[all-commits] [llvm/llvm-project] ed4dae: [mlir][spirv][gpu] Add conversion for load/store/m...

[Jakub Kuderski via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ed4daeaa131dcb33f2836f24f286fc8d757862a6
      https://github.com/llvm/llvm-project/commit/ed4daeaa131dcb33f2836f24f286fc8d757862a6
  Author: Jakub Kuderski <jakub at nod-labs.com>
  Date:   2023-09-14 (Thu, 14 Sep 2023)

  Changed paths:
    M mlir/include/mlir/Conversion/GPUToSPIRV/GPUToSPIRV.h
    M mlir/include/mlir/Conversion/Passes.td
    M mlir/include/mlir/Dialect/SPIRV/IR/SPIRVCooperativeMatrixOps.td
    M mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRVPass.cpp
    M mlir/lib/Conversion/GPUToSPIRV/WmmaOpsToSPIRV.cpp
    A mlir/test/Conversion/GPUToSPIRV/wmma-ops-to-spirv-khr-coop-matrix.mlir
    M mlir/test/Conversion/GPUToSPIRV/wmma-ops-to-spirv-nv-coop-matrix.mlir

  Log Message:
  -----------
  [mlir][spirv][gpu] Add conversion for load/store/mad coop matrix ops (#66311)

This is plugged in as an alternative lowering path in the gpu to spirv
dialect conversion. Add custom op builders for coop matrix ops to make
the create functions nicer to work with and less error-prone. The latter
is accomplished by following the op syntax and also requiring stride to
be a constant op to avoid confusion around the order of arguments.

The remaining lowering patterns will be added in a future patch.