[Mlir-commits] [mlir] [MLIR][ROCDL] Remove patterns for ops supported as intrinsics in the AMDGPU backend (PR #102971)

[Krzysztof Drewniak]

krzysz00 wrote:

This patch does change the lowering to use the compiler intrinsics. The OCML functions in question were, for f32, redundant wrappers for the intrinsics that the compiler team wants to remove.

For bf16 exp and log, I'd almost think that an LLVM patch is in order

https://github.com/llvm/llvm-project/pull/102971