Before this change, the variadic function template version of
WrapInFunction would be selected when passing a utils::VectorRef<const
ast::Statement*>, even though an overload exists for that type. The
reason is that during type deduction, the compiler will select templates
over non-templates in its overload set. The only way around this was to
avoid type-deduction by explicitly casting the argument to
utils::VectorRef<const ast::Statement*>.
This CL adds a CanWrapInStatement metafunction that evaluates to true if
the arg type is one that could be passed to
ProgramBuilder::WrapInStatement. This is used to SFINAE in the variadic
args version of WrapInFunction.
Change-Id: I8aa3d69e2ce7324fd60b1b2a5906a51d51b549a3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/115502
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
This change is a necessary to support workgroupUniformLoad in a
following patch. Otherwise, there is no change to the set of shaders
that are accepted or rejected by the analysis.
We now distinguish between uniformity requirements on the contents of
a pointer parameter versus the pointer value itself when generating
tags for function parameters.
Whilst processing an expression, if we see a sem::Load node we pass a
flag down through child expressions to indicate that we will be
loading from the result. When processing an identifier expression, we
can then select between adding an edge to the source of the
pointer/reference versus the contents of the root identifier that it
corresponds to.
Since the pointers passed to atomic builtins can be uniform, we
special-case them to capture the fact that their return value is
always considered non-uniform.
The arrayLength builtin no longer needs special-casing.
Added many tests to cover various cases that are now captured
differently in the graph. There are two cases that are disabled as
they require variable pointers to trigger the uniformity violation.
Bug: tint:1780
Change-Id: I03edb65f22a6ffb0e7daf8b2f590f5de898e6262
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114861
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
Reviewed-by: David Neto <dneto@google.com>
Variables declared in for-loop initializers were not being tracked
properly across iterations as a check was wrongly determining them to
be declared inside the loop body.
Also fixes an issue where variables declared in for-loop initializers
were still considered to be in scope after loop exit.
Change-Id: I2ce3a519be45c8daba31bf00e8b2614f0bd6a2de
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114364
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Auto-Submit: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
When useing Bitcast to or from a class type, gcc warns even if the type
is trivially copyable. Fixed this by static_asserting that both types
are trivially copyable, and casting the pointers to std::byte*.
Change-Id: Ibb420f2dcdd35cfb187d74983fa8ab9b50d10c85
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/115180
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Previously this toggle was implemented only for the Metal backend, but
a need for it was identified on Android as well. This change moves the
implementation to the backend-agnostic command encoder recording so that
it works for all backends. Fundamentally it's still doing the same
thing, however: Swapping resolve targets that point at a non-zero mip
level or layer with a temporary texture and then performing a copy once
the render pass has ended.
Bug: dawn:1569
Change-Id: I292860cc74f653b2880e727d2ef3a7dfa3f10b91
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/106040
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Brandon Jones <bajones@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
This immediate rejection has been implemented in Native but
hasn't been yet in Wire. This commit adds the implementation
to Wire.
Also the commit changes the MapAsync callback firing timing
if pending map buffer is unmapped or destroyed. With this
commit the callback will be fired immediately Unmap or
Destroy is called to match the WebGPU spec. Currently the
callback is fired when the client receives a response from
server but it mismatches the spec.
Change-Id: Ia48d62be31912fd0384e23271e9de516f9d71d6c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/113607
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Takahiro <hogehoge@gachapin.jp>
Reviewed-by: Austin Eng <enga@chromium.org>
The resolver now wraps sem::Expression objects with a sem::Load object
anywhere that the load rule is invoked. sem::Expression provides an
`UnwrapLoad()` method that returns the inner expression (or
passthrough, if no load is present), which is analaguous to
Type::UnwrapRef().
The logic for alias analysis in `RegisterLoadIfNeeded` has been folded
into the new `Resolver::Load` method.
Fixed up many transforms and tests. The only difference in output is
for a single SPIR-V backend test, where some IDs have changed due to
slight re-ordering of when expressions are generated.
There may be further clean-ups possible (e.g. removing unnecessary
calls to `UnwrapRef`, and simplifying places in the SPIR-V writer or
transforms that deal with memory accesses), but these can be addressed
in future patches.
Fixed: tint:1654
Change-Id: I69adecfe9251faae46546b64d0cdc29eea26cd4e
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/99706
Commit-Queue: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
2D array texture may corrupt on some Intel devices, making out-of-bound
texture access and memory information leak. It's a critical security
issue. Intel driver team suggested the 24K extra memory approach in
order to mitigate the security issue before.
However, the texture corruption issue (and even the correctness issue)
can be worked around via allocating a few extra layers. And patches
have already been merged in Dawn, with a lot tests for verification. The
24K extra memory for each texture is actually incorrect and unnecessary.
So this patch removes relevant code in Dawn.
This patch mainly reverts some code of this patch below:
https://dawn-review.googlesource.com/c/dawn/+/96220
Bug: dawn:1507
Change-Id: Ic3239115ad4c74bdee928577ccbb20f1e35d13c3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114641
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Yunchao He <yunchao.he@intel.com>
Don't use dynamic casting in hashing and equality. These should be fast, and dynamic casting is expensive.
Add type::UniqueNode for things that need de-duplicating and bin the types on construction.
Replace some use of SFINAE with constexpr.
Also fixes a build failure for x86.
Bug: oss-fuzz:54184
Change-Id: Ic1b0708394f9f5703fc179a2c31ce18bd07e196c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114760
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
2D array textures with non-color formats like depth/stencil formats
are always fine. Workaround is not needed.
Multisample textures are treated as array textures from the
perspective of texture memory layout on Intel Gen12 and each sample
acts like a layer. However, multisample textures are fine.
Workaround is not needed.
Bug: dawn:1507
Change-Id: I1e5cd6a4e46503f67e4c1ffe2133e2e8fb121016
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/113740
Reviewed-by: Austin Eng <enga@chromium.org>
Kokoro: Austin Eng <enga@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Austin Eng <enga@chromium.org>
Makes it clear that the index being reported by these messages is the
group index and not the binding number. This was a point of confusion
in on the bug.
Additionally, adds more information to the error message regarding
buffer sizes being too small for the current pipeline. Now includes the
pipeline name and buffer size as well as the minimum required size. Also
includes a note explaining that uniform buffer bindings must be a
multiple of 16. (This recently changed and cause several existing
samples to break for non-obvious reasons.)
The error message still does not contain the buffer or binding number,
which would be helpful. This is because we currently lack a way to look
up the binding index from the packed index that this error is generated
with.
Bug: dawn:1604
Change-Id: Ibb2b44bc9e1583ddef34d703e83bcf64ed7a3aa2
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/113602
Commit-Queue: Brandon Jones <bajones@chromium.org>
Auto-Submit: Brandon Jones <bajones@chromium.org>
Reviewed-by: Kai Ninomiya <kainino@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
This CL updates Constant::As to be Constant::ValueAs. Now that Constant
inherits from CastableBase, there is already an As method on
CastableBase. This makes the override inside Constant confusing and
potentially incorrect.
Bug: tint:1718
Change-Id: I4f73971801e95225a99a5a993124c04194d0d7d6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114360
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Dan Sinclair <dsinclair@chromium.org>
This Cl splits the concrete constant implementations out of the
const_eval.cc file and into individual files. The classes are left in
the resolver namespace and will have a namespace update in a followup
CL.
Bug: tint:1718
Change-Id: I54539b6aa06f09aff39a1b1331d89f67a3594791
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/114160
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
