intel/mlir-extensions

Support atomic operations for low precision

Opened this issue · 0 comments

tye1 commented

For triton usage, we would like to request the support of atomic operations for low precisions, including bf16, fp16 and INT8.