dan sinclair
dan sinclair
Initialization to 0 is tested in https://github.com/gpuweb/cts/blob/main/src/webgpu/shader/execution/zero_init.spec.ts
* Builtin shadowing tests: https://github.com/gpuweb/cts/blob/main/src/webgpu/shader/validation/parse/shadow_builtins.spec.ts * Other shadowing tests: https://github.com/gpuweb/cts/blob/main/src/webgpu/shader/execution/shadow.spec.ts
```groovy fn foo() { let h = 0x1p-200f; } ``` was the test case from the other issue. Small magnitude which should fail as not representable.
Also, hex float values which are positive and negative infinity should be a shader error
The WPT support in bikeshed requires the tests to be in WPT (the paths are relative, not absolute). So, unless we moved into WPT, or updated bikeshed to allow a...
This was discussed at the last meeting: https://docs.google.com/document/d/1vaICnZrSK_3xexJhJsMGkj_vtie2WHnav1gADmEP-88/edit#bookmark=id.gbgc1udsz1hm The relevant part being: `KG: Sound like consensus that the cautious path is to make it an error at shader creation time....
@sagecram
From the description, it sounds like there are 4 separate PRs in here? Can we split them apart so they're easier to review? What is the reason behind the `INCLUDE`...
Scanning through the gitlab PR, the `INCLUDE` pulls out a few lines, but also makes it harder to read the individual scripts, so from someone just reading the `.amber` files...
This is the first part of the review for what I've gotten through so far. Is there a document on pipeline libraries I can read? I'm a bit confused by...