[Mlir-commits] [mlir] Add Lowerings for GPU WMMA F16/F32 ops to ROCDL dialect (PR #69357)

Tue Dec 19 08:05:59 PST 2023

krzysz00 wrote:

@navdeepkk-polymagelabs  I'd definitely like the testing issues fixed, and ... how much work have you done on moving to lowering to amdgpu.wmma + memref + ... ? I don't want to have to maintain this second lowering path, hence me wanting to shove things through `amdgpu.wmma`. 

If you really want this in, I'd like a timeline for when you'll do the followup PR that I can nag you about.

https://github.com/llvm/llvm-project/pull/69357