AMDGPU/SI: Add support for AMD code object version 2.

[• tstellarAMD]

Summary

Version 2 is now the default. If you want to emit version 1, use
the amdgcn--amdhsa-amdcov1 triple.

[arsenm]

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

107–109

This looks like it should be a select on the value to pass to the function

lib/Target/AMDGPU/AMDGPUSubtarget.h

328–329 ↗

(On Diff #54250)

Why are the extra checks needed here?

[• tstellarAMD]

Drop support for code object version 1.

[kzhuravl]

LGTM

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

void AMDGPUAsmPrinter::EmitStartOfAsmFile(Module &M) {
   // in the module.
   std::unique_ptr<MCSubtargetInfo> STI(TM.getTarget().createMCSubtargetInfo(
         TM.getTargetTriple().str(), TM.getTargetCPU(),
         TM.getTargetFeatureString()));
 
   AMDGPUTargetStreamer *TS =
       static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer());
 
-  TS->EmitDirectiveHSACodeObjectVersion(1, 0);
+  TS->EmitDirectiveHSACodeObjectVersion(2, 0);
+
   AMDGPU::IsaVersion ISA = AMDGPU::getIsaVersion(STI->getFeatureBits());
   TS->EmitDirectiveHSACodeObjectISA(ISA.Major, ISA.Minor, ISA.Stepping,
                                     "AMD", "AMDGPU");
 }
 
 void AMDGPUAsmPrinter::EmitFunctionBodyStart() {
   const AMDGPUSubtarget &STM = MF->getSubtarget<AMDGPUSubtarget>();
   SIProgramInfo KernelInfo;
   if (STM.isAmdHsaOS()) {
AMDGPUTargetStreamer *TS =
         static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer());
     TS->EmitAMDGPUSymbolType(CurrentFnSym->getName(),
                              ELF::STT_AMDGPU_HSA_KERNEL);
   }
 
   AsmPrinter::EmitFunctionEntryLabel();
 }
 
-static bool isModuleLinkage(const GlobalValue *GV) {
-  switch (GV->getLinkage()) {
-  case GlobalValue::LinkOnceODRLinkage:
-  case GlobalValue::LinkOnceAnyLinkage:
-  case GlobalValue::InternalLinkage:
-  case GlobalValue::CommonLinkage:
-   return true;
-  case GlobalValue::ExternalLinkage:
-   return false;
-  default: llvm_unreachable("unknown linkage type");
-  }
-}
-
 void AMDGPUAsmPrinter::EmitGlobalVariable(const GlobalVariable *GV) {
 
-  if (TM.getTargetTriple().getOS() != Triple::AMDHSA) {
-    AsmPrinter::EmitGlobalVariable(GV);
-    return;
-  }
-
-  if (GV->isDeclaration() || GV->getLinkage() == GlobalValue::PrivateLinkage) {
-    AsmPrinter::EmitGlobalVariable(GV);
-    return;
-  }
-
   // Group segment variables aren't emitted in HSA.
   if (AMDGPU::isGroupSegment(GV))
     return;
 
-  AMDGPUTargetStreamer *TS =
-      static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer());
-  if (isModuleLinkage(GV)) {
-    TS->EmitAMDGPUHsaModuleScopeGlobal(GV->getName());
-  } else {
-    TS->EmitAMDGPUHsaProgramScopeGlobal(GV->getName());
-  }
-
-  MCSymbolELF *GVSym = cast<MCSymbolELF>(getSymbol(GV));
-  const DataLayout &DL = getDataLayout();
-
-  // Emit the size
-  uint64_t Size = DL.getTypeAllocSize(GV->getType()->getElementType());
-  OutStreamer->emitELFSize(GVSym, MCConstantExpr::create(Size, OutContext));
-  OutStreamer->PushSection();
-  OutStreamer->SwitchSection(
-      getObjFileLowering().SectionForGlobal(GV, *Mang, TM));
-  const Constant *C = GV->getInitializer();
-  OutStreamer->EmitLabel(GVSym);
-  EmitGlobalConstant(DL, C);
-  OutStreamer->PopSection();
+  AsmPrinter::EmitGlobalVariable(GV);
 }
 
 bool AMDGPUAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
 
   // The starting address of all shader programs must be 256 bytes aligned.
   MF.setAlignment(8);
 
   SetupMachineFunction(MF);
void AMDGPUAsmPrinter::EmitAmdKernelCodeT(const MachineFunction &MF,
   header.kernarg_segment_byte_size = MFI->ABIArgOffset;
   header.wavefront_sgpr_count = KernelInfo.NumSGPR;
   header.workitem_vgpr_count = KernelInfo.NumVGPR;
   header.workitem_private_segment_byte_size = KernelInfo.ScratchSize;
   header.workgroup_group_segment_byte_size = KernelInfo.LDSSize;
 
   AMDGPUTargetStreamer *TS =
       static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer());
+
+  OutStreamer->SwitchSection(getObjFileLowering().getTextSection());
   TS->EmitAMDKernelCodeT(header);
 }
 
 bool AMDGPUAsmPrinter::PrintAsmOperand(const MachineInstr *MI, unsigned OpNo,
                                        unsigned AsmVariant,
                                        const char *ExtraCode, raw_ostream &O) {
   if (ExtraCode && ExtraCode[0]) {
     if (ExtraCode[1] != 0)

lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

extern "C" void LLVMInitializeAMDGPUTarget() {
   initializeSIAnnotateControlFlowPass(*PR);
   initializeSIInsertNopsPass(*PR);
   initializeSIInsertWaitsPass(*PR);
   initializeSIWholeQuadModePass(*PR);
   initializeSILowerControlFlowPass(*PR);
 }
 
 static std::unique_ptr<TargetLoweringObjectFile> createTLOF(const Triple &TT) {
-  if (TT.getOS() == Triple::AMDHSA)
-    return make_unique<AMDGPUHSATargetObjectFile>();
-
   return make_unique<AMDGPUTargetObjectFile>();
 }
 
 static ScheduleDAGInstrs *createR600MachineScheduler(MachineSchedContext *C) {
   return new ScheduleDAGMILive(C, make_unique<R600SchedStrategy>());
 }
 
 static MachineSchedRegistry

lib/Target/AMDGPU/AMDGPUTargetObjectFile.h

 
 class AMDGPUTargetObjectFile : public TargetLoweringObjectFileELF {
   public:
     MCSection *SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
                                       Mangler &Mang,
                                       const TargetMachine &TM) const override;
 };
 
-class AMDGPUHSATargetObjectFile final : public AMDGPUTargetObjectFile {
-private:
-  MCSection *DataGlobalAgentSection;
-  MCSection *DataGlobalProgramSection;
-  MCSection *RodataReadonlyAgentSection;
-
-  bool isAgentAllocationSection(const char *SectionName) const;
-  bool isAgentAllocation(const GlobalValue *GV) const;
-  bool isProgramAllocation(const GlobalValue *GV) const;
-
-public:
-  void Initialize(MCContext &Ctx, const TargetMachine &TM) override;
-
-  MCSection *SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
-                                    Mangler &Mang,
-                                    const TargetMachine &TM) const override;
-};
-
 } // end namespace llvm
 
 #endif

lib/Target/AMDGPU/AMDGPUTargetObjectFile.cpp

MCSection *AMDGPUTargetObjectFile::SelectSectionForGlobal(const GlobalValue *GV,
                                                           SectionKind Kind,
                                                           Mangler &Mang,
                                                 const TargetMachine &TM) const {
   if (Kind.isReadOnly() && AMDGPU::isReadOnlySegment(GV))
     return TextSection;
 
