Index: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
===================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -627,13 +627,39 @@
   // Ensure that the value being stored is something that can be memset'able a
   // byte at a time like "0" or "-1" or any width, as well as things like
   // 0xA0A0A0A0 and 0.0.
-  if (Value *ByteVal = isBytewiseValue(SI->getOperand(0)))
+  auto *V = SI->getOperand(0);
+  if (Value *ByteVal = isBytewiseValue(V)) {
     if (Instruction *I = tryMergingIntoMemset(SI, SI->getPointerOperand(),
                                               ByteVal)) {
       BBI = I->getIterator(); // Don't invalidate iterator.
       return true;
     }
 
+    // If we have an aggregate, we try to promote it to memset regardless
+    // of opportunity for merging as it can expose optimization opportunities
+    // in subsequent passes.
+    auto *T = V->getType();
+    if (T->isAggregateType()) {
+      uint64_t Size = DL.getTypeStoreSize(T);
+      unsigned Align = SI->getAlignment();
+      if (!Align)
+        Align = DL.getABITypeAlignment(T);
+      IRBuilder<> Builder(SI);
+      auto *M = Builder.CreateMemSet(SI->getPointerOperand(), ByteVal,
+                                     Size, Align, SI->isVolatile());
+
+      DEBUG(dbgs() << "Promoting " << *SI << " to " << *M << "\n");
+
+      MD->removeInstruction(SI);
+      SI->eraseFromParent();
+      NumMemSetInfer++;
+
+      // Make sure we do not invalidate the iterator.
+      BBI = M->getIterator();
+      return true;
+    }
+  }
+
   return false;
 }
 
Index: llvm/trunk/test/Transforms/MemCpyOpt/fca2memcpy.ll
===================================================================
--- llvm/trunk/test/Transforms/MemCpyOpt/fca2memcpy.ll
+++ llvm/trunk/test/Transforms/MemCpyOpt/fca2memcpy.ll
@@ -3,7 +3,7 @@
 target datalayout = "e-i64:64-f80:128-n8:16:32:64"
 target triple = "x86_64-unknown-linux-gnu"
 
-%S = type { i8*, i32 }
+%S = type { i8*, i8, i32 }
 
 define void @copy(%S* %src, %S* %dst) {
 ; CHECK-LABEL: copy
@@ -37,8 +37,10 @@
 
 define void @destroysrc(%S* %src, %S* %dst) {
 ; CHECK-LABEL: destroysrc
-; CHECK-NOT: call
-; CHECK: ret void
+; CHECK: load %S, %S* %src
+; CHECK: call void @llvm.memset.p0i8.i64
+; CHECK-NEXT: store %S %1, %S* %dst
+; CHECK-NEXT: ret void
   %1 = load %S, %S* %src
   store %S zeroinitializer, %S* %src
   store %S %1, %S* %dst
@@ -49,7 +51,7 @@
 ; CHECK-LABEL: destroynoaliassrc
 ; CHECK-NOT: load
 ; CHECK: call void @llvm.memcpy.p0i8.p0i8.i64
-; CHECK-NEXT: store %S zeroinitializer, %S* %src
+; CHECK-NEXT: call void @llvm.memset.p0i8.i64
 ; CHECK-NEXT: ret void
   %1 = load %S, %S* %src
   store %S zeroinitializer, %S* %src