[PATCH] D89693: [AArch64] Favor post-increments

[Sjoerd Meijer via Phabricator via llvm-commits]

SjoerdMeijer created this revision.
SjoerdMeijer added reviewers: fhahn, efriedma, dmgreen, samparker.
Herald added subscribers: danielkiss, hiraditya, kristof.beyls.
Herald added a project: LLVM.
SjoerdMeijer requested review of this revision.

This implements target hook `shouldFavorPostInc` for AArch64, which is queried in LoopStrengthReduce that can bring loops in a better form to generate post-increments.

Motivating examples are similar to: test/CodeGen/AArch64/arm64-scaled_iv.ll. I have changed that first from an opt to an llc test locally, and the diff here is against this local change. Using that example, more efficient is to transform this:

  LBB0_1:
      ldr     d0, [x1, x8]
      ldr     d1, [x10, x8]
      fmul    d0, d0, d1
      str     d0, [x9, x8]
      add     x8, x8, #8
      cmp     w8, #152
      b.ne    LBB0_1

into this:

  LBB0_1:
          ldr     d0, [x1], #8
          ldr     d1, [x9], #8
          subs    w10, w10, #1
          fmul    d0, d0, d1
          str     d0, [x8], #8
          b.ne    LBB0_1

Which improves one benchmark with 1.2%, and didn't show any changes in another which I also didn't expect to be impacted (just checking as a sanity check).


https://reviews.llvm.org/D89693

Files:
  llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
  llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
  llvm/test/CodeGen/AArch64/arm64-scaled_iv.ll
  llvm/test/CodeGen/AArch64/ragreedy-local-interval-cost.ll
  llvm/test/CodeGen/AArch64/shrink-wrapping-vla.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D89693.299030.patch
Type: text/x-patch
Size: 12540 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20201019/9861e5d1/attachment.bin>