   return TargetLoweringObjectFileELF::SelectSectionForGlobal(GV, Kind, Mang, TM);
 }
-
-//===----------------------------------------------------------------------===//
-// HSA Object File
-//===----------------------------------------------------------------------===//
-
-
-void AMDGPUHSATargetObjectFile::Initialize(MCContext &Ctx,
-                                           const TargetMachine &TM){
-  TargetLoweringObjectFileELF::Initialize(Ctx, TM);
-  InitializeELF(TM.Options.UseInitArray);
-
-  TextSection = AMDGPU::getHSATextSection(Ctx);
-
-  DataGlobalAgentSection = AMDGPU::getHSADataGlobalAgentSection(Ctx);
-  DataGlobalProgramSection = AMDGPU::getHSADataGlobalProgramSection(Ctx);
-
-  RodataReadonlyAgentSection = AMDGPU::getHSARodataReadonlyAgentSection(Ctx);
-}
-
-bool AMDGPUHSATargetObjectFile::isAgentAllocationSection(
-    const char *SectionName) const {
-  return cast<MCSectionELF>(DataGlobalAgentSection)
-      ->getSectionName()
-      .equals(SectionName);
-}
-
-bool AMDGPUHSATargetObjectFile::isAgentAllocation(const GlobalValue *GV) const {
-  // Read-only segments can only have agent allocation.
-  return AMDGPU::isReadOnlySegment(GV) ||
-         (AMDGPU::isGlobalSegment(GV) && GV->hasSection() &&
-          isAgentAllocationSection(GV->getSection()));
-}
-
-bool AMDGPUHSATargetObjectFile::isProgramAllocation(
-    const GlobalValue *GV) const {
-  // The default for global segments is program allocation.
-  return AMDGPU::isGlobalSegment(GV) && !isAgentAllocation(GV);
-}
-
-MCSection *AMDGPUHSATargetObjectFile::SelectSectionForGlobal(
-                                        const GlobalValue *GV, SectionKind Kind,
-                                        Mangler &Mang,
-                                        const TargetMachine &TM) const {
-  if (Kind.isText() && !GV->hasComdat())
-    return getTextSection();
-
-  if (AMDGPU::isGlobalSegment(GV)) {
-    if (isAgentAllocation(GV))
-      return DataGlobalAgentSection;
-
-    if (isProgramAllocation(GV))
-      return DataGlobalProgramSection;
-  }
-
-  return AMDGPUTargetObjectFile::SelectSectionForGlobal(GV, Kind, Mang, TM);
-}

lib/Target/AMDGPU/MCTargetDesc/AMDGPUELFStreamer.h

 class MCSubtargetInfo;
 
 class AMDGPUELFStreamer : public MCELFStreamer {
 public:
   AMDGPUELFStreamer(MCContext &Context, MCAsmBackend &MAB, raw_pwrite_stream &OS,
                   MCCodeEmitter *Emitter)
       : MCELFStreamer(Context, MAB, OS, Emitter) { }
 
-  virtual void InitSections(bool NoExecStac) override;
 };
 
 MCELFStreamer *createAMDGPUELFStreamer(MCContext &Context, MCAsmBackend &MAB,
                                      raw_pwrite_stream &OS,
                                      MCCodeEmitter *Emitter, bool RelaxAll);
 } // namespace llvm.
 
 #endif

lib/Target/AMDGPU/MCTargetDesc/AMDGPUELFStreamer.cpp

 //===-------- AMDGPUELFStreamer.cpp - ELF Object Output -------------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPUELFStreamer.h"
 #include "Utils/AMDGPUBaseInfo.h"
 
 using namespace llvm;
 
-void AMDGPUELFStreamer::InitSections(bool NoExecStack) {
-  // Start with the .hsatext section by default.
-  SwitchSection(AMDGPU::getHSATextSection(getContext()));
-}
-
 MCELFStreamer *llvm::createAMDGPUELFStreamer(MCContext &Context,
                                            MCAsmBackend &MAB,
                                            raw_pwrite_stream &OS,
                                            MCCodeEmitter *Emitter,
                                            bool RelaxAll) {
   return new AMDGPUELFStreamer(Context, MAB, OS, Emitter);
 }

lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp

AMDGPUTargetELFStreamer::EmitDirectiveHSACodeObjectISA(uint32_t Major,
   OS.PopSection();
 }
 
 void
 AMDGPUTargetELFStreamer::EmitAMDKernelCodeT(const amd_kernel_code_t &Header) {
 
   MCStreamer &OS = getStreamer();
   OS.PushSection();
-  // The MCObjectFileInfo that is available to the assembler is a generic
-  // implementation and not AMDGPUHSATargetObjectFile, so we can't use
-  // MCObjectFileInfo::getTextSection() here for fetching the HSATextSection.
-  OS.SwitchSection(AMDGPU::getHSATextSection(OS.getContext()));
   OS.EmitBytes(StringRef((const char*)&Header, sizeof(Header)));
   OS.PopSection();
 }
 
 void AMDGPUTargetELFStreamer::EmitAMDGPUSymbolType(StringRef SymbolName,
                                                    unsigned Type) {
   MCSymbolELF *Symbol = cast<MCSymbolELF>(
       getStreamer().getContext().getOrCreateSymbol(SymbolName));

test/CodeGen/AMDGPU/global-constant.ll

 ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=NOHSA %s
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=kaveri -verify-machineinstrs < %s | FileCheck -check-prefix=GCN -check-prefix=HSA %s
 
 @readonly = private unnamed_addr addrspace(2) constant [4 x float] [float 0.0, float 1.0, float 2.0, float 3.0]
 @readonly2 = private unnamed_addr addrspace(2) constant [4 x float] [float 4.0, float 5.0, float 6.0, float 7.0]
 
 ; GCN-LABEL: {{^}}main:
 ; GCN: s_getpc_b64 s{{\[}}[[PC0_LO:[0-9]+]]:[[PC0_HI:[0-9]+]]{{\]}}
 ; GCN-NEXT: s_add_u32 s{{[0-9]+}}, s[[PC0_LO]], readonly
 ; GCN: s_addc_u32 s{{[0-9]+}}, s[[PC0_HI]], 0
 ; GCN: s_getpc_b64 s{{\[}}[[PC1_LO:[0-9]+]]:[[PC1_HI:[0-9]+]]{{\]}}
 ; GCN-NEXT: s_add_u32 s{{[0-9]+}}, s[[PC1_LO]], readonly
 ; GCN: s_addc_u32 s{{[0-9]+}}, s[[PC1_HI]], 0
 ; NOHSA: .text
-; HSA: .hsatext
+; HSA: .text
 ; GCN: readonly:
 ; GCN: readonly2:
 define void @main(i32 %index, float addrspace(1)* %out) {
   %ptr = getelementptr [4 x float], [4 x float] addrspace(2) * @readonly, i32 0, i32 %index
   %val = load float, float addrspace(2)* %ptr
   store float %val, float addrspace(1)* %out
   %ptr2 = getelementptr [4 x float], [4 x float] addrspace(2) * @readonly2, i32 0, i32 %index
   %val2 = load float, float addrspace(2)* %ptr2
   store float %val2, float addrspace(1)* %out
   ret void
 }
 

test/CodeGen/AMDGPU/hsa-globals.ll

 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | FileCheck --check-prefix=ASM %s
-; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | llvm-mc -filetype=obj -triple amdgcn--amdhsa -mcpu=kaveri | llvm-readobj -symbols -s | FileCheck %s --check-prefix=ELF
 
 @linkonce_odr_global_program = linkonce_odr addrspace(1) global i32 0
 @linkonce_global_program = linkonce addrspace(1) global i32 0
 @internal_global_program = internal addrspace(1) global i32 0
 @common_global_program = common addrspace(1) global i32 0
 @external_global_program = addrspace(1) global i32 0
 
-@linkonce_odr_global_agent = linkonce_odr addrspace(1) global i32 0, section ".hsadata_global_agent"
-@linkonce_global_agent = linkonce addrspace(1) global i32 0, section ".hsadata_global_agent"
-@internal_global_agent = internal addrspace(1) global i32 0, section ".hsadata_global_agent"
-@common_global_agent = common addrspace(1) global i32 0, section ".hsadata_global_agent"
-@external_global_agent = addrspace(1) global i32 0, section ".hsadata_global_agent"
-
 @internal_readonly = internal unnamed_addr addrspace(2) constant i32 0
 @external_readonly = unnamed_addr addrspace(2) constant i32 0
 
 define void @test() {
   ret void
 }
 
-; ASM: .amdgpu_hsa_module_global linkonce_odr_global
-; ASM: .size linkonce_odr_global_program, 4
-; ASM: .hsadata_global_program
+; ASM: .type linkonce_odr_global_program,@object
+; ASM: .section .bss,#alloc,#write
+; ASM: .weak linkonce_odr_global_program
 ; ASM: linkonce_odr_global_program:
 ; ASM: .long 0
+; ASM: .size linkonce_odr_global_program, 4
 
-; ASM: .amdgpu_hsa_module_global linkonce_global
-; ASM: .size linkonce_global_program, 4
-; ASM: .hsadata_global_program
+; ASM: .type linkonce_global_program,@object
+; ASM: .weak linkonce_global_program
 ; ASM: linkonce_global_program:
 ; ASM: .long 0
+; ASM: .size linkonce_global_program, 4
 
-; ASM: .amdgpu_hsa_module_global internal_global
-; ASM: .size internal_global_program, 4
-; ASM: .hsadata_global_program
-; ASM: internal_global_program:
-; ASM: .long 0
+; ASM: .type internal_global_program,@object
+; ASM: .local internal_global_program
+; ASM: .comm internal_global_program,4,2
 
-; ASM: .amdgpu_hsa_module_global common_global
-; ASM: .size common_global_program, 4
-; ASM: .hsadata_global_program
-; ASM: common_global_program:
-; ASM: .long 0
+; ASM: .type common_global_program,@object
+; ASM: .comm common_global_program,4,2
 
-; ASM: .amdgpu_hsa_program_global external_global
-; ASM: .size external_global_program, 4
-; ASM: .hsadata_global_program
 ; ASM: external_global_program:
 ; ASM: .long 0
+; ASM: .size external_global_program, 4
 
-; ASM: .amdgpu_hsa_module_global linkonce_odr_global
-; ASM: .size linkonce_odr_global_agent, 4
-; ASM: .hsadata_global_agent
-; ASM: linkonce_odr_global_agent:
-; ASM: .long 0
-
-; ASM: .amdgpu_hsa_module_global linkonce_global
-; ASM: .size linkonce_global_agent, 4
-; ASM: .hsadata_global_agent
-; ASM: linkonce_global_agent:
-; ASM: .long 0
-
-; ASM: .amdgpu_hsa_module_global internal_global
-; ASM: .size internal_global_agent, 4
-; ASM: .hsadata_global_agent
-; ASM: internal_global_agent:
-; ASM: .long 0
-
-; ASM: .amdgpu_hsa_module_global common_global
-; ASM: .size common_global_agent, 4
-; ASM: .hsadata_global_agent
-; ASM: common_global_agent:
-; ASM: .long 0
-
-; ASM: .amdgpu_hsa_program_global external_global
-; ASM: .size external_global_agent, 4
-; ASM: .hsadata_global_agent
-; ASM: external_global_agent:
-; ASM: .long 0
-
-; ASM: .amdgpu_hsa_module_global internal_readonly
-; ASM: .size internal_readonly, 4
-; ASM: .hsatext
+; ASM: .type internal_readonly,@object
+; ASM: .text
 ; ASM: internal_readonly:
 ; ASM: .long 0
+; ASM: .size internal_readonly, 4
 
-; ASM: .amdgpu_hsa_program_global external_readonly
-; ASM: .size external_readonly, 4
-; ASM: .hsatext
+; ASM: .type external_readonly,@object
+; ASM: .globl external_readonly
 ; ASM: external_readonly:
 ; ASM: .long 0
-
-; ELF: Section {
-; ELF: Name: .hsadata_global_program
-; ELF: Type: SHT_PROGBITS (0x1)
-; ELF: Flags [ (0x100003)
-; ELF: SHF_ALLOC (0x2)
-; ELF: SHF_AMDGPU_HSA_GLOBAL (0x100000)
-; ELF: SHF_WRITE (0x1)
-; ELF: ]
-; ELF: }
-
-; ELF: Section {
-; ELF: Name: .hsadata_global_agent
-; ELF: Type: SHT_PROGBITS (0x1)
-; ELF: Flags [ (0x900003)
-; ELF: SHF_ALLOC (0x2)
-; ELF: SHF_AMDGPU_HSA_AGENT (0x800000)
-; ELF: SHF_AMDGPU_HSA_GLOBAL (0x100000)
-; ELF: SHF_WRITE (0x1)
-; ELF: ]
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: common_global_agent
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_agent
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: common_global_program
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_program
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: internal_global_agent
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Type: Object
-; ELF: Section: .hsadata_global_agent
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: internal_global_program
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Type: Object
-; ELF: Section: .hsadata_global_program
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: internal_readonly
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Type: Object
-; ELF: Section: .hsatext
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: linkonce_global_agent
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_agent
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: linkonce_global_program
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_program
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: linkonce_odr_global_agent
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_agent
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: linkonce_odr_global_program
-; ELF: Size: 4
-; ELF: Binding: Local
-; ELF: Section: .hsadata_global_program
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: external_global_agent
-; ELF: Size: 4
-; ELF: Binding: Global
-; ELF: Type: Object
-; ELF: Section: .hsadata_global_agent
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: external_global_program
-; ELF: Size: 4
-; ELF: Binding: Global
-; ELF: Type: Object
-; ELF: Section: .hsadata_global_program
-; ELF: }
-
-; ELF: Symbol {
-; ELF: Name: external_readonly
-; ELF: Size: 4
-; ELF: Binding: Global
-; ELF: Type: Object
-; ELF: Section: .hsatext
-; ELF: }
+; ASM: .size external_readonly, 4

test/CodeGen/AMDGPU/hsa-note-no-func.ll

 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | FileCheck --check-prefix=HSA --check-prefix=HSA-CI %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=carrizo | FileCheck --check-prefix=HSA --check-prefix=HSA-VI %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=fiji | FileCheck --check-prefix=HSA --check-prefix=HSA-FIJI %s
 
-; HSA: .hsa_code_object_version 1,0
+; HSA: .hsa_code_object_version 2,0
 ; HSA-CI: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
 ; HSA-VI: .hsa_code_object_isa 8,0,1,"AMD","AMDGPU"
 ; HSA-FIJI: .hsa_code_object_isa 8,0,3,"AMD","AMDGPU"

test/CodeGen/AMDGPU/hsa.ll

 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | FileCheck --check-prefix=HSA %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri -mattr=-flat-for-global | FileCheck --check-prefix=HSA-CI %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=carrizo  | FileCheck --check-prefix=HSA %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=carrizo -mattr=-flat-for-global | FileCheck --check-prefix=HSA-VI %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri -filetype=obj | llvm-readobj -symbols -s -sd | FileCheck --check-prefix=ELF %s
 ; RUN: llc < %s -mtriple=amdgcn--amdhsa -mcpu=kaveri | llvm-mc -filetype=obj -triple amdgcn--amdhsa -mcpu=kaveri | llvm-readobj -symbols -s -sd | FileCheck %s --check-prefix=ELF
 
 ; The SHT_NOTE section contains the output from the .hsa_code_object_*
 ; directives.
 
 ; ELF: Section {
-; ELF: Name: .hsatext
+; ELF: Name: .text
 ; ELF: Type: SHT_PROGBITS (0x1)
-; ELF: Flags [ (0xC00007)
+; ELF: Flags [ (0x6)
 ; ELF: SHF_ALLOC (0x2)
-; ELF: SHF_AMDGPU_HSA_AGENT (0x800000)
-; ELF: SHF_AMDGPU_HSA_CODE (0x400000)
 ; ELF: SHF_EXECINSTR (0x4)
-; ELF: SHF_WRITE (0x1)
 ; ELF: }
 
 ; ELF: SHT_NOTE
 ; ELF: 0000: 04000000 08000000 01000000 414D4400
-; ELF: 0010: 01000000 00000000 04000000 1B000000
+; ELF: 0010: 02000000 00000000 04000000 1B000000
 ; ELF: 0020: 03000000 414D4400 04000700 07000000
 ; ELF: 0030: 00000000 00000000 414D4400 414D4447
 ; ELF: 0040: 50550000
 
 ; ELF: Symbol {
 ; ELF: Name: simple
 ; ELF: Size: 288
 ; ELF: Type: AMDGPU_HSA_KERNEL (0xA)
 ; ELF: }
 
-; HSA: .hsa_code_object_version 1,0
+; HSA: .hsa_code_object_version 2,0
 ; HSA-CI: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
 ; HSA-VI: .hsa_code_object_isa 8,0,1,"AMD","AMDGPU"
 
-; HSA: .hsatext
+; HSA: .text
 
 ; HSA: .amdgpu_hsa_kernel simple
 ; HSA: {{^}}simple:
 ; HSA: .amd_kernel_code_t
 ; HSA: enable_sgpr_private_segment_buffer = 1
 ; HSA: enable_sgpr_kernarg_segment_ptr = 1
 ; HSA: .end_amd_kernel_code_t
 ; HSA: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[4:5], 0x0