[all-commits] [llvm/llvm-project] 6b9c18: [mlir][spirv] Handle non-innerprod float vector ad...

[Jakub Kuderski via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 6b9c186b2d4c8bd315034a9655a28d32bcf745ab
      https://github.com/llvm/llvm-project/commit/6b9c186b2d4c8bd315034a9655a28d32bcf745ab
  Author: Jakub Kuderski <jakub at nod-labs.com>
  Date:   2023-11-27 (Mon, 27 Nov 2023)

  Changed paths:
    M mlir/lib/Conversion/VectorToSPIRV/VectorToSPIRV.cpp
    M mlir/test/Conversion/VectorToSPIRV/vector-to-spirv.mlir

  Log Message:
  -----------
  [mlir][spirv] Handle non-innerprod float vector add reductions (#73476)

Instead of extracting all individual vector components and performing a
scalar summation, use `spirv.Dot` with the original reduction operand
and a vector constant of all ones.