David Neto

At the moment the WHLSL spec permits certain arrayed texture resource. See https://github.com/gpuweb/WHLSL/blob/master/Spec/source/index.rst Copying from there: ``` Texture1D RWTexture1D Texture1DArray RWTexture1DArray Texture2D RWTexture2D Texture2DArray RWTexture2DArray Texture3D RWTexture3D TextureCube TextureDepth2D RWTextureDepth2D...

In particular, that leaves out CubeArray. For index clamping CubeArray has an odd corner case where in Vulkan/SPIR-V the layer count and face index is folded together into the last...

It's also worth taking a second closer look, in case we missed something the first time around.

After #3 was merged, the following items from https://github.com/gpuweb/gpuweb/issues/34 remain to be addressed in spirv-execution-env.md: - division by zero - (not in the old issue): signed integer division, of INT_MIN...

The old issue explicitly did not deal with floating point exceptional cases. But the environment spec should bound the behaviours.

After the last community group meeting I started looking at what formal semantics was available for JavaScript. I'd want to use a formalism that's checkable. The JavaScript spec has prose...

Yes, I think this would be generally useful as a SPIR-V validation step. But the messages would be terrible (today) since we don't carry source level debug info into SPIR-V....

Placing this in Milestone 2, with other issues for WGSL for programming-in-the large.

To be explicit about it: * For array *values*, each modification is a conceptual copy. So swapping is two replacements. But I don't think this is the case you care...

cross-link to subgroups issue #4306