[PATCH] D101507: [SimplifyCFG] Ignore llvm.assume when looking for side effects

Fri Apr 30 08:23:39 PDT 2021

tejohnson added a comment.

In D101507#2728127 <https://reviews.llvm.org/D101507#2728127>, @nikic wrote:

> LGTM, summary needs an update though.

Ack. fixing momentarily.

================
Comment at: llvm/test/Transforms/PhaseOrdering/d83507-knowledge-retention-bug.ll:3
+; RUN: opt -loop-rotate -instcombine -enable-knowledge-retention -S < %s  | FileCheck %s
+; RUN: opt -passes='loop(loop-rotate),instcombine' -enable-knowledge-retention -S < %s  | FileCheck %s

----------------
nikic wrote:
> Would it be possible to instead adjust the test to have a non-undef IV?
Indeed, I removed the undef and the test still works (confirmed by temporarily reverting its fix again). I went ahead and left my change to only invoke the 2 passes of interest though - it seems like this is probably better to avoid the test being affected by future changes to various unrelated optimization passes.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101507/new/

https://reviews.llvm.org/D101507