[all-commits] [llvm/llvm-project] bd07c2: [AArch64] Prefer fmov over orr v.16b when copying ...

[David Green via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: bd07c2e266f65acb0204198ae1a441bf10499cb2
      https://github.com/llvm/llvm-project/commit/bd07c2e266f65acb0204198ae1a441bf10499cb2
  Author: David Green <david.green at arm.com>
  Date:   2021-08-03 (Tue, 03 Aug 2021)

  Changed paths:
    M llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
    M llvm/test/CodeGen/AArch64/arm64-AdvSIMD-Scalar.ll
    M llvm/test/CodeGen/AArch64/arm64-aapcs.ll
    M llvm/test/CodeGen/AArch64/arm64-neon-copy.ll
    M llvm/test/CodeGen/AArch64/arm64-vcvt_f.ll
    M llvm/test/CodeGen/AArch64/arm64-vector-insertion.ll
    M llvm/test/CodeGen/AArch64/bf16-vector-bitcast.ll
    M llvm/test/CodeGen/AArch64/bf16-vector-shuffle.ll
    M llvm/test/CodeGen/AArch64/bitcast-promote-widen.ll
    M llvm/test/CodeGen/AArch64/combine-comparisons-by-cse.ll
    M llvm/test/CodeGen/AArch64/f16-instructions.ll
    M llvm/test/CodeGen/AArch64/fadd-combines.ll
    M llvm/test/CodeGen/AArch64/fast-isel-select.ll
    M llvm/test/CodeGen/AArch64/fp16-v4-instructions.ll
    M llvm/test/CodeGen/AArch64/fp16-vector-bitcast.ll
    M llvm/test/CodeGen/AArch64/fptosi-sat-scalar.ll
    M llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll
    M llvm/test/CodeGen/AArch64/fptoui-sat-scalar.ll
    M llvm/test/CodeGen/AArch64/fptoui-sat-vector.ll
    M llvm/test/CodeGen/AArch64/machine-combiner.ll
    M llvm/test/CodeGen/AArch64/mla_mls_merge.ll
    M llvm/test/CodeGen/AArch64/neon-mla-mls.ll
    M llvm/test/CodeGen/AArch64/popcount.ll
    M llvm/test/CodeGen/AArch64/sqrt-fastmath.ll
    M llvm/test/CodeGen/AArch64/swift-return.ll
    M llvm/test/CodeGen/AArch64/urem-vector-lkk.ll
    M llvm/test/CodeGen/AArch64/vec-libcalls.ll
    M llvm/test/Transforms/LoopStrengthReduce/AArch64/small-constant.ll

  Log Message:
  -----------
  [AArch64] Prefer fmov over orr v.16b when copying f32/f64

This changes the lowering of f32 and f64 COPY from a 128bit vector ORR to
a fmov of the appropriate type. At least on some CPU's with 64bit NEON
data paths this is expected to be faster, and shouldn't be slower on any
CPU that treats fmov as a register rename.

Differential Revision: https://reviews.llvm.org/D106365