diff options
| author | Christopher Bate <cbate@nvidia.com> | 2022-06-02 20:06:24 -0600 |
|---|---|---|
| committer | Christopher Bate <cbate@nvidia.com> | 2022-06-06 13:18:08 -0600 |
| commit | cca662b8495537b67f7b70090a33fc36154de81c (patch) | |
| tree | 0c3a6c559bc6384d5578703040d57f8c4e196727 /clang/lib/CodeGen/ObjectFilePCHContainerOperations.cpp | |
| parent | 66bd14697b11045462534a84b904d956e93ccf06 (diff) | |
| download | llvm-cca662b8495537b67f7b70090a33fc36154de81c.zip llvm-cca662b8495537b67f7b70090a33fc36154de81c.tar.gz llvm-cca662b8495537b67f7b70090a33fc36154de81c.tar.bz2 | |
[mlir][linalg] add conv_2d_nhwc_fhwc named op
This operation should be supported as a named op because
when the operands are viewed as having canonical layouts
with decreasing strides, then the "reduction" dimensions
of the filter (h, w, and c) are contiguous relative to each
output channel. When lowered to a matrix multiplication,
this layout is the simplest to deal with, and thus future
transforms/vectorizations of `conv2d` may find using this
named op convenient.
Differential Revision: https://reviews.llvm.org/D126995
Diffstat (limited to 'clang/lib/CodeGen/ObjectFilePCHContainerOperations.cpp')
0 files changed, 0 insertions, 0 deletions