Skip to content

Commit

Permalink
AMDGPU: Bring elf flags in sync with the spec
Browse files Browse the repository at this point in the history 
- Add MACH flags
- Add XNACK flag
- Add reserved flags
- Minor cleanups in docs

Differential Revision: https://reviews.llvm.org/D43356

llvm-svn: 325399
  • Loading branch information
@kzhuravl
kzhuravl committed Feb 16, 2018
1 parent 08868e4 commit 9122a63
Show file tree
Hide file tree
Showing 21 changed files with 942 additions and 273 deletions.
85 changes: 44 additions & 41 deletions llvm/docs/AMDGPUUsage.rst
View file
Expand Up @@ -100,23 +100,23 @@ names from both the *Processor* and *Alternative Processor* can be used.
**Radeon HD 5000 Series (Evergreen)** [AMD-RADEON-HD-5000]_
-----------------------------------------------------------------------------------
``cedar`` ``r600`` dGPU
``cypress`` ``r600`` dGPU
``juniper`` ``r600`` dGPU
``redwood`` ``r600`` dGPU
``sumo`` ``r600`` dGPU
``juniper`` ``r600`` dGPU
``cypress`` ``r600`` dGPU
**Radeon HD 6000 Series (Northern Islands)** [AMD-RADEON-HD-6000]_
-----------------------------------------------------------------------------------
``barts`` ``r600`` dGPU
``turks`` ``r600`` dGPU
``caicos`` ``r600`` dGPU
``cayman`` ``r600`` dGPU
``turks`` ``r600`` dGPU
**GCN GFX6 (Southern Islands (SI))** [AMD-GCN-GFX6]_
-----------------------------------------------------------------------------------
``gfx600`` - ``tahiti`` ``amdgcn`` dGPU
``gfx601`` - ``pitcairn`` ``amdgcn`` dGPU
- ``verde``
``gfx601`` - ``hainan`` ``amdgcn`` dGPU
- ``oland``
- ``hainan``
- ``pitcairn``
- ``verde``
**GCN GFX7 (Sea Islands (CI))** [AMD-GCN-GFX7]_
-----------------------------------------------------------------------------------
``gfx700`` - ``kaveri`` ``amdgcn`` APU - A6-7000
Expand Down Expand Up @@ -170,8 +170,8 @@ names from both the *Processor* and *Alternative Processor* can be used.
\ ``amdgcn`` APU - xnack - E2-9010
[on] - A6-9210
- A9-9410
``gfx802`` - ``tonga`` ``amdgcn`` dGPU - xnack ROCm - FirePro S7150
- ``iceland`` [off] - FirePro S7100
``gfx802`` - ``iceland`` ``amdgcn`` dGPU - xnack ROCm - FirePro S7150
- ``tonga`` [off] - FirePro S7100
- FirePro W7100
- Radeon R285
- Radeon R9 380
Expand Down Expand Up @@ -514,39 +514,42 @@ The AMDGPU backend uses the following ELF header:
Name Value Description (see
:ref:`amdgpu-processor-table`)
================================= ========== =============================
``EF_AMDGPU_MACH_NONE`` 0 *not specified*
``EF_AMDGPU_MACH_R600_R600`` 1 ``r600``
``EF_AMDGPU_MACH_R600_R630`` 2 ``r630``
``EF_AMDGPU_MACH_R600_RS880`` 3 ``rs880``
``EF_AMDGPU_MACH_R600_RV670`` 4 ``rv670``
``EF_AMDGPU_MACH_R600_RV710`` 5 ``rv710``
``EF_AMDGPU_MACH_R600_RV730`` 6 ``rv730``
``EF_AMDGPU_MACH_R600_RV770`` 7 ``rv770``
``EF_AMDGPU_MACH_R600_CEDAR`` 8 ``cedar``
``EF_AMDGPU_MACH_R600_REDWOOD`` 9 ``redwood``
``EF_AMDGPU_MACH_R600_SUMO`` 10 ``sumo``
``EF_AMDGPU_MACH_R600_JUNIPER`` 11 ``juniper``
``EF_AMDGPU_MACH_R600_CYPRESS`` 12 ``cypress``
``EF_AMDGPU_MACH_R600_BARTS`` 13 ``barts``
``EF_AMDGPU_MACH_R600_TURKS`` 14 ``turks``
``EF_AMDGPU_MACH_R600_CAICOS`` 15 ``caicos``
``EF_AMDGPU_MACH_R600_CAYMAN`` 16 ``cayman``
*reserved* 17-31 Reserved for ``r600``
architecture processors.
``EF_AMDGPU_MACH_AMDGCN_GFX600`` 32 ``gfx600``
``EF_AMDGPU_MACH_AMDGCN_GFX601`` 33 ``gfx601``
``EF_AMDGPU_MACH_AMDGCN_GFX700`` 34 ``gfx700``
``EF_AMDGPU_MACH_AMDGCN_GFX701`` 35 ``gfx701``
``EF_AMDGPU_MACH_AMDGCN_GFX702`` 36 ``gfx702``
``EF_AMDGPU_MACH_AMDGCN_GFX703`` 37 ``gfx703``
``EF_AMDGPU_MACH_AMDGCN_GFX704`` 38 ``gfx704``
*reserved* 39 Reserved.
``EF_AMDGPU_MACH_AMDGCN_GFX801`` 40 ``gfx801``
``EF_AMDGPU_MACH_AMDGCN_GFX802`` 41 ``gfx802``
``EF_AMDGPU_MACH_AMDGCN_GFX803`` 42 ``gfx803``
``EF_AMDGPU_MACH_AMDGCN_GFX810`` 43 ``gfx810``
``EF_AMDGPU_MACH_AMDGCN_GFX900`` 44 ``gfx900``
``EF_AMDGPU_MACH_AMDGCN_GFX902`` 45 ``gfx902``
``EF_AMDGPU_MACH_NONE`` 0x000 *not specified*
``EF_AMDGPU_MACH_R600_R600`` 0x001 ``r600``
``EF_AMDGPU_MACH_R600_R630`` 0x002 ``r630``
``EF_AMDGPU_MACH_R600_RS880`` 0x003 ``rs880``
``EF_AMDGPU_MACH_R600_RV670`` 0x004 ``rv670``
``EF_AMDGPU_MACH_R600_RV710`` 0x005 ``rv710``
``EF_AMDGPU_MACH_R600_RV730`` 0x006 ``rv730``
``EF_AMDGPU_MACH_R600_RV770`` 0x007 ``rv770``
``EF_AMDGPU_MACH_R600_CEDAR`` 0x008 ``cedar``
``EF_AMDGPU_MACH_R600_CYPRESS`` 0x009 ``cypress``
``EF_AMDGPU_MACH_R600_JUNIPER`` 0x00a ``juniper``
``EF_AMDGPU_MACH_R600_REDWOOD`` 0x00b ``redwood``
``EF_AMDGPU_MACH_R600_SUMO`` 0x00c ``sumo``
``EF_AMDGPU_MACH_R600_BARTS`` 0x00d ``barts``
``EF_AMDGPU_MACH_R600_CAICOS`` 0x00e ``caicos``
``EF_AMDGPU_MACH_R600_CAYMAN`` 0x00f ``cayman``
``EF_AMDGPU_MACH_R600_TURKS`` 0x010 ``turks``
*reserved* 0x011 - Reserved for ``r600``
0x01f architecture processors.
``EF_AMDGPU_MACH_AMDGCN_GFX600`` 0x020 ``gfx600``
``EF_AMDGPU_MACH_AMDGCN_GFX601`` 0x021 ``gfx601``
``EF_AMDGPU_MACH_AMDGCN_GFX700`` 0x022 ``gfx700``
``EF_AMDGPU_MACH_AMDGCN_GFX701`` 0x023 ``gfx701``
``EF_AMDGPU_MACH_AMDGCN_GFX702`` 0x024 ``gfx702``
``EF_AMDGPU_MACH_AMDGCN_GFX703`` 0x025 ``gfx703``
``EF_AMDGPU_MACH_AMDGCN_GFX704`` 0x026 ``gfx704``
*reserved* 0x027 Reserved.
``EF_AMDGPU_MACH_AMDGCN_GFX801`` 0x028 ``gfx801``
``EF_AMDGPU_MACH_AMDGCN_GFX802`` 0x029 ``gfx802``
``EF_AMDGPU_MACH_AMDGCN_GFX803`` 0x02a ``gfx803``
``EF_AMDGPU_MACH_AMDGCN_GFX810`` 0x02b ``gfx810``
``EF_AMDGPU_MACH_AMDGCN_GFX900`` 0x02c ``gfx900``
``EF_AMDGPU_MACH_AMDGCN_GFX902`` 0x02d ``gfx902``
*reserved* 0x02e Reserved.
*reserved* 0x02f Reserved.
*reserved* 0x030 Reserved.
================================= ========== =============================

Sections
Expand Down
69 changes: 64 additions & 5 deletions llvm/include/llvm/BinaryFormat/ELF.h
View file
Expand Up @@ -651,11 +651,70 @@ enum {

// AMDGPU specific e_flags.
enum : unsigned {
// AMDGPU machine architectures.
EF_AMDGPU_ARCH_NONE = 0x00000000, // None/unknown.
EF_AMDGPU_ARCH_R600 = 0x00000001, // AMD HD2XXX-HD6XXX GPUs.
EF_AMDGPU_ARCH_GCN = 0x00000002, // AMD GCN GFX6+ GPUs.
EF_AMDGPU_ARCH = 0x0000000f // EF_AMDGPU_ARCH_XXX selection mask.
// Processor selection mask for EF_AMDGPU_MACH_* values.
EF_AMDGPU_MACH = 0x0ff,

// Not specified processor.
EF_AMDGPU_MACH_NONE = 0x000,

// R600-based processors.
EF_AMDGPU_MACH_R600_FIRST = 0x001,
EF_AMDGPU_MACH_R600_LAST = 0x010,
// Radeon HD 2000/3000 Series (R600).
EF_AMDGPU_MACH_R600_R600 = 0x001,
EF_AMDGPU_MACH_R600_R630 = 0x002,
EF_AMDGPU_MACH_R600_RS880 = 0x003,
EF_AMDGPU_MACH_R600_RV670 = 0x004,
// Radeon HD 4000 Series (R700).
EF_AMDGPU_MACH_R600_RV710 = 0x005,
EF_AMDGPU_MACH_R600_RV730 = 0x006,
EF_AMDGPU_MACH_R600_RV770 = 0x007,
// Radeon HD 5000 Series (Evergreen).
EF_AMDGPU_MACH_R600_CEDAR = 0x008,
EF_AMDGPU_MACH_R600_CYPRESS = 0x009,
EF_AMDGPU_MACH_R600_JUNIPER = 0x00a,
EF_AMDGPU_MACH_R600_REDWOOD = 0x00b,
EF_AMDGPU_MACH_R600_SUMO = 0x00c,
// Radeon HD 6000 Series (Northern Islands).
EF_AMDGPU_MACH_R600_BARTS = 0x00d,
EF_AMDGPU_MACH_R600_CAICOS = 0x00e,
EF_AMDGPU_MACH_R600_CAYMAN = 0x00f,
EF_AMDGPU_MACH_R600_TURKS = 0x010,

// Reserved for R600-based processors.
EF_AMDGPU_MACH_R600_RESERVED_FIRST = 0x011,
EF_AMDGPU_MACH_R600_RESERVED_LAST = 0x01f,

// AMDGCN-based processors.
EF_AMDGPU_MACH_AMDGCN_FIRST = 0x020,
EF_AMDGPU_MACH_AMDGCN_LAST = 0x02d,
// AMDGCN GFX6.
EF_AMDGPU_MACH_AMDGCN_GFX600 = 0x020,
EF_AMDGPU_MACH_AMDGCN_GFX601 = 0x021,
// AMDGCN GFX7.
EF_AMDGPU_MACH_AMDGCN_GFX700 = 0x022,
EF_AMDGPU_MACH_AMDGCN_GFX701 = 0x023,
EF_AMDGPU_MACH_AMDGCN_GFX702 = 0x024,
EF_AMDGPU_MACH_AMDGCN_GFX703 = 0x025,
EF_AMDGPU_MACH_AMDGCN_GFX704 = 0x026,
// AMDGCN GFX8.
EF_AMDGPU_MACH_AMDGCN_GFX801 = 0x028,
EF_AMDGPU_MACH_AMDGCN_GFX802 = 0x029,
EF_AMDGPU_MACH_AMDGCN_GFX803 = 0x02a,
EF_AMDGPU_MACH_AMDGCN_GFX810 = 0x02b,
// AMDGCN GFX9.
EF_AMDGPU_MACH_AMDGCN_GFX900 = 0x02c,
EF_AMDGPU_MACH_AMDGCN_GFX902 = 0x02d,

// Reserved for AMDGCN-based processors.
EF_AMDGPU_MACH_AMDGCN_RESERVED0 = 0x027,
EF_AMDGPU_MACH_AMDGCN_RESERVED1 = 0x02e,
EF_AMDGPU_MACH_AMDGCN_RESERVED2 = 0x02f,
EF_AMDGPU_MACH_AMDGCN_RESERVED3 = 0x030,

// Indicates if the xnack target feature is enabled for all code contained in
// the object.
EF_AMDGPU_XNACK = 0x100,
};

// ELF Relocation types for AMDGPU
Expand Down
14 changes: 7 additions & 7 deletions llvm/include/llvm/Object/ELFObjectFile.h
View file
Expand Up @@ -1087,15 +1087,15 @@ template <class ELFT> Triple::ArchType ELFObjectFile<ELFT>::getArch() const {
if (!IsLittleEndian)
return Triple::UnknownArch;

unsigned EFlags = EF.getHeader()->e_flags;
switch (EFlags & ELF::EF_AMDGPU_ARCH) {
case ELF::EF_AMDGPU_ARCH_R600:
unsigned MACH = EF.getHeader()->e_flags & ELF::EF_AMDGPU_MACH;
if (MACH >= ELF::EF_AMDGPU_MACH_R600_FIRST &&
MACH <= ELF::EF_AMDGPU_MACH_R600_LAST)
return Triple::r600;
case ELF::EF_AMDGPU_ARCH_GCN:
if (MACH >= ELF::EF_AMDGPU_MACH_AMDGCN_FIRST &&
MACH <= ELF::EF_AMDGPU_MACH_AMDGCN_LAST)
return Triple::amdgcn;
default:
return Triple::UnknownArch;
}

return Triple::UnknownArch;
}

case ELF::EM_BPF:
Expand Down
33 changes: 31 additions & 2 deletions llvm/lib/ObjectYAML/ELFYAML.cpp
View file
Expand Up @@ -369,8 +369,37 @@ void ScalarBitSetTraits<ELFYAML::ELF_EF>::bitset(IO &IO,
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);
BCaseMask(EF_AMDGPU_MACH_NONE, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_R600, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_R630, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_RS880, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_RV670, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_RV710, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_RV730, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_RV770, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_CEDAR, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_CYPRESS, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_JUNIPER, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_REDWOOD, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_SUMO, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_BARTS, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_CAICOS, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_CAYMAN, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_R600_TURKS, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX600, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX601, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX700, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX701, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX702, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX703, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX704, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX801, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX802, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX803, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX810, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX900, EF_AMDGPU_MACH);
BCaseMask(EF_AMDGPU_MACH_AMDGCN_GFX902, EF_AMDGPU_MACH);
BCase(EF_AMDGPU_XNACK);
break;
case ELF::EM_X86_64:
break;
Expand Down
24 changes: 0 additions & 24 deletions llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUELFStreamer.cpp
View file
Expand Up @@ -15,30 +15,6 @@

using namespace llvm;

AMDGPUELFStreamer::AMDGPUELFStreamer(const Triple &T, MCContext &Context,
std::unique_ptr<MCAsmBackend> MAB,
raw_pwrite_stream &OS,
std::unique_ptr<MCCodeEmitter> Emitter)
: MCELFStreamer(Context, std::move(MAB), OS, std::move(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, std::unique_ptr<MCAsmBackend> MAB,
raw_pwrite_stream &OS, std::unique_ptr<MCCodeEmitter> Emitter,
Expand Down
3 changes: 2 additions & 1 deletion llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUELFStreamer.h
View file
Expand Up @@ -27,7 +27,8 @@ class AMDGPUELFStreamer : public MCELFStreamer {
public:
AMDGPUELFStreamer(const Triple &T, MCContext &Context,
std::unique_ptr<MCAsmBackend> MAB, raw_pwrite_stream &OS,
std::unique_ptr<MCCodeEmitter> Emitter);
std::unique_ptr<MCCodeEmitter> Emitter)
: MCELFStreamer(Context, std::move(MAB), OS, std::move(Emitter)) {}
};

MCELFStreamer *createAMDGPUELFStreamer(const Triple &T, MCContext &Context,
Expand Down
4 changes: 3 additions & 1 deletion llvm/lib/Target/AMDGPU/MCTargetDesc/AMDGPUMCTargetDesc.cpp
View file
Expand Up @@ -76,7 +76,7 @@ static MCTargetStreamer *createAMDGPUAsmTargetStreamer(MCStreamer &S,
static MCTargetStreamer * createAMDGPUObjectTargetStreamer(
MCStreamer &S,
const MCSubtargetInfo &STI) {
return new AMDGPUTargetELFStreamer(S);
return new AMDGPUTargetELFStreamer(S, STI);
}

static MCStreamer *createMCStreamer(const Triple &T, MCContext &Context,
Expand All @@ -103,6 +103,8 @@ extern "C" void LLVMInitializeAMDGPUTargetMC() {
// R600 specific registration
TargetRegistry::RegisterMCCodeEmitter(getTheAMDGPUTarget(),
createR600MCCodeEmitter);
TargetRegistry::RegisterObjectTargetStreamer(
getTheAMDGPUTarget(), createAMDGPUObjectTargetStreamer);

// GCN specific registration
TargetRegistry::RegisterMCCodeEmitter(getTheGCNTarget(),
Expand Down

0 comments on commit 9122a63

Please sign in to comment.