Ben Clayton

Probably not. I'm not aware of marl being used in any project with exceptions enabled, and I certainly didn't consider exceptions during implementation. Sorry.

> Would it be not possible to just have one row? E.g., > > ```wgsl > @const @must_use fn mat2x2(v1 : vec2, v2 : vec2) -> mat2x2 > ``` We...

> The standardization group discussed this like a year ago, but we couldn't come to a conclusion about how they should be exposed. You're right that it's worth a revisit,...

> Possibly the usages of access_mode, texel_format and address_space but I haven't checked the grammar yet. These are used by type template arguments, and to support future user-templated code, I...

Yes. And known backend compilers that can produce severely broken output with non-finites. We will likely expose non-finites with a future extension, which will be available for drivers and hardware...

As mentioned in https://github.com/gpuweb/gpuweb/issues/4435, it would be nice to clarify that bit 0 is the LSB.

You're right. I missed the spec change in https://github.com/gpuweb/gpuweb/pull/2031, and only became aware of this with our discussion yesterday. With that, declaring an array-size-from-let would be a more sensible way...

> For example arrayLength() is the only way solve it when you don’t know exact array size in _generalised_ function. But module-scope constants is the thing that was in GLSL...

I personally think `arrayLength()` aligns the best with today's WGSL. It would require two new things: * `typeof(expr)` - Resolves to the type of `expr`, **without** evaluating `expr`. There's probably...

@dj2 - see https://github.com/gpuweb/gpuweb/pull/2081#issuecomment-915052204.