[pull] main from gpuweb:main #18
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}
textureLoad and textureGather and textureGatherComapre do not use GPU mix weights so don't load them. This will make the these test run on Intel Mac compute stage where the weights are bad.
Compat will not have rgba32float as a render target support except as an optional feature so refactor to not use it.
It's a little confusing but depth textures used for "sampling" are not filterable but used for "comparisons" they are. | | texture_?d<f32> with depth texture | texture_depth_?? | | ----------------------------- | ---------------------------------- | ---------------- | | sampler `'linear'` | bad | bad | | sampler_comparison `'linear'` | N/A | OK |
This PR fix the expectation of fragment shader subgroup_invocation_id. Any invocation id (including 0) within subgroup can be assigned to inactivate invocations, and ids of active invocations can go larger than active invocations number but still smaller than subgroup size. This PR also fix the draw call for fragment subgroup tests.
texture_depth_xxx can only be used with non-filtering samplers or comparison samples, not with filtering samplers. see table in #4026 Also enabled more depth tests
This makes toneMapping optional which required updating a few tests to use the nullish coalescing operator when accessing it. New vertex formats are added that require updating capability_info.ts (and tests for unorm8x4-bgra is added at the same time). Finally dictionary types for copies are renamed throughout the CTS even if deprecated aliases are available.
Cube arrays are unsupported in Compatibility mode.
Co-authored-by: Kai Ninomiya <[email protected]>
This CL modify the shader for fragment subgroup_size built-in test to prevent using broadcasting (from invocation 0), because any invocation id can be inactivate for fragment shader. This CL use ballot instead and do more validation within the shader.
* test 2d, 2d-array, 3d textures * test fragment stages * test out of range values
* Change fragment builtin tests to avoid non-uniform branches around subgroup operations * Add diagnostic filters to execution tests relying on non-uniformity
See CL/diff for details Also fixed an issue with copyBufferToTextureViaRender that it was using the source format to decide whether to render to a depth or stencil format but it should be using the destination format for that.
Co-authored-by: Peter McNeeley <[email protected]>
This bug broke all textureXXCompare tests with unencodable depth formats.
* Fix readTextureToTextureViews for multisampled textures I'm not sure when this bug was introduced. These tests (textureLoad:multisampled) used to pass. Fixed the bug and added a test to hopefully find it quicker if it breaks in the future. Also added a test for createTextureWithRandomDataAndGetTexels with a generator. This was fixed in a previous PR without a a specific test as the fact that it failed other tests showed it needed fixing but I decided it's probably best to keep these tests too. * Update src/webgpu/shader/execution/expression/call/builtin/texture_utils.ts Co-authored-by: Kai Ninomiya <[email protected]>
* Do not query weights for `textureSampleBaseClampToEdge` `textureSampleBaseClampToEdge` does not need the GPU weights. Skip loading them. This will let the tests run on Mac Intel where the weights are bad.
* Test texture mix weights separately Before this change, the texture builtin utils would query how the local GPU mixes texture mip levels. It expected that to be close to a linear interpolation which is true for most GPUs. The known exceptions are Mac AMD and Mac Intel in a compute stage. It would check how far from linear iterpolation the weights are. If the weights were too far off linear interpolation it would assert. With this case, those asserts for linear interpolation are moved from all the tests that use these weights to a separate test. This allows the texture builtin tests to run further on devices that have bad mix weights with the new test at least documenting a particular GPU's failure to implement linear interpolation between mip levels.
We sample a compressed texture on the GPU and compare that to expected. If they don't match we try to identify which texels were sampled. This is done by doing a binary search, creating textures with some pixels white and some black and then sampling the texture to see if those white pixels contributed to the sample values. We can't do this for compressed textures because we can not easily create a compressed texture with white and black pixels exactly where we want them to appear. So, for compressed textures we just use rgba8unorm when identifying sample points. That can be useful to help find major bugs either in the CTS itself or in the GPU that are true for all texture formats. Unfortunately, if the bug is specificially with compressed textures this doesn't help. So, print out a warning that the sample points might not match. Also, change the diagram to show blocks. Eexample when blockWidth = 2, blockHeight = 2 ``` +===+===+===+===+ # a | # | # +---+---+---+---+ # | # | # +===+===+===+===+ # | # | # +---+---+---+---+ # | # | b # +===+===+===+===+ ```
Compat: Don't test depth textures with non-comparison samplers
These should genearte a validation error in compat but not in core.
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 : )