Files
clang-p2996/llvm/lib/ExecutionEngine/Orc/IndirectionUtils.cpp
wanglei 15218a1a48 [ORC] Add lazy jit support for LoongArch64
This patch adds resolver, indirection and trampoline stubs for
loongarch64, allowing lazy compilation to work.

It assumes hard float feature exists.

Depends on D141036

Reviewed By: lhames

Differential Revision: https://reviews.llvm.org/D141102
2023-01-21 17:49:36 +08:00

475 lines
16 KiB
C++

//===---- IndirectionUtils.cpp - Utilities for call indirection in Orc ----===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include "llvm/ExecutionEngine/Orc/IndirectionUtils.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/Triple.h"
#include "llvm/ExecutionEngine/JITLink/x86_64.h"
#include "llvm/ExecutionEngine/Orc/OrcABISupport.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/MC/MCDisassembler/MCDisassembler.h"
#include "llvm/MC/MCInstrAnalysis.h"
#include "llvm/Support/Format.h"
#include "llvm/Transforms/Utils/Cloning.h"
#include <sstream>
#define DEBUG_TYPE "orc"
using namespace llvm;
using namespace llvm::orc;
namespace {
class CompileCallbackMaterializationUnit : public orc::MaterializationUnit {
public:
using CompileFunction = JITCompileCallbackManager::CompileFunction;
CompileCallbackMaterializationUnit(SymbolStringPtr Name,
CompileFunction Compile)
: MaterializationUnit(Interface(
SymbolFlagsMap({{Name, JITSymbolFlags::Exported}}), nullptr)),
Name(std::move(Name)), Compile(std::move(Compile)) {}
StringRef getName() const override { return "<Compile Callbacks>"; }
private:
void materialize(std::unique_ptr<MaterializationResponsibility> R) override {
SymbolMap Result;
Result[Name] = JITEvaluatedSymbol(Compile(), JITSymbolFlags::Exported);
// No dependencies, so these calls cannot fail.
cantFail(R->notifyResolved(Result));
cantFail(R->notifyEmitted());
}
void discard(const JITDylib &JD, const SymbolStringPtr &Name) override {
llvm_unreachable("Discard should never occur on a LMU?");
}
SymbolStringPtr Name;
CompileFunction Compile;
};
} // namespace
namespace llvm {
namespace orc {
TrampolinePool::~TrampolinePool() = default;
void IndirectStubsManager::anchor() {}
Expected<JITTargetAddress>
JITCompileCallbackManager::getCompileCallback(CompileFunction Compile) {
if (auto TrampolineAddr = TP->getTrampoline()) {
auto CallbackName =
ES.intern(std::string("cc") + std::to_string(++NextCallbackId));
std::lock_guard<std::mutex> Lock(CCMgrMutex);
AddrToSymbol[*TrampolineAddr] = CallbackName;
cantFail(
CallbacksJD.define(std::make_unique<CompileCallbackMaterializationUnit>(
std::move(CallbackName), std::move(Compile))));
return *TrampolineAddr;
} else
return TrampolineAddr.takeError();
}
JITTargetAddress JITCompileCallbackManager::executeCompileCallback(
JITTargetAddress TrampolineAddr) {
SymbolStringPtr Name;
{
std::unique_lock<std::mutex> Lock(CCMgrMutex);
auto I = AddrToSymbol.find(TrampolineAddr);
// If this address is not associated with a compile callback then report an
// error to the execution session and return ErrorHandlerAddress to the
// callee.
if (I == AddrToSymbol.end()) {
Lock.unlock();
std::string ErrMsg;
{
raw_string_ostream ErrMsgStream(ErrMsg);
ErrMsgStream << "No compile callback for trampoline at "
<< format("0x%016" PRIx64, TrampolineAddr);
}
ES.reportError(
make_error<StringError>(std::move(ErrMsg), inconvertibleErrorCode()));
return ErrorHandlerAddress;
} else
Name = I->second;
}
if (auto Sym =
ES.lookup(makeJITDylibSearchOrder(
&CallbacksJD, JITDylibLookupFlags::MatchAllSymbols),
Name))
return Sym->getAddress();
else {
llvm::dbgs() << "Didn't find callback.\n";
// If anything goes wrong materializing Sym then report it to the session
// and return the ErrorHandlerAddress;
ES.reportError(Sym.takeError());
return ErrorHandlerAddress;
}
}
Expected<std::unique_ptr<JITCompileCallbackManager>>
createLocalCompileCallbackManager(const Triple &T, ExecutionSession &ES,
JITTargetAddress ErrorHandlerAddress) {
switch (T.getArch()) {
default:
return make_error<StringError>(
std::string("No callback manager available for ") + T.str(),
inconvertibleErrorCode());
case Triple::aarch64:
case Triple::aarch64_32: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcAArch64> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::x86: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcI386> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::loongarch64: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcLoongArch64> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::mips: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcMips32Be> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::mipsel: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcMips32Le> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::mips64:
case Triple::mips64el: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcMips64> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::riscv64: {
typedef orc::LocalJITCompileCallbackManager<orc::OrcRiscv64> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
case Triple::x86_64: {
if (T.getOS() == Triple::OSType::Win32) {
typedef orc::LocalJITCompileCallbackManager<orc::OrcX86_64_Win32> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
} else {
typedef orc::LocalJITCompileCallbackManager<orc::OrcX86_64_SysV> CCMgrT;
return CCMgrT::Create(ES, ErrorHandlerAddress);
}
}
}
}
std::function<std::unique_ptr<IndirectStubsManager>()>
createLocalIndirectStubsManagerBuilder(const Triple &T) {
switch (T.getArch()) {
default:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcGenericABI>>();
};
case Triple::aarch64:
case Triple::aarch64_32:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcAArch64>>();
};
case Triple::x86:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcI386>>();
};
case Triple::loongarch64:
return []() {
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcLoongArch64>>();
};
case Triple::mips:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcMips32Be>>();
};
case Triple::mipsel:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcMips32Le>>();
};
case Triple::mips64:
case Triple::mips64el:
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcMips64>>();
};
case Triple::riscv64:
return []() {
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcRiscv64>>();
};
case Triple::x86_64:
if (T.getOS() == Triple::OSType::Win32) {
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcX86_64_Win32>>();
};
} else {
return [](){
return std::make_unique<
orc::LocalIndirectStubsManager<orc::OrcX86_64_SysV>>();
};
}
}
}
Constant* createIRTypedAddress(FunctionType &FT, JITTargetAddress Addr) {
Constant *AddrIntVal =
ConstantInt::get(Type::getInt64Ty(FT.getContext()), Addr);
Constant *AddrPtrVal =
ConstantExpr::getCast(Instruction::IntToPtr, AddrIntVal,
PointerType::get(&FT, 0));
return AddrPtrVal;
}
GlobalVariable* createImplPointer(PointerType &PT, Module &M,
const Twine &Name, Constant *Initializer) {
auto IP = new GlobalVariable(M, &PT, false, GlobalValue::ExternalLinkage,
Initializer, Name, nullptr,
GlobalValue::NotThreadLocal, 0, true);
IP->setVisibility(GlobalValue::HiddenVisibility);
return IP;
}
void makeStub(Function &F, Value &ImplPointer) {
assert(F.isDeclaration() && "Can't turn a definition into a stub.");
assert(F.getParent() && "Function isn't in a module.");
Module &M = *F.getParent();
BasicBlock *EntryBlock = BasicBlock::Create(M.getContext(), "entry", &F);
IRBuilder<> Builder(EntryBlock);
LoadInst *ImplAddr = Builder.CreateLoad(F.getType(), &ImplPointer);
std::vector<Value*> CallArgs;
for (auto &A : F.args())
CallArgs.push_back(&A);
CallInst *Call = Builder.CreateCall(F.getFunctionType(), ImplAddr, CallArgs);
Call->setTailCall();
Call->setAttributes(F.getAttributes());
if (F.getReturnType()->isVoidTy())
Builder.CreateRetVoid();
else
Builder.CreateRet(Call);
}
std::vector<GlobalValue *> SymbolLinkagePromoter::operator()(Module &M) {
std::vector<GlobalValue *> PromotedGlobals;
for (auto &GV : M.global_values()) {
bool Promoted = true;
// Rename if necessary.
if (!GV.hasName())
GV.setName("__orc_anon." + Twine(NextId++));
else if (GV.getName().startswith("\01L"))
GV.setName("__" + GV.getName().substr(1) + "." + Twine(NextId++));
else if (GV.hasLocalLinkage())
GV.setName("__orc_lcl." + GV.getName() + "." + Twine(NextId++));
else
Promoted = false;
if (GV.hasLocalLinkage()) {
GV.setLinkage(GlobalValue::ExternalLinkage);
GV.setVisibility(GlobalValue::HiddenVisibility);
Promoted = true;
}
GV.setUnnamedAddr(GlobalValue::UnnamedAddr::None);
if (Promoted)
PromotedGlobals.push_back(&GV);
}
return PromotedGlobals;
}
Function* cloneFunctionDecl(Module &Dst, const Function &F,
ValueToValueMapTy *VMap) {
Function *NewF =
Function::Create(cast<FunctionType>(F.getValueType()),
F.getLinkage(), F.getName(), &Dst);
NewF->copyAttributesFrom(&F);
if (VMap) {
(*VMap)[&F] = NewF;
auto NewArgI = NewF->arg_begin();
for (auto ArgI = F.arg_begin(), ArgE = F.arg_end(); ArgI != ArgE;
++ArgI, ++NewArgI)
(*VMap)[&*ArgI] = &*NewArgI;
}
return NewF;
}
void moveFunctionBody(Function &OrigF, ValueToValueMapTy &VMap,
ValueMaterializer *Materializer,
Function *NewF) {
assert(!OrigF.isDeclaration() && "Nothing to move");
if (!NewF)
NewF = cast<Function>(VMap[&OrigF]);
else
assert(VMap[&OrigF] == NewF && "Incorrect function mapping in VMap.");
assert(NewF && "Function mapping missing from VMap.");
assert(NewF->getParent() != OrigF.getParent() &&
"moveFunctionBody should only be used to move bodies between "
"modules.");
SmallVector<ReturnInst *, 8> Returns; // Ignore returns cloned.
CloneFunctionInto(NewF, &OrigF, VMap,
CloneFunctionChangeType::DifferentModule, Returns, "",
nullptr, nullptr, Materializer);
OrigF.deleteBody();
}
GlobalVariable* cloneGlobalVariableDecl(Module &Dst, const GlobalVariable &GV,
ValueToValueMapTy *VMap) {
GlobalVariable *NewGV = new GlobalVariable(
Dst, GV.getValueType(), GV.isConstant(),
GV.getLinkage(), nullptr, GV.getName(), nullptr,
GV.getThreadLocalMode(), GV.getType()->getAddressSpace());
NewGV->copyAttributesFrom(&GV);
if (VMap)
(*VMap)[&GV] = NewGV;
return NewGV;
}
void moveGlobalVariableInitializer(GlobalVariable &OrigGV,
ValueToValueMapTy &VMap,
ValueMaterializer *Materializer,
GlobalVariable *NewGV) {
assert(OrigGV.hasInitializer() && "Nothing to move");
if (!NewGV)
NewGV = cast<GlobalVariable>(VMap[&OrigGV]);
else
assert(VMap[&OrigGV] == NewGV &&
"Incorrect global variable mapping in VMap.");
assert(NewGV->getParent() != OrigGV.getParent() &&
"moveGlobalVariableInitializer should only be used to move "
"initializers between modules");
NewGV->setInitializer(MapValue(OrigGV.getInitializer(), VMap, RF_None,
nullptr, Materializer));
}
GlobalAlias* cloneGlobalAliasDecl(Module &Dst, const GlobalAlias &OrigA,
ValueToValueMapTy &VMap) {
assert(OrigA.getAliasee() && "Original alias doesn't have an aliasee?");
auto *NewA = GlobalAlias::create(OrigA.getValueType(),
OrigA.getType()->getPointerAddressSpace(),
OrigA.getLinkage(), OrigA.getName(), &Dst);
NewA->copyAttributesFrom(&OrigA);
VMap[&OrigA] = NewA;
return NewA;
}
void cloneModuleFlagsMetadata(Module &Dst, const Module &Src,
ValueToValueMapTy &VMap) {
auto *MFs = Src.getModuleFlagsMetadata();
if (!MFs)
return;
for (auto *MF : MFs->operands())
Dst.addModuleFlag(MapMetadata(MF, VMap));
}
Error addFunctionPointerRelocationsToCurrentSymbol(jitlink::Symbol &Sym,
jitlink::LinkGraph &G,
MCDisassembler &Disassembler,
MCInstrAnalysis &MIA) {
// AArch64 appears to already come with the necessary relocations. Among other
// architectures, only x86_64 is currently implemented here.
if (G.getTargetTriple().getArch() != Triple::x86_64)
return Error::success();
raw_null_ostream CommentStream;
auto &STI = Disassembler.getSubtargetInfo();
// Determine the function bounds
auto &B = Sym.getBlock();
assert(!B.isZeroFill() && "expected content block");
auto SymAddress = Sym.getAddress();
auto SymStartInBlock =
(const uint8_t *)B.getContent().data() + Sym.getOffset();
auto SymSize = Sym.getSize() ? Sym.getSize() : B.getSize() - Sym.getOffset();
auto Content = ArrayRef(SymStartInBlock, SymSize);
LLVM_DEBUG(dbgs() << "Adding self-relocations to " << Sym.getName() << "\n");
SmallDenseSet<uintptr_t, 8> ExistingRelocations;
for (auto &E : B.edges()) {
if (E.isRelocation())
ExistingRelocations.insert(E.getOffset());
}
size_t I = 0;
while (I < Content.size()) {
MCInst Instr;
uint64_t InstrSize = 0;
uint64_t InstrStart = SymAddress.getValue() + I;
auto DecodeStatus = Disassembler.getInstruction(
Instr, InstrSize, Content.drop_front(I), InstrStart, CommentStream);
if (DecodeStatus != MCDisassembler::Success) {
LLVM_DEBUG(dbgs() << "Aborting due to disassembly failure at address "
<< InstrStart);
return make_error<StringError>(
formatv("failed to disassemble at address {0:x16}", InstrStart),
inconvertibleErrorCode());
}
// Advance to the next instruction.
I += InstrSize;
// Check for a PC-relative address equal to the symbol itself.
auto PCRelAddr =
MIA.evaluateMemoryOperandAddress(Instr, &STI, InstrStart, InstrSize);
if (!PCRelAddr || *PCRelAddr != SymAddress.getValue())
continue;
auto RelocOffInInstr =
MIA.getMemoryOperandRelocationOffset(Instr, InstrSize);
if (!RelocOffInInstr || InstrSize - *RelocOffInInstr != 4) {
LLVM_DEBUG(dbgs() << "Skipping unknown self-relocation at "
<< InstrStart);
continue;
}
auto RelocOffInBlock = orc::ExecutorAddr(InstrStart) + *RelocOffInInstr -
SymAddress + Sym.getOffset();
if (ExistingRelocations.contains(RelocOffInBlock))
continue;
LLVM_DEBUG(dbgs() << "Adding delta32 self-relocation at " << InstrStart);
B.addEdge(jitlink::x86_64::Delta32, RelocOffInBlock, Sym, /*Addend=*/-4);
}
return Error::success();
}
} // End namespace orc.
} // End namespace llvm.