Actually check memcpy lengths, instead of just commenting about

how they should be checked.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@123999 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 0d3c5c7bdc985bcc29fb421b7ecba750ee1063a5..acddf0878b0edd348713db9c48e35b89a6a4e9e1 100644 (file)
--- a/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -690,8 +690,10 @@ bool MemCpyOpt::processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep,
   
   // Second, the length of the memcpy's must be the same, or the preceeding one
   // must be larger than the following one.
-  ConstantInt *C1 = dyn_cast<ConstantInt>(MDep->getLength());
-  if (!C1) return false;
+  ConstantInt *MDepLen = dyn_cast<ConstantInt>(MDep->getLength());
+  ConstantInt *MLen = dyn_cast<ConstantInt>(M->getLength());
+  if (!MDepLen || !MLen || MDepLen->getZExtValue() < MLen->getZExtValue())
+    return false;
   
   AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
 

diff --git a/test/Transforms/MemCpyOpt/smaller.ll b/test/Transforms/MemCpyOpt/smaller.ll
new file mode 100644 (file)
index 0000000..1d35582
--- /dev/null
+++ b/test/Transforms/MemCpyOpt/smaller.ll
@@ -0,0 +1,28 @@
+; RUN: opt -memcpyopt -S < %s | FileCheck %s
+; rdar://8875553
+
+; Memcpyopt shouldn't optimize the second memcpy using the first
+; because the first has a smaller size.
+
+; CHECK: call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp, i8* getelementptr inbounds (%struct.s* @cell, i32 0, i32 0, i32 0), i32 16, i32 4, i1 false)
+
+target datalayout = "e-p:32:32:32"
+
+%struct.s = type { [11 x i8], i32 }
+
+@.str = private constant [11 x i8] c"0123456789\00"
+@cell = external global %struct.s
+
+declare void @check(%struct.s* byval %p) nounwind
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8* nocapture, i8* nocapture, i32, i32, i1) nounwind
+
+define void @foo() nounwind {
+entry:
+  %agg.tmp = alloca %struct.s, align 4
+  store i32 99, i32* getelementptr inbounds (%struct.s* @cell, i32 0, i32 1), align 4
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* getelementptr inbounds (%struct.s* @cell, i32 0, i32 0, i32 0), i8* getelementptr inbounds ([11 x i8]* @.str, i32 0, i32 0), i32 11, i32 1, i1 false)
+  %tmp = getelementptr inbounds %struct.s* %agg.tmp, i32 0, i32 0, i32 0
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* %tmp, i8* getelementptr inbounds (%struct.s* @cell, i32 0, i32 0, i32 0), i32 16, i32 4, i1 false)
+  call void @check(%struct.s* byval %agg.tmp)
+  ret void
+}