[MemCpyOpt] Use the raw i8* dest when optimizing memset+memcpy.

MemIntrinsic::getDest() looks through pointer casts, and using it
directly when building the new GEP+memset results in stuff like:

  %0 = getelementptr i64* %p, i32 16
  %1 = bitcast i64* %0 to i8*
  call ..memset(i8* %1, ...)

instead of the correct:

  %0 = bitcast i64* %p to i8*
  %1 = getelementptr i8* %0, i32 16
  call ..memset(i8* %1, ...)

Instead, use getRawDest, which just gives you the i8* value.
While there, use the memcpy's dest, as it's live anyway.

In most cases, when the optimization triggers, the memset and memcpy
sizes are the same, so the built memset is 0-sized and eliminated.
The problem occurs when they're different.

Fixes a regression caused by r235232: PR23300.

llvm-svn: 235419

1 files changed, 2 insertions, 1 deletions

diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 563e49c0190..8465ffd5064 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -860,7 +860,8 @@ bool MemCpyOpt::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
   if (MemSet->getDest() != MemCpy->getDest())
     return false;
 
-  Value *Dest = MemSet->getDest();
+  // Use the same i8* dest as the memcpy, killing the memset dest if different.
+  Value *Dest = MemCpy->getRawDest();
   Value *DestSize = MemSet->getLength();
   Value *SrcSize = MemCpy->getLength();