AMDGPU/R600: Don't use REGISTER_{LOAD,STORE} ISD nodes

[jvesely]

Summary

This helps eliminate few redundant MOVs from private AS loads/stores
I think we'd need to be smarter about MMO for truncating store/extending loads to eliminate all of them.

This also makes it much easier to switch to MEM_{RD,WR}_SCRATCH once mesa bits are done.

[• tstellarAMD]

LGTM.

[jvesely]

Looks like I'll need to handle aliasing properly, before going ahead with this one.

[jvesely]

rebase and improve tests

[jvesely]

ping

[arsenm]

LGTM. Can the node definitions be deleted now?

lib/Target/AMDGPU/R600ISelLowering.cpp

1468

The dump should be removed from here

[jvesely]

In D24746#638053, @arsenm wrote:

LGTM. Can the node definitions be deleted now?

yes. I kept them in case they might be useful in the future, but they are not used by anything atm.

lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp

   if ((C = dyn_cast<ConstantSDNode>(Addr))) {
     Base = CurDAG->getRegister(AMDGPU::INDIRECT_BASE_ADDR, MVT::i32);
     Offset = CurDAG->getTargetConstant(C->getZExtValue(), DL, MVT::i32);
+  } else if ((Addr.getOpcode() == AMDGPUISD::DWORDADDR) &&
+             (C = dyn_cast<ConstantSDNode>(Addr.getOperand(0)))) {
+    Base = CurDAG->getRegister(AMDGPU::INDIRECT_BASE_ADDR, MVT::i32);
+    Offset = CurDAG->getTargetConstant(C->getZExtValue(), DL, MVT::i32);
   } else if ((Addr.getOpcode() == ISD::ADD || Addr.getOpcode() == ISD::OR) &&
             (C = dyn_cast<ConstantSDNode>(Addr.getOperand(1)))) {
     Base = Addr.getOperand(0);

lib/Target/AMDGPU/AMDGPUISelLowering.cpp

   setOperationAction(ISD::STORE, MVT::v2f64, Promote);
   AddPromotedToType(ISD::STORE, MVT::v2f64, MVT::v4i32);
 
-  setTruncStoreAction(MVT::v2i32, MVT::v2i8, Custom);
-  setTruncStoreAction(MVT::v2i32, MVT::v2i16, Custom);
-
-  setTruncStoreAction(MVT::v4i32, MVT::v4i8, Custom);
-  setTruncStoreAction(MVT::v4i32, MVT::v4i16, Expand);
-
-  setTruncStoreAction(MVT::v8i32, MVT::v8i16, Expand);
-  setTruncStoreAction(MVT::v16i32, MVT::v16i8, Expand);
-  setTruncStoreAction(MVT::v16i32, MVT::v16i16, Expand);
-
   setTruncStoreAction(MVT::i64, MVT::i1, Expand);
   setTruncStoreAction(MVT::i64, MVT::i8, Expand);
   setTruncStoreAction(MVT::i64, MVT::i16, Expand);

lib/Target/AMDGPU/R600ISelLowering.cpp

 
   setTruncStoreAction(MVT::i32, MVT::i8, Custom);
   setTruncStoreAction(MVT::i32, MVT::i16, Custom);
+  // We need to include these since trunc STORES to PRIVATE need
+  // special handling to accommodate RMW
+  setTruncStoreAction(MVT::v2i32, MVT::v2i16, Custom);
+  setTruncStoreAction(MVT::v4i32, MVT::v4i16, Custom);
+  setTruncStoreAction(MVT::v8i32, MVT::v8i16, Custom);
+  setTruncStoreAction(MVT::v16i32, MVT::v16i16, Custom);
+  setTruncStoreAction(MVT::v32i32, MVT::v32i16, Custom);
+  setTruncStoreAction(MVT::v2i32, MVT::v2i8, Custom);
+  setTruncStoreAction(MVT::v4i32, MVT::v4i8, Custom);
+  setTruncStoreAction(MVT::v8i32, MVT::v8i8, Custom);
+  setTruncStoreAction(MVT::v16i32, MVT::v16i8, Custom);
+  setTruncStoreAction(MVT::v32i32, MVT::v32i8, Custom);
 
   // Workaround for LegalizeDAG asserting on expansion of i1 vector stores.
   setTruncStoreAction(MVT::v2i32, MVT::v2i1, Expand);

 SDValue R600TargetLowering::lowerPrivateTruncStore(StoreSDNode *Store,
                                                    SelectionDAG &DAG) const {
   SDLoc DL(Store);
+  //TODO: Who creates the i8 stores?
+  assert(Store->isTruncatingStore()
+         || Store->getValue().getValueType() == MVT::i8);
+  assert(Store->getAddressSpace() == AMDGPUAS::PRIVATE_ADDRESS);
 
-  unsigned Mask = 0;
+  SDValue Mask;
   if (Store->getMemoryVT() == MVT::i8) {
-    Mask = 0xff;
+    assert(Store->getAlignment() >= 1);
+    Mask = DAG.getConstant(0xff, DL, MVT::i32);
   } else if (Store->getMemoryVT() == MVT::i16) {
-    Mask = 0xffff;
+    assert(Store->getAlignment() >= 2);
+    Mask = DAG.getConstant(0xffff, DL, MVT::i32);;
+  } else {
+    llvm_unreachable("Unsupported private trunc store");
   }
 
   SDValue Chain = Store->getChain();
   SDValue BasePtr = Store->getBasePtr();
+  SDValue Offset = Store->getOffset();
   EVT MemVT = Store->getMemoryVT();
 
-  SDValue Ptr = DAG.getNode(ISD::SRL, DL, MVT::i32, BasePtr,
-                            DAG.getConstant(2, DL, MVT::i32));
-  SDValue Dst = DAG.getNode(AMDGPUISD::REGISTER_LOAD, DL, MVT::i32,
-                            Chain, Ptr,
-                            DAG.getTargetConstant(0, DL, MVT::i32));
+  SDValue LoadPtr = BasePtr;
+  if (!Offset.isUndef()) {
+    LoadPtr = DAG.getNode(ISD::ADD, DL, MVT::i32, BasePtr, Offset);
+  }
+
+  // Get dword location
+  // TODO: this should be eliminated by the future SHR ptr, 2
+  SDValue Ptr = DAG.getNode(ISD::AND, DL, MVT::i32, LoadPtr,
+                            DAG.getConstant(0xfffffffc, DL, MVT::i32));
+
+  // Load dword
+  // TODO: can we be smarter about machine pointer info?
+  SDValue Dst = DAG.getLoad(MVT::i32, DL, Chain, Ptr, MachinePointerInfo());
 
-  SDValue ByteIdx = DAG.getNode(ISD::AND, DL, MVT::i32, BasePtr,
+  Chain = Dst.getValue(1);
+
+  // Get offset in dword
+  SDValue ByteIdx = DAG.getNode(ISD::AND, DL, MVT::i32, LoadPtr,
                                 DAG.getConstant(0x3, DL, MVT::i32));
 
+  // Convert byte offset to bit shift
   SDValue ShiftAmt = DAG.getNode(ISD::SHL, DL, MVT::i32, ByteIdx,
                                  DAG.getConstant(3, DL, MVT::i32));
 
+  // TODO: Contrary to the name of the functiom,
+  // it also handles sub i32 non-truncating stores (like i1)
   SDValue SExtValue = DAG.getNode(ISD::SIGN_EXTEND, DL, MVT::i32,
                                   Store->getValue());
 
+  // Mask the value to the right type
   SDValue MaskedValue = DAG.getZeroExtendInReg(SExtValue, DL, MemVT);
 
+  // Shift the value in place
   SDValue ShiftedValue = DAG.getNode(ISD::SHL, DL, MVT::i32,
                                      MaskedValue, ShiftAmt);
 
-  SDValue DstMask = DAG.getNode(ISD::SHL, DL, MVT::i32,
-                                DAG.getConstant(Mask, DL, MVT::i32),
-                                ShiftAmt);
-  DstMask = DAG.getNode(ISD::XOR, DL, MVT::i32, DstMask,
-                        DAG.getConstant(0xffffffff, DL, MVT::i32));
+  // Shift the mask in place
+  SDValue DstMask = DAG.getNode(ISD::SHL, DL, MVT::i32, Mask, ShiftAmt);
+
+  // Invert the mask. NOTE: if we had native ROL instructions we could
+  // use inverted mask
+  DstMask = DAG.getNOT(DL, DstMask, MVT::i32);
+
+  // Cleanup the target bits
   Dst = DAG.getNode(ISD::AND, DL, MVT::i32, Dst, DstMask);
 
+  // Add the new bits
   SDValue Value = DAG.getNode(ISD::OR, DL, MVT::i32, Dst, ShiftedValue);
-  return DAG.getNode(AMDGPUISD::REGISTER_STORE, DL, MVT::Other,
-                     Chain, Value, Ptr,
-                     DAG.getTargetConstant(0, DL, MVT::i32));
+
+  // Store dword
+  // TODO: Can we be smarter about MachinePointerInfo?
+  return DAG.getStore(Chain, DL, Value, Ptr, MachinePointerInfo());
 }
 
