alan-baker

Vulkan specifically goes out of it's way to mention that mapping being local id and subgroup id is not reliable. I have not seen it fail for one dimensional workgroup,...

Adding a proposal for the WGSL side of things. I outlined this at the previous F2F, but don't think I ever wrote it into any issue. ### WGSL Proposal ####...

> > New built-in that synchronizes texture memory in a workgroup. > > What would the function signature on this be? Would it be global for all textures? Or would...

WGSL meeting minutes 2023-10-10 * AB: Google would also like to ship this * AB: have an experimental implementation with subset of formats available behind flags in chrome * AB:...

Vulkan doesn't require bitcast to preserve a NaN and I have seen implementations that don't do so. I wouldn't be opposed to a note in bitcast that calls this out.

> @alan-baker is there some reference as to how bitcast works in Vulkan that you can point to? > > It is really surprising to Apple / WebKit that this...

> Oh thanks for that explanation @alan-baker , that made me realize the difference in our implementation. > > We compile with `.fastMathEnabled = YES` since nothing in the WGSL...

Workgroup operations are out of scope for this issue, but feel free to file a new issue with that request so we can track it. I expect the answer will...

Chrome implemented minimal experimental extensions to experiment how much portability exists with subgroup operations. # Extension ## WGSL The following additions were made to WGSL: * subgroupBallot - an unpredicated...

Made a proposal in #4368 based on the previous work.