mirror of https://github.com/llvm/torch-mlir
aca33f1742
I've upstreamed the necessary quantized linalg Op with the "channel-first" ordering used by torch (https://github.com/llvm/llvm-project/pull/107740) for 2d convolution. This patch changes the lowering for the quantized 2d case of `aten.convolution` accordingly, which saves three transpositions per convolution (input, weights, result) and therefore removes the requirement to try to optimize these away in downstream passes. |
||
|---|---|---|
| .. | ||
| CMakeLists.txt | ||
| DataMovement.cpp | ||
| IndirectDataMovement.cpp | ||
| Linear.cpp | ||
| Pooling.cpp | ||
| PopulatePatterns.h | ||
| Random.cpp | ||
| Reduction.cpp | ||
| TensorConstructors.cpp | ||
| TensorScalarInterop.cpp | ||
| TorchToLinalg.cpp | ||
| Uncategorized.cpp | ||
| Utils.cpp | ||