[all-commits] [llvm/llvm-project] ac3779: Enable Custom Lowering for fabs.v8f16 on AVX (#71730)

[David Li via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ac3779e92ef9405fd2c602a08e8031f7b8aeedd8
      https://github.com/llvm/llvm-project/commit/ac3779e92ef9405fd2c602a08e8031f7b8aeedd8
  Author: David Li <57157229+david-xl at users.noreply.github.com>
  Date:   2023-11-16 (Thu, 16 Nov 2023)

  Changed paths:
    M llvm/lib/Target/X86/X86ISelLowering.cpp
    M llvm/test/CodeGen/X86/vec_fabs.ll

  Log Message:
  -----------
  Enable Custom Lowering for fabs.v8f16 on AVX (#71730)

[X86]: Enable custom lowering for fabs.v8f16 on AVX

Currently, custom lowering of fabs.v8f16 requires AVX512FP16, which is
too restrictive. For v8f16 fabs lowering, no instructions in AVX512FP16
are needed. Without the fix, horribly inefficient code is generated
without AVX512FP16. Note instcombiner generates calls to intrinsics
@llvm.fabs.v8f16 when simplifyping AND <8 x half> operations.