* Fixed DXC installation so that we download the latest release, and
patch it with the latest artifact build. This ensures that the correct
dxil.dll is next to dxc.exe, rather than having dxc.exe pick up whatever
dxil.dll is in PATH.
* To run tests with FXC requires the correct d3dcompiler_47.dll. To do
this, I modified the script to download the Windows SDK, and install
only the feature that contains the DLL.
* To avoid issues with DXC and FXC finding exes/dlls in PATH, modified
the script to temporarily set PATH as needed.
Bug: tint:940
Change-Id: Ic83b0b97f72ccc1f37b4b495eae6a6ab3421a95a
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/72560
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Vulkan allows them, but WGSL does not.
A duplicate decoration is purely redundant.
A parameterized decoratio with different parameterization is
an inconsistency and a semantic error, at least for currently defined
SPIR-V decorations.
So for each target, only take the first decoration of each kind.
Fixed: tint:1337
Change-Id: I6ed5c39cf2e213c695cb8217ed1b97814da3db56
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/72500
Reviewed-by: Ben Clayton <bclayton@google.com>
Auto-Submit: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: James Price <jrprice@google.com>
These argument names were probably renamed as part of a mass rename of
the "callback" category to "function pointer". The name "callback" more
clearly represents what the argument is used for, so revert back to it.
Bug: None
Change-Id: Id7539667f5daa87aec32ce8c90c20275d2fe80fd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72485
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
FXC fails to compile code that assigns to dynamically-indexed fixed-size
arrays in structs on internal shader variables with:
error X3500: array reference cannot be used as an l-value; not natively
addressable
This CL detects this case, and transforms such assignments into copying
out the array to a local variable, assigning to that local, and then
copying the array back.
Also manually regenerate SKIPs for HLSL/FXC after this change, which
fixes 30 tests. Also exposes some "compilation aborted unexpectedly" now
that "array reference cannot be used as an l-value" has been fixed. For
tests that fail for both DXC and FXC, updating SKIPs to the DXC one to
help distinguish actual FXC bugs from valid errors.
Bug: tint:998
Bug: tint:1206
Change-Id: I09204d8d81ab27d1c257538ad702414ccc386543
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/71620
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.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 patch tries to use the data directly in deserialized buffer
when the data is "data-only" and doesn't affect the control flow instead
of allocating and copying into a temporary buffer. Due to the protection
of TOCTOU attacks, currently we only set "wire_is_data_only" on the
parameter "data" in Queue.WriteBuffer() and Queue.WriteTexture().
With this patch, the performance of dawn_perf_tests
BufferUploadPerf.Run/*_WriteBuffer_BufferSize_* with "-w" will
be greatly improved (~20%) when the upload buffer size is greater than
1MB.
BUG=chromium:1266727
Change-Id: I7a9d54c9b505975235ee37aa72ee97f082ad3aa3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/72063
Reviewed-by: Austin Eng <enga@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
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>
Skipping those that are block-decorated is not correct, as
block-decorated structures can also have non-buffer usages. This is
even clearer now that WGSL has removed the block attribute.
Bug: tint:1324
Change-Id: I6484766a5c541d39e2dc08beb3ae7b889759a3fb
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/72083
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Any struct which is used as the store type of a buffer variable needs
to have a block decoration. If that struct is nested inside an array
or another struct, we wrap it inside another struct first.
This removes the SPIR-V backend's reliance on the [[block]] attribute,
which will soon be deprecated and removed.
Bug: tint:1324
Change-Id: Ib6ad54f24a3e4a090da9faeed699f266abcb66ba
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/72082
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This transform does not use the top-level ctx.Clone() method, so we
need to manually copy the applied transforms over to the destination
program.
This fixes the GLSL backend, where this transform was dropping the
applied transforms and breaking a required transform dependency later
in the chain.
Change-Id: I3db688fe35dfbe67468b9941f739cb2ecf3a843b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/72220
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
This fixes errors like "error X3500: array reference cannot be used as
an l-value; not natively addressable". Note that FXC treats matrices
like arrays. We still get this error for dynamically indexed arrays in
structs.
Also improved HLSL assign tests, and add missing ones for vector
indexing.
Manually removed 20 e2e skip hlsl SKIP files that are now passing with
this change.
Bug: tint:1333
Change-Id: If23881a667857a4d4ec6881e72666af0a666ef10
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/71982
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
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>
