In a typical graphics application it is a common usage to update some uniforms once per draw,
and such uniforms include the word positions, orientations, and so on. In the current state of
WebGPU, this means that for each draw call we have to create a new bind group to set the right
uniform values. Bind group creation is expected to be more expensive than
recording draws because a memory allocation is required.
The functionality of dynamic buffer offset is to reduce the number of bind groups that need to
be created.
The patch implements dynamic buffer offset on D3D backend using root descriptor.
Bug=dawn:55
Change-Id: Ia713a4edb3c0ab8f3bba048d7813f343e9dee166
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9040
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Introduces the buddy system for fast power-of-two sub-allocation.
BUG=dawn:27
Change-Id: I56836ae317ecc5a91d8341c843fc37d4f91fb5af
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9260
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
bindgroup->GetLayout() was called before checking the bindgroup is an
error, causing an assert to fire in Debug. Adds a regression unittest.
BUG=dawn:196
Change-Id: I58e95879a7f2a7cf5c47c7b228f7e3b5bb72d8ea
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9560
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
In a typical graphics application it is a common usage to update some uniforms once per draw,
and such uniforms include the word positions, orientations, and so on. In the current state of
WebGPU, this means that for each draw call we have to create a new bind group to set the right
uniform values. Bind group creation is expected to be more expensive than
recording draws because a memory allocation is required.
The functionality of dynamic buffer offset is to reduce the number of bind groups that need to
be created.
The patch implements dynamic buffer offset on OpenGL backend using glBindBufferRange and adds
validation to check whether visibility of resources are none.
Bug=dawn:55
Change-Id: I77e10a9677d1737f377301ee89e29d904c91c298
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9540
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
WebGPU have a 2D pipeline layout plus a vertex buffer table while Metal
has a single vertex buffer table that contains everything (including
uniform and storage buffers). Previously the space for vertex buffers
was statically allocated in that table which made the last vertex buffer
go out of bound of the Metal vertex buffer table.
This fixes the issue by packing all the vertex buffers that are used
right after the vertex buffers used by the bind groups. This is a
drive-by fix found while looking at reserving Metal vertex buffer 30 to
contain the shader storage buffer lengths.
BUG=dawn:195
Change-Id: If5c67bbc0d15c976793ef43889e50e4a360217d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9387
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Add test for many buffer writes.
BUG=dawn:27
Change-Id: I026c3b9a6d30c7c6634b89db9798902b036150c3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9280
Commit-Queue: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
This patch factors the CommandAllocator, CommandIterator, and error
handling out of CommandEncoderBase so it can later be used by the
RenderBundleEncoder.
Bug: dawn:154
Change-Id: Ia4f8c3ce7f432f0887b619bd8090aa9bec7330fc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9181
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Implement SetViewport on the rest backends: D3D12 and Metal
BUG=dawn:53
TEST=dawn_end2end_tests
Change-Id: I2e552867080eb8c35db641d1a61b0341d2a0a048
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9020
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
This CL tests integration of the MemoryTransferService with buffer mapping.
It tests the basic success and error cases for buffer mapping, and it tests
mocked failures of each fallible MemoryTransferService method that an embedder
could implement.
Change-Id: Iece660fb49664cc6a09a0b0b8dbe59e2882a6017
Bug: dawn:156
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8841
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch adds MemoryTransfer client/server interfaces and
uses it to implement data transfers for buffer mapping.
This patch also provides a default "inline" implementation of
the MemoryTransfer which is used if the embedder does not
provide one on initialization.
Because implementations of MemoryTransfer perform their own
serialization, a skip_serialize option is added to WireCmd records.
Bug: dawn:156
Change-Id: I2fa035517628a3ad465b0bc18a6ffc477e2bd67f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8642
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Currently Dawn always doubles the size of the largest ring buffer
everytime when DynamicUploader::Allocate() is called, which is not
necessary and will cause out-of-memory errors when this function is
called for too many times. This time fixes this behaviour by only
doubling the size of the largest ring buffer when it is not big enough
for the request buffer size.
BUG=dawn:108
TEST=dawn_end2end_tests
Change-Id: I734493dbc632ca8e9a30c1ceeb0b7b9e0474656b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9160
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch meant to add tests for SetViewport to verify that the
viewport we set in one render pass will not impact the viewport(s)
in other render pass(es).
These newly added tests exposed a potential error on OpenGL backend.
This patch also implemented SetViewport and fixed the bug on OpenGL.
BUG=dawn:53, dawn:189
TEST=dawn_end2end_tests
Change-Id: Ia2e74f003c02acdcf983b24a3386efab30793652
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9140
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This is needed for two reasons:
- TextureBase and TextureViewBase stored Formats by value which isn't too
much overhead at this time but will get bigger in the future.
- The OpenGL backends needs its own GLFormat structure to store data about
each format which will eventually contain complicated logic to detect
support in the GL driver so it shouldn't be duplicated in Textures.
The computations of the information about Format is moved from being done
whenever they are needed to being precomputed at DeviceBase initialization.
This makes each format have a constant "index" in that can be used in the
backends to address their own structure, for example a GLFormat table.
Also some DeviceBase pointers were made const for validation.
BUG=dawn:128
Change-Id: I37d1e9c739b87cddcea09cb1759e175704d90f9e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9101
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
The existing implementation of Result with tagged pointers was not able
to handle constant pointers for the result. This is required in
follow-up CLs to return internal formats in a ResultOrError.
This CL extracts the tagged pointer logic out of Result<T*, E*> so it
can be shared with Result<const T*, E*>.
Tests are also added to cover Result<const T*, E*>.
BUG=dawn:128
Change-Id: Id19ae8e1153bcfcaf94d95ac314faf2b23af6f91
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/9100
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Following WebGPU spec change at https://github.com/gpuweb/gpuweb/issues/331,
bind groups in Dawn now use the whole size of the buffer if binding buffer
size is UINT64_MAX.
Bug: dawn:22
Change-Id: If28d905e634432755dad5c67c69eadedcee53dfe
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8863
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: François Beaufort <beaufort.francois@gmail.com>
OpenGL doesn't allow to query PCI information, but we can get vendor name through glGetString(GL_VENDOR), then match it with vendor id.
BUG=dawn:184
Change-Id: Icdcdc6bfdb5cd561b5057e8b2eab06df2dd5cea1
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8940
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
This patch adds the support of all BC formats on D3D12, Metal and
Vulkan and related end2end tests.
BUG=dawn:42
TEST=dawn_end2end_tests
Change-Id: I50caf687d3cfee7df80070952f55f96dc363a830
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8900
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
The main aim of this CL is to add tests for SetViewport. But it
also implements the feature on Vulkan backend to verify the tests.
BUG=dawn:53
TEST=dawn_end2end_tests
Change-Id: I9578b2d9650cb952ce67cf91f79e9682c9932ee3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8820
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
Binding layout uses an attribute to mark dynamic resource. And this should be taken into
consideration when determining whether two bindingInfo are equal.
Otherwise, dawn cache system may take two different bind group layouts(they have same binding
resource type, masks and visibilities but one bind group has dynamic resource and the other
not) as the same. This will return wrong bind group layout object to user and cause validation
error when user tries to use dynamic buffer offset.
BUG=dawn:188
Change-Id: I11b158724c7592718af434b48c9e3f2f9bb72b93
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8901
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
These additions to dawn.json don't require changes to users of the C++
API so they are done now to prepare for webgpu.h. Validation is added
that forbids using the new enum values and non-default values for
members until they are implemented.
BUG=dawn:22
Change-Id: Ie6f3384243d6874e7aca4b94733409755a066c93
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8862
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Previously validation of the combination of dynamic with the binding
type was only done when creating the bind group when it should have been
done when the bind group layout is created.
BUG=dawn:55
Change-Id: I976f7e052f9737929fc05908af50e6ad5ceef397
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8861
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Instead it is replaced by one enum in dawn_native and another in utils.
BUG=dawn:22
Change-Id: I094a40c8d4e22b704e59aea60cbefd1f05c5352a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8800
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This patch is to verify the framebuffer coordinate in Dawn (and WebGPU).
It shows that gl_FragCoord(0, 0) is at the top left.
BUG=dawn:43
TEST=dawn_end2end_tests
Change-Id: Ifd16488169c986d0c2c85eb92283c69e70ad33ae
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8780
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This also adds the extra validation needed to support non-renderable
formats, as well as tests for it
BUG=dawn:128
Change-Id: I3bc79b641aa0fd5e3358f89a87f2e457d0ecc58a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8760
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
This patch adds the support of BC5 formats on Metal and the related
dawn_end2end_tests to verify Dawn works correctly when (bufferSize -
bufferOffset < bytesPerImage * copyExtent.depth), which is the special
case of buffer-to-texture and texture-to-buffer copies on Metal.
BUG=dawn:42
TEST=dawn_end2end_tests
Change-Id: I27c384d0d8d2bb908f1ad15c2451fd23c1313598
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8720
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
These are the CommandEncoder, ComputePass and CommandBuffer descriptors
that contains nothing but a debug name for now but are important for
later extensibility. Defaults are added so the C++ API doesn't require
the descriptors to be passed as arguments.
Also renames variables named "info" for RenderPassDescriptor to
"descriptor" as is now the standard in the codebase.
BUG=dawn:22
Change-Id: I9de4cfbbce952d01fb79ed1d9f34825a6fa174f9
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8686
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
WebGPU remove dynamic-uniform-buffer and dynamic-storage-buffer but add a new attribute in
BindgroupLayoutBinding to record whether a buffer resource is dynamic.
Dawn need to align with this change.
BUG=dawn:180
Change-Id: I873ad2ec75575e72d184f89a6e3698dff6df50d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8520
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
The CL that introduced this formats reordered their components based on
the name of the Vulkan formats, but it turns out that WebGPU lists
components from low bits to high-bits while Vulkan PACK32 formats are
listed from high to low. So the format component orders match, except
for RGB10A2 which is actually BGR10A2 in Vulkan. Instead the Vulkan
backend is updated to correctly use the RGB10A2 format.
BUG=dawn:128
Change-Id: If7f045f020429c44c84b9aed34a4a80107208d5c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8601
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
This is to match the equivalent rename in WebGPU's IDL.
BUG=dawn:22
Change-Id: Ibdf75869e58835d984df559878b884c55310a403
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8623
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Yunchao He <yunchao.he@intel.com>
This is to match the name in the WeBGPU IDL.
BUG=dawn:22
Change-Id: I0a4a7b62afa30cfd699d90efb2dd0c085778883b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8624
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Yunchao He <yunchao.he@intel.com>
This removes the R5G6B5 format that's only available on iOS on Metal and
leaves A2RGB10 unimplemented because it has inverted R and G channels
compared to Metal's RGB10A2 format.
BUG=dawn:128
Change-Id: I9b960d00f281b4d32318bad3360826b8db5bcb72
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8600
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Jiawei Shao <jiawei.shao@intel.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This is to match the WebGPU IDL.
BUG=dawn:22
Change-Id: I63e32fe980c1727562055221ea2769a670461e93
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8625
Reviewed-by: Yunchao He <yunchao.he@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Now end2end tests always run with first adapter (the last adapter for
Metal because of dawn:29) on multi-GPU systems. Add
'--adapter-vendor-id' flag to make it run with specified adapter.
BUG=dawn:181
Change-Id: Iaff3ab92a502f88b24e4aa8a5ed3cb4d1d5925e3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8460
Commit-Queue: Hao Li <hao.x.li@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch implements Culling and FrontFace on backends, and add tests too.
This test also verified that we couldn't invert FrontFace on Metal backend.
Otherwise, the tests would fail on all HWs.
But we do need to invert CCW/CW on OpenGL backend. Because Y axis is up in
OpenGL while Y axis is down in WebGPU.
Bug=dawn:43
Change-Id: I7dd0922477397a13c5f7208e104ff352a673a556
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8420
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
This patch adds the support of BC5 formats on D3D12 backend. On D3D12,
the "rowPitch" refers to the number of bytes in a row of blocks (which
covers up to 4 scanlines at once) for the textures in BC formats.
This patch also adds the related end2end tests for four typical cases of
B2T and T2B copies on D3D12:
(1) copyBytesPerRowPitch + byteOffsetPerRowPitch <= rowPitch and
texelOffset.y == 0
(2) copyBytesPerRowPitch + byteOffsetPerRowPitch <= rowPitch and
texelOffset.y > 0
(3) copyBytesPerRowPitch + byteOffsetPerRowPitch > rowPitch
(4) texelOffset.z > 0
BUG=dawn:42
TEST=dawn_end2end_tests
Change-Id: If27ab3e56596e25c1c5be787ca021c0748021a46
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8541
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
The root causes of these failures are as follows:
1. 'fragmentStoresAndAtomics' feature is not enabled when we create the Vulkan device.
2. The binding value of dynamic buffer offset end2end test not set correctly.
For failure reason 1, this patch enabled fragmentStoresAndAtomics.
For failure reason 2, this patch modify dawn validation logic in SetBindGroup to check
binding size and update binding size in dynamic buffer offset end2end test.
BUG=dawn:170
Change-Id: I46f12453d4c83d9d3c7de6e183442cf516335f2f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8320
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch implements RenderPassEncoder::SetViewport. It also adds
validation code, and validation tests as well.
BUG=dawn:53
TEST=dawn_unittests
Change-Id: Idc27f4098761a57fc10e53be82bd3c4b35b53cb4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8542
Commit-Queue: Yunchao He <yunchao.he@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch supports BC5 formats on Vulkan backends and adds related
Dawn end2end tests.
For the textures with BC formats, they could have non-multiple-of-4
sizes on the non-zero mipmap levels in sampling, but we are still
required to provide texture data in complete 4x4 blocks in texture
copies because that is the size of which they are stored in GPU memory.
In this patch, we refer the term "physical memory size" as the memory
size of the texture subresource in GPU memory, and the term "virtual
memory size" as the size used in texture sampling. As Dawn requires
the Extent3D in texture copies must fit in the physical memory size,
while Vulkan requires it must fit in the virtual memory size, this
patch recalculates the imageExtent to ensure it always follow this
Vulkan validation rules.
For Dawn end2end tests, note that we use pure green and pure red for
the textures because BC5 does not support SRGB formats. Furthermore,
"CopyPartofTextureSubResourceIntoNonZeroMipmapLevel" is skipped in
this patch because there is an issue on the T2T copies from a region
within the virtual size of one texture to another one that exceeds
the virtual size of another texture in Vulkan SPEC.
BUG=dawn:42
TEST=dawn_end2end_tests
Change-Id: I17518cd335fb13125cb753bbf879bc06eb20e426
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/8260
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
