projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e49ae62) | patch
[mlir][ROCDL] Define MLIR wrappers around new MFMA intrinsics
authorKrzysztof Drewniak <Krzysztof.Drewniak@amd.com>
Fri, 17 Jun 2022 18:31:05 +0000 (18:31 +0000)
committerKrzysztof Drewniak <Krzysztof.Drewniak@amd.com>
Mon, 20 Jun 2022 15:03:45 +0000 (15:03 +0000)
commit8e61fdc72780a17ef3b03c2526bba9272ce303c5
tree6acac03ec58e00ab5c4321437d90800cbdf04a91tree | snapshot
parente49ae6284c38c7fef93b3b72af4c89a6e4836a45commit | diff
[mlir][ROCDL] Define MLIR wrappers around new MFMA intrinsics

In order to support newer hardware, define wrappers around MFMA
intrinsics that have not previously been exposed in the ROCDL dialect.

A `amdgpu.mfma` wrapper around these instructions is in development
and will provide a more user-friendly interface to them.

Reviewed By: ThomasRaoux

Differential Revision: https://reviews.llvm.org/D128079
mlir/include/mlir/Dialect/LLVMIR/ROCDLOps.td diff | blob | history
mlir/test/Dialect/LLVMIR/rocdl.mlir diff | blob | history
Domain: System / Toolchain;