diff --git a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -155,6 +155,7 @@ class AMDGPUOperand : public MCParsedAsmOperand {
     ImmTySdwaSrc1Sel,
     ImmTySdwaDstUnused,
     ImmTyDMask,
+    ImmTyDim,
     ImmTyUNorm,
     ImmTyDA,
     ImmTyR128A16,
@@ -296,6 +297,7 @@ class AMDGPUOperand : public MCParsedAsmOperand {
   bool isClampSI() const { return isImmTy(ImmTyClampSI); }
   bool isOModSI() const { return isImmTy(ImmTyOModSI); }
   bool isDMask() const { return isImmTy(ImmTyDMask); }
+  bool isDim() const { return isImmTy(ImmTyDim); }
   bool isUNorm() const { return isImmTy(ImmTyUNorm); }
   bool isDA() const { return isImmTy(ImmTyDA); }
   bool isR128A16() const { return isImmTy(ImmTyR128A16); }
@@ -695,6 +697,7 @@ class AMDGPUOperand : public MCParsedAsmOperand {
     case ImmTySdwaSrc1Sel: OS << "SdwaSrc1Sel"; break;
     case ImmTySdwaDstUnused: OS << "SdwaDstUnused"; break;
     case ImmTyDMask: OS << "DMask"; break;
+    case ImmTyDim: OS << "Dim"; break;
     case ImmTyUNorm: OS << "UNorm"; break;
     case ImmTyDA: OS << "DA"; break;
     case ImmTyR128A16: OS << "R128A16"; break;
@@ -926,6 +929,10 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
   enum AMDGPUMatchResultTy {
     Match_PreferE32 = FIRST_TARGET_MATCH_RESULT_TY
   };
+  enum OperandMode {
+    OperandMode_Default,
+    OperandMode_NSA,
+  };
 
   using OptionalImmIndexMap = std::map<AMDGPUOperand::ImmTy, unsigned>;
 
@@ -1065,7 +1072,8 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
                                uint64_t &ErrorInfo,
                                bool MatchingInlineAsm) override;
   bool ParseDirective(AsmToken DirectiveID) override;
-  OperandMatchResultTy parseOperand(OperandVector &Operands, StringRef Mnemonic);
+  OperandMatchResultTy parseOperand(OperandVector &Operands, StringRef Mnemonic,
+                                    OperandMode Mode = OperandMode_Default);
   StringRef parseMnemonicSuffix(StringRef Name);
   bool ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
                         SMLoc NameLoc, OperandVector &Operands) override;
@@ -1133,7 +1141,9 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
   bool validateMIMGAtomicDMask(const MCInst &Inst);
   bool validateMIMGGatherDMask(const MCInst &Inst);
   bool validateMIMGDataSize(const MCInst &Inst);
+  bool validateMIMGAddrSize(const MCInst &Inst);
   bool validateMIMGD16(const MCInst &Inst);
+  bool validateMIMGDim(const MCInst &Inst);
   bool validateLdsDirect(const MCInst &Inst);
   bool usesConstantBus(const MCInst &Inst, unsigned OpIdx);
   bool isInlineConstant(const MCInst &Inst, unsigned OpIdx) const;
@@ -1211,6 +1221,7 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
                bool IsAtomic = false);
   void cvtMIMGAtomic(MCInst &Inst, const OperandVector &Operands);
 
+  OperandMatchResultTy parseDim(OperandVector &Operands);
   OperandMatchResultTy parseDPPCtrl(OperandVector &Operands);
   AMDGPUOperand::Ptr defaultRowMask() const;
   AMDGPUOperand::Ptr defaultBankMask() const;
@@ -2565,6 +2576,46 @@ bool AMDGPUAsmParser::validateMIMGDataSize(const MCInst &Inst) {
   return (VDataSize / 4) == DataSize + TFESize;
 }
 
+bool AMDGPUAsmParser::validateMIMGAddrSize(const MCInst &Inst) {
+  const unsigned Opc = Inst.getOpcode();
+  const MCInstrDesc &Desc = MII.get(Opc);
+
+  if ((Desc.TSFlags & SIInstrFlags::MIMG) == 0 || !isGFX10())
+    return true;
+
+  const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(Opc);
+  const AMDGPU::MIMGBaseOpcodeInfo *BaseOpcode =
+      AMDGPU::getMIMGBaseOpcodeInfo(Info->BaseOpcode);
+  int VAddr0Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::vaddr0);
+  int SrsrcIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::srsrc);
+  int DimIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::dim);
+
+  assert(VAddr0Idx != -1);
+  assert(SrsrcIdx != -1);
+  assert(DimIdx != -1);
+  assert(SrsrcIdx > VAddr0Idx);
+
+  unsigned Dim = Inst.getOperand(DimIdx).getImm();
+  const AMDGPU::MIMGDimInfo *DimInfo = AMDGPU::getMIMGDimInfoByEncoding(Dim);
+  bool IsNSA = SrsrcIdx - VAddr0Idx > 1;
+  unsigned VAddrSize =
+      IsNSA ? SrsrcIdx - VAddr0Idx
+            : AMDGPU::getRegOperandSize(getMRI(), Desc, VAddr0Idx) / 4;
+
+  unsigned AddrSize = BaseOpcode->NumExtraArgs +
+                      (BaseOpcode->Gradients ? DimInfo->NumGradients : 0) +
+                      (BaseOpcode->Coordinates ? DimInfo->NumCoords : 0) +
+                      (BaseOpcode->LodOrClampOrMip ? 1 : 0);
+  if (!IsNSA) {
+    if (AddrSize > 8)
+      AddrSize = 16;
+    else if (AddrSize > 4)
+      AddrSize = 8;
+  }
+
+  return VAddrSize == AddrSize;
+}
+
 bool AMDGPUAsmParser::validateMIMGAtomicDMask(const MCInst &Inst) {
 
   const unsigned Opc = Inst.getOpcode();
@@ -2621,6 +2672,24 @@ bool AMDGPUAsmParser::validateMIMGD16(const MCInst &Inst) {
   return true;
 }
 
+bool AMDGPUAsmParser::validateMIMGDim(const MCInst &Inst) {
+  const unsigned Opc = Inst.getOpcode();
+  const MCInstrDesc &Desc = MII.get(Opc);
+
+  if ((Desc.TSFlags & SIInstrFlags::MIMG) == 0)
+    return true;
+
+  int DimIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::dim);
+  if (DimIdx < 0)
+    return true;
+
+  long Imm = Inst.getOperand(DimIdx).getImm();
+  if (Imm < 0 || Imm >= 8)
+    return false;
+
+  return true;
+}
+
 static bool IsRevOpcode(const unsigned Opcode)
 {
   switch (Opcode) {
@@ -2853,11 +2922,20 @@ bool AMDGPUAsmParser::validateInstruction(const MCInst &Inst,
       "d16 modifier is not supported on this GPU");
     return false;
   }
+  if (!validateMIMGDim(Inst)) {
+    Error(IDLoc, "dim modifier is required on this GPU");
+    return false;
+  }
   if (!validateMIMGDataSize(Inst)) {
     Error(IDLoc,
       "image data size does not match dmask and tfe");
     return false;
   }
+  if (!validateMIMGAddrSize(Inst)) {
+    Error(IDLoc,
+      "image address size does not match dim and a16");
+    return false;
+  }
   if (!validateMIMGAtomicDMask(Inst)) {
     Error(IDLoc,
       "invalid atomic image dmask");
@@ -3217,6 +3295,24 @@ bool AMDGPUAsmParser::ParseDirectiveAMDHSAKernel() {
                                  IDRange);
       PARSE_BITS_ENTRY(KD.compute_pgm_rsrc1, COMPUTE_PGM_RSRC1_FP16_OVFL, Val,
                        ValRange);
+    } else if (ID == ".amdhsa_workgroup_processor_mode") {
+      if (IVersion.Major < 10)
+        return getParser().Error(IDRange.Start, "directive requires gfx10+",
+                                 IDRange);
+      PARSE_BITS_ENTRY(KD.compute_pgm_rsrc1, COMPUTE_PGM_RSRC1_WGP_MODE, Val,
+                       ValRange);
+    } else if (ID == ".amdhsa_memory_ordered") {
+      if (IVersion.Major < 10)
+        return getParser().Error(IDRange.Start, "directive requires gfx10+",
+                                 IDRange);
+      PARSE_BITS_ENTRY(KD.compute_pgm_rsrc1, COMPUTE_PGM_RSRC1_MEM_ORDERED, Val,
+                       ValRange);
+    } else if (ID == ".amdhsa_forward_progress") {
+      if (IVersion.Major < 10)
+        return getParser().Error(IDRange.Start, "directive requires gfx10+",
+                                 IDRange);
+      PARSE_BITS_ENTRY(KD.compute_pgm_rsrc1, COMPUTE_PGM_RSRC1_FWD_PROGRESS, Val,
+                       ValRange);
     } else if (ID == ".amdhsa_exception_fp_ieee_invalid_op") {
       PARSE_BITS_ENTRY(
           KD.compute_pgm_rsrc2,
@@ -3370,6 +3466,22 @@ bool AMDGPUAsmParser::ParseAMDKernelCodeTValue(StringRef ID,
     return TokError(Err.str());
   }
   Lex();
+
+  if (ID == "enable_wgp_mode") {
+    if (G_00B848_WGP_MODE(Header.compute_pgm_resource_registers) && !isGFX10())
+      return TokError("enable_wgp_mode=1 is only allowed on GFX10+");
+  }
+
+  if (ID == "enable_mem_ordered") {
+    if (G_00B848_MEM_ORDERED(Header.compute_pgm_resource_registers) && !isGFX10())
+      return TokError("enable_mem_ordered=1 is only allowed on GFX10+");
+  }
+
+  if (ID == "enable_fwd_progress") {
+    if (G_00B848_FWD_PROGRESS(Header.compute_pgm_resource_registers) && !isGFX10())
+      return TokError("enable_fwd_progress=1 is only allowed on GFX10+");
+  }
+
   return false;
 }
 
@@ -3669,7 +3781,8 @@ bool AMDGPUAsmParser::subtargetHasRegister(const MCRegisterInfo &MRI,
 }
 
 OperandMatchResultTy
-AMDGPUAsmParser::parseOperand(OperandVector &Operands, StringRef Mnemonic) {
+AMDGPUAsmParser::parseOperand(OperandVector &Operands, StringRef Mnemonic,
+                              OperandMode Mode) {
   // Try to parse with a custom parser
   OperandMatchResultTy ResTy = MatchOperandParserImpl(Operands, Mnemonic);
 
@@ -3683,6 +3796,35 @@ AMDGPUAsmParser::parseOperand(OperandVector &Operands, StringRef Mnemonic) {
       getLexer().is(AsmToken::EndOfStatement))
     return ResTy;
 
+  if (Mode == OperandMode_NSA && getLexer().is(AsmToken::LBrac)) {
+    unsigned Prefix = Operands.size();
+    SMLoc LBraceLoc = getTok().getLoc();
+    Parser.Lex(); // eat the '['
+
+    for (;;) {
+      ResTy = parseReg(Operands);
+      if (ResTy != MatchOperand_Success)
+        return ResTy;
+
+      if (getLexer().is(AsmToken::RBrac))
+        break;
+
+      if (getLexer().isNot(AsmToken::Comma))
+        return MatchOperand_ParseFail;
+      Parser.Lex();
+    }
+
+    if (Operands.size() - Prefix > 1) {
+      Operands.insert(Operands.begin() + Prefix,
+                      AMDGPUOperand::CreateToken(this, "[", LBraceLoc));
+      Operands.push_back(AMDGPUOperand::CreateToken(this, "]",
+                                                    getTok().getLoc()));
+    }
+
+    Parser.Lex(); // eat the ']'
+    return MatchOperand_Success;
+  }
+
   ResTy = parseRegOrImm(Operands);
 
   if (ResTy == MatchOperand_Success || ResTy == MatchOperand_ParseFail)
@@ -3736,8 +3878,13 @@ bool AMDGPUAsmParser::ParseInstruction(ParseInstructionInfo &Info,
   Name = parseMnemonicSuffix(Name);
   Operands.push_back(AMDGPUOperand::CreateToken(this, Name, NameLoc));
 
+  bool IsMIMG = Name.startswith("image_");
+
   while (!getLexer().is(AsmToken::EndOfStatement)) {
-    OperandMatchResultTy Res = parseOperand(Operands, Name);
+    OperandMode Mode = OperandMode_Default;
+    if (IsMIMG && isGFX10() && Operands.size() == 2)
+      Mode = OperandMode_NSA;
+    OperandMatchResultTy Res = parseOperand(Operands, Name, Mode);
 
     // Eat the comma or space if there is one.
     if (getLexer().is(AsmToken::Comma))
@@ -5275,14 +5422,16 @@ void AMDGPUAsmParser::cvtMIMG(MCInst &Inst, const OperandVector &Operands,
       Op.addRegOperands(Inst, 1);
     } else if (Op.isImmModifier()) {
       OptionalIdx[Op.getImmTy()] = I;
-    } else {
+    } else if (!Op.isToken()) {
       llvm_unreachable("unexpected operand type");
     }
   }
 
   bool IsGFX10 = isGFX10();
 
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDMask);
+  if (IsGFX10)
+    addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDim, -1);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyUNorm);
   if (IsGFX10)
     addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDLC);
@@ -5291,7 +5440,8 @@ void AMDGPUAsmParser::cvtMIMG(MCInst &Inst, const OperandVector &Operands,
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyR128A16);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyTFE);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyLWE);
-  addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDA);
+  if (!IsGFX10)
+    addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyDA);
   addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyD16);
 }
 
@@ -5389,7 +5539,7 @@ static const OptionalOperand AMDGPUOptionalOperandTable[] = {
   {"offset",  AMDGPUOperand::ImmTyOffset, false, nullptr},
   {"inst_offset", AMDGPUOperand::ImmTyInstOffset, false, nullptr},
   {"dlc",     AMDGPUOperand::ImmTyDLC, true, nullptr},
-  {"dfmt",    AMDGPUOperand::ImmTyFORMAT, false, nullptr},
+  {"format",  AMDGPUOperand::ImmTyFORMAT, false, nullptr},
   {"glc",     AMDGPUOperand::ImmTyGLC, true, nullptr},
   {"slc",     AMDGPUOperand::ImmTySLC, true, nullptr},
   {"tfe",     AMDGPUOperand::ImmTyTFE, true, nullptr},
@@ -5404,6 +5554,7 @@ static const OptionalOperand AMDGPUOptionalOperandTable[] = {
   {"lwe",     AMDGPUOperand::ImmTyLWE,   true, nullptr},
   {"d16",     AMDGPUOperand::ImmTyD16,   true, nullptr},
   {"dmask",   AMDGPUOperand::ImmTyDMask, false, nullptr},
+  {"dim",     AMDGPUOperand::ImmTyDim,   false, nullptr},
   {"row_mask",   AMDGPUOperand::ImmTyDppRowMask, false, nullptr},
   {"bank_mask",  AMDGPUOperand::ImmTyDppBankMask, false, nullptr},
   {"bound_ctrl", AMDGPUOperand::ImmTyDppBoundCtrl, false, ConvertBoundCtrl},
@@ -5472,7 +5623,9 @@ OperandMatchResultTy AMDGPUAsmParser::parseOptionalOpr(OperandVector &Operands)
                Op.Type == AMDGPUOperand::ImmTyNegHi) {
       res = parseOperandArrayWithPrefix(Op.Name, Operands, Op.Type,
                                         Op.ConvertResult);
-    } else if (Op.Type == AMDGPUOperand::ImmTyFORMAT) {
+    } else if (Op.Type == AMDGPUOperand::ImmTyDim) {
+      res = parseDim(Operands);
+    } else if (Op.Type == AMDGPUOperand::ImmTyFORMAT && !isGFX10()) {
       res = parseDfmtNfmt(Operands);
     } else {
       res = parseIntWithPrefix(Op.Name, Operands, Op.Type, Op.ConvertResult);
@@ -5758,6 +5911,52 @@ bool AMDGPUOperand::isU16Imm() const {
   return isImm() && isUInt<16>(getImm());
 }
 
+OperandMatchResultTy AMDGPUAsmParser::parseDim(OperandVector &Operands) {
+  if (!isGFX10())
+    return MatchOperand_NoMatch;
+
+  SMLoc S = Parser.getTok().getLoc();
+
+  if (getLexer().isNot(AsmToken::Identifier))
+    return MatchOperand_NoMatch;
+  if (getLexer().getTok().getString() != "dim")
+    return MatchOperand_NoMatch;
+
+  Parser.Lex();
+  if (getLexer().isNot(AsmToken::Colon))
+    return MatchOperand_ParseFail;
+
+  Parser.Lex();
+
+  // We want to allow "dim:1D" etc., but the initial 1 is tokenized as an
+  // integer.
+  std::string Token;
+  if (getLexer().is(AsmToken::Integer)) {
+    SMLoc Loc = getLexer().getTok().getEndLoc();
+    Token = getLexer().getTok().getString();
+    Parser.Lex();
+    if (getLexer().getTok().getLoc() != Loc)
+      return MatchOperand_ParseFail;
+  }
+  if (getLexer().isNot(AsmToken::Identifier))
+    return MatchOperand_ParseFail;
+  Token += getLexer().getTok().getString();
+
+  StringRef DimId = Token;
+  if (DimId.startswith("SQ_RSRC_IMG_"))
+    DimId = DimId.substr(12);
+
+  const AMDGPU::MIMGDimInfo *DimInfo = AMDGPU::getMIMGDimInfoByAsmSuffix(DimId);
+  if (!DimInfo)
+    return MatchOperand_ParseFail;
+
+  Parser.Lex();
+
+  Operands.push_back(AMDGPUOperand::CreateImm(this, DimInfo->Encoding, S,
+                                              AMDGPUOperand::ImmTyDim));
+  return MatchOperand_Success;
+}
+
 OperandMatchResultTy
 AMDGPUAsmParser::parseDPPCtrl(OperandVector &Operands) {
   using namespace AMDGPU::DPP;

diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
@@ -290,7 +290,26 @@ DecodeStatus AMDGPUDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
   }
 
   if (Res && (MCII->get(MI.getOpcode()).TSFlags & SIInstrFlags::MIMG)) {
-    Res = convertMIMGInst(MI);
+    int VAddr0Idx =
+        AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::vaddr0);
+    int RsrcIdx =
+        AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::srsrc);
+    unsigned NSAArgs = RsrcIdx - VAddr0Idx - 1;
+    if (VAddr0Idx >= 0 && NSAArgs > 0) {
+      unsigned NSAWords = (NSAArgs + 3) / 4;
+      if (Bytes.size() < 4 * NSAWords) {
+        Res = MCDisassembler::Fail;
+      } else {
+        for (unsigned i = 0; i < NSAArgs; ++i) {
+          MI.insert(MI.begin() + VAddr0Idx + 1 + i,
+                    decodeOperand_VGPR_32(Bytes[i]));
+        }
+        Bytes = Bytes.slice(4 * NSAWords);
+      }
+    }
+
+    if (Res)
+      Res = convertMIMGInst(MI);
   }
 
   if (Res && IsSDWA)
@@ -339,17 +358,18 @@ DecodeStatus AMDGPUDisassembler::convertSDWAInst(MCInst &MI) const {
   return MCDisassembler::Success;
 }
 
-// Note that MIMG format provides no information about VADDR size.
-// Consequently, decoded instructions always show address
-// as if it has 1 dword, which could be not really so.
+// Note that before gfx10, the MIMG encoding provided no information about
+// VADDR size. Consequently, decoded instructions always show address as if it
+// has 1 dword, which could be not really so.
 DecodeStatus AMDGPUDisassembler::convertMIMGInst(MCInst &MI) const {
 
   int VDstIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                            AMDGPU::OpName::vdst);
 
   int VDataIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                             AMDGPU::OpName::vdata);
-
+  int VAddr0Idx =
+      AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::vaddr0);
   int DMaskIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
                                             AMDGPU::OpName::dmask);
 
@@ -362,16 +382,42 @@ DecodeStatus AMDGPUDisassembler::convertMIMGInst(MCInst &MI) const {
   assert(DMaskIdx != -1);
   assert(TFEIdx != -1);
 
+  const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(MI.getOpcode());
   bool IsAtomic = (VDstIdx != -1);
   bool IsGather4 = MCII->get(MI.getOpcode()).TSFlags & SIInstrFlags::Gather4;
 
-  unsigned DMask = MI.getOperand(DMaskIdx).getImm() & 0xf;
-  if (DMask == 0)
-    return MCDisassembler::Success;
+  bool IsNSA = false;
+  unsigned AddrSize = Info->VAddrDwords;
+
+  if (STI.getFeatureBits()[AMDGPU::FeatureGFX10]) {
+    unsigned DimIdx =
+        AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::dim);
+    const AMDGPU::MIMGBaseOpcodeInfo *BaseOpcode =
+        AMDGPU::getMIMGBaseOpcodeInfo(Info->BaseOpcode);
+    const AMDGPU::MIMGDimInfo *Dim =
+        AMDGPU::getMIMGDimInfoByEncoding(MI.getOperand(DimIdx).getImm());
+
+    AddrSize = BaseOpcode->NumExtraArgs +
+               (BaseOpcode->Gradients ? Dim->NumGradients : 0) +
+               (BaseOpcode->Coordinates ? Dim->NumCoords : 0) +
+               (BaseOpcode->LodOrClampOrMip ? 1 : 0);
+    IsNSA = Info->MIMGEncoding == AMDGPU::MIMGEncGfx10NSA;
+    if (!IsNSA) {
+      if (AddrSize > 8)
+        AddrSize = 16;
+      else if (AddrSize > 4)
+        AddrSize = 8;
+    } else {
+      if (AddrSize > Info->VAddrDwords) {
+        // The NSA encoding does not contain enough operands for the combination
+        // of base opcode / dimension. Should this be an error?
+        return MCDisassembler::Success;
+      }
+    }
+  }
 
-  unsigned DstSize = IsGather4 ? 4 : countPopulation(DMask);
-  if (DstSize == 1)
-    return MCDisassembler::Success;
+  unsigned DMask = MI.getOperand(DMaskIdx).getImm() & 0xf;
+  unsigned DstSize = IsGather4 ? 4 : std::max(countPopulation(DMask), 1u);
 
   bool D16 = D16Idx >= 0 && MI.getOperand(D16Idx).getImm();
   if (D16 && AMDGPU::hasPackedD16(STI)) {
@@ -382,44 +428,64 @@ DecodeStatus AMDGPUDisassembler::convertMIMGInst(MCInst &MI) const {
   if (MI.getOperand(TFEIdx).getImm())
     return MCDisassembler::Success;
 
-  int NewOpcode = -1;
+  if (DstSize == Info->VDataDwords && AddrSize == Info->VAddrDwords)
+    return MCDisassembler::Success;
 
-  if (IsGather4) {
-    if (D16 && AMDGPU::hasPackedD16(STI))
-      NewOpcode = AMDGPU::getMaskedMIMGOp(MI.getOpcode(), 2);
-    else
+  int NewOpcode =
+      AMDGPU::getMIMGOpcode(Info->BaseOpcode, Info->MIMGEncoding, DstSize, AddrSize);
+  if (NewOpcode == -1)
+    return MCDisassembler::Success;
+
+  // Widen the register to the correct number of enabled channels.
+  unsigned NewVdata = AMDGPU::NoRegister;
+  if (DstSize != Info->VDataDwords) {
+    auto DataRCID = MCII->get(NewOpcode).OpInfo[VDataIdx].RegClass;
+
+    // Get first subregister of VData
+    unsigned Vdata0 = MI.getOperand(VDataIdx).getReg();
+    unsigned VdataSub0 = MRI.getSubReg(Vdata0, AMDGPU::sub0);
+    Vdata0 = (VdataSub0 != 0)? VdataSub0 : Vdata0;
+
+    NewVdata = MRI.getMatchingSuperReg(Vdata0, AMDGPU::sub0,
+                                       &MRI.getRegClass(DataRCID));
+    if (NewVdata == AMDGPU::NoRegister) {
+      // It's possible to encode this such that the low register + enabled
+      // components exceeds the register count.
       return MCDisassembler::Success;
-  } else {
-    NewOpcode = AMDGPU::getMaskedMIMGOp(MI.getOpcode(), DstSize);
-    if (NewOpcode == -1)
+    }
+  }
+
+  unsigned NewVAddr0 = AMDGPU::NoRegister;
+  if (STI.getFeatureBits()[AMDGPU::FeatureGFX10] && !IsNSA &&
+      AddrSize != Info->VAddrDwords) {
+    unsigned VAddr0 = MI.getOperand(VAddr0Idx).getReg();
+    unsigned VAddrSub0 = MRI.getSubReg(VAddr0, AMDGPU::sub0);
+    VAddr0 = (VAddrSub0 != 0) ? VAddrSub0 : VAddr0;
+
+    auto AddrRCID = MCII->get(NewOpcode).OpInfo[VAddr0Idx].RegClass;
+    NewVAddr0 = MRI.getMatchingSuperReg(VAddr0, AMDGPU::sub0,
+                                        &MRI.getRegClass(AddrRCID));
+    if (NewVAddr0 == AMDGPU::NoRegister)
       return MCDisassembler::Success;
   }
 
-  auto RCID = MCII->get(NewOpcode).OpInfo[VDataIdx].RegClass;
+  MI.setOpcode(NewOpcode);
 
-  // Get first subregister of VData
-  unsigned Vdata0 = MI.getOperand(VDataIdx).getReg();
-  unsigned VdataSub0 = MRI.getSubReg(Vdata0, AMDGPU::sub0);
-  Vdata0 = (VdataSub0 != 0)? VdataSub0 : Vdata0;
+  if (NewVdata != AMDGPU::NoRegister) {
+    MI.getOperand(VDataIdx) = MCOperand::createReg(NewVdata);
 
-  // Widen the register to the correct number of enabled channels.
-  auto NewVdata = MRI.getMatchingSuperReg(Vdata0, AMDGPU::sub0,
-                                          &MRI.getRegClass(RCID));
-  if (NewVdata == AMDGPU::NoRegister) {
-    // It's possible to encode this such that the low register + enabled
-    // components exceeds the register count.
-    return MCDisassembler::Success;
+    if (IsAtomic) {
+      // Atomic operations have an additional operand (a copy of data)
+      MI.getOperand(VDstIdx) = MCOperand::createReg(NewVdata);
+    }
   }
 
-  MI.setOpcode(NewOpcode);
-  // vaddr will be always appear as a single VGPR. This will look different than
-  // how it is usually emitted because the number of register components is not
-  // in the instruction encoding.
-  MI.getOperand(VDataIdx) = MCOperand::createReg(NewVdata);
-
-  if (IsAtomic) {
-    // Atomic operations have an additional operand (a copy of data)
-    MI.getOperand(VDstIdx) = MCOperand::createReg(NewVdata);
+  if (NewVAddr0 != AMDGPU::NoRegister) {
+    MI.getOperand(VAddr0Idx) = MCOperand::createReg(NewVAddr0);
+  } else if (IsNSA) {
+    assert(AddrSize <= Info->VAddrDwords);
+    MI.erase(MI.begin() + VAddr0Idx + AddrSize,
+             MI.begin() + VAddr0Idx + Info->VAddrDwords);
   }
 
   return MCDisassembler::Success;

diff --git a/llvm/lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.cpp b/llvm/lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.cpp
@@ -208,6 +208,18 @@ void AMDGPUInstPrinter::printDMask(const MCInst *MI, unsigned OpNo,
   }
 }
 
+void AMDGPUInstPrinter::printDim(const MCInst *MI, unsigned OpNo,
+                                 const MCSubtargetInfo &STI, raw_ostream &O) {
+  unsigned Dim = MI->getOperand(OpNo).getImm();
+  O << " dim:SQ_RSRC_IMG_";
+
+  const AMDGPU::MIMGDimInfo *DimInfo = AMDGPU::getMIMGDimInfoByEncoding(Dim);
+  if (DimInfo)
+    O << DimInfo->AsmSuffix;
+  else
+    O << Dim;
+}
+
 void AMDGPUInstPrinter::printUNorm(const MCInst *MI, unsigned OpNo,
                                    const MCSubtargetInfo &STI, raw_ostream &O) {
   printNamedBit(MI, OpNo, O, "unorm");
@@ -254,8 +266,12 @@ void AMDGPUInstPrinter::printFORMAT(const MCInst *MI, unsigned OpNo,
                                     const MCSubtargetInfo &STI,
                                     raw_ostream &O) {
   if (unsigned Val = MI->getOperand(OpNo).getImm()) {
-    O << " dfmt:" << (Val & 15);
-    O << ", nfmt:" << (Val >> 4);
+    if (AMDGPU::isGFX10(STI))
+      O << " format:" << Val;
+    else {
+      O << " dfmt:" << (Val & 15);
+      O << ", nfmt:" << (Val >> 4);
+    }
   }
 }
 

diff --git a/llvm/lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.h b/llvm/lib/Target/AMDGPU/InstPrinter/AMDGPUInstPrinter.h
@@ -78,6 +78,8 @@ class AMDGPUInstPrinter : public MCInstPrinter {
                 raw_ostream &O);
   void printDMask(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,
                   raw_ostream &O);
+  void printDim(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,
+                raw_ostream &O);
   void printUNorm(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,
                   raw_ostream &O);
   void printDA(const MCInst *MI, unsigned OpNo, const MCSubtargetInfo &STI,

diff --git a/llvm/lib/Target/AMDGPU/MCTargetDesc/SIMCCodeEmitter.cpp b/llvm/lib/Target/AMDGPU/MCTargetDesc/SIMCCodeEmitter.cpp
@@ -16,6 +16,7 @@
 #include "MCTargetDesc/AMDGPUFixupKinds.h"
 #include "MCTargetDesc/AMDGPUMCCodeEmitter.h"
 #include "MCTargetDesc/AMDGPUMCTargetDesc.h"
+#include "SIDefines.h"
 #include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/MC/MCCodeEmitter.h"
 #include "llvm/MC/MCContext.h"
@@ -273,7 +274,25 @@ void SIMCCodeEmitter::encodeInstruction(const MCInst &MI, raw_ostream &OS,
     OS.write((uint8_t) ((Encoding >> (8 * i)) & 0xff));
   }
 
-  if (bytes > 4)
+  // NSA encoding.
+  if (AMDGPU::isGFX10(STI) && Desc.TSFlags & SIInstrFlags::MIMG) {
+    int vaddr0 = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
+                                            AMDGPU::OpName::vaddr0);
+    int srsrc = AMDGPU::getNamedOperandIdx(MI.getOpcode(),
+                                           AMDGPU::OpName::srsrc);
+    assert(vaddr0 >= 0 && srsrc > vaddr0);
+    unsigned NumExtraAddrs = srsrc - vaddr0 - 1;
+    unsigned NumPadding = (-NumExtraAddrs) & 3;
+
+    for (unsigned i = 0; i < NumExtraAddrs; ++i)
+      OS.write((uint8_t)getMachineOpValue(MI, MI.getOperand(vaddr0 + 1 + i),
+                                          Fixups, STI));
+    for (unsigned i = 0; i < NumPadding; ++i)
+      OS.write(0);
+  }
+
+  if ((bytes > 8 && STI.getFeatureBits()[AMDGPU::FeatureVOP3Literal]) ||
+      (bytes > 4 && !STI.getFeatureBits()[AMDGPU::FeatureVOP3Literal]))
     return;
 
   // Check for additional literals in SRC0/1/2 (Op 1/2/3)
@@ -428,7 +447,8 @@ uint64_t SIMCCodeEmitter::getMachineOpValue(const MCInst &MI,
   const MCInstrDesc &Desc = MCII.get(MI.getOpcode());
   if (AMDGPU::isSISrcOperand(Desc, OpNo)) {
     uint32_t Enc = getLitEncoding(MO, Desc.OpInfo[OpNo], STI);
-    if (Enc != ~0U && (Enc != 255 || Desc.getSize() == 4))
+    if (Enc != ~0U &&
+        (Enc != 255 || Desc.getSize() == 4 || Desc.getSize() == 8))
       return Enc;
 
   } else if (MO.isImm())

diff --git a/llvm/lib/Target/AMDGPU/MIMGInstructions.td b/llvm/lib/Target/AMDGPU/MIMGInstructions.td
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -1003,6 +1003,13 @@ bool SITargetLowering::isLegalFlatAddressingMode(const AddrMode &AM) const {
   // GFX9 added a 13-bit signed offset. When using regular flat instructions,
   // the sign bit is ignored and is treated as a 12-bit unsigned offset.
 
+  // GFX10 shrinked signed offset to 12 bits. When using regular flat
+  // instructions, the sign bit is also ignored and is treated as 11-bit
+  // unsigned offset.
+
+  if (Subtarget->getGeneration() >= AMDGPUSubtarget::GFX10)
+    return isUInt<11>(AM.BaseOffs) && AM.Scale == 0;
+
   // Just r + i
   return isUInt<12>(AM.BaseOffs) && AM.Scale == 0;
 }
@@ -2828,8 +2835,9 @@ unsigned SITargetLowering::getRegisterByName(const char* RegName, EVT VT,
 
   }
 
-  if (Subtarget->getGeneration() == AMDGPUSubtarget::SOUTHERN_ISLANDS &&
-      Subtarget->getRegisterInfo()->regsOverlap(Reg, AMDGPU::FLAT_SCR)) {
+  if ((Subtarget->getGeneration() == AMDGPUSubtarget::SOUTHERN_ISLANDS ||
+       Subtarget->getGeneration() >= AMDGPUSubtarget::GFX10) &&
+       Subtarget->getRegisterInfo()->regsOverlap(Reg, AMDGPU::FLAT_SCR)) {
     report_fatal_error(Twine("invalid register \""
                              + StringRef(RegName)  + "\" for subtarget."));
   }
@@ -4656,7 +4664,7 @@ static SDValue getBuildDwordsVector(SelectionDAG &DAG, SDLoc DL,
 }
 
 static bool parseCachePolicy(SDValue CachePolicy, SelectionDAG &DAG,
-                             SDValue *GLC, SDValue *SLC) {
+                             SDValue *GLC, SDValue *SLC, SDValue *DLC) {
   auto CachePolicyConst = cast<ConstantSDNode>(CachePolicy.getNode());
 
   uint64_t Value = CachePolicyConst->getZExtValue();
@@ -4669,6 +4677,10 @@ static bool parseCachePolicy(SDValue CachePolicy, SelectionDAG &DAG,
     *SLC = DAG.getTargetConstant((Value & 0x2) ? 1 : 0, DL, MVT::i32);
     Value &= ~(uint64_t)0x2;
   }
+  if (DLC) {
+    *DLC = DAG.getTargetConstant((Value & 0x4) ? 1 : 0, DL, MVT::i32);
+    Value &= ~(uint64_t)0x4;
+  }
 
   return Value == 0;
 }
@@ -4786,6 +4798,7 @@ SDValue SITargetLowering::lowerImage(SDValue Op,
   const AMDGPU::MIMGLZMappingInfo *LZMappingInfo =
       AMDGPU::getMIMGLZMappingInfo(Intr->BaseOpcode);
   unsigned IntrOpcode = Intr->BaseOpcode;
+  bool IsGFX10 = Subtarget->getGeneration() >= AMDGPUSubtarget::GFX10;
 
   SmallVector<EVT, 3> ResultTypes(Op->value_begin(), Op->value_end());
   SmallVector<EVT, 3> OrigResultTypes(Op->value_begin(), Op->value_end());
@@ -4924,7 +4937,22 @@ SDValue SITargetLowering::lowerImage(SDValue Op,
       VAddrs.push_back(Op.getOperand(AddrIdx + i));
   }
 
-  SDValue VAddr = getBuildDwordsVector(DAG, DL, VAddrs);
+  // If the register allocator cannot place the address registers contiguously
+  // without introducing moves, then using the non-sequential address encoding
+  // is always preferable, since it saves VALU instructions and is usually a
+  // wash in terms of code size or even better.
+  //
+  // However, we currently have no way of hinting to the register allocator that
+  // MIMG addresses should be placed contiguously when it is possible to do so,
+  // so force non-NSA for the common 2-address case as a heuristic.
+  //
+  // SIShrinkInstructions will convert NSA encodings to non-NSA after register
+  // allocation when possible.
+  bool UseNSA =
+      ST->hasFeature(AMDGPU::FeatureNSAEncoding) && VAddrs.size() >= 3;
+  SDValue VAddr;
+  if (!UseNSA)
+    VAddr = getBuildDwordsVector(DAG, DL, VAddrs);
 
   SDValue True = DAG.getTargetConstant(1, DL, MVT::i1);
   SDValue False = DAG.getTargetConstant(0, DL, MVT::i1);
@@ -4987,45 +5015,66 @@ SDValue SITargetLowering::lowerImage(SDValue Op,
 
   SDValue GLC;
   SDValue SLC;
+  SDValue DLC;
   if (BaseOpcode->Atomic) {
     GLC = True; // TODO no-return optimization
-    if (!parseCachePolicy(Op.getOperand(CtrlIdx + 1), DAG, nullptr, &SLC))
+    if (!parseCachePolicy(Op.getOperand(CtrlIdx + 1), DAG, nullptr, &SLC,
+                          IsGFX10 ? &DLC : nullptr))
       return Op;
   } else {
-    if (!parseCachePolicy(Op.getOperand(CtrlIdx + 1), DAG, &GLC, &SLC))
+    if (!parseCachePolicy(Op.getOperand(CtrlIdx + 1), DAG, &GLC, &SLC,
+                          IsGFX10 ? &DLC : nullptr))
       return Op;
   }
 
-  SmallVector<SDValue, 14> Ops;
+  SmallVector<SDValue, 26> Ops;
   if (BaseOpcode->Store || BaseOpcode->Atomic)
     Ops.push_back(VData); // vdata
-  Ops.push_back(VAddr);
+  if (UseNSA) {
+    for (const SDValue &Addr : VAddrs)
+      Ops.push_back(Addr);
+  } else {
+    Ops.push_back(VAddr);
+  }
   Ops.push_back(Op.getOperand(AddrIdx + NumVAddrs)); // rsrc
   if (BaseOpcode->Sampler)
     Ops.push_back(Op.getOperand(AddrIdx + NumVAddrs + 1)); // sampler
   Ops.push_back(DAG.getTargetConstant(DMask, DL, MVT::i32));
+  if (IsGFX10)
+    Ops.push_back(DAG.getTargetConstant(DimInfo->Encoding, DL, MVT::i32));
   Ops.push_back(Unorm);
+  if (IsGFX10)
+    Ops.push_back(DLC);
   Ops.push_back(GLC);
   Ops.push_back(SLC);
   Ops.push_back(IsA16 &&  // a16 or r128
                 ST->hasFeature(AMDGPU::FeatureR128A16) ? True : False);
   Ops.push_back(TFE); // tfe
   Ops.push_back(LWE); // lwe
-  Ops.push_back(DimInfo->DA ? True : False);
+  if (!IsGFX10)
+    Ops.push_back(DimInfo->DA ? True : False);
   if (BaseOpcode->HasD16)
     Ops.push_back(IsD16 ? True : False);
   if (isa<MemSDNode>(Op))
     Ops.push_back(Op.getOperand(0)); // chain
 
-  int NumVAddrDwords = VAddr.getValueType().getSizeInBits() / 32;
+  int NumVAddrDwords =
+      UseNSA ? VAddrs.size() : VAddr.getValueType().getSizeInBits() / 32;
   int Opcode = -1;
 
-  if (Subtarget->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS)
-    Opcode = AMDGPU::getMIMGOpcode(IntrOpcode, AMDGPU::MIMGEncGfx8,
-                                   NumVDataDwords, NumVAddrDwords);
-  if (Opcode == -1)
-    Opcode = AMDGPU::getMIMGOpcode(IntrOpcode, AMDGPU::MIMGEncGfx6,
+  if (IsGFX10) {
+    Opcode = AMDGPU::getMIMGOpcode(IntrOpcode,
+                                   UseNSA ? AMDGPU::MIMGEncGfx10NSA
+                                          : AMDGPU::MIMGEncGfx10Default,
                                    NumVDataDwords, NumVAddrDwords);
+  } else {
+    if (Subtarget->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS)
+      Opcode = AMDGPU::getMIMGOpcode(IntrOpcode, AMDGPU::MIMGEncGfx8,
+                                     NumVDataDwords, NumVAddrDwords);
+    if (Opcode == -1)
+      Opcode = AMDGPU::getMIMGOpcode(IntrOpcode, AMDGPU::MIMGEncGfx6,
+                                     NumVDataDwords, NumVAddrDwords);
+  }
   assert(Opcode != -1);
 
   MachineSDNode *NewNode = DAG.getMachineNode(Opcode, DL, ResultTypes, Ops);

diff --git a/llvm/lib/Target/AMDGPU/SIInstrFormats.td b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
@@ -246,38 +246,58 @@ class VINTRPe <bits<2> op> : Enc32 {
   let Inst{31-26} = 0x32; // encoding
 }
 
-class MIMGe <bits<7> op> : Enc64 {
+class MIMGe : Enc64 {
   bits<8> vdata;
   bits<4> dmask;
   bits<1> unorm;
   bits<1> glc;
-  bits<1> da;
   bits<1> r128;
   bits<1> tfe;
   bits<1> lwe;
   bits<1> slc;
   bit d16;
-  bits<8> vaddr;
   bits<7> srsrc;
   bits<7> ssamp;
 
   let Inst{11-8} = dmask;
   let Inst{12} = unorm;
   let Inst{13} = glc;
-  let Inst{14} = da;
   let Inst{15} = r128;
   let Inst{16} = tfe;
   let Inst{17} = lwe;
-  let Inst{24-18} = op;
   let Inst{25} = slc;
   let Inst{31-26} = 0x3c;
-  let Inst{39-32} = vaddr;
   let Inst{47-40} = vdata;
   let Inst{52-48} = srsrc{6-2};
   let Inst{57-53} = ssamp{6-2};
   let Inst{63} = d16;
 }
 
+class MIMGe_gfx6789 <bits<7> op> : MIMGe {
+  bits<8> vaddr;
+  bits<1> da;
+
+  let Inst{14} = da;
+  let Inst{24-18} = op;
+  let Inst{39-32} = vaddr;
+}
+
+class MIMGe_gfx10 <bits<8> op> : MIMGe {
+  bits<8> vaddr0;
+  bits<3> dim;
+  bits<2> nsa;
+  bits<1> dlc;
+  bits<1> a16 = 0; // TODO: this should be an operand
+
+  let Inst{0} = op{7};
+  let Inst{2-1} = nsa;
+  let Inst{5-3} = dim;
+  let Inst{7} = dlc;
+  let Inst{24-18} = op{6-0};
+  let Inst{39-32} = vaddr0;
+  let Inst{62} = a16;
+}
+
 class EXPe : Enc64 {
   bits<4> en;
   bits<6> tgt;

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -3223,6 +3223,53 @@ bool SIInstrInfo::verifyInstruction(const MachineInstr &MI,
     }
   }
 
+  if (isMIMG(MI)) {
+    const MachineOperand *DimOp = getNamedOperand(MI, AMDGPU::OpName::dim);
+    if (DimOp) {
+      int VAddr0Idx = AMDGPU::getNamedOperandIdx(Opcode,
+                                                 AMDGPU::OpName::vaddr0);
+      int SRsrcIdx = AMDGPU::getNamedOperandIdx(Opcode, AMDGPU::OpName::srsrc);
+      const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(Opcode);
+      const AMDGPU::MIMGBaseOpcodeInfo *BaseOpcode =
+          AMDGPU::getMIMGBaseOpcodeInfo(Info->BaseOpcode);
+      const AMDGPU::MIMGDimInfo *Dim =
+          AMDGPU::getMIMGDimInfoByEncoding(DimOp->getImm());
+
+      if (!Dim) {
+        ErrInfo = "dim is out of range";
+        return false;
+      }
+
+      bool IsNSA = SRsrcIdx - VAddr0Idx > 1;
+      unsigned AddrWords = BaseOpcode->NumExtraArgs +
+                           (BaseOpcode->Gradients ? Dim->NumGradients : 0) +
+                           (BaseOpcode->Coordinates ? Dim->NumCoords : 0) +
+                           (BaseOpcode->LodOrClampOrMip ? 1 : 0);
+
+      unsigned VAddrWords;
+      if (IsNSA) {
+        VAddrWords = SRsrcIdx - VAddr0Idx;
+      } else {
+        const TargetRegisterClass *RC = getOpRegClass(MI, VAddr0Idx);
+        VAddrWords = MRI.getTargetRegisterInfo()->getRegSizeInBits(*RC) / 32;
+        if (AddrWords > 8)
+          AddrWords = 16;
+        else if (AddrWords > 4)
+          AddrWords = 8;
+        else if (AddrWords == 3 && VAddrWords == 4) {
+          // CodeGen uses the V4 variant of instructions for three addresses,
+          // because the selection DAG does not support non-power-of-two types.
+          AddrWords = 4;
+        }
+      }
+
+      if (VAddrWords != AddrWords) {
+        ErrInfo = "bad vaddr size";
+        return false;
+      }
+    }
+  }
+
   const MachineOperand *DppCt = getNamedOperand(MI, AMDGPU::OpName::dpp_ctrl);
   if (DppCt) {
     using namespace AMDGPU::DPP;
@@ -5356,25 +5403,35 @@ unsigned SIInstrInfo::getInstSizeInBytes(const MachineInstr &MI) const {
       return DescSize; // No operands.
 
     if (isLiteralConstantLike(MI.getOperand(Src0Idx), Desc.OpInfo[Src0Idx]))
-      return DescSize + 4;
+      return isVOP3(MI) ? 12 : (DescSize + 4);
 
     int Src1Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src1);
     if (Src1Idx == -1)
       return DescSize;
 
     if (isLiteralConstantLike(MI.getOperand(Src1Idx), Desc.OpInfo[Src1Idx]))
-      return DescSize + 4;
+      return isVOP3(MI) ? 12 : (DescSize + 4);
 
     int Src2Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src2);
     if (Src2Idx == -1)
       return DescSize;
 
     if (isLiteralConstantLike(MI.getOperand(Src2Idx), Desc.OpInfo[Src2Idx]))
-      return DescSize + 4;
+      return isVOP3(MI) ? 12 : (DescSize + 4);
 
     return DescSize;
   }
 
+  // Check whether we have extra NSA words.
+  if (isMIMG(MI)) {
+    int VAddr0Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::vaddr0);
+    if (VAddr0Idx < 0)
+      return 8;
+
+    int RSrcIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::srsrc);
+    return 8 + 4 * ((RSrcIdx - VAddr0Idx + 2) / 4);
+  }
+
   switch (Opc) {
   case TargetOpcode::IMPLICIT_DEF:
   case TargetOpcode::KILL:

diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -845,6 +845,7 @@ def exp_vm : NamedOperandBit<"ExpVM", NamedMatchClass<"ExpVM">>;
 def FORMAT : NamedOperandU8<"FORMAT", NamedMatchClass<"FORMAT">>;
 
 def DMask : NamedOperandU16<"DMask", NamedMatchClass<"DMask">>;
+def Dim : NamedOperandU8<"Dim", NamedMatchClass<"Dim", 0>>;
 
 def dpp_ctrl : NamedOperandU32<"DPPCtrl", NamedMatchClass<"DPPCtrl", 0>>;
 def row_mask : NamedOperandU32<"RowMask", NamedMatchClass<"RowMask">>;

diff --git a/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp b/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
@@ -38,6 +38,8 @@ class SIShrinkInstructions : public MachineFunctionPass {
 public:
   static char ID;
 
+  void shrinkMIMG(MachineInstr &MI);
+
 public:
   SIShrinkInstructions() : MachineFunctionPass(ID) {
   }
@@ -211,6 +213,96 @@ static void shrinkScalarCompare(const SIInstrInfo *TII, MachineInstr &MI) {
   }
 }
 
+// Shrink NSA encoded instructions with contiguous VGPRs to non-NSA encoding.
+void SIShrinkInstructions::shrinkMIMG(MachineInstr &MI) {
+  const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(MI.getOpcode());
+  if (Info->MIMGEncoding != AMDGPU::MIMGEncGfx10NSA)
+    return;
+
+  MachineFunction *MF = MI.getParent()->getParent();
+  const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+  const SIInstrInfo *TII = ST.getInstrInfo();
+  const SIRegisterInfo &TRI = TII->getRegisterInfo();
+  int VAddr0Idx =
+      AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::vaddr0);
+  unsigned NewAddrDwords = Info->VAddrDwords;
+  const TargetRegisterClass *RC;
+
+  if (Info->VAddrDwords == 2) {
+    RC = &AMDGPU::VReg_64RegClass;
+  } else if (Info->VAddrDwords == 3) {
+    RC = &AMDGPU::VReg_96RegClass;
+  } else if (Info->VAddrDwords == 4) {
+    RC = &AMDGPU::VReg_128RegClass;
+  } else if (Info->VAddrDwords <= 8) {
+    RC = &AMDGPU::VReg_256RegClass;
+    NewAddrDwords = 8;
+  } else {
+    RC = &AMDGPU::VReg_512RegClass;
+    NewAddrDwords = 16;
+  }
+
+  unsigned VgprBase = 0;
+  bool IsUndef = true;
+  bool IsKill = NewAddrDwords == Info->VAddrDwords;
+  for (unsigned i = 0; i < Info->VAddrDwords; ++i) {
+    const MachineOperand &Op = MI.getOperand(VAddr0Idx + i);
+    unsigned Vgpr = TRI.getHWRegIndex(Op.getReg());
+
+    if (i == 0) {
+      VgprBase = Vgpr;
+    } else if (VgprBase + i != Vgpr)
+      return;
+
+    if (!Op.isUndef())
+      IsUndef = false;
+    if (!Op.isKill())
+      IsKill = false;
+  }
+
+  if (VgprBase + NewAddrDwords > 256)
+    return;
+
+  // Further check for implicit tied operands - this may be present if TFE is
+  // enabled
+  int TFEIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::tfe);
+  int LWEIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::lwe);
+  unsigned TFEVal = MI.getOperand(TFEIdx).getImm();
+  unsigned LWEVal = MI.getOperand(LWEIdx).getImm();
+  int ToUntie = -1;
+  if (TFEVal || LWEVal) {
+    // TFE/LWE is enabled so we need to deal with an implicit tied operand
+    for (unsigned i = LWEIdx + 1, e = MI.getNumOperands(); i != e; ++i) {
+      if (MI.getOperand(i).isReg() && MI.getOperand(i).isTied() &&
+          MI.getOperand(i).isImplicit()) {
+        // This is the tied operand
+        assert(
+            ToUntie == -1 &&
+            "found more than one tied implicit operand when expecting only 1");
+        ToUntie = i;
+        MI.untieRegOperand(ToUntie);
+      }
+    }
+  }
+
+  unsigned NewOpcode =
+      AMDGPU::getMIMGOpcode(Info->BaseOpcode, AMDGPU::MIMGEncGfx10Default,
+                            Info->VDataDwords, NewAddrDwords);
+  MI.setDesc(TII->get(NewOpcode));
+  MI.getOperand(VAddr0Idx).setReg(RC->getRegister(VgprBase));
+  MI.getOperand(VAddr0Idx).setIsUndef(IsUndef);
+  MI.getOperand(VAddr0Idx).setIsKill(IsKill);
+
+  for (unsigned i = 1; i < Info->VAddrDwords; ++i)
+    MI.RemoveOperand(VAddr0Idx + 1);
+
+  if (ToUntie >= 0) {
+    MI.tieOperands(
+        AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::vdata),
+        ToUntie - (Info->VAddrDwords - 1));
+  }
+}
+
 /// Attempt to shink AND/OR/XOR operations requiring non-inlineable literals.
 /// For AND or OR, try using S_BITSET{0,1} to clear or set bits.
 /// If the inverse of the immediate is legal, use ANDN2, ORN2 or
@@ -597,6 +689,14 @@ bool SIShrinkInstructions::runOnMachineFunction(MachineFunction &MF) {
           continue;
       }
 
+      if (TII->isMIMG(MI.getOpcode()) &&
+          ST.getGeneration() >= AMDGPUSubtarget::GFX10 &&
+          MF.getProperties().hasProperty(
+              MachineFunctionProperties::Property::NoVRegs)) {
+        shrinkMIMG(MI);
+        continue;
+      }
+
       if (!TII->hasVALU32BitEncoding(MI.getOpcode()))
         continue;
 

diff --git a/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h b/llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h
@@ -200,6 +200,8 @@ struct MIMGDimInfo {
   uint8_t NumCoords;
   uint8_t NumGradients;
   bool DA;
+  uint8_t Encoding;
+  const char *AsmSuffix;
 };
 
 LLVM_READONLY

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.atomic.dim.ll
@@ -1,8 +1,10 @@
 ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefix=GCN %s
 
 ; GCN-LABEL: {{^}}atomic_swap_1d:
-; GCN: image_atomic_swap v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_swap v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_swap_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -11,7 +13,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_1d:
-; GCN: image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_add_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -20,7 +23,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_sub_1d:
-; GCN: image_atomic_sub v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_sub v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_sub v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_sub_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -29,7 +33,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_smin_1d:
-; GCN: image_atomic_smin v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_smin v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_smin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_smin_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -38,7 +43,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_umin_1d:
-; GCN: image_atomic_umin v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_umin v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_umin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_umin_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -47,7 +53,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_smax_1d:
-; GCN: image_atomic_smax v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_smax v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_smax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_smax_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -56,7 +63,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_umax_1d:
-; GCN: image_atomic_umax v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_umax v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_umax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_umax_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -65,7 +73,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_and_1d:
-; GCN: image_atomic_and v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_and v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_and v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_and_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.and.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -74,7 +83,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_or_1d:
-; GCN: image_atomic_or v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_or v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_or v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_or_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.or.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -83,7 +93,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_xor_1d:
-; GCN: image_atomic_xor v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_xor v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_xor v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_xor_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -92,7 +103,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_inc_1d:
-; GCN: image_atomic_inc v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_inc v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_inc v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_inc_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -101,7 +113,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_dec_1d:
-; GCN: image_atomic_dec v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_dec v0, v1, s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_dec v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_dec_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -110,7 +123,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_cmpswap_1d:
-; GCN: image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 unorm glc{{$}}
+; GFX6789: image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 unorm glc{{$}}
+; GFX10: image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps float @atomic_cmpswap_1d(<8 x i32> inreg %rsrc, i32 %cmp, i32 %swap, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i32(i32 %cmp, i32 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -119,7 +133,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_2d:
-; GCN: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm glc ;
 define amdgpu_ps float @atomic_add_2d(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i32(i32 %data, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
@@ -128,7 +143,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_3d:
-; GCN: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm glc ;
 define amdgpu_ps float @atomic_add_3d(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t, i32 %r) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.3d.i32.i32(i32 %data, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
@@ -137,7 +153,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_cube:
-; GCN: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX6789: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX10: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_CUBE unorm glc ;
 define amdgpu_ps float @atomic_add_cube(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t, i32 %face) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.cube.i32.i32(i32 %data, i32 %s, i32 %t, i32 %face, <8 x i32> %rsrc, i32 0, i32 0)
@@ -146,7 +163,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_1darray:
-; GCN: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX6789: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX10: image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc ;
 define amdgpu_ps float @atomic_add_1darray(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %slice) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i32(i32 %data, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
@@ -155,7 +173,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_2darray:
-; GCN: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX6789: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX10: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc ;
 define amdgpu_ps float @atomic_add_2darray(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t, i32 %slice) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i32(i32 %data, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
@@ -164,7 +183,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_2dmsaa:
-; GCN: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX6789: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc{{$}}
+; GFX10: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA unorm glc ;
 define amdgpu_ps float @atomic_add_2dmsaa(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t, i32 %fragid) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i32(i32 %data, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
@@ -173,7 +193,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_2darraymsaa:
-; GCN: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX6789: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 unorm glc da{{$}}
+; GFX10: image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc ;
 define amdgpu_ps float @atomic_add_2darraymsaa(<8 x i32> inreg %rsrc, i32 %data, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i32(i32 %data, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
@@ -182,7 +203,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}atomic_add_1d_slc:
-; GCN: image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc slc{{$}}
+; GFX6789: image_atomic_add v0, v1, s[0:7] dmask:0x1 unorm glc slc{{$}}
+; GFX10: image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc slc ;
 define amdgpu_ps float @atomic_add_1d_slc(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.d16.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.d16.dim.ll
@@ -1,9 +1,11 @@
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX81 %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX9 %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=GCN,UNPACKED,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX81,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX9,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}image_load_f16:
-; GCN: image_load v0, v[0:1], s[0:7] dmask:0x1 unorm d16{{$}}
+; GFX89: image_load v0, v[0:1], s[0:7] dmask:0x1 unorm d16{{$}}
+; GFX10: image_load v0, v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps half @image_load_f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
 main_body:
   %tex = call half @llvm.amdgcn.image.load.2d.f16.i32(i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
@@ -13,6 +15,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_load_v2f16:
 ; UNPACKED: image_load v[0:1], v[0:1], s[0:7] dmask:0x3 unorm d16{{$}}
 ; PACKED: image_load v0, v[0:1], s[0:7] dmask:0x3 unorm d16{{$}}
+; GFX10: image_load v0, v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps float @image_load_v2f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
 main_body:
   %tex = call <2 x half> @llvm.amdgcn.image.load.2d.v2f16.i32(i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
@@ -23,6 +26,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_load_v4f16:
 ; UNPACKED: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
 ; PACKED: image_load v[0:1], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
+; GFX10: image_load v[0:1], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps <2 x float> @image_load_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
 main_body:
   %tex = call <4 x half> @llvm.amdgcn.image.load.2d.v4f16.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
@@ -33,6 +37,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_load_mip_v4f16:
 ; UNPACKED: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm d16{{$}}
 ; PACKED: image_load_mip v[0:1], v[0:3], s[0:7] dmask:0xf unorm d16{{$}}
+; GFX10: image_load_mip v[0:1], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps <2 x float> @image_load_mip_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
 main_body:
   %tex = call <4 x half> @llvm.amdgcn.image.load.mip.2d.v4f16.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -43,6 +48,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_load_3d_v2f16:
 ; UNPACKED: image_load v[0:1], v[0:3], s[0:7] dmask:0x3 unorm d16{{$}}
 ; PACKED: image_load v0, v[0:3], s[0:7] dmask:0x3 unorm d16{{$}}
+; GFX10: image_load v0, v[0:2], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm d16{{$}}
 define amdgpu_ps float @image_load_3d_v2f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %r) {
 main_body:
   %tex = call <2 x half> @llvm.amdgcn.image.load.3d.v2f16.i32(i32 3, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
@@ -51,7 +57,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}image_store_f16
-; GCN: image_store v2, v[0:1], s[0:7] dmask:0x1 unorm d16{{$}}
+; GFX89: image_store v2, v[0:1], s[0:7] dmask:0x1 unorm d16{{$}}
+; GFX10: image_store v2, v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps void @image_store_f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, half %data) {
 main_body:
   call void @llvm.amdgcn.image.store.2d.f16.i32(half %data, i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
@@ -63,6 +70,7 @@ main_body:
 ; UNPACKED: v_and_b32_e32
 ; UNPACKED: image_store v[{{[0-9:]+}}], v[0:1], s[0:7] dmask:0x3 unorm d16{{$}}
 ; PACKED: image_store v2, v[0:1], s[0:7] dmask:0x3 unorm d16{{$}}
+; GFX10: image_store v2, v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps void @image_store_v2f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, float %in) {
 main_body:
   %data = bitcast float %in to <2 x half>
@@ -77,6 +85,7 @@ main_body:
 ; UNPACKED: v_and_b32_e32
 ; UNPACKED: image_store v[{{[0-9:]+}}], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
 ; PACKED: image_store v[2:3], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
+; GFX10: image_store v[2:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm d16{{$}}
 define amdgpu_ps void @image_store_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <2 x float> %in) {
 main_body:
   %data = bitcast <2 x float> %in to <4 x half>
@@ -91,6 +100,7 @@ main_body:
 ; UNPACKED: v_and_b32_e32
 ; UNPACKED: image_store_mip v[{{[0-9:]+}}], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
 ; PACKED: image_store_mip v[2:3], v[0:1], s[0:7] dmask:0xf unorm d16{{$}}
+; GFX10: image_store_mip v[2:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm d16{{$}}
 define amdgpu_ps void @image_store_mip_1d_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %mip, <2 x float> %in) {
 main_body:
   %data = bitcast <2 x float> %in to <4 x half>

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.d16.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.d16.dim.ll
@@ -1,10 +1,12 @@
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX81 %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX9 %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=GCN,UNPACKED,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX81,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX9,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}image_gather4_b_2d_v4f16:
 ; UNPACKED: image_gather4_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x4 d16{{$}}
 ; PACKED: image_gather4_b v[0:1], v[0:3], s[0:7], s[8:11] dmask:0x4 d16{{$}}
+; GFX10: image_gather4_b v[0:1], v[0:2], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D d16{{$}}
 define amdgpu_ps <2 x float> @image_gather4_b_2d_v4f16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %s, float %t) {
 main_body:
   %tex = call <4 x half> @llvm.amdgcn.image.gather4.b.2d.v4f16.f32.f32(i32 4, float %bias, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.dim.ll
@@ -1,136 +1,154 @@
-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX6789 %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX6789 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}gather4_2d:
-; GCN: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_cube:
-; GCN: image_gather4 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 da{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 da{{$}}
+; GFX10: image_gather4 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_CUBE ;
 define amdgpu_ps <4 x float> @gather4_cube(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t, float %face) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.cube.v4f32.f32(i32 1, float %s, float %t, float %face, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_2darray:
-; GCN: image_gather4 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 da{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 da{{$}}
+; GFX10: image_gather4 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY ;
 define amdgpu_ps <4 x float> @gather4_2darray(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t, float %slice) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2darray.v4f32.f32(i32 1, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_2d:
-; GCN: image_gather4_c v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.2d.v4f32.f32(i32 1, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_cl_2d:
-; GCN: image_gather4_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.cl.2d.v4f32.f32(i32 1, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_cl_2d:
-; GCN: image_gather4_c_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.2d.v4f32.f32(i32 1, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_b_2d:
-; GCN: image_gather4_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_b v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_b_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.2d.v4f32.f32.f32(i32 1, float %bias, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_b_2d:
-; GCN: image_gather4_c_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_b_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %zcompare, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_b_cl_2d:
-; GCN: image_gather4_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_b_cl_2d:
-; GCN: image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_l_2d:
-; GCN: image_gather4_l v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_l v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_l v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_l_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.l.2d.v4f32.f32(i32 1, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_l_2d:
-; GCN: image_gather4_c_l v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c_l v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_l v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_l_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s, float %t, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.l.2d.v4f32.f32(i32 1, float %zcompare, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_lz_2d:
-; GCN: image_gather4_lz v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_lz v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_lz v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_lz_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.lz.2d.v4f32.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_lz_2d:
-; GCN: image_gather4_c_lz v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX6789: image_gather4_c_lz v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_lz v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_lz_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.lz.2d.v4f32.f32(i32 1, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_2d_dmask_2:
-; GCN: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x2{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x2{{$}}
+; GFX10: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x2 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_2d_dmask_2(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f32(i32 2, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_2d_dmask_4:
-; GCN: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x4{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x4{{$}}
+; GFX10: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_2d_dmask_4(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f32(i32 4, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_2d_dmask_8:
-; GCN: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x8{{$}}
+; GFX6789: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x8{{$}}
+; GFX10: image_gather4 v[0:3], v[0:1], s[0:7], s[8:11] dmask:0x8 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_2d_dmask_8(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.2d.v4f32.f32(i32 8, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.getlod.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.getlod.dim.ll
@@ -1,8 +1,10 @@
-; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=GCN %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck --check-prefix=GCN %s
+; RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck --check-prefixes=GCN,PRE-GFX10 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck --check-prefixes=GCN,PRE-GFX10 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck --check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}getlod_1d:
-; GCN: image_get_lod v[0:3], v0, s[0:7], s[8:11] dmask:0xf{{$}}
+; PRE-GFX10: image_get_lod v[0:3], v0, s[0:7], s[8:11] dmask:0xf{{$}}
+; GFX10: image_get_lod v[0:3], v0, s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GCN: s_waitcnt vmcnt(0)
 define amdgpu_ps <4 x float> @getlod_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s) {
 main_body:
@@ -11,7 +13,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}getlod_2d:
-; GCN: image_get_lod v[0:1], v[0:1], s[0:7], s[8:11] dmask:0x3{{$}}
+; PRE-GFX10: image_get_lod v[0:1], v[0:1], s[0:7], s[8:11] dmask:0x3{{$}}
+; GFX10: image_get_lod v[0:1], v[0:1], s[0:7], s[8:11] dmask:0x3 dim:SQ_RSRC_IMG_2D
 ; GCN: s_waitcnt vmcnt(0)
 define amdgpu_ps <2 x float> @getlod_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.nsa.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.nsa.ll
@@ -0,0 +1,91 @@
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-nsa-encoding -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=GCN,NONSA %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=GCN,NSA %s
+
+; GCN-LABEL: {{^}}sample_2d:
+;
+; TODO: use NSA here
+; GCN: v_mov_b32_e32 v2, v0
+;
+; GCN: image_sample v[0:3], v[1:2],
+define amdgpu_ps <4 x float> @sample_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %t, float %s) {
+main_body:
+  %v = call <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32(i32 15, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  ret <4 x float> %v
+}
+
+; GCN-LABEL: {{^}}sample_3d:
+; NONSA: v_mov_b32_e32 v3, v0
+; NONSA: image_sample v[0:3], v[1:4],
+; NSA: image_sample v[0:3], [v1, v2, v0],
+define amdgpu_ps <4 x float> @sample_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %r, float %s, float %t) {
+main_body:
+  %v = call <4 x float> @llvm.amdgcn.image.sample.3d.v4f32.f32(i32 15, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  ret <4 x float> %v
+}
+
+; GCN-LABEL: {{^}}sample_d_3d:
+; NSA: image_sample_d v[0:3], [v3, v8, v7, v5, v4, v6, v0, v2, v1],
+define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %r, float %t, float %dsdh, float %dtdv, float %dsdv, float %drdv, float %drdh, float %dtdh) {
+main_body:
+  %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f32(i32 15, float %dsdh, float %dtdh, float %drdh, float %dsdv, float %dtdv, float %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  ret <4 x float> %v
+}
+
+; GCN-LABEL: {{^}}sample_contig_nsa:
+; GCN: image_sample_c_l v0, v[0:7],
+; NSA: image_sample v1, [v6, v7, v5],
+define amdgpu_ps <2 x float> @sample_contig_nsa(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s1, float %t1, float %r1, float %lod, float %r2, float %s2, float %t2) {
+main_body:
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %r.0 = insertelement <2 x float> undef, float %v1, i32 0
+  %r = insertelement <2 x float> %r.0, float %v2, i32 1
+  ret <2 x float> %r
+}
+
+; GCN-LABEL: {{^}}sample_nsa_nsa:
+; NSA: image_sample_c_l v0, [v1, v2, v3, v4, v0],
+; NSA: image_sample v1, [v6, v7, v5],
+define amdgpu_ps <2 x float> @sample_nsa_nsa(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %lod, float %zcompare, float %s1, float %t1, float %r1, float %r2, float %s2, float %t2) {
+main_body:
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %r.0 = insertelement <2 x float> undef, float %v1, i32 0
+  %r = insertelement <2 x float> %r.0, float %v2, i32 1
+  ret <2 x float> %r
+}
+
+; GCN-LABEL: {{^}}sample_nsa_contig:
+; NSA: image_sample_c_l v0, [v1, v2, v3, v4, v0],
+; NSA: image_sample v1, v[5:7],
+define amdgpu_ps <2 x float> @sample_nsa_contig(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %lod, float %zcompare, float %s1, float %t1, float %r1, float %s2, float %t2, float %r2) {
+main_body:
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %r.0 = insertelement <2 x float> undef, float %v1, i32 0
+  %r = insertelement <2 x float> %r.0, float %v2, i32 1
+  ret <2 x float> %r
+}
+
+; GCN-LABEL: {{^}}sample_contig_contig:
+; GCN: image_sample_c_l v0, v[0:7],
+; NSA: image_sample v1, v[5:7],
+; NONSA: image_sample v1, v[5:8],
+define amdgpu_ps <2 x float> @sample_contig_contig(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s1, float %t1, float %r1, float %lod, float %s2, float %t2, float %r2) {
+main_body:
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %r.0 = insertelement <2 x float> undef, float %v1, i32 0
+  %r = insertelement <2 x float> %r.0, float %v2, i32 1
+  ret <2 x float> %r
+}
+
+
+declare <4 x float> @llvm.amdgcn.image.sample.2d.v4f32.f32(i32, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
+declare <4 x float> @llvm.amdgcn.image.sample.3d.v4f32.f32(i32, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
+declare <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f32(i32, float, float, float, float, float, float, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
+
+declare float @llvm.amdgcn.image.sample.3d.f32.f32(i32, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
+declare float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32, float, float, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32) #1
+
+attributes #1 = { nounwind readonly }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.d16.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.d16.dim.ll
@@ -1,9 +1,11 @@
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX81 %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX9 %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=GCN,UNPACKED,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX81,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -check-prefixes=GCN,PACKED,GFX9,GFX89 %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}image_sample_2d_f16:
-; GCN: image_sample v0, v[0:1], s[0:7], s[8:11] dmask:0x1 d16{{$}}
+; GFX89: image_sample v0, v[0:1], s[0:7], s[8:11] dmask:0x1 d16{{$}}
+; GFX10: image_sample v0, v[0:1], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D d16{{$}}
 define amdgpu_ps half @image_sample_2d_f16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t) {
 main_body:
   %tex = call half @llvm.amdgcn.image.sample.2d.f16.f32(i32 1, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
@@ -26,6 +28,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_sample_c_d_1d_v2f16:
 ; UNPACKED: image_sample_c_d v[0:1], v[0:3], s[0:7], s[8:11] dmask:0x3 d16{{$}}
 ; PACKED: image_sample_c_d v0, v[0:3], s[0:7], s[8:11] dmask:0x3 d16{{$}}
+; GFX10: image_sample_c_d v0, v[0:3], s[0:7], s[8:11] dmask:0x3 dim:SQ_RSRC_IMG_1D d16{{$}}
 define amdgpu_ps float @image_sample_c_d_1d_v2f16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, float %s) {
 main_body:
   %tex = call <2 x half> @llvm.amdgcn.image.sample.c.d.1d.v2f16.f32.f32(i32 3, float %zcompare, float %dsdh, float %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
@@ -52,6 +55,7 @@ main_body:
 ; GCN-LABEL: {{^}}image_sample_b_2d_v4f16:
 ; UNPACKED: image_sample_b v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf d16{{$}}
 ; PACKED: image_sample_b v[0:1], v[0:3], s[0:7], s[8:11] dmask:0xf d16{{$}}
+; GFX10: image_sample_b v[0:1], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D d16{{$}}
 define amdgpu_ps <2 x float> @image_sample_b_2d_v4f16(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %s, float %t) {
 main_body:
   %tex = call <4 x half> @llvm.amdgcn.image.sample.b.2d.v4f16.f32.f32(i32 15, float %bias, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.load.d16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.load.d16.ll
@@ -1,20 +1,23 @@
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -show-mc-encoding | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -show-mc-encoding | FileCheck -enable-var-scope -check-prefixes=GCN,UNPACKED,PREGFX10,PREGFX10-UNPACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,GFX10,GFX10-PACKED %s
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_x:
-; GCN: tbuffer_load_format_d16_x v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0
+; PREGFX10: tbuffer_load_format_d16_x v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], dfmt:6, nfmt:1, 0
+; GFX10: tbuffer_load_format_d16_x v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}], format:22, 0
 define amdgpu_ps half @tbuffer_load_d16_x(<4 x i32> inreg %rsrc) {
 main_body:
   %data = call half @llvm.amdgcn.raw.tbuffer.load.f16(<4 x i32> %rsrc, i32 0, i32 0, i32 22, i32 0)
   ret half %data
 }
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_xy:
-; UNPACKED: tbuffer_load_format_d16_xy v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0
+; PREGFX10-UNPACKED: tbuffer_load_format_d16_xy v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}], dfmt:6, nfmt:1, 0
 ; UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
 
-; PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], off, s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0
+; PREGFX10-PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], dfmt:6, nfmt:1, 0
+; GFX10-PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], format:22, 0
 ; PACKED: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v[[FULL]]
 define amdgpu_ps half @tbuffer_load_d16_xy(<4 x i32> inreg %rsrc) {
 main_body:
@@ -24,10 +27,12 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_xyzw:
-; UNPACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0
+; PREGFX10-UNPACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}], dfmt:6, nfmt:1, 0
+; GFX10-UNPACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}], format:22, 0
 ; UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
 
-; PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0
+; PREGFX10-PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}], dfmt:6, nfmt:1, 0
+; GFX10-PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}], format:22, 0
 ; PACKED: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v[[HI]]
 define amdgpu_ps half @tbuffer_load_d16_xyzw(<4 x i32> inreg %rsrc) {
 main_body:

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.store.d16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.store.d16.ll
@@ -1,36 +1,39 @@
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx810 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX81 %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED -check-prefix=GFX9 %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,UNPACKED,PREGFX10,PREGFX10-UNPACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,GFX10,GFX10-PACKED %s
 
 
 ; GCN-LABEL: {{^}}tbuffer_store_d16_x:
-; GCN: s_load_dwordx4
-; GCN: s_load_dword s[[S_LO:[0-9]+]]
-; GCN: v_mov_b32_e32 v[[V_LO:[0-9]+]], s[[S_LO]]
-; GCN: tbuffer_store_format_d16_x v[[V_LO]], off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; GCN-DAG: s_load_dwordx4
+; GCN-DAG: s_load_dword s[[S_LO:[0-9]+]]
+; GCN-DAG: v_mov_b32_e32 v[[V_LO:[0-9]+]], s[[S_LO]]
+; PREGFX10: tbuffer_store_format_d16_x v[[V_LO]], off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; GFX10: tbuffer_store_format_d16_x v[[V_LO]], off, s[{{[0-9]+:[0-9]+}}],  format:33, 0
 define amdgpu_kernel void @tbuffer_store_d16_x(<4 x i32> %rsrc, half %data) {
 main_body:
   call void @llvm.amdgcn.raw.tbuffer.store.f16(half %data, <4 x i32> %rsrc, i32 0, i32 0, i32 33, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}tbuffer_store_d16_xy:
-; GCN: s_load_dword [[S_DATA:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, 0x10
+; GCN: s_load_dword [[S_DATA:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}},
 ; UNPACKED-DAG: s_lshr_b32 [[SHR:s[0-9]+]], [[S_DATA]], 16
 ; UNPACKED-DAG: s_and_b32 [[MASKED:s[0-9]+]], [[S_DATA]], 0xffff{{$}}
 ; UNPACKED-DAG: v_mov_b32_e32 v[[V_LO:[0-9]+]], [[MASKED]]
 ; UNPACKED-DAG: v_mov_b32_e32 v[[V_HI:[0-9]+]], [[SHR]]
-; UNPACKED: tbuffer_store_format_d16_xy v{{\[}}[[V_LO]]:[[V_HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; PREGFX10-UNPACKED: tbuffer_store_format_d16_xy v{{\[}}[[V_LO]]:[[V_HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
 
-; PACKED: tbuffer_store_format_d16_xy v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; PREGFX10-PACKED: tbuffer_store_format_d16_xy v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; GFX10-PACKED: tbuffer_store_format_d16_xy v{{[0-9]+}}, off, s[{{[0-9]+:[0-9]+}}],  format:33, 0
 define amdgpu_kernel void @tbuffer_store_d16_xy(<4 x i32> %rsrc, <2 x half> %data) {
 main_body:
   call void @llvm.amdgcn.raw.tbuffer.store.v2f16(<2 x half> %data, <4 x i32> %rsrc, i32 0, i32 0, i32 33, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}tbuffer_store_d16_xyzw:
-; GCN-DAG: s_load_dwordx2 s{{\[}}[[S_DATA_0:[0-9]+]]:[[S_DATA_1:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}}, 0x10
+; GCN-DAG: s_load_dwordx2 s{{\[}}[[S_DATA_0:[0-9]+]]:[[S_DATA_1:[0-9]+]]{{\]}}, s{{\[[0-9]+:[0-9]+\]}},
 
 ; UNPACKED-DAG: s_mov_b32 [[K:s[0-9]+]], 0xffff{{$}}
 ; UNPACKED-DAG: s_lshr_b32 [[SHR0:s[0-9]+]], s[[S_DATA_0]], 16
@@ -40,12 +43,13 @@ main_body:
 
 ; UNPACKED-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], [[MASKED0]]
 ; UNPACKED-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], [[SHR1]]
-; UNPACKED: tbuffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; PREGFX10-UNPACKED: tbuffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
 
 
 ; PACKED-DAG: v_mov_b32_e32 v[[LO:[0-9]+]], s[[S_DATA_0]]
 ; PACKED-DAG: v_mov_b32_e32 v[[HI:[0-9]+]], s[[S_DATA_1]]
-; PACKED: tbuffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; PREGFX10-PACKED: tbuffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  dfmt:1,  nfmt:2, 0
+; GFX10-PACKED: tbuffer_store_format_d16_xyzw v{{\[}}[[LO]]:[[HI]]{{\]}}, off, s[{{[0-9]+:[0-9]+}}],  format:33, 0
 define amdgpu_kernel void @tbuffer_store_d16_xyzw(<4 x i32> %rsrc, <4 x half> %data) {
 main_body:
   call void @llvm.amdgcn.raw.tbuffer.store.v4f16(<4 x half> %data, <4 x i32> %rsrc, i32 0, i32 0, i32 33, i32 0)

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.store.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.tbuffer.store.ll
@@ -1,11 +1,16 @@
-;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck -check-prefixes=GCN,VERDE %s
-;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefix=GCN %s
+;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck -check-prefixes=GCN,VERDE,PREGFX10 %s
+;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck -check-prefixes=GCN,PREGFX10 %s
+;RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}tbuffer_store:
-; GCN: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:12, nfmt:2, 0
-; GCN: tbuffer_store_format_xyzw v[4:7], off, s[0:3], dfmt:13, nfmt:3, 0 glc
-; GCN: tbuffer_store_format_xyzw v[8:11], off, s[0:3], dfmt:14, nfmt:4, 0 slc
-; GCN: tbuffer_store_format_xyzw v[8:11], off, s[0:3], dfmt:14, nfmt:4, 0
+; PREGFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:12, nfmt:2, 0
+; PREGFX10: tbuffer_store_format_xyzw v[4:7], off, s[0:3], dfmt:13, nfmt:3, 0 glc
+; PREGFX10: tbuffer_store_format_xyzw v[8:11], off, s[0:3], dfmt:14, nfmt:4, 0 slc
+; PREGFX10: tbuffer_store_format_xyzw v[8:11], off, s[0:3], dfmt:14, nfmt:4, 0 glc
+; GFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], format:44, 0
+; GFX10: tbuffer_store_format_xyzw v[4:7], off, s[0:3], format:61, 0 glc
+; GFX10: tbuffer_store_format_xyzw v[8:11], off, s[0:3], format:78, 0 slc
+; GFX10: tbuffer_store_format_xyzw v[8:11], off, s[0:3], format:78, 0 glc dlc
 define amdgpu_ps void @tbuffer_store(<4 x i32> inreg, <4 x float>, <4 x float>, <4 x float>) {
 main_body:
   %in1 = bitcast <4 x float> %1 to <4 x i32>
@@ -14,12 +19,13 @@ main_body:
   call void @llvm.amdgcn.raw.tbuffer.store.v4i32(<4 x i32> %in1, <4 x i32> %0, i32 0, i32 0, i32 44, i32 0)
   call void @llvm.amdgcn.raw.tbuffer.store.v4i32(<4 x i32> %in2, <4 x i32> %0, i32 0, i32 0, i32 61, i32 1)
   call void @llvm.amdgcn.raw.tbuffer.store.v4i32(<4 x i32> %in3, <4 x i32> %0, i32 0, i32 0, i32 78, i32 2)
-  call void @llvm.amdgcn.raw.tbuffer.store.v4f32(<4 x float> %3, <4 x i32> %0, i32 0, i32 0, i32 78, i32 0)
+  call void @llvm.amdgcn.raw.tbuffer.store.v4f32(<4 x float> %3, <4 x i32> %0, i32 0, i32 0, i32 78, i32 5)
   ret void
 }
 
 ; GCN-LABEL: {{^}}tbuffer_store_immoffs:
-; GCN: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:5, nfmt:7, 0 offset:42
+; PREGFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:5, nfmt:7, 0 offset:42
+; GFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], format:117, 0 offset:42
 define amdgpu_ps void @tbuffer_store_immoffs(<4 x i32> inreg, <4 x float>) {
 main_body:
   %in1 = bitcast <4 x float> %1 to <4 x i32>
@@ -28,7 +34,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}tbuffer_store_scalar_and_imm_offs:
-; GCN: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:5, nfmt:7, {{s[0-9]+}} offset:42
+; PREGFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], dfmt:5, nfmt:7, {{s[0-9]+}} offset:42
+; GFX10: tbuffer_store_format_xyzw v[0:3], off, s[0:3], format:117, {{s[0-9]+}} offset:42
 define amdgpu_ps void @tbuffer_store_scalar_and_imm_offs(<4 x i32> inreg, <4 x float> %vdata, i32 inreg %soffset) {
 main_body:
   %in1 = bitcast <4 x float> %vdata to <4 x i32>
@@ -37,7 +44,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}buffer_store_ofs:
-; GCN: tbuffer_store_format_xyzw v[0:3], v4, s[0:3], dfmt:3, nfmt:7, 0 offen
+; PREGFX10: tbuffer_store_format_xyzw v[0:3], v4, s[0:3], dfmt:3, nfmt:7, 0 offen
+; GFX10: tbuffer_store_format_xyzw v[0:3], v4, s[0:3], format:115, 0 offen
 define amdgpu_ps void @buffer_store_ofs(<4 x i32> inreg, <4 x float> %vdata, i32 %voffset) {
 main_body:
   %in1 = bitcast <4 x float> %vdata to <4 x i32>
@@ -46,7 +54,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}buffer_store_x1:
-; GCN: tbuffer_store_format_x v0, off, s[0:3], dfmt:13, nfmt:7, 0
+; PREGFX10: tbuffer_store_format_x v0, off, s[0:3], dfmt:13, nfmt:7, 0
+; GFX10: tbuffer_store_format_x v0, off, s[0:3], format:125, 0
 define amdgpu_ps void @buffer_store_x1(<4 x i32> inreg %rsrc, float %data) {
 main_body:
   %data.i = bitcast float %data to i32
@@ -55,7 +64,8 @@ main_body:
 }
 
 ; GCN-LABEL: {{^}}buffer_store_x2:
-; GCN: tbuffer_store_format_xy v[0:1], off, s[0:3], dfmt:1, nfmt:2, 0
+; PREGFX10: tbuffer_store_format_xy v[0:1], off, s[0:3], dfmt:1, nfmt:2, 0
+; GFX10: tbuffer_store_format_xy v[0:1], off, s[0:3], format:33, 0
 define amdgpu_ps void @buffer_store_x2(<4 x i32> inreg %rsrc, <2 x float> %data) {
 main_body:
   %data.i = bitcast <2 x float> %data to <2 x i32>

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.tbuffer.load.d16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.struct.tbuffer.load.d16.ll
@@ -1,10 +1,12 @@
-; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -show-mc-encoding | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=UNPACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED %s
-; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefix=GCN -check-prefix=PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs -show-mc-encoding | FileCheck -enable-var-scope -check-prefixes=GCN,UNPACKED,PREGFX10,PREGFX10-UNPACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx810 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx900 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,PREGFX10,PREGFX10-PACKED %s
+; RUN: llc < %s -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs | FileCheck -enable-var-scope -check-prefixes=GCN,PACKED,GFX10,GFX10-PACKED %s
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_x:
 ; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0
-; GCN: tbuffer_load_format_d16_x v{{[0-9]+}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; PREGFX10: tbuffer_load_format_d16_x v{{[0-9]+}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; GFX10: tbuffer_load_format_d16_x v{{[0-9]+}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  format:22, 0 idxen
 define amdgpu_ps half @tbuffer_load_d16_x(<4 x i32> inreg %rsrc) {
 main_body:
   %data = call half @llvm.amdgcn.struct.tbuffer.load.f16(<4 x i32> %rsrc, i32 0, i32 0, i32 0, i32 22, i32 0)
@@ -13,10 +15,11 @@ main_body:
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_xy:
 ; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0
-; UNPACKED: tbuffer_load_format_d16_xy v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
-; UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
+; PREGFX10-UNPACKED: tbuffer_load_format_d16_xy v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; PREGFX10-UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
 
-; PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; PREGFX10-PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; GFX10-PACKED: tbuffer_load_format_d16_xy v[[FULL:[0-9]+]], [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  format:22, 0 idxen
 ; PACKED: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v[[FULL]]
 define amdgpu_ps half @tbuffer_load_d16_xy(<4 x i32> inreg %rsrc) {
 main_body:
@@ -27,10 +30,11 @@ main_body:
 
 ; GCN-LABEL: {{^}}tbuffer_load_d16_xyzw:
 ; GCN: v_mov_b32_e32 [[ZEROREG:v[0-9]+]], 0
-; UNPACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
-; UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
+; PREGFX10-UNPACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; PREGFX10-UNPACKED: v_mov_b32_e32 v{{[0-9]+}}, v[[HI]]
 
-; PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; PREGFX10-PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  dfmt:6,  nfmt:1, 0 idxen
+; GFX10-PACKED: tbuffer_load_format_d16_xyzw v{{\[}}{{[0-9]+}}:[[HI:[0-9]+]]{{\]}}, [[ZEROREG]], s[{{[0-9]+:[0-9]+}}],  format:22, 0 idxen
 ; PACKED: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v[[HI]]
 define amdgpu_ps half @tbuffer_load_d16_xyzw(<4 x i32> inreg %rsrc) {
 main_body: