Consider tier1 formats language feature in RO/RW storage texture test

[andyleiserson]

Changes the readonly_and_readwrite_storage_textures test to consider the tier1 formats language feature.

Issue: none

---

Requirements for PR author:

• All missing test coverage is tracked with "TODO" or .unimplemented().
• New helpers are /** documented */ and new helper files are found in helper_index.txt.
• Test behaves as expected in a WebGPU implementation. (If not passing, explain above.)
• Test have be tested with compatibility mode validation enabled and behave as expected. (If not passing, explain above.)

Requirements for reviewer sign-off:

• Tests are properly located.
• Test descriptions are accurate and complete.
• Tests provide complete coverage (including validation control cases). Missing coverage MUST be covered by TODOs.
• Tests avoid over-parameterization (see case count report).

When landing this PR, be sure to make any necessary issue status updates.

[github-actions]

Results for build job (at db90237):

-webgpu:shader,validation,parse,requires:wgsl_matches_api:* - 10 cases, 10 subcases (~1/case)
+webgpu:shader,validation,parse,requires:wgsl_matches_api:* - 11 cases, 11 subcases (~1/case)
-TOTAL: 280650 cases, 2321809 subcases
+TOTAL: 280651 cases, 2321810 subcases

[kainino0x]

Hm........ in theory, the correct thing to do here would be to check wgslLanguageFeatures and not the device features. Because:

1. It's not just the new formats that should be skipped when the browser doesn't have support, it's also the existing formats that were newly-added to WGSL.
2. We should be running this test if the browser (compiler) supports the format, even if the device doesn't.

But we didn't add a wgsl language feature! Filed gpuweb/gpuweb#5524

[andyleiserson]

Updated with a check of the new language feature, currently commented out.

I'm not sure the best way to proceed here:

1. Merge with the new check enabled, even though the test can't pass on implementations yet.
2. Merge with the new check commented out, and file an issue to revisit.
3. Leave this PR open and wait for at least one implementation to catch up.

Maybe there should be a test specifically for the texture_formats_tier1 language feature? Although I think it would just be a subset of this test, unless we do something like specify that texture_formats_tier1 may be implemented only when readonly_and_readwrite_storage_textures is also implemented, in which case we could remove the tier1 formats from this test entirely.

[kainino0x]

We should go ahead and add some tests that fail. We can have separate test that JUST tests that using the requires directive or one of the 23 new format identifiers is allowed if-and-only-if the WGSL feature is available. That test will fail, but it won't prevent the others from running.

Then, with that out of the way, I think we can make this test complete now:

• If the format requires tier1 for storage (so the 23 formats, not just the 6 new ones):
  • Skip the test unless: the device feature is available OR the wgsl feature is available

This way it will start running in more places as soon as the wgsl feature is enabled (when the compiler supports the feature but the hardware does not).

WDYT, does this sound right? (Also let me know if you want to tackle all of this or not. Thanks!)

[andyleiserson]

Yes, that sounds right to me. I will try to pick this back up and revise the RO/RW test to work as you suggest, plus add a test specifically for the tier1 extension. The requires directive is covered by a different test, not sure if I will get to that one, but if I don't I'll at least file an issue.

[andyleiserson]

Updated this PR, and filed #4612 for more follow-up.

I actually don't think we need to skip this test under any condition. What I made it do is expect shader compilation to fail only if neither the device nor language feature is present. This allows the test to continue passing in Chrome and Safari. It fails in Firefox but that's because we currently allow these types in WGSL despite not exposing the tier1 and tier2 device features. One part of #4612 is to update the test to expect fail only based on language features, once there is some support.