[pull] main from gpuweb:main #13
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![pull[bot]](https://avatars.githubusercontent.com/in/12910?s=60&v=4){kind=small}
* Add a test which prints what environment it runs in This is very useful for making sure tests are running in the environment (main thread vs various worker types) that we expect them to be. Issue: https://crbug.com/330573129 * Print a bunch more info and print it to console.log * nits * hashes * second try at evading wpt lint * Simplify printing by printing everything as JSON
* Tests that function calls evaluate arguments left to right * Includes && and || operators * Includes updates through pointers
These tests were either not skipping themselves when running in workers (where the DOM is not available), or they were unnecessarily using DOM features where they didn't intend to. This makes these two tests pass or skip correctly in workers (at least, pass as much as the non-worker versions do). Issue: https://crbug.com/331351978
We were relying on postMessage to turn LogMessageWithStack objects into Error objects, but this was kind of magic, and didn't work well - in particular, the name was lost. Flattening this into a plain object before postMessage makes it much less magic. Now logs sent from workers print correctly, showing for example "DEBUG:" instead of "Error:". Issue: none
Checks the following context dependent names: * Attribute names * Built-in value names * Diagnostic severity control * Diagnostic triggering rules * Enable extensions * Language extensions * Swizzles * Interpolation type * Interpolation sampling Built-in value names and interpolation type and sampling are made context dependent in gpuweb/gpuweb#4559.
* WGSL validation test for textureGather/compare
This fixes issues finding the service worker files both on gpuweb.github.io and in Chromium's automated testing setup. Also: - Add the base resource path to print_environment - some tangentially-related comment cleanup. Issue: #3583, https://crbug.com/330596242
* wgsl: Add tests for operator precedence Includes execution and validation tests. * Add literal and no-whitespace parameterizations
Rather than just throwing an error and letting the runtime break down when this happens (which in Chromium's testing causes timeouts). I decided to fail instead of skip for now because probably nothing should actually rely on the skip - and failing loudly is less error-prone. Issue: #3591
* Fix wrong metadata in video resource meta lists H264 HFlip/VFlip marked metadata as vp9 wrongly. Fixed this issue by this PR. * Also fix some typo in README
* Uses const_assert to test correct overloads
* wgsl: Add execution tests for derivative builtins * wgsl: Add non-uniform discards to derivative tests * Address review comments - Remove unncessary subtraction of 0.5 from position - Cover viewport with a single triangle instead of a quad
Shared/service workers don't have requestAnimationFrame. - Update webgpu:web_platform,canvas,getCurrentTexture:expiry:* to skip the requestAnimationFrame cases. - Update webgpu:web_platform,canvas,getCurrentTexture:multiple_frames:* to not use requestAnimationFrame unnecessarily.
There's no reason to iterate the subcases if the test is unimplemented. When someone goes and implements the test, they can deal with any overparameterization problems. This cuts log spew from tests like `webgpu:shader,execution,expression,call,builtin,textureSampleBias:arrayed_2d_coords:`
* WGSL execution tests for textureSample
* Initial change * Validation tests for insertBits. (Some minor updates on utils) * Address comments * Update tests a bit to use subcases and prune some combinations
* WGSL builtin textureLoad for texture_2d tests
…ns by test functions (#3844)
* Add CTS test for new depthBias validation * Lint * Apply Kai's suggestsions * Fix TS error
Signed-off-by: sagudev <[email protected]> Co-authored-by: Brandon Jones <[email protected]>
It is a shader/pipeline-creation error if the number of bits to shift is greater than *or equal* to the bit width of the value to shift. This change corrects the expectations for the equals case, and adds a function reference to the pipeline subcase so it isn't stripped out. These tests will pass in Dawn following: https://dawn-review.googlesource.com/c/dawn/+/199075
When testing accesses into uniform address space, the array tests pack scalars into a 4-element vector. This is a big hack, to accommodate the 16byte alignment requirement for uniform address space. Previously the robust access tests had a subtle tight coupling to this decision. This change adds a 'accessSuffixes' field to the array type info to indicate how to access individual scalar elements under test. So in the uniform case, when a vec4 is artificially created, return a list of the .x .y .z .w components to get the original scalar values back. This allows client tests to smoothly generalize. Update the robust access tests to this change. It's the only affected test. Issue: #3405
This is the last step in the code cleanup for adding f16 to generateTypes Fixed: #3405
According to the spec, division by zero is invalid even if only the rhs can be constant evaluated for integral types. However, for FP types, it is only an error if the whole expression can be const-evaled.
