[InstCombine] avoid infinite loop from shuffle-extract-insert sequence (PR30923)

Removing the limitation in visitInsertElementInst() causes several regressions
because we're not prepared to fold sequences of shuffles or inserts and extracts
separated by shuffles. Fixing that appears to be a difficult mission because we
are purposely trying to avoid creating shuffles with arbitrary shuffle masks
because some targets may choke on those.

https://llvm.org/bugs/show_bug.cgi?id=30923

llvm-svn: 286423

1 files changed, 8 insertions, 0 deletions

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
index 6c3aa3ad6c9..732a786ceb7 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -413,6 +413,14 @@ static void replaceExtractElements(InsertElementInst *InsElt,
   if (InsertionBlock != InsElt->getParent())
     return;
 
+  // TODO: This restriction matches the check in visitInsertElementInst() and
+  // prevents an infinite loop caused by not turning the extract/insert pair
+  // into a shuffle. We really should not need either check, but we're lacking
+  // folds for shufflevectors because we're afraid to generate shuffle masks
+  // that the backend can't handle.
+  if (InsElt->hasOneUse() && isa<InsertElementInst>(InsElt->user_back()))
+    return;
+
   auto *WideVec = new ShuffleVectorInst(ExtVecOp, UndefValue::get(ExtVecType),
                                         ConstantVector::get(ExtendMask));