Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
@@ -160,10 +160,31 @@
 
 bool AMDGPUInstructionSelector::selectG_STORE(MachineInstr &I) const {
   MachineBasicBlock *BB = I.getParent();
+  MachineFunction *MF = BB->getParent();
+  MachineRegisterInfo &MRI = MF->getRegInfo();
   DebugLoc DL = I.getDebugLoc();
+  unsigned StoreSize = RBI.getSizeInBits(I.getOperand(0).getReg(), MRI, TRI);
+  unsigned Opcode;
 
   // FIXME: Select store instruction based on address space
-  MachineInstr *Flat = BuildMI(*BB, &I, DL, TII.get(AMDGPU::FLAT_STORE_DWORD))
+  switch (StoreSize) {
+  default:
+    return false;
+  case 32:
+    Opcode = AMDGPU::FLAT_STORE_DWORD;
+    break;
+  case 64:
+    Opcode = AMDGPU::FLAT_STORE_DWORDX2;
+    break;
+  case 96:
+    Opcode = AMDGPU::FLAT_STORE_DWORDX3;
+    break;
+  case 128:
+    Opcode = AMDGPU::FLAT_STORE_DWORDX4;
+    break;
+  }
+
+  MachineInstr *Flat = BuildMI(*BB, &I, DL, TII.get(Opcode))
           .add(I.getOperand(1))
           .add(I.getOperand(0))
           .addImm(0)  // offset
Index: llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -1579,6 +1579,12 @@
   case 64:
     return RB->getID() == AMDGPU::VGPRRegBankID ? &AMDGPU::VReg_64RegClass :
                                                    &AMDGPU::SReg_64_XEXECRegClass;
+  case 96:
+    return RB->getID() == AMDGPU::VGPRRegBankID ? &AMDGPU::VReg_96RegClass :
+                                                  nullptr;
+  case 128:
+    return RB->getID() == AMDGPU::VGPRRegBankID ? &AMDGPU::VReg_128RegClass :
+                                                  &AMDGPU::SReg_128RegClass;
   default:
     llvm_unreachable("not implemented");
   }
Index: llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
===================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
+++ llvm/trunk/test/CodeGen/AMDGPU/GlobalISel/inst-select-store-flat.mir
@@ -4,7 +4,10 @@
 # REQUIRES: global-isel
 
 --- |
-  define amdgpu_kernel void @global_addrspace(i32 addrspace(1)* %global0) { ret void }
+  define amdgpu_kernel void @global_addrspace(i32 addrspace(1)* %global0,
+                                              i64 addrspace(1)* %global1,
+                                              i96 addrspace(1)* %global2,
+                                              i128 addrspace(1)* %global3) { ret void }
 ...
 ---
 
@@ -14,16 +17,28 @@
 
 # GCN: global_addrspace
 # GCN: [[PTR:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
-# GCN: [[VAL:%[0-9]+]]:vgpr_32 = COPY $vgpr2
-# GCN: FLAT_STORE_DWORD [[PTR]], [[VAL]], 0, 0, 0
+# GCN: [[VAL4:%[0-9]+]]:vgpr_32 = COPY $vgpr2
+# GCN: [[VAL8:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4
+# GCN: [[VAL12:%[0-9]+]]:vreg_96 = COPY $vgpr5_vgpr6_vgpr7
+# GCN: [[VAL16:%[0-9]+]]:vreg_128 = COPY $vgpr8_vgpr9_vgpr10_vgpr11
+# GCN: FLAT_STORE_DWORD [[PTR]], [[VAL4]], 0, 0, 0
+# GCN: FLAT_STORE_DWORDX2 [[PTR]], [[VAL8]], 0, 0, 0
+# GCN: FLAT_STORE_DWORDX3 [[PTR]], [[VAL12]], 0, 0, 0
+# GCN: FLAT_STORE_DWORDX4 [[PTR]], [[VAL16]], 0, 0, 0
 
 body: |
   bb.0:
-    liveins:  $vgpr0_vgpr1, $vgpr2
+    liveins:  $vgpr0_vgpr1, $vgpr2, $vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10_vgpr11
 
     %0:vgpr(p1) = COPY $vgpr0_vgpr1
     %1:vgpr(s32) = COPY $vgpr2
+    %2:vgpr(s64) = COPY $vgpr3_vgpr4
+    %3:vgpr(s96) = COPY $vgpr5_vgpr6_vgpr7
+    %4:vgpr(s128) = COPY $vgpr8_vgpr9_vgpr10_vgpr11
     G_STORE %1, %0 :: (store 4 into %ir.global0)
+    G_STORE %2, %0 :: (store 8 into %ir.global1)
+    G_STORE %3, %0 :: (store 12 into %ir.global2, align 16)
+    G_STORE %4, %0 :: (store 16 into %ir.global3)
 
 ...
 ---