[all-commits] [llvm/llvm-project] b10ecf: [SLP]Represent externally used values as original ...

[Alexey Bataev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: b10ecfa914dd1bc2013584917d0505ba5f15f75c
      https://github.com/llvm/llvm-project/commit/b10ecfa914dd1bc2013584917d0505ba5f15f75c
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2024-08-12 (Mon, 12 Aug 2024)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/external-non-inst-use.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/memory-runtime-checks.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/multiple_reduction.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/scalarization-overhead.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/shuffle-vectors-mask-size.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/slp-fma-loss.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose-inseltpoison.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/transpose.ll
    M llvm/test/Transforms/SLPVectorizer/AArch64/vectorize-free-extracts-inserts.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/mixed-extracts-types.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/strided-unsupported-type.ll
    M llvm/test/Transforms/SLPVectorizer/X86/PR32086.ll
    M llvm/test/Transforms/SLPVectorizer/X86/PR35628_1.ll
    M llvm/test/Transforms/SLPVectorizer/X86/PR39774.ll
    M llvm/test/Transforms/SLPVectorizer/X86/PR40310.ll
    M llvm/test/Transforms/SLPVectorizer/X86/compare-reduce.ll
    M llvm/test/Transforms/SLPVectorizer/X86/crash_vectorizeTree.ll
    M llvm/test/Transforms/SLPVectorizer/X86/cse_extractelement.ll
    M llvm/test/Transforms/SLPVectorizer/X86/diamond.ll
    M llvm/test/Transforms/SLPVectorizer/X86/external-user-instruction-minbitwidth.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extract-many-users-buildvector.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extract-subvector-long-input.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extract-vectorized-operand.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extract_in_tree_user.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extractelement-multi-register-use.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extractelement-single-use-many-nodes.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extractelements-vector-ops-shuffle.ll
    M llvm/test/Transforms/SLPVectorizer/X86/extracts-with-undefs.ll
    M llvm/test/Transforms/SLPVectorizer/X86/gathered-delayed-nodes-with-reused-user.ll
    M llvm/test/Transforms/SLPVectorizer/X86/gep-with-extractelement-many-users.ll
    M llvm/test/Transforms/SLPVectorizer/X86/geps-non-pow-2.ll
    M llvm/test/Transforms/SLPVectorizer/X86/horizontal-list.ll
    M llvm/test/Transforms/SLPVectorizer/X86/insertelement-uses-vectorized-index.ll
    M llvm/test/Transforms/SLPVectorizer/X86/lookahead.ll
    M llvm/test/Transforms/SLPVectorizer/X86/minbitwidth-drop-wrapping-flags.ll
    M llvm/test/Transforms/SLPVectorizer/X86/operandorder.ll
    M llvm/test/Transforms/SLPVectorizer/X86/ordering-bug.ll
    M llvm/test/Transforms/SLPVectorizer/X86/pr27163.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reduced-value-replace-extractelement.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reduction-value-in-tree.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reorder-node.ll
    M llvm/test/Transforms/SLPVectorizer/X86/reordering-single-phi.ll
    M llvm/test/Transforms/SLPVectorizer/X86/replaced-external-in-reduction.ll
    M llvm/test/Transforms/SLPVectorizer/X86/same-scalar-in-same-phi-extract.ll
    M llvm/test/Transforms/SLPVectorizer/X86/scalarization-overhead.ll
    M llvm/test/Transforms/SLPVectorizer/X86/slp-throttle.ll

  Log Message:
  -----------
  [SLP]Represent externally used values as original scalars, if profitable.

Currently SLP vectorizer tries to keep only GEPs as scalar, if they are
vectorized but used externally. Same approach can be used for all scalar
values. This patch tries to keep original scalars if all its operands
remain scalar or externally used, the cost of the original scalar is
lower than the cost of the extractelement instruction, or if the number
of externally used scalars in the same entry is power of 2. Last
criterion allows better revectorization for multiply used scalars.

Reviewers: RKSimon

Reviewed By: RKSimon

Pull Request: https://github.com/llvm/llvm-project/pull/100904



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications