[all-commits] [llvm/llvm-project] 8b140d: [AArch64][GlobalISel] Add select to and combines (...

David Green via All-commits all-commits at lists.llvm.org
Thu Jun 4 09:11:07 PDT 2026


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 8b140d425a48f90a858394c4186d4bb7826c1376
      https://github.com/llvm/llvm-project/commit/8b140d425a48f90a858394c4186d4bb7826c1376
  Author: David Green <david.green at arm.com>
  Date:   2026-06-04 (Thu, 04 Jun 2026)

  Changed paths:
    M llvm/include/llvm/Target/GlobalISel/Combine.td
    M llvm/lib/Target/AArch64/AArch64Combine.td
    A llvm/test/CodeGen/AArch64/GlobalISel/combine-select-zero.mir
    M llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-divergent-i1-phis-no-lane-mask-merging.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-divergent-i1-used-outside-loop.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.ll

  Log Message:
  -----------
  [AArch64][GlobalISel] Add select to and combines (#200131)

This adds combines for
// select c, x, 0 -> and c, x
// select c, 0, x -> and (not c), x
// select (not c), x, y -> select c, y, x

We need to freeze the value in the first two. The second is only
profitable if hasAndNot, so it excluded from all_combines.
https://alive2.llvm.org/ce/z/eG-aHT

This helps alleviate regressions when G_SELECT is made legal for vector
operations under AArch64. The AMD tests I am not sure about - let me
know if they look worse. The third combine is mostly useful
post-legalize.



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


More information about the All-commits mailing list