[PATCH] D81375: [InstCombine] Simplify boolean Phis with const inputs using CFG

[Max Kazantsev via Phabricator via llvm-commits]

mkazantsev marked an inline comment as done.
mkazantsev added inline comments.


================
Comment at: llvm/test/Transforms/PhaseOrdering/inlining-alignment-assumptions.ll:43
 ; ASSUMPTIONS-ON:       true1:
-; ASSUMPTIONS-ON-NEXT:    [[C_PR:%.*]] = phi i1 [ false, [[FALSE1]] ], [ true, [[TMP0:%.*]] ]
+; ASSUMPTIONS-ON-NEXT:    [[C_PR:%.*]] = phi i1 [ [[C]], [[FALSE1]] ], [ true, [[TMP0:%.*]] ]
 ; ASSUMPTIONS-ON-NEXT:    [[PTRINT:%.*]] = ptrtoint i64* [[PTR:%.*]] to i64
----------------
mkazantsev wrote:
> xbolva00 wrote:
> > Unwanted
> SimplifyCFG does something stupid here it seems. Not related to InstCombine.
Added a test on that:
```
./test/Transforms/SimplifyCFG/unprofitable-pr.ll
```
SimplifyCFG bases its choice on block size, and sometimes does something particularly stupid.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D81375/new/

https://reviews.llvm.org/D81375