In a typical application, most draws will use different uniforms values for
things like the world position and orientation. In the current state of WebGPU
this means that a new bind group needs to be created for each draw to set the
right uniforms. Bind group creation is expected to be more expensive than
recording draws because they incur an allocation.
This feature is to reduce the number of bind groups that need to be
created.
The patch implements dynamic buffer offset on vulkan backend. But Vulkan
takes dynamic offset as uint32_t type, which is not the same size as the VkDeviceSize
used to create buffers so we cast them for now.
Bug=dawn:55
Change-Id: I6163866feb040d1a653f9a20d2ce22d80509968e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7461
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Should prevent some uninteresting fuzzer failures.
Bug: dawn:37
Change-Id: I6b2bef01deb700239e02ef5afa9313023ec83899
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7483
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Forces texture to clear to non-zero on creation to test the logic of
lazy clearing.
Bug: dawn:145
Change-Id: Ia738bf03be29dc620caf5f6d04b57c29b6919e85
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7280
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
These tests cannot be run using the wire because the wire client
device is not a real Metal device.
Bug: dawn:158
Change-Id: I97dec7f10fa19d118a72a965f57ee6703cc57765
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7540
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
In a typical application, most draws will use different uniforms values for
things like the world position and orientation. In the current state of WebGPU
this means that a new bind group needs to be created for each draw to set the
right uniforms. Bind group creation is expected to be more expensive than
recording draws because they incur an allocation.
This feature is to reduce the number of bind groups that need to be
created.
The patch implemented dynamic buffer offset on metal backend.
Bug=dawn:55
Change-Id: I34f397ec9ec10fa5b8e8a1c029d2d6a19e8dc0ee
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7320
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
This is only a renaming: change VertexInput to VertexBuffer, and
change InputState to VertexInput.
The next two patches will do as follows:
1) change the structure of vertex input descriptor related stuff.
2) change num to count.
BUG=dawn:80, dawn:107
Change-Id: Ie76aa653a527759a9c3b4a4792e3254689f053b8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7420
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
Implemention of InsertDebugMarker, PushDebugGroup and PopDebugGroup
for D3D12 using PIX event runtime.
Bug: dawn:44
Change-Id: I488f4638777afad3420ba96b350d9f19f2cd80dc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7400
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Brandon Jones <brandon1.jones@intel.com>
For samplers with NaN mLodMinClamp or mLodMaxClamp, the equality
operator returned false.
Adds checks for finite values, and also early returns if two pointers
are equal.
Bug: dawn:143, chromium:965633
Change-Id: Id5998d6eec275af0fbe30e3b4fcb3eed4fe64c6a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7401
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Idan Raiter <idanr@google.com>
Current Metal drivers have a bug that doing MSAA resolve with non-zero
ResolveLevel and ResolveSlice cannot work correctly. This patch adds a
workaround for this bug that first resolving into a temporary one-level
2D texture, then copying the result into the true resolve target.
Although the end2end test ResolveIntoOneMipmapLevelOf2DTexture and
ResolveInto2DArrayTexture can pass on the try bot using AMD GPU, we
find they fail on some other AMD GPUs (Macbook Pro 2018), so currently
this workaround is enabled on all Metal backends.
BUG=dawn:56
TEST=dawn_end2end_tests
Change-Id: Ie85858e58ff486d49dc11c270d8b6d95e216fd42
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6780
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Forces texture to clear to nonzero on creation to test the logic of
lazy clearing.
Bug: dawn:145
Change-Id: Ib1f6e8f961927008e3c09960ad0219f975d86cc6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7240
Commit-Queue: Natasha Lee <natlee@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
In a typical application, most draws will use different uniforms values for
things like the world position and orientation. In the current state of WebGPU
this means that a new bind group needs to be created for each draw to set the
right uniforms. Bind group creation is expected to be more expensive than
recording draws because they incur an allocation.
This feature is to reduce the number of bind groups that need to be
created.
The patch implemented dynamic buffer offset validation logics and adding unittests.
BUG=dawn:55
Change-Id: If6200a87bfedba825abcbfe60f336eab2e27226a
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7100
Commit-Queue: Shaobo Yan <shaobo.yan@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch fixes a crash issue that on D3D12 backends Dawn will crash
when we create too many views from cbvSrvUav or sampler descriptor heaps
as both of the heaps have fixed sizes.
To fix this issue we count the number of views that are required
in these two heaps before allocating memory for them, so that the
memory allocation from these two heaps will always succeed.
BUG=dawn:149
Change-Id: I7d42bf2ad677c6ecfa3ce1b3e471bc0d9258f266
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7260
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This patch refactors the implementation of the emulation on "store and
MSAA resolve" store operation by moving all the function calls related
to this toggle into one recursive function EncoderRenderPass(). This
refactoring will also make it easier to implement more workarounds on
Metal render pass.
BUG=dawn:56
Change-Id: Ifc737407001e55863835ab994b735e088beda8c6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7220
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This is the first command to return a struct. This patch also
updates the code generator to support structure return values.
Bug: dawn:7
Change-Id: Ie8acec895c0ec88429672138ffc900032fbbc447
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/4780
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Run end2end tests with argument '-d' or '--enable-backend-validation'
Change-Id: I34a3f453dcd3a57d76301801ae2abe42f847cb1b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7140
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
Created toggle to force texture clearing to 1 bits in order to
test the logic of lazy clearing.
Bug: dawn:145
Change-Id: I83bc32f046159c709c426b77458fbdf115f7bfd0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7120
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Natasha Lee <natlee@microsoft.com>
This patch fixes a bug on the OpenGL backend by setting default values
for dynamic states (blend color, viewport, scissor rectangle, etc) before
clearing the attachments of the render pass so that the full area of the
attachments can be cleared correctly.
BUG=dawn:126, dawn:137, dawn:141
TEST=dawn_end2end_tests
Change-Id: I9ddf74a67413ed51f83a69a3dc2c2377a0622cac
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7080
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Austin Eng <enga@chromium.org>
Update stride value to max(attribute.offset + sizeof(attribute) for
each attribute) when input stride is 0 in Metal backend
BUG=dawn:75
TEST=dawn_end2end_tests
Change-Id: I449532807f7fba1bca22106f2c38d496b8cbf01f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7020
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Hao Li <hao.x.li@intel.com>
Mip map level starts with the size of texture image and halves the
size until a 1x1 dimension.
BUG=chromium:954769
Change-Id: I59e51af216d65c721a3d2c2023306883e524820d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/7000
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Yizhou Jiang <yizhou.jiang@intel.com>
This patch fixes an issue in the check of passed OpenGL fences.
Currently Dawn checks if an OpenGL fence is passed with glGetSynciv(sync,
GL_SYNC_CONDITION, sizeof(GLint), &length, &status), which always returns
GL_SYNC_GPU_COMMANDS_COMPLETE and is meaningless. This patch uses
glClientWaitSync() to check if a fence is passed instead.
BUG=dawn:137
Change-Id: Iff6b18bcc7155b06bde98ff68c7ed104a27bfe55
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6800
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Static_casts are prefered over reinterpret_casts for better type
safety
Change-Id: I190cbee293591ebf8ab8035e900c081848eb1f30
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6921
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This change runs Dawn DEPS trybots as experimental. It allows us to know
when Dawn changes will break Chromium and the autoroller. They are
experimental because we expect Chromium integration to break when Dawn's
API changes.
Bug: chromium:948902
Change-Id: Ifb2ac253fefc90c68a4b56d5895099e1ddf596e7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6880
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kenneth Russell <kbr@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
Error objects that are cached still have mIsBlueprint to false and would
try to uncache themselves, which isn't valid.
Also adds a regression test.
BUG=dawn:143
Change-Id: Icd8bab52982b8520519d52296fffa6ed85088b8e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6900
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
This is the first step to do pipeline deduplication. It also introduces
tests for deduplication.
BUG=dawn:143
Change-Id: Ib22496f543f8d1f9cfde04f725612504132c7d72
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6861
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Code surrounding the swapchain needs to Flush the wire so that the
native swapchain implementation is initialized before GetPreferredFormat
is called. Likewise the swapchain should be destroyed before the last
FlushWire otherwise the native swapchain will still be using the window
at the beginning of the next test.
Also fixes the newly added ASSERT_DEVICE_ERROR not working with -w.
BUG=dawn:143
Change-Id: Ie3d6fa8c9236f68f5bff1ad7abd112ab25035843
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6860
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Adds an additional buffer and frame to the D3D12 swap chain to
unthrottle fps in situations where we can't maintain a constant 60fps.
Bug: dawn:142
Change-Id: I06ef697ca37608dc51816dd99de5534cb8c9b843
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6840
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Rafael Cintron <rafael.cintron@microsoft.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Change Vulkan Swap Chain from VK_PRESENT_MODE_IMMEDIATE_KHR to
VK_PRESENT_MODE_FIFO_KHR. Some common GPUs don't support immediate mode,
so we should default to a value that is required to be supported.
Bug: dawn:90
Change-Id: I66dec23141e69637ccfa8a2ed0ee651114b9ddcf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6820
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
This patch implements "store and MSAA resolve" store operation on Metal
drivers that don't support MTLStoreActionStoreAndMultisampleResolve with
a workaround that does MSAA resolve in another render pass.
Driver workaround is one type of Dawn Toggles. Dawn Toggles will include
other optional optimizations and features that can be configured to use
or not when we create Dawn Devices.
As all Metal try bots don't need this toggle, to better test this
patch on the try bots:
1. We add the support of forcing enabling a workaround when starting an
Dawn end2end test so that we can test the workaround on the platforms
where the workaround is disabled.
2. We add an optional parameter DeviceDescriptor to CreateDevice() so
that we can custom the toggles the Dawn device should use.
This patch also adds the support of querying toggle details from Instance
and the names of the toggles in use from Device. These APIs are tested in
the Dawn unittests added in this patch.
BUG=dawn:56
TEST=dawn_end2end_tests
TEST=dawn_unittests
Change-Id: Iae31d2ded6057eee638b6099d3061e9d78b04d55
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6620
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
This is implementing the feature on a single backend and without tests
so we can get it in the hands of people trying WebGPU quickly.
BUG=dawn:43
Change-Id: I4d0611efd21dd1af053288957f137febc78a74e8
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6721
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Now MultisampledRenderingWithDepthTest can pass on Windows Intel
Vulkan drivers with the fix in
https://dawn-review.googlesource.com/c/dawn/+/6640.
BUG=dawn:56
TEST=dawn_end2end_tests
Change-Id: If83d8183960960119e728d0372f5d1c5c11d849e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6680
Commit-Queue: Kai Ninomiya <kainino@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
This test was using two separate depth-stencil textures for sequential
render passes. The second render pass is supposed to load the texture
from the first but it was instead loading a new uninitialized texture.
Bug: dawn:133
Change-Id: I615ec4502a6f26bd039187b0c1942ffbb806b549
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/6640
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Yan, Shaobo
Kai Ninomiya
Natasha Lee
Austin Eng
Yunchao He
Brandon Jones
Idan Raiter
Jiawei Shao
Li Hao
Corentin Wallez
Bryan Bernhart
Yizhou Jiang
Rafael Cintron