[llvm-commits] [llvm] r142150 - in /llvm/trunk: lib/Transforms/Scalar/MemCpyOptimizer.cpp test/Transforms/MemCpyOpt/memcpy.ll

[Nick Lewycky]

Author: nicholas
Date: Sun Oct 16 15:13:32 2011
New Revision: 142150

URL: http://llvm.org/viewvc/llvm-project?rev=142150&view=rev
Log:
When looking for dependencies on the src pointer, scan the src pointer. Scanning
on the memcpy call will pull up other unrelated stuff. Fixes PR11142.

Modified:
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
    llvm/trunk/test/Transforms/MemCpyOpt/memcpy.ll

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=142150&r1=142149&r2=142150&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Sun Oct 16 15:13:32 2011
@@ -806,21 +806,26 @@
   //   a) memcpy-memcpy xform which exposes redundance for DSE.
   //   b) call-memcpy xform for return slot optimization.
   MemDepResult DepInfo = MD->getDependency(M);
-  if (!DepInfo.isClobber())
-    return false;
-  
-  if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(DepInfo.getInst()))
-    return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
-    
-  if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
-    if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
-                             CopySize->getZExtValue(), C)) {
-      MD->removeInstruction(M);
-      M->eraseFromParent();
-      return true;
+  if (DepInfo.isClobber()) {
+    if (CallInst *C = dyn_cast<CallInst>(DepInfo.getInst())) {
+      if (performCallSlotOptzn(M, M->getDest(), M->getSource(),
+                               CopySize->getZExtValue(), C)) {
+        MD->removeInstruction(M);
+        M->eraseFromParent();
+        return true;
+      }
     }
   }
   
+  AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
+  AliasAnalysis::Location SrcLoc = AA.getLocationForSource(M);
+  MemDepResult SrcDepInfo = MD->getPointerDependencyFrom(SrcLoc, true,
+                                                         M, M->getParent());
+  if (SrcDepInfo.isClobber()) {
+    if (MemCpyInst *MDep = dyn_cast<MemCpyInst>(SrcDepInfo.getInst()))
+      return processMemCpyMemCpyDependence(M, MDep, CopySize->getZExtValue());
+  }
+
   return false;
 }
 

Modified: llvm/trunk/test/Transforms/MemCpyOpt/memcpy.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/MemCpyOpt/memcpy.ll?rev=142150&r1=142149&r2=142150&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/MemCpyOpt/memcpy.ll (original)
+++ llvm/trunk/test/Transforms/MemCpyOpt/memcpy.ll Sun Oct 16 15:13:32 2011
@@ -59,7 +59,7 @@
   call void @llvm.memcpy.p0i8.p0i8.i32(i8* %agg.result2, i8* %x.01, i32 32, i32 16, i1 false)
   ret void
 ; CHECK: @test3
-; CHECK-NEXT: %agg.result2 = bitcast 
+; CHECK-NEXT: %agg.result1 = bitcast 
 ; CHECK-NEXT: call void @llvm.memcpy
 ; CHECK-NEXT: ret void
 }
@@ -130,3 +130,21 @@
 
 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
 
+; PR11142 - When looking for a memcpy-memcpy dependency, don't get stuck on
+; instructions between the memcpy's that only affect the destination pointer.
+ at test8.str = internal constant [7 x i8] c"ABCDEF\00"
+
+define void @test8() {
+; CHECK: test8
+  %A = tail call i8* @malloc(i32 10)
+  %B = getelementptr inbounds i8* %A, i64 2
+  tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %B, i8* getelementptr inbounds ([7 x i8]* @test8.str, i64 0, i64 0), i32 7, i32 1, i1 false)
+; CHECK: tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %D, i8* getelementptr
+  %C = tail call i8* @malloc(i32 10)
+  %D = getelementptr inbounds i8* %C, i64 2
+  tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %D, i8* %B, i32 7, i32 1, i1 false)
+; CHECK: tail call void @llvm.memcpy.p0i8.p0i8.i32(i8* %D, i8* getelementptr
+  ret void
+}
+
+declare noalias i8* @malloc(i32)