[llvm-commits] [llvm] r170018 - in /llvm/trunk: lib/CodeGen/SelectionDAG/SelectionDAG.cpp test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll

[Evan Cheng]

Author: evancheng
Date: Wed Dec 12 14:43:23 2012
New Revision: 170018

URL: http://llvm.org/viewvc/llvm-project?rev=170018&view=rev
Log:
Fix a logic bug in inline expansion of memcpy / memset with an overlapping
load / store pair. It's not legal to use a wider load than the size of
the remaining bytes if it's the first pair of load / store.

Added:
    llvm/trunk/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=170018&r1=170017&r2=170018&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Wed Dec 12 14:43:23 2012
@@ -3469,9 +3469,6 @@
 
   unsigned NumMemOps = 0;
   while (Size != 0) {
-    if (++NumMemOps > Limit)
-      return false;
-
     unsigned VTSize = VT.getSizeInBits() / 8;
     while (VTSize > Size) {
       // For now, only use non-vector load / store's for the left-over pieces.
@@ -3507,7 +3504,8 @@
       // FIXME: Only does this for 64-bit or more since we don't have proper
       // cost model for unaligned load / store.
       bool Fast;
-      if (AllowOverlap && VTSize >= 8 && NewVTSize < Size &&
+      if (NumMemOps && AllowOverlap &&
+          VTSize >= 8 && NewVTSize < Size &&
           TLI.allowsUnalignedMemoryAccesses(VT, &Fast) && Fast)
         VTSize = Size;
       else {
@@ -3516,6 +3514,9 @@
       }
     }
 
+    if (++NumMemOps > Limit)
+      return false;
+
     MemOps.push_back(VT);
     Size -= VTSize;
   }

Added: llvm/trunk/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll?rev=170018&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll (added)
+++ llvm/trunk/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll Wed Dec 12 14:43:23 2012
@@ -0,0 +1,11 @@
+; RUN: llc -march=mips64el -mcpu=mips64r2 < %s
+
+ at .str = private unnamed_addr constant [7 x i8] c"hello\0A\00", align 1
+
+define void @t(i8* %ptr) {
+entry:
+  tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %ptr, i8* getelementptr inbounds ([7 x i8]* @.str, i64 0, i64 0), i64 7, i32 1, i1 false)
+  ret void
+}
+
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind