P4 to Git Change 1070977 by ssahasra@ssahasra_opencl_windows on 2014/08/28 01:10:11
ECR #333753 - unify online/offline linkers The code for "FixUpModule" from the online linker is now moved to a common file under llvm/lib. This replaces the copy present in llvm/tools/llvm-link, thus unifying the two linkers. Reviewed by Stanislav Mekhanoshin, Yaxun Liu (Sam) Passes smoke, smoke_clang and precheckin. Also passes OpenCL 2.0 conformance tests. Affected files ... ... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/common/linker.cpp#109 edit ... //depot/stg/opencl/drivers/opencl/compiler/llvm/include/llvm/AMDFixupKernelModule.h#1 add ... //depot/stg/opencl/drivers/opencl/compiler/llvm/include/llvm/AMDUtils.h#1 add ... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Linker/AMDFixupKernelModule.cpp#1 add ... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Transforms/Utils/AMDUtils.cpp#1 add ... //depot/stg/opencl/drivers/opencl/compiler/llvm/tools/llvm-link/AMDFixUpModule.cpp#12 delete ... //depot/stg/opencl/drivers/opencl/compiler/llvm/tools/llvm-link/llvm-link.cpp#48 edit
このコミットが含まれているのは:
foreman
@@ -20,8 +20,10 @@
|
||||
#include "llvm/GlobalValue.h"
|
||||
#include "llvm/GlobalVariable.h"
|
||||
|
||||
#include "llvm/AMDFixupKernelModule.h"
|
||||
#include "llvm/AMDResolveLinker.h"
|
||||
#include "llvm/AMDPrelinkOpt.h"
|
||||
#include "llvm/AMDUtils.h"
|
||||
#include "llvm/ADT/Triple.h"
|
||||
#include "llvm/ADT/StringMap.h"
|
||||
#include "llvm/Analysis/AMDLocalArrayUsage.h"
|
||||
@@ -695,144 +697,6 @@ void amdcl::OCLLinker::fixupOldTriple(llvm::Module *module)
|
||||
}
|
||||
}
|
||||
|
||||
static unsigned getSPIRVersion(const llvm::Module *M) {
|
||||
const llvm::NamedMDNode *SPIRVersion
|
||||
= M->getNamedMetadata("opencl.spir.version");
|
||||
|
||||
if (!SPIRVersion) return 0; // not SPIR
|
||||
|
||||
// When multiple llvm modules are linked together to create a single module
|
||||
// Metadata's of llvm modules are added into destination module and
|
||||
// it results in a more than one SPIR MDNode value.
|
||||
// Marking this fix as temporary and it will be tracked in bugzilla id 9775
|
||||
// FIXME: Uncomment the line below
|
||||
// assert(SPIRVersion->getNumOperands() == 1);
|
||||
assert(SPIRVersion->getNumOperands() > 0);
|
||||
if (SPIRVersion->getNumOperands() > 1) {
|
||||
DEBUG_WITH_TYPE("linkTriple",
|
||||
llvm::dbgs() << "[CheckSPIRVersion] "
|
||||
"Too many arguments to SPIR version MDNode\n");
|
||||
}
|
||||
|
||||
const llvm::MDNode *VersionMD = SPIRVersion->getOperand(0);
|
||||
assert(VersionMD->getNumOperands() == 2);
|
||||
|
||||
const llvm::ConstantInt *CMajor
|
||||
= llvm::cast<llvm::ConstantInt>(VersionMD->getOperand(0));
|
||||
assert(CMajor->getType()->getIntegerBitWidth() == 32);
|
||||
unsigned VersionMajor = CMajor->getZExtValue();
|
||||
|
||||
const llvm::ConstantInt *CMinor
|
||||
= llvm::cast<llvm::ConstantInt>(VersionMD->getOperand(1));
|
||||
assert(CMinor->getType()->getIntegerBitWidth() == 32);
|
||||
unsigned VersionMinor = CMinor->getZExtValue();
|
||||
|
||||
return (VersionMajor * 100) + (VersionMinor * 10);
|
||||
}
|
||||
|
||||
//Modify module for targets before linking.
|
||||
//Report error by buildLog.
|
||||
//Return false on error.
|
||||
static bool fixUpModule(llvm::Module *M,
|
||||
llvm::StringRef TargetTriple,
|
||||
llvm::StringRef TargetLayout,
|
||||
bool RunSPIRLoader,
|
||||
bool DemangleBuiltins,
|
||||
bool RunEDGAdapter,
|
||||
bool SetSPIRCallingConv,
|
||||
bool RunX86Adpater) {
|
||||
llvm::PassManager Passes;
|
||||
|
||||
DEBUG_WITH_TYPE("linkTriple", llvm::dbgs() <<
|
||||
"[fixUpModule] module triple: " << M->getTargetTriple() <<
|
||||
" target triple: " << TargetTriple);
|
||||
llvm::Triple triple(M->getTargetTriple());
|
||||
#if OPENCL_MAJOR < 2
|
||||
if (triple.getArch() == llvm::Triple::spir ||
|
||||
triple.getArch() == llvm::Triple::spir64 ||
|
||||
triple.getArch() == llvm::Triple::x86 ||
|
||||
triple.getArch() == llvm::Triple::x86_64 ||
|
||||
M->getTargetTriple().empty())
|
||||
#endif
|
||||
{
|
||||
M->setTargetTriple(TargetTriple);
|
||||
M->setDataLayout(TargetLayout);
|
||||
}
|
||||
#if OPENCL_MAJOR < 2
|
||||
if (M->getTargetTriple() != TargetTriple) {
|
||||
//ToDo: There is bug 9996 in compiler library about converting BIF30 to BIF21
|
||||
//which causes regressions in ocltst if the following check is enabled.
|
||||
//Fix the bugs then enable the following check
|
||||
#if 0
|
||||
llvm::dbgs() << "Internal Error: Inconsistent module and library target\n";
|
||||
return false;
|
||||
#else
|
||||
llvm::dbgs() << "WARNING: Inconsistent module and library target\n";
|
||||
return true;
|
||||
#endif
|
||||
}
|
||||
#endif
|
||||
|
||||
Passes.add(new llvm::DataLayout(M));
|
||||
|
||||
Passes.add(llvm::createAMDLowerAtomicsPass());
|
||||
|
||||
if (getSPIRVersion(M) >= 200) {
|
||||
Passes.add(llvm::createAMDPrintfRuntimeBinding());
|
||||
Passes.add(llvm::createAMDLowerPipeBuiltinsPass());
|
||||
Passes.add(llvm::createAMDLowerEnqueueKernelPass());
|
||||
Passes.add(llvm::createAMDGenerateDevEnqMetadataPass());
|
||||
}
|
||||
|
||||
if (RunEDGAdapter) {
|
||||
assert(!RunSPIRLoader);
|
||||
Passes.add(llvm::createAMDEDGToIA64TranslatorPass(SetSPIRCallingConv));
|
||||
}
|
||||
|
||||
if (RunSPIRLoader) {
|
||||
assert(!RunEDGAdapter);
|
||||
Passes.add(llvm::createSPIRLoader(DemangleBuiltins));
|
||||
}
|
||||
|
||||
if (RunX86Adpater) {
|
||||
// One of them should run before the AMDX86Adapter Pass.
|
||||
assert(RunSPIRLoader || RunEDGAdapter);
|
||||
Passes.add(llvm::createAMDX86AdapterPass());
|
||||
}
|
||||
|
||||
Passes.run(*M);
|
||||
return true;
|
||||
}
|
||||
|
||||
static bool isSPIRTriple(const llvm::Triple &Triple) {
|
||||
return Triple.getArch() == llvm::Triple::spir
|
||||
|| Triple.getArch() == llvm::Triple::spir64;
|
||||
}
|
||||
|
||||
static bool isAMDILTriple(const llvm::Triple &Triple) {
|
||||
return Triple.getArch() == llvm::Triple::amdil
|
||||
|| Triple.getArch() == llvm::Triple::amdil64;
|
||||
}
|
||||
|
||||
static bool isX86Triple(const llvm::Triple &Triple) {
|
||||
return Triple.getArch() == llvm::Triple::x86
|
||||
|| Triple.getArch() == llvm::Triple::x86_64;
|
||||
}
|
||||
|
||||
static bool isHSAILTriple(const llvm::Triple &Triple) {
|
||||
return Triple.getArch() == llvm::Triple::hsail
|
||||
|| Triple.getArch() == llvm::Triple::hsail_64;
|
||||
}
|
||||
|
||||
static void CheckSPIRVersionForTarget(const llvm::Module *M,
|
||||
const llvm::Triple &TargetTriple) {
|
||||
unsigned SPIRVersion = getSPIRVersion(M);
|
||||
if (SPIRVersion >= 200)
|
||||
assert(!isAMDILTriple(TargetTriple));
|
||||
else
|
||||
assert(SPIRVersion == 120);
|
||||
}
|
||||
|
||||
// On 64 bit device, aclBinary target is set to 64 bit by default. When 32 bit
|
||||
// LLVM or SPIR binary is loaded, aclBinary target needs to be modified to
|
||||
// match LLVM or SPIR bitness.
|
||||
@@ -918,7 +782,7 @@ checkAndFixAclBinaryTarget(llvm::Module* module, aclBinary* elf,
|
||||
#endif
|
||||
}
|
||||
|
||||
int
|
||||
int
|
||||
amdcl::OCLLinker::link(llvm::Module* input, std::vector<llvm::Module*> &libs)
|
||||
{
|
||||
bool IsGPUTarget = isGpuTarget(Elf()->target);
|
||||
@@ -1037,96 +901,13 @@ amdcl::OCLLinker::link(llvm::Module* input, std::vector<llvm::Module*> &libs)
|
||||
#endif
|
||||
|
||||
|
||||
// Under various situations, the LLVM dialect used in the kernel
|
||||
// module does not match the dialect used in the builtin library. We
|
||||
// need to fix-up the kernel module to eliminate this mismatch.
|
||||
//
|
||||
// SPIRLoader is required to consume a SPIR kernel:
|
||||
// SPIR 1.2 on all targets.
|
||||
// SPIR 2.0 on x86 and HSAIL only.
|
||||
//
|
||||
// The AMDIL libary is compiled by EDG, and hence it does not use
|
||||
// the SPIR mangling scheme. To allow a SPIR 1.2 kernel to link with
|
||||
// this library, the SPIRLoader must fix the mangling in the kernel.
|
||||
//
|
||||
// EDGAdapter is required to consume a non-SPIR (EDG) kernel on x86
|
||||
// and HSAIL targets. The builtins library for these targets are
|
||||
// built by Clang, but OpenCL 1.2 kernels are compiled by EDG.
|
||||
//
|
||||
// A non-SPIR kernel module is not expected on the HSAIL target in a
|
||||
// normal OpenCL 2.0 build. We should actually flag an error if this
|
||||
// occurs, but we let it through to facilitate custom builds created
|
||||
// to test this combination. In this situation, the EDGAdapter must
|
||||
// additionally set the calling conventions correctly, because the
|
||||
// HSAIL library is in SPIR format.
|
||||
//
|
||||
// RunX86Adpater is required to run only on the CPU path. It is
|
||||
// expected to the solve the link issues between the user kernel
|
||||
// (SPIR/EDG) vs. Clang compiled x86 builtins library.
|
||||
if (!llvm::fixupKernelModule(LLVMBinary(), LibTargetTriple, LibDataLayout))
|
||||
return 1;
|
||||
|
||||
// Enabled for:
|
||||
bool RunSPIRLoader = false; // SPIR -> x86/HSAIL/AMDIL
|
||||
bool DemangleBuiltins = false; // SPIR -> AMDIL
|
||||
bool RunEDGAdapter = false; // EDG -> x86/HSAIL
|
||||
bool SetSPIRCallingConv = false; // EDG -> HSAIL
|
||||
bool RunX86Adapter = false; // SPIR/EDG -> x86
|
||||
bool LowerToPreciseFunctions = false;
|
||||
|
||||
llvm::Triple ModuleTriple(LLVMBinary()->getTargetTriple());
|
||||
llvm::Triple TargetTriple(LibTargetTriple);
|
||||
|
||||
|
||||
if (isSPIRTriple(ModuleTriple)) {
|
||||
CheckSPIRVersionForTarget(LLVMBinary(), TargetTriple);
|
||||
RunSPIRLoader = true;
|
||||
#if OPENCL_MAJOR >= 2 // this will become default
|
||||
DemangleBuiltins |= isAMDILTriple(TargetTriple);
|
||||
#ifdef BUILD_HSA_TARGET // special case for HSA build
|
||||
DemangleBuiltins |= isHSAILTriple(TargetTriple);
|
||||
#endif
|
||||
// Never demangle for x86 target on 200 build.
|
||||
#else // OpenCL 1.2 build (this will go away)
|
||||
DemangleBuiltins = true;
|
||||
#endif
|
||||
} else {
|
||||
#if OPENCL_MAJOR >= 2
|
||||
// Decide if we need to adapt the non-SPIR (EDG) kernel module.
|
||||
//
|
||||
// FIXME: Remove the #ifdef when x86 and HSAIL libraries are
|
||||
// always built by Clang.
|
||||
#ifndef BUILD_HSA_TARGET
|
||||
// Run the adapter for HSAIL, only if this is an ORCA build!
|
||||
//
|
||||
// On an HSA build, the HSAIL library is always built with EDG.
|
||||
// This assumption must match the settings in
|
||||
// "opencl/library/hsa/hsail/build/Makefile.hsail"
|
||||
RunEDGAdapter |= isHSAILTriple(TargetTriple);
|
||||
#endif
|
||||
// HSAIL requires SPIR calling conventions since the library is in
|
||||
// SPIR format. This doesn't matter if the EDGAdapter is not run.
|
||||
SetSPIRCallingConv = isHSAILTriple(TargetTriple);
|
||||
|
||||
// Run the EDG Adapter if OPENCL_MAJOR >= 2 and for x86 target.
|
||||
RunEDGAdapter |= isX86Triple(TargetTriple);
|
||||
#endif // OPENCL_MAJOR >= 2
|
||||
}
|
||||
|
||||
// X86Adapter should run for both EDG generated LLVM IR and SPIR for x86 path.
|
||||
// FIXME: Remove the #ifdef when x86 is always built by Clang on
|
||||
// OpenCL 1.2 builds.
|
||||
#if OPENCL_MAJOR >=2
|
||||
RunX86Adapter = isX86Triple(TargetTriple);
|
||||
// For HSAIL targets, when the option -cl-fp32-correctly-rounded-divide-sqrt
|
||||
// lower divide and sqrt functions to precise HSAIL builtin library functions.
|
||||
LowerToPreciseFunctions = (isHSAILTriple(TargetTriple)
|
||||
&& Options()->oVariables->FP32RoundDivideSqrt);
|
||||
#endif
|
||||
|
||||
if (!fixUpModule(LLVMBinary(), LibTargetTriple, LibDataLayout,
|
||||
RunSPIRLoader, DemangleBuiltins,
|
||||
RunEDGAdapter, SetSPIRCallingConv,
|
||||
RunX86Adapter))
|
||||
return 1;
|
||||
bool LowerToPreciseFunctions = (isHSAILTriple(llvm::Triple(LibTargetTriple)) &&
|
||||
Options()->oVariables->FP32RoundDivideSqrt);
|
||||
|
||||
// Before doing anything else, quickly optimize Module
|
||||
if (Options()->oVariables->OptLevel) {
|
||||
|
||||
新しいイシューから参照
ユーザーをブロックする