 SDValue R600TargetLowering::LowerSTORE(SDValue Op, SelectionDAG &DAG) const {
   StoreSDNode *StoreNode = cast<StoreSDNode>(Op);
   unsigned AS = StoreNode->getAddressSpace();
+
+  SDValue Chain = StoreNode->getChain();
+  SDValue Ptr = StoreNode->getBasePtr();
   SDValue Value = StoreNode->getValue();
-  EVT ValueVT = Value.getValueType();
+
+  EVT VT = Value.getValueType();
   EVT MemVT = StoreNode->getMemoryVT();
-  unsigned Align = StoreNode->getAlignment();
+  EVT PtrVT = Ptr.getValueType();
 
+  SDLoc DL(Op);
+
+  // Neither LOCAL nor PRIVATE can do vectors at the moment
   if ((AS == AMDGPUAS::LOCAL_ADDRESS || AS == AMDGPUAS::PRIVATE_ADDRESS) &&
-      ValueVT.isVector()) {
-    return SplitVectorStore(Op, DAG);
+      VT.isVector()) {
+    return scalarizeVectorStore(StoreNode, DAG);
   }
 
-  // Private AS needs special fixes
-  if (Align < MemVT.getStoreSize() && (AS != AMDGPUAS::PRIVATE_ADDRESS) &&
+  unsigned Align = StoreNode->getAlignment();
+  if (Align < MemVT.getStoreSize() &&
       !allowsMisalignedMemoryAccesses(MemVT, AS, Align, nullptr)) {
     return expandUnalignedStore(StoreNode, DAG);
   }
 
-  SDLoc DL(Op);
-  SDValue Chain = StoreNode->getChain();
-  SDValue Ptr = StoreNode->getBasePtr();
+  SDValue DWordAddr = DAG.getNode(ISD::SRL, DL, PtrVT, Ptr,
+                                  DAG.getConstant(2, DL, PtrVT));
 
   if (AS == AMDGPUAS::GLOBAL_ADDRESS) {
     // It is beneficial to create MSKOR here instead of combiner to avoid
     // artificial dependencies introduced by RMW
     if (StoreNode->isTruncatingStore()) {
-      EVT VT = Value.getValueType();
       assert(VT.bitsLE(MVT::i32));
       SDValue MaskConstant;
       if (MemVT == MVT::i8) {

         assert(StoreNode->getAlignment() >= 2);
         MaskConstant = DAG.getConstant(0xFFFF, DL, MVT::i32);
       }
-      SDValue DWordAddr = DAG.getNode(ISD::SRL, DL, VT, Ptr,
-                                      DAG.getConstant(2, DL, MVT::i32));
-      SDValue ByteIndex = DAG.getNode(ISD::AND, DL, Ptr.getValueType(), Ptr,
-                                      DAG.getConstant(0x00000003, DL, VT));
+
+      SDValue ByteIndex = DAG.getNode(ISD::AND, DL, PtrVT, Ptr,
+                                      DAG.getConstant(0x00000003, DL, PtrVT));
+      SDValue BitShift = DAG.getNode(ISD::SHL, DL, VT, ByteIndex,
+                                     DAG.getConstant(3, DL, VT));
+
+      // Put the mask in correct place
+      SDValue Mask = DAG.getNode(ISD::SHL, DL, VT, MaskConstant, BitShift);
+
+      // Put the mask in correct place
       SDValue TruncValue = DAG.getNode(ISD::AND, DL, VT, Value, MaskConstant);
-      SDValue Shift = DAG.getNode(ISD::SHL, DL, VT, ByteIndex,
-                                   DAG.getConstant(3, DL, VT));
-      SDValue ShiftedValue = DAG.getNode(ISD::SHL, DL, VT, TruncValue, Shift);
-      SDValue Mask = DAG.getNode(ISD::SHL, DL, VT, MaskConstant, Shift);
+      SDValue ShiftedValue = DAG.getNode(ISD::SHL, DL, VT, TruncValue, BitShift);
+
       // XXX: If we add a 64-bit ZW register class, then we could use a 2 x i32
       // vector instead.
       SDValue Src[4] = {

       return DAG.getMemIntrinsicNode(AMDGPUISD::STORE_MSKOR, DL,
                                      Op->getVTList(), Args, MemVT,
                                      StoreNode->getMemOperand());
-    } else if (Ptr->getOpcode() != AMDGPUISD::DWORDADDR &&
-               ValueVT.bitsGE(MVT::i32)) {
+    } else if (Ptr->getOpcode() != AMDGPUISD::DWORDADDR && VT.bitsGE(MVT::i32)) {
       // Convert pointer from byte address to dword address.
-      Ptr = DAG.getNode(AMDGPUISD::DWORDADDR, DL, Ptr.getValueType(),
-                        DAG.getNode(ISD::SRL, DL, Ptr.getValueType(),
-                                    Ptr, DAG.getConstant(2, DL, MVT::i32)));
+      Ptr = DAG.getNode(AMDGPUISD::DWORDADDR, DL, PtrVT, DWordAddr);
 
       if (StoreNode->isTruncatingStore() || StoreNode->isIndexed()) {
         llvm_unreachable("Truncated and indexed stores not supported yet");

     }
   }
 
+  // GLOBAL_ADDRESS has been handled above, LOCAL_ADDRESS allows all sizes
   if (AS != AMDGPUAS::PRIVATE_ADDRESS)
     return SDValue();
 
   if (MemVT.bitsLT(MVT::i32))
     return lowerPrivateTruncStore(StoreNode, DAG);
 
-  // Lowering for indirect addressing
-  const MachineFunction &MF = DAG.getMachineFunction();
-  const R600FrameLowering *TFL = getSubtarget()->getFrameLowering();
-  unsigned StackWidth = TFL->getStackWidth(MF);
-
-  Ptr = stackPtrToRegIndex(Ptr, StackWidth, DAG);
-
-  if (ValueVT.isVector()) {
-    unsigned NumElemVT = ValueVT.getVectorNumElements();
-    EVT ElemVT = ValueVT.getVectorElementType();
-    SmallVector<SDValue, 4> Stores(NumElemVT);
-
-    assert(NumElemVT >= StackWidth && "Stack width cannot be greater than "
-                                      "vector width in load");
-
-    for (unsigned i = 0; i < NumElemVT; ++i) {
-      unsigned Channel, PtrIncr;
-      getStackAddress(StackWidth, i, Channel, PtrIncr);
-      Ptr = DAG.getNode(ISD::ADD, DL, MVT::i32, Ptr,
-                        DAG.getConstant(PtrIncr, DL, MVT::i32));
-      SDValue Elem = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, ElemVT,
-                                 Value, DAG.getConstant(i, DL, MVT::i32));
-
-      Stores[i] = DAG.getNode(AMDGPUISD::REGISTER_STORE, DL, MVT::Other,
-                              Chain, Elem, Ptr,
-                              DAG.getTargetConstant(Channel, DL, MVT::i32));
-    }
-     Chain =  DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Stores);
-   } else {
-    if (ValueVT == MVT::i8) {
-      Value = DAG.getNode(ISD::ZERO_EXTEND, DL, MVT::i32, Value);
-    }
-    Chain = DAG.getNode(AMDGPUISD::REGISTER_STORE, DL, MVT::Other, Chain, Value, Ptr,
-    DAG.getTargetConstant(0, DL, MVT::i32)); // Channel
+  // Standard i32+ store, tag it with DWORDADDR to note that the address
+  // has been shifted
+  if (Ptr.getOpcode() != AMDGPUISD::DWORDADDR) {
+    Ptr = DAG.getNode(AMDGPUISD::DWORDADDR, DL, PtrVT, DWordAddr);
+    return DAG.getStore(Chain, DL, Value, Ptr, StoreNode->getMemOperand());
   }
 
-  return Chain;
+  // Tagged i32+ stores will be matched by patterns
+  return SDValue();
 }
 
 // return (512 + (kc_bank << 12)

   LoadSDNode *Load = cast<LoadSDNode>(Op);
   ISD::LoadExtType ExtType = Load->getExtensionType();
   EVT MemVT = Load->getMemoryVT();
+  assert(Load->getAlignment() >= MemVT.getStoreSize());
 
-  // <SI && AS=PRIVATE && EXTLOAD && size < 32bit,
-  // register (2-)byte extract.
+  SDValue BasePtr = Load->getBasePtr();
+  SDValue Chain = Load->getChain();
+  SDValue Offset = Load->getOffset();
 
-  // Get Register holding the target.
-  SDValue Ptr = DAG.getNode(ISD::SRL, DL, MVT::i32, Load->getBasePtr(),
-                            DAG.getConstant(2, DL, MVT::i32));
-  // Load the Register.
-  SDValue Ret = DAG.getNode(AMDGPUISD::REGISTER_LOAD, DL, Op.getValueType(),
-                            Load->getChain(),
-                            Ptr,
-                            DAG.getTargetConstant(0, DL, MVT::i32),
-                            Op.getOperand(2));
+  SDValue LoadPtr = BasePtr;
+  if (!Offset.isUndef()) {
+    LoadPtr = DAG.getNode(ISD::ADD, DL, MVT::i32, BasePtr, Offset);
+  }
+
+  // Get dword location
+  // NOTE: this should be eliminated by the future SHR ptr, 2
+  SDValue Ptr = DAG.getNode(ISD::AND, DL, MVT::i32, LoadPtr,
+                            DAG.getConstant(0xfffffffc, DL, MVT::i32));
+
+  // Load dword
+  // TODO: can we be smarter about machine pointer info?
+  SDValue Read = DAG.getLoad(MVT::i32, DL, Chain, Ptr, MachinePointerInfo());
 
   // Get offset within the register.
   SDValue ByteIdx = DAG.getNode(ISD::AND, DL, MVT::i32,
-                                Load->getBasePtr(),
-                                DAG.getConstant(0x3, DL, MVT::i32));
+                                LoadPtr, DAG.getConstant(0x3, DL, MVT::i32));
 
   // Bit offset of target byte (byteIdx * 8).
   SDValue ShiftAmt = DAG.getNode(ISD::SHL, DL, MVT::i32, ByteIdx,
                                  DAG.getConstant(3, DL, MVT::i32));
 
   // Shift to the right.
-  Ret = DAG.getNode(ISD::SRL, DL, MVT::i32, Ret, ShiftAmt);
+  SDValue Ret = DAG.getNode(ISD::SRL, DL, MVT::i32, Read, ShiftAmt);
 
   // Eliminate the upper bits by setting them to ...
   EVT MemEltVT = MemVT.getScalarType();
 
-  // ... ones.
-  if (ExtType == ISD::SEXTLOAD) {
+  if (ExtType == ISD::SEXTLOAD) { // ... ones.
     SDValue MemEltVTNode = DAG.getValueType(MemEltVT);
-
-    SDValue Ops[] = {
-      DAG.getNode(ISD::SIGN_EXTEND_INREG, DL, MVT::i32, Ret, MemEltVTNode),
-      Load->getChain()
-    };
-
-    return DAG.getMergeValues(Ops, DL);
+    Ret = DAG.getNode(ISD::SIGN_EXTEND_INREG, DL, MVT::i32, Ret, MemEltVTNode);
+  } else { // ... or zeros.
+    Ret = DAG.getZeroExtendInReg(Ret, DL, MemEltVT);
   }
 
-  // ... or zeros.
   SDValue Ops[] = {
-    DAG.getZeroExtendInReg(Ret, DL, MemEltVT),
-    Load->getChain()
+    Ret,
+    Read.getValue(1) // This should be our output chain
   };
 
   return DAG.getMergeValues(Ops, DL);

   SDValue Chain = LoadNode->getChain();
   SDValue Ptr = LoadNode->getBasePtr();
 
-  if (LoadNode->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS && VT.isVector()) {
-    SDValue MergedValues[2] = {
-      scalarizeVectorLoad(LoadNode, DAG),
-      Chain
-    };
-    return DAG.getMergeValues(MergedValues, DL);
+  if ((LoadNode->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS ||
+      LoadNode->getAddressSpace() == AMDGPUAS::PRIVATE_ADDRESS) &&
+      VT.isVector()) {
+      return scalarizeVectorLoad(LoadNode, DAG);
   }
 
   int ConstantBlock = ConstantAddressBlock(LoadNode->getAddressSpace());

     return DAG.getMergeValues(MergedValues, DL);
   }
 
-  SDValue LoweredLoad;
-
   // For most operations returning SDValue() will result in the node being
   // expanded by the DAG Legalizer. This is not the case for ISD::LOAD, so we
   // need to manually expand loads that may be legal in some address spaces and

     return SDValue();
   }
 
-  // Lowering for indirect addressing
-  const MachineFunction &MF = DAG.getMachineFunction();
-  const R600FrameLowering *TFL = getSubtarget()->getFrameLowering();
-  unsigned StackWidth = TFL->getStackWidth(MF);
-
-  Ptr = stackPtrToRegIndex(Ptr, StackWidth, DAG);
-
-  if (VT.isVector()) {
-    unsigned NumElemVT = VT.getVectorNumElements();
-    EVT ElemVT = VT.getVectorElementType();
-    SDValue Loads[4];
-
-    assert(NumElemVT <= 4);
-    assert(NumElemVT >= StackWidth && "Stack width cannot be greater than "
-                                      "vector width in load");
-
-    for (unsigned i = 0; i < NumElemVT; ++i) {
-      unsigned Channel, PtrIncr;
-      getStackAddress(StackWidth, i, Channel, PtrIncr);
-      Ptr = DAG.getNode(ISD::ADD, DL, MVT::i32, Ptr,
-                        DAG.getConstant(PtrIncr, DL, MVT::i32));
-      Loads[i] = DAG.getNode(AMDGPUISD::REGISTER_LOAD, DL, ElemVT,
-                             Chain, Ptr,
-                             DAG.getTargetConstant(Channel, DL, MVT::i32),
-                             Op.getOperand(2));
-    }
-    EVT TargetVT = EVT::getVectorVT(*DAG.getContext(), ElemVT, NumElemVT);
-    LoweredLoad = DAG.getBuildVector(TargetVT, DL, makeArrayRef(Loads, NumElemVT));
-  } else {
-    LoweredLoad = DAG.getNode(AMDGPUISD::REGISTER_LOAD, DL, VT,
-                              Chain, Ptr,
-                              DAG.getTargetConstant(0, DL, MVT::i32), // Channel
-                              Op.getOperand(2));
+  // DWORDADDR ISD marks already shifted address
+  if (Ptr.getOpcode() != AMDGPUISD::DWORDADDR) {
+    assert(VT == MVT::i32);
+    Ptr = DAG.getNode(ISD::SRL, DL, MVT::i32, Ptr, DAG.getConstant(2, DL, MVT::i32));
+    Ptr = DAG.getNode(AMDGPUISD::DWORDADDR, DL, MVT::i32, Ptr);
+    return DAG.getLoad(MVT::i32, DL, Chain, Ptr, LoadNode->getMemOperand());
   }
-
-  SDValue Ops[2] = {
-    LoweredLoad,
-    Chain
-  };
-
-  return DAG.getMergeValues(Ops, DL);
+  return SDValue();
 }
 
 SDValue R600TargetLowering::LowerBRCOND(SDValue Op, SelectionDAG &DAG) const {

lib/Target/AMDGPU/R600Instructions.td

 
 defm R600_ : RegisterLoadStore <R600_Reg32, FRAMEri, ADDRIndirect>;
 
+// Hardcode channel to 0
+// NOTE: LSHR is not available here. LSHR is per family instruction
+def : Pat <
+  (i32 (load_private ADDRIndirect:$addr) ),
+  (R600_RegisterLoad FRAMEri:$addr, (i32 0))
+>;
+def : Pat <
+  (store_private i32:$val, ADDRIndirect:$addr),
+  (R600_RegisterStore i32:$val, FRAMEri:$addr, (i32 0))
+>;
+
 
 //===----------------------------------------------------------------------===//
 // Pseudo instructions

lib/Target/AMDGPU/SIISelLowering.cpp

   setOperationAction(ISD::STORE, MVT::v16i32, Custom);
   setOperationAction(ISD::STORE, MVT::i1, Custom);
 
+  setTruncStoreAction(MVT::v2i32, MVT::v2i16, Expand);
+  setTruncStoreAction(MVT::v4i32, MVT::v4i16, Expand);
+  setTruncStoreAction(MVT::v8i32, MVT::v8i16, Expand);
+  setTruncStoreAction(MVT::v16i32, MVT::v16i16, Expand);
+  setTruncStoreAction(MVT::v32i32, MVT::v32i16, Expand);
+  setTruncStoreAction(MVT::v2i32, MVT::v2i8, Expand);
+  setTruncStoreAction(MVT::v4i32, MVT::v4i8, Expand);
+  setTruncStoreAction(MVT::v8i32, MVT::v8i8, Expand);
+  setTruncStoreAction(MVT::v16i32, MVT::v16i8, Expand);
+  setTruncStoreAction(MVT::v32i32, MVT::v32i8, Expand);
+
+
   setOperationAction(ISD::GlobalAddress, MVT::i32, Custom);
   setOperationAction(ISD::GlobalAddress, MVT::i64, Custom);
   setOperationAction(ISD::ConstantPool, MVT::v2i64, Expand);

test/CodeGen/AMDGPU/load-constant-i16.ll

 
 ; v2i16 is naturally 4 byte aligned
 ; EG: VTX_READ_32 [[DST:T[0-9]\.[XYZW]]], [[DST]], 0, #1
-; TODO: This should use DST, but for some there are redundant MOVs
-; EG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, literal
+; EG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], literal
+; EG: 16
 ; EG: 16
 define void @constant_zextload_v2i16_to_v2i32(<2 x i32> addrspace(1)* %out, <2 x i16> addrspace(2)* %in) #0 {
   %load = load <2 x i16>, <2 x i16> addrspace(2)* %in

 ; GCN-DAG: s_sext_i32_i16
 
 ; v2i16 is naturally 4 byte aligned
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XY, {{T[0-9].[XYZW]}},
 ; EG: VTX_READ_32 [[DST:T[0-9]\.[XYZW]]], [[DST]], 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; TODO: We should also use ASHR instead of LSHR + BFE
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].X, [[DST]], 0.0, literal
+; TODO: We should use ASHR instead of LSHR + BFE
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].Y, {{PV\.[XYZW]}}, 0.0, literal
 ; EG-DAG: 16
 ; EG-DAG: 16
 define void @constant_sextload_v2i16_to_v2i32(<2 x i32> addrspace(1)* %out, <2 x i16> addrspace(2)* %in) #0 {

   ret void
 }
 
-; FUNC-LABEL: {{^}}constant_constant_zextload_v3i16_to_v3i32:
+; FUNC-LABEL: {{^}}constant_zextload_v3i16_to_v3i32:
 ; GCN: s_load_dwordx2
 
 ; v3i16 is naturally 8 byte aligned
-; EG-DAG: VTX_READ_32 [[DST_HI:T[0-9]\.[XYZW]]], [[DST_HI]], 0, #1
-; EG-DAG: VTX_READ_16 [[DST_LO:T[0-9]\.[XYZW]]], [[DST_LO]], 4, #1
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XY, {{T[0-9].[XYZW]}},
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].X, {{T[0-9].[XYZW]}},
+; EG: CF_END
+; EG-DAG: VTX_READ_32 [[DST_LO:T[0-9]\.[XYZW]]], {{T[0-9]\.[XYZW]}}, 0, #1
+; EG-DAG: VTX_READ_16 [[DST_HI:T[0-9]\.[XYZW]]], {{T[0-9]\.[XYZW]}}, 4, #1
 ; TODO: This should use DST, but for some there are redundant MOVs
-; EG: LSHR {{[* ]*}}{{T[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG: 16
-define void @constant_constant_zextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(2)* %in) {
+; EG-DAG: LSHR {{[* ]*}}[[ST_LO]].Y, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: 16
+; EG-DAG: AND_INT {{[* ]*}}[[ST_LO]].X, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST_HI]].X, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: 65535
+; EG-DAG: 65535
+define void @constant_zextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(2)* %in) {
 entry:
   %ld = load <3 x i16>, <3 x i16> addrspace(2)* %in
   %ext = zext <3 x i16> %ld to <3 x i32>

   ret void
 }
 
-; FUNC-LABEL: {{^}}constant_constant_sextload_v3i16_to_v3i32:
+; FUNC-LABEL: {{^}}constant_sextload_v3i16_to_v3i32:
 ; GCN: s_load_dwordx2
 
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XY, {{T[0-9].[XYZW]}},
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].X, {{T[0-9].[XYZW]}},
 ; v3i16 is naturally 8 byte aligned
-; EG-DAG: VTX_READ_32 [[DST_HI:T[0-9]\.[XYZW]]], [[DST_HI]], 0, #1
-; EG-DAG: VTX_READ_16 [[DST_LO:T[0-9]\.[XYZW]]], [[DST_LO]], 4, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: ASHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, 0.0, literal
+; EG-DAG: VTX_READ_32 [[DST_HI:T[0-9]\.[XYZW]]], [[PTR:T[0-9]\.[XYZW]]], 0, #1
+; EG-DAG: VTX_READ_16 [[DST_LO:T[0-9]\.[XYZW]]], {{T[0-9]\.[XYZW]}}, 4, #1
+; EG-DAG: ASHR {{[* ]*}}[[ST_LO]].Y, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_LO]].X, {{T[0-9]\.[XYZW]}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_HI]].X, {{T[0-9]\.[XYZW]}}, 0.0, literal
 ; EG-DAG: 16
 ; EG-DAG: 16
-define void @constant_constant_sextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(2)* %in) {
+define void @constant_sextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(2)* %in) {
 entry:
   %ld = load <3 x i16>, <3 x i16> addrspace(2)* %in
   %ext = sext <3 x i16> %ld to <3 x i32>

   ret void
 }
 
-; FUNC-LABEL: {{^}}constant_constant_zextload_v4i16_to_v4i32:
+; FUNC-LABEL: {{^}}constant_zextload_v4i16_to_v4i32:
 ; GCN: s_load_dwordx2
 ; GCN-DAG: s_and_b32
 ; GCN-DAG: s_lshr_b32
 
 ; v4i16 is naturally 8 byte aligned
-; EG: VTX_READ_64 [[DST:T[0-9]\.XY]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, literal
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XYZW, {{T[0-9].[XYZW]}}
+; EG: VTX_READ_64 [[LD:T[0-9]]].XY, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: This should use LD, but for some there are redundant MOVs
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST]].Y, {{.*\.[XYZW]}}, literal
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST]].W, {{.*\.[XYZW]}}, literal
 ; EG-DAG: 16
-; EG-DAG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG-DAG: AND_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
 ; EG-DAG: 16
-define void @constant_constant_zextload_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i16> addrspace(2)* %in) #0 {
+; EG-DAG: AND_INT {{[* ]*}}[[ST]].X, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST]].Z, {{T[0-9]\.[XYZW]}}, literal
+; EG-DAG: 65535
+; EG-DAG: 65535
+define void @constant_zextload_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i16> addrspace(2)* %in) #0 {
   %load = load <4 x i16>, <4 x i16> addrspace(2)* %in
   %ext = zext <4 x i16> %load to <4 x i32>
   store <4 x i32> %ext, <4 x i32> addrspace(1)* %out

 ; GCN-DAG: s_sext_i32_i16
 
 ; v4i16 is naturally 8 byte aligned
-; EG: VTX_READ_64 [[DST:T[0-9]\.XY]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EG: VTX_READ_64 [[DST:T[0-9]]].XY, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: This should use LD, but for some there are redundant MOVs
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].X, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].Z, {{.*}}, 0.0, literal
 ; TODO: We should use ASHR instead of LSHR + BFE
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].Y, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST]].W, {{.*}}, 0.0, literal
 ; EG-DAG: 16
 ; EG-DAG: 16
 ; EG-DAG: 16

 ; GCN-DAG: s_lshr_b32
 
 ; v8i16 is naturally 16 byte aligned
-; EG: VTX_READ_128 [[DST:T[0-9]\.XYZW]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].XYZW, {{T[0-9]+.[XYZW]}},
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XYZW, {{T[0-9]+.[XYZW]}},
+; EG: VTX_READ_128 [[DST:T[0-9]]].XYZW, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: These should use LSHR instead of BFE_UINT
+; TODO: This should use DST, but for some there are redundant MOVs
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST_LO]].Y, {{.*}}, literal
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST_LO]].W, {{.*}}, literal
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST_HI]].Y, {{.*}}, literal
+; EG-DAG: BFE_UINT {{[* ]*}}[[ST_HI]].W, {{.*}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST_LO]].X, {{.*}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST_LO]].Z, {{.*}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST_HI]].X, {{.*}}, literal
+; EG-DAG: AND_INT {{[* ]*}}[[ST_HI]].Z, {{.*}}, literal
 ; EG-DAG: 16
 ; EG-DAG: 16
 ; EG-DAG: 16
 ; EG-DAG: 16
+; EG-DAG: 65535
+; EG-DAG: 65535
+; EG-DAG: 65535
+; EG-DAG: 65535
 define void @constant_zextload_v8i16_to_v8i32(<8 x i32> addrspace(1)* %out, <8 x i16> addrspace(2)* %in) #0 {
   %load = load <8 x i16>, <8 x i16> addrspace(2)* %in
   %ext = zext <8 x i16> %load to <8 x i32>

 ; GCN-DAG: s_sext_i32_i16
 
 ; v8i16 is naturally 16 byte aligned
-; EG: VTX_READ_128 [[DST:T[0-9]\.XYZW]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; TODO: We should use ASHR instead of LSHR + BFE
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].XYZW, {{T[0-9]+.[XYZW]}},
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XYZW, {{T[0-9]+.[XYZW]}},
+; EG: VTX_READ_128 [[DST:T[0-9]]].XYZW, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: 4 of these should use ASHR instead of LSHR + BFE_INT
+; TODO: This should use DST, but for some there are redundant MOVs
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_LO]].Y, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_LO]].W, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_HI]].Y, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_HI]].W, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_LO]].X, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_LO]].Z, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_HI]].X, {{.*}}, 0.0, literal
+; EG-DAG: BFE_INT {{[* ]*}}[[ST_HI]].Z, {{.*}}, 0.0, literal
 ; EG-DAG: 16
 ; EG-DAG: 16
 ; EG-DAG: 16

 
 ; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 ; EG: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
-; TODO: Why not 15 ?
+; TODO: These could be expanded earlier using ASHR 15
 ; EG: 31
 define void @constant_sextload_i16_to_i64(i64 addrspace(1)* %out, i16 addrspace(2)* %in) #0 {
   %a = load i16, i16 addrspace(2)* %in

 
 ; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 ; EG: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
-; TODO: Why not 15 ?
+; TODO: These could be expanded earlier using ASHR 15
 ; EG: 31
 define void @constant_sextload_v1i16_to_v1i64(<1 x i64> addrspace(1)* %out, <1 x i16> addrspace(2)* %in) #0 {
   %load = load <1 x i16>, <1 x i16> addrspace(2)* %in

test/CodeGen/AMDGPU/load-global-i16.ll

 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-SI,FUNC %s
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-HSA,FUNC %s
 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GCN-NOHSA,GCN-NOHSA-VI,FUNC %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
-; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=EGCM -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=CM -check-prefix=EGCM -check-prefix=FUNC %s
 
 ; FIXME: r600 is broken because the bigger testcases spill and it's not implemented
 

 ; GCN-NOHSA: buffer_load_ushort v{{[0-9]+}}
 ; GCN-HSA: flat_load_ushort
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 define void @global_load_i16(i16 addrspace(1)* %out, i16 addrspace(1)* %in) {
 entry:
   %ld = load i16, i16 addrspace(1)* %in

 ; GCN-NOHSA: buffer_load_dword v
 ; GCN-HSA: flat_load_dword v
 
-; EG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 define void @global_load_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <2 x i16>, <2 x i16> addrspace(1)* %in

 ; GCN-NOHSA: buffer_load_dwordx2 v
 ; GCN-HSA: flat_load_dwordx2 v
 
-; EG-DAG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 4, #1
+; EGCM-DAG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 4, #1
 define void @global_load_v3i16(<3 x i16> addrspace(1)* %out, <3 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <3 x i16>, <3 x i16> addrspace(1)* %in

 ; GCN-NOHSA: buffer_load_dwordx2
 ; GCN-HSA: flat_load_dwordx2
 
-; EG: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
 define void @global_load_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <4 x i16>, <4 x i16> addrspace(1)* %in

 ; GCN-NOHSA: buffer_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
 define void @global_load_v8i16(<8 x i16> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <8 x i16>, <8 x i16> addrspace(1)* %in

 ; GCN-HSA: flat_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
 define void @global_load_v16i16(<16 x i16> addrspace(1)* %out, <16 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <16 x i16>, <16 x i16> addrspace(1)* %in

 ; GCN-HSA: flat_load_ushort
 ; GCN-HSA: flat_store_dword
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 define void @global_zextload_i16_to_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
   %a = load i16, i16 addrspace(1)* %in
   %ext = zext i16 %a to i32

 ; GCN-HSA: flat_load_sshort
 ; GCN-HSA: flat_store_dword
 
-; EG: VTX_READ_16 [[DST:T[0-9]\.[XYZW]]], T{{[0-9]+}}.X, 0, #1
-; EG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], 0.0, literal
-; EG: 16
+; EGCM: VTX_READ_16 [[DST:T[0-9]\.[XYZW]]], T{{[0-9]+}}.X, 0, #1
+; EGCM: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], 0.0, literal
+; EGCM: 16
 define void @global_sextload_i16_to_i32(i32 addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
   %a = load i16, i16 addrspace(1)* %in
   %ext = sext i16 %a to i32

 ; GCN-NOHSA: buffer_load_ushort
 ; GCN-HSA: flat_load_ushort
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 define void @global_zextload_v1i16_to_v1i32(<1 x i32> addrspace(1)* %out, <1 x i16> addrspace(1)* %in) #0 {
   %load = load <1 x i16>, <1 x i16> addrspace(1)* %in
   %ext = zext <1 x i16> %load to <1 x i32>

 ; GCN-NOHSA: buffer_load_sshort
 ; GCN-HSA: flat_load_sshort
 
-; EG: VTX_READ_16 [[DST:T[0-9]\.[XYZW]]], T{{[0-9]+}}.X, 0, #1
-; EG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], 0.0, literal
-; EG: 16
+; EGCM: VTX_READ_16 [[DST:T[0-9]\.[XYZW]]], T{{[0-9]+}}.X, 0, #1
+; EGCM: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], 0.0, literal
+; EGCM: 16
 define void @global_sextload_v1i16_to_v1i32(<1 x i32> addrspace(1)* %out, <1 x i16> addrspace(1)* %in) #0 {
   %load = load <1 x i16>, <1 x i16> addrspace(1)* %in
   %ext = sext <1 x i16> %load to <1 x i32>

 ; GCN-NOHSA: buffer_load_dword
 ; GCN-HSA: flat_load_dword
 
-; EG: VTX_READ_32 [[DST:T[0-9]\.[XYZW]]], [[DST]], 0, #1
-; TODO: This should use DST, but for some there are redundant MOVs
-; EG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, literal
-; EG: 16
+; EGCM: VTX_READ_32 [[DST:T[0-9]\.[XYZW]]], [[DST]], 0, #1
+; EGCM: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, [[DST]], literal
+; EGCM: 16
 define void @global_zextload_v2i16_to_v2i32(<2 x i32> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
   %load = load <2 x i16>, <2 x i16> addrspace(1)* %in
   %ext = zext <2 x i16> %load to <2 x i32>

 
 ; GCN-HSA: flat_load_dword
 
-; EG: VTX_READ_32 [[DST:T[0-9]\.[XYZW]]], [[DST]], 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; TODO: We should also use ASHR instead of LSHR + BFE
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, 0.0, literal
-; EG-DAG: 16
-; EG-DAG: 16
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XY, {{T[0-9]\.[XYZW]}},
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EGCM: VTX_READ_32 [[DST:T[0-9].[XYZW]]], [[DST]], 0, #1
+; TODO: This should use ASHR instead of LSHR + BFE
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].X, [[DST]], 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].Y, {{PV.[XYZW]}}, 0.0, literal
+; EGCM-DAG: 16
+; EGCM-DAG: 16
 define void @global_sextload_v2i16_to_v2i32(<2 x i32> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
   %load = load <2 x i16>, <2 x i16> addrspace(1)* %in
   %ext = sext <2 x i16> %load to <2 x i32>

   ret void
 }
 
-; FUNC-LABEL: {{^}}global_global_zextload_v3i16_to_v3i32:
+; FUNC-LABEL: {{^}}global_zextload_v3i16_to_v3i32:
 ; GCN-NOHSA: buffer_load_dwordx2
 ; GCN-HSA: flat_load_dwordx2
 
-; EG-DAG: VTX_READ_32 [[DST_HI:T[0-9]\.[XYZW]]], [[DST_HI]], 0, #1
-; EG-DAG: VTX_READ_16 [[DST_LO:T[0-9]\.[XYZW]]], [[DST_LO]], 4, #1
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST_HI:T[0-9]]].X, {{T[0-9]\.[XYZW]}}
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST_LO:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].X, {{T[0-9]\.[XYZW]}},
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XY, {{T[0-9]\.[XYZW]}},
+; EGCM-DAG: VTX_READ_32 [[DST_LO:T[0-9]\.[XYZW]]], {{T[0-9]\.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_16 [[DST_HI:T[0-9]\.[XYZW]]], {{T[0-9]\.[XYZW]}}, 4, #1
 ; TODO: This should use DST, but for some there are redundant MOVs
-; EG: LSHR {{[* ]*}}{{T[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG: 16
-define void @global_global_zextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(1)* %in) {
+; EGCM: LSHR {{[* ]*}}[[ST_LO]].Y, {{T[0-9]\.[XYZW]}}, literal
+; EGCM: 16
+; EGCM: AND_INT {{[* ]*}}[[ST_LO]].X, {{T[0-9]\.[XYZW]}}, literal
+; EGCM: AND_INT {{[* ]*}}[[ST_HI]].X, [[DST_HI]], literal
+define void @global_zextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <3 x i16>, <3 x i16> addrspace(1)* %in
   %ext = zext <3 x i16> %ld to <3 x i32>

   ret void
 }
 
-; FUNC-LABEL: {{^}}global_global_sextload_v3i16_to_v3i32:
+; FUNC-LABEL: {{^}}global_sextload_v3i16_to_v3i32:
 ; GCN-NOHSA: buffer_load_dwordx2
 ; GCN-HSA: flat_load_dwordx2
 
-; EG-DAG: VTX_READ_32 [[DST_HI:T[0-9]\.[XYZW]]], [[DST_HI]], 0, #1
-; EG-DAG: VTX_READ_16 [[DST_LO:T[0-9]\.[XYZW]]], [[DST_LO]], 4, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: ASHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{PV.[XYZW]}}, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, 0.0, literal
-; EG-DAG: 16
-; EG-DAG: 16
-define void @global_global_sextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(1)* %in) {
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST_HI:T[0-9]]].X, {{T[0-9]\.[XYZW]}}
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST_LO:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].X, {{T[0-9]\.[XYZW]}},
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XY, {{T[0-9]\.[XYZW]}},
+; EGCM-DAG: VTX_READ_32 [[DST_LO:T[0-9]\.[XYZW]]], {{T[0-9].[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_16 [[DST_HI:T[0-9]\.[XYZW]]], {{T[0-9].[XYZW]}}, 4, #1
+; TODO: This should use DST, but for some there are redundant MOVs
+; EGCM-DAG: ASHR {{[* ]*}}[[ST_LO]].Y, {{T[0-9]\.[XYZW]}}, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_LO]].X, {{T[0-9]\.[XYZW]}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_HI]].X, [[DST_HI]], 0.0, literal
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+define void @global_sextload_v3i16_to_v3i32(<3 x i32> addrspace(1)* %out, <3 x i16> addrspace(1)* %in) {
 entry:
   %ld = load <3 x i16>, <3 x i16> addrspace(1)* %in
   %ext = sext <3 x i16> %ld to <3 x i32>

   ret void
 }
 
-; FUNC-LABEL: {{^}}global_global_zextload_v4i16_to_v4i32:
+; FUNC-LABEL: {{^}}global_zextload_v4i16_to_v4i32:
 ; GCN-NOHSA: buffer_load_dwordx2
 
 ; GCN-HSA: flat_load_dwordx2
 
-; EG: VTX_READ_64 [[DST:T[0-9]\.XY]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG-DAG: 16
-; EG-DAG: BFE_UINT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG-DAG: AND_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{T[0-9].[XYZW]}}, literal
-; EG-DAG: 16
-define void @global_global_zextload_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) #0 {
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EGCM: VTX_READ_64 [[DST:T[0-9]]].XY, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: This should use DST, but for some there are redundant MOVs
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST]].Y, {{.*}}, literal
+; EGCM-DAG: 16
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST]].W, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST]].X, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST]].Z, {{.*}}, literal
+; EGCM-DAG: 16
+define void @global_zextload_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) #0 {
   %load = load <4 x i16>, <4 x i16> addrspace(1)* %in
   %ext = zext <4 x i16> %load to <4 x i32>
   store <4 x i32> %ext, <4 x i32> addrspace(1)* %out

 
 ; GCN-HSA: flat_load_dwordx2
 
-; EG: VTX_READ_64 [[DST:T[0-9]\.XY]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
+; CM: MEM_RAT_CACHELESS STORE_DWORD [[ST:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG: MEM_RAT_CACHELESS STORE_RAW [[ST:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EGCM: VTX_READ_64 [[DST:T[0-9]]].XY, {{T[0-9].[XYZW]}}, 0, #1
 ; TODO: We should use ASHR instead of LSHR + BFE
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: BFE_INT {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, 0.0, literal
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
+; TODO: This should use DST, but for some there are redundant MOVs
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].X, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].Y, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].Z, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST]].W, {{.*}}, 0.0, literal
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
 define void @global_sextload_v4i16_to_v4i32(<4 x i32> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) #0 {
   %load = load <4 x i16>, <4 x i16> addrspace(1)* %in
   %ext = sext <4 x i16> %load to <4 x i32>

 ; GCN-NOHSA: buffer_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG: VTX_READ_128 [[DST:T[0-9]\.XYZW]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
+; CM-DAG: MEM_RAT_CACHELESS STORE_DWORD [[ST_LO:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; CM-DAG: MEM_RAT_CACHELESS STORE_DWORD [[ST_HI:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EGCM: CF_END
+; EGCM: VTX_READ_128 [[DST:T[0-9]]].XYZW, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: These should use LSHR instead of BFE_UINT
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST_LO]].Y, {{.*}}, literal
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST_LO]].W, {{.*}}, literal
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST_HI]].Y, {{.*}}, literal
+; EGCM-DAG: BFE_UINT {{[* ]*}}[[ST_HI]].W, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST_LO]].X, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST_LO]].Z, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST_HI]].X, {{.*}}, literal
+; EGCM-DAG: AND_INT {{[* ]*}}[[ST_HI]].Z, {{.*}}, literal
+; EGCM-DAG: 65535
+; EGCM-DAG: 65535
+; EGCM-DAG: 65535
+; EGCM-DAG: 65535
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
 define void @global_zextload_v8i16_to_v8i32(<8 x i32> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) #0 {
   %load = load <8 x i16>, <8 x i16> addrspace(1)* %in
   %ext = zext <8 x i16> %load to <8 x i32>

 ; GCN-NOHSA: buffer_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG: VTX_READ_128 [[DST:T[0-9]\.XYZW]], {{T[0-9].[XYZW]}}, 0, #1
-; TODO: These should use DST, but for some there are redundant MOVs
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: LSHR {{[* ]*}}T{{[0-9].[XYZW]}}, {{.*}}, literal
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
-; EG-DAG: 16
+; CM-DAG: MEM_RAT_CACHELESS STORE_DWORD [[ST_LO:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; CM-DAG: MEM_RAT_CACHELESS STORE_DWORD [[ST_HI:T[0-9]]], {{T[0-9]\.[XYZW]}}
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_LO:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EG-DAG: MEM_RAT_CACHELESS STORE_RAW [[ST_HI:T[0-9]]].XYZW, {{T[0-9]\.[XYZW]}},
+; EGCM: CF_END
+; EGCM: VTX_READ_128 [[DST:T[0-9]]].XYZW, {{T[0-9].[XYZW]}}, 0, #1
+; TODO: These should use ASHR instead of LSHR + BFE_INT
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_LO]].Y, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_LO]].W, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_HI]].Y, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_HI]].W, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_LO]].X, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_LO]].Z, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_HI]].X, {{.*}}, 0.0, literal
+; EGCM-DAG: BFE_INT {{[* ]*}}[[ST_HI]].Z, {{.*}}, 0.0, literal
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
+; EGCM-DAG: 16
 define void @global_sextload_v8i16_to_v8i32(<8 x i32> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) #0 {
   %load = load <8 x i16>, <8 x i16> addrspace(1)* %in
   %ext = sext <8 x i16> %load to <8 x i32>

 ; GCN-HSA: flat_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
 define void @global_zextload_v16i16_to_v16i32(<16 x i32> addrspace(1)* %out, <16 x i16> addrspace(1)* %in) #0 {
   %load = load <16 x i16>, <16 x i16> addrspace(1)* %in
   %ext = zext <16 x i16> %load to <16 x i32>

 
 ; FUNC-LABEL: {{^}}global_sextload_v16i16_to_v16i32:
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
 define void @global_sextload_v16i16_to_v16i32(<16 x i32> addrspace(1)* %out, <16 x i16> addrspace(1)* %in) #0 {
   %load = load <16 x i16>, <16 x i16> addrspace(1)* %in
   %ext = sext <16 x i16> %load to <16 x i32>

 ; GCN-HSA: flat_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
 define void @global_zextload_v32i16_to_v32i32(<32 x i32> addrspace(1)* %out, <32 x i16> addrspace(1)* %in) #0 {
   %load = load <32 x i16>, <32 x i16> addrspace(1)* %in
   %ext = zext <32 x i16> %load to <32 x i32>

 ; GCN-HSA: flat_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
 define void @global_sextload_v32i16_to_v32i32(<32 x i32> addrspace(1)* %out, <32 x i16> addrspace(1)* %in) #0 {
   %load = load <32 x i16>, <32 x i16> addrspace(1)* %in
   %ext = sext <32 x i16> %load to <32 x i32>

 ; GCN-HSA: flat_load_dwordx4
 ; GCN-HSA: flat_load_dwordx4
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 64, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 80, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 96, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 112, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 64, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 80, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 96, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 112, #1
 define void @global_zextload_v64i16_to_v64i32(<64 x i32> addrspace(1)* %out, <64 x i16> addrspace(1)* %in) #0 {
   %load = load <64 x i16>, <64 x i16> addrspace(1)* %in
   %ext = zext <64 x i16> %load to <64 x i32>

 
 ; FUNC-LABEL: {{^}}global_sextload_v64i16_to_v64i32:
 
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 64, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 80, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 96, #1
-; EG-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 112, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 0, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 16, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 32, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 48, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 64, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 80, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 96, #1
+; EGCM-DAG: VTX_READ_128 {{T[0-9]+\.XYZW}}, {{T[0-9]+.[XYZW]}}, 112, #1
 define void @global_sextload_v64i16_to_v64i32(<64 x i32> addrspace(1)* %out, <64 x i16> addrspace(1)* %in) #0 {
   %load = load <64 x i16>, <64 x i16> addrspace(1)* %in
   %ext = sext <64 x i16> %load to <64 x i32>

 ; GCN-NOHSA: buffer_store_dwordx2 v{{\[}}[[LO]]:[[HI]]]
 ; GCN-HSA: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[LO]]:[[HI]]{{\]}}
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
-; EG: MOV {{.*}}, 0.0
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: MOV {{.*}}, 0.0
 define void @global_zextload_i16_to_i64(i64 addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
   %a = load i16, i16 addrspace(1)* %in
   %ext = zext i16 %a to i64

 ; GCN-NOHSA: buffer_store_dwordx2 v{{\[}}[[LO]]:[[HI]]]
 ; GCN-HSA: flat_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{\[}}[[LO]]:[[HI]]{{\]}}
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
-; EG: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
-; TODO: Why not 15 ?
-; EG: 31
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
+; TODO: These could be expanded earlier using ASHR 15
+; EGCM: 31
 define void @global_sextload_i16_to_i64(i64 addrspace(1)* %out, i16 addrspace(1)* %in) #0 {
   %a = load i16, i16 addrspace(1)* %in
   %ext = sext i16 %a to i64

 
 ; FUNC-LABEL: {{^}}global_zextload_v1i16_to_v1i64:
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
-; EG: MOV {{.*}}, 0.0
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: MOV {{.*}}, 0.0
 define void @global_zextload_v1i16_to_v1i64(<1 x i64> addrspace(1)* %out, <1 x i16> addrspace(1)* %in) #0 {
   %load = load <1 x i16>, <1 x i16> addrspace(1)* %in
   %ext = zext <1 x i16> %load to <1 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v1i16_to_v1i64:
 
-; EG: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
-; EG: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
-; TODO: Why not 15 ?
-; EG: 31
+; EGCM: VTX_READ_16 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: ASHR {{\**}} {{T[0-9]\.[XYZW]}}, {{.*}}, literal
+; TODO: These could be expanded earlier using ASHR 15
+; EGCM: 31
 define void @global_sextload_v1i16_to_v1i64(<1 x i64> addrspace(1)* %out, <1 x i16> addrspace(1)* %in) #0 {
   %load = load <1 x i16>, <1 x i16> addrspace(1)* %in
   %ext = sext <1 x i16> %load to <1 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v2i16_to_v2i64:
 
-; EG: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_32 T{{[0-9]+}}.X, T{{[0-9]+}}.X, 0, #1
 define void @global_sextload_v2i16_to_v2i64(<2 x i64> addrspace(1)* %out, <2 x i16> addrspace(1)* %in) #0 {
   %load = load <2 x i16>, <2 x i16> addrspace(1)* %in
   %ext = sext <2 x i16> %load to <2 x i64>

 
 ; FUNC-LABEL: {{^}}global_zextload_v4i16_to_v4i64:
 
-; EG: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
 define void @global_zextload_v4i16_to_v4i64(<4 x i64> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) #0 {
   %load = load <4 x i16>, <4 x i16> addrspace(1)* %in
   %ext = zext <4 x i16> %load to <4 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v4i16_to_v4i64:
 
-; EG: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_64 T{{[0-9]+}}.XY, T{{[0-9]+}}.X, 0, #1
 define void @global_sextload_v4i16_to_v4i64(<4 x i64> addrspace(1)* %out, <4 x i16> addrspace(1)* %in) #0 {
   %load = load <4 x i16>, <4 x i16> addrspace(1)* %in
   %ext = sext <4 x i16> %load to <4 x i64>

 
 ; FUNC-LABEL: {{^}}global_zextload_v8i16_to_v8i64:
 
-; EG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
 define void @global_zextload_v8i16_to_v8i64(<8 x i64> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) #0 {
   %load = load <8 x i16>, <8 x i16> addrspace(1)* %in
   %ext = zext <8 x i16> %load to <8 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v8i16_to_v8i64:
 
-; EG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
 define void @global_sextload_v8i16_to_v8i64(<8 x i64> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) #0 {
   %load = load <8 x i16>, <8 x i16> addrspace(1)* %in
   %ext = sext <8 x i16> %load to <8 x i64>

 
 ; FUNC-LABEL: {{^}}global_zextload_v16i16_to_v16i64:
 
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
 define void @global_zextload_v16i16_to_v16i64(<16 x i64> addrspace(1)* %out, <16 x i16> addrspace(1)* %in) #0 {
   %load = load <16 x i16>, <16 x i16> addrspace(1)* %in
   %ext = zext <16 x i16> %load to <16 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v16i16_to_v16i64:
 
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
 define void @global_sextload_v16i16_to_v16i64(<16 x i64> addrspace(1)* %out, <16 x i16> addrspace(1)* %in) #0 {
   %load = load <16 x i16>, <16 x i16> addrspace(1)* %in
   %ext = sext <16 x i16> %load to <16 x i64>

 
 ; FUNC-LABEL: {{^}}global_zextload_v32i16_to_v32i64:
 
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 32, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 48, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 32, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 48, #1
 define void @global_zextload_v32i16_to_v32i64(<32 x i64> addrspace(1)* %out, <32 x i16> addrspace(1)* %in) #0 {
   %load = load <32 x i16>, <32 x i16> addrspace(1)* %in
   %ext = zext <32 x i16> %load to <32 x i64>

 
 ; FUNC-LABEL: {{^}}global_sextload_v32i16_to_v32i64:
 
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 32, #1
-; EG-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 48, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 0, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 16, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 32, #1
+; EGCM-DAG: VTX_READ_128 T{{[0-9]+}}.XYZW, T{{[0-9]+}}.X, 48, #1
 define void @global_sextload_v32i16_to_v32i64(<32 x i64> addrspace(1)* %out, <32 x i16> addrspace(1)* %in) #0 {
   %load = load <32 x i16>, <32 x i16> addrspace(1)* %in
   %ext = sext <32 x i16> %load to <32 x i64>

test/CodeGen/AMDGPU/store-private.ll

+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=cayman < %s | FileCheck -check-prefix=CM -check-prefix=FUNC %s
+
+; FUNC-LABEL: {{^}}store_i1:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_byte
+define void @store_i1(i1 addrspace(0)* %out) {
+entry:
+  store i1 true, i1 addrspace(0)* %out
+  ret void
+}
+
+; i8 store
+; FUNC-LABEL: {{^}}store_i8:
+; EG: LSHR * [[ADDRESS:T[0-9]\.[XYZW]]], KC0[2].Y, literal.x
+; EG-NEXT: 2
+; EG: MOVA_INT * AR.x (MASKED)
+; EG: MOV [[OLD:T[0-9]\.[XYZW]]], {{.*}}AR.x
+
+; IG 0: Get the byte index and truncate the value
+; EG: AND_INT * T{{[0-9]}}.[[BI_CHAN:[XYZW]]], KC0[2].Y, literal.x
+; EG: LSHL * T{{[0-9]}}.[[SHIFT_CHAN:[XYZW]]], PV.[[BI_CHAN]], literal.x
+; EG-NEXT: 3(4.203895e-45)
+; EG: AND_INT * T{{[0-9]}}.[[TRUNC_CHAN:[XYZW]]], KC0[2].Z, literal.x
+; EG-NEXT: 255(3.573311e-43)
+
+; EG: NOT_INT
+; EG: AND_INT {{[\* ]*}}[[CLR_CHAN:T[0-9]\.[XYZW]]], {{.*}}[[OLD]]
+; EG: OR_INT * [[RES:T[0-9]\.[XYZW]]]
+; TODO: Is the reload necessary?
+; EG: MOVA_INT * AR.x (MASKED), [[ADDRESS]]
+; EG: MOV * T(0 + AR.x).X+, [[RES]]
+
+; SI: buffer_store_byte
+
+define void @store_i8(i8 addrspace(0)* %out, i8 %in) {
+entry:
+  store i8 %in, i8 addrspace(0)* %out
+  ret void
+}
+
+; i16 store
+; FUNC-LABEL: {{^}}store_i16:
+; EG: LSHR * [[ADDRESS:T[0-9]\.[XYZW]]], KC0[2].Y, literal.x
+; EG-NEXT: 2
+; EG: MOVA_INT * AR.x (MASKED)
+; EG: MOV [[OLD:T[0-9]\.[XYZW]]], {{.*}}AR.x
+
+; IG 0: Get the byte index and truncate the value
+; EG: AND_INT * T{{[0-9]}}.[[BI_CHAN:[XYZW]]], KC0[2].Y, literal.x
+; EG: LSHL * T{{[0-9]}}.[[SHIFT_CHAN:[XYZW]]], PV.[[BI_CHAN]], literal.x
+; EG-NEXT: 3(4.203895e-45)
+; EG: AND_INT * T{{[0-9]}}.[[TRUNC_CHAN:[XYZW]]], KC0[2].Z, literal.x
+; EG-NEXT: 65535(9.183409e-41)
+
+; EG: NOT_INT
+; EG: AND_INT {{[\* ]*}}[[CLR_CHAN:T[0-9]\.[XYZW]]], {{.*}}[[OLD]]
+; EG: OR_INT * [[RES:T[0-9]\.[XYZW]]]
+; TODO: Is the reload necessary?
+; EG: MOVA_INT * AR.x (MASKED), [[ADDRESS]]
+; EG: MOV * T(0 + AR.x).X+, [[RES]]
+
+; SI: buffer_store_short
+define void @store_i16(i16 addrspace(0)* %out, i16 %in) {
+entry:
+  store i16 %in, i16 addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_i24:
+; SI: s_lshr_b32 s{{[0-9]+}}, s{{[0-9]+}}, 16
+; SI-DAG: buffer_store_byte
+; SI-DAG: buffer_store_short
+
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store can be eliminated
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store can be eliminated
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+define void @store_i24(i24 addrspace(0)* %out, i24 %in) {
+entry:
+  store i24 %in, i24 addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_i25:
+; SI: s_and_b32 [[AND:s[0-9]+]], s{{[0-9]+}}, 0x1ffffff{{$}}
+; SI: v_mov_b32_e32 [[VAND:v[0-9]+]], [[AND]]
+; SI: buffer_store_dword [[VAND]]
+
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG-NOT: MOVA_INT
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM-NOT: MOVA_INT
+define void @store_i25(i25 addrspace(0)* %out, i25 %in) {
+entry:
+  store i25 %in, i25 addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v2i8:
+; v2i8 is naturally 2B aligned, treat as i16
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG-NOT: MOVA_INT
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM-NOT: MOVA_INT
+
+; SI: buffer_store_short
+define void @store_v2i8(<2 x i8> addrspace(0)* %out, <2 x i32> %in) {
+entry:
+  %0 = trunc <2 x i32> %in to <2 x i8>
+  store <2 x i8> %0, <2 x i8> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v2i8_unaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_byte
+define void @store_v2i8_unaligned(<2 x i8> addrspace(0)* %out, <2 x i32> %in) {
+entry:
+  %0 = trunc <2 x i32> %in to <2 x i8>
+  store <2 x i8> %0, <2 x i8> addrspace(0)* %out, align 1
+  ret void
+}
+
+
+; FUNC-LABEL: {{^}}store_v2i16:
+; v2i8 is naturally 2B aligned, treat as i16
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG-NOT: MOVA_INT
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM-NOT: MOVA_INT
+
+; SI: buffer_store_dword
+define void @store_v2i16(<2 x i16> addrspace(0)* %out, <2 x i32> %in) {
+entry:
+  %0 = trunc <2 x i32> %in to <2 x i16>
+  store <2 x i16> %0, <2 x i16> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v2i16_unaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_short
+; SI: buffer_store_short
+define void @store_v2i16_unaligned(<2 x i16> addrspace(0)* %out, <2 x i32> %in) {
+entry:
+  %0 = trunc <2 x i32> %in to <2 x i16>
+  store <2 x i16> %0, <2 x i16> addrspace(0)* %out, align 2
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i8:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG-NOT: MOVA_INT
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM-NOT: MOVA_INT
+
+; SI: buffer_store_dword
+define void @store_v4i8(<4 x i8> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  %0 = trunc <4 x i32> %in to <4 x i8>
+  store <4 x i8> %0, <4 x i8> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i8_unaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI-NOT: buffer_store_dword
+define void @store_v4i8_unaligned(<4 x i8> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  %0 = trunc <4 x i32> %in to <4 x i8>
+  store <4 x i8> %0, <4 x i8> addrspace(0)* %out, align 1
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v8i8_unaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI: buffer_store_byte
+; SI-NOT: buffer_store_dword
+define void @store_v8i8_unaligned(<8 x i8> addrspace(0)* %out, <8 x i32> %in) {
+entry:
+  %0 = trunc <8 x i32> %in to <8 x i8>
+  store <8 x i8> %0, <8 x i8> addrspace(0)* %out, align 1
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i8_halfaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; TODO: This load and store cannot be eliminated,
+;       they might be different locations
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_short
+; SI: buffer_store_short
+; SI-NOT: buffer_store_dword
+define void @store_v4i8_halfaligned(<4 x i8> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  %0 = trunc <4 x i32> %in to <4 x i8>
+  store <4 x i8> %0, <4 x i8> addrspace(0)* %out, align 2
+  ret void
+}
+
+; floating-point store
+; FUNC-LABEL: {{^}}store_f32:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_dword
+
+define void @store_f32(float addrspace(0)* %out, float %in) {
+  store float %in, float addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i16:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x2?
+; XSI: buffer_store_dwordx2
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @store_v4i16(<4 x i16> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  %0 = trunc <4 x i32> %in to <4 x i16>
+  store <4 x i16> %0, <4 x i16> addrspace(0)* %out
+  ret void
+}
+
+; vec2 floating-point stores
+; FUNC-LABEL: {{^}}store_v2f32:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x2?
+; XSI: buffer_store_dwordx2
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+
+define void @store_v2f32(<2 x float> addrspace(0)* %out, float %a, float %b) {
+entry:
+  %0 = insertelement <2 x float> <float 0.0, float 0.0>, float %a, i32 0
+  %1 = insertelement <2 x float> %0, float %b, i32 1
+  store <2 x float> %1, <2 x float> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v3i32:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x2?
+; XSI-DAG: buffer_store_dwordx2
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+
+define void @store_v3i32(<3 x i32> addrspace(0)* %out, <3 x i32> %a) nounwind {
+  store <3 x i32> %a, <3 x i32> addrspace(0)* %out, align 16
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i32:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x4?
+; XSI: buffer_store_dwordx4
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @store_v4i32(<4 x i32> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  store <4 x i32> %in, <4 x i32> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_v4i32_unaligned:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x4?
+; XSI: buffer_store_dwordx4
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @store_v4i32_unaligned(<4 x i32> addrspace(0)* %out, <4 x i32> %in) {
+entry:
+  store <4 x i32> %in, <4 x i32> addrspace(0)* %out, align 4
+  ret void
+}
+
+; v4f32 store
+; FUNC-LABEL: {{^}}store_v4f32:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x4?
+; XSI: buffer_store_dwordx4
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @store_v4f32(<4 x float> addrspace(0)* %out, <4 x float> addrspace(0)* %in) {
+  %1 = load <4 x float>, <4 x float> addrspace(0) * %in
+  store <4 x float> %1, <4 x float> addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_i64_i8:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_byte
+define void @store_i64_i8(i8 addrspace(0)* %out, i64 %in) {
+entry:
+  %0 = trunc i64 %in to i8
+  store i8 %0, i8 addrspace(0)* %out
+  ret void
+}
+
+; FUNC-LABEL: {{^}}store_i64_i16:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}{{T[0-9]+\.[XYZW]}}, T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; SI: buffer_store_short
+define void @store_i64_i16(i16 addrspace(0)* %out, i64 %in) {
+entry:
+  %0 = trunc i64 %in to i16
+  store i16 %0, i16 addrspace(0)* %out
+  ret void
+}
+
+; The stores in this function are combined by the optimizer to create a
+; 64-bit store with 32-bit alignment.  This is legal and the legalizer
+; should not try to split the 64-bit store back into 2 32-bit stores.
+
+; FUNC-LABEL: {{^}}vecload2:
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x2?
+; XSI: buffer_store_dwordx2
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @vecload2(i32 addrspace(0)* nocapture %out, i32 addrspace(2)* nocapture %mem) #0 {
+entry:
+  %0 = load i32, i32 addrspace(2)* %mem, align 4
+  %arrayidx1.i = getelementptr inbounds i32, i32 addrspace(2)* %mem, i64 1
+  %1 = load i32, i32 addrspace(2)* %arrayidx1.i, align 4
+  store i32 %0, i32 addrspace(0)* %out, align 4
+  %arrayidx1 = getelementptr inbounds i32, i32 addrspace(0)* %out, i64 1
+  store i32 %1, i32 addrspace(0)* %arrayidx1, align 4
+  ret void
+}
+
+; When i128 was a legal type this program generated cannot select errors:
+
+; FUNC-LABEL: {{^}}"i128-const-store":
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; EG: MOVA_INT
+; EG: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+; CM: MOVA_INT
+; CM: MOV {{[\* ]*}}T(0 + AR.x).X+,
+
+;TODO: why not x4?
+; XSI: buffer_store_dwordx4
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+; SI: buffer_store_dword
+define void @i128-const-store(i32 addrspace(0)* %out) {
+entry:
+  store i32 1, i32 addrspace(0)* %out, align 4
+  %arrayidx2 = getelementptr inbounds i32, i32 addrspace(0)* %out, i64 1
+  store i32 1, i32 addrspace(0)* %arrayidx2, align 4
+  %arrayidx4 = getelementptr inbounds i32, i32 addrspace(0)* %out, i64 2
+  store i32 2, i32 addrspace(0)* %arrayidx4, align 4
+  %arrayidx6 = getelementptr inbounds i32, i32 addrspace(0)* %out, i64 3
+  store i32 2, i32 addrspace(0)* %arrayidx6, align 4
+  ret void
+}
+
+
+attributes #0 = { nounwind }