[all-commits] [llvm/llvm-project] fc0aac: [SLP]Improve analysis/emission of vector operands ...

[Andrew V. Tischenko via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: fc0aacf324b172309ee10334e2213e1b3df8ea45
      https://github.com/llvm/llvm-project/commit/fc0aacf324b172309ee10334e2213e1b3df8ea45
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2021-11-26 (Fri, 26 Nov 2021)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose.ll
    M llvm/test/Transforms/SLPVectorizer/X86/gather-move-out-of-loop.ll
    M llvm/test/Transforms/SLPVectorizer/X86/remark_alternate.ll

  Log Message:
  -----------
  [SLP]Improve analysis/emission of vector operands for alternate nodes.

Compiler has an analysis for perfect diamond matching but it does not
support nodes with main/alternate opcodes. The problem is that the
scalars themselves are different and might not match directly with other
nodes, but operands and main/alternate opcodes might match and compiler
might reuse some previously emitted vector instructions. Need to include
this analysis in the cost model and actual vector instructions emission
process.

Differential Revision: https://reviews.llvm.org/D114101