[PATCH] D57568: [SCEV] Don't bother preserving LCSSA in SCEV

Fri Feb 1 06:09:40 PST 2019

fhahn accepted this revision.
fhahn added a comment.
This revision is now accepted and ready to land.

LCSSA form is a property of the IR. As we do not change the IR here, there is no way we could destroy the LCSSA form of the underlying IR. If transformations use the extra info to destroy LCSSA form, they should be fixed.

LGTM, but please wait a bit with committing, in case the other reviewers have additional comments or I missed something.

================
Comment at: lib/Analysis/ScalarEvolution.cpp:5342

   // If the PHI has a single incoming value, follow that value, unless the
   // PHI's incoming blocks are in a different loop, in which case doing so
----------------
This comment is obsolete now.

================
Comment at: test/Transforms/LoopStrengthReduce/funclet.ll:48
 ; CHECK: cleanuppad within none []
-; CHECK-NEXT: ptrtoint i8* %phi2 to i32
+; CHECK-NEXT: br

----------------
not really related to this patch, but it is not really clear what gets checked here and elsewhere. I think it would be great if we could add a clear test case where this change is beneficial.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57568/new/

https://reviews.llvm.org/D57568