[all-commits] [llvm/llvm-project] 73e14d: [X86] combineConcatVectorOps - recursively call co...

[Simon Pilgrim via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 73e14de207a3aa0fa071fa56756e8e316edf5227
      https://github.com/llvm/llvm-project/commit/73e14de207a3aa0fa071fa56756e8e316edf5227
  Author: Simon Pilgrim <llvm-dev at redking.me.uk>
  Date:   2025-03-08 (Sat, 08 Mar 2025)

  Changed paths:
    M llvm/lib/Target/X86/X86ISelLowering.cpp
    M llvm/test/CodeGen/X86/matrix-multiply.ll

  Log Message:
  -----------
  [X86] combineConcatVectorOps - recursively call combineConcatVectorOps instead of predicting when ops will freely concat (#130275)

The IsConcatFree helper is limited to estimates on where concatenating the subvector operands is beneficial, this patch replaces FADD/FSUB/FMUL concatenation checks with a recursive call to combineConcatVectorOps to see if it will profitably concatenate further up the chain.

Other opcodes can be moved to using the CombineSubOperand helper in future patches.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications