[all-commits] [llvm/llvm-project] 7c3cda: [AArch64][SVE] Prefer SIMD&FP variant of clast[ab]

[Cullen Rhodes via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 7c3cda551ac702de4eb8899180aa715896020d43
      https://github.com/llvm/llvm-project/commit/7c3cda551ac702de4eb8899180aa715896020d43
  Author: Cullen Rhodes <cullen.rhodes at arm.com>
  Date:   2022-07-13 (Wed, 13 Jul 2022)

  Changed paths:
    M clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clasta.c
    M clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_clastb.c
    M llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
    A llvm/test/Transforms/InstCombine/AArch64/sve-intrinsic-opts-clast.ll

  Log Message:
  -----------
  [AArch64][SVE] Prefer SIMD&FP variant of clast[ab]

The scalar variant with GPR source/dest has considerably higher latency
than the SIMD&FP scalar variant across a variety of micro-architectures:

  Core           Scalar    SIMD&FP
  --------------------------------
  Neoverse V1     9 cyc      3 cyc
  Neoverse N2     8 cyc      3 cyc
  Cortex A510     8 cyc      4 cyc
  A64FX          29 cyc      6 cyc