#ifndef BITSTREAM_READER_H
#define BITSTREAM_READER_H
#include "llvm/Bitcode/BitCodes.h"
#include <climits>
#include <string>
#include <vector>
namespace llvm {
class Deserializer;
class BitstreamReader {
public:
struct BlockInfo {
unsigned BlockID;
std::vector<BitCodeAbbrev*> Abbrevs;
std::string Name;
std::vector<std::pair<unsigned, std::string> > RecordNames;
};
private:
const unsigned char *FirstChar, *LastChar;
std::vector<BlockInfo> BlockInfoRecords;
bool IgnoreBlockInfoNames;
BitstreamReader(const BitstreamReader&); void operator=(const BitstreamReader&); public:
BitstreamReader() : FirstChar(0), LastChar(0), IgnoreBlockInfoNames(true) {
}
BitstreamReader(const unsigned char *Start, const unsigned char *End) {
IgnoreBlockInfoNames = true;
init(Start, End);
}
void init(const unsigned char *Start, const unsigned char *End) {
FirstChar = Start;
LastChar = End;
assert(((End-Start) & 3) == 0 &&"Bitcode stream not a multiple of 4 bytes");
}
~BitstreamReader() {
while (!BlockInfoRecords.empty()) {
BlockInfo &Info = BlockInfoRecords.back();
for (unsigned i = 0, e = static_cast<unsigned>(Info.Abbrevs.size());
i != e; ++i)
Info.Abbrevs[i]->dropRef();
BlockInfoRecords.pop_back();
}
}
const unsigned char *getFirstChar() const { return FirstChar; }
const unsigned char *getLastChar() const { return LastChar; }
void CollectBlockInfoNames() { IgnoreBlockInfoNames = false; }
bool isIgnoringBlockInfoNames() { return IgnoreBlockInfoNames; }
bool hasBlockInfoRecords() const { return !BlockInfoRecords.empty(); }
const BlockInfo *getBlockInfo(unsigned BlockID) const {
if (!BlockInfoRecords.empty() && BlockInfoRecords.back().BlockID == BlockID)
return &BlockInfoRecords.back();
for (unsigned i = 0, e = static_cast<unsigned>(BlockInfoRecords.size());
i != e; ++i)
if (BlockInfoRecords[i].BlockID == BlockID)
return &BlockInfoRecords[i];
return 0;
}
BlockInfo &getOrCreateBlockInfo(unsigned BlockID) {
if (const BlockInfo *BI = getBlockInfo(BlockID))
return *const_cast<BlockInfo*>(BI);
BlockInfoRecords.push_back(BlockInfo());
BlockInfoRecords.back().BlockID = BlockID;
return BlockInfoRecords.back();
}
};
class BitstreamCursor {
friend class Deserializer;
BitstreamReader *BitStream;
const unsigned char *NextChar;
uint32_t CurWord;
unsigned BitsInCurWord;
unsigned CurCodeSize;
std::vector<BitCodeAbbrev*> CurAbbrevs;
struct Block {
unsigned PrevCodeSize;
std::vector<BitCodeAbbrev*> PrevAbbrevs;
explicit Block(unsigned PCS) : PrevCodeSize(PCS) {}
};
SmallVector<Block, 8> BlockScope;
public:
BitstreamCursor() : BitStream(0), NextChar(0) {
}
BitstreamCursor(const BitstreamCursor &RHS) : BitStream(0), NextChar(0) {
operator=(RHS);
}
explicit BitstreamCursor(BitstreamReader &R) : BitStream(&R) {
NextChar = R.getFirstChar();
assert(NextChar && "Bitstream not initialized yet");
CurWord = 0;
BitsInCurWord = 0;
CurCodeSize = 2;
}
void init(BitstreamReader &R) {
freeState();
BitStream = &R;
NextChar = R.getFirstChar();
assert(NextChar && "Bitstream not initialized yet");
CurWord = 0;
BitsInCurWord = 0;
CurCodeSize = 2;
}
~BitstreamCursor() {
freeState();
}
void operator=(const BitstreamCursor &RHS) {
freeState();
BitStream = RHS.BitStream;
NextChar = RHS.NextChar;
CurWord = RHS.CurWord;
BitsInCurWord = RHS.BitsInCurWord;
CurCodeSize = RHS.CurCodeSize;
CurAbbrevs = RHS.CurAbbrevs;
for (unsigned i = 0, e = static_cast<unsigned>(CurAbbrevs.size());
i != e; ++i)
CurAbbrevs[i]->addRef();
for (unsigned S = 0, e = static_cast<unsigned>(BlockScope.size());
S != e; ++S) {
std::vector<BitCodeAbbrev*> &Abbrevs = BlockScope[S].PrevAbbrevs;
for (unsigned i = 0, e = static_cast<unsigned>(Abbrevs.size());
i != e; ++i)
Abbrevs[i]->addRef();
}
}
void freeState() {
for (unsigned i = 0, e = static_cast<unsigned>(CurAbbrevs.size());
i != e; ++i)
CurAbbrevs[i]->dropRef();
CurAbbrevs.clear();
for (unsigned S = 0, e = static_cast<unsigned>(BlockScope.size());
S != e; ++S) {
std::vector<BitCodeAbbrev*> &Abbrevs = BlockScope[S].PrevAbbrevs;
for (unsigned i = 0, e = static_cast<unsigned>(Abbrevs.size());
i != e; ++i)
Abbrevs[i]->dropRef();
}
BlockScope.clear();
}
unsigned GetAbbrevIDWidth() const { return CurCodeSize; }
bool AtEndOfStream() const {
return NextChar == BitStream->getLastChar() && BitsInCurWord == 0;
}
uint64_t GetCurrentBitNo() const {
return (NextChar-BitStream->getFirstChar())*CHAR_BIT - BitsInCurWord;
}
BitstreamReader *getBitStreamReader() {
return BitStream;
}
const BitstreamReader *getBitStreamReader() const {
return BitStream;
}
void JumpToBit(uint64_t BitNo) {
uintptr_t ByteNo = uintptr_t(BitNo/8) & ~3;
uintptr_t WordBitNo = uintptr_t(BitNo) & 31;
assert(ByteNo <= (uintptr_t)(BitStream->getLastChar()-
BitStream->getFirstChar()) &&
"Invalid location");
NextChar = BitStream->getFirstChar()+ByteNo;
BitsInCurWord = 0;
CurWord = 0;
if (WordBitNo)
Read(static_cast<unsigned>(WordBitNo));
}
uint32_t Read(unsigned NumBits) {
assert(NumBits <= 32 && "Cannot return more than 32 bits!");
if (BitsInCurWord >= NumBits) {
uint32_t R = CurWord & ((1U << NumBits)-1);
CurWord >>= NumBits;
BitsInCurWord -= NumBits;
return R;
}
if (NextChar == BitStream->getLastChar()) {
CurWord = 0;
BitsInCurWord = 0;
return 0;
}
unsigned R = CurWord;
CurWord = (NextChar[0] << 0) | (NextChar[1] << 8) |
(NextChar[2] << 16) | (NextChar[3] << 24);
NextChar += 4;
unsigned BitsLeft = NumBits-BitsInCurWord;
R |= (CurWord & (~0U >> (32-BitsLeft))) << BitsInCurWord;
if (BitsLeft != 32)
CurWord >>= BitsLeft;
else
CurWord = 0;
BitsInCurWord = 32-BitsLeft;
return R;
}
uint64_t Read64(unsigned NumBits) {
if (NumBits <= 32) return Read(NumBits);
uint64_t V = Read(32);
return V | (uint64_t)Read(NumBits-32) << 32;
}
uint32_t ReadVBR(unsigned NumBits) {
uint32_t Piece = Read(NumBits);
if ((Piece & (1U << (NumBits-1))) == 0)
return Piece;
uint32_t Result = 0;
unsigned NextBit = 0;
while (1) {
Result |= (Piece & ((1U << (NumBits-1))-1)) << NextBit;
if ((Piece & (1U << (NumBits-1))) == 0)
return Result;
NextBit += NumBits-1;
Piece = Read(NumBits);
}
}
uint64_t ReadVBR64(unsigned NumBits) {
uint32_t Piece = Read(NumBits);
if ((Piece & (1U << (NumBits-1))) == 0)
return uint64_t(Piece);
uint64_t Result = 0;
unsigned NextBit = 0;
while (1) {
Result |= uint64_t(Piece & ((1U << (NumBits-1))-1)) << NextBit;
if ((Piece & (1U << (NumBits-1))) == 0)
return Result;
NextBit += NumBits-1;
Piece = Read(NumBits);
}
}
void SkipToWord() {
BitsInCurWord = 0;
CurWord = 0;
}
unsigned ReadCode() {
return Read(CurCodeSize);
}
unsigned ReadSubBlockID() {
return ReadVBR(bitc::BlockIDWidth);
}
bool SkipBlock() {
ReadVBR(bitc::CodeLenWidth);
SkipToWord();
unsigned NumWords = Read(bitc::BlockSizeWidth);
if (AtEndOfStream() || NextChar+NumWords*4 > BitStream->getLastChar())
return true;
NextChar += NumWords*4;
return false;
}
bool EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = 0) {
BlockScope.push_back(Block(CurCodeSize));
BlockScope.back().PrevAbbrevs.swap(CurAbbrevs);
if (const BitstreamReader::BlockInfo *Info =
BitStream->getBlockInfo(BlockID)) {
for (unsigned i = 0, e = static_cast<unsigned>(Info->Abbrevs.size());
i != e; ++i) {
CurAbbrevs.push_back(Info->Abbrevs[i]);
CurAbbrevs.back()->addRef();
}
}
CurCodeSize = ReadVBR(bitc::CodeLenWidth);
SkipToWord();
unsigned NumWords = Read(bitc::BlockSizeWidth);
if (NumWordsP) *NumWordsP = NumWords;
if (CurCodeSize == 0 || AtEndOfStream() ||
NextChar+NumWords*4 > BitStream->getLastChar())
return true;
return false;
}
bool ReadBlockEnd() {
if (BlockScope.empty()) return true;
SkipToWord();
PopBlockScope();
return false;
}
private:
void PopBlockScope() {
CurCodeSize = BlockScope.back().PrevCodeSize;
for (unsigned i = 0, e = static_cast<unsigned>(CurAbbrevs.size());
i != e; ++i)
CurAbbrevs[i]->dropRef();
BlockScope.back().PrevAbbrevs.swap(CurAbbrevs);
BlockScope.pop_back();
}
private:
void ReadAbbreviatedLiteral(const BitCodeAbbrevOp &Op,
SmallVectorImpl<uint64_t> &Vals) {
assert(Op.isLiteral() && "Not a literal");
Vals.push_back(Op.getLiteralValue());
}
void ReadAbbreviatedField(const BitCodeAbbrevOp &Op,
SmallVectorImpl<uint64_t> &Vals) {
assert(!Op.isLiteral() && "Use ReadAbbreviatedLiteral for literals!");
switch (Op.getEncoding()) {
default: assert(0 && "Unknown encoding!");
case BitCodeAbbrevOp::Fixed:
Vals.push_back(Read((unsigned)Op.getEncodingData()));
break;
case BitCodeAbbrevOp::VBR:
Vals.push_back(ReadVBR64((unsigned)Op.getEncodingData()));
break;
case BitCodeAbbrevOp::Char6:
Vals.push_back(BitCodeAbbrevOp::DecodeChar6(Read(6)));
break;
}
}
public:
const BitCodeAbbrev *getAbbrev(unsigned AbbrevID) {
unsigned AbbrevNo = AbbrevID-bitc::FIRST_APPLICATION_ABBREV;
assert(AbbrevNo < CurAbbrevs.size() && "Invalid abbrev #!");
return CurAbbrevs[AbbrevNo];
}
unsigned ReadRecord(unsigned AbbrevID, SmallVectorImpl<uint64_t> &Vals,
const char **BlobStart = 0, unsigned *BlobLen = 0) {
if (AbbrevID == bitc::UNABBREV_RECORD) {
unsigned Code = ReadVBR(6);
unsigned NumElts = ReadVBR(6);
for (unsigned i = 0; i != NumElts; ++i)
Vals.push_back(ReadVBR64(6));
return Code;
}
const BitCodeAbbrev *Abbv = getAbbrev(AbbrevID);
for (unsigned i = 0, e = Abbv->getNumOperandInfos(); i != e; ++i) {
const BitCodeAbbrevOp &Op = Abbv->getOperandInfo(i);
if (Op.isLiteral()) {
ReadAbbreviatedLiteral(Op, Vals);
} else if (Op.getEncoding() == BitCodeAbbrevOp::Array) {
unsigned NumElts = ReadVBR(6);
assert(i+2 == e && "array op not second to last?");
const BitCodeAbbrevOp &EltEnc = Abbv->getOperandInfo(++i);
for (; NumElts; --NumElts)
ReadAbbreviatedField(EltEnc, Vals);
} else if (Op.getEncoding() == BitCodeAbbrevOp::Blob) {
unsigned NumElts = ReadVBR(6);
SkipToWord();
const unsigned char *NewEnd = NextChar+((NumElts+3)&~3);
if (NewEnd > BitStream->getLastChar()) {
Vals.append(NumElts, 0);
NextChar = BitStream->getLastChar();
break;
}
if (BlobStart) {
*BlobStart = (const char*)NextChar;
*BlobLen = NumElts;
} else {
for (; NumElts; ++NextChar, --NumElts)
Vals.push_back(*NextChar);
}
NextChar = NewEnd;
} else {
ReadAbbreviatedField(Op, Vals);
}
}
unsigned Code = (unsigned)Vals[0];
Vals.erase(Vals.begin());
return Code;
}
unsigned ReadRecord(unsigned AbbrevID, SmallVectorImpl<uint64_t> &Vals,
const char *&BlobStart, unsigned &BlobLen) {
return ReadRecord(AbbrevID, Vals, &BlobStart, &BlobLen);
}
void ReadAbbrevRecord() {
BitCodeAbbrev *Abbv = new BitCodeAbbrev();
unsigned NumOpInfo = ReadVBR(5);
for (unsigned i = 0; i != NumOpInfo; ++i) {
bool IsLiteral = Read(1) ? true : false;
if (IsLiteral) {
Abbv->Add(BitCodeAbbrevOp(ReadVBR64(8)));
continue;
}
BitCodeAbbrevOp::Encoding E = (BitCodeAbbrevOp::Encoding)Read(3);
if (BitCodeAbbrevOp::hasEncodingData(E))
Abbv->Add(BitCodeAbbrevOp(E, ReadVBR64(5)));
else
Abbv->Add(BitCodeAbbrevOp(E));
}
CurAbbrevs.push_back(Abbv);
}
public:
bool ReadBlockInfoBlock() {
if (BitStream->hasBlockInfoRecords())
return SkipBlock();
if (EnterSubBlock(bitc::BLOCKINFO_BLOCK_ID)) return true;
SmallVector<uint64_t, 64> Record;
BitstreamReader::BlockInfo *CurBlockInfo = 0;
while (1) {
unsigned Code = ReadCode();
if (Code == bitc::END_BLOCK)
return ReadBlockEnd();
if (Code == bitc::ENTER_SUBBLOCK) {
ReadSubBlockID();
if (SkipBlock()) return true;
continue;
}
if (Code == bitc::DEFINE_ABBREV) {
if (!CurBlockInfo) return true;
ReadAbbrevRecord();
BitCodeAbbrev *Abbv = CurAbbrevs.back();
CurAbbrevs.pop_back();
CurBlockInfo->Abbrevs.push_back(Abbv);
continue;
}
Record.clear();
switch (ReadRecord(Code, Record)) {
default: break; case bitc::BLOCKINFO_CODE_SETBID:
if (Record.size() < 1) return true;
CurBlockInfo = &BitStream->getOrCreateBlockInfo((unsigned)Record[0]);
break;
case bitc::BLOCKINFO_CODE_BLOCKNAME: {
if (!CurBlockInfo) return true;
if (BitStream->isIgnoringBlockInfoNames()) break; std::string Name;
for (unsigned i = 0, e = Record.size(); i != e; ++i)
Name += (char)Record[i];
CurBlockInfo->Name = Name;
break;
}
case bitc::BLOCKINFO_CODE_SETRECORDNAME: {
if (!CurBlockInfo) return true;
if (BitStream->isIgnoringBlockInfoNames()) break; std::string Name;
for (unsigned i = 1, e = Record.size(); i != e; ++i)
Name += (char)Record[i];
CurBlockInfo->RecordNames.push_back(std::make_pair((unsigned)Record[0],
Name));
break;
}
}
}
}
};
}
#endif