[all-commits] [llvm/llvm-project] 2ac5eb: [SLP]Do not emit extract elements for insertelemen...

[Andrew V. Tischenko via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 2ac5ebedeac438f26bb0439c25a0069fd72914b0
      https://github.com/llvm/llvm-project/commit/2ac5ebedeac438f26bb0439c25a0069fd72914b0
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2022-05-23 (Mon, 23 May 2022)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/loadorder.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/tsc-s116.ll
    M llvm/test/Transforms/SLPVectorizer/X86/buildvector-same-lane-insert.ll
    M llvm/test/Transforms/SLPVectorizer/X86/buildvector-shuffle.ll
    M llvm/test/Transforms/SLPVectorizer/X86/cmp-as-alternate-ops.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_7zip.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_bullet3.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_cmpop.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_exceed_scheduling.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_lencod.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_scheduling-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_scheduling.ll
    M llvm/test/Transforms/SLPVectorizer/X86/horizontal-list.ll
    M llvm/test/Transforms/SLPVectorizer/X86/horizontal-minmax.ll
    M llvm/test/Transforms/SLPVectorizer/X86/insert-shuffle.ll
    M llvm/test/Transforms/SLPVectorizer/X86/jumbled-load-multiuse.ll
    M llvm/test/Transforms/SLPVectorizer/X86/jumbled-load.ll
    M llvm/test/Transforms/SLPVectorizer/X86/jumbled_store_crash.ll
    M llvm/test/Transforms/SLPVectorizer/X86/load-merge-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/X86/load-merge.ll
    M llvm/test/Transforms/SLPVectorizer/X86/lookahead.ll
    M llvm/test/Transforms/SLPVectorizer/X86/ordering-bug.ll
    M llvm/test/Transforms/SLPVectorizer/X86/phi.ll
    M llvm/test/Transforms/SLPVectorizer/X86/pr42022-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/X86/pr42022.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reduction-same-vals.ll
    M llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reorder_diamond_match.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reordered-top-scalars.ll
    M llvm/test/Transforms/SLPVectorizer/X86/vec_list_bias-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/X86/vec_list_bias.ll
    M llvm/test/Transforms/SLPVectorizer/X86/vectorize-widest-phis.ll

  Log Message:
  -----------
  [SLP]Do not emit extract elements for insertelements users, replace with shuffles directly.

SLP vectorizer emits extracts for externally used vectorized scalars and
estimates the cost for each such extract. But in many cases these
scalars are input for insertelement instructions, forming buildvector,
and instead of extractelement/insertelement pair we can emit/cost
estimate shuffle(s) cost and generate series of shuffles, which can be
further optimized.

Tested using test-suite (+SPEC2017), the tests passed, SLP was able to
generate/vectorize more instructions in many cases and it allowed to reduce
number of re-vectorization attempts (where we could try to vectorize
buildector insertelements again and again).

Differential Revision: https://reviews.llvm.org/D107966