aboutsummaryrefslogtreecommitdiff
path: root/lib/Bytecode/Analyzer/Parser.cpp
diff options
context:
space:
mode:
authorReid Spencer <rspencer@reidspencer.com>2004-06-10 08:09:13 +0000
committerReid Spencer <rspencer@reidspencer.com>2004-06-10 08:09:13 +0000
commit00c28a7481324a4804badbdbabb555c0d94e66f3 (patch)
tree92aa0212394ef44bc0a170cd319f3887db44ff30 /lib/Bytecode/Analyzer/Parser.cpp
parent43f38677f565a485e80e9254ad08e955e1bf0b73 (diff)
Implemented the bulk of the functionality. Cleaned up the code.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14113 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Bytecode/Analyzer/Parser.cpp')
-rw-r--r--lib/Bytecode/Analyzer/Parser.cpp510
1 files changed, 335 insertions, 175 deletions
diff --git a/lib/Bytecode/Analyzer/Parser.cpp b/lib/Bytecode/Analyzer/Parser.cpp
index 743db6c315..80800e75a6 100644
--- a/lib/Bytecode/Analyzer/Parser.cpp
+++ b/lib/Bytecode/Analyzer/Parser.cpp
@@ -1,4 +1,4 @@
-//===- Reader.cpp - Code to read bytecode files ---------------------------===//
+//===- Parser.cpp - Code to parse bytecode files --------------------------===//
//
// The LLVM Compiler Infrastructure
//
@@ -7,7 +7,7 @@
//
//===----------------------------------------------------------------------===//
//
-// This library implements the functionality defined in llvm/Bytecode/Reader.h
+// This library implements the functionality defined in llvm/Bytecode/Parser.h
//
// Note that this library should be as fast as possible, reentrant, and
// threadsafe!!
@@ -17,7 +17,6 @@
//===----------------------------------------------------------------------===//
#include "AnalyzerInternals.h"
-#include "ReaderPrimitives.h"
#include "llvm/Module.h"
#include "llvm/Bytecode/Format.h"
#include "Support/StringExtras.h"
@@ -37,40 +36,128 @@ using namespace llvm;
#define BCR_TRACE(n, X)
#endif
-#define PARSE_ERROR(inserters) \
- { \
+#define PARSE_ERROR(inserters) { \
std::ostringstream errormsg; \
errormsg << inserters; \
if ( ! handler->handleError( errormsg.str() ) ) \
throw std::string(errormsg.str()); \
}
+inline bool AbstractBytecodeParser::moreInBlock() {
+ return At < BlockEnd;
+}
-inline void AbstractBytecodeParser::readBlock(const unsigned char *&Buf,
- const unsigned char *EndBuf,
- unsigned &Type, unsigned &Size)
-{
- Type = read(Buf, EndBuf);
- Size = read(Buf, EndBuf);
+inline void AbstractBytecodeParser::checkPastBlockEnd(const char * block_name) {
+ if ( At > BlockEnd )
+ PARSE_ERROR("Attempt to read past the end of " << block_name << " block.");
}
-const Type *AbstractBytecodeParser::getType(unsigned ID) {
- //cerr << "Looking up Type ID: " << ID << "\n";
+inline void AbstractBytecodeParser::align32() {
+ BufPtr Save = At;
+ At = (const unsigned char *)((unsigned long)(At+3) & (~3UL));
+ if ( reportAlignment && At > Save ) handler->handleAlignment( At - Save );
+ if (At > BlockEnd)
+ throw std::string("Ran out of data while aligning!");
+}
+
+inline unsigned AbstractBytecodeParser::read_uint() {
+ if (At+4 > BlockEnd)
+ throw std::string("Ran out of data reading uint!");
+ At += 4;
+ return At[-4] | (At[-3] << 8) | (At[-2] << 16) | (At[-1] << 24);
+}
+
+inline unsigned AbstractBytecodeParser::read_vbr_uint() {
+ unsigned Shift = 0;
+ unsigned Result = 0;
+ BufPtr Save = At;
+
+ do {
+ if (At == BlockEnd)
+ throw std::string("Ran out of data reading vbr_uint!");
+ Result |= (unsigned)((*At++) & 0x7F) << Shift;
+ Shift += 7;
+ } while (At[-1] & 0x80);
+ if (reportVBR)
+ handler->handleVBR32(At-Save);
+ return Result;
+}
+
+inline uint64_t AbstractBytecodeParser::read_vbr_uint64() {
+ unsigned Shift = 0;
+ uint64_t Result = 0;
+ BufPtr Save = At;
+
+ do {
+ if (At == BlockEnd)
+ throw std::string("Ran out of data reading vbr_uint64!");
+ Result |= (uint64_t)((*At++) & 0x7F) << Shift;
+ Shift += 7;
+ } while (At[-1] & 0x80);
+ if (reportVBR)
+ handler->handleVBR64(At-Save);
+ return Result;
+}
- if (ID < Type::FirstDerivedTyID)
- if (const Type *T = Type::getPrimitiveType((Type::PrimitiveID)ID))
- return T; // Asked for a primitive type...
+inline int64_t AbstractBytecodeParser::read_vbr_int64() {
+ uint64_t R = read_vbr_uint64();
+ if (R & 1) {
+ if (R != 1)
+ return -(int64_t)(R >> 1);
+ else // There is no such thing as -0 with integers. "-0" really means
+ // 0x8000000000000000.
+ return 1LL << 63;
+ } else
+ return (int64_t)(R >> 1);
+}
- // Otherwise, derived types need offset...
- ID -= Type::FirstDerivedTyID;
+inline std::string AbstractBytecodeParser::read_str() {
+ unsigned Size = read_vbr_uint();
+ const unsigned char *OldAt = At;
+ At += Size;
+ if (At > BlockEnd) // Size invalid?
+ throw std::string("Ran out of data reading a string!");
+ return std::string((char*)OldAt, Size);
+}
+
+inline void AbstractBytecodeParser::read_data(void *Ptr, void *End) {
+ unsigned char *Start = (unsigned char *)Ptr;
+ unsigned Amount = (unsigned char *)End - Start;
+ if (At+Amount > BlockEnd)
+ throw std::string("Ran out of data!");
+ std::copy(At, At+Amount, Start);
+ At += Amount;
+}
- if (!CompactionTypeTable.empty()) {
- if (ID >= CompactionTypeTable.size())
- PARSE_ERROR("Type ID out of range for compaction table!");
- return CompactionTypeTable[ID];
+inline void AbstractBytecodeParser::readBlock(unsigned &Type, unsigned &Size) {
+ Type = read_uint();
+ Size = read_uint();
+ BlockStart = At;
+ if ( At + Size > BlockEnd )
+ throw std::string("Attempt to size a block past end of memory");
+ BlockEnd = At + Size;
+ if ( reportBlocks ) {
+ handler->handleBlock( Type, BlockStart, Size );
}
+}
+
+const Type *AbstractBytecodeParser::getType(unsigned ID) {
+//cerr << "Looking up Type ID: " << ID << "\n";
+
+if (ID < Type::FirstDerivedTyID)
+ if (const Type *T = Type::getPrimitiveType((Type::PrimitiveID)ID))
+ return T; // Asked for a primitive type...
+
+// Otherwise, derived types need offset...
+ID -= Type::FirstDerivedTyID;
- // Is it a module-level type?
+if (!CompactionTypeTable.empty()) {
+ if (ID >= CompactionTypeTable.size())
+ PARSE_ERROR("Type ID out of range for compaction table!");
+ return CompactionTypeTable[ID];
+}
+
+// Is it a module-level type?
if (ID < ModuleTypes.size())
return ModuleTypes[ID].get();
@@ -83,12 +170,12 @@ const Type *AbstractBytecodeParser::getType(unsigned ID) {
return Type::VoidTy;
}
-bool AbstractBytecodeParser::ParseInstruction(BufPtr& Buf, BufPtr EndBuf,
- std::vector<unsigned> &Operands) {
+bool AbstractBytecodeParser::ParseInstruction(std::vector<unsigned> &Operands) {
+ BufPtr SaveAt = At;
Operands.clear();
unsigned iType = 0;
unsigned Opcode = 0;
- unsigned Op = read(Buf, EndBuf);
+ unsigned Op = read_uint();
// bits Instruction format: Common to all formats
// --------------------------
@@ -134,61 +221,56 @@ bool AbstractBytecodeParser::ParseInstruction(BufPtr& Buf, BufPtr EndBuf,
Operands[2] = (Op >> 26) & 63;
break;
case 0:
- Buf -= 4; // Hrm, try this again...
- Opcode = read_vbr_uint(Buf, EndBuf);
+ At -= 4; // Hrm, try this again...
+ Opcode = read_vbr_uint();
Opcode >>= 2;
- iType = read_vbr_uint(Buf, EndBuf);
+ iType = read_vbr_uint();
- unsigned NumOperands = read_vbr_uint(Buf, EndBuf);
+ unsigned NumOperands = read_vbr_uint();
Operands.resize(NumOperands);
if (NumOperands == 0)
PARSE_ERROR("Zero-argument instruction found; this is invalid.");
for (unsigned i = 0; i != NumOperands; ++i)
- Operands[i] = read_vbr_uint(Buf, EndBuf);
- align32(Buf, EndBuf);
+ Operands[i] = read_vbr_uint();
+ align32();
break;
}
- return handler->handleInstruction(Opcode, getType(iType), Operands);
+ return handler->handleInstruction(Opcode, getType(iType), Operands, At-SaveAt);
}
/// ParseBasicBlock - In LLVM 1.0 bytecode files, we used to output one
/// basicblock at a time. This method reads in one of the basicblock packets.
-void AbstractBytecodeParser::ParseBasicBlock(BufPtr &Buf,
- BufPtr EndBuf,
- unsigned BlockNo) {
+void AbstractBytecodeParser::ParseBasicBlock( unsigned BlockNo) {
handler->handleBasicBlockBegin( BlockNo );
std::vector<unsigned> Args;
bool is_terminating = false;
- while (Buf < EndBuf)
- is_terminating = ParseInstruction(Buf, EndBuf, Args);
+ while ( moreInBlock() )
+ is_terminating = ParseInstruction(Args);
if ( ! is_terminating )
- PARSE_ERROR(
- "Failed to recognize instruction as terminating at end of block");
+ PARSE_ERROR("Non-terminated basic block found!");
handler->handleBasicBlockEnd( BlockNo );
}
-
/// ParseInstructionList - Parse all of the BasicBlock's & Instruction's in the
/// body of a function. In post 1.0 bytecode files, we no longer emit basic
/// block individually, in order to avoid per-basic-block overhead.
-unsigned AbstractBytecodeParser::ParseInstructionList( BufPtr &Buf,
- BufPtr EndBuf) {
+unsigned AbstractBytecodeParser::ParseInstructionList() {
unsigned BlockNo = 0;
std::vector<unsigned> Args;
- while (Buf < EndBuf) {
+ while ( moreInBlock() ) {
handler->handleBasicBlockBegin( BlockNo );
// Read instructions into this basic block until we get to a terminator
bool is_terminating = false;
- while (Buf < EndBuf && !is_terminating )
- is_terminating = ParseInstruction(Buf, EndBuf, Args ) ;
+ while (moreInBlock() && !is_terminating )
+ is_terminating = ParseInstruction(Args ) ;
if (!is_terminating)
PARSE_ERROR( "Non-terminated basic block found!");
@@ -199,36 +281,34 @@ unsigned AbstractBytecodeParser::ParseInstructionList( BufPtr &Buf,
return BlockNo;
}
-void AbstractBytecodeParser::ParseSymbolTable(BufPtr &Buf, BufPtr EndBuf) {
+void AbstractBytecodeParser::ParseSymbolTable() {
handler->handleSymbolTableBegin();
- while (Buf < EndBuf) {
+ while ( moreInBlock() ) {
// Symtab block header: [num entries][type id number]
- unsigned NumEntries = read_vbr_uint(Buf, EndBuf);
- unsigned Typ = read_vbr_uint(Buf, EndBuf);
+ unsigned NumEntries = read_vbr_uint();
+ unsigned Typ = read_vbr_uint();
const Type *Ty = getType(Typ);
handler->handleSymbolTablePlane( Typ, NumEntries, Ty );
for (unsigned i = 0; i != NumEntries; ++i) {
// Symtab entry: [def slot #][name]
- unsigned slot = read_vbr_uint(Buf, EndBuf);
- std::string Name = read_str(Buf, EndBuf);
+ unsigned slot = read_vbr_uint();
+ std::string Name = read_str();
if (Typ == Type::TypeTyID)
handler->handleSymbolTableType( i, slot, Name );
else
- handler->handleSymbolTableValue( i, slot, Name );
+ handler->handleSymbolTableValue( i, slot, Name );
}
}
-
- if (Buf > EndBuf)
- PARSE_ERROR("Tried to read past end of buffer while reading symbol table.");
+ checkPastBlockEnd("Symbol Table");
handler->handleSymbolTableEnd();
}
-void AbstractBytecodeParser::ParseFunctionLazily(BufPtr &Buf, BufPtr EndBuf) {
+void AbstractBytecodeParser::ParseFunctionLazily() {
if (FunctionSignatureList.empty())
throw std::string("FunctionSignatureList empty!");
@@ -236,9 +316,10 @@ void AbstractBytecodeParser::ParseFunctionLazily(BufPtr &Buf, BufPtr EndBuf) {
FunctionSignatureList.pop_back();
// Save the information for future reading of the function
- LazyFunctionLoadMap[FType] = LazyFunctionInfo(Buf, EndBuf);
+ LazyFunctionLoadMap[FType] = LazyFunctionInfo(BlockStart, BlockEnd);
+
// Pretend we've `parsed' this function
- Buf = EndBuf;
+ At = BlockEnd;
}
void AbstractBytecodeParser::ParseNextFunction(Type* FType) {
@@ -251,21 +332,20 @@ void AbstractBytecodeParser::ParseNextFunction(Type* FType) {
return;
}
- BufPtr Buf = Fi->second.Buf;
- BufPtr EndBuf = Fi->second.EndBuf;
+ BlockStart = At = Fi->second.Buf;
+ BlockEnd = Fi->second.Buf;
assert(Fi->first == FType);
LazyFunctionLoadMap.erase(Fi);
- this->ParseFunctionBody( FType, Buf, EndBuf );
+ this->ParseFunctionBody( FType );
}
-void AbstractBytecodeParser::ParseFunctionBody(const Type* FType,
- BufPtr &Buf, BufPtr EndBuf ) {
+void AbstractBytecodeParser::ParseFunctionBody(const Type* FType ) {
GlobalValue::LinkageTypes Linkage = GlobalValue::ExternalLinkage;
- unsigned LinkageType = read_vbr_uint(Buf, EndBuf);
+ unsigned LinkageType = read_vbr_uint();
switch (LinkageType) {
case 0: Linkage = GlobalValue::ExternalLinkage; break;
case 1: Linkage = GlobalValue::WeakLinkage; break;
@@ -284,43 +364,45 @@ void AbstractBytecodeParser::ParseFunctionBody(const Type* FType,
unsigned BlockNum = 0;
bool InsertedArguments = false;
- while (Buf < EndBuf) {
+ BufPtr MyEnd = BlockEnd;
+ while ( At < MyEnd ) {
unsigned Type, Size;
- BufPtr OldBuf = Buf;
- readBlock(Buf, EndBuf, Type, Size);
+ BufPtr OldAt = At;
+ readBlock(Type, Size);
switch (Type) {
case BytecodeFormat::ConstantPool:
- ParseConstantPool(Buf, Buf+Size, FunctionTypes );
+ ParseConstantPool(FunctionTypes );
break;
case BytecodeFormat::CompactionTable:
- ParseCompactionTable(Buf, Buf+Size);
+ ParseCompactionTable();
break;
case BytecodeFormat::BasicBlock:
- ParseBasicBlock(Buf, Buf+Size, BlockNum++);
+ ParseBasicBlock(BlockNum++);
break;
case BytecodeFormat::InstructionList:
if (BlockNum)
- PARSE_ERROR("InstructionList must come before basic blocks!");
- BlockNum = ParseInstructionList(Buf, Buf+Size);
+ PARSE_ERROR("InstructionList must come before basic blocks!");
+ BlockNum = ParseInstructionList();
break;
case BytecodeFormat::SymbolTable:
- ParseSymbolTable(Buf, Buf+Size );
+ ParseSymbolTable();
break;
default:
- Buf += Size;
- if (OldBuf > Buf)
- PARSE_ERROR("Wrapped around reading bytecode");
+ At += Size;
+ if (OldAt > At)
+ PARSE_ERROR("Wrapped around reading bytecode");
break;
}
+ BlockEnd = MyEnd;
// Malformed bc file if read past end of block.
- align32(Buf, EndBuf);
+ align32();
}
handler->handleFunctionEnd(FType);
@@ -336,21 +418,24 @@ void AbstractBytecodeParser::ParseAllFunctionBodies() {
while ( Fi != Fe ) {
const Type* FType = Fi->first;
- this->ParseFunctionBody(FType, Fi->second.Buf, Fi->second.EndBuf);
+ BlockStart = At = Fi->second.Buf;
+ BlockEnd = Fi->second.EndBuf;
+ this->ParseFunctionBody(FType);
+ ++Fi;
}
}
-void AbstractBytecodeParser::ParseCompactionTable(BufPtr &Buf, BufPtr End) {
+void AbstractBytecodeParser::ParseCompactionTable() {
handler->handleCompactionTableBegin();
- while (Buf != End) {
- unsigned NumEntries = read_vbr_uint(Buf, End);
+ while ( moreInBlock() ) {
+ unsigned NumEntries = read_vbr_uint();
unsigned Ty;
if ((NumEntries & 3) == 3) {
NumEntries >>= 2;
- Ty = read_vbr_uint(Buf, End);
+ Ty = read_vbr_uint();
} else {
Ty = NumEntries >> 2;
NumEntries &= 3;
@@ -360,25 +445,24 @@ void AbstractBytecodeParser::ParseCompactionTable(BufPtr &Buf, BufPtr End) {
if (Ty == Type::TypeTyID) {
for (unsigned i = 0; i != NumEntries; ++i) {
- unsigned TypeSlot = read_vbr_uint(Buf,End);
+ unsigned TypeSlot = read_vbr_uint();
const Type *Typ = getGlobalTableType(TypeSlot);
- handler->handleCompactionTableType( i, TypeSlot, Typ );
+ handler->handleCompactionTableType( i, TypeSlot, Typ );
}
} else {
const Type *Typ = getType(Ty);
// Push the implicit zero
for (unsigned i = 0; i != NumEntries; ++i) {
- unsigned ValSlot = read_vbr_uint(Buf, End);
- handler->handleCompactionTableValue( i, ValSlot, Typ );
+ unsigned ValSlot = read_vbr_uint();
+ handler->handleCompactionTableValue( i, ValSlot, Typ );
}
}
}
handler->handleCompactionTableEnd();
}
-const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
- const unsigned char *EndBuf) {
- unsigned PrimType = read_vbr_uint(Buf, EndBuf);
+const Type *AbstractBytecodeParser::ParseTypeConstant() {
+ unsigned PrimType = read_vbr_uint();
const Type *Val = 0;
if ((Val = Type::getPrimitiveType((Type::PrimitiveID)PrimType)))
@@ -386,13 +470,13 @@ const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
switch (PrimType) {
case Type::FunctionTyID: {
- const Type *RetType = getType(read_vbr_uint(Buf, EndBuf));
+ const Type *RetType = getType(read_vbr_uint());
- unsigned NumParams = read_vbr_uint(Buf, EndBuf);
+ unsigned NumParams = read_vbr_uint();
std::vector<const Type*> Params;
while (NumParams--)
- Params.push_back(getType(read_vbr_uint(Buf, EndBuf)));
+ Params.push_back(getType(read_vbr_uint()));
bool isVarArg = Params.size() && Params.back() == Type::VoidTy;
if (isVarArg) Params.pop_back();
@@ -402,10 +486,10 @@ const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
return result;
}
case Type::ArrayTyID: {
- unsigned ElTyp = read_vbr_uint(Buf, EndBuf);
+ unsigned ElTyp = read_vbr_uint();
const Type *ElementType = getType(ElTyp);
- unsigned NumElements = read_vbr_uint(Buf, EndBuf);
+ unsigned NumElements = read_vbr_uint();
BCR_TRACE(5, "Array Type Constant #" << ElTyp << " size="
<< NumElements << "\n");
@@ -415,10 +499,10 @@ const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
}
case Type::StructTyID: {
std::vector<const Type*> Elements;
- unsigned Typ = read_vbr_uint(Buf, EndBuf);
+ unsigned Typ = read_vbr_uint();
while (Typ) { // List is terminated by void/0 typeid
Elements.push_back(getType(Typ));
- Typ = read_vbr_uint(Buf, EndBuf);
+ Typ = read_vbr_uint();
}
Type* result = StructType::get(Elements);
@@ -426,7 +510,7 @@ const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
return result;
}
case Type::PointerTyID: {
- unsigned ElTyp = read_vbr_uint(Buf, EndBuf);
+ unsigned ElTyp = read_vbr_uint();
BCR_TRACE(5, "Pointer Type Constant #" << ElTyp << "\n");
Type* result = PointerType::get(getType(ElTyp));
handler->handleType( result );
@@ -455,10 +539,9 @@ const Type *AbstractBytecodeParser::ParseTypeConstant(const unsigned char *&Buf,
// something and when we reread the type later, we can replace the opaque type
// with a new resolved concrete type.
//
-void AbstractBytecodeParser::ParseTypeConstants(const unsigned char *&Buf,
- const unsigned char *EndBuf,
- TypeListTy &Tab,
- unsigned NumEntries) {
+void AbstractBytecodeParser::ParseTypeConstants(
+ TypeListTy &Tab, unsigned NumEntries
+) {
assert(Tab.size() == 0 && "should not have read type constants in before!");
// Insert a bunch of opaque types to be resolved later...
@@ -470,7 +553,7 @@ void AbstractBytecodeParser::ParseTypeConstants(const unsigned char *&Buf,
// opaque types just inserted.
//
for (unsigned i = 0; i != NumEntries; ++i) {
- const Type *NewTy = ParseTypeConstant(Buf, EndBuf), *OldTy = Tab[i].get();
+ const Type *NewTy = ParseTypeConstant(), *OldTy = Tab[i].get();
if (NewTy == 0) throw std::string("Couldn't parse type!");
BCR_TRACE(4, "#" << i << ": Read Type Constant: '" << NewTy <<
"' Replacing: " << OldTy << "\n");
@@ -497,18 +580,16 @@ void AbstractBytecodeParser::ParseTypeConstants(const unsigned char *&Buf,
}
-void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
- const unsigned char *EndBuf,
- unsigned TypeID) {
+void AbstractBytecodeParser::ParseConstantValue(unsigned TypeID) {
// We must check for a ConstantExpr before switching by type because
// a ConstantExpr can be of any type, and has no explicit value.
//
// 0 if not expr; numArgs if is expr
- unsigned isExprNumArgs = read_vbr_uint(Buf, EndBuf);
+ unsigned isExprNumArgs = read_vbr_uint();
if (isExprNumArgs) {
- unsigned Opcode = read_vbr_uint(Buf, EndBuf);
+ unsigned Opcode = read_vbr_uint();
const Type* Typ = getType(TypeID);
// FIXME: Encoding of constant exprs could be much more compact!
@@ -517,8 +598,8 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
// Read the slot number and types of each of the arguments
for (unsigned i = 0; i != isExprNumArgs; ++i) {
- unsigned ArgValSlot = read_vbr_uint(Buf, EndBuf);
- unsigned ArgTypeSlot = read_vbr_uint(Buf, EndBuf);
+ unsigned ArgValSlot = read_vbr_uint();
+ unsigned ArgTypeSlot = read_vbr_uint();
BCR_TRACE(4, "CE Arg " << i << ": Type: '" << *getType(ArgTypeSlot)
<< "' slot: " << ArgValSlot << "\n");
@@ -534,7 +615,7 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
const Type *Ty = getType(TypeID);
switch (Ty->getPrimitiveID()) {
case Type::BoolTyID: {
- unsigned Val = read_vbr_uint(Buf, EndBuf);
+ unsigned Val = read_vbr_uint();
if (Val != 0 && Val != 1)
PARSE_ERROR("Invalid boolean value read.");
@@ -545,7 +626,7 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
case Type::UByteTyID: // Unsigned integer types...
case Type::UShortTyID:
case Type::UIntTyID: {
- unsigned Val = read_vbr_uint(Buf, EndBuf);
+ unsigned Val = read_vbr_uint();
if (!ConstantUInt::isValueValidForType(Ty, Val))
throw std::string("Invalid unsigned byte/short/int read.");
handler->handleConstantValue( ConstantUInt::get(Ty, Val) );
@@ -553,7 +634,7 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
}
case Type::ULongTyID: {
- handler->handleConstantValue( ConstantUInt::get(Ty, read_vbr_uint64(Buf, EndBuf)) );
+ handler->handleConstantValue( ConstantUInt::get(Ty, read_vbr_uint64()) );
break;
}
@@ -561,7 +642,7 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
case Type::ShortTyID:
case Type::IntTyID: {
case Type::LongTyID:
- int64_t Val = read_vbr_int64(Buf, EndBuf);
+ int64_t Val = read_vbr_int64();
if (!ConstantSInt::isValueValidForType(Ty, Val))
throw std::string("Invalid signed byte/short/int/long read.");
handler->handleConstantValue( ConstantSInt::get(Ty, Val) );
@@ -570,14 +651,14 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
case Type::FloatTyID: {
float F;
- input_data(Buf, EndBuf, &F, &F+1);
+ read_data(&F, &F+1);
handler->handleConstantValue( ConstantFP::get(Ty, F) );
break;
}
case Type::DoubleTyID: {
double Val;
- input_data(Buf, EndBuf, &Val, &Val+1);
+ read_data(&Val, &Val+1);
handler->handleConstantValue( ConstantFP::get(Ty, Val) );
break;
}
@@ -592,7 +673,7 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
std::vector<unsigned> Elements;
Elements.reserve(NumElements);
while (NumElements--) // Read all of the elements of the constant.
- Elements.push_back(read_vbr_uint(Buf, EndBuf));
+ Elements.push_back(read_vbr_uint());
handler->handleConstantArray( AT, Elements );
break;
@@ -603,15 +684,16 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
std::vector<unsigned> Elements;
Elements.reserve(ST->getNumElements());
for (unsigned i = 0; i != ST->getNumElements(); ++i)
- Elements.push_back(read_vbr_uint(Buf, EndBuf));
-
+ Elements.push_back(read_vbr_uint());
handler->handleConstantStruct( ST, Elements );
+ break;
}
case Type::PointerTyID: { // ConstantPointerRef value...
const PointerType *PT = cast<PointerType>(Ty);
- unsigned Slot = read_vbr_uint(Buf, EndBuf);
+ unsigned Slot = read_vbr_uint();
handler->handleConstantPointer( PT, Slot );
+ break;
}
default:
@@ -620,16 +702,13 @@ void AbstractBytecodeParser::ParseConstantValue(const unsigned char *&Buf,
}
}
-void AbstractBytecodeParser::ParseGlobalTypes(const unsigned char *&Buf,
- const unsigned char *EndBuf) {
- ParseConstantPool(Buf, EndBuf, ModuleTypes);
+void AbstractBytecodeParser::ParseGlobalTypes() {
+ ParseConstantPool(ModuleTypes);
}
-void AbstractBytecodeParser::ParseStringConstants(const unsigned char *&Buf,
- const unsigned char *EndBuf,
- unsigned NumEntries ){
+void AbstractBytecodeParser::ParseStringConstants(unsigned NumEntries ){
for (; NumEntries; --NumEntries) {
- unsigned Typ = read_vbr_uint(Buf, EndBuf);
+ unsigned Typ = read_vbr_uint();
const Type *Ty = getType(Typ);
if (!isa<ArrayType>(Ty))
throw std::string("String constant data invalid!");
@@ -641,7 +720,7 @@ void AbstractBytecodeParser::ParseStringConstants(const unsigned char *&Buf,
// Read character data. The type tells us how long the string is.
char Data[ATy->getNumElements()];
- input_data(Buf, EndBuf, Data, Data+ATy->getNumElements());
+ read_data(Data, Data+ATy->getNumElements());
std::vector<Constant*> Elements(ATy->getNumElements());
if (ATy->getElementType() == Type::SByteTy)
@@ -658,35 +737,33 @@ void AbstractBytecodeParser::ParseStringConstants(const unsigned char *&Buf,
}
-void AbstractBytecodeParser::ParseConstantPool(const unsigned char *&Buf,
- const unsigned char *EndBuf,
- TypeListTy &TypeTab) {
- while (Buf < EndBuf) {
- unsigned NumEntries = read_vbr_uint(Buf, EndBuf);
- unsigned Typ = read_vbr_uint(Buf, EndBuf);
+void AbstractBytecodeParser::ParseConstantPool( TypeListTy &TypeTab) {
+ while ( moreInBlock() ) {
+ unsigned NumEntries = read_vbr_uint();
+ unsigned Typ = read_vbr_uint();
if (Typ == Type::TypeTyID) {
- ParseTypeConstants(Buf, EndBuf, TypeTab, NumEntries);
+ ParseTypeConstants(TypeTab, NumEntries);
} else if (Typ == Type::VoidTyID) {
- ParseStringConstants(Buf, EndBuf, NumEntries);
+ ParseStringConstants(NumEntries);
} else {
BCR_TRACE(3, "Type: '" << *getType(Typ) << "' NumEntries: "
<< NumEntries << "\n");
for (unsigned i = 0; i < NumEntries; ++i) {
- ParseConstantValue(Buf, EndBuf, Typ);
+ ParseConstantValue(Typ);
}
}
}
- if (Buf > EndBuf) PARSE_ERROR("Read past end of buffer.");
+ checkPastBlockEnd("Constant Pool");
}
-void AbstractBytecodeParser::ParseModuleGlobalInfo(BufPtr &Buf, BufPtr End) {
+void AbstractBytecodeParser::ParseModuleGlobalInfo() {
handler->handleModuleGlobalsBegin();
// Read global variables...
- unsigned VarType = read_vbr_uint(Buf, End);
+ unsigned VarType = read_vbr_uint();
while (VarType != Type::VoidTyID) { // List is terminated by Void
// VarType Fields: bit0 = isConstant, bit1 = hasInitializer, bit2,3,4 =
// Linkage, bit4+ = slot#
@@ -721,17 +798,17 @@ void AbstractBytecodeParser::ParseModuleGlobalInfo(BufPtr &Buf, BufPtr End) {
// Create the global variable...
if (hasInitializer) {
- unsigned initSlot = read_vbr_uint(Buf,End);
+ unsigned initSlot = read_vbr_uint();
handler->handleInitializedGV( ElTy, isConstant, Linkage, initSlot );
} else
handler->handleGlobalVariable( ElTy, isConstant, Linkage );
// Get next item
- VarType = read_vbr_uint(Buf, End);
+ VarType = read_vbr_uint();
}
// Read the function objects for all of the functions that are coming
- unsigned FnSignature = read_vbr_uint(Buf, End);
+ unsigned FnSignature = read_vbr_uint();
while (FnSignature != Type::VoidTyID) { // List is terminated by Void
const Type *Ty = getType(FnSignature);
if (!isa<PointerType>(Ty) ||
@@ -750,22 +827,26 @@ void AbstractBytecodeParser::ParseModuleGlobalInfo(BufPtr &Buf, BufPtr End) {
handler->handleFunctionDeclaration(Ty);
// Get Next function signature
- FnSignature = read_vbr_uint(Buf, End);
+ FnSignature = read_vbr_uint();
}
if (hasInconsistentModuleGlobalInfo)
- align32(Buf, End);
+ align32();
+
+ // Now that the function signature list is set up, reverse it so that we can
+ // remove elements efficiently from the back of the vector.
+ std::reverse(FunctionSignatureList.begin(), FunctionSignatureList.end());
// This is for future proofing... in the future extra fields may be added that
// we don't understand, so we transparently ignore them.
//
- Buf = End;
+ At = BlockEnd;
handler->handleModuleGlobalsEnd();
}
-void AbstractBytecodeParser::ParseVersionInfo(BufPtr &Buf, BufPtr EndBuf) {
- unsigned Version = read_vbr_uint(Buf, EndBuf);
+void AbstractBytecodeParser::ParseVersionInfo() {
+ unsigned Version = read_vbr_uint();
// Unpack version number: low four bits are for flags, top bits = version
Module::Endianness Endianness;
@@ -814,85 +895,164 @@ void AbstractBytecodeParser::ParseVersionInfo(BufPtr &Buf, BufPtr EndBuf) {
handler->handleVersionInfo(RevisionNum, Endianness, PointerSize );
}
-void AbstractBytecodeParser::ParseModule(BufPtr &Buf, BufPtr EndBuf ) {
+void AbstractBytecodeParser::ParseModule() {
unsigned Type, Size;
- readBlock(Buf, EndBuf, Type, Size);
- if (Type != BytecodeFormat::Module || Buf+Size != EndBuf)
- // Hrm, not a class?
- PARSE_ERROR("Expected Module block! B: " << unsigned(intptr_t(Buf)) <<
- ", S: " << Size << " E: " << unsigned(intptr_t(EndBuf)));
+
+ FunctionSignatureList.clear(); // Just in case...
// Read into instance variables...
- ParseVersionInfo(Buf, EndBuf);
- align32(Buf, EndBuf);
+ ParseVersionInfo();
+ align32(); /// FIXME: Is this redundant? VI is first and 4 bytes!
bool SeenModuleGlobalInfo = false;
bool SeenGlobalTypePlane = false;
- while (Buf < EndBuf) {
- BufPtr OldBuf = Buf;
- readBlock(Buf, EndBuf, Type, Size);
+ BufPtr MyEnd = BlockEnd;
+ while (At < MyEnd) {
+ BufPtr OldAt = At;
+ readBlock(Type, Size);
switch (Type) {
case BytecodeFormat::GlobalTypePlane:
if ( SeenGlobalTypePlane )
- PARSE_ERROR("Two GlobalTypePlane Blocks Encountered!");
+ PARSE_ERROR("Two GlobalTypePlane Blocks Encountered!");
- ParseGlobalTypes(Buf, Buf+Size);
+ ParseGlobalTypes();
SeenGlobalTypePlane = true;
break;
case BytecodeFormat::ModuleGlobalInfo:
if ( SeenModuleGlobalInfo )
- PARSE_ERROR("Two ModuleGlobalInfo Blocks Encountered!");
- ParseModuleGlobalInfo(Buf, Buf+Size);
+ PARSE_ERROR("Two ModuleGlobalInfo Blocks Encountered!");
+ ParseModuleGlobalInfo();
SeenModuleGlobalInfo = true;
break;
case BytecodeFormat::ConstantPool:
- ParseConstantPool(Buf, Buf+Size, ModuleTypes);
+ ParseConstantPool(ModuleTypes);
break;
case BytecodeFormat::Function:
- ParseFunctionLazily(Buf, Buf+Size);
+ ParseFunctionLazily();
break;
case BytecodeFormat::SymbolTable:
- ParseSymbolTable(Buf, Buf+Size );
+ ParseSymbolTable();
break;
default:
- Buf += Size;
- if (OldBuf > Buf)
- {
- PARSE_ERROR("Unexpected Block of Type" << Type << "encountered!" );
+ At += Size;
+ if (OldAt > At) {
+ PARSE_ERROR("Unexpected Block of Type" << Type << "encountered!" );
}
break;
}
- align32(Buf, EndBuf);
+ BlockEnd = MyEnd;
+ align32();
}
+
+ /// Make sure we pulled them all out. If we didn't then there's a declaration
+ /// but a missing body. That's not allowed.
+ if (!FunctionSignatureList.empty())
+ throw std::string(
+ "Function declared, but bytecode stream ended before definition");
}
void AbstractBytecodeParser::ParseBytecode(
- BufPtr Buf, unsigned Length,
+ BufPtr b, unsigned Length,
const std::string &ModuleID) {
+ At = MemStart = BlockStart = b;
+ MemEnd = BlockEnd = b + Length;
handler->handleStart();
- unsigned char *EndBuf = (unsigned char*)(Buf + Length);
// Read and check signature...
- unsigned Sig = read(Buf, EndBuf);
+ unsigned Sig = read_uint();
if (Sig != ('l' | ('l' << 8) | ('v' << 16) | ('m' << 24))) {
PARSE_ERROR("Invalid bytecode signature: " << Sig);
}
handler->handleModuleBegin(ModuleID);
- this->ParseModule(Buf, EndBuf);
+ unsigned Type, Size;
+ readBlock(Type, Size);
+ if ( Type != BytecodeFormat::Module ) {
+ PARSE_ERROR("Expected Module Block! At: " << unsigned(intptr_t(At))
+ << ", Type:" << Type << ", Size:" << Size);
+ }
+ if ( At + Size != MemEnd ) {
+ PARSE_ERROR("Invalid Top Level Block Length! At: "
+ << unsigned(intptr_t(At)) << ", Type:" << Type << ", Size:" << Size);
+ }
+ this->ParseModule();
handler->handleModuleEnd(ModuleID);
handler->handleFinish();
}
+//===----------------------------------------------------------------------===//
+//=== Default Implementations of Handler Methods
+//===----------------------------------------------------------------------===//
+
+bool BytecodeHandler::handleError(const std::string& str ) { return false; }
+void BytecodeHandler::handleStart() { }
+void BytecodeHandler::handleFinish() { }
+void BytecodeHandler::handleModuleBegin(const std::string& id) { }
+void BytecodeHandler::handleModuleEnd(const std::string& id) { }
+void BytecodeHandler::handleVersionInfo( unsigned char RevisionNum,
+ Module::Endianness Endianness, Module::PointerSize PointerSize) { }
+void BytecodeHandler::handleModuleGlobalsBegin() { }
+void BytecodeHandler::handleGlobalVariable(
+ const Type* ElemType, bool isConstant, GlobalValue::LinkageTypes ) { }
+void BytecodeHandler::handleInitializedGV(
+ const Type* ElemType, bool isConstant, GlobalValue::LinkageTypes,
+ unsigned initSlot) {}
+void BytecodeHandler::handleType( const Type* Ty ) {}
+void BytecodeHandler::handleFunctionDeclaration(
+ const Type* FuncType) {}
+void BytecodeHandler::handleModuleGlobalsEnd() { }
+void BytecodeHandler::handleCompactionTableBegin() { }
+void BytecodeHandler::handleCompactionTablePlane( unsigned Ty,
+ unsigned NumEntries) {}
+void BytecodeHandler::handleCompactionTableType( unsigned i, unsigned TypSlot,
+ const Type* ) {}
+void BytecodeHandler::handleCompactionTableValue( unsigned i, unsigned ValSlot,
+ const Type* ) {}
+void BytecodeHandler::handleCompactionTableEnd() { }
+void BytecodeHandler::handleSymbolTableBegin() { }
+void BytecodeHandler::handleSymbolTablePlane( unsigned Ty, unsigned NumEntries,
+ const Type* Typ) { }
+void BytecodeHandler::handleSymbolTableType( unsigned i, unsigned slot,
+ const std::string& name ) { }
+void BytecodeHandler::handleSymbolTableValue( unsigned i, unsigned slot,
+ const std::string& name ) { }
+void BytecodeHandler::handleSymbolTableEnd() { }
<