[all-commits] [llvm/llvm-project] 295eda: [AArch64][GlobalISel] Better vecreduce.fadd loweri...

[David Green via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 295edaab13326d842f8298ad07e9a5f9e5d3b131
      https://github.com/llvm/llvm-project/commit/295edaab13326d842f8298ad07e9a5f9e5d3b131
  Author: David Green <david.green at arm.com>
  Date:   2023-11-27 (Mon, 27 Nov 2023)

  Changed paths:
    M llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
    M llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
    M llvm/test/CodeGen/AArch64/vecreduce-fadd.ll

  Log Message:
  -----------
  [AArch64][GlobalISel] Better vecreduce.fadd lowering. (PR #73294)

This changes the fadd legalization to handle fp16 types, and treats more types
as legal so that the backend can produce the correct patterns. This is
currently a missing identity fold for `fadd x -0.0 -> x`