[Mlir-commits] [mlir] [AMD][ROCDL][AMDGPU] Support packed conversions fp8/bf8->bf16 and fp8/bf8->fp32 (PR #131850)

[Krzysztof Drewniak]

krzysz00 wrote:

> There are no ->bf16 instructions on gfx942. They're only available on gfx950.

So where in the code do we prevent those instructions from being emitted on gfx942?

https://github.com/llvm/llvm-project/pull/131850