[PATCH] D120481: [AArch64] Try to re-use extended operand for SETCC with v16i8 operands.

[Florian Hahn via Phabricator via llvm-commits]

fhahn created this revision.
fhahn added reviewers: t.p.northover, ab, dmgreen, paquette.
Herald added subscribers: hiraditya, kristof.beyls.
fhahn requested review of this revision.
Herald added a project: LLVM.

Try to re-use an already extended operand for SetCC with v16i8 operands
feeding na extended select. Doing so avoids requiring another full
extension of the SET_CC result when lowering the select.

This improves lowering for certain extend/cmp/select patterns operating
on v16i8, by replacing 6 instructions for the extra extension with 4
separate selects.

This improves the generated code for loops like the one below in
combination with D96522 <https://reviews.llvm.org/D96522>.

  int foo(uint8_t *p, int N) {
    unsigned long long sum = 0;
    for (int i = 0; i < N ; i++, p++) {
      unsigned int v = *p;
      sum += (v < 127) ? v : 256 - v;
    }
    return sum;
  }

https://clang.godbolt.org/z/Wco866MjY

On the AArch64 cores I have access to, the patch improves performance of
the vector loop by ~10%.

This could be generalized per follow-ups, but the initial version
targets one of the more important cases in combination with D96522 <https://reviews.llvm.org/D96522>.

Alive2 model:

- unsigned: https://alive2.llvm.org/ce/z/iEwLU3
- signed: https://alive2.llvm.org/ce/z/aMBega


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D120481

Files:
  llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
  llvm/test/CodeGen/AArch64/vselect-ext.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D120481.411110.patch
Type: text/x-patch
Size: 5843 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220224/81a91169/attachment.bin>