[all-commits] [llvm/llvm-project] cca662: [mlir][linalg] add conv_2d_nhwc_fhwc named op

[Chris via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: cca662b8495537b67f7b70090a33fc36154de81c
      https://github.com/llvm/llvm-project/commit/cca662b8495537b67f7b70090a33fc36154de81c
  Author: Christopher Bate <cbate at nvidia.com>
  Date:   2022-06-06 (Mon, 06 Jun 2022)

  Changed paths:
    M mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
    M mlir/test/Dialect/Linalg/named-ops.mlir

  Log Message:
  -----------
  [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