1 //===-- llvm-bcanalyzer.cpp - Bitcode Analyzer --------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This tool may be invoked in the following manner:
11 // llvm-bcanalyzer [options] - Read LLVM bitcode from stdin
12 // llvm-bcanalyzer [options] x.bc - Read LLVM bitcode from the x.bc file
15 // --help - Output information about command line switches
16 // --dump - Dump low-level bitcode structure in readable format
18 // This tool provides analytical information about a bitcode file. It is
19 // intended as an aid to developers of bitcode reading and writing software. It
20 // produces on std::out a summary of the bitcode file that shows various
21 // statistics about the contents of the file. By default this information is
22 // detailed and contains information about individual bitcode blocks and the
23 // functions in the module.
24 // The tool is also able to print a bitcode file in a straight forward text
25 // format that shows the containment and relationships of the information in
26 // the bitcode file (-dump option).
28 //===----------------------------------------------------------------------===//
30 #include "llvm/Bitcode/BitstreamReader.h"
31 #include "llvm/ADT/Optional.h"
32 #include "llvm/Bitcode/LLVMBitCodes.h"
33 #include "llvm/Bitcode/ReaderWriter.h"
34 #include "llvm/IR/Verifier.h"
35 #include "llvm/Support/CommandLine.h"
36 #include "llvm/Support/Format.h"
37 #include "llvm/Support/ManagedStatic.h"
38 #include "llvm/Support/MemoryBuffer.h"
39 #include "llvm/Support/PrettyStackTrace.h"
40 #include "llvm/Support/Signals.h"
41 #include "llvm/Support/raw_ostream.h"
45 #include <system_error>
48 static cl::opt<std::string>
49 InputFilename(cl::Positional, cl::desc("<input bitcode>"), cl::init("-"));
51 static cl::opt<bool> Dump("dump", cl::desc("Dump low level bitcode trace"));
53 //===----------------------------------------------------------------------===//
54 // Bitcode specific analysis.
55 //===----------------------------------------------------------------------===//
57 static cl::opt<bool> NoHistogram("disable-histogram",
58 cl::desc("Do not print per-code histogram"));
61 NonSymbolic("non-symbolic",
62 cl::desc("Emit numeric info in dump even if"
63 " symbolic info is available"));
65 static cl::opt<std::string>
66 BlockInfoFilename("block-info",
67 cl::desc("Use the BLOCK_INFO from the given file"));
70 ShowBinaryBlobs("show-binary-blobs",
71 cl::desc("Print binary blobs using hex escapes"));
75 /// CurStreamTypeType - A type for CurStreamType
76 enum CurStreamTypeType {
83 /// GetBlockName - Return a symbolic block name if known, otherwise return
85 static const char *GetBlockName(unsigned BlockID,
86 const BitstreamReader &StreamFile,
87 CurStreamTypeType CurStreamType) {
88 // Standard blocks for all bitcode files.
89 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) {
90 if (BlockID == bitc::BLOCKINFO_BLOCK_ID)
91 return "BLOCKINFO_BLOCK";
95 // Check to see if we have a blockinfo record for this block, with a name.
96 if (const BitstreamReader::BlockInfo *Info =
97 StreamFile.getBlockInfo(BlockID)) {
98 if (!Info->Name.empty())
99 return Info->Name.c_str();
103 if (CurStreamType != LLVMIRBitstream) return nullptr;
106 default: return nullptr;
107 case bitc::MODULE_BLOCK_ID: return "MODULE_BLOCK";
108 case bitc::PARAMATTR_BLOCK_ID: return "PARAMATTR_BLOCK";
109 case bitc::PARAMATTR_GROUP_BLOCK_ID: return "PARAMATTR_GROUP_BLOCK_ID";
110 case bitc::TYPE_BLOCK_ID_NEW: return "TYPE_BLOCK_ID";
111 case bitc::CONSTANTS_BLOCK_ID: return "CONSTANTS_BLOCK";
112 case bitc::FUNCTION_BLOCK_ID: return "FUNCTION_BLOCK";
113 case bitc::IDENTIFICATION_BLOCK_ID:
114 return "IDENTIFICATION_BLOCK_ID";
115 case bitc::VALUE_SYMTAB_BLOCK_ID: return "VALUE_SYMTAB";
116 case bitc::METADATA_BLOCK_ID: return "METADATA_BLOCK";
117 case bitc::METADATA_ATTACHMENT_ID: return "METADATA_ATTACHMENT_BLOCK";
118 case bitc::USELIST_BLOCK_ID: return "USELIST_BLOCK_ID";
119 case bitc::FUNCTION_SUMMARY_BLOCK_ID:
120 return "FUNCTION_SUMMARY_BLOCK";
121 case bitc::MODULE_STRTAB_BLOCK_ID: return "MODULE_STRTAB_BLOCK";
125 /// GetCodeName - Return a symbolic code name if known, otherwise return
127 static const char *GetCodeName(unsigned CodeID, unsigned BlockID,
128 const BitstreamReader &StreamFile,
129 CurStreamTypeType CurStreamType) {
130 // Standard blocks for all bitcode files.
131 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) {
132 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) {
134 default: return nullptr;
135 case bitc::BLOCKINFO_CODE_SETBID: return "SETBID";
136 case bitc::BLOCKINFO_CODE_BLOCKNAME: return "BLOCKNAME";
137 case bitc::BLOCKINFO_CODE_SETRECORDNAME: return "SETRECORDNAME";
143 // Check to see if we have a blockinfo record for this record, with a name.
144 if (const BitstreamReader::BlockInfo *Info =
145 StreamFile.getBlockInfo(BlockID)) {
146 for (unsigned i = 0, e = Info->RecordNames.size(); i != e; ++i)
147 if (Info->RecordNames[i].first == CodeID)
148 return Info->RecordNames[i].second.c_str();
152 if (CurStreamType != LLVMIRBitstream) return nullptr;
154 #define STRINGIFY_CODE(PREFIX, CODE) \
155 case bitc::PREFIX##_##CODE: \
158 default: return nullptr;
159 case bitc::MODULE_BLOCK_ID:
161 default: return nullptr;
162 STRINGIFY_CODE(MODULE_CODE, VERSION)
163 STRINGIFY_CODE(MODULE_CODE, TRIPLE)
164 STRINGIFY_CODE(MODULE_CODE, DATALAYOUT)
165 STRINGIFY_CODE(MODULE_CODE, ASM)
166 STRINGIFY_CODE(MODULE_CODE, SECTIONNAME)
167 STRINGIFY_CODE(MODULE_CODE, DEPLIB) // FIXME: Remove in 4.0
168 STRINGIFY_CODE(MODULE_CODE, GLOBALVAR)
169 STRINGIFY_CODE(MODULE_CODE, FUNCTION)
170 STRINGIFY_CODE(MODULE_CODE, ALIAS)
171 STRINGIFY_CODE(MODULE_CODE, PURGEVALS)
172 STRINGIFY_CODE(MODULE_CODE, GCNAME)
173 STRINGIFY_CODE(MODULE_CODE, VSTOFFSET)
175 case bitc::IDENTIFICATION_BLOCK_ID:
179 STRINGIFY_CODE(IDENTIFICATION_CODE, STRING)
180 STRINGIFY_CODE(IDENTIFICATION_CODE, EPOCH)
182 case bitc::PARAMATTR_BLOCK_ID:
184 default: return nullptr;
185 // FIXME: Should these be different?
186 case bitc::PARAMATTR_CODE_ENTRY_OLD: return "ENTRY";
187 case bitc::PARAMATTR_CODE_ENTRY: return "ENTRY";
188 case bitc::PARAMATTR_GRP_CODE_ENTRY: return "ENTRY";
190 case bitc::TYPE_BLOCK_ID_NEW:
192 default: return nullptr;
193 STRINGIFY_CODE(TYPE_CODE, NUMENTRY)
194 STRINGIFY_CODE(TYPE_CODE, VOID)
195 STRINGIFY_CODE(TYPE_CODE, FLOAT)
196 STRINGIFY_CODE(TYPE_CODE, DOUBLE)
197 STRINGIFY_CODE(TYPE_CODE, LABEL)
198 STRINGIFY_CODE(TYPE_CODE, OPAQUE)
199 STRINGIFY_CODE(TYPE_CODE, INTEGER)
200 STRINGIFY_CODE(TYPE_CODE, POINTER)
201 STRINGIFY_CODE(TYPE_CODE, ARRAY)
202 STRINGIFY_CODE(TYPE_CODE, VECTOR)
203 STRINGIFY_CODE(TYPE_CODE, X86_FP80)
204 STRINGIFY_CODE(TYPE_CODE, FP128)
205 STRINGIFY_CODE(TYPE_CODE, PPC_FP128)
206 STRINGIFY_CODE(TYPE_CODE, METADATA)
207 STRINGIFY_CODE(TYPE_CODE, STRUCT_ANON)
208 STRINGIFY_CODE(TYPE_CODE, STRUCT_NAME)
209 STRINGIFY_CODE(TYPE_CODE, STRUCT_NAMED)
210 STRINGIFY_CODE(TYPE_CODE, FUNCTION)
213 case bitc::CONSTANTS_BLOCK_ID:
215 default: return nullptr;
216 STRINGIFY_CODE(CST_CODE, SETTYPE)
217 STRINGIFY_CODE(CST_CODE, NULL)
218 STRINGIFY_CODE(CST_CODE, UNDEF)
219 STRINGIFY_CODE(CST_CODE, INTEGER)
220 STRINGIFY_CODE(CST_CODE, WIDE_INTEGER)
221 STRINGIFY_CODE(CST_CODE, FLOAT)
222 STRINGIFY_CODE(CST_CODE, AGGREGATE)
223 STRINGIFY_CODE(CST_CODE, STRING)
224 STRINGIFY_CODE(CST_CODE, CSTRING)
225 STRINGIFY_CODE(CST_CODE, CE_BINOP)
226 STRINGIFY_CODE(CST_CODE, CE_CAST)
227 STRINGIFY_CODE(CST_CODE, CE_GEP)
228 STRINGIFY_CODE(CST_CODE, CE_INBOUNDS_GEP)
229 STRINGIFY_CODE(CST_CODE, CE_SELECT)
230 STRINGIFY_CODE(CST_CODE, CE_EXTRACTELT)
231 STRINGIFY_CODE(CST_CODE, CE_INSERTELT)
232 STRINGIFY_CODE(CST_CODE, CE_SHUFFLEVEC)
233 STRINGIFY_CODE(CST_CODE, CE_CMP)
234 STRINGIFY_CODE(CST_CODE, INLINEASM)
235 STRINGIFY_CODE(CST_CODE, CE_SHUFVEC_EX)
236 case bitc::CST_CODE_BLOCKADDRESS: return "CST_CODE_BLOCKADDRESS";
237 STRINGIFY_CODE(CST_CODE, DATA)
239 case bitc::FUNCTION_BLOCK_ID:
241 default: return nullptr;
242 STRINGIFY_CODE(FUNC_CODE, DECLAREBLOCKS)
243 STRINGIFY_CODE(FUNC_CODE, INST_BINOP)
244 STRINGIFY_CODE(FUNC_CODE, INST_CAST)
245 STRINGIFY_CODE(FUNC_CODE, INST_GEP_OLD)
246 STRINGIFY_CODE(FUNC_CODE, INST_INBOUNDS_GEP_OLD)
247 STRINGIFY_CODE(FUNC_CODE, INST_SELECT)
248 STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTELT)
249 STRINGIFY_CODE(FUNC_CODE, INST_INSERTELT)
250 STRINGIFY_CODE(FUNC_CODE, INST_SHUFFLEVEC)
251 STRINGIFY_CODE(FUNC_CODE, INST_CMP)
252 STRINGIFY_CODE(FUNC_CODE, INST_RET)
253 STRINGIFY_CODE(FUNC_CODE, INST_BR)
254 STRINGIFY_CODE(FUNC_CODE, INST_SWITCH)
255 STRINGIFY_CODE(FUNC_CODE, INST_INVOKE)
256 STRINGIFY_CODE(FUNC_CODE, INST_UNREACHABLE)
257 STRINGIFY_CODE(FUNC_CODE, INST_CLEANUPRET)
258 STRINGIFY_CODE(FUNC_CODE, INST_CATCHRET)
259 STRINGIFY_CODE(FUNC_CODE, INST_CATCHPAD)
260 STRINGIFY_CODE(FUNC_CODE, INST_CLEANUPENDPAD)
261 STRINGIFY_CODE(FUNC_CODE, INST_CATCHENDPAD)
262 STRINGIFY_CODE(FUNC_CODE, INST_TERMINATEPAD)
263 STRINGIFY_CODE(FUNC_CODE, INST_PHI)
264 STRINGIFY_CODE(FUNC_CODE, INST_ALLOCA)
265 STRINGIFY_CODE(FUNC_CODE, INST_LOAD)
266 STRINGIFY_CODE(FUNC_CODE, INST_VAARG)
267 STRINGIFY_CODE(FUNC_CODE, INST_STORE)
268 STRINGIFY_CODE(FUNC_CODE, INST_EXTRACTVAL)
269 STRINGIFY_CODE(FUNC_CODE, INST_INSERTVAL)
270 STRINGIFY_CODE(FUNC_CODE, INST_CMP2)
271 STRINGIFY_CODE(FUNC_CODE, INST_VSELECT)
272 STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC_AGAIN)
273 STRINGIFY_CODE(FUNC_CODE, INST_CALL)
274 STRINGIFY_CODE(FUNC_CODE, DEBUG_LOC)
275 STRINGIFY_CODE(FUNC_CODE, INST_GEP)
277 case bitc::VALUE_SYMTAB_BLOCK_ID:
279 default: return nullptr;
280 STRINGIFY_CODE(VST_CODE, ENTRY)
281 STRINGIFY_CODE(VST_CODE, BBENTRY)
282 STRINGIFY_CODE(VST_CODE, FNENTRY)
283 STRINGIFY_CODE(VST_CODE, COMBINED_FNENTRY)
285 case bitc::MODULE_STRTAB_BLOCK_ID:
289 STRINGIFY_CODE(MST_CODE, ENTRY)
291 case bitc::FUNCTION_SUMMARY_BLOCK_ID:
295 STRINGIFY_CODE(FS_CODE, PERMODULE_ENTRY)
296 STRINGIFY_CODE(FS_CODE, COMBINED_ENTRY)
298 case bitc::METADATA_ATTACHMENT_ID:
300 default:return nullptr;
301 STRINGIFY_CODE(METADATA, ATTACHMENT)
303 case bitc::METADATA_BLOCK_ID:
305 default:return nullptr;
306 STRINGIFY_CODE(METADATA, STRING)
307 STRINGIFY_CODE(METADATA, NAME)
308 STRINGIFY_CODE(METADATA, KIND)
309 STRINGIFY_CODE(METADATA, NODE)
310 STRINGIFY_CODE(METADATA, VALUE)
311 STRINGIFY_CODE(METADATA, OLD_NODE)
312 STRINGIFY_CODE(METADATA, OLD_FN_NODE)
313 STRINGIFY_CODE(METADATA, NAMED_NODE)
314 STRINGIFY_CODE(METADATA, DISTINCT_NODE)
315 STRINGIFY_CODE(METADATA, LOCATION)
316 STRINGIFY_CODE(METADATA, GENERIC_DEBUG)
317 STRINGIFY_CODE(METADATA, SUBRANGE)
318 STRINGIFY_CODE(METADATA, ENUMERATOR)
319 STRINGIFY_CODE(METADATA, BASIC_TYPE)
320 STRINGIFY_CODE(METADATA, FILE)
321 STRINGIFY_CODE(METADATA, DERIVED_TYPE)
322 STRINGIFY_CODE(METADATA, COMPOSITE_TYPE)
323 STRINGIFY_CODE(METADATA, SUBROUTINE_TYPE)
324 STRINGIFY_CODE(METADATA, COMPILE_UNIT)
325 STRINGIFY_CODE(METADATA, SUBPROGRAM)
326 STRINGIFY_CODE(METADATA, LEXICAL_BLOCK)
327 STRINGIFY_CODE(METADATA, LEXICAL_BLOCK_FILE)
328 STRINGIFY_CODE(METADATA, NAMESPACE)
329 STRINGIFY_CODE(METADATA, TEMPLATE_TYPE)
330 STRINGIFY_CODE(METADATA, TEMPLATE_VALUE)
331 STRINGIFY_CODE(METADATA, GLOBAL_VAR)
332 STRINGIFY_CODE(METADATA, LOCAL_VAR)
333 STRINGIFY_CODE(METADATA, EXPRESSION)
334 STRINGIFY_CODE(METADATA, OBJC_PROPERTY)
335 STRINGIFY_CODE(METADATA, IMPORTED_ENTITY)
336 STRINGIFY_CODE(METADATA, MODULE)
338 case bitc::USELIST_BLOCK_ID:
340 default:return nullptr;
341 case bitc::USELIST_CODE_DEFAULT: return "USELIST_CODE_DEFAULT";
342 case bitc::USELIST_CODE_BB: return "USELIST_CODE_BB";
345 #undef STRINGIFY_CODE
348 struct PerRecordStats {
349 unsigned NumInstances;
353 PerRecordStats() : NumInstances(0), NumAbbrev(0), TotalBits(0) {}
356 struct PerBlockIDStats {
357 /// NumInstances - This the number of times this block ID has been seen.
358 unsigned NumInstances;
360 /// NumBits - The total size in bits of all of these blocks.
363 /// NumSubBlocks - The total number of blocks these blocks contain.
364 unsigned NumSubBlocks;
366 /// NumAbbrevs - The total number of abbreviations.
369 /// NumRecords - The total number of records these blocks contain, and the
370 /// number that are abbreviated.
371 unsigned NumRecords, NumAbbreviatedRecords;
373 /// CodeFreq - Keep track of the number of times we see each code.
374 std::vector<PerRecordStats> CodeFreq;
377 : NumInstances(0), NumBits(0),
378 NumSubBlocks(0), NumAbbrevs(0), NumRecords(0), NumAbbreviatedRecords(0) {}
381 static std::map<unsigned, PerBlockIDStats> BlockIDStats;
385 /// Error - All bitcode analysis errors go through this function, making this a
386 /// good place to breakpoint if debugging.
387 static bool Error(const Twine &Err) {
388 errs() << Err << "\n";
392 /// ParseBlock - Read a block, updating statistics, etc.
393 static bool ParseBlock(BitstreamCursor &Stream, unsigned BlockID,
394 unsigned IndentLevel, CurStreamTypeType CurStreamType) {
395 std::string Indent(IndentLevel*2, ' ');
396 uint64_t BlockBitStart = Stream.GetCurrentBitNo();
398 // Get the statistics for this BlockID.
399 PerBlockIDStats &BlockStats = BlockIDStats[BlockID];
401 BlockStats.NumInstances++;
403 // BLOCKINFO is a special part of the stream.
404 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) {
405 if (Dump) outs() << Indent << "<BLOCKINFO_BLOCK/>\n";
406 if (Stream.ReadBlockInfoBlock())
407 return Error("Malformed BlockInfoBlock");
408 uint64_t BlockBitEnd = Stream.GetCurrentBitNo();
409 BlockStats.NumBits += BlockBitEnd-BlockBitStart;
413 unsigned NumWords = 0;
414 if (Stream.EnterSubBlock(BlockID, &NumWords))
415 return Error("Malformed block record");
417 const char *BlockName = nullptr;
419 outs() << Indent << "<";
420 if ((BlockName = GetBlockName(BlockID, *Stream.getBitStreamReader(),
424 outs() << "UnknownBlock" << BlockID;
426 if (NonSymbolic && BlockName)
427 outs() << " BlockID=" << BlockID;
429 outs() << " NumWords=" << NumWords
430 << " BlockCodeSize=" << Stream.getAbbrevIDWidth() << ">\n";
433 SmallVector<uint64_t, 64> Record;
435 // Read all the records for this block.
437 if (Stream.AtEndOfStream())
438 return Error("Premature end of bitstream");
440 uint64_t RecordStartBit = Stream.GetCurrentBitNo();
442 BitstreamEntry Entry =
443 Stream.advance(BitstreamCursor::AF_DontAutoprocessAbbrevs);
445 switch (Entry.Kind) {
446 case BitstreamEntry::Error:
447 return Error("malformed bitcode file");
448 case BitstreamEntry::EndBlock: {
449 uint64_t BlockBitEnd = Stream.GetCurrentBitNo();
450 BlockStats.NumBits += BlockBitEnd-BlockBitStart;
452 outs() << Indent << "</";
454 outs() << BlockName << ">\n";
456 outs() << "UnknownBlock" << BlockID << ">\n";
461 case BitstreamEntry::SubBlock: {
462 uint64_t SubBlockBitStart = Stream.GetCurrentBitNo();
463 if (ParseBlock(Stream, Entry.ID, IndentLevel+1, CurStreamType))
465 ++BlockStats.NumSubBlocks;
466 uint64_t SubBlockBitEnd = Stream.GetCurrentBitNo();
468 // Don't include subblock sizes in the size of this block.
469 BlockBitStart += SubBlockBitEnd-SubBlockBitStart;
472 case BitstreamEntry::Record:
473 // The interesting case.
477 if (Entry.ID == bitc::DEFINE_ABBREV) {
478 Stream.ReadAbbrevRecord();
479 ++BlockStats.NumAbbrevs;
485 ++BlockStats.NumRecords;
488 unsigned Code = Stream.readRecord(Entry.ID, Record, &Blob);
490 // Increment the # occurrences of this code.
491 if (BlockStats.CodeFreq.size() <= Code)
492 BlockStats.CodeFreq.resize(Code+1);
493 BlockStats.CodeFreq[Code].NumInstances++;
494 BlockStats.CodeFreq[Code].TotalBits +=
495 Stream.GetCurrentBitNo()-RecordStartBit;
496 if (Entry.ID != bitc::UNABBREV_RECORD) {
497 BlockStats.CodeFreq[Code].NumAbbrev++;
498 ++BlockStats.NumAbbreviatedRecords;
502 outs() << Indent << " <";
503 if (const char *CodeName =
504 GetCodeName(Code, BlockID, *Stream.getBitStreamReader(),
508 outs() << "UnknownCode" << Code;
510 GetCodeName(Code, BlockID, *Stream.getBitStreamReader(),
512 outs() << " codeid=" << Code;
513 const BitCodeAbbrev *Abbv = nullptr;
514 if (Entry.ID != bitc::UNABBREV_RECORD) {
515 Abbv = Stream.getAbbrev(Entry.ID);
516 outs() << " abbrevid=" << Entry.ID;
519 for (unsigned i = 0, e = Record.size(); i != e; ++i)
520 outs() << " op" << i << "=" << (int64_t)Record[i];
525 for (unsigned i = 1, e = Abbv->getNumOperandInfos(); i != e; ++i) {
526 const BitCodeAbbrevOp &Op = Abbv->getