AMDGPU: Add and set AMDGPU-specific e_flags

[t-tye]

include/llvm/BinaryFormat/ELF.h

651

// AMDGPU machine architectures.

652–655

Suggest moving to the lower 4 bits.

[t-tye]

include/llvm/Object/ELFObjectFile.h

1067

Suggest including:

if (!IsLittleEndian)
   return Triple::UnknownArch;

[kzhuravl]

Address review feedback and add r600 test.

[t-tye]

LGTM

include/llvm/BinaryFormat/ELF.h

 #include "ELFRelocs/Sparc.def"
 };
 
 // ELF Relocation types for WebAssembly
 enum {
 #include "ELFRelocs/WebAssembly.def"
 };
 
+// AMDGPU Specific e_flags
+enum : unsigned {
+  EF_AMDGPU_ARCH_NONE = 0x00000000, // None/unknown.
+  EF_AMDGPU_ARCH_R600 = 0x10000000, // AMD HD2XXX-HD6XXX GPUs.
+  EF_AMDGPU_ARCH_GCN = 0x20000000,  // AMD GCN GFX6+ GPUs.
+  EF_AMDGPU_ARCH = 0xf0000000       // EF_AMDGPU_ARCH_XXX selection mask.
+};
+
 // ELF Relocation types for AMDGPU
 enum {
 #include "ELFRelocs/AMDGPU.def"
 };
 
 // ELF Relocation types for BPF
 enum {
 #include "ELFRelocs/BPF.def"

include/llvm/Object/ELFObjectFile.h

case ELF::EM_SPARCV9:
     return Triple::sparcv9;
   case ELF::EM_WEBASSEMBLY:
     switch (EF.getHeader()->e_ident[ELF::EI_CLASS]) {
     case ELF::ELFCLASS32: return Triple::wasm32;
     case ELF::ELFCLASS64: return Triple::wasm64;
     default: return Triple::UnknownArch;
     }
 
-  case ELF::EM_AMDGPU:
-    if (EF.getHeader()->e_ident[ELF::EI_CLASS] != ELF::ELFCLASS64)
-      return Triple::UnknownArch;
-    if (!IsLittleEndian)
-      return Triple::UnknownArch;
-
-    // TODO: Determine r600/amdgcn architecture based e_flags.
+  case ELF::EM_AMDGPU: {
+    unsigned EFlags = EF.getHeader()->e_flags;
+    switch (EFlags & ELF::EF_AMDGPU_ARCH) {
+    case ELF::EF_AMDGPU_ARCH_R600:
+      return Triple::r600;
+    case ELF::EF_AMDGPU_ARCH_GCN:
       return Triple::amdgcn;
+    default:
+      return Triple::UnknownArch;
+    }
+  }
 
   case ELF::EM_BPF:
     return IsLittleEndian ? Triple::bpfel : Triple::bpfeb;
 
   default:
     return Triple::UnknownArch;
   }
 }

lib/ObjectYAML/ELFYAML.cpp

 
 void ScalarEnumerationTraits<ELFYAML::ELF_ELFOSABI>::enumeration(
     IO &IO, ELFYAML::ELF_ELFOSABI &Value) {
 #define ECase(X) IO.enumCase(Value, #X, ELF::X)
   ECase(ELFOSABI_NONE);
   ECase(ELFOSABI_HPUX);
   ECase(ELFOSABI_NETBSD);
   ECase(ELFOSABI_GNU);
-  ECase(ELFOSABI_GNU);
   ECase(ELFOSABI_HURD);
   ECase(ELFOSABI_SOLARIS);
   ECase(ELFOSABI_AIX);
   ECase(ELFOSABI_IRIX);
   ECase(ELFOSABI_FREEBSD);
   ECase(ELFOSABI_TRU64);
   ECase(ELFOSABI_MODESTO);
   ECase(ELFOSABI_OPENBSD);
case ELF::EM_RISCV:
     BCase(EF_RISCV_RVC);
     BCaseMask(EF_RISCV_FLOAT_ABI_SOFT, EF_RISCV_FLOAT_ABI);
     BCaseMask(EF_RISCV_FLOAT_ABI_SINGLE, EF_RISCV_FLOAT_ABI);
     BCaseMask(EF_RISCV_FLOAT_ABI_DOUBLE, EF_RISCV_FLOAT_ABI);
     BCaseMask(EF_RISCV_FLOAT_ABI_QUAD, EF_RISCV_FLOAT_ABI);
     BCase(EF_RISCV_RVE);
     break;
   case ELF::EM_AMDGPU:
+    BCaseMask(EF_AMDGPU_ARCH_R600, EF_AMDGPU_ARCH);
+    BCaseMask(EF_AMDGPU_ARCH_GCN, EF_AMDGPU_ARCH);
+    break;
   case ELF::EM_X86_64:
     break;
   default:
     llvm_unreachable("Unsupported architecture");
   }
 #undef BCase
 #undef BCaseMask
 }

lib/Target/AMDGPU/MCTargetDesc/AMDGPUELFStreamer.h

 namespace llvm {
 class MCAsmBackend;
 class MCCodeEmitter;
 class MCContext;
 class MCSubtargetInfo;
 
 class AMDGPUELFStreamer : public MCELFStreamer {
 public:
-  AMDGPUELFStreamer(MCContext &Context, MCAsmBackend &MAB, raw_pwrite_stream &OS,
-                  MCCodeEmitter *Emitter)
-      : MCELFStreamer(Context, MAB, OS, Emitter) { }
-
+  AMDGPUELFStreamer(const Triple &T, MCContext &Context, MCAsmBackend &MAB,
+                    raw_pwrite_stream &OS, MCCodeEmitter *Emitter);
 };
 
-MCELFStreamer *createAMDGPUELFStreamer(MCContext &Context, MCAsmBackend &MAB,
-                                     raw_pwrite_stream &OS,
+MCELFStreamer *createAMDGPUELFStreamer(const Triple &T, 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"
+#include "llvm/BinaryFormat/ELF.h"
 
 using namespace llvm;
 
-MCELFStreamer *llvm::createAMDGPUELFStreamer(MCContext &Context,
+AMDGPUELFStreamer::AMDGPUELFStreamer(const Triple &T, MCContext &Context,
+                                     MCAsmBackend &MAB, raw_pwrite_stream &OS,
+                                     MCCodeEmitter *Emitter)
+    : MCELFStreamer(Context, MAB, OS, Emitter) {
+  unsigned Arch = ELF::EF_AMDGPU_ARCH_NONE;
+  switch (T.getArch()) {
+  case Triple::r600:
+    Arch = ELF::EF_AMDGPU_ARCH_R600;
+    break;
+  case Triple::amdgcn:
+    Arch = ELF::EF_AMDGPU_ARCH_GCN;
+    break;
+  default:
+    break;
+  }
+
+  MCAssembler &MCA = getAssembler();
+  unsigned EFlags = MCA.getELFHeaderEFlags();
+  EFlags &= ~ELF::EF_AMDGPU_ARCH;
+  EFlags |= Arch;
+  MCA.setELFHeaderEFlags(EFlags);
+}
+
+MCELFStreamer *llvm::createAMDGPUELFStreamer(const Triple &T, MCContext &Context,
                                              MCAsmBackend &MAB,
                                              raw_pwrite_stream &OS,
                                              MCCodeEmitter *Emitter,
                                              bool RelaxAll) {
-  return new AMDGPUELFStreamer(Context, MAB, OS, Emitter);
+  return new AMDGPUELFStreamer(T, Context, MAB, OS, Emitter);
 }

lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCTargetDesc.cpp

static MCTargetStreamer * createAMDGPUObjectTargetStreamer(
                                                    MCStreamer &S,
                                                    const MCSubtargetInfo &STI) {
   return new AMDGPUTargetELFStreamer(S);
 }
 
 static MCStreamer *createMCStreamer(const Triple &T, MCContext &Context,
                                     MCAsmBackend &MAB, raw_pwrite_stream &OS,
                                     MCCodeEmitter *Emitter, bool RelaxAll) {
-  if (T.getOS() == Triple::AMDHSA)
-    return createAMDGPUELFStreamer(Context, MAB, OS, Emitter, RelaxAll);
-
-  return createELFStreamer(Context, MAB, OS, Emitter, RelaxAll);
+  return createAMDGPUELFStreamer(T, Context, MAB, OS, Emitter, RelaxAll);
 }
 
 extern "C" void LLVMInitializeAMDGPUTargetMC() {
   for (Target *T : {&getTheAMDGPUTarget(), &getTheGCNTarget()}) {
     RegisterMCAsmInfo<AMDGPUMCAsmInfo> X(*T);
 
     TargetRegistry::RegisterMCInstrInfo(*T, createAMDGPUMCInstrInfo);
     TargetRegistry::RegisterMCRegInfo(*T, createAMDGPUMCRegisterInfo);

test/CodeGen/AMDGPU/elf-header.ll

 ; RUN: llc -mtriple=amdgcn-amd-amdpal -filetype=obj < %s | llvm-readobj -file-headers - | FileCheck --check-prefix=GCN --check-prefix=GCN-OSABI-PAL %s
 ; RUN: llc -mtriple=amdgcn-unknown-amdpal -filetype=obj < %s | llvm-readobj -file-headers - | FileCheck --check-prefix=GCN --check-prefix=GCN-OSABI-PAL %s
 
 ; RUN: llc -mtriple=amdgcn--mesa3d -filetype=obj < %s | llvm-readobj -file-headers - | FileCheck --check-prefix=GCN --check-prefix=GCN-OSABI-MESA3D %s
 ; RUN: llc -mtriple=amdgcn-amd-mesa3d -filetype=obj < %s | llvm-readobj -file-headers - | FileCheck --check-prefix=GCN --check-prefix=GCN-OSABI-MESA3D %s
 ; RUN: llc -mtriple=amdgcn-unknown-mesa3d -filetype=obj < %s | llvm-readobj -file-headers - | FileCheck --check-prefix=GCN --check-prefix=GCN-OSABI-MESA3D %s
 
 ; R600: Format: ELF32-amdgpu
-; R600: Arch: unknown
+; R600: Arch: r600
 ; R600: AddressSize: 32bit
-; GCN: Format: ELF64-amdgpu
-; GCN: Arch: amdgcn
-; GCN: AddressSize: 64bit
+; GCN:  Format: ELF64-amdgpu
+; GCN:  Arch: amdgcn
+; GCN:  AddressSize: 64bit
 
 ; R600-OSABI-NONE:  OS/ABI: SystemV (0x0)
 ; GCN-OSABI-NONE:   OS/ABI: SystemV (0x0)
 ; GCN-OSABI-HSA:    OS/ABI: AMDGPU_HSA (0x40)
 ; GCN-OSABI-PAL:    OS/ABI: AMDGPU_PAL (0x41)
 ; GCN-OSABI-MESA3D: OS/ABI: AMDGPU_MESA3D (0x42)
 
 ; R600: Machine: EM_AMDGPU (0xE0)
-; GCN: Machine: EM_AMDGPU (0xE0)
+; R600: Flags [ (0x10000000)
+; R600:   EF_AMDGPU_ARCH_R600 (0x10000000)
+; R600: ]
+; GCN:  Machine: EM_AMDGPU (0xE0)
+; GCN:  Flags [ (0x20000000)
+; GCN:    EF_AMDGPU_ARCH_GCN (0x20000000)
+; GCN:  ]
 
 define amdgpu_kernel void @elf_header() {
   ret void
-}
+}
 

test/Object/AMDGPU/elf64-amdgcn-amdhsa-definitions.yaml

 # RUN: yaml2obj %s > %t.o
-# RUN: llvm-readobj -s -file-headers %t.o | FileCheck %s
+# RUN: llvm-readobj -s -file-headers %t.o | FileCheck --check-prefix=ELF %s
+# RUN: obj2yaml %t.o | FileCheck --check-prefix=YAML %s
 
-# CHECK: Format: ELF64-amdgpu
-# CHECK: Arch: amdgcn
-# CHECK: ElfHeader {
-# CHECK:   Ident {
-# CHECK:     OS/ABI: AMDGPU_HSA (0x40)
-# CHECK:     ABIVersion: 0
-# CHECK:   }
-# CHECK:   Machine: EM_AMDGPU (0xE0)
-# CHECK: }
+# ELF: Format: ELF64-amdgpu
+# ELF: Arch: amdgcn
+# ELF: ElfHeader {
+# ELF:   Ident {
+# ELF:     OS/ABI: AMDGPU_HSA (0x40)
+# ELF:     ABIVersion: 0
+# ELF:   }
+# ELF:   Machine: EM_AMDGPU (0xE0)
+# ELF:   Flags [ (0x20000000)
+# ELF:     EF_AMDGPU_ARCH_GCN (0x20000000)
+# ELF:   ]
+# ELF: }
+
+# YAML: FileHeader
+# YAML:   Class:   ELFCLASS64
+# YAML:   Data:    ELFDATA2LSB
+# YAML:   OSABI:   ELFOSABI_AMDGPU_HSA
+# YAML:   Type:    ET_REL
+# YAML:   Machine: EM_AMDGPU
+# YAML:   Flags:   [ EF_AMDGPU_ARCH_GCN ]
 
 --- !ELF
 FileHeader:
   Class:   ELFCLASS64
   Data:    ELFDATA2LSB
+  OSABI:   ELFOSABI_AMDGPU_HSA
   Type:    ET_REL
   Machine: EM_AMDGPU
-  OSABI:   ELFOSABI_AMDGPU_HSA
+  Flags:   [ EF_AMDGPU_ARCH_GCN ]
 ...

test/Object/AMDGPU/elf64-amdgcn-amdpal-definitions.yaml

 # RUN: yaml2obj %s > %t.o
-# RUN: llvm-readobj -s -file-headers %t.o | FileCheck %s
+# RUN: llvm-readobj -s -file-headers %t.o | FileCheck --check-prefix=ELF %s
+# RUN: obj2yaml %t.o | FileCheck --check-prefix=YAML %s
 
-# CHECK: Format: ELF64-amdgpu
-# CHECK: Arch: amdgcn
-# CHECK: ElfHeader {
-# CHECK:   Ident {
-# CHECK:     OS/ABI: AMDGPU_PAL (0x41)
-# CHECK:     ABIVersion: 0
-# CHECK:   }
-# CHECK:   Machine: EM_AMDGPU (0xE0)
-# CHECK: }
+# ELF: Format: ELF64-amdgpu
+# ELF: Arch: amdgcn
+# ELF: ElfHeader {
+# ELF:   Ident {
+# ELF:     OS/ABI: AMDGPU_PAL (0x41)
+# ELF:     ABIVersion: 0
+# ELF:   }
+# ELF:   Machine: EM_AMDGPU (0xE0)
+# ELF:   Flags [ (0x20000000)
+# ELF:     EF_AMDGPU_ARCH_GCN (0x20000000)
+# ELF:   ]
+# ELF: }
+
+# YAML: FileHeader
+# YAML:   Class:   ELFCLASS64
+# YAML:   Data:    ELFDATA2LSB
+# YAML:   OSABI:   ELFOSABI_AMDGPU_PAL
+# YAML:   Type:    ET_REL
+# YAML:   Machine: EM_AMDGPU
+# YAML:   Flags:   [ EF_AMDGPU_ARCH_GCN ]
 
 --- !ELF
 FileHeader:
   Class:   ELFCLASS64
   Data:    ELFDATA2LSB
+  OSABI:   ELFOSABI_AMDGPU_PAL
   Type:    ET_REL
   Machine: EM_AMDGPU
-  OSABI:   ELFOSABI_AMDGPU_PAL
+  Flags:   [ EF_AMDGPU_ARCH_GCN ]
 ...

test/Object/AMDGPU/elf64-amdgcn-mesa3d-definitions.yaml

 # RUN: yaml2obj %s > %t.o
-# RUN: llvm-readobj -s -file-headers %t.o | FileCheck %s
+# RUN: llvm-readobj -s -file-headers %t.o | FileCheck --check-prefix=ELF %s
+# RUN: obj2yaml %t.o | FileCheck --check-prefix=YAML %s
 
-# CHECK: Format: ELF64-amdgpu
-# CHECK: Arch: amdgcn
-# CHECK: ElfHeader {
-# CHECK:   Ident {
-# CHECK:     OS/ABI: AMDGPU_MESA3D (0x42)
-# CHECK:     ABIVersion: 0
-# CHECK:   }
-# CHECK:   Machine: EM_AMDGPU (0xE0)
-# CHECK: }
+# ELF: Format: ELF64-amdgpu
+# ELF: Arch: amdgcn
+# ELF: ElfHeader {
+# ELF:   Ident {
+# ELF:     OS/ABI: AMDGPU_MESA3D (0x42)
+# ELF:     ABIVersion: 0
+# ELF:   }
+# ELF:   Machine: EM_AMDGPU (0xE0)
+# ELF:   Flags [ (0x20000000)
+# ELF:     EF_AMDGPU_ARCH_GCN (0x20000000)
+# ELF:   ]
+# ELF: }
+
+# YAML: FileHeader
+# YAML:   Class:   ELFCLASS64
+# YAML:   Data:    ELFDATA2LSB
+# YAML:   OSABI:   ELFOSABI_AMDGPU_MESA3D
+# YAML:   Type:    ET_REL
+# YAML:   Machine: EM_AMDGPU
+# YAML:   Flags:   [ EF_AMDGPU_ARCH_GCN ]
 
 --- !ELF
 FileHeader:
   Class:   ELFCLASS64
   Data:    ELFDATA2LSB
+  OSABI:   ELFOSABI_AMDGPU_MESA3D
   Type:    ET_REL
   Machine: EM_AMDGPU
-  OSABI:   ELFOSABI_AMDGPU_MESA3D
+  Flags:   [ EF_AMDGPU_ARCH_GCN ]
 ...

test/tools/llvm-readobj/amdgpu-elf-definitions.test

 RUN: llvm-readobj -file-headers -program-headers -sections -symbols %p/Inputs/trivial.obj.elf-amdhsa-gfx803 | FileCheck %s
 
 CHECK: Format: ELF64-amdgpu
-CHECK: Arch: amdgcn
+CHECK: Arch: unknown
 CHECK: ElfHeader {
 CHECK:   Ident {
 CHECK:     OS/ABI: AMDGPU_HSA (0x40)
 CHECK:     ABIVersion: 0
 CHECK:   }
 CHECK:   Machine: EM_AMDGPU (0xE0)
 CHECK: }

tools/llvm-readobj/ELFDumper.cpp

static const EnumEntry<unsigned> ElfHeaderMipsFlags[] = {
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_32),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_64),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_32R2),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_64R2),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_32R6),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_MIPS_ARCH_64R6)
 };
 
+static const EnumEntry<unsigned> ElfHeaderAMDGPUFlags[] = {
+  LLVM_READOBJ_ENUM_ENT(ELF, EF_AMDGPU_ARCH_NONE),
+  LLVM_READOBJ_ENUM_ENT(ELF, EF_AMDGPU_ARCH_R600),
+  LLVM_READOBJ_ENUM_ENT(ELF, EF_AMDGPU_ARCH_GCN)
+};
+
 static const EnumEntry<unsigned> ElfHeaderRISCVFlags[] = {
   LLVM_READOBJ_ENUM_ENT(ELF, EF_RISCV_RVC),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_RISCV_FLOAT_ABI_SINGLE),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_RISCV_FLOAT_ABI_DOUBLE),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_RISCV_FLOAT_ABI_QUAD),
   LLVM_READOBJ_ENUM_ENT(ELF, EF_RISCV_RVE)
 };
 
const Elf_Ehdr *e = Obj->getHeader();
     W.printNumber("Version", e->e_version);
     W.printHex("Entry", e->e_entry);
     W.printHex("ProgramHeaderOffset", e->e_phoff);
     W.printHex("SectionHeaderOffset", e->e_shoff);
     if (e->e_machine == EM_MIPS)
       W.printFlags("Flags", e->e_flags, makeArrayRef(ElfHeaderMipsFlags),
                    unsigned(ELF::EF_MIPS_ARCH), unsigned(ELF::EF_MIPS_ABI),
                    unsigned(ELF::EF_MIPS_MACH));
+    else if (e->e_machine == EM_AMDGPU)
+      W.printFlags("Flags", e->e_flags, makeArrayRef(ElfHeaderAMDGPUFlags),
+                   unsigned(ELF::EF_AMDGPU_ARCH));
     else if (e->e_machine == EM_RISCV)
       W.printFlags("Flags", e->e_flags, makeArrayRef(ElfHeaderRISCVFlags));
     else
       W.printFlags("Flags", e->e_flags);
     W.printNumber("HeaderSize", e->e_ehsize);
     W.printNumber("ProgramHeaderEntrySize", e->e_phentsize);
     W.printNumber("ProgramHeaderCount", e->e_phnum);
     W.printNumber("SectionHeaderEntrySize", e->e_shentsize);