[all-commits] [llvm/llvm-project] cca173: [LoopInterchange] Identify unsafe instructions for...

[Ryotaro Kasuga via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: cca173885706f6c728b46f9035f034de4f39a4ba
      https://github.com/llvm/llvm-project/commit/cca173885706f6c728b46f9035f034de4f39a4ba
  Author: Ryotaro Kasuga <kasuga.ryotaro at fujitsu.com>
  Date:   2026-06-03 (Wed, 03 Jun 2026)

  Changed paths:
    M llvm/lib/Transforms/Scalar/LoopInterchange.cpp
    M llvm/test/Transforms/LoopInterchange/atomic-memory-ordering.ll
    M llvm/test/Transforms/LoopInterchange/call-instructions-remarks.ll
    M llvm/test/Transforms/LoopInterchange/invoke.ll

  Log Message:
  -----------
  [LoopInterchange] Identify unsafe instructions for interchange (#201402)

In LoopInterchange, all instructions in the loops are traversed and
checked during the legality phase, and the pass bails out if it finds an
instruction that is unsafe to interchange. However, previously it only
handled call instructions and ignored all others, leaving several kinds
of instruction that should have been detected but were not.
This patch fixes the issue by restructuring the legality check, in
particular replacing instruction‑specific handling with calling general
Instruction APIs.

Fixes #200913.



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