Index: lib/CodeGen/SelectionDAG/DAGCombiner.cpp
===================================================================
--- lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -16088,6 +16088,16 @@
   if (Op1->isInvariant() && Op0->writeMem())
     return false;
 
+  unsigned NumBytes0 = Op0->getMemoryVT().getSizeInBits() >> 3;
+  unsigned NumBytes1 = Op1->getMemoryVT().getSizeInBits() >> 3;
+
+  // Check for BaseIndexOffset matching.
+  BaseIndexOffset BasePtr0 = BaseIndexOffset::match(Op0->getBasePtr(), DAG);
+  BaseIndexOffset BasePtr1 = BaseIndexOffset::match(Op1->getBasePtr(), DAG);
+  if (BasePtr0.equalBaseIndex(BasePtr1))
+    return !((BasePtr0.Offset + NumBytes0 <= BasePtr1.Offset) ||
+             (BasePtr1.Offset + NumBytes1 <= BasePtr0.Offset));
+
   // Gather base node and offset information.
   SDValue Base0, Base1;
   int64_t Offset0, Offset1;
@@ -16099,8 +16109,6 @@
                                       Base1, Offset1, GV1, CV1);
 
   // If they have the same base address, then check to see if they overlap.
-  unsigned NumBytes0 = Op0->getMemoryVT().getSizeInBits() >> 3;
-  unsigned NumBytes1 = Op1->getMemoryVT().getSizeInBits() >> 3;
   if (Base0 == Base1 || (GV0 && (GV0 == GV1)) || (CV0 && (CV0 == CV1)))
     return !((Offset0 + NumBytes0) <= Offset1 ||
              (Offset1 + NumBytes1) <= Offset0);
Index: test/CodeGen/X86/i256-add.ll
===================================================================
--- test/CodeGen/X86/i256-add.ll
+++ test/CodeGen/X86/i256-add.ll
@@ -12,34 +12,35 @@
 ; X32-NEXT:    subl $12, %esp
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl 8(%ecx), %edx
-; X32-NEXT:    movl (%ecx), %ebx
-; X32-NEXT:    movl 4(%ecx), %edi
+; X32-NEXT:    movl 8(%ecx), %edi
+; X32-NEXT:    movl (%ecx), %edx
+; X32-NEXT:    movl 4(%ecx), %ebx
 ; X32-NEXT:    movl 28(%eax), %esi
 ; X32-NEXT:    movl %esi, {{[0-9]+}}(%esp) # 4-byte Spill
 ; X32-NEXT:    movl 24(%eax), %ebp
-; X32-NEXT:    addl (%eax), %ebx
-; X32-NEXT:    adcl 4(%eax), %edi
-; X32-NEXT:    adcl 8(%eax), %edx
+; X32-NEXT:    addl (%eax), %edx
 ; X32-NEXT:    movl %edx, {{[0-9]+}}(%esp) # 4-byte Spill
-; X32-NEXT:    movl 20(%eax), %esi
+; X32-NEXT:    adcl 4(%eax), %ebx
+; X32-NEXT:    adcl 8(%eax), %edi
+; X32-NEXT:    movl %edi, (%esp) # 4-byte Spill
+; X32-NEXT:    movl 20(%eax), %edi
 ; X32-NEXT:    movl 12(%eax), %edx
-; X32-NEXT:    movl 16(%eax), %eax
+; X32-NEXT:    movl 16(%eax), %esi
 ; X32-NEXT:    adcl 12(%ecx), %edx
-; X32-NEXT:    adcl 16(%ecx), %eax
-; X32-NEXT:    adcl 20(%ecx), %esi
-; X32-NEXT:    adcl 24(%ecx), %ebp
-; X32-NEXT:    movl %ebp, (%esp) # 4-byte Spill
+; X32-NEXT:    adcl 16(%ecx), %esi
+; X32-NEXT:    adcl 20(%ecx), %edi
+; X32-NEXT:    movl %ebp, %eax
+; X32-NEXT:    adcl 24(%ecx), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebp # 4-byte Reload
 ; X32-NEXT:    adcl %ebp, 28(%ecx)
+; X32-NEXT:    movl (%esp), %ebp # 4-byte Reload
+; X32-NEXT:    movl %ebp, 8(%ecx)
+; X32-NEXT:    movl %ebx, 4(%ecx)
+; X32-NEXT:    movl {{[0-9]+}}(%esp), %ebx # 4-byte Reload
 ; X32-NEXT:    movl %ebx, (%ecx)
-; X32-NEXT:    movl %edi, 4(%ecx)
-; X32-NEXT:    movl {{[0-9]+}}(%esp), %edi # 4-byte Reload
-; X32-NEXT:    movl %edi, 8(%ecx)
 ; X32-NEXT:    movl %edx, 12(%ecx)
-; X32-NEXT:    movl %eax, 16(%ecx)
-; X32-NEXT:    movl %esi, 20(%ecx)
-; X32-NEXT:    movl (%esp), %eax # 4-byte Reload
+; X32-NEXT:    movl %esi, 16(%ecx)
+; X32-NEXT:    movl %edi, 20(%ecx)
 ; X32-NEXT:    movl %eax, 24(%ecx)
 ; X32-NEXT:    addl $12, %esp
 ; X32-NEXT:    popl %esi
@@ -58,9 +59,9 @@
 ; X64-NEXT:    adcq 8(%rsi), %rdx
 ; X64-NEXT:    adcq 16(%rsi), %rax
 ; X64-NEXT:    adcq %r8, 24(%rdi)
-; X64-NEXT:    movq %rcx, (%rdi)
-; X64-NEXT:    movq %rdx, 8(%rdi)
 ; X64-NEXT:    movq %rax, 16(%rdi)
+; X64-NEXT:    movq %rdx, 8(%rdi)
+; X64-NEXT:    movq %rcx, (%rdi)
 ; X64-NEXT:    retq
   %a = load i256, i256* %p
   %b = load i256, i256* %q
@@ -96,9 +97,9 @@
 ; X32-NEXT:    sbbl 24(%esi), %eax
 ; X32-NEXT:    movl 28(%esi), %esi
 ; X32-NEXT:    sbbl %esi, 28(%ecx)
-; X32-NEXT:    movl %ebx, (%ecx)
-; X32-NEXT:    movl %ebp, 4(%ecx)
 ; X32-NEXT:    movl %edi, 8(%ecx)
+; X32-NEXT:    movl %ebp, 4(%ecx)
+; X32-NEXT:    movl %ebx, (%ecx)
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %esi # 4-byte Reload
 ; X32-NEXT:    movl %esi, 12(%ecx)
 ; X32-NEXT:    movl (%esp), %esi # 4-byte Reload
@@ -122,9 +123,9 @@
 ; X64-NEXT:    sbbq 8(%rsi), %rdx
 ; X64-NEXT:    sbbq 16(%rsi), %rax
 ; X64-NEXT:    sbbq %r8, 24(%rdi)
-; X64-NEXT:    movq %rcx, (%rdi)
-; X64-NEXT:    movq %rdx, 8(%rdi)
 ; X64-NEXT:    movq %rax, 16(%rdi)
+; X64-NEXT:    movq %rdx, 8(%rdi)
+; X64-NEXT:    movq %rcx, (%rdi)
 ; X64-NEXT:    retq
   %a = load i256, i256* %p
   %b = load i256, i256* %q
Index: test/CodeGen/X86/merge_store.ll
===================================================================
--- test/CodeGen/X86/merge_store.ll
+++ test/CodeGen/X86/merge_store.ll
@@ -29,17 +29,8 @@
   ret void
 }
 
-
-
 ;; CHECK-LABEL: indexed-store-merge
-
-;; We should be able to merge the 4 consecutive stores.  
-;; FIXMECHECK: movl	$0, 2(%rsi,%rdi)
-
-;; CHECK: movb	$0, 2(%rsi,%rdi)
-;; CHECK: movb	$0, 3(%rsi,%rdi)
-;; CHECK: movb	$0, 4(%rsi,%rdi)
-;; CHECK: movb	$0, 5(%rsi,%rdi)
+;; CHECK: movl	$0, 2(%rsi,%rdi)
 ;; CHECK: movb	$0, (%rsi)
 define void @indexed-store-merge(i64 %p, i8* %v) {
 entry: