diff --git a/lib/Target/R600/AMDGPUISelDAGToDAG.cpp b/lib/Target/R600/AMDGPUISelDAGToDAG.cpp
index 968ca6c5413..36296f52e68 100644
--- a/lib/Target/R600/AMDGPUISelDAGToDAG.cpp
+++ b/lib/Target/R600/AMDGPUISelDAGToDAG.cpp
@@ -438,6 +438,31 @@ SDNode *AMDGPUDAGToDAGISel::Select(SDNode *N) {
     break;
   }
 
+  case ISD::STORE: {
+    // Handle i64 stores here for the same reason mentioned above for loads.
+    StoreSDNode *ST = cast<StoreSDNode>(N);
+    SDValue Value = ST->getValue();
+    if (Value.getValueType() != MVT::i64 || ST->isTruncatingStore())
+      break;
+
+    SDValue NewValue = CurDAG->getNode(ISD::BITCAST, SDLoc(N),
+                                      MVT::v2i32, Value);
+    SDValue NewStore = CurDAG->getStore(ST->getChain(), SDLoc(N), NewValue,
+                                        ST->getBasePtr(), ST->getMemOperand());
+
+    CurDAG->ReplaceAllUsesOfValueWith(SDValue(N, 0), NewStore);
+
+    if (NewValue.getOpcode() == ISD::BITCAST) {
+      Select(NewStore.getNode());
+      return SelectCode(NewValue.getNode());
+    }
+
+    // getNode() may fold the bitcast if its input was another bitcast.  If that
+    // happens we should only select the new store.
+    N = NewStore.getNode();
+    break;
+  }
+
   case AMDGPUISD::REGISTER_LOAD: {
     if (Subtarget->getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS)
       break;
diff --git a/lib/Target/R600/AMDGPUISelLowering.cpp b/lib/Target/R600/AMDGPUISelLowering.cpp
index d3aac9707ef..507e399b878 100644
--- a/lib/Target/R600/AMDGPUISelLowering.cpp
+++ b/lib/Target/R600/AMDGPUISelLowering.cpp
@@ -141,9 +141,6 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM,
   setOperationAction(ISD::STORE, MVT::v2f32, Promote);
   AddPromotedToType(ISD::STORE, MVT::v2f32, MVT::v2i32);
 
-  setOperationAction(ISD::STORE, MVT::i64, Promote);
-  AddPromotedToType(ISD::STORE, MVT::i64, MVT::v2i32);
-
   setOperationAction(ISD::STORE, MVT::v4f32, Promote);
   AddPromotedToType(ISD::STORE, MVT::v4f32, MVT::v4i32);
 
diff --git a/test/CodeGen/R600/unaligned-load-store.ll b/test/CodeGen/R600/unaligned-load-store.ll
index 1187ff246f6..54ac608e190 100644
--- a/test/CodeGen/R600/unaligned-load-store.ll
+++ b/test/CodeGen/R600/unaligned-load-store.ll
@@ -35,7 +35,8 @@ define void @unaligned_load_store_i32_global(i32 addrspace(1)* %p, i32 addrspace
 ; SI: ds_read_u8
 ; SI: ds_read_u8
 ; SI: ds_read_u8
-; SI: ds_write2_b32
+; SI: ds_write_b32
+; SI: ds_write_b32
 ; SI: s_endpgm
 define void @unaligned_load_store_i64_local(i64 addrspace(3)* %p, i64 addrspace(3)* %r) {
   %v = load i64 addrspace(3)* %p, align 1
@@ -52,7 +53,8 @@ define void @unaligned_load_store_i64_local(i64 addrspace(3)* %p, i64 addrspace(
 ; SI: buffer_load_ubyte
 ; SI: buffer_load_ubyte
 ; SI: buffer_load_ubyte
-; SI: buffer_store_dwordx2
+; SI: buffer_store_dword
+; SI: buffer_store_dword
 define void @unaligned_load_store_i64_global(i64 addrspace(1)* %p, i64 addrspace(1)* %r) {
   %v = load i64 addrspace(1)* %p, align 1
   store i64 %v, i64 addrspace(1)* %r, align 1