[Mlir-commits] [mlir] [mlir][spirv] Add bfloat16 support (PR #141458)

[Darren Wihandi]

fairywreath wrote:

> Thanks for contributing this.
> 
> Do we emit errors when using unsupported arithmetic ops etc. with bf16? Could we have tests that cover this?

Yes, originally there was only 1 FMul test. I have added more tests to ensure errors are emitted for arithmetic, subgroup ops, and atomics.

https://github.com/llvm/llvm-project/pull/141458