AMDGPU/SI: Correctly emit program global variables when targeting HSA

[arsenm]

LGTM

[• tstellarAMD]

Fixing handling of linkage and allocation types

[arsenm]

lib/Target/AMDGPU/AMDGPUHSATargetObjectFile.cpp

58–62

The isAgentAllocation vs. isProgramAllocation can be moved under a single isGlobalSegment

[• tstellarAMD]

Address review comment.

[arsenm]

LGTM

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

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::InternalLinkage:
+  case GlobalValue::CommonLinkage:
+   return true;
+  case GlobalValue::ExternalLinkage:
+   return false;
+  default: llvm_unreachable("unknown linkage type");
+  }
+}
+
 void AMDGPUAsmPrinter::EmitGlobalVariable(const GlobalVariable *GV) {
 
   const Module *M = GV->getParent();
   if (Triple(M->getTargetTriple()).getOS() != Triple::AMDHSA ||
-      !AMDGPU::isGroupSegment(GV))
-    return AsmPrinter::EmitGlobalVariable(GV);
+      GV->isDeclaration() || AMDGPU::isReadOnlySegment(GV)) {
+    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());
+  }
+
+  const DataLayout &DL = getDataLayout();
+  OutStreamer->PushSection();
+  OutStreamer->SwitchSection(
+      getObjFileLowering().SectionForGlobal(GV, *Mang, TM));
+  MCSymbol *GVSym = getSymbol(GV);
+  const Constant *C = GV->getInitializer();
+  OutStreamer->EmitLabel(GVSym);
+  EmitGlobalConstant(DL, C);
+  OutStreamer->PopSection();
 }
 
 bool AMDGPUAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
 
   // The starting address of all shader programs must be 256 bytes aligned.
   MF.setAlignment(8);
 
   SetupMachineFunction(MF);

lib/Target/AMDGPU/AMDGPUHSATargetObjectFile.h

 #define LLVM_LIB_TARGET_AMDGPU_AMDGPUHSATARGETOBJECTFILE_H
 
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
 #include "llvm/Target/TargetMachine.h"
 
 namespace llvm {
 
 class AMDGPUHSATargetObjectFile final : public TargetLoweringObjectFileELF {
+private:
+  MCSection *DataGlobalAgentSection;
+  MCSection *DataGlobalProgramSection;
+
+  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/AMDGPUHSATargetObjectFile.cpp

 //===-- AMDGPUHSATargetObjectFile.cpp - AMDGPU Object Files ---------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #include "AMDGPUHSATargetObjectFile.h"
+#include "AMDGPU.h"
 #include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCSectionELF.h"
 #include "llvm/Support/ELF.h"
 
 using namespace llvm;
 
 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);
+
+}
+
+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) && isAgentAllocation(GV))
+    return DataGlobalAgentSection;
+
+  if (AMDGPU::isGlobalSegment(GV) && isProgramAllocation(GV))
+    return DataGlobalProgramSection;
+
   return TargetLoweringObjectFileELF::SelectSectionForGlobal(GV, Kind, Mang, TM);
 }

lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp

private:
   bool ParseDirectiveMajorMinor(uint32_t &Major, uint32_t &Minor);
   bool ParseDirectiveHSACodeObjectVersion();
   bool ParseDirectiveHSACodeObjectISA();
   bool ParseAMDKernelCodeTValue(StringRef ID, amd_kernel_code_t &Header);
   bool ParseDirectiveAMDKernelCodeT();
   bool ParseSectionDirectiveHSAText();
   bool subtargetHasRegister(const MCRegisterInfo &MRI, unsigned RegNo) const;
   bool ParseDirectiveAMDGPUHsaKernel();
+  bool ParseDirectiveAMDGPUHsaModuleGlobal();
+  bool ParseDirectiveAMDGPUHsaProgramGlobal();
+  bool ParseSectionDirectiveHSADataGlobalAgent();
+  bool ParseSectionDirectiveHSADataGlobalProgram();
 
 public:
 public:
   enum AMDGPUMatchResultTy {
     Match_PreferE32 = FIRST_TARGET_MATCH_RESULT_TY
   };
 
   AMDGPUAsmParser(MCSubtargetInfo &STI, MCAsmParser &_Parser,
bool AMDGPUAsmParser::ParseDirectiveAMDGPUHsaKernel() {
   StringRef KernelName = Parser.getTok().getString();
 
   getTargetStreamer().EmitAMDGPUSymbolType(KernelName,
                                            ELF::STT_AMDGPU_HSA_KERNEL);
   Lex();
   return false;
 }
 
+bool AMDGPUAsmParser::ParseDirectiveAMDGPUHsaModuleGlobal() {
+  if (getLexer().isNot(AsmToken::Identifier))
+    return TokError("expected symbol name");
+
+  StringRef GlobalName = Parser.getTok().getIdentifier();
+
+  getTargetStreamer().EmitAMDGPUHsaModuleScopeGlobal(GlobalName);
+  Lex();
+  return false;
+}
+
+bool AMDGPUAsmParser::ParseDirectiveAMDGPUHsaProgramGlobal() {
+  if (getLexer().isNot(AsmToken::Identifier))
+    return TokError("expected symbol name");
+
+  StringRef GlobalName = Parser.getTok().getIdentifier();
+
+  getTargetStreamer().EmitAMDGPUHsaProgramScopeGlobal(GlobalName);
+  Lex();
+  return false;
+}
+
+bool AMDGPUAsmParser::ParseSectionDirectiveHSADataGlobalAgent() {
+  getParser().getStreamer().SwitchSection(
+      AMDGPU::getHSADataGlobalAgentSection(getContext()));
+  return false;
+}
+
+bool AMDGPUAsmParser::ParseSectionDirectiveHSADataGlobalProgram() {
+  getParser().getStreamer().SwitchSection(
+      AMDGPU::getHSADataGlobalProgramSection(getContext()));
+  return false;
+}
+
 bool AMDGPUAsmParser::ParseDirective(AsmToken DirectiveID) {
   StringRef IDVal = DirectiveID.getString();
 
   if (IDVal == ".hsa_code_object_version")
     return ParseDirectiveHSACodeObjectVersion();
 
   if (IDVal == ".hsa_code_object_isa")
     return ParseDirectiveHSACodeObjectISA();
 
   if (IDVal == ".amd_kernel_code_t")
     return ParseDirectiveAMDKernelCodeT();
 
   if (IDVal == ".hsatext" || IDVal == ".text")
     return ParseSectionDirectiveHSAText();
 
   if (IDVal == ".amdgpu_hsa_kernel")
     return ParseDirectiveAMDGPUHsaKernel();
 
+  if (IDVal == ".amdgpu_hsa_module_global")
+    return ParseDirectiveAMDGPUHsaModuleGlobal();
+
+  if (IDVal == ".amdgpu_hsa_program_global")
+    return ParseDirectiveAMDGPUHsaProgramGlobal();
+
+  if (IDVal == ".hsadata_global_agent")
+    return ParseSectionDirectiveHSADataGlobalAgent();
+
+  if (IDVal == ".hsadata_global_program")
+    return ParseSectionDirectiveHSADataGlobalProgram();
+
   return true;
 }
 
 bool AMDGPUAsmParser::subtargetHasRegister(const MCRegisterInfo &MRI,
                                            unsigned RegNo) const {
   if (!isVI())
     return true;
 

lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCAsmInfo.cpp

AMDGPUMCAsmInfo::AMDGPUMCAsmInfo(const Triple &TT) : MCAsmInfoELF() {
   HasDotTypeDotSizeDirective = false;
   HasNoDeadStrip = true;
   WeakRefDirective = ".weakref\t";
   //===--- Dwarf Emission Directives -----------------------------------===//
   SupportsDebugInformation = true;
 }
 
 bool AMDGPUMCAsmInfo::shouldOmitSectionDirective(StringRef SectionName) const {
-  return SectionName == ".hsatext" ||
+  return SectionName == ".hsatext" || SectionName == ".hsadata_global_agent" ||
+         SectionName == ".hsadata_global_program" ||
          MCAsmInfo::shouldOmitSectionDirective(SectionName);
 }

lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.h

public:
   virtual void EmitDirectiveHSACodeObjectISA(uint32_t Major, uint32_t Minor,
                                              uint32_t Stepping,
                                              StringRef VendorName,
                                              StringRef ArchName) = 0;
 
   virtual void EmitAMDKernelCodeT(const amd_kernel_code_t &Header) = 0;
 
   virtual void EmitAMDGPUSymbolType(StringRef SymbolName, unsigned Type) = 0;
+
+  virtual void EmitAMDGPUHsaModuleScopeGlobal(StringRef GlobalName) = 0;
+
+  virtual void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) = 0;
 };
 
 class AMDGPUTargetAsmStreamer : public AMDGPUTargetStreamer {
   formatted_raw_ostream &OS;
 public:
   AMDGPUTargetAsmStreamer(MCStreamer &S, formatted_raw_ostream &OS);
   void EmitDirectiveHSACodeObjectVersion(uint32_t Major,
                                          uint32_t Minor) override;
 
   void EmitDirectiveHSACodeObjectISA(uint32_t Major, uint32_t Minor,
                                      uint32_t Stepping, StringRef VendorName,
                                      StringRef ArchName) override;
 
   void EmitAMDKernelCodeT(const amd_kernel_code_t &Header) override;
 
   void EmitAMDGPUSymbolType(StringRef SymbolName, unsigned Type) override;
+
+  void EmitAMDGPUHsaModuleScopeGlobal(StringRef GlobalName) override;
+
+  void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) override;
 };
 
 class AMDGPUTargetELFStreamer : public AMDGPUTargetStreamer {
 
   enum NoteType {
     NT_AMDGPU_HSA_CODE_OBJECT_VERSION = 1,
     NT_AMDGPU_HSA_HSAIL = 2,
     NT_AMDGPU_HSA_ISA = 3,
public:
 
   void EmitDirectiveHSACodeObjectISA(uint32_t Major, uint32_t Minor,
                                      uint32_t Stepping, StringRef VendorName,
                                      StringRef ArchName) override;
 
   void EmitAMDKernelCodeT(const amd_kernel_code_t &Header) override;
 
   void EmitAMDGPUSymbolType(StringRef SymbolName, unsigned Type) override;
+
+  void EmitAMDGPUHsaModuleScopeGlobal(StringRef GlobalName) override;
+
+  void EmitAMDGPUHsaProgramScopeGlobal(StringRef GlobalName) override;
 };
 
 }
 #endif

lib/Target/AMDGPU/MCTargetDesc/AMDGPUTargetStreamer.cpp

void AMDGPUTargetAsmStreamer::EmitAMDGPUSymbolType(StringRef SymbolName,
   switch (Type) {
     default: llvm_unreachable("Invalid AMDGPU symbol type");
     case ELF::STT_AMDGPU_HSA_KERNEL:
       OS << "\t.amdgpu_hsa_kernel " << SymbolName << '\n' ;
       break;
   }
 }
 
+void AMDGPUTargetAsmStreamer::EmitAMDGPUHsaModuleScopeGlobal(
+    StringRef GlobalName) {
+  OS << "\t.amdgpu_hsa_module_global " << GlobalName << '\n';
+}
+
+void AMDGPUTargetAsmStreamer::EmitAMDGPUHsaProgramScopeGlobal(
+    StringRef GlobalName) {
+  OS << "\t.amdgpu_hsa_program_global " << GlobalName << '\n';
+}
+
 //===----------------------------------------------------------------------===//
 // AMDGPUTargetELFStreamer
 //===----------------------------------------------------------------------===//
 
 AMDGPUTargetELFStreamer::AMDGPUTargetELFStreamer(MCStreamer &S)
     : AMDGPUTargetStreamer(S), Streamer(S) { }
 
 MCELFStreamer &AMDGPUTargetELFStreamer::getStreamer() {

 }
 
 void AMDGPUTargetELFStreamer::EmitAMDGPUSymbolType(StringRef SymbolName,
                                                    unsigned Type) {
   MCSymbolELF *Symbol = cast<MCSymbolELF>(
       getStreamer().getContext().getOrCreateSymbol(SymbolName));
   Symbol->setType(ELF::STT_AMDGPU_HSA_KERNEL);
 }
+
+void AMDGPUTargetELFStreamer::EmitAMDGPUHsaModuleScopeGlobal(
+    StringRef GlobalName) {
+
+  MCSymbolELF *Symbol = cast<MCSymbolELF>(
+      getStreamer().getContext().getOrCreateSymbol(GlobalName));
+  Symbol->setType(ELF::STT_OBJECT);
+  Symbol->setBinding(ELF::STB_LOCAL);
+}
+
+void AMDGPUTargetELFStreamer::EmitAMDGPUHsaProgramScopeGlobal(
+    StringRef GlobalName) {
+
+  MCSymbolELF *Symbol = cast<MCSymbolELF>(
+      getStreamer().getContext().getOrCreateSymbol(GlobalName));
+  Symbol->setType(ELF::STT_OBJECT);
+  Symbol->setBinding(ELF::STB_GLOBAL);
+}

lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.h

struct IsaVersion {
   unsigned Stepping;
 };
 
 IsaVersion getIsaVersion(const FeatureBitset &Features);
 void initDefaultAMDKernelCodeT(amd_kernel_code_t &Header,
                                const FeatureBitset &Features);
 MCSection *getHSATextSection(MCContext &Ctx);
 
+MCSection *getHSADataGlobalAgentSection(MCContext &Ctx);
+
+MCSection *getHSADataGlobalProgramSection(MCContext &Ctx);
+
 bool isGroupSegment(const GlobalValue *GV);
+bool isGlobalSegment(const GlobalValue *GV);
+bool isReadOnlySegment(const GlobalValue *GV);
 
 } // end namespace AMDGPU
 } // end namespace llvm
 
 #endif

lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp

 MCSection *getHSATextSection(MCContext &Ctx) {
   return Ctx.getELFSection(".hsatext", ELF::SHT_PROGBITS,
                            ELF::SHF_ALLOC | ELF::SHF_WRITE |
                            ELF::SHF_EXECINSTR |
                            ELF::SHF_AMDGPU_HSA_AGENT |
                            ELF::SHF_AMDGPU_HSA_CODE);
 }
 
+MCSection *getHSADataGlobalAgentSection(MCContext &Ctx) {
+  return Ctx.getELFSection(".hsadata_global_agent", ELF::SHT_PROGBITS,
+                           ELF::SHF_ALLOC | ELF::SHF_WRITE |
+                           ELF::SHF_AMDGPU_HSA_GLOBAL |
+                           ELF::SHF_AMDGPU_HSA_AGENT);
+}
+
+MCSection *getHSADataGlobalProgramSection(MCContext &Ctx) {
+  return  Ctx.getELFSection(".hsadata_global_program", ELF::SHT_PROGBITS,
+                            ELF::SHF_ALLOC | ELF::SHF_WRITE |
+                            ELF::SHF_AMDGPU_HSA_GLOBAL);
+}
+
 bool isGroupSegment(const GlobalValue *GV) {
   return GV->getType()->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS;
 }
 
+bool isGlobalSegment(const GlobalValue *GV) {
+  return GV->getType()->getAddressSpace() == AMDGPUAS::GLOBAL_ADDRESS;
+}
+
+bool isReadOnlySegment(const GlobalValue *GV) {
+  return GV->getType()->getAddressSpace() == AMDGPUAS::CONSTANT_ADDRESS;
+}
+
 } // End namespace AMDGPU
 } // End namespace llvm

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
+
+@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
+
+@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"
+
+define void @test() {
+  ret void
+}
+
+; ASM: .amdgpu_hsa_module_global internal_global
+; ASM: .hsadata_global_program
+; ASM: internal_global_program:
+; ASM: .long 0
+
+; ASM: .amdgpu_hsa_module_global common_global
+; ASM: .hsadata_global_program
+; ASM: common_global_program:
+; ASM: .long 0
+
+; ASM: .amdgpu_hsa_program_global external_global
+; ASM: .hsadata_global_program
+; ASM: external_global_program:
+; ASM: .long 0
+
+; ASM: .amdgpu_hsa_module_global internal_global
+; ASM: .hsadata_global_agent
+; ASM: internal_global_agent:
+; ASM: .long 0
+
+; ASM: .amdgpu_hsa_module_global common_global
+; ASM: .hsadata_global_agent
+; ASM: common_global_agent:
+; ASM: .long 0
+
+; ASM: .amdgpu_hsa_program_global external_global
+; ASM: .hsadata_global_agent
+; ASM: external_global_agent:
+; 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: Binding: Local
+; ELF: Section: .hsadata_global_agent
+; ELF: }
+
+; ELF: Symbol {
+; ELF: Name: common_global_program
+; ELF: Binding: Local
+; ELF: Section: .hsadata_global_program
+; ELF: }
+
+; ELF: Symbol {
+; ELF: Name: internal_global_agent
+; ELF: Binding: Local
+; ELF: Type: Object
+; ELF: Section: .hsadata_global_agent
+; ELF: }
+
+; ELF: Symbol {
+; ELF: Name: internal_global_program
+; ELF: Binding: Local
+; ELF: Type: Object
+; ELF: Section: .hsadata_global_program
+; ELF: }
+
+; ELF: Symbol {
+; ELF: Name: external_global_agent
+; ELF: Binding: Global
+; ELF: Type: Object
+; ELF: Section: .hsadata_global_agent
+; ELF: }
+
+; ELF: Symbol {
+; ELF: Name: external_global_program
+; ELF: Binding: Global
+; ELF: Type: Object
+; ELF: Section: .hsadata_global_program
+; ELF: }