enhance memcpyopt to zap memcpy's that have the same src/dst.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@121362 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 6f93e326ba126df9ef960c6cee4502d57d322d50..4c487e0a345dc9f5047b230e5345d1ede65640c1 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -762,6 +762,14 @@ bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
   ConstantInt *CopySize = dyn_cast<ConstantInt>(M->getLength());
   if (CopySize == 0 || M->isVolatile()) return false;
 
+  // If the source and destination of the memcpy are the same, then zap it.
+  if (M->getSource() == M->getDest()) {
+    MD->removeInstruction(M);
+    M->eraseFromParent();
+    return false;
+  }
+  
+  
   // The are two possible optimizations we can do for memcpy:
   //   a) memcpy-memcpy xform which exposes redundance for DSE.
   //   b) call-memcpy xform for return slot optimization.
@@ -773,10 +781,11 @@ bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
     return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
     
   if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
-    bool changed = performCallSlotOptzn(M, M->getDest(), M->getSource(),
-                                        CopySize->getZExtValue(), C);
-    if (changed) M->eraseFromParent();
-    return changed;
+    if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
+                             CopySize->getZExtValue(), C)) {
+      M->eraseFromParent();
+      return true;
+    }
   }
   return false;
 }

diff --git a/test/Transforms/MemCpyOpt/memcpy.ll b/test/Transforms/MemCpyOpt/memcpy.ll
index 16b80a628a3c1dbbe070f61cecfb55c4490f10d5..b387d32a7d5e8319082715feb698744bb1ba601e 100644 (file)
--- a/test/Transforms/MemCpyOpt/memcpy.ll
+++ b/test/Transforms/MemCpyOpt/memcpy.ll
@@ -100,3 +100,12 @@ entry:
   ; CHECK: store i8 4
   ; CHECK: call void @test5a(%struct.S* byval align 16 %y)
 }
+
+;; Noop memcpy should be zapped.
+define void @test6(i8 *%P) {
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %P, i64 8, i32 4, i1 false)
+  ret void
+; CHECK: @test6
+; CHECK-NEXT: ret void
+}
+