Does gpu.cpp support f64 or i64?

[ghostplant]

I wrote some kernels based on this framework, and I found datatypes for f64 and i64 are not defined.

[MichealReed]

Looks like float32 and halfs, maybe a contribution could be added to https://github.com/AnswerDotAI/gpu.cpp/tree/main/numeric_types for f64?

[ghostplant]

I don't think so. Unlike f16, datatype f64 is not defined in dawn so maybe more contribution should be in dawn?

[MichealReed]

Thanks for investigating. Maybe you could reach out to the author here to ask about contributing this?

https://developer.chrome.com/blog/new-in-webgpu-120

Dawn uses Gerrit here

https://dawn-review.googlesource.com/q/status:open+-is:wip

I have not even begun to digest their roadmap though.

[MichealReed]

Well, seems wgsl doesn't support f64 yet? https://webgpufundamentals.org/webgpu/lessons/webgpu-wgsl-function-reference.html

seems possible gfx-rs/wgpu#1154