* wgsl: Add coverage for swizzle on a pointer In addition to a new execution test, also add coverage to the alias analysis and uniformity analysis tests. Both of these new tests failed prior to a recent Tint change that fixed a bug with these swizzles. * Change for review * Use skipIf for execution test * Add uniform version of swizzle uniformity test
Bug: crbug.com/346817929
) * Add API validation tests about dual source blending and color attachments This patch adds validation tests about using dual source blending and color attachments. When dual source blending is used, there can only be exactly one color attachment, whether its write mask is 0 or not, which is required by Metal. * Fix assignment to vector elements
This CL tests 50 texels per subcase where as before it was testing 1 texel per subcase. This is more inline with the original design for texture_utils.ts doTextureCalls and checkCallResults and means more tests but less subcases which is usually faster. #3871
* Cover GPUExternalTexture with WebCam Cases Add utils to capture frame from camera if env supported. And compared the rendering result with 2d context draw image to cover uploading webcam frames. * Update src/webgpu/web_platform/external_texture/video.spec.ts Co-authored-by: Corentin Wallez <[email protected]> * Remove extra type declaration. --------- Co-authored-by: Corentin Wallez <[email protected]>
* Add validation tests on blend factors with Src1 and @blend_src This patch adds the validation tests on the use of blend factors with `Src1` and `@blend_src` in the fragment shader. When the blend factor uses `Src1`, `@blend_src` must be used in the fragment shader whether the corresponding colorWriteMask is 0 or not, as is required by Metal. * Small fix
…src1 (#3879) * Add validation tests on the blend factors using the alpha channel of src1 * For merge
* Add dual source blending blend factors to `kBlendFactors` This patch adds all the blend factors defined in the extension "dual-source-blending" to `kBlendFactors` in capability_info.ts so that all the tests about blend factors can test these newly added blend factors. This patch also removes the validation tests about using the alpha channel of `Src1` because it doesn't follow the latest SPEC: only when the color blend factor uses the alpha channel of `Src` must the fragment output be `vec4`. When the color target doesn't have the alpha channel, the alpha blend factor is ignored so it is OK to set it as if it "uses" the alpha channel of `Src`. Note that the new validation test against `Src1-alpha` will be added into 'pipeline_output_targets,blend' after the fix is landed in the browser. * Improve the test descriptions and messages in `TODO`
* Validation * enables * built-in values * built-in functions: broadcast and ballot * Execution * built-in values * built-in functions: broadcast and ballot
…sh:* (#3887) Fix the case error which set the expired/destroyed external texture wrongly.
* Add validation tests on the access of the alpha channel of Src1 This patch adds validation tests on the use of `src1-alpha` in the blend factor. When the blend factor uses the alpha channel of src1, the fragment output must be vec4. This patch also implements function `IsDualSourceBlendingFactor()` to check if a blend factor needs the extension "dual-source-blending". This patch also adds the support of dual source blending in the helper function `getFragmentShaderCodeWithOutput()` to simplify the shaders used in dual source blending tests. * Fix `IsDualSourceBlendingFactor()` * Allow `IsDualSourceBlendingFactor()` accepts null value as parameter * More fixes
Cube and cube-arrays are not thoroughly tested at the moment. See TODO in comments.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot]
Can you help keep this open source service alive? 💖 Please sponsor : )