 'uint32_t' types. It cannot support constant length.
CopyTextureForBrowserOptions transfers conversion parameters and
conversion matrix with a constant length. This CL adds this ability
to meet such requirements.
Bug: dawn:1140
Change-Id: I6eeabbc55cc3853fe15e33bdd44060b16f6096bf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72580
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This is so we can implement the adapter/device APIs fully
on dawn_wire.
Bug: dawn:689
Change-Id: I47f68157d081f359f871e0efe0d974dfe53de7d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71521
Reviewed-by: Loko Kung <lokokung@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Adds upstream instance/adapter APIs. In dawn_native, the basic APIs
to get limits and properties are implemented, but requestAdapter and
requestDevice are not. In dawn_wire, nothing is implemented, but the
stub definitions are put in place, as well the mechanism to inject
WGPUInstance into the wire.
There is a lifetime concern with WGPUInstance and WGPUAdapter on the
wire in that we need to ensure that the client cannot free the
instance or adapter while they are in use. In the near term, this is
not a problem because Chromium will always hold ownership of the
instance and adapters outside of the wire - i.e. it won't inject and
then release ownership.
Bug: dawn:160, dawn:689
Change-Id: Id904272983f23babc9177bc163d78c4fa1044da0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71520
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Loko Kung <lokokung@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
This CL add color space conversion bases for CopyTextureForBrowser.
Theoretically, it could support any color space conversion. But
test cases only cover (Srgb, DisplayP3) to (Srgb).
It could be expanded to more color spaces conversions.
Bug: dawn:1140
Change-Id: I332e6d1f7cf2424fd5f5af83c71fa45c98d2d8ac
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/70780
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
On Intel platforms, all planes in fact have same dma-buf, so the
DISJOINT bit shouldn't be used to create the vkimage.
For multi-planar formats, VkImageDrmFormatModifierListCreateInfoEXT
has to be used instead of
VkImageDrmFormatModifierExplicitCreateInfoEXT.
Bug: chromium:1258986
Change-Id: I25306a438e7ba9fd981848e63068e486bbddf11d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/68961
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jie A Chen <jie.a.chen@intel.com>
This will hopefully help catch issues early when people forget to set
pNext before creating the PNextChainBuilder.
Bug: dawn:1223
Change-Id: Ic6b9704aeaa20731e4f7de4d1ac0207d4110c720
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71762
Reviewed-by: Loko Kung <lokokung@google.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
In RenderPipelineDescriptor.DepthStencilState, if depth test or depth
write is enabled, the texture format must have depth aspect. Likewise,
if stencil test or stencil write is enabled, the texture format must
have stencil aspect.
Bug: dawn:1226
Change-Id: I9d7efb25675ff2c90704fa45703fb542bab6f1f5
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72101
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
It requires to revise quite a few flags in order to implement
readonly depth/stencil attachment on Vulkan. For example:
- VkAccessFlags,
- VkPipelineStageFlags,
- VkImageUsageFlags,
- and the most important: VkImageLayout.
These revised flags need to be applied to many Vulkan objects.
For examples:
- VkImageMemoryBarriers,
- IMAGE_LAYOUT revisions in descriptor set (bindings),
render pass, and subpass,
- and loadOp/storeOp revisions in render pass and subpass.
This change also does some workarounds in order to make Vulkan
validation layers happy. For example:
- use DEPTH_STENCIL_READ_ONLY image layout for binding a
depth/stencil texture,
- add DEPTH_STENCIL_ATTACHMENT_BIT image usage for binding a
depth/stencil texture.
Note that STORE_OP_STORE is used for depth/stencil's storeOp for
readonly attachment. It leads to Vulkan validation error. This
change igores that error and let it proceeds and it works well.
Bug: dawn:485
Change-Id: Ie247c9cbffbd837984b0933a905632ab5ad8862d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/70280
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This library binds directly to dawn_native and implements
webgpu.h. It may be built as a single library so it can
be easily used in other projects.
Bug: dawn:1220
Change-Id: I73be8c6455922fa526efd1600446cc46b07e82ed
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/53887
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Bug was a result of an external BGL reference that lingered after device was destroyed leading to a bad read on the device's FencedDeleter when the BGL reference was finally released. Fix just makes sure that the previous code path runs during the device destruction instead of afterwards.
- Removes passthrough call in BGL to the allocator and instead has the device keep track of the allocator directly so that the list can be used to both deallocate bind groups and bind group layouts at the end.
- Makes the allocator an ObjectBase so that we can have an explicit copy of the device since getting it from the layout can be dangerous now that the allocator may outlive the layout.
Bug: chromium:1276928
Change-Id: Ibca5e3c313fc0c0980ecaaa9ad2c871e204ac153
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71860
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Loko Kung <lokokung@google.com>
Updates validation logic to match the recent changes in
https://github.com/gpuweb/gpuweb/pull/2385 that allows stripIndexFormat
to be undefined at pipeline creation time, even for strip topologies.
Non indexed draw calls are valid with such pipelines. Indexed draw calls
fail validation at draw time.
Bug: dawn:1224
Change-Id: I28ff78eac726d46f99a099ffb2338b5da81a4a88
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72000
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
The EnumClassBitmasks is used by dawn/api_cpp.h that needs to be common.
Define a macro to export the operators from dawn to other various namespace.
BUG=dawn:1201
Change-Id: I20badd54e844fead6ecf12546a2c9e0afa2fd83f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71900
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Junwei Fu <junwei.fu@intel.com>
Remove wgpu::ErrorFilter::None from Dawn as it is removed from the
specification.
Bug: dawn:1206
Change-Id: If8ec2722cf1b2bad380011f191f296f0e591646d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71607
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
This patch fixes a bug in the allocation of internal staging buffer
for Queue::WriteTexture() that we must ensure the buffer offset to
be 4 bytes when calling Queue::WriteTexture() on depth stencil
textures as is restricted by Vulkan SPEC.
BUG=dawn:1213
TEST=dawn_end2end_tests
Change-Id: Ia2d073ef12d48baff42fca97005c1185c9560f1c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/71605
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Loko Kung
Brandon Jones
Corentin Wallez
Austin Eng
Sunny Sachanandani
Jiawei Shao
Ben Clayton
Shaobo Yan
Li Hao
fujunwei
Yunchao He
James Price
jchen10
Zhaoming Jiang