[all-commits] [llvm/llvm-project] a7993b: [LV] Don't skip VPlan cost model for div/rem instr...

[David Sherwood via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: a7993befd9a6c80ca1bc07ae5db2b644eae235a7
      https://github.com/llvm/llvm-project/commit/a7993befd9a6c80ca1bc07ae5db2b644eae235a7
  Author: David Sherwood <david.sherwood at arm.com>
  Date:   2026-04-14 (Tue, 14 Apr 2026)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
    M llvm/test/Transforms/LoopVectorize/AArch64/aarch64-predication.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/predication_costs.ll

  Log Message:
  -----------
  [LV] Don't skip VPlan cost model for div/rem instructions (#187056)

In LoopVectorizationPlanner::precomputeCosts we are skipping calculation
of costs using the VPlan cost model, instead preferring to use the
legacy costs. This helps to prevent the legacy and vplan cost model
assert firing, but really we should be encouraging full use of the VPlan
cost model.

I've created this initial PR to stop skipping the computation costs for
udiv/urem/sdiv/srem instructions. The VPlan costs seem to match up
nicely.

I intend to follow up with more PRs to move more opcodes across.



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