X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=include%2Fllvm%2FBitcode%2FBitstreamReader.h;h=c0cf6cde887f12d032256a9ffb4472e8bd0dbfc0;hb=3011017eccb93033ce44146c76d98ff3b84642b4;hp=b7ae47d1e62bc4fa3b231cc0f071f19b38316f42;hpb=0370cc6399022b687f898c9edc5e98482252997c;p=oota-llvm.git diff --git a/include/llvm/Bitcode/BitstreamReader.h b/include/llvm/Bitcode/BitstreamReader.h index b7ae47d1e62..c0cf6cde887 100644 --- a/include/llvm/Bitcode/BitstreamReader.h +++ b/include/llvm/Bitcode/BitstreamReader.h @@ -12,88 +12,90 @@ // //===----------------------------------------------------------------------===// -#ifndef BITSTREAM_READER_H -#define BITSTREAM_READER_H +#ifndef LLVM_BITCODE_BITSTREAMREADER_H +#define LLVM_BITCODE_BITSTREAMREADER_H #include "llvm/Bitcode/BitCodes.h" +#include "llvm/Support/Endian.h" +#include "llvm/Support/StreamingMemoryObject.h" #include +#include #include namespace llvm { - class Deserializer; - +/// This class is used to read from an LLVM bitcode stream, maintaining +/// information that is global to decoding the entire file. While a file is +/// being read, multiple cursors can be independently advanced or skipped around +/// within the file. These are represented by the BitstreamCursor class. class BitstreamReader { public: - /// BlockInfo - This contains information emitted to BLOCKINFO_BLOCK blocks. - /// These describe abbreviations that all blocks of the specified ID inherit. + /// This contains information emitted to BLOCKINFO_BLOCK blocks. These + /// describe abbreviations that all blocks of the specified ID inherit. struct BlockInfo { unsigned BlockID; - std::vector Abbrevs; + std::vector> Abbrevs; std::string Name; - + std::vector > RecordNames; }; private: - /// FirstChar/LastChar - This remembers the first and last bytes of the - /// stream. - const unsigned char *FirstChar, *LastChar; - + std::unique_ptr BitcodeBytes; + std::vector BlockInfoRecords; - /// IgnoreBlockInfoNames - This is set to true if we don't care about the - /// block/record name information in the BlockInfo block. Only llvm-bcanalyzer - /// uses this. + /// This is set to true if we don't care about the block/record name + /// information in the BlockInfo block. Only llvm-bcanalyzer uses this. bool IgnoreBlockInfoNames; - - BitstreamReader(const BitstreamReader&); // NOT IMPLEMENTED - void operator=(const BitstreamReader&); // NOT IMPLEMENTED + + BitstreamReader(const BitstreamReader&) = delete; + void operator=(const BitstreamReader&) = delete; public: - BitstreamReader() : FirstChar(0), LastChar(0), IgnoreBlockInfoNames(true) { + BitstreamReader() : IgnoreBlockInfoNames(true) { } - BitstreamReader(const unsigned char *Start, const unsigned char *End) { - IgnoreBlockInfoNames = true; + BitstreamReader(const unsigned char *Start, const unsigned char *End) + : IgnoreBlockInfoNames(true) { init(Start, End); } + BitstreamReader(std::unique_ptr BitcodeBytes) + : BitcodeBytes(std::move(BitcodeBytes)), IgnoreBlockInfoNames(true) {} + + BitstreamReader(BitstreamReader &&Other) { + *this = std::move(Other); + } + + BitstreamReader &operator=(BitstreamReader &&Other) { + BitcodeBytes = std::move(Other.BitcodeBytes); + // Explicitly swap block info, so that nothing gets destroyed twice. + std::swap(BlockInfoRecords, Other.BlockInfoRecords); + IgnoreBlockInfoNames = Other.IgnoreBlockInfoNames; + return *this; + } + 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"); + BitcodeBytes.reset(getNonStreamedMemoryObject(Start, End)); } - ~BitstreamReader() { - // Free the BlockInfoRecords. - while (!BlockInfoRecords.empty()) { - BlockInfo &Info = BlockInfoRecords.back(); - // Free blockinfo abbrev info. - for (unsigned i = 0, e = static_cast(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; } + MemoryObject &getBitcodeBytes() { return *BitcodeBytes; } - /// CollectBlockInfoNames - This is called by clients that want block/record - /// name information. + /// This is called by clients that want block/record name information. void CollectBlockInfoNames() { IgnoreBlockInfoNames = false; } bool isIgnoringBlockInfoNames() { return IgnoreBlockInfoNames; } - + //===--------------------------------------------------------------------===// // Block Manipulation //===--------------------------------------------------------------------===// - /// hasBlockInfoRecords - Return true if we've already read and processed the - /// block info block for this Bitstream. We only process it for the first - /// cursor that walks over it. + /// Return true if we've already read and processed the block info block for + /// this Bitstream. We only process it for the first cursor that walks over + /// it. bool hasBlockInfoRecords() const { return !BlockInfoRecords.empty(); } - - /// getBlockInfo - If there is block info for the specified ID, return it, - /// otherwise return null. + + /// If there is block info for the specified ID, return it, otherwise return + /// null. const BlockInfo *getBlockInfo(unsigned BlockID) const { // Common case, the most recent entry matches BlockID. if (!BlockInfoRecords.empty() && BlockInfoRecords.back().BlockID == BlockID) @@ -103,7 +105,7 @@ public: i != e; ++i) if (BlockInfoRecords[i].BlockID == BlockID) return &BlockInfoRecords[i]; - return 0; + return nullptr; } BlockInfo &getOrCreateBlockInfo(unsigned BlockID) { @@ -111,199 +113,271 @@ public: return *const_cast(BI); // Otherwise, add a new record. - BlockInfoRecords.push_back(BlockInfo()); + BlockInfoRecords.emplace_back(); BlockInfoRecords.back().BlockID = BlockID; return BlockInfoRecords.back(); } + /// Takes block info from the other bitstream reader. + /// + /// This is a "take" operation because BlockInfo records are non-trivial, and + /// indeed rather expensive. + void takeBlockInfo(BitstreamReader &&Other) { + assert(!hasBlockInfoRecords()); + BlockInfoRecords = std::move(Other.BlockInfoRecords); + } +}; + +/// When advancing through a bitstream cursor, each advance can discover a few +/// different kinds of entries: +struct BitstreamEntry { + enum { + Error, // Malformed bitcode was found. + EndBlock, // We've reached the end of the current block, (or the end of the + // file, which is treated like a series of EndBlock records. + SubBlock, // This is the start of a new subblock of a specific ID. + Record // This is a record with a specific AbbrevID. + } Kind; + + unsigned ID; + + static BitstreamEntry getError() { + BitstreamEntry E; E.Kind = Error; return E; + } + static BitstreamEntry getEndBlock() { + BitstreamEntry E; E.Kind = EndBlock; return E; + } + static BitstreamEntry getSubBlock(unsigned ID) { + BitstreamEntry E; E.Kind = SubBlock; E.ID = ID; return E; + } + static BitstreamEntry getRecord(unsigned AbbrevID) { + BitstreamEntry E; E.Kind = Record; E.ID = AbbrevID; return E; + } }; +/// This represents a position within a bitcode file. There may be multiple +/// independent cursors reading within one bitstream, each maintaining their own +/// local state. +/// +/// Unlike iterators, BitstreamCursors are heavy-weight objects that should not +/// be passed by value. class BitstreamCursor { - friend class Deserializer; BitstreamReader *BitStream; - const unsigned char *NextChar; - - /// CurWord - This is the current data we have pulled from the stream but have - /// not returned to the client. - uint32_t CurWord; - - /// BitsInCurWord - This is the number of bits in CurWord that are valid. This - /// is always from [0...31] inclusive. + size_t NextChar; + + // The size of the bicode. 0 if we don't know it yet. + size_t Size; + + /// This is the current data we have pulled from the stream but have not + /// returned to the client. This is specifically and intentionally defined to + /// follow the word size of the host machine for efficiency. We use word_t in + /// places that are aware of this to make it perfectly explicit what is going + /// on. + typedef size_t word_t; + word_t CurWord; + + /// This is the number of bits in CurWord that are valid. This is always from + /// [0...bits_of(size_t)-1] inclusive. unsigned BitsInCurWord; - - // CurCodeSize - This is the declared size of code values used for the current - // block, in bits. + + // This is the declared size of code values used for the current block, in + // bits. unsigned CurCodeSize; - - /// CurAbbrevs - Abbrevs installed at in this block. - std::vector CurAbbrevs; - + + /// Abbrevs installed at in this block. + std::vector> CurAbbrevs; + struct Block { unsigned PrevCodeSize; - std::vector PrevAbbrevs; + std::vector> PrevAbbrevs; explicit Block(unsigned PCS) : PrevCodeSize(PCS) {} }; - - /// BlockScope - This tracks the codesize of parent blocks. + + /// This tracks the codesize of parent blocks. SmallVector 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) { + static const size_t MaxChunkSize = sizeof(word_t) * 8; + + BitstreamCursor() { init(nullptr); } + + explicit BitstreamCursor(BitstreamReader &R) { init(&R); } + + void init(BitstreamReader *R) { freeState(); - - BitStream = &R; - NextChar = R.getFirstChar(); - assert(NextChar && "Bitstream not initialized yet"); - CurWord = 0; + + BitStream = R; + NextChar = 0; + Size = 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; - - // Copy abbreviations, and bump ref counts. - CurAbbrevs = RHS.CurAbbrevs; - for (unsigned i = 0, e = static_cast(CurAbbrevs.size()); - i != e; ++i) - CurAbbrevs[i]->addRef(); - - // Copy block scope and bump ref counts. - for (unsigned S = 0, e = static_cast(BlockScope.size()); - S != e; ++S) { - std::vector &Abbrevs = BlockScope[S].PrevAbbrevs; - for (unsigned i = 0, e = static_cast(Abbrevs.size()); - i != e; ++i) - Abbrevs[i]->addRef(); - } - } - - void freeState() { - // Free all the Abbrevs. - for (unsigned i = 0, e = static_cast(CurAbbrevs.size()); - i != e; ++i) - CurAbbrevs[i]->dropRef(); - CurAbbrevs.clear(); - - // Free all the Abbrevs in the block scope. - for (unsigned S = 0, e = static_cast(BlockScope.size()); - S != e; ++S) { - std::vector &Abbrevs = BlockScope[S].PrevAbbrevs; - for (unsigned i = 0, e = static_cast(Abbrevs.size()); - i != e; ++i) - Abbrevs[i]->dropRef(); - } - BlockScope.clear(); + + void freeState(); + + bool canSkipToPos(size_t pos) const { + // pos can be skipped to if it is a valid address or one byte past the end. + return pos == 0 || BitStream->getBitcodeBytes().isValidAddress( + static_cast(pos - 1)); } - - /// GetAbbrevIDWidth - Return the number of bits used to encode an abbrev #. - unsigned GetAbbrevIDWidth() const { return CurCodeSize; } - - bool AtEndOfStream() const { - return NextChar == BitStream->getLastChar() && BitsInCurWord == 0; + + bool AtEndOfStream() { + if (BitsInCurWord != 0) + return false; + if (Size != 0) + return Size == NextChar; + fillCurWord(); + return BitsInCurWord == 0; } - - /// GetCurrentBitNo - Return the bit # of the bit we are reading. + + /// Return the number of bits used to encode an abbrev #. + unsigned getAbbrevIDWidth() const { return CurCodeSize; } + + /// Return the bit # of the bit we are reading. uint64_t GetCurrentBitNo() const { - return (NextChar-BitStream->getFirstChar())*CHAR_BIT - BitsInCurWord; + return NextChar*CHAR_BIT - BitsInCurWord; } - + BitstreamReader *getBitStreamReader() { return BitStream; } const BitstreamReader *getBitStreamReader() const { return BitStream; } - - - /// JumpToBit - Reset the stream to the specified bit number. + + /// Flags that modify the behavior of advance(). + enum { + /// If this flag is used, the advance() method does not automatically pop + /// the block scope when the end of a block is reached. + AF_DontPopBlockAtEnd = 1, + + /// If this flag is used, abbrev entries are returned just like normal + /// records. + AF_DontAutoprocessAbbrevs = 2 + }; + + /// Advance the current bitstream, returning the next entry in the stream. + BitstreamEntry advance(unsigned Flags = 0) { + while (1) { + unsigned Code = ReadCode(); + if (Code == bitc::END_BLOCK) { + // Pop the end of the block unless Flags tells us not to. + if (!(Flags & AF_DontPopBlockAtEnd) && ReadBlockEnd()) + return BitstreamEntry::getError(); + return BitstreamEntry::getEndBlock(); + } + + if (Code == bitc::ENTER_SUBBLOCK) + return BitstreamEntry::getSubBlock(ReadSubBlockID()); + + if (Code == bitc::DEFINE_ABBREV && + !(Flags & AF_DontAutoprocessAbbrevs)) { + // We read and accumulate abbrev's, the client can't do anything with + // them anyway. + ReadAbbrevRecord(); + continue; + } + + return BitstreamEntry::getRecord(Code); + } + } + + /// This is a convenience function for clients that don't expect any + /// subblocks. This just skips over them automatically. + BitstreamEntry advanceSkippingSubblocks(unsigned Flags = 0) { + while (1) { + // If we found a normal entry, return it. + BitstreamEntry Entry = advance(Flags); + if (Entry.Kind != BitstreamEntry::SubBlock) + return Entry; + + // If we found a sub-block, just skip over it and check the next entry. + if (SkipBlock()) + return BitstreamEntry::getError(); + } + } + + /// Reset the stream to the specified bit number. 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"); - + size_t ByteNo = size_t(BitNo/8) & ~(sizeof(word_t)-1); + unsigned WordBitNo = unsigned(BitNo & (sizeof(word_t)*8-1)); + assert(canSkipToPos(ByteNo) && "Invalid location"); + // Move the cursor to the right word. - NextChar = BitStream->getFirstChar()+ByteNo; + NextChar = ByteNo; BitsInCurWord = 0; - CurWord = 0; - + // Skip over any bits that are already consumed. if (WordBitNo) - Read(static_cast(WordBitNo)); + Read(WordBitNo); + } + + void fillCurWord() { + if (Size != 0 && NextChar >= Size) + report_fatal_error("Unexpected end of file"); + + // Read the next word from the stream. + uint8_t Array[sizeof(word_t)] = {0}; + + uint64_t BytesRead = + BitStream->getBitcodeBytes().readBytes(Array, sizeof(Array), NextChar); + + // If we run out of data, stop at the end of the stream. + if (BytesRead == 0) { + CurWord = 0; + BitsInCurWord = 0; + Size = NextChar; + return; + } + + CurWord = + support::endian::read( + Array); + NextChar += BytesRead; + BitsInCurWord = BytesRead * 8; } - - - uint32_t Read(unsigned NumBits) { + + word_t Read(unsigned NumBits) { + static const unsigned BitsInWord = MaxChunkSize; + + assert(NumBits && NumBits <= BitsInWord && + "Cannot return zero or more than BitsInWord bits!"); + + static const unsigned Mask = sizeof(word_t) > 4 ? 0x3f : 0x1f; + // If the field is fully contained by CurWord, return it quickly. if (BitsInCurWord >= NumBits) { - uint32_t R = CurWord & ((1U << NumBits)-1); - CurWord >>= NumBits; + word_t R = CurWord & (~word_t(0) >> (BitsInWord - NumBits)); + + // Use a mask to avoid undefined behavior. + CurWord >>= (NumBits & Mask); + BitsInCurWord -= NumBits; return R; } + word_t R = BitsInCurWord ? CurWord : 0; + unsigned BitsLeft = NumBits - BitsInCurWord; + + fillCurWord(); + // If we run out of data, stop at the end of the stream. - if (NextChar == BitStream->getLastChar()) { - CurWord = 0; - BitsInCurWord = 0; + if (BitsLeft > BitsInCurWord) return 0; - } - unsigned R = CurWord; + word_t R2 = CurWord & (~word_t(0) >> (BitsInWord - BitsLeft)); - // Read the next word from the stream. - CurWord = (NextChar[0] << 0) | (NextChar[1] << 8) | - (NextChar[2] << 16) | (NextChar[3] << 24); - NextChar += 4; + // Use a mask to avoid undefined behavior. + CurWord >>= (BitsLeft & Mask); - // Extract NumBits-BitsInCurWord from what we just read. - unsigned BitsLeft = NumBits-BitsInCurWord; + BitsInCurWord -= BitsLeft; - // Be careful here, BitsLeft is in the range [1..32] inclusive. - R |= (CurWord & (~0U >> (32-BitsLeft))) << BitsInCurWord; + R |= R2 << (NumBits - BitsLeft); - // BitsLeft bits have just been used up from CurWord. - 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) @@ -322,15 +396,17 @@ public: } } + // Read a VBR that may have a value up to 64-bits in size. The chunk size of + // the VBR must still be <= 32 bits though. uint64_t ReadVBR64(unsigned NumBits) { - uint64_t Piece = Read(NumBits); + uint32_t Piece = Read(NumBits); if ((Piece & (1U << (NumBits-1))) == 0) - return Piece; + return uint64_t(Piece); uint64_t Result = 0; unsigned NextBit = 0; while (1) { - Result |= (Piece & ((1U << (NumBits-1))-1)) << NextBit; + Result |= uint64_t(Piece & ((1U << (NumBits-1))-1)) << NextBit; if ((Piece & (1U << (NumBits-1))) == 0) return Result; @@ -340,10 +416,20 @@ public: } } - void SkipToWord() { +private: + void SkipToFourByteBoundary() { + // If word_t is 64-bits and if we've read less than 32 bits, just dump + // the bits we have up to the next 32-bit boundary. + if (sizeof(word_t) > 4 && + BitsInCurWord >= 32) { + CurWord >>= BitsInCurWord-32; + BitsInCurWord = 32; + return; + } + BitsInCurWord = 0; - CurWord = 0; } +public: unsigned ReadCode() { return Read(CurCodeSize); @@ -353,286 +439,81 @@ public: // Block header: // [ENTER_SUBBLOCK, blockid, newcodelen, , blocklen] - /// ReadSubBlockID - Having read the ENTER_SUBBLOCK code, read the BlockID for - /// the block. + /// Having read the ENTER_SUBBLOCK code, read the BlockID for the block. unsigned ReadSubBlockID() { return ReadVBR(bitc::BlockIDWidth); } - /// SkipBlock - Having read the ENTER_SUBBLOCK abbrevid and a BlockID, skip - /// over the body of this block. If the block record is malformed, return - /// true. + /// Having read the ENTER_SUBBLOCK abbrevid and a BlockID, skip over the body + /// of this block. If the block record is malformed, return true. bool SkipBlock() { // Read and ignore the codelen value. Since we are skipping this block, we // don't care what code widths are used inside of it. ReadVBR(bitc::CodeLenWidth); - SkipToWord(); - unsigned NumWords = Read(bitc::BlockSizeWidth); + SkipToFourByteBoundary(); + unsigned NumFourBytes = Read(bitc::BlockSizeWidth); // Check that the block wasn't partially defined, and that the offset isn't // bogus. - if (AtEndOfStream() || NextChar+NumWords*4 > BitStream->getLastChar()) + size_t SkipTo = GetCurrentBitNo() + NumFourBytes*4*8; + if (AtEndOfStream() || !canSkipToPos(SkipTo/8)) return true; - NextChar += NumWords*4; + JumpToBit(SkipTo); return false; } - /// EnterSubBlock - Having read the ENTER_SUBBLOCK abbrevid, enter - /// the block, and return true if the block is valid. - bool EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = 0) { - // Save the current block's state on BlockScope. - BlockScope.push_back(Block(CurCodeSize)); - BlockScope.back().PrevAbbrevs.swap(CurAbbrevs); - - // Add the abbrevs specific to this block to the CurAbbrevs list. - if (const BitstreamReader::BlockInfo *Info = - BitStream->getBlockInfo(BlockID)) { - for (unsigned i = 0, e = static_cast(Info->Abbrevs.size()); - i != e; ++i) { - CurAbbrevs.push_back(Info->Abbrevs[i]); - CurAbbrevs.back()->addRef(); - } - } - - // Get the codesize of this block. - CurCodeSize = ReadVBR(bitc::CodeLenWidth); - SkipToWord(); - unsigned NumWords = Read(bitc::BlockSizeWidth); - if (NumWordsP) *NumWordsP = NumWords; - - // Validate that this block is sane. - if (CurCodeSize == 0 || AtEndOfStream() || - NextChar+NumWords*4 > BitStream->getLastChar()) - return true; - - return false; - } + /// Having read the ENTER_SUBBLOCK abbrevid, enter the block, and return true + /// if the block has an error. + bool EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = nullptr); bool ReadBlockEnd() { if (BlockScope.empty()) return true; // Block tail: // [END_BLOCK, ] - SkipToWord(); + SkipToFourByteBoundary(); - PopBlockScope(); + popBlockScope(); return false; } private: - void PopBlockScope() { - CurCodeSize = BlockScope.back().PrevCodeSize; - // Delete abbrevs from popped scope. - for (unsigned i = 0, e = static_cast(CurAbbrevs.size()); - i != e; ++i) - CurAbbrevs[i]->dropRef(); + void popBlockScope() { + CurCodeSize = BlockScope.back().PrevCodeSize; - BlockScope.back().PrevAbbrevs.swap(CurAbbrevs); + CurAbbrevs = std::move(BlockScope.back().PrevAbbrevs); BlockScope.pop_back(); } - //===--------------------------------------------------------------------===// + //===--------------------------------------------------------------------===// // Record Processing //===--------------------------------------------------------------------===// -private: - void ReadAbbreviatedLiteral(const BitCodeAbbrevOp &Op, - SmallVectorImpl &Vals) { - assert(Op.isLiteral() && "Not a literal"); - // If the abbrev specifies the literal value to use, use it. - Vals.push_back(Op.getLiteralValue()); - } - - void ReadAbbreviatedField(const BitCodeAbbrevOp &Op, - SmallVectorImpl &Vals) { - assert(!Op.isLiteral() && "Use ReadAbbreviatedLiteral for literals!"); - - // Decode the value as we are commanded. - 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: - - /// getAbbrev - Return the abbreviation for the specified AbbrevId. + /// Return the abbreviation for the specified AbbrevId. 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 &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) { - // Array case. Read the number of elements as a vbr6. - unsigned NumElts = ReadVBR(6); - - // Get the element encoding. - assert(i+2 == e && "array op not second to last?"); - const BitCodeAbbrevOp &EltEnc = Abbv->getOperandInfo(++i); - - // Read all the elements. - for (; NumElts; --NumElts) - ReadAbbreviatedField(EltEnc, Vals); - } else if (Op.getEncoding() == BitCodeAbbrevOp::Blob) { - // Blob case. Read the number of bytes as a vbr6. - unsigned NumElts = ReadVBR(6); - SkipToWord(); // 32-bit alignment - - // Figure out where the end of this blob will be including tail padding. - const unsigned char *NewEnd = NextChar+((NumElts+3)&~3); - - // If this would read off the end of the bitcode file, just set the - // record to empty and return. - if (NewEnd > BitStream->getLastChar()) { - Vals.append(NumElts, 0); - NextChar = BitStream->getLastChar(); - break; - } - - // Otherwise, read the number of bytes. If we can return a reference to - // the data, do so to avoid copying it. - if (BlobStart) { - *BlobStart = (const char*)NextChar; - *BlobLen = NumElts; - } else { - for (; NumElts; ++NextChar, --NumElts) - Vals.push_back(*NextChar); - } - // Skip over tail padding. - NextChar = NewEnd; - } else { - ReadAbbreviatedField(Op, Vals); - } - } - - unsigned Code = (unsigned)Vals[0]; - Vals.erase(Vals.begin()); - return Code; + unsigned AbbrevNo = AbbrevID - bitc::FIRST_APPLICATION_ABBREV; + if (AbbrevNo >= CurAbbrevs.size()) + report_fatal_error("Invalid abbrev number"); + return CurAbbrevs[AbbrevNo].get(); } - unsigned ReadRecord(unsigned AbbrevID, SmallVectorImpl &Vals, - const char *&BlobStart, unsigned &BlobLen) { - return ReadRecord(AbbrevID, Vals, &BlobStart, &BlobLen); - } + /// Read the current record and discard it. + void skipRecord(unsigned AbbrevID); + + unsigned readRecord(unsigned AbbrevID, SmallVectorImpl &Vals, + StringRef *Blob = nullptr); - //===--------------------------------------------------------------------===// // Abbrev Processing //===--------------------------------------------------------------------===// + void ReadAbbrevRecord(); - 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 this is the second stream to get to the block info block, skip it. - if (BitStream->hasBlockInfoRecords()) - return SkipBlock(); - - if (EnterSubBlock(bitc::BLOCKINFO_BLOCK_ID)) return true; - - SmallVector Record; - BitstreamReader::BlockInfo *CurBlockInfo = 0; - - // Read all the records for this module. - while (1) { - unsigned Code = ReadCode(); - if (Code == bitc::END_BLOCK) - return ReadBlockEnd(); - if (Code == bitc::ENTER_SUBBLOCK) { - ReadSubBlockID(); - if (SkipBlock()) return true; - continue; - } - - // Read abbrev records, associate them with CurBID. - if (Code == bitc::DEFINE_ABBREV) { - if (!CurBlockInfo) return true; - ReadAbbrevRecord(); - - // ReadAbbrevRecord installs the abbrev in CurAbbrevs. Move it to the - // appropriate BlockInfo. - BitCodeAbbrev *Abbv = CurAbbrevs.back(); - CurAbbrevs.pop_back(); - CurBlockInfo->Abbrevs.push_back(Abbv); - continue; - } - - // Read a record. - Record.clear(); - switch (ReadRecord(Code, Record)) { - default: break; // Default behavior, ignore unknown content. - 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; // Ignore name. - 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; // Ignore name. - 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; - } - } - } - } + bool ReadBlockInfoBlock(); }; - + } // End llvm namespace #endif