This replaces the @override keyword that was previously used for
pipeline-overridable constants.
Support for pipeline-overridable constants in Dawn is hidden behind
the "disallow-unsafe-APIs" toggle, so we can make this change directly
instead of deprecating and continuing to support the old syntax.
Fixed: tint:1403
Change-Id: I9b2957a1e33b12772bfe449c0f3a31d929f8aa8b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/80480
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
That is, bring back both VarForDynamicIndex and
PromoteInitializersToConstVar. This is not a complete revert, though:
* VarForDynamicIndex no longer depends on ForLoopToLoop
* Both can cope with hoisting from "else if"
* More unit tests were added in the interim
Delete PromoteSideEffectsToDecl for now. This may be brought back to
handle ensuring order of evaluation.
Bug: tint:1300
Change-Id: I8bbae46377ec4603cc02c1eb3f0661a8461a19fa
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/79940
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
This significantly simplifies the calling code from Dawn.
It does require moving the SingleEntryPoint transform above the
renamer, since we need to call SEP with the original name. That is
probably also an optimization, since we do the subsequent transforms
with a culled AST.
Bug: tint:1268
Change-Id: I1f48ec0238ea76310742c60b119311747dd3dbac
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/80020
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
As the resolver currently enforces in-order declarations, this does not
change the declaration order from iterating over the
ast::Module::GlobalDeclarations.
The MSL backend has been changed to use the
sem::Module::DependencyOrderedDeclarations list instead of looping over
different declaration types separately.
Bug: tint:1266
Change-Id: I698d612032285311017bfceab3c42adae1928a0e
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/79767
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
This toggle controls if the fuzzer will throw a fatal error in the
case that the shader becomes invalid.
Currently the fuzzers do no guarantee that the options that are
provided are correct/valid, so there are many uninteresting cases that
become invalid due to the limited nature of the fuzzers, not due to
bugs in the code. The default off state of this toggle will suppress
this noise.
Once https://bugs.chromium.org/p/tint/issues/detail?id=1356 is
implemented this toggle can be default on.
BUG=tint:1357,chromium:1294533
Change-Id: I7170e5a30691105c97e20d8337aadf81ac2bc3bc
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/79840
Reviewed-by: Ben Clayton <bclayton@google.com>
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
This is a check that has been added to newer versions of clang and is
tripping for me locally. The actual issue is in code being generated
by protobuf.
Updating protobufs has cross-dependency issues with spriv-tools, so is
non-trivial. There is already a special case suppression for internal
protobuf issues, so I am just adding to the carve out.
BUG=tint:1419
Change-Id: I3ecd111a778fb4c65a113382ded8d6160deab462
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/79841
Auto-Submit: Ryan Harrison <rharrison@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ryan Harrison <rharrison@chromium.org>
Use a bloom-filter-style early rejection to eliminate whole blocks of
case statements from the switch type checks. Much like IsAnyOf(), the
list of types considered are recursively tested as a whole and
then binary-chopped if there's a potential match, until we test the
individual switch case types.
Bug: tint:1383
Change-Id: I5b30f19ea070e8352bf6b9363f133da906013182
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/78544
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
In GLSL, gl_SampleID and gl_SampleMask[In] require the
GL_OES_sample_variables extension, so output:
"#extension GL_OES_sample_variables : require"
in the header if those builtins is used.
Note that extensions must be inserted before the default precision
declaration, but helpers must be inserted after it, so we set a flag
and emit extensions, then the precision declaration, then helpers.
Further fixes:
- use gl_SampleMaskIn for input builtins, gl_SampleMask for output,
necessitating the addition of a storage class to GLSLBuiltinToString()
- fix the handling of gl_SampleMaskIn: it's array<i32> in GLSL, not
array<u32> as in SPIR-V
- centralize conversions for GLSL builtins used as input variables in
FromGLSLBuiltin()
- implement bitcasts on assignment to GLSL builtin output variables,
centralized in ToGLSLBuiltin()
- update the extension handling in the GLSL writer to check for both
sample_index and sample_mask.
- call UnwrapRef() in GLSL's EmitBitcast(). In the test case, we were
not recognizing the argument as a uint, yielding float() instead of
uintBitsToFloat().
Bug: tint:1408, tint:1412, tint:1414
Change-Id: Ie01541eb6e7cdf4e21347341f988bff916346797
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/78920
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Transform any SPIR-V that has an array with a custom stride:
@stride(S) array<T, N>
into:
struct strided_arr {
@size(S) er : T;
};
array<strided_arr, N>
Also remove any @stride decorations that match the default array stride.
Bug: tint:1394
Bug: tint:1381
Change-Id: I8be8f3a76c5335fdb2bc5183388366091dbc7642
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/78781
Reviewed-by: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
A type dispatch helper with replaces chains of:
if (auto* a = obj->As<A>()) {
...
} else if (auto* b = obj->As<B>()) {
...
} else {
...
}
with:
Switch(obj,
[&](A* a) { ... },
[&](B* b) { ... },
[&](Default) { ... });
This new helper provides greater opportunities for optimizations, avoids
scoping issues with if-else blocks, and is slightly cleaner (IMO).
Bug: tint:1383
Change-Id: Ice469a03342ef57cbcf65f69753e4b528ac50137
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/78543
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
The fields `has_location_decoration` and `location_decoration` are references to `has_location_attribute` and `location_attribute`, respectively, which act as an alias for a field-rename.
The default copy-constructor was copying across the reference (pointer) to the source StageVariable, instead of making a reference to the target's fields. This is causing Bad Things™ to happen in downstream projects.
Credit to dneto@ for identifying the issue, and for making this fix (which I've stolen for this CL).
Change-Id: Id0570e4e389f28980fa5aea29ad63b5100373dc5
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/79200
Reviewed-by: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
The CombineSamplers transform was incorrectly flagging StorageTexture
(which in GLSL ends up as image2D) as needing to be combined with a
sampler, or at least renamed. This is incorrect: StorageTexture never
has an associated sampler, so don't try to pair it up and just output
it as image* in GLSL.
In GLSL, textureLoad (aka texelFetch) of depth textures is not allowed.
The fix is to bind the depth texture as the corresponding f32 texture
instead (e.g., texture_depth_2d -> texture_2d<f32>,
texture_depth_cube -> texture_cube<f32>, etc). This requires changing
both the uniform globals and function parameter types. We're now going
to receive a vec4 instead of a float from texelFetch, so add a ".x"
member accessor to retrieve the first component. (Note that we don't
do this inside a CallStatement since this gives the CloneContext
indigestion, and CallStatement is going to ignore the result of the
call anyway.)
We were failing to find the dummy samplers that Dawn creates for the
calls that actually do require a dummy sampler, since the old Inspector
implementation of GetSamplerTextureUses() does not find them. The fix
is to implement a new Inspector call to return the texture/sampler
pairs the Resolver found during resolution. This will include the
dummy sampler as a null variable pointer.
In order to identify the placeholder sampler, we pass in a BindingPair
to represent it. When we discover a null sampler in the variable pair,
we return the passed-in placeholder binding point to the caller (Dawn).
(Dawn will use a group of kMaxBindGroups, to ensure that it never
collides with an existing sampler.)
Bug: tint:1298
Change-Id: I82e142c2b4318608c27a9fa9521c27f15a6214cd
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/78820
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
