forked from OSchip/llvm-project
53d6988171
Switch r.h.s operand layout (n1, k1, n0, k0) -> (n1, k1, k0, n0) which is more consistant with scalar-vector products vectorization and elementates operand transpose. Reviewed By: rsuderman Differential Revision: https://reviews.llvm.org/D107307 |
||
|---|---|---|
| .. | ||
| mlir | ||
| .style.yapf | ||
| CMakeLists.txt | ||
| requirements.txt | ||