1 //===- Reader.cpp - Code to read bytecode files ---------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file was developed by the LLVM research group and is distributed under
6 // the University of Illinois Open Source License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This library implements the functionality defined in llvm/Bytecode/Reader.h
12 // Note that this library should be as fast as possible, reentrant, and
15 // TODO: Allow passing in an option to ignore the symbol table
17 //===----------------------------------------------------------------------===//
20 #include "llvm/Bytecode/BytecodeHandler.h"
21 #include "llvm/BasicBlock.h"
22 #include "llvm/Constants.h"
23 #include "llvm/Instructions.h"
24 #include "llvm/SymbolTable.h"
25 #include "llvm/Bytecode/Format.h"
26 #include "llvm/Support/GetElementPtrTypeIterator.h"
27 #include "Support/StringExtras.h"
34 /// @brief A class for maintaining the slot number definition
35 /// as a placeholder for the actual definition for forward constants defs.
36 class ConstantPlaceHolder : public ConstantExpr {
38 ConstantPlaceHolder(); // DO NOT IMPLEMENT
39 void operator=(const ConstantPlaceHolder &); // DO NOT IMPLEMENT
41 ConstantPlaceHolder(const Type *Ty, unsigned id)
42 : ConstantExpr(Instruction::UserOp1, Constant::getNullValue(Ty), Ty),
44 unsigned getID() { return ID; }
49 // Provide some details on error
50 inline void BytecodeReader::error(std::string err) {
52 err += itostr(RevisionNum) ;
54 err += itostr(At-MemStart);
59 //===----------------------------------------------------------------------===//
60 // Bytecode Reading Methods
61 //===----------------------------------------------------------------------===//
63 /// Determine if the current block being read contains any more data.
64 inline bool BytecodeReader::moreInBlock() {
68 /// Throw an error if we've read past the end of the current block
69 inline void BytecodeReader::checkPastBlockEnd(const char * block_name) {
71 error(std::string("Attempt to read past the end of ") + block_name + " block.");
74 /// Align the buffer position to a 32 bit boundary
75 inline void BytecodeReader::align32() {
77 At = (const unsigned char *)((unsigned long)(At+3) & (~3UL));
79 if (Handler) Handler->handleAlignment(At - Save);
81 error("Ran out of data while aligning!");
84 /// Read a whole unsigned integer
85 inline unsigned BytecodeReader::read_uint() {
87 error("Ran out of data reading uint!");
89 return At[-4] | (At[-3] << 8) | (At[-2] << 16) | (At[-1] << 24);
92 /// Read a variable-bit-rate encoded unsigned integer
93 inline unsigned BytecodeReader::read_vbr_uint() {
100 error("Ran out of data reading vbr_uint!");
101 Result |= (unsigned)((*At++) & 0x7F) << Shift;
103 } while (At[-1] & 0x80);
104 if (Handler) Handler->handleVBR32(At-Save);
108 /// Read a variable-bit-rate encoded unsigned 64-bit integer.
109 inline uint64_t BytecodeReader::read_vbr_uint64() {
116 error("Ran out of data reading vbr_uint64!");
117 Result |= (uint64_t)((*At++) & 0x7F) << Shift;
119 } while (At[-1] & 0x80);
120 if (Handler) Handler->handleVBR64(At-Save);
124 /// Read a variable-bit-rate encoded signed 64-bit integer.
125 inline int64_t BytecodeReader::read_vbr_int64() {
126 uint64_t R = read_vbr_uint64();
129 return -(int64_t)(R >> 1);
130 else // There is no such thing as -0 with integers. "-0" really means
131 // 0x8000000000000000.
134 return (int64_t)(R >> 1);
137 /// Read a pascal-style string (length followed by text)
138 inline std::string BytecodeReader::read_str() {
139 unsigned Size = read_vbr_uint();
140 const unsigned char *OldAt = At;
142 if (At > BlockEnd) // Size invalid?
143 error("Ran out of data reading a string!");
144 return std::string((char*)OldAt, Size);
147 /// Read an arbitrary block of data
148 inline void BytecodeReader::read_data(void *Ptr, void *End) {
149 unsigned char *Start = (unsigned char *)Ptr;
150 unsigned Amount = (unsigned char *)End - Start;
151 if (At+Amount > BlockEnd)
152 error("Ran out of data!");
153 std::copy(At, At+Amount, Start);
157 /// Read a float value in little-endian order
158 inline void BytecodeReader::read_float(float& FloatVal) {
159 if (hasPlatformSpecificFloatingPoint) {
160 read_data(&FloatVal, &FloatVal+1);
162 /// FIXME: This isn't optimal, it has size problems on some platforms
163 /// where FP is not IEEE.
168 FloatUnion.i = At[0] | (At[1] << 8) | (At[2] << 16) | (At[3] << 24);
169 At+=sizeof(uint32_t);
170 FloatVal = FloatUnion.f;
174 /// Read a double value in little-endian order
175 inline void BytecodeReader::read_double(double& DoubleVal) {
176 if (hasPlatformSpecificFloatingPoint) {
177 read_data(&DoubleVal, &DoubleVal+1);
179 /// FIXME: This isn't optimal, it has size problems on some platforms
180 /// where FP is not IEEE.
185 DoubleUnion.i = At[0] | (At[1] << 8) | (At[2] << 16) | (At[3] << 24) |
186 (uint64_t(At[4]) << 32) | (uint64_t(At[5]) << 40) |
187 (uint64_t(At[6]) << 48) | (uint64_t(At[7]) << 56);
188 At+=sizeof(uint64_t);
189 DoubleVal = DoubleUnion.d;
193 /// Read a block header and obtain its type and size
194 inline void BytecodeReader::read_block(unsigned &Type, unsigned &Size) {
195 if ( hasLongBlockHeaders ) {
199 case BytecodeFormat::Reserved_DoNotUse :
200 error("Reserved_DoNotUse used as Module Type?");
201 Type = BytecodeFormat::Module; break;
202 case BytecodeFormat::Module:
203 Type = BytecodeFormat::ModuleBlockID; break;
204 case BytecodeFormat::Function:
205 Type = BytecodeFormat::FunctionBlockID; break;
206 case BytecodeFormat::ConstantPool:
207 Type = BytecodeFormat::ConstantPoolBlockID; break;
208 case BytecodeFormat::SymbolTable:
209 Type = BytecodeFormat::SymbolTableBlockID; break;
210 case BytecodeFormat::ModuleGlobalInfo:
211 Type = BytecodeFormat::ModuleGlobalInfoBlockID; break;
212 case BytecodeFormat::GlobalTypePlane:
213 Type = BytecodeFormat::GlobalTypePlaneBlockID; break;
214 case BytecodeFormat::InstructionList:
215 Type = BytecodeFormat::InstructionListBlockID; break;
216 case BytecodeFormat::CompactionTable:
217 Type = BytecodeFormat::CompactionTableBlockID; break;
218 case BytecodeFormat::BasicBlock:
219 /// This block type isn't used after version 1.1. However, we have to
220 /// still allow the value in case this is an old bc format file.
221 /// We just let its value creep thru.
224 error("Invalid module type found: " + utostr(Type));
229 Type = Size & 0x1F; // mask low order five bits
230 Size >>= 5; // get rid of five low order bits, leaving high 27
233 if (At + Size > BlockEnd)
234 error("Attempt to size a block past end of memory");
235 BlockEnd = At + Size;
236 if (Handler) Handler->handleBlock(Type, BlockStart, Size);
240 /// In LLVM 1.2 and before, Types were derived from Value and so they were
241 /// written as part of the type planes along with any other Value. In LLVM
242 /// 1.3 this changed so that Type does not derive from Value. Consequently,
243 /// the BytecodeReader's containers for Values can't contain Types because
244 /// there's no inheritance relationship. This means that the "Type Type"
245 /// plane is defunct along with the Type::TypeTyID TypeID. In LLVM 1.3
246 /// whenever a bytecode construct must have both types and values together,
247 /// the types are always read/written first and then the Values. Furthermore
248 /// since Type::TypeTyID no longer exists, its value (12) now corresponds to
249 /// Type::LabelTyID. In order to overcome this we must "sanitize" all the
250 /// type TypeIDs we encounter. For LLVM 1.3 bytecode files, there's no change.
251 /// For LLVM 1.2 and before, this function will decrement the type id by
252 /// one to account for the missing Type::TypeTyID enumerator if the value is
253 /// larger than 12 (Type::LabelTyID). If the value is exactly 12, then this
254 /// function returns true, otherwise false. This helps detect situations
255 /// where the pre 1.3 bytecode is indicating that what follows is a type.
256 /// @returns true iff type id corresponds to pre 1.3 "type type"
257 inline bool BytecodeReader::sanitizeTypeId(unsigned &TypeId) {
258 if (hasTypeDerivedFromValue) { /// do nothing if 1.3 or later
259 if (TypeId == Type::LabelTyID) {
260 TypeId = Type::VoidTyID; // sanitize it
261 return true; // indicate we got TypeTyID in pre 1.3 bytecode
262 } else if (TypeId > Type::LabelTyID)
263 --TypeId; // shift all planes down because type type plane is missing
268 /// Reads a vbr uint to read in a type id and does the necessary
269 /// conversion on it by calling sanitizeTypeId.
270 /// @returns true iff \p TypeId read corresponds to a pre 1.3 "type type"
271 /// @see sanitizeTypeId
272 inline bool BytecodeReader::read_typeid(unsigned &TypeId) {
273 TypeId = read_vbr_uint();
274 if ( !has32BitTypes )
275 if ( TypeId == 0x00FFFFFF )
276 TypeId = read_vbr_uint();
277 return sanitizeTypeId(TypeId);
280 //===----------------------------------------------------------------------===//
282 //===----------------------------------------------------------------------===//
284 /// Determine if a type id has an implicit null value
285 inline bool BytecodeReader::hasImplicitNull(unsigned TyID) {
286 if (!hasExplicitPrimitiveZeros)
287 return TyID != Type::LabelTyID && TyID != Type::VoidTyID;
288 return TyID >= Type::FirstDerivedTyID;
291 /// Obtain a type given a typeid and account for things like compaction tables,
292 /// function level vs module level, and the offsetting for the primitive types.
293 const Type *BytecodeReader::getType(unsigned ID) {
294 if (ID < Type::FirstDerivedTyID)
295 if (const Type *T = Type::getPrimitiveType((Type::TypeID)ID))
296 return T; // Asked for a primitive type...
298 // Otherwise, derived types need offset...
299 ID -= Type::FirstDerivedTyID;
301 if (!CompactionTypes.empty()) {
302 if (ID >= CompactionTypes.size())
303 error("Type ID out of range for compaction table!");
304 return CompactionTypes[ID];
307 // Is it a module-level type?
308 if (ID < ModuleTypes.size())
309 return ModuleTypes[ID].get();
311 // Nope, is it a function-level type?
312 ID -= ModuleTypes.size();
313 if (ID < FunctionTypes.size())
314 return FunctionTypes[ID].get();
316 error("Illegal type reference!");
320 /// Get a sanitized type id. This just makes sure that the \p ID
321 /// is both sanitized and not the "type type" of pre-1.3 bytecode.
322 /// @see sanitizeTypeId
323 inline const Type* BytecodeReader::getSanitizedType(unsigned& ID) {
324 if (sanitizeTypeId(ID))
325 error("Invalid type id encountered");
329 /// This method just saves some coding. It uses read_typeid to read
330 /// in a sanitized type id, errors that its not the type type, and
331 /// then calls getType to return the type value.
332 inline const Type* BytecodeReader::readSanitizedType() {
335 error("Invalid type id encountered");
339 /// Get the slot number associated with a type accounting for primitive
340 /// types, compaction tables, and function level vs module level.
341 unsigned BytecodeReader::getTypeSlot(const Type *Ty) {
342 if (Ty->isPrimitiveType())
343 return Ty->getTypeID();
345 // Scan the compaction table for the type if needed.
346 if (!CompactionTypes.empty()) {
347 std::vector<const Type*>::const_iterator I =
348 find(CompactionTypes.begin(), CompactionTypes.end(), Ty);
350 if (I == CompactionTypes.end())
351 error("Couldn't find type specified in compaction table!");
352 return Type::FirstDerivedTyID + (&*I - &CompactionTypes[0]);
355 // Check the function level types first...
356 TypeListTy::iterator I = find(FunctionTypes.begin(), FunctionTypes.end(), Ty);
358 if (I != FunctionTypes.end())
359 return Type::FirstDerivedTyID + ModuleTypes.size() +
360 (&*I - &FunctionTypes[0]);
362 // Check the module level types now...
363 I = find(ModuleTypes.begin(), ModuleTypes.end(), Ty);
364 if (I == ModuleTypes.end())
365 error("Didn't find type in ModuleTypes.");
366 return Type::FirstDerivedTyID + (&*I - &ModuleTypes[0]);
369 /// This is just like getType, but when a compaction table is in use, it is
370 /// ignored. It also ignores function level types.
372 const Type *BytecodeReader::getGlobalTableType(unsigned Slot) {
373 if (Slot < Type::FirstDerivedTyID) {
374 const Type *Ty = Type::getPrimitiveType((Type::TypeID)Slot);
376 error("Not a primitive type ID?");
379 Slot -= Type::FirstDerivedTyID;
380 if (Slot >= ModuleTypes.size())
381 error("Illegal compaction table type reference!");
382 return ModuleTypes[Slot];
385 /// This is just like getTypeSlot, but when a compaction table is in use, it
386 /// is ignored. It also ignores function level types.
387 unsigned BytecodeReader::getGlobalTableTypeSlot(const Type *Ty) {
388 if (Ty->isPrimitiveType())
389 return Ty->getTypeID();
390 TypeListTy::iterator I = find(ModuleTypes.begin(),
391 ModuleTypes.end(), Ty);
392 if (I == ModuleTypes.end())
393 error("Didn't find type in ModuleTypes.");
394 return Type::FirstDerivedTyID + (&*I - &ModuleTypes[0]);
397 /// Retrieve a value of a given type and slot number, possibly creating
398 /// it if it doesn't already exist.
399 Value * BytecodeReader::getValue(unsigned type, unsigned oNum, bool Create) {
400 assert(type != Type::LabelTyID && "getValue() cannot get blocks!");
403 // If there is a compaction table active, it defines the low-level numbers.
404 // If not, the module values define the low-level numbers.
405 if (CompactionValues.size() > type && !CompactionValues[type].empty()) {
406 if (Num < CompactionValues[type].size())
407 return CompactionValues[type][Num];
408 Num -= CompactionValues[type].size();
410 // By default, the global type id is the type id passed in
411 unsigned GlobalTyID = type;
413 // If the type plane was compactified, figure out the global type ID
414 // by adding the derived type ids and the distance.
415 if (!CompactionTypes.empty() && type >= Type::FirstDerivedTyID) {
416 const Type *Ty = CompactionTypes[type-Type::FirstDerivedTyID];
417 TypeListTy::iterator I =
418 find(ModuleTypes.begin(), ModuleTypes.end(), Ty);
419 assert(I != ModuleTypes.end());
420 GlobalTyID = Type::FirstDerivedTyID + (&*I - &ModuleTypes[0]);
423 if (hasImplicitNull(GlobalTyID)) {
425 return Constant::getNullValue(getType(type));
429 if (GlobalTyID < ModuleValues.size() && ModuleValues[GlobalTyID]) {
430 if (Num < ModuleValues[GlobalTyID]->size())
431 return ModuleValues[GlobalTyID]->getOperand(Num);
432 Num -= ModuleValues[GlobalTyID]->size();
436 if (FunctionValues.size() > type &&
437 FunctionValues[type] &&
438 Num < FunctionValues[type]->size())
439 return FunctionValues[type]->getOperand(Num);
441 if (!Create) return 0; // Do not create a placeholder?
443 std::pair<unsigned,unsigned> KeyValue(type, oNum);
444 ForwardReferenceMap::iterator I = ForwardReferences.lower_bound(KeyValue);
445 if (I != ForwardReferences.end() && I->first == KeyValue)
446 return I->second; // We have already created this placeholder
448 Value *Val = new Argument(getType(type));
449 ForwardReferences.insert(I, std::make_pair(KeyValue, Val));
453 /// This is just like getValue, but when a compaction table is in use, it
454 /// is ignored. Also, no forward references or other fancy features are
456 Value* BytecodeReader::getGlobalTableValue(const Type *Ty, unsigned SlotNo) {
457 // FIXME: getTypeSlot is inefficient!
458 unsigned TyID = getGlobalTableTypeSlot(Ty);
460 if (TyID != Type::LabelTyID) {
462 return Constant::getNullValue(Ty);
466 if (TyID >= ModuleValues.size() || ModuleValues[TyID] == 0 ||
467 SlotNo >= ModuleValues[TyID]->size()) {
468 error("Corrupt compaction table entry!"
469 + utostr(TyID) + ", " + utostr(SlotNo) + ": "
470 + utostr(ModuleValues.size()) + ", "
471 + utohexstr(intptr_t((void*)ModuleValues[TyID])) + ", "
472 + utostr(ModuleValues[TyID]->size()));
474 return ModuleValues[TyID]->getOperand(SlotNo);
477 /// Just like getValue, except that it returns a null pointer
478 /// only on error. It always returns a constant (meaning that if the value is
479 /// defined, but is not a constant, that is an error). If the specified
480 /// constant hasn't been parsed yet, a placeholder is defined and used.
481 /// Later, after the real value is parsed, the placeholder is eliminated.
482 Constant* BytecodeReader::getConstantValue(unsigned TypeSlot, unsigned Slot) {
483 if (Value *V = getValue(TypeSlot, Slot, false))
484 if (Constant *C = dyn_cast<Constant>(V))
485 return C; // If we already have the value parsed, just return it
487 error("Value for slot " + utostr(Slot) +
488 " is expected to be a constant!");
490 const Type *Ty = getType(TypeSlot);
491 std::pair<const Type*, unsigned> Key(Ty, Slot);
492 ConstantRefsType::iterator I = ConstantFwdRefs.lower_bound(Key);
494 if (I != ConstantFwdRefs.end() && I->first == Key) {
497 // Create a placeholder for the constant reference and
498 // keep track of the fact that we have a forward ref to recycle it
499 Constant *C = new ConstantPlaceHolder(Ty, Slot);
501 // Keep track of the fact that we have a forward ref to recycle it
502 ConstantFwdRefs.insert(I, std::make_pair(Key, C));
507 //===----------------------------------------------------------------------===//
508 // IR Construction Methods
509 //===----------------------------------------------------------------------===//
511 /// As values are created, they are inserted into the appropriate place
512 /// with this method. The ValueTable argument must be one of ModuleValues
513 /// or FunctionValues data members of this class.
514 unsigned BytecodeReader::insertValue(Value *Val, unsigned type,
515 ValueTable &ValueTab) {
516 assert((!isa<Constant>(Val) || !cast<Constant>(Val)->isNullValue()) ||
517 !hasImplicitNull(type) &&
518 "Cannot read null values from bytecode!");
520 if (ValueTab.size() <= type)
521 ValueTab.resize(type+1);
523 if (!ValueTab[type]) ValueTab[type] = new ValueList();
525 ValueTab[type]->push_back(Val);
527 bool HasOffset = hasImplicitNull(type);
528 return ValueTab[type]->size()-1 + HasOffset;
531 /// Insert the arguments of a function as new values in the reader.
532 void BytecodeReader::insertArguments(Function* F) {
533 const FunctionType *FT = F->getFunctionType();
534 Function::aiterator AI = F->abegin();
535 for (FunctionType::param_iterator It = FT->param_begin();
536 It != FT->param_end(); ++It, ++AI)
537 insertValue(AI, getTypeSlot(AI->getType()), FunctionValues);
540 //===----------------------------------------------------------------------===//
541 // Bytecode Parsing Methods
542 //===----------------------------------------------------------------------===//
544 /// This method parses a single instruction. The instruction is
545 /// inserted at the end of the \p BB provided. The arguments of
546 /// the instruction are provided in the \p Args vector.
547 void BytecodeReader::ParseInstruction(std::vector<unsigned> &Oprnds,
551 // Clear instruction data
555 unsigned Op = read_uint();
557 // bits Instruction format: Common to all formats
558 // --------------------------
559 // 01-00: Opcode type, fixed to 1.
561 Opcode = (Op >> 2) & 63;
562 Oprnds.resize((Op >> 0) & 03);
564 // Extract the operands
565 switch (Oprnds.size()) {
567 // bits Instruction format:
568 // --------------------------
569 // 19-08: Resulting type plane
570 // 31-20: Operand #1 (if set to (2^12-1), then zero operands)
572 iType = (Op >> 8) & 4095;
573 Oprnds[0] = (Op >> 20) & 4095;
574 if (Oprnds[0] == 4095) // Handle special encoding for 0 operands...
578 // bits Instruction format:
579 // --------------------------
580 // 15-08: Resulting type plane
584 iType = (Op >> 8) & 255;
585 Oprnds[0] = (Op >> 16) & 255;
586 Oprnds[1] = (Op >> 24) & 255;
589 // bits Instruction format:
590 // --------------------------
591 // 13-08: Resulting type plane
596 iType = (Op >> 8) & 63;
597 Oprnds[0] = (Op >> 14) & 63;
598 Oprnds[1] = (Op >> 20) & 63;
599 Oprnds[2] = (Op >> 26) & 63;
602 At -= 4; // Hrm, try this again...
603 Opcode = read_vbr_uint();
605 iType = read_vbr_uint();
607 unsigned NumOprnds = read_vbr_uint();
608 Oprnds.resize(NumOprnds);
611 error("Zero-argument instruction found; this is invalid.");
613 for (unsigned i = 0; i != NumOprnds; ++i)
614 Oprnds[i] = read_vbr_uint();
619 const Type *InstTy = getSanitizedType(iType);
621 // We have enough info to inform the handler now.
622 if (Handler) Handler->handleInstruction(Opcode, InstTy, Oprnds, At-SaveAt);
624 // Declare the resulting instruction we'll build.
625 Instruction *Result = 0;
627 // Handle binary operators
628 if (Opcode >= Instruction::BinaryOpsBegin &&
629 Opcode < Instruction::BinaryOpsEnd && Oprnds.size() == 2)
630 Result = BinaryOperator::create((Instruction::BinaryOps)Opcode,
631 getValue(iType, Oprnds[0]),
632 getValue(iType, Oprnds[1]));
637 error("Illegal instruction read!");
639 case Instruction::VAArg:
640 Result = new VAArgInst(getValue(iType, Oprnds[0]),
641 getSanitizedType(Oprnds[1]));
643 case Instruction::VANext:
644 Result = new VANextInst(getValue(iType, Oprnds[0]),
645 getSanitizedType(Oprnds[1]));
647 case Instruction::Cast:
648 Result = new CastInst(getValue(iType, Oprnds[0]),
649 getSanitizedType(Oprnds[1]));
651 case Instruction::Select:
652 Result = new SelectInst(getValue(Type::BoolTyID, Oprnds[0]),
653 getValue(iType, Oprnds[1]),
654 getValue(iType, Oprnds[2]));
656 case Instruction::PHI: {
657 if (Oprnds.size() == 0 || (Oprnds.size() & 1))
658 error("Invalid phi node encountered!");
660 PHINode *PN = new PHINode(InstTy);
661 PN->op_reserve(Oprnds.size());
662 for (unsigned i = 0, e = Oprnds.size(); i != e; i += 2)
663 PN->addIncoming(getValue(iType, Oprnds[i]), getBasicBlock(Oprnds[i+1]));
668 case Instruction::Shl:
669 case Instruction::Shr:
670 Result = new ShiftInst((Instruction::OtherOps)Opcode,
671 getValue(iType, Oprnds[0]),
672 getValue(Type::UByteTyID, Oprnds[1]));
674 case Instruction::Ret:
675 if (Oprnds.size() == 0)
676 Result = new ReturnInst();
677 else if (Oprnds.size() == 1)
678 Result = new ReturnInst(getValue(iType, Oprnds[0]));
680 error("Unrecognized instruction!");
683 case Instruction::Br:
684 if (Oprnds.size() == 1)
685 Result = new BranchInst(getBasicBlock(Oprnds[0]));
686 else if (Oprnds.size() == 3)
687 Result = new BranchInst(getBasicBlock(Oprnds[0]),
688 getBasicBlock(Oprnds[1]), getValue(Type::BoolTyID , Oprnds[2]));
690 error("Invalid number of operands for a 'br' instruction!");
692 case Instruction::Switch: {
693 if (Oprnds.size() & 1)
694 error("Switch statement with odd number of arguments!");
696 SwitchInst *I = new SwitchInst(getValue(iType, Oprnds[0]),
697 getBasicBlock(Oprnds[1]));
698 for (unsigned i = 2, e = Oprnds.size(); i != e; i += 2)
699 I->addCase(cast<Constant>(getValue(iType, Oprnds[i])),
700 getBasicBlock(Oprnds[i+1]));
705 case Instruction::Call: {
706 if (Oprnds.size() == 0)
707 error("Invalid call instruction encountered!");
709 Value *F = getValue(iType, Oprnds[0]);
711 // Check to make sure we have a pointer to function type
712 const PointerType *PTy = dyn_cast<PointerType>(F->getType());
713 if (PTy == 0) error("Call to non function pointer value!");
714 const FunctionType *FTy = dyn_cast<FunctionType>(PTy->getElementType());
715 if (FTy == 0) error("Call to non function pointer value!");
717 std::vector<Value *> Params;
718 if (!FTy->isVarArg()) {
719 FunctionType::param_iterator It = FTy->param_begin();
721 for (unsigned i = 1, e = Oprnds.size(); i != e; ++i) {
722 if (It == FTy->param_end())
723 error("Invalid call instruction!");
724 Params.push_back(getValue(getTypeSlot(*It++), Oprnds[i]));
726 if (It != FTy->param_end())
727 error("Invalid call instruction!");
729 Oprnds.erase(Oprnds.begin(), Oprnds.begin()+1);
731 unsigned FirstVariableOperand;
732 if (Oprnds.size() < FTy->getNumParams())
733 error("Call instruction missing operands!");
735 // Read all of the fixed arguments
736 for (unsigned i = 0, e = FTy->getNumParams(); i != e; ++i)
737 Params.push_back(getValue(getTypeSlot(FTy->getParamType(i)),Oprnds[i]));
739 FirstVariableOperand = FTy->getNumParams();
741 if ((Oprnds.size()-FirstVariableOperand) & 1) // Must be pairs of type/value
742 error("Invalid call instruction!");
744 for (unsigned i = FirstVariableOperand, e = Oprnds.size();
746 Params.push_back(getValue(Oprnds[i], Oprnds[i+1]));
749 Result = new CallInst(F, Params);
752 case Instruction::Invoke: {
753 if (Oprnds.size() < 3)
754 error("Invalid invoke instruction!");
755 Value *F = getValue(iType, Oprnds[0]);
757 // Check to make sure we have a pointer to function type
758 const PointerType *PTy = dyn_cast<PointerType>(F->getType());
760 error("Invoke to non function pointer value!");
761 const FunctionType *FTy = dyn_cast<FunctionType>(PTy->getElementType());
763 error("Invoke to non function pointer value!");
765 std::vector<Value *> Params;
766 BasicBlock *Normal, *Except;
768 if (!FTy->isVarArg()) {
769 Normal = getBasicBlock(Oprnds[1]);
770 Except = getBasicBlock(Oprnds[2]);
772 FunctionType::param_iterator It = FTy->param_begin();
773 for (unsigned i = 3, e = Oprnds.size(); i != e; ++i) {
774 if (It == FTy->param_end())
775 error("Invalid invoke instruction!");
776 Params.push_back(getValue(getTypeSlot(*It++), Oprnds[i]));
778 if (It != FTy->param_end())
779 error("Invalid invoke instruction!");
781 Oprnds.erase(Oprnds.begin(), Oprnds.begin()+1);
783 Normal = getBasicBlock(Oprnds[0]);
784 Except = getBasicBlock(Oprnds[1]);
786 unsigned FirstVariableArgument = FTy->getNumParams()+2;
787 for (unsigned i = 2; i != FirstVariableArgument; ++i)
788 Params.push_back(getValue(getTypeSlot(FTy->getParamType(i-2)),
791 if (Oprnds.size()-FirstVariableArgument & 1) // Must be type/value pairs
792 error("Invalid invoke instruction!");
794 for (unsigned i = FirstVariableArgument; i < Oprnds.size(); i += 2)
795 Params.push_back(getValue(Oprnds[i], Oprnds[i+1]));
798 Result = new InvokeInst(F, Normal, Except, Params);
801 case Instruction::Malloc:
802 if (Oprnds.size() > 2)
803 error("Invalid malloc instruction!");
804 if (!isa<PointerType>(InstTy))
805 error("Invalid malloc instruction!");
807 Result = new MallocInst(cast<PointerType>(InstTy)->getElementType(),
808 Oprnds.size() ? getValue(Type::UIntTyID,
812 case Instruction::Alloca:
813 if (Oprnds.size() > 2)
814 error("Invalid alloca instruction!");
815 if (!isa<PointerType>(InstTy))
816 error("Invalid alloca instruction!");
818 Result = new AllocaInst(cast<PointerType>(InstTy)->getElementType(),
819 Oprnds.size() ? getValue(Type::UIntTyID,
822 case Instruction::Free:
823 if (!isa<PointerType>(InstTy))
824 error("Invalid free instruction!");
825 Result = new FreeInst(getValue(iType, Oprnds[0]));
827 case Instruction::GetElementPtr: {
828 if (Oprnds.size() == 0 || !isa<PointerType>(InstTy))
829 error("Invalid getelementptr instruction!");
831 std::vector<Value*> Idx;
833 const Type *NextTy = InstTy;
834 for (unsigned i = 1, e = Oprnds.size(); i != e; ++i) {
835 const CompositeType *TopTy = dyn_cast_or_null<CompositeType>(NextTy);
837 error("Invalid getelementptr instruction!");
839 unsigned ValIdx = Oprnds[i];
841 if (!hasRestrictedGEPTypes) {
842 // Struct indices are always uints, sequential type indices can be any
843 // of the 32 or 64-bit integer types. The actual choice of type is
844 // encoded in the low two bits of the slot number.
845 if (isa<StructType>(TopTy))
846 IdxTy = Type::UIntTyID;
848 switch (ValIdx & 3) {
850 case 0: IdxTy = Type::UIntTyID; break;
851 case 1: IdxTy = Type::IntTyID; break;
852 case 2: IdxTy = Type::ULongTyID; break;
853 case 3: IdxTy = Type::LongTyID; break;
858 IdxTy = isa<StructType>(TopTy) ? Type::UByteTyID : Type::LongTyID;
861 Idx.push_back(getValue(IdxTy, ValIdx));
863 // Convert ubyte struct indices into uint struct indices.
864 if (isa<StructType>(TopTy) && hasRestrictedGEPTypes)
865 if (ConstantUInt *C = dyn_cast<ConstantUInt>(Idx.back()))
866 Idx[Idx.size()-1] = ConstantExpr::getCast(C, Type::UIntTy);
868 NextTy = GetElementPtrInst::getIndexedType(InstTy, Idx, true);
871 Result = new GetElementPtrInst(getValue(iType, Oprnds[0]), Idx);
875 case 62: // volatile load
876 case Instruction::Load:
877 if (Oprnds.size() != 1 || !isa<PointerType>(InstTy))
878 error("Invalid load instruction!");
879 Result = new LoadInst(getValue(iType, Oprnds[0]), "", Opcode == 62);
882 case 63: // volatile store
883 case Instruction::Store: {
884 if (!isa<PointerType>(InstTy) || Oprnds.size() != 2)
885 error("Invalid store instruction!");
887 Value *Ptr = getValue(iType, Oprnds[1]);
888 const Type *ValTy = cast<PointerType>(Ptr->getType())->getElementType();
889 Result = new StoreInst(getValue(getTypeSlot(ValTy), Oprnds[0]), Ptr,
893 case Instruction::Unwind:
894 if (Oprnds.size() != 0)
895 error("Invalid unwind instruction!");
896 Result = new UnwindInst();
898 } // end switch(Opcode)
901 if (Result->getType() == InstTy)
904 TypeSlot = getTypeSlot(Result->getType());
906 insertValue(Result, TypeSlot, FunctionValues);
907 BB->getInstList().push_back(Result);
910 /// Get a particular numbered basic block, which might be a forward reference.
911 /// This works together with ParseBasicBlock to handle these forward references
912 /// in a clean manner. This function is used when constructing phi, br, switch,
913 /// and other instructions that reference basic blocks. Blocks are numbered
914 /// sequentially as they appear in the function.
915 BasicBlock *BytecodeReader::getBasicBlock(unsigned ID) {
916 // Make sure there is room in the table...
917 if (ParsedBasicBlocks.size() <= ID) ParsedBasicBlocks.resize(ID+1);
919 // First check to see if this is a backwards reference, i.e., ParseBasicBlock
920 // has already created this block, or if the forward reference has already
922 if (ParsedBasicBlocks[ID])
923 return ParsedBasicBlocks[ID];
925 // Otherwise, the basic block has not yet been created. Do so and add it to
926 // the ParsedBasicBlocks list.
927 return ParsedBasicBlocks[ID] = new BasicBlock();
930 /// In LLVM 1.0 bytecode files, we used to output one basicblock at a time.
931 /// This method reads in one of the basicblock packets. This method is not used
932 /// for bytecode files after LLVM 1.0
933 /// @returns The basic block constructed.
934 BasicBlock *BytecodeReader::ParseBasicBlock(unsigned BlockNo) {
935 if (Handler) Handler->handleBasicBlockBegin(BlockNo);
939 if (ParsedBasicBlocks.size() == BlockNo)
940 ParsedBasicBlocks.push_back(BB = new BasicBlock());
941 else if (ParsedBasicBlocks[BlockNo] == 0)
942 BB = ParsedBasicBlocks[BlockNo] = new BasicBlock();
944 BB = ParsedBasicBlocks[BlockNo];
946 std::vector<unsigned> Operands;
947 while (moreInBlock())
948 ParseInstruction(Operands, BB);
950 if (Handler) Handler->handleBasicBlockEnd(BlockNo);
954 /// Parse all of the BasicBlock's & Instruction's in the body of a function.
955 /// In post 1.0 bytecode files, we no longer emit basic block individually,
956 /// in order to avoid per-basic-block overhead.
957 /// @returns Rhe number of basic blocks encountered.
958 unsigned BytecodeReader::ParseInstructionList(Function* F) {
959 unsigned BlockNo = 0;
960 std::vector<unsigned> Args;
962 while (moreInBlock()) {
963 if (Handler) Handler->handleBasicBlockBegin(BlockNo);
965 if (ParsedBasicBlocks.size() == BlockNo)
966 ParsedBasicBlocks.push_back(BB = new BasicBlock());
967 else if (ParsedBasicBlocks[BlockNo] == 0)
968 BB = ParsedBasicBlocks[BlockNo] = new BasicBlock();
970 BB = ParsedBasicBlocks[BlockNo];
972 F->getBasicBlockList().push_back(BB);
974 // Read instructions into this basic block until we get to a terminator
975 while (moreInBlock() && !BB->getTerminator())
976 ParseInstruction(Args, BB);
978 if (!BB->getTerminator())
979 error("Non-terminated basic block found!");
981 if (Handler) Handler->handleBasicBlockEnd(BlockNo-1);
987 /// Parse a symbol table. This works for both module level and function
988 /// level symbol tables. For function level symbol tables, the CurrentFunction
989 /// parameter must be non-zero and the ST parameter must correspond to
990 /// CurrentFunction's symbol table. For Module level symbol tables, the
991 /// CurrentFunction argument must be zero.
992 void BytecodeReader::ParseSymbolTable(Function *CurrentFunction,
994 if (Handler) Handler->handleSymbolTableBegin(CurrentFunction,ST);
996 // Allow efficient basic block lookup by number.
997 std::vector<BasicBlock*> BBMap;
999 for (Function::iterator I = CurrentFunction->begin(),
1000 E = CurrentFunction->end(); I != E; ++I)
1003 /// In LLVM 1.3 we write types separately from values so
1004 /// The types are always first in the symbol table. This is
1005 /// because Type no longer derives from Value.
1006 if (!hasTypeDerivedFromValue) {
1007 // Symtab block header: [num entries]
1008 unsigned NumEntries = read_vbr_uint();
1009 for (unsigned i = 0; i < NumEntries; ++i) {
1010 // Symtab entry: [def slot #][name]
1011 unsigned slot = read_vbr_uint();
1012 std::string Name = read_str();
1013 const Type* T = getType(slot);
1014 ST->insert(Name, T);
1018 while (moreInBlock()) {
1019 // Symtab block header: [num entries][type id number]
1020 unsigned NumEntries = read_vbr_uint();
1022 bool isTypeType = read_typeid(Typ);
1023 const Type *Ty = getType(Typ);
1025 for (unsigned i = 0; i != NumEntries; ++i) {
1026 // Symtab entry: [def slot #][name]
1027 unsigned slot = read_vbr_uint();
1028 std::string Name = read_str();
1030 // if we're reading a pre 1.3 bytecode file and the type plane
1031 // is the "type type", handle it here
1033 const Type* T = getType(slot);
1035 error("Failed type look-up for name '" + Name + "'");
1036 ST->insert(Name, T);
1037 continue; // code below must be short circuited
1040 if (Typ == Type::LabelTyID) {
1041 if (slot < BBMap.size())
1044 V = getValue(Typ, slot, false); // Find mapping...
1047 error("Failed value look-up for name '" + Name + "'");
1048 V->setName(Name, ST);
1052 checkPastBlockEnd("Symbol Table");
1053 if (Handler) Handler->handleSymbolTableEnd();
1056 /// Read in the types portion of a compaction table.
1057 void BytecodeReader::ParseCompactionTypes(unsigned NumEntries) {
1058 for (unsigned i = 0; i != NumEntries; ++i) {
1059 unsigned TypeSlot = 0;
1060 if (read_typeid(TypeSlot))
1061 error("Invalid type in compaction table: type type");
1062 const Type *Typ = getGlobalTableType(TypeSlot);
1063 CompactionTypes.push_back(Typ);
1064 if (Handler) Handler->handleCompactionTableType(i, TypeSlot, Typ);
1068 /// Parse a compaction table.
1069 void BytecodeReader::ParseCompactionTable() {
1071 // Notify handler that we're beginning a compaction table.
1072 if (Handler) Handler->handleCompactionTableBegin();
1074 // In LLVM 1.3 Type no longer derives from Value. So,
1075 // we always write them first in the compaction table
1076 // because they can't occupy a "type plane" where the
1078 if (! hasTypeDerivedFromValue) {
1079 unsigned NumEntries = read_vbr_uint();
1080 ParseCompactionTypes(NumEntries);
1083 // Compaction tables live in separate blocks so we have to loop
1084 // until we've read the whole thing.
1085 while (moreInBlock()) {
1086 // Read the number of Value* entries in the compaction table
1087 unsigned NumEntries = read_vbr_uint();
1089 unsigned isTypeType = false;
1091 // Decode the type from value read in. Most compaction table
1092 // planes will have one or two entries in them. If that's the
1093 // case then the length is encoded in the bottom two bits and
1094 // the higher bits encode the type. This saves another VBR value.
1095 if ((NumEntries & 3) == 3) {
1096 // In this case, both low-order bits are set (value 3). This
1097 // is a signal that the typeid follows.
1099 isTypeType = read_typeid(Ty);
1101 // In this case, the low-order bits specify the number of entries
1102 // and the high order bits specify the type.
1103 Ty = NumEntries >> 2;
1104 isTypeType = sanitizeTypeId(Ty);
1108 // if we're reading a pre 1.3 bytecode file and the type plane
1109 // is the "type type", handle it here
1111 ParseCompactionTypes(NumEntries);
1113 // Make sure we have enough room for the plane
1114 if (Ty >= CompactionValues.size())
1115 CompactionValues.resize(Ty+1);
1117 // Make sure the plane is empty or we have some kind of error
1118 if (!CompactionValues[Ty].empty())
1119 error("Compaction table plane contains multiple entries!");
1121 // Notify handler about the plane
1122 if (Handler) Handler->handleCompactionTablePlane(Ty, NumEntries);
1124 // Convert the type slot to a type
1125 const Type *Typ = getType(Ty);
1127 // Push the implicit zero
1128 CompactionValues[Ty].push_back(Constant::getNullValue(Typ));
1130 // Read in each of the entries, put them in the compaction table
1131 // and notify the handler that we have a new compaction table value.
1132 for (unsigned i = 0; i != NumEntries; ++i) {
1133 unsigned ValSlot = read_vbr_uint();
1134 Value *V = getGlobalTableValue(Typ, ValSlot);
1135 CompactionValues[Ty].push_back(V);
1136 if (Handler) Handler->handleCompactionTableValue(i, Ty, ValSlot, Typ);
1140 // Notify handler that the compaction table is done.
1141 if (Handler) Handler->handleCompactionTableEnd();
1144 // Parse a single type. The typeid is read in first. If its a primitive type
1145 // then nothing else needs to be read, we know how to instantiate it. If its
1146 // a derived type, then additional data is read to fill out the type
1148 const Type *BytecodeReader::ParseType() {
1149 unsigned PrimType = 0;
1150 if (read_typeid(PrimType))
1151 error("Invalid type (type type) in type constants!");
1153 const Type *Result = 0;
1154 if ((Result = Type::getPrimitiveType((Type::TypeID)PrimType)))
1158 case Type::FunctionTyID: {
1159 const Type *RetType = readSanitizedType();
1161 unsigned NumParams = read_vbr_uint();
1163 std::vector<const Type*> Params;
1165 Params.push_back(readSanitizedType());
1167 bool isVarArg = Params.size() && Params.back() == Type::VoidTy;
1168 if (isVarArg) Params.pop_back();
1170 Result = FunctionType::get(RetType, Params, isVarArg);
1173 case Type::ArrayTyID: {
1174 const Type *ElementType = readSanitizedType();
1175 unsigned NumElements = read_vbr_uint();
1176 Result = ArrayType::get(ElementType, NumElements);
1179 case Type::StructTyID: {
1180 std::vector<const Type*> Elements;
1182 if (read_typeid(Typ))
1183 error("Invalid element type (type type) for structure!");
1185 while (Typ) { // List is terminated by void/0 typeid
1186 Elements.push_back(getType(Typ));
1187 if (read_typeid(Typ))
1188 error("Invalid element type (type type) for structure!");
1191 Result = StructType::get(Elements);
1194 case Type::PointerTyID: {
1195 Result = PointerType::get(readSanitizedType());
1199 case Type::OpaqueTyID: {
1200 Result = OpaqueType::get();
1205 error("Don't know how to deserialize primitive type " + utostr(PrimType));
1208 if (Handler) Handler->handleType(Result);
1212 // ParseType - We have to use this weird code to handle recursive
1213 // types. We know that recursive types will only reference the current slab of
1214 // values in the type plane, but they can forward reference types before they
1215 // have been read. For example, Type #0 might be '{ Ty#1 }' and Type #1 might
1216 // be 'Ty#0*'. When reading Type #0, type number one doesn't exist. To fix
1217 // this ugly problem, we pessimistically insert an opaque type for each type we
1218 // are about to read. This means that forward references will resolve to
1219 // something and when we reread the type later, we can replace the opaque type
1220 // with a new resolved concrete type.
1222 void BytecodeReader::ParseTypes(TypeListTy &Tab, unsigned NumEntries){
1223 assert(Tab.size() == 0 && "should not have read type constants in before!");
1225 // Insert a bunch of opaque types to be resolved later...
1226 Tab.reserve(NumEntries);
1227 for (unsigned i = 0; i != NumEntries; ++i)
1228 Tab.push_back(OpaqueType::get());
1230 // Loop through reading all of the types. Forward types will make use of the
1231 // opaque types just inserted.
1233 for (unsigned i = 0; i != NumEntries; ++i) {
1234 const Type* NewTy = ParseType();
1235 const Type* OldTy = Tab[i].get();
1237 error("Couldn't parse type!");
1239 // Don't directly push the new type on the Tab. Instead we want to replace
1240 // the opaque type we previously inserted with the new concrete value. This
1241 // approach helps with forward references to types. The refinement from the
1242 // abstract (opaque) type to the new type causes all uses of the abstract
1243 // type to use the concrete type (NewTy). This will also cause the opaque
1244 // type to be deleted.
1245 cast<DerivedType>(const_cast<Type*>(OldTy))->refineAbstractTypeTo(NewTy);
1247 // This should have replaced the old opaque type with the new type in the
1248 // value table... or with a preexisting type that was already in the system.
1249 // Let's just make sure it did.
1250 assert(Tab[i] != OldTy && "refineAbstractType didn't work!");
1254 /// Parse a single constant value
1255 Constant *BytecodeReader::ParseConstantValue(unsigned TypeID) {
1256 // We must check for a ConstantExpr before switching by type because
1257 // a ConstantExpr can be of any type, and has no explicit value.
1259 // 0 if not expr; numArgs if is expr
1260 unsigned isExprNumArgs = read_vbr_uint();
1262 if (isExprNumArgs) {
1263 // FIXME: Encoding of constant exprs could be much more compact!
1264 std::vector<Constant*> ArgVec;
1265 ArgVec.reserve(isExprNumArgs);
1266 unsigned Opcode = read_vbr_uint();
1268 // Read the slot number and types of each of the arguments
1269 for (unsigned i = 0; i != isExprNumArgs; ++i) {
1270 unsigned ArgValSlot = read_vbr_uint();
1271 unsigned ArgTypeSlot = 0;
1272 if (read_typeid(ArgTypeSlot))
1273 error("Invalid argument type (type type) for constant value");
1275 // Get the arg value from its slot if it exists, otherwise a placeholder
1276 ArgVec.push_back(getConstantValue(ArgTypeSlot, ArgValSlot));
1279 // Construct a ConstantExpr of the appropriate kind
1280 if (isExprNumArgs == 1) { // All one-operand expressions
1281 if (Opcode != Instruction::Cast)
1282 error("Only Cast instruction has one argument for ConstantExpr");
1284 Constant* Result = ConstantExpr::getCast(ArgVec[0], getType(TypeID));
1285 if (Handler) Handler->handleConstantExpression(Opcode, ArgVec, Result);
1287 } else if (Opcode == Instruction::GetElementPtr) { // GetElementPtr
1288 std::vector<Constant*> IdxList(ArgVec.begin()+1, ArgVec.end());
1290 if (hasRestrictedGEPTypes) {
1291 const Type *BaseTy = ArgVec[0]->getType();
1292 generic_gep_type_iterator<std::vector<Constant*>::iterator>
1293 GTI = gep_type_begin(BaseTy, IdxList.begin(), IdxList.end()),
1294 E = gep_type_end(BaseTy, IdxList.begin(), IdxList.end());
1295 for (unsigned i = 0; GTI != E; ++GTI, ++i)
1296 if (isa<StructType>(*GTI)) {
1297 if (IdxList[i]->getType() != Type::UByteTy)
1298 error("Invalid index for getelementptr!");
1299 IdxList[i] = ConstantExpr::getCast(IdxList[i], Type::UIntTy);
1303 Constant* Result = ConstantExpr::getGetElementPtr(ArgVec[0], IdxList);
1304 if (Handler) Handler->handleConstantExpression(Opcode, ArgVec, Result);
1306 } else if (Opcode == Instruction::Select) {
1307 if (ArgVec.size() != 3)
1308 error("Select instruction must have three arguments.");
1309 Constant* Result = ConstantExpr::getSelect(ArgVec[0], ArgVec[1],
1311 if (Handler) Handler->handleConstantExpression(Opcode, ArgVec, Result);
1313 } else { // All other 2-operand expressions
1314 Constant* Result = ConstantExpr::get(Opcode, ArgVec[0], ArgVec[1]);
1315 if (Handler) Handler->handleConstantExpression(Opcode, ArgVec, Result);
1320 // Ok, not an ConstantExpr. We now know how to read the given type...
1321 const Type *Ty = getType(TypeID);
1322 switch (Ty->getTypeID()) {
1323 case Type::BoolTyID: {
1324 unsigned Val = read_vbr_uint();
1325 if (Val != 0 && Val != 1)
1326 error("Invalid boolean value read.");
1327 Constant* Result = ConstantBool::get(Val == 1);
1328 if (Handler) Handler->handleConstantValue(Result);
1332 case Type::UByteTyID: // Unsigned integer types...
1333 case Type::UShortTyID:
1334 case Type::UIntTyID: {
1335 unsigned Val = read_vbr_uint();
1336 if (!ConstantUInt::isValueValidForType(Ty, Val))
1337 error("Invalid unsigned byte/short/int read.");
1338 Constant* Result = ConstantUInt::get(Ty, Val);
1339 if (Handler) Handler->handleConstantValue(Result);
1343 case Type::ULongTyID: {
1344 Constant* Result = ConstantUInt::get(Ty, read_vbr_uint64());
1345 if (Handler) Handler->handleConstantValue(Result);
1349 case Type::SByteTyID: // Signed integer types...
1350 case Type::ShortTyID:
1351 case Type::IntTyID: {
1352 case Type::LongTyID:
1353 int64_t Val = read_vbr_int64();
1354 if (!ConstantSInt::isValueValidForType(Ty, Val))
1355 error("Invalid signed byte/short/int/long read.");
1356 Constant* Result = ConstantSInt::get(Ty, Val);
1357 if (Handler) Handler->handleConstantValue(Result);
1361 case Type::FloatTyID: {
1364 Constant* Result = ConstantFP::get(Ty, Val);
1365 if (Handler) Handler->handleConstantValue(Result);
1369 case Type::DoubleTyID: {
1372 Constant* Result = ConstantFP::get(Ty, Val);
1373 if (Handler) Handler->handleConstantValue(Result);
1377 case Type::ArrayTyID: {
1378 const ArrayType *AT = cast<ArrayType>(Ty);
1379 unsigned NumElements = AT->getNumElements();
1380 unsigned TypeSlot = getTypeSlot(AT->getElementType());
1381 std::vector<Constant*> Elements;
1382 Elements.reserve(NumElements);
1383 while (NumElements--) // Read all of the elements of the constant.
1384 Elements.push_back(getConstantValue(TypeSlot,
1386 Constant* Result = ConstantArray::get(AT, Elements);
1387 if (Handler) Handler->handleConstantArray(AT, Elements, TypeSlot, Result);
1391 case Type::StructTyID: {
1392 const StructType *ST = cast<StructType>(Ty);
1394 std::vector<Constant *> Elements;
1395 Elements.reserve(ST->getNumElements());
1396 for (unsigned i = 0; i != ST->getNumElements(); ++i)
1397 Elements.push_back(getConstantValue(ST->getElementType(i),
1400 Constant* Result = ConstantStruct::get(ST, Elements);
1401 if (Handler) Handler->handleConstantStruct(ST, Elements, Result);
1405 case Type::PointerTyID: { // ConstantPointerRef value...
1406 const PointerType *PT = cast<PointerType>(Ty);
1407 unsigned Slot = read_vbr_uint();
1409 // Check to see if we have already read this global variable...
1410 Value *Val = getValue(TypeID, Slot, false);
1413 if (!(GV = dyn_cast<GlobalValue>(Val)))
1414 error("GlobalValue not in ValueTable!");
1416 error("Forward references are not allowed here.");
1419 if (Handler) Handler->handleConstantPointer(PT, Slot, GV );
1424 error("Don't know how to deserialize constant value of type '" +
1425 Ty->getDescription());
1431 /// Resolve references for constants. This function resolves the forward
1432 /// referenced constants in the ConstantFwdRefs map. It uses the
1433 /// replaceAllUsesWith method of Value class to substitute the placeholder
1434 /// instance with the actual instance.
1435 void BytecodeReader::ResolveReferencesToConstant(Constant *NewV, unsigned Slot){
1436 ConstantRefsType::iterator I =
1437 ConstantFwdRefs.find(std::make_pair(NewV->getType(), Slot));
1438 if (I == ConstantFwdRefs.end()) return; // Never forward referenced?
1440 Value *PH = I->second; // Get the placeholder...
1441 PH->replaceAllUsesWith(NewV);
1442 delete PH; // Delete the old placeholder
1443 ConstantFwdRefs.erase(I); // Remove the map entry for it
1446 /// Parse the constant strings section.
1447 void BytecodeReader::ParseStringConstants(unsigned NumEntries, ValueTable &Tab){
1448 for (; NumEntries; --NumEntries) {
1450 if (read_typeid(Typ))
1451 error("Invalid type (type type) for string constant");
1452 const Type *Ty = getType(Typ);
1453 if (!isa<ArrayType>(Ty))
1454 error("String constant data invalid!");
1456 const ArrayType *ATy = cast<ArrayType>(Ty);
1457 if (ATy->getElementType() != Type::SByteTy &&
1458 ATy->getElementType() != Type::UByteTy)
1459 error("String constant data invalid!");
1461 // Read character data. The type tells us how long the string is.
1462 char Data[ATy->getNumElements()];
1463 read_data(Data, Data+ATy->getNumElements());
1465 std::vector<Constant*> Elements(ATy->getNumElements());
1466 if (ATy->getElementType() == Type::SByteTy)
1467 for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i)
1468 Elements[i] = ConstantSInt::get(Type::SByteTy, (signed char)Data[i]);
1470 for (unsigned i = 0, e = ATy->getNumElements(); i != e; ++i)
1471 Elements[i] = ConstantUInt::get(Type::UByteTy, (unsigned char)Data[i]);
1473 // Create the constant, inserting it as needed.
1474 Constant *C = ConstantArray::get(ATy, Elements);
1475 unsigned Slot = insertValue(C, Typ, Tab);
1476 ResolveReferencesToConstant(C, Slot);
1477 if (Handler) Handler->handleConstantString(cast<ConstantArray>(C));
1481 /// Parse the constant pool.
1482 void BytecodeReader::ParseConstantPool(ValueTable &Tab,
1483 TypeListTy &TypeTab,
1485 if (Handler) Handler->handleGlobalConstantsBegin();
1487 /// In LLVM 1.3 Type does not derive from Value so the types
1488 /// do not occupy a plane. Consequently, we read the types
1489 /// first in the constant pool.
1490 if (isFunction && !hasTypeDerivedFromValue) {
1491 unsigned NumEntries = read_vbr_uint();
1492 ParseTypes(TypeTab, NumEntries);
1495 while (moreInBlock()) {
1496 unsigned NumEntries = read_vbr_uint();
1498 bool isTypeType = read_typeid(Typ);
1500 /// In LLVM 1.2 and before, Types were written to the
1501 /// bytecode file in the "Type Type" plane (#12).
1502 /// In 1.3 plane 12 is now the label plane. Handle this here.
1504 ParseTypes(TypeTab, NumEntries);
1505 } else if (Typ == Type::VoidTyID) {
1506 /// Use of Type::VoidTyID is a misnomer. It actually means
1507 /// that the following plane is constant strings
1508 assert(&Tab == &ModuleValues && "Cannot read strings in functions!");
1509 ParseStringConstants(NumEntries, Tab);
1511 for (unsigned i = 0; i < NumEntries; ++i) {
1512 Constant *C = ParseConstantValue(Typ);
1513 assert(C && "ParseConstantValue returned NULL!");
1514 unsigned Slot = insertValue(C, Typ, Tab);
1516 // If we are reading a function constant table, make sure that we adjust
1517 // the slot number to be the real global constant number.
1519 if (&Tab != &ModuleValues && Typ < ModuleValues.size() &&
1521 Slot += ModuleValues[Typ]->size();
1522 ResolveReferencesToConstant(C, Slot);
1526 checkPastBlockEnd("Constant Pool");
1527 if (Handler) Handler->handleGlobalConstantsEnd();
1530 /// Parse the contents of a function. Note that this function can be
1531 /// called lazily by materializeFunction
1532 /// @see materializeFunction
1533 void BytecodeReader::ParseFunctionBody(Function* F) {
1535 unsigned FuncSize = BlockEnd - At;
1536 GlobalValue::LinkageTypes Linkage = GlobalValue::ExternalLinkage;
1538 unsigned LinkageType = read_vbr_uint();
1539 switch (LinkageType) {
1540 case 0: Linkage = GlobalValue::ExternalLinkage; break;
1541 case 1: Linkage = GlobalValue::WeakLinkage; break;
1542 case 2: Linkage = GlobalValue::AppendingLinkage; break;
1543 case 3: Linkage = GlobalValue::InternalLinkage; break;
1544 case 4: Linkage = GlobalValue::LinkOnceLinkage; break;
1546 error("Invalid linkage type for Function.");
1547 Linkage = GlobalValue::InternalLinkage;
1551 F->setLinkage(Linkage);
1552 if (Handler) Handler->handleFunctionBegin(F,FuncSize);
1554 // Keep track of how many basic blocks we have read in...
1555 unsigned BlockNum = 0;
1556 bool InsertedArguments = false;
1558 BufPtr MyEnd = BlockEnd;
1559 while (At < MyEnd) {
1560 unsigned Type, Size;
1562 read_block(Type, Size);
1565 case BytecodeFormat::ConstantPoolBlockID:
1566 if (!InsertedArguments) {
1567 // Insert arguments into the value table before we parse the first basic
1568 // block in the function, but after we potentially read in the
1569 // compaction table.
1571 InsertedArguments = true;
1574 ParseConstantPool(FunctionValues, FunctionTypes, true);
1577 case BytecodeFormat::CompactionTableBlockID:
1578 ParseCompactionTable();
1581 case BytecodeFormat::BasicBlock: {
1582 if (!InsertedArguments) {
1583 // Insert arguments into the value table before we parse the first basic
1584 // block in the function, but after we potentially read in the
1585 // compaction table.
1587 InsertedArguments = true;
1590 BasicBlock *BB = ParseBasicBlock(BlockNum++);
1591 F->getBasicBlockList().push_back(BB);
1595 case BytecodeFormat::InstructionListBlockID: {
1596 // Insert arguments into the value table before we parse the instruction
1597 // list for the function, but after we potentially read in the compaction
1599 if (!InsertedArguments) {
1601 InsertedArguments = true;
1605 error("Already parsed basic blocks!");
1606 BlockNum = ParseInstructionList(F);
1610 case BytecodeFormat::SymbolTableBlockID:
1611 ParseSymbolTable(F, &F->getSymbolTable());
1617 error("Wrapped around reading bytecode.");
1622 // Malformed bc file if read past end of block.
1626 // Make sure there were no references to non-existant basic blocks.
1627 if (BlockNum != ParsedBasicBlocks.size())
1628 error("Illegal basic block operand reference");
1630 ParsedBasicBlocks.clear();
1632 // Resolve forward references. Replace any uses of a forward reference value
1633 // with the real value.
1635 // replaceAllUsesWith is very inefficient for instructions which have a LARGE
1636 // number of operands. PHI nodes often have forward references, and can also
1637 // often have a very large number of operands.
1639 // FIXME: REEVALUATE. replaceAllUsesWith is _much_ faster now, and this code
1640 // should be simplified back to using it!
1642 std::map<Value*, Value*> ForwardRefMapping;
1643 for (std::map<std::pair<unsigned,unsigned>, Value*>::iterator
1644 I = ForwardReferences.begin(), E = ForwardReferences.end();
1646 ForwardRefMapping[I->second] = getValue(I->first.first, I->first.second,
1649 for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
1650 for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
1651 for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
1652 if (Argument *A = dyn_cast<Argument>(I->getOperand(i))) {
1653 std::map<Value*, Value*>::iterator It = ForwardRefMapping.find(A);
1654 if (It != ForwardRefMapping.end()) I->setOperand(i, It->second);
1657 while (!ForwardReferences.empty()) {
1658 std::map<std::pair<unsigned,unsigned>, Value*>::iterator I =
1659 ForwardReferences.begin();
1660 Value *PlaceHolder = I->second;
1661 ForwardReferences.erase(I);
1663 // Now that all the uses are gone, delete the placeholder...
1664 // If we couldn't find a def (error case), then leak a little
1665 // memory, because otherwise we can't remove all uses!
1669 // Clear out function-level types...
1670 FunctionTypes.clear();
1671 CompactionTypes.clear();
1672 CompactionValues.clear();
1673 freeTable(FunctionValues);
1675 if (Handler) Handler->handleFunctionEnd(F);
1678 /// This function parses LLVM functions lazily. It obtains the type of the
1679 /// function and records where the body of the function is in the bytecode
1680 /// buffer. The caller can then use the ParseNextFunction and
1681 /// ParseAllFunctionBodies to get handler events for the functions.
1682 void BytecodeReader::ParseFunctionLazily() {
1683 if (FunctionSignatureList.empty())
1684 error("FunctionSignatureList empty!");
1686 Function *Func = FunctionSignatureList.back();
1687 FunctionSignatureList.pop_back();
1689 // Save the information for future reading of the function
1690 LazyFunctionLoadMap[Func] = LazyFunctionInfo(BlockStart, BlockEnd);
1692 // Pretend we've `parsed' this function
1696 /// The ParserFunction method lazily parses one function. Use this method to
1697 /// casue the parser to parse a specific function in the module. Note that
1698 /// this will remove the function from what is to be included by
1699 /// ParseAllFunctionBodies.
1700 /// @see ParseAllFunctionBodies
1701 /// @see ParseBytecode
1702 void BytecodeReader::ParseFunction(Function* Func) {
1703 // Find {start, end} pointers and slot in the map. If not there, we're done.
1704 LazyFunctionMap::iterator Fi = LazyFunctionLoadMap.find(Func);
1706 // Make sure we found it
1707 if (Fi == LazyFunctionLoadMap.end()) {
1708 error("Unrecognized function of type " + Func->getType()->getDescription());
1712 BlockStart = At = Fi->second.Buf;
1713 BlockEnd = Fi->second.EndBuf;
1714 assert(Fi->first == Func && "Found wrong function?");
1716 LazyFunctionLoadMap.erase(Fi);
1718 this->ParseFunctionBody(Func);
1721 /// The ParseAllFunctionBodies method parses through all the previously
1722 /// unparsed functions in the bytecode file. If you want to completely parse
1723 /// a bytecode file, this method should be called after Parsebytecode because
1724 /// Parsebytecode only records the locations in the bytecode file of where
1725 /// the function definitions are located. This function uses that information
1726 /// to materialize the functions.
1727 /// @see ParseBytecode
1728 void BytecodeReader::ParseAllFunctionBodies() {
1729 LazyFunctionMap::iterator Fi = LazyFunctionLoadMap.begin();
1730 LazyFunctionMap::iterator Fe = LazyFunctionLoadMap.end();
1733 Function* Func = Fi->first;
1734 BlockStart = At = Fi->second.Buf;
1735 BlockEnd = Fi->second.EndBuf;
1736 this->ParseFunctionBody(Func);
1741 /// Parse the global type list
1742 void BytecodeReader::ParseGlobalTypes() {
1743 // Read the number of types
1744 unsigned NumEntries = read_vbr_uint();
1746 // Ignore the type plane identifier for types if the bc file is pre 1.3
1747 if (hasTypeDerivedFromValue)
1750 ParseTypes(ModuleTypes, NumEntries);
1753 /// Parse the Global info (types, global vars, constants)
1754 void BytecodeReader::ParseModuleGlobalInfo() {
1756 if (Handler) Handler->handleModuleGlobalsBegin();
1758 // Read global variables...
1759 unsigned VarType = read_vbr_uint();
1760 while (VarType != Type::VoidTyID) { // List is terminated by Void
1761 // VarType Fields: bit0 = isConstant, bit1 = hasInitializer, bit2,3,4 =
1762 // Linkage, bit4+ = slot#
1763 unsigned SlotNo = VarType >> 5;
1764 if (sanitizeTypeId(SlotNo))
1765 error("Invalid type (type type) for global var!");
1766 unsigned LinkageID = (VarType >> 2) & 7;
1767 bool isConstant = VarType & 1;
1768 bool hasInitializer = VarType & 2;
1769 GlobalValue::LinkageTypes Linkage;
1771 switch (LinkageID) {
1772 case 0: Linkage = GlobalValue::ExternalLinkage; break;
1773 case 1: Linkage = GlobalValue::WeakLinkage; break;
1774 case 2: Linkage = GlobalValue::AppendingLinkage; break;
1775 case 3: Linkage = GlobalValue::InternalLinkage; break;
1776 case 4: Linkage = GlobalValue::LinkOnceLinkage; break;
1778 error("Unknown linkage type: " + utostr(LinkageID));
1779 Linkage = GlobalValue::InternalLinkage;
1783 const Type *Ty = getType(SlotNo);
1785 error("Global has no type! SlotNo=" + utostr(SlotNo));
1788 if (!isa<PointerType>(Ty)) {
1789 error("Global not a pointer type! Ty= " + Ty->getDescription());
1792 const Type *ElTy = cast<PointerType>(Ty)->getElementType();
1794 // Create the global variable...
1795 GlobalVariable *GV = new GlobalVariable(ElTy, isConstant, Linkage,
1797 insertValue(GV, SlotNo, ModuleValues);
1799 unsigned initSlot = 0;
1800 if (hasInitializer) {
1801 initSlot = read_vbr_uint();
1802 GlobalInits.push_back(std::make_pair(GV, initSlot));
1805 // Notify handler about the global value.
1806 if (Handler) Handler->handleGlobalVariable(ElTy, isConstant, Linkage, SlotNo, initSlot);
1809 VarType = read_vbr_uint();
1812 // Read the function objects for all of the functions that are coming
1813 unsigned FnSignature = 0;
1814 if (read_typeid(FnSignature))
1815 error("Invalid function type (type type) found");
1817 while (FnSignature != Type::VoidTyID) { // List is terminated by Void
1818 const Type *Ty = getType(FnSignature);
1819 if (!isa<PointerType>(Ty) ||
1820 !isa<FunctionType>(cast<PointerType>(Ty)->getElementType())) {
1821 error("Function not a pointer to function type! Ty = " +
1822 Ty->getDescription());
1823 // FIXME: what should Ty be if handler continues?
1826 // We create functions by passing the underlying FunctionType to create...
1827 const FunctionType* FTy =
1828 cast<FunctionType>(cast<PointerType>(Ty)->getElementType());
1830 // Insert the place hodler
1831 Function* Func = new Function(FTy, GlobalValue::InternalLinkage,
1833 insertValue(Func, FnSignature, ModuleValues);
1835 // Save this for later so we know type of lazily instantiated functions
1836 FunctionSignatureList.push_back(Func);
1838 if (Handler) Handler->handleFunctionDeclaration(Func);
1840 // Get Next function signature
1841 if (read_typeid(FnSignature))
1842 error("Invalid function type (type type) found");
1845 // Now that the function signature list is set up, reverse it so that we can
1846 // remove elements efficiently from the back of the vector.
1847 std::reverse(FunctionSignatureList.begin(), FunctionSignatureList.end());
1849 // If this bytecode format has dependent library information in it ..
1850 if (!hasNoDependentLibraries) {
1851 // Read in the number of dependent library items that follow
1852 unsigned num_dep_libs = read_vbr_uint();
1853 std::string dep_lib;
1854 while( num_dep_libs-- ) {
1855 dep_lib = read_str();
1856 TheModule->linsert(dep_lib);
1859 // Read target triple and place into the module
1860 std::string triple = read_str();
1861 TheModule->setTargetTriple(triple);
1864 if (hasInconsistentModuleGlobalInfo)
1867 // This is for future proofing... in the future extra fields may be added that
1868 // we don't understand, so we transparently ignore them.
1872 if (Handler) Handler->handleModuleGlobalsEnd();
1875 /// Parse the version information and decode it by setting flags on the
1876 /// Reader that enable backward compatibility of the reader.
1877 void BytecodeReader::ParseVersionInfo() {
1878 unsigned Version = read_vbr_uint();
1880 // Unpack version number: low four bits are for flags, top bits = version
1881 Module::Endianness Endianness;
1882 Module::PointerSize PointerSize;
1883 Endianness = (Version & 1) ? Module::BigEndian : Module::LittleEndian;
1884 PointerSize = (Version & 2) ? Module::Pointer64 : Module::Pointer32;
1886 bool hasNoEndianness = Version & 4;
1887 bool hasNoPointerSize = Version & 8;
1889 RevisionNum = Version >> 4;
1891 // Default values for the current bytecode version
1892 hasInconsistentModuleGlobalInfo = false;
1893 hasExplicitPrimitiveZeros = false;
1894 hasRestrictedGEPTypes = false;
1895 hasTypeDerivedFromValue = false;
1896 hasLongBlockHeaders = false;
1897 hasPlatformSpecificFloatingPoint = false;
1898 has32BitTypes = false;
1899 hasNoDependentLibraries = false;
1901 switch (RevisionNum) {
1902 case 0: // LLVM 1.0, 1.1 release version
1903 // Base LLVM 1.0 bytecode format.
1904 hasInconsistentModuleGlobalInfo = true;
1905 hasExplicitPrimitiveZeros = true;
1909 case 1: // LLVM 1.2 release version
1910 // LLVM 1.2 added explicit support for emitting strings efficiently.
1912 // Also, it fixed the problem where the size of the ModuleGlobalInfo block
1913 // included the size for the alignment at the end, where the rest of the
1916 // LLVM 1.2 and before required that GEP indices be ubyte constants for
1917 // structures and longs for sequential types.
1918 hasRestrictedGEPTypes = true;
1920 // LLVM 1.2 and before had the Type class derive from Value class. This
1921 // changed in release 1.3 and consequently LLVM 1.3 bytecode files are
1922 // written differently because Types can no longer be part of the
1923 // type planes for Values.
1924 hasTypeDerivedFromValue = true;
1928 case 2: /// 1.2.5 (mid-release) version
1930 /// LLVM 1.2 and earlier had two-word block headers. This is a bit wasteful,
1931 /// especially for small files where the 8 bytes per block is a large fraction
1932 /// of the total block size. In LLVM 1.3, the block type and length are
1933 /// compressed into a single 32-bit unsigned integer. 27 bits for length, 5
1934 /// bits for block type.
1935 hasLongBlockHeaders = true;
1937 /// LLVM 1.2 and earlier wrote floating point values in a platform specific
1938 /// bit ordering. This was fixed in LLVM 1.3, but we still need to be backwards
1940 hasPlatformSpecificFloatingPoint = true;
1942 /// LLVM 1.2 and earlier wrote type slot numbers as vbr_uint32. In LLVM 1.3
1943 /// this has been reduced to vbr_uint24. It shouldn't make much difference
1944 /// since we haven't run into a module with > 24 million types, but for safety
1945 /// the 24-bit restriction has been enforced in 1.3 to free some bits in
1946 /// various places and to ensure consistency.
1947 has32BitTypes = true;
1949 /// LLVM 1.2 and earlier did not provide a target triple nor a list of
1950 /// libraries on which the bytecode is dependent. LLVM 1.3 provides these
1951 /// features, for use in future versions of LLVM.
1952 hasNoDependentLibraries = true;
1955 case 3: // LLVM 1.3 release version
1959 error("Unknown bytecode version number: " + itostr(RevisionNum));
1962 if (hasNoEndianness) Endianness = Module::AnyEndianness;
1963 if (hasNoPointerSize) PointerSize = Module::AnyPointerSize;
1965 TheModule->setEndianness(Endianness);
1966 TheModule->setPointerSize(PointerSize);
1968 if (Handler) Handler->handleVersionInfo(RevisionNum, Endianness, PointerSize);
1971 /// Parse a whole module.
1972 void BytecodeReader::ParseModule() {
1973 unsigned Type, Size;
1975 FunctionSignatureList.clear(); // Just in case...
1977 // Read into instance variables...
1981 bool SeenModuleGlobalInfo = false;
1982 bool SeenGlobalTypePlane = false;
1983 BufPtr MyEnd = BlockEnd;
1984 while (At < MyEnd) {
1986 read_block(Type, Size);
1990 case BytecodeFormat::GlobalTypePlaneBlockID:
1991 if (SeenGlobalTypePlane)
1992 error("Two GlobalTypePlane Blocks Encountered!");
1995 SeenGlobalTypePlane = true;
1998 case BytecodeFormat::ModuleGlobalInfoBlockID:
1999 if (SeenModuleGlobalInfo)
2000 error("Two ModuleGlobalInfo Blocks Encountered!");
2001 ParseModuleGlobalInfo();
2002 SeenModuleGlobalInfo = true;
2005 case BytecodeFormat::ConstantPoolBlockID:
2006 ParseConstantPool(ModuleValues, ModuleTypes,false);
2009 case BytecodeFormat::FunctionBlockID:
2010 ParseFunctionLazily();
2013 case BytecodeFormat::SymbolTableBlockID:
2014 ParseSymbolTable(0, &TheModule->getSymbolTable());
2020 error("Unexpected Block of Type #" + utostr(Type) + " encountered!");
2028 // After the module constant pool has been read, we can safely initialize
2029 // global variables...
2030 while (!GlobalInits.empty()) {
2031 GlobalVariable *GV = GlobalInits.back().first;
2032 unsigned Slot = GlobalInits.back().second;
2033 GlobalInits.pop_back();
2035 // Look up the initializer value...
2036 // FIXME: Preserve this type ID!
2038 const llvm::PointerType* GVType = GV->getType();
2039 unsigned TypeSlot = getTypeSlot(GVType->getElementType());
2040 if (Constant *CV = getConstantValue(TypeSlot, Slot)) {
2041 if (GV->hasInitializer())
2042 error("Global *already* has an initializer?!");
2043 if (Handler) Handler->handleGlobalInitializer(GV,CV);
2044 GV->setInitializer(CV);
2046 error("Cannot find initializer value.");
2049 /// Make sure we pulled them all out. If we didn't then there's a declaration
2050 /// but a missing body. That's not allowed.
2051 if (!FunctionSignatureList.empty())
2052 error("Function declared, but bytecode stream ended before definition");
2055 /// This function completely parses a bytecode buffer given by the \p Buf
2056 /// and \p Length parameters.
2057 void BytecodeReader::ParseBytecode(BufPtr Buf, unsigned Length,
2058 const std::string &ModuleID,
2059 bool processFunctions) {
2062 At = MemStart = BlockStart = Buf;
2063 MemEnd = BlockEnd = Buf + Length;
2065 // Create the module
2066 TheModule = new Module(ModuleID);
2068 if (Handler) Handler->handleStart(TheModule, Length);
2070 // Read and check signature...
2071 unsigned Sig = read_uint();
2072 if (Sig != ('l' | ('l' << 8) | ('v' << 16) | ('m' << 24))) {
2073 error("Invalid bytecode signature: " + utostr(Sig));
2076 // Tell the handler we're starting a module
2077 if (Handler) Handler->handleModuleBegin(ModuleID);
2079 // Get the module block and size and verify. This is handled specially
2080 // because the module block/size is always written in long format. Other
2081 // blocks are written in short format so the read_block method is used.
2082 unsigned Type, Size;
2085 if (Type != BytecodeFormat::ModuleBlockID) {
2086 error("Expected Module Block! Type:" + utostr(Type) + ", Size:"
2089 if (At + Size != MemEnd) {
2090 error("Invalid Top Level Block Length! Type:" + utostr(Type)
2091 + ", Size:" + utostr(Size));
2094 // Parse the module contents
2095 this->ParseModule();
2097 // Check for missing functions
2099 error("Function expected, but bytecode stream ended!");
2101 // Process all the function bodies now, if requested
2102 if (processFunctions)
2103 ParseAllFunctionBodies();
2105 // Tell the handler we're done with the module
2107 Handler->handleModuleEnd(ModuleID);
2109 // Tell the handler we're finished the parse
2110 if (Handler) Handler->handleFinish();
2112 } catch (std::string& errstr) {
2113 if (Handler) Handler->handleError(errstr);
2119 std::string msg("Unknown Exception Occurred");
2120 if (Handler) Handler->handleError(msg);
2128 //===----------------------------------------------------------------------===//
2129 //=== Default Implementations of Handler Methods
2130 //===----------------------------------------------------------------------===//
2132 BytecodeHandler::~BytecodeHandler() {}