[AMDGPU] gfx1010 GCNNSAReassign pass

[rampitec]

Summary

Convert NSA into non-NSA images.

[kzhuravl]

LGTM

[kzhuravl]

lib/Target/AMDGPU/GCNNSAReassign.cpp

5–6 ↗

(On Diff #197437)

Needs to be updated to match the new license. Can be changed before submitting. Thanks.

llvm/trunk/lib/Target/AMDGPU/AMDGPU.h

 
 Pass *createAMDGPUFunctionInliningPass();
 void initializeAMDGPUInlinerPass(PassRegistry&);
 
 ModulePass *createAMDGPUOpenCLEnqueuedBlockLoweringPass();
 void initializeAMDGPUOpenCLEnqueuedBlockLoweringPass(PassRegistry &);
 extern char &AMDGPUOpenCLEnqueuedBlockLoweringID;
 
+void initializeGCNNSAReassignPass(PassRegistry &);
+extern char &GCNNSAReassignID;
+
 Target &getTheAMDGPUTarget();
 Target &getTheGCNTarget();
 
 namespace AMDGPU {
 enum TargetIndex {
   TI_CONSTDATA_START,
   TI_SCRATCH_RSRC_DWORD0,
   TI_SCRATCH_RSRC_DWORD1,

llvm/trunk/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

static cl::opt<bool> EnableLibCallSimplify(
   cl::Hidden);
 
 static cl::opt<bool> EnableLowerKernelArguments(
   "amdgpu-ir-lower-kernel-arguments",
   cl::desc("Lower kernel argument loads in IR pass"),
   cl::init(true),
   cl::Hidden);
 
+static cl::opt<bool> EnableRegReassign(
+  "amdgpu-reassign-regs",
+  cl::desc("Enable register reassign optimizations on gfx10+"),
+  cl::init(true),
+  cl::Hidden);
+
 // Enable atomic optimization
 static cl::opt<bool> EnableAtomicOptimizations(
   "amdgpu-atomic-optimizations",
   cl::desc("Enable atomic optimizations"),
   cl::init(false),
   cl::Hidden);
 
 // Enable Mode register optimization
extern "C" void LLVMInitializeAMDGPUTarget() {
   initializeSIPreAllocateWWMRegsPass(*PR);
   initializeSIFormMemoryClausesPass(*PR);
   initializeAMDGPUUnifyDivergentExitNodesPass(*PR);
   initializeAMDGPUAAWrapperPassPass(*PR);
   initializeAMDGPUExternalAAWrapperPass(*PR);
   initializeAMDGPUUseNativeCallsPass(*PR);
   initializeAMDGPUSimplifyLibCallsPass(*PR);
   initializeAMDGPUInlinerPass(*PR);
+  initializeGCNNSAReassignPass(*PR);
 }
 
 static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
   return llvm::make_unique<AMDGPUTargetObjectFile>();
 }
 
 static ScheduleDAGInstrs *createR600MachineScheduler(MachineSchedContext *C) {
   return new ScheduleDAGMILive(C, llvm::make_unique<R600SchedStrategy>());
public:
   bool addInstSelector() override;
   bool addIRTranslator() override;
   bool addLegalizeMachineIR() override;
   bool addRegBankSelect() override;
   bool addGlobalInstructionSelect() override;
   void addFastRegAlloc() override;
   void addOptimizedRegAlloc() override;
   void addPreRegAlloc() override;
+  bool addPreRewrite() override;
   void addPostRegAlloc() override;
   void addPreSched2() override;
   void addPreEmitPass() override;
 };
 
 } // end anonymous namespace
 
 void AMDGPUPassConfig::addEarlyCSEOrGVNPass() {
void GCNPassConfig::addOptimizedRegAlloc() {
   insertPass(&RegisterCoalescerID, &SIPreAllocateWWMRegsID, false);
 
   if (EnableDCEInRA)
     insertPass(&RenameIndependentSubregsID, &DeadMachineInstructionElimID);
 
   TargetPassConfig::addOptimizedRegAlloc();
 }
 
+bool GCNPassConfig::addPreRewrite() {
+  if (EnableRegReassign) {
+    addPass(&GCNNSAReassignID);
+  }
+  return true;
+}
+
 void GCNPassConfig::addPostRegAlloc() {
   addPass(&SIFixVGPRCopiesID);
   if (getOptLevel() > CodeGenOpt::None)
     addPass(&SIOptimizeExecMaskingID);
   TargetPassConfig::addPostRegAlloc();
 }
 
 void GCNPassConfig::addPreSched2() {

llvm/trunk/lib/Target/AMDGPU/CMakeLists.txt

add_llvm_target(AMDGPUCodeGen
   SIMemoryLegalizer.cpp
   SIOptimizeExecMasking.cpp
   SIOptimizeExecMaskingPreRA.cpp
   SIPeepholeSDWA.cpp
   SIRegisterInfo.cpp
   SIShrinkInstructions.cpp
   SIWholeQuadMode.cpp
   GCNILPSched.cpp
+  GCNNSAReassign.cpp
   GCNDPPCombine.cpp
   SIModeRegister.cpp
   )
 
 add_subdirectory(AsmParser)
 add_subdirectory(Disassembler)
 add_subdirectory(InstPrinter)
 add_subdirectory(MCTargetDesc)
 add_subdirectory(TargetInfo)
 add_subdirectory(Utils)

llvm/trunk/lib/Target/AMDGPU/GCNNSAReassign.cpp

+//===-- GCNNSAReassign.cpp - Reassign registers in NSA unstructions -------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+/// \file
+/// \brief Try to reassign registers on GFX10+ from non-sequential to sequential
+/// in NSA image instructions. Later SIShrinkInstructions pass will relace NSA
+/// with sequential versions where possible.
+///
+//===----------------------------------------------------------------------===//
+
+#include "AMDGPU.h"
+#include "AMDGPUSubtarget.h"
+#include "SIInstrInfo.h"
+#include "SIMachineFunctionInfo.h"
+#include "llvm/ADT/Statistic.h"
+#include "llvm/CodeGen/LiveInterval.h"
+#include "llvm/CodeGen/LiveIntervals.h"
+#include "llvm/CodeGen/LiveRegMatrix.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/VirtRegMap.h"
+#include "llvm/Support/MathExtras.h"
+#include <algorithm>
+
+using namespace llvm;
+
+#define DEBUG_TYPE "amdgpu-nsa-reassign"
+
+STATISTIC(NumNSAInstructions,
+          "Number of NSA instructions with non-sequential address found");
+STATISTIC(NumNSAConverted,
+          "Number of NSA instructions changed to sequential");
+
+namespace {
+
+class GCNNSAReassign : public MachineFunctionPass {
+public:
+  static char ID;
+
+  GCNNSAReassign() : MachineFunctionPass(ID) {
+    initializeGCNNSAReassignPass(*PassRegistry::getPassRegistry());
+  }
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+
+  StringRef getPassName() const override { return "GCN NSA Reassign"; }
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.addRequired<LiveIntervals>();
+    AU.addRequired<VirtRegMap>();
+    AU.addRequired<LiveRegMatrix>();
+    AU.setPreservesAll();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
+private:
+  typedef enum {
+    NOT_NSA,        // Not an NSA instruction
+    FIXED,          // NSA which we cannot modify
+    NON_CONTIGUOUS, // NSA with non-sequential address which we can try
+                    // to optimize.
+    CONTIGUOUS      // NSA with all sequential address registers
+  } NSA_Status;
+
+  const GCNSubtarget *ST;
+
+  const MachineRegisterInfo *MRI;
+
+  const SIRegisterInfo *TRI;
+
+  VirtRegMap *VRM;
+
+  LiveRegMatrix *LRM;
+
+  LiveIntervals *LIS;
+
+  unsigned MaxNumVGPRs;
+
+  const MCPhysReg *CSRegs;
+
+  NSA_Status CheckNSA(const MachineInstr &MI, bool Fast = false) const;
+
+  bool tryAssignRegisters(SmallVectorImpl<LiveInterval *> &Intervals,
+                          unsigned StartReg) const;
+
+  bool canAssign(unsigned StartReg, unsigned NumRegs) const;
+
+  bool scavengeRegs(SmallVectorImpl<LiveInterval *> &Intervals) const;
+};
+
+} // End anonymous namespace.
+
+INITIALIZE_PASS_BEGIN(GCNNSAReassign, DEBUG_TYPE, "GCN NSA Reassign",
+                      false, false)
+INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
+INITIALIZE_PASS_DEPENDENCY(VirtRegMap)
+INITIALIZE_PASS_DEPENDENCY(LiveRegMatrix)
+INITIALIZE_PASS_END(GCNNSAReassign, DEBUG_TYPE, "GCN NSA Reassign",
+                    false, false)
+
+
+char GCNNSAReassign::ID = 0;
+
+char &llvm::GCNNSAReassignID = GCNNSAReassign::ID;
+
+bool
+GCNNSAReassign::tryAssignRegisters(SmallVectorImpl<LiveInterval *> &Intervals,
+                                   unsigned StartReg) const {
+  unsigned NumRegs = Intervals.size();
+
+  for (unsigned N = 0; N < NumRegs; ++N)
+    if (VRM->hasPhys(Intervals[N]->reg))
+      LRM->unassign(*Intervals[N]);
+
+  for (unsigned N = 0; N < NumRegs; ++N)
+    if (LRM->checkInterference(*Intervals[N], StartReg + N))
+      return false;
+
+  for (unsigned N = 0; N < NumRegs; ++N)
+    LRM->assign(*Intervals[N], StartReg + N);
+
+  return true;
+}
+
+bool GCNNSAReassign::canAssign(unsigned StartReg, unsigned NumRegs) const {
+  for (unsigned N = 0; N < NumRegs; ++N) {
+    unsigned Reg = StartReg + N;
+    if (!MRI->isAllocatable(Reg))
+      return false;
+
+    for (unsigned I = 0; CSRegs[I]; ++I)
+      if (TRI->isSubRegisterEq(Reg, CSRegs[I]) &&
+          !LRM->isPhysRegUsed(CSRegs[I]))
+      return false;
+  }
+
+  return true;
+}
+
+bool
+GCNNSAReassign::scavengeRegs(SmallVectorImpl<LiveInterval *> &Intervals) const {
+  unsigned NumRegs = Intervals.size();
+
+  if (NumRegs > MaxNumVGPRs)
+    return false;
+  unsigned MaxReg = MaxNumVGPRs - NumRegs + AMDGPU::VGPR0;
+
+  for (unsigned Reg = AMDGPU::VGPR0; Reg <= MaxReg; ++Reg) {
+    if (!canAssign(Reg, NumRegs))
+      continue;
+
+    if (tryAssignRegisters(Intervals, Reg))
+      return true;
+  }
+
+  return false;
+}
+
+GCNNSAReassign::NSA_Status
+GCNNSAReassign::CheckNSA(const MachineInstr &MI, bool Fast) const {
+  const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(MI.getOpcode());
+  if (!Info || Info->MIMGEncoding != AMDGPU::MIMGEncGfx10NSA)
+    return NSA_Status::NOT_NSA;
+
+  int VAddr0Idx =
+    AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::vaddr0);
+
+  unsigned VgprBase = 0;
+  bool NSA = false;
+  for (unsigned I = 0; I < Info->VAddrDwords; ++I) {
+    const MachineOperand &Op = MI.getOperand(VAddr0Idx + I);
+    unsigned Reg = Op.getReg();
+    if (TargetRegisterInfo::isPhysicalRegister(Reg) || !VRM->isAssignedReg(Reg))
+      return NSA_Status::FIXED;
+
+    unsigned PhysReg = VRM->getPhys(Reg);
+
+    if (!Fast) {
+      if (!PhysReg)
+        return NSA_Status::FIXED;
+
+      // Bail if address is not a VGPR32. That should be possible to extend the
+      // optimization to work with subregs of a wider register tuples, but the
+      // logic to find free registers will be much more complicated with much
+      // less chances for success. That seems reasonable to assume that in most
+      // cases a tuple is used because a vector variable contains different
+      // parts of an address and it is either already consequitive or cannot
+      // be reassigned if not. If needed it is better to rely on register
+      // coalescer to process such address tuples.
+      if (MRI->getRegClass(Reg) != &AMDGPU::VGPR_32RegClass || Op.getSubReg())
+        return NSA_Status::FIXED;
+
+      const MachineInstr *Def = MRI->getUniqueVRegDef(Reg);
+
+      if (Def && Def->isCopy() && Def->getOperand(1).getReg() == PhysReg)
+        return NSA_Status::FIXED;
+
+      for (auto U : MRI->use_nodbg_operands(Reg)) {
+        if (U.isImplicit())
+          return NSA_Status::FIXED;
+        const MachineInstr *UseInst = U.getParent();
+        if (UseInst->isCopy() && UseInst->getOperand(0).getReg() == PhysReg)
+          return NSA_Status::FIXED;
+      }
+
+      if (!LIS->hasInterval(Reg))
+        return NSA_Status::FIXED;
+    }
+
+    if (I == 0)
+      VgprBase = PhysReg;
+    else if (VgprBase + I != PhysReg)
+      NSA = true;
+  }
+
+  return NSA ? NSA_Status::NON_CONTIGUOUS : NSA_Status::CONTIGUOUS;
+}
+
+bool GCNNSAReassign::runOnMachineFunction(MachineFunction &MF) {
+  ST = &MF.getSubtarget<GCNSubtarget>();
+  if (ST->getGeneration() < GCNSubtarget::GFX10)
+    return false;
+
+  MRI = &MF.getRegInfo();
+  TRI = ST->getRegisterInfo();
+  VRM = &getAnalysis<VirtRegMap>();
+  LRM = &getAnalysis<LiveRegMatrix>();
+  LIS = &getAnalysis<LiveIntervals>();
+
+  const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  MaxNumVGPRs = ST->getMaxNumVGPRs(MF);
+  MaxNumVGPRs = std::min(ST->getMaxNumVGPRs(MFI->getOccupancy()), MaxNumVGPRs);
+  CSRegs = TRI->getCalleeSavedRegs(&MF);
+
+  using Candidate = std::pair<const MachineInstr*, bool>;
+  SmallVector<Candidate, 32> Candidates;
+  for (const MachineBasicBlock &MBB : MF) {
+    for (const MachineInstr &MI : MBB) {
+      switch (CheckNSA(MI)) {
+      default:
+        continue;
+      case NSA_Status::CONTIGUOUS:
+        Candidates.push_back(std::make_pair(&MI, true));
+        break;
+      case NSA_Status::NON_CONTIGUOUS:
+        Candidates.push_back(std::make_pair(&MI, false));
+        ++NumNSAInstructions;
+        break;
+      }
+    }
+  }
+
+  bool Changed = false;
+  for (auto &C : Candidates) {
+    if (C.second)
+      continue;
+
+    const MachineInstr *MI = C.first;
+    if (CheckNSA(*MI, true) == NSA_Status::CONTIGUOUS) {
+      // Already happen to be fixed.
+      C.second = true;
+      ++NumNSAConverted;
+      continue;
+    }
+
+    const AMDGPU::MIMGInfo *Info = AMDGPU::getMIMGInfo(MI->getOpcode());
+    int VAddr0Idx =
+      AMDGPU::getNamedOperandIdx(MI->getOpcode(), AMDGPU::OpName::vaddr0);
+
+    SmallVector<LiveInterval *, 16> Intervals;
+    SmallVector<unsigned, 16> OrigRegs;
+    SlotIndex MinInd, MaxInd;
+    for (unsigned I = 0; I < Info->VAddrDwords; ++I) {
+      const MachineOperand &Op = MI->getOperand(VAddr0Idx + I);
+      unsigned Reg = Op.getReg();
+      LiveInterval *LI = &LIS->getInterval(Reg);
+      if (llvm::find(Intervals, LI) != Intervals.end()) {
+        // Same register used, unable to make sequential
+        Intervals.clear();
+        break;
+      }
+      Intervals.push_back(LI);
+      OrigRegs.push_back(VRM->getPhys(Reg));
+      MinInd = I ? std::min(MinInd, LI->beginIndex()) : LI->beginIndex();
+      MaxInd = I ? std::max(MaxInd, LI->endIndex()) : LI->endIndex();
+    }
+
+    if (Intervals.empty())
+      continue;
+
+    LLVM_DEBUG(dbgs() << "Attempting to reassign NSA: " << *MI
+                      << "\tOriginal allocation:\t";
+               for(auto *LI : Intervals)
+                 dbgs() << " " << llvm::printReg((VRM->getPhys(LI->reg)), TRI);
+               dbgs() << '\n');
+
+    bool Success = scavengeRegs(Intervals);
+    if (!Success) {
+      LLVM_DEBUG(dbgs() << "\tCannot reallocate.\n");
+      if (VRM->hasPhys(Intervals.back()->reg)) // Did not change allocation.
+        continue;
+    } else {
+      // Check we did not make it worse for other instructions.
+      auto I = std::lower_bound(Candidates.begin(), &C, MinInd,
+                                [this](const Candidate &C, SlotIndex I) {
+                                  return LIS->getInstructionIndex(*C.first) < I;
+                                });
+      for (auto E = Candidates.end(); Success && I != E &&
+              LIS->getInstructionIndex(*I->first) < MaxInd; ++I) {
+        if (I->second && CheckNSA(*I->first, true) < NSA_Status::CONTIGUOUS) {
+          Success = false;
+          LLVM_DEBUG(dbgs() << "\tNSA conversion conflict with " << *I->first);
+        }
+      }
+    }
+
+    if (!Success) {
+      for (unsigned I = 0; I < Info->VAddrDwords; ++I)
+        if (VRM->hasPhys(Intervals[I]->reg))
+          LRM->unassign(*Intervals[I]);
+
+      for (unsigned I = 0; I < Info->VAddrDwords; ++I)
+        LRM->assign(*Intervals[I], OrigRegs[I]);
+
+      continue;
+    }
+
+    C.second = true;
+    ++NumNSAConverted;
+    LLVM_DEBUG(dbgs() << "\tNew allocation:\t\t ["
+                 << llvm::printReg((VRM->getPhys(Intervals.front()->reg)), TRI)
+                 << " : "
+                 << llvm::printReg((VRM->getPhys(Intervals.back()->reg)), TRI)
+                 << "]\n");
+    Changed = true;
+  }
+
+  return Changed;
+}

llvm/trunk/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll

-; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,SIVI,PRT %s
-; RUN: llc -march=amdgcn -mcpu=fiji  -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,SIVI,PRT %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,PRT %s
+; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX6789,SI %s
+; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX6789,GFX8910,SIVI,PRT %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX6789,PRT %s
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-enable-prt-strict-null -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,NOPRT %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -show-mc-encoding < %s | FileCheck -check-prefixes=GCN,GFX10 %s
 
 ; GCN-LABEL: {{^}}load_1d:
-; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps <4 x float> @load_1d(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_1d_tfe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX6789: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX10: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm tfe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_1d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_1d_lwe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf unorm lwe{{$}}
+; GFX6789: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf unorm lwe{{$}}
+; GFX10: image_load v[0:4], v{{[0-9]+}}, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_1d_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<4 x float>, i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_2d:
-; GCN: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm ;
 define amdgpu_ps <4 x float> @load_2d(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_2d_tfe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_2d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_3d:
-; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm ;
 define amdgpu_ps <4 x float> @load_3d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %r) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_3d_tfe_lwe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm tfe lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_3d_tfe_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %r) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_cube:
-; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm ;
 define amdgpu_ps <4 x float> @load_cube(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.cube.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_cube_lwe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_cube_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_1darray:
-; GCN: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @load_1darray(<8 x i32> inreg %rsrc, i32 %s, i32 %slice) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.1darray.v4f32.i32(i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_1darray_tfe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm tfe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_1darray_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %slice) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1darray.v4f32i32.i32(i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_2darray:
-; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @load_2darray(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.2darray.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_2darray_lwe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_2darray_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_2dmsaa:
-; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm ;
 define amdgpu_ps <4 x float> @load_2dmsaa(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %fragid) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.2dmsaa.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_2dmsaa_both:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm tfe lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_2dmsaa_both(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %fragid) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_2darraymsaa:
-; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm ;
 define amdgpu_ps <4 x float> @load_2darraymsaa(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_2darraymsaa_tfe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
+; GFX6789: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
+; GFX10: image_load v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_2darraymsaa_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darraymsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_mip_1d:
-; GCN: image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps <4 x float> @load_mip_1d(<8 x i32> inreg %rsrc, i32 %s, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.1d.v4f32.i32(i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_mip_1d_lwe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe{{$}}
+; GFX6789: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe{{$}}
+; GFX10: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm lwe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_mip_1d_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %mip) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.1d.v4f32i32.i32(i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_mip_2d:
-; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load_mip v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm ;
 define amdgpu_ps <4 x float> @load_mip_2d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.2d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_mip_2d_tfe:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v4, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
 ; NOPRT-NOT: v_mov_b32_e32 v3
-; GCN: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX6789: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
+; GFX10: image_load_mip v[0:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe ;
 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
 define amdgpu_ps <4 x float> @load_mip_2d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %mip) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4

 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v3, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
 ; NOPRT-NOT: v_mov_b32_e32 v2
-; GCN: image_load v[0:3], v{{[0-9]+}}, s[0:7] dmask:0x7 unorm tfe{{$}}
+; GFX6789: image_load v[0:3], v{{[0-9]+}}, s[0:7] dmask:0x7 unorm tfe{{$}}
+; GFX10: image_load v[0:3], v{{[0-9]+}}, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm tfe ;
 ; SIVI: buffer_store_dword v3, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v3
 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask3(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 7, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_1d_tfe_V4_dmask2:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v2, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
 ; NOPRT-NOT: v_mov_b32_e32 v1
-; GCN: image_load v[0:2], v{{[0-9]+}}, s[0:7] dmask:0x6 unorm tfe{{$}}
+; GFX6789: image_load v[0:2], v{{[0-9]+}}, s[0:7] dmask:0x6 unorm tfe{{$}}
+; GFX10: image_load v[0:2], v{{[0-9]+}}, s[0:7] dmask:0x6 dim:SQ_RSRC_IMG_1D unorm tfe ;
 ; SIVI: buffer_store_dword v2, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v2
 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask2(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 6, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_1d_tfe_V4_dmask1:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v1, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
-; GCN: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
+; GFX6789: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
+; GFX10: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_1D unorm tfe ;
 ; SIVI: buffer_store_dword v1, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v1
 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask1(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
   %v.err = extractvalue {<4 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <4 x float> %v.vec
 }
 
 ; GCN-LABEL: {{^}}load_1d_tfe_V2_dmask1:
 ; PRT: v_mov_b32_e32 v0, 0
 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
 ; NOPRT: v_mov_b32_e32 v1, 0
 ; NOPRT-NOT: v_mov_b32_e32 v0
-; GCN: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
+; GFX6789: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
+; GFX10: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_1D unorm tfe ;
 ; SIVI: buffer_store_dword v1, off, s[8:11], 0
 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v1
 define amdgpu_ps <2 x float> @load_1d_tfe_V2_dmask1(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
 main_body:
   %v = call {<2 x float>,i32} @llvm.amdgcn.image.load.1d.v2f32i32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
   %v.vec = extractvalue {<2 x float>, i32} %v, 0
   %v.err = extractvalue {<2 x float>, i32} %v, 1
   store i32 %v.err, i32 addrspace(1)* %out, align 4
   ret <2 x float> %v.vec
 }
 
 
 ; GCN-LABEL: {{^}}load_mip_3d:
-; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm ;
 define amdgpu_ps <4 x float> @load_mip_3d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %r, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.3d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %r, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_mip_cube:
-; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm ;
 define amdgpu_ps <4 x float> @load_mip_cube(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.cube.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_mip_1darray:
-; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load_mip v[0:3], v[0:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @load_mip_1darray(<8 x i32> inreg %rsrc, i32 %s, i32 %slice, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.1darray.v4f32.i32(i32 15, i32 %s, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_mip_2darray:
-; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @load_mip_2darray(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.mip.2darray.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}store_1d:
-; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store v[0:3], v4, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_2d:
-; GCN: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm ;
 define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t) {
 main_body:
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_3d:
-; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm ;
 define amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %r) {
 main_body:
   call void @llvm.amdgcn.image.store.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_cube:
-; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm ;
 define amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice) {
 main_body:
   call void @llvm.amdgcn.image.store.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_1darray:
-; GCN: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm ;
 define amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %slice) {
 main_body:
   call void @llvm.amdgcn.image.store.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_2darray:
-; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm ;
 define amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice) {
 main_body:
   call void @llvm.amdgcn.image.store.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_2dmsaa:
-; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm ;
 define amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %fragid) {
 main_body:
   call void @llvm.amdgcn.image.store.2dmsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_2darraymsaa:
-; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm ;
 define amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
 main_body:
   call void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_1d:
-; GCN: image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_2d:
-; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm ;
 define amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_3d:
-; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm ;
 define amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %r, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_cube:
-; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm ;
 define amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_1darray:
-; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm ;
 define amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %slice, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_mip_2darray:
-; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm ;
 define amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %mip) {
 main_body:
   call void @llvm.amdgcn.image.store.mip.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}getresinfo_1d:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps <4 x float> @getresinfo_1d(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_2d:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm ;
 define amdgpu_ps <4 x float> @getresinfo_2d(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_3d:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm ;
 define amdgpu_ps <4 x float> @getresinfo_3d(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.3d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_cube:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm ;
 define amdgpu_ps <4 x float> @getresinfo_cube(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.cube.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_1darray:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @getresinfo_1darray(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1darray.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_2darray:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm ;
 define amdgpu_ps <4 x float> @getresinfo_2darray(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darray.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_2dmsaa:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm ;
 define amdgpu_ps <4 x float> @getresinfo_2dmsaa(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2dmsaa.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}getresinfo_2darraymsaa:
-; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX6789: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
+; GFX10: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm ;
 define amdgpu_ps <4 x float> @getresinfo_2darraymsaa(<8 x i32> inreg %rsrc, i32 %mip) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darraymsaa.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_1d_V1:
-; GCN: image_load v0, v0, s[0:7] dmask:0x8 unorm{{$}}
+; GFX6789: image_load v0, v0, s[0:7] dmask:0x8 unorm{{$}}
+; GFX10: image_load v0, v0, s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps float @load_1d_V1(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call float @llvm.amdgcn.image.load.1d.f32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret float %v
 }
 
 ; GCN-LABEL: {{^}}load_1d_V2:
-; GCN: image_load v[0:1], v0, s[0:7] dmask:0x9 unorm{{$}}
+; GFX6789: image_load v[0:1], v0, s[0:7] dmask:0x9 unorm{{$}}
+; GFX10: image_load v[0:1], v0, s[0:7] dmask:0x9 dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps <2 x float> @load_1d_V2(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i32(i32 9, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret <2 x float> %v
 }
 
 ; GCN-LABEL: {{^}}store_1d_V1:
-; GCN: image_store v0, v1, s[0:7] dmask:0x2 unorm{{$}}
+; GFX6789: image_store v0, v1, s[0:7] dmask:0x2 unorm{{$}}
+; GFX10: image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.f32.i32(float %vdata, i32 2, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_1d_V2:
-; GCN: image_store v[0:1], v2, s[0:7] dmask:0xc unorm{{$}}
+; GFX6789: image_store v[0:1], v2, s[0:7] dmask:0xc unorm{{$}}
+; GFX10: image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm ;
 define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v2f32.i32(<2 x float> %vdata, i32 12, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 ; GCN-LABEL: {{^}}load_1d_glc:
-; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc{{$}}
+; GFX6789: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc{{$}}
+; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 1)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_1d_slc:
-; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm slc{{$}}
+; GFX6789: image_load v[0:3], v0, s[0:7] dmask:0xf unorm slc{{$}}
+; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc ;
 define amdgpu_ps <4 x float> @load_1d_slc(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}load_1d_glc_slc:
-; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc slc{{$}}
+; GFX6789: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc slc{{$}}
+; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc ;
 define amdgpu_ps <4 x float> @load_1d_glc_slc(<8 x i32> inreg %rsrc, i32 %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 3)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}store_1d_glc:
-; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc{{$}}
+; GFX6789: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc{{$}}
+; GFX10: image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc ;
 define amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 1)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_1d_slc:
-; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm slc{{$}}
+; GFX6789: image_store v[0:3], v4, s[0:7] dmask:0xf unorm slc{{$}}
+; GFX10: image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc ;
 define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
   ret void
 }
 
 ; GCN-LABEL: {{^}}store_1d_glc_slc:
-; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc slc{{$}}
+; GFX6789: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc slc{{$}}
+; GFX10: image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc ;
 define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 3)
   ret void
 }
 
 ; GCN-LABEL: {{^}}getresinfo_dmask0:
 ; GCN-NOT: image
 ; GCN: ; return to shader part epilog
 define amdgpu_ps <4 x float> @getresinfo_dmask0(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %mip) #0 {
 main_body:
   %r = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i32(i32 0, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret <4 x float> %r
 }
 
 ; Ideally, the register allocator would avoid the wait here
 ;
 ; GCN-LABEL: {{^}}image_store_wait:
-; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm
+; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf
 ; SI: s_waitcnt expcnt(0)
-; GCN: image_load v[0:3], v4, s[8:15] dmask:0xf unorm
+; GCN: image_load v[0:3], v4, s[8:15] dmask:0xf
 ; GCN: s_waitcnt vmcnt(0)
-; GCN: image_store v[0:3], v4, s[16:23] dmask:0xf unorm
+; GCN: image_store v[0:3], v4, s[16:23] dmask:0xf
 define amdgpu_ps void @image_store_wait(<8 x i32> inreg %arg, <8 x i32> inreg %arg1, <8 x i32> inreg %arg2, <4 x float> %arg3, i32 %arg4) #0 {
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %arg3, i32 15, i32 %arg4, <8 x i32> %arg, i32 0, i32 0)
   %data = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %arg4, <8 x i32> %arg1, i32 0, i32 0)
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %data, i32 15, i32 %arg4, <8 x i32> %arg2, i32 0, i32 0)
   ret void
 }
 
 ; SI won't merge ds memory operations, because of the signed offset bug, so
-; we only have check lines for VI.
-; VI-LABEL: image_load_mmo
-; VI: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
-; VI: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
+; we only have check lines for VI+.
+; GFX8910-LABEL: image_load_mmo
+; GFX8910: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
+; GFX8910: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
 define amdgpu_ps float @image_load_mmo(<8 x i32> inreg %rsrc, float addrspace(3)* %lds, <2 x i32> %c) #0 {
   store float 0.000000e+00, float addrspace(3)* %lds
   %c0 = extractelement <2 x i32> %c, i32 0
   %c1 = extractelement <2 x i32> %c, i32 1
   %tex = call float @llvm.amdgcn.image.load.2d.f32.i32(i32 1, i32 %c0, i32 %c1, <8 x i32> %rsrc, i32 0, i32 0)
   %tmp2 = getelementptr float, float addrspace(3)* %lds, i32 4
   store float 0.000000e+00, float addrspace(3)* %tmp2
   ret float %tex

llvm/trunk/test/CodeGen/AMDGPU/nsa-reassign.ll

+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+
+; GCN-LABEL: {{^}}sample_contig_nsa:
+; GCN-DAG: image_sample_c_l v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}],
+; GCN-DAG: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}],
+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:
+  %zcompare.1 = fadd float %zcompare, 1.0
+  %s1.1 = fadd float %s1, 1.0
+  %t1.1 = fadd float %t1, 1.0
+  %r1.1 = fadd float %r1, 1.0
+  %s2.1 = fadd float %s2, 1.0
+  %t2.1 = fadd float %t2, 1.0
+  %r2.1 = fadd float %r2, 1.0
+  %lod.1 = fadd float %lod, 1.0
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare.1, float %s1.1, float %t1.1, float %r1.1, float %lod.1, <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 %s2.1, float %t2.1, float %r2.1, <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_nsa_10vgprs:
+; GCN-DAG: image_sample_c_l v{{[0-9]+}}, [{{v[0-9]+, v[0-9]+, v[0-9]+, v[0-9]+, v[0-9]+}}],
+; GCN-DAG: image_sample v{{[0-9]+}}, [{{v[0-9]+, v[0-9]+, v[0-9]+}}],
+define amdgpu_ps <2 x float> @sample_contig_nsa_10vgprs(<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) #0 {
+main_body:
+  %zcompare.1 = fadd float %zcompare, 1.0
+  %s1.1 = fadd float %s1, 1.0
+  %t1.1 = fadd float %t1, 1.0
+  %r1.1 = fadd float %r1, 1.0
+  %s2.1 = fadd float %s2, 1.0
+  %t2.1 = fadd float %t2, 1.0
+  %r2.1 = fadd float %r2, 1.0
+  %lod.1 = fadd float %lod, 1.0
+  %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare.1, float %s1.1, float %t1.1, float %r1.1, float %lod.1, <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 %s2.1, float %t2.1, float %r2.1, <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_nsa_conflict:
+; GCN-DAG: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}],
+; GCN-DAG: image_sample v{{[0-9]+}}, [{{v[0-9]+, v[0-9]+, v[0-9]+}}],
+define amdgpu_ps <2 x float> @sample_contig_nsa_conflict(<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:
+  %zcompare.1 = fadd float %zcompare, 1.0
+  %s1.1 = fadd float %s1, 1.0
+  %t1.1 = fadd float %t1, 1.0
+  %r1.1 = fadd float %r1, 1.0
+  %s2.1 = fadd float %s2, 1.0
+  %t2.1 = fadd float %t2, 1.0
+  %r2.1 = fadd float %r2, 1.0
+  %lod.1 = fadd float %lod, 1.0
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2.1, float %t2.1, float %r2.1, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  %v1 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %t2.1, float %s2.1, float %r2.1, <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_nsa_same_addr:
+; GCN-DAG: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}],
+; GCN-DAG: image_sample v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}],
+define amdgpu_ps <2 x float> @sample_contig_nsa_same_addr(<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:
+  %zcompare.1 = fadd float %zcompare, 1.0
+  %s1.1 = fadd float %s1, 1.0
+  %t1.1 = fadd float %t1, 1.0
+  %r1.1 = fadd float %r1, 1.0
+  %s2.1 = fadd float %s2, 1.0
+  %t2.1 = fadd float %t2, 1.0
+  %r2.1 = fadd float %r2, 1.0
+  %lod.1 = fadd float %lod, 1.0
+  %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2.1, float %t2.1, float %r2.1, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 1)
+  %v1 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2.1, float %t2.1, float %r2.1, <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_nsa_same_reg:
+; GCN-DAG: image_sample v{{[0-9]+}}, [{{v[0-9]+, v[0-9]+, v[0-9]+}}],
+define amdgpu_ps float @sample_contig_nsa_same_reg(<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:
+  %zcompare.1 = fadd float %zcompare, 1.0
+  %s1.1 = fadd float %s1, 1.0
+  %t1.1 = fadd float %t1, 1.0
+  %r1.1 = fadd float %r1, 1.0
+  %s2.1 = fadd float %s2, 1.0
+  %t2.1 = fadd float %t2, 1.0
+  %r2.1 = fadd float %r2, 1.0
+  %lod.1 = fadd float %lod, 1.0
+  %v = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %t2.1, float %t2.1, float %r2.1, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
+  ret float %v
+}
+
+declare float @llvm.amdgcn.image.sample.3d.f32.f32(i32, float, float, float, <8 x i32>, <4 x i32>, i1, i32, i32)
+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)
+
+attributes #0 = {"amdgpu-num-vgpr"="10"}