Index: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
===================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -346,6 +346,7 @@
                               uint64_t cpyLen, unsigned cpyAlign, CallInst *C);
     bool processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep,
                                        uint64_t MSize);
+    bool processMemSetMemCpyDependence(MemCpyInst *M, MemSetInst *MDep);
     bool processByValArgument(CallSite CS, unsigned ArgNo);
     Instruction *tryMergingIntoMemset(Instruction *I, Value *StartPtr,
                                       Value *ByteVal);
@@ -839,6 +840,53 @@
   return true;
 }
 
+/// We've found that the (upward scanning) memory dependence of \p MemCpy is
+/// \p MemSet.  Try to simplify \p MemSet to only set the trailing bytes that
+/// weren't copied over by \p MemCpy.
+///
+/// In other words, transform:
+/// \code
+///   memset(dst, c, dst_size);
+///   memcpy(dst, src, src_size);
+/// \endcode
+/// into:
+/// \code
+///   memcpy(dst, src, src_size);
+///   memset(dst + src_size, c, dst_size <= src_size ? 0 : dst_size - src_size);
+/// \endcode
+bool MemCpyOpt::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
+                                              MemSetInst *MemSet) {
+  // We can only transform memset/memcpy with the same destination.
+  if (MemSet->getDest() != MemCpy->getDest())
+    return false;
+
+  Value *Dest = MemSet->getDest();
+  Value *DestSize = MemSet->getLength();
+  Value *SrcSize = MemCpy->getLength();
+
+  // By default, create an unaligned memset.
+  unsigned Align = 1;
+  // If Dest is aligned, and SrcSize is constant, use the minimum alignment
+  // of the sum.
+  const unsigned DestAlign =
+      std::max(MemSet->getAlignment(), MemCpy->getAlignment());
+  if (DestAlign > 1)
+    if (ConstantInt *SrcSizeC = dyn_cast<ConstantInt>(SrcSize))
+      Align = MinAlign(SrcSizeC->getZExtValue(), DestAlign);
+
+  IRBuilder<> Builder(MemCpy->getNextNode());
+
+  Value *MemsetLen =
+      Builder.CreateSelect(Builder.CreateICmpULE(DestSize, SrcSize),
+                           ConstantInt::getNullValue(DestSize->getType()),
+                           Builder.CreateSub(DestSize, SrcSize));
+  Builder.CreateMemSet(Builder.CreateGEP(Dest, SrcSize), MemSet->getOperand(1),
+                       MemsetLen, Align);
+
+  MD->removeInstruction(MemSet);
+  MemSet->eraseFromParent();
+  return true;
+}
 
 /// processMemCpy - perform simplification of memcpy's.  If we have memcpy A
 /// which copies X to Y, and memcpy B which copies Y to Z, then we can rewrite
@@ -869,6 +917,17 @@
         return true;
       }
 
+  AliasAnalysis::Location SrcLoc = AliasAnalysis::getLocationForSource(M);
+  MemDepResult SrcDepInfo = MD->getPointerDependencyFrom(SrcLoc, true,
+                                                         M, M->getParent());
+
+  // Try to turn a partially redundant memset + memcpy into
+  // memcpy + smaller memset.  We don't need the memcpy size for this.
+  if (SrcDepInfo.isClobber())
+    if (MemSetInst *MDep = dyn_cast<MemSetInst>(SrcDepInfo.getInst()))
+      if (processMemSetMemCpyDependence(M, MDep))
+        return true;
+
   // The optimizations after this point require the memcpy size.
   ConstantInt *CopySize = dyn_cast<ConstantInt>(M->getLength());
   if (!CopySize) return false;
@@ -892,9 +951,6 @@
     }
   }
 
-  AliasAnalysis::Location SrcLoc = AliasAnalysis::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());
Index: llvm/trunk/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
===================================================================
--- llvm/trunk/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
+++ llvm/trunk/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll
@@ -0,0 +1,54 @@
+; RUN: opt -memcpyopt -S %s | FileCheck %s
+
+target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
+
+; CHECK-LABEL: define void @test
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %src_size, i32 1, i1 false)
+; CHECK-DAG: [[DST:%[0-9]+]] = getelementptr i8, i8* %dst, i64 %src_size
+; CHECK-DAG: [[ULE:%[0-9]+]] = icmp ule i64 %dst_size, %src_size
+; CHECK-DAG: [[SIZEDIFF:%[0-9]+]] = sub i64 %dst_size, %src_size
+; CHECK-DAG: [[SIZE:%[0-9]+]] = select i1 [[ULE]], i64 0, i64 [[SIZEDIFF]]
+; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* [[DST]], i8 0, i64 [[SIZE]], i32 1, i1 false)
+; CHECK-NEXT: ret void
+define void @test(i8* %src, i64 %src_size, i8* %dst, i64 %dst_size) {
+  call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 1, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 %src_size, i32 1, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: define void @test_align_same
+; CHECK: call void @llvm.memset.p0i8.i64(i8* {{.*}}, i8 0, i64 {{.*}}, i32 8, i1 false)
+define void @test_align_same(i8* %src, i8* %dst, i64 %dst_size) {
+  call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 8, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 80, i32 1, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: define void @test_align_min
+; CHECK: call void @llvm.memset.p0i8.i64(i8* {{.*}}, i8 0, i64 {{.*}}, i32 4, i1 false)
+define void @test_align_min(i8* %src, i8* %dst, i64 %dst_size) {
+  call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 8, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 36, i32 1, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: define void @test_align_memcpy
+; CHECK: call void @llvm.memset.p0i8.i64(i8* {{.*}}, i8 0, i64 {{.*}}, i32 8, i1 false)
+define void @test_align_memcpy(i8* %src, i8* %dst, i64 %dst_size) {
+  call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 1, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst, i8* %src, i64 80, i32 8, i1 false)
+  ret void
+}
+
+; CHECK-LABEL: define void @test_different_dst
+; CHECK-NEXT: call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 1, i1 false)
+; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src, i64 %src_size, i32 1, i1 false)
+; CHECK-NEXT: ret void
+define void @test_different_dst(i8* %dst2, i8* %src, i64 %src_size, i8* %dst, i64 %dst_size) {
+  call void @llvm.memset.p0i8.i64(i8* %dst, i8 0, i64 %dst_size, i32 1, i1 false)
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %dst2, i8* %src, i64 %src_size, i32 1, i1 false)
+  ret void
+}
+
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i32, i1)