[all-commits] [llvm/llvm-project] cdae90: [SLP]Pass operand info and context to TTI cost que...

[Alexey Bataev via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: cdae908ca061cc4bbf0d43a428107770102d460f
      https://github.com/llvm/llvm-project/commit/cdae908ca061cc4bbf0d43a428107770102d460f
  Author: Alexey Bataev <a.bataev at outlook.com>
  Date:   2026-09-21 (Mon, 21 Sep 2026)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    M llvm/lib/Transforms/Vectorize/SLPVectorizer/SLPMemoryUtils.cpp
    M llvm/test/Transforms/SLPVectorizer/AArch64/externally-used-copyables.ll
    M llvm/test/Transforms/SLPVectorizer/AMDGPU/elementwise-fma-operand1.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/non-commutative-second-arg-only-copyable.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/reduced-copyable-element.ll
    M llvm/test/Transforms/SLPVectorizer/RISCV/strided-loads-with-external-indices.ll
    M llvm/test/Transforms/SLPVectorizer/SystemZ/non-power-2-subvector-extract.ll
    M llvm/test/Transforms/SLPVectorizer/X86/phi-multi-same-nodes.ll

  Log Message:
  -----------
  [SLP]Pass operand info and context to TTI cost queries

Feed operand value info, the context instruction, cast context hints,
and the actual select condition predicate to the TTI cost queries, so
target discounts that depend on them (fused fmul/fma pricing, folded
read-modify-write, constant or uniform operands) apply to both scalar
and vector sides of the SLP cost model. Also make the reversed-store
query describe the stored value instead of the pointer, per the
getMemoryOpCost contract.

Assisted-by: Cursor

Reviewers: RKSimon, bababuck

Pull Request: https://github.com/llvm/llvm-project/pull/224931



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