Teach CodeGenPrep to look past bitcast when it's duplicating return instruction

into predecessor blocks to enable tail call optimization.

rdar://11958338


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160894 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 34 insertions, 7 deletions

diff --git a/lib/Transforms/Scalar/CodeGenPrepare.cpp b/lib/Transforms/Scalar/CodeGenPrepare.cpp
index f425cd0..4b4a8c5 100644
--- a/lib/Transforms/Scalar/CodeGenPrepare.cpp
+++ b/lib/Transforms/Scalar/CodeGenPrepare.cpp
@@ -645,10 +645,18 @@ bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
   if (!TLI)
     return false;
 
+  PHINode *PN = 0;
+  BitCastInst *BCI = 0;
   Value *V = RI->getReturnValue();
-  PHINode *PN = V ? dyn_cast<PHINode>(V) : NULL;
-  if (V && !PN)
-    return false;
+  if (V) {
+    BCI = dyn_cast<BitCastInst>(V);
+    if (BCI)
+      V = BCI->getOperand(0);
+
+    PN = dyn_cast<PHINode>(V);
+    if (!PN)
+      return false;
+  }
 
   BasicBlock *BB = RI->getParent();
   if (PN && PN->getParent() != BB)
@@ -666,6 +674,9 @@ bool CodeGenPrepare::DupRetToEnableTailCallOpts(ReturnInst *RI) {
   if (PN) {
     BasicBlock::iterator BI = BB->begin();
     do { ++BI; } while (isa<DbgInfoIntrinsic>(BI));
+    if (&*BI == BCI)
+      // Also skip over the bitcast.
+      ++BI;
     if (&*BI != RI)
       return false;
   } else {
diff --git a/lib/Transforms/Utils/BasicBlockUtils.cpp b/lib/Transforms/Utils/BasicBlockUtils.cpp
index 5576432..2679b93 100644
--- a/lib/Transforms/Utils/BasicBlockUtils.cpp
+++ b/lib/Transforms/Utils/BasicBlockUtils.cpp
@@ -659,10 +659,26 @@ ReturnInst *llvm::FoldReturnIntoUncondBranch(ReturnInst *RI, BasicBlock *BB,
   // If the return instruction returns a value, and if the value was a
   // PHI node in "BB", propagate the right value into the return.
   for (User::op_iterator i = NewRet->op_begin(), e = NewRet->op_end();
-       i != e; ++i)
-    if (PHINode *PN = dyn_cast<PHINode>(*i))
-      if (PN->getParent() == BB)
-        *i = PN->getIncomingValueForBlock(Pred);
+       i != e; ++i) {
+    Value *V = *i;
+    Instruction *NewBC = 0;
+    if (BitCastInst *BCI = dyn_cast<BitCastInst>(V)) {
+      // Return value might be bitcasted. Clone and insert it before the
+      // return instruction.
+      V = BCI->getOperand(0);
+      NewBC = BCI->clone();
+      Pred->getInstList().insert(NewRet, NewBC);
+      *i = NewBC;
+    }
+    if (PHINode *PN = dyn_cast<PHINode>(V)) {
+      if (PN->getParent() == BB) {
+        if (NewBC)
+          NewBC->setOperand(0, PN->getIncomingValueForBlock(Pred));
+        else
+          *i = PN->getIncomingValueForBlock(Pred);
+      }
+    }
+  }
       
   // Update any PHI nodes in the returning block to realize that we no
   // longer branch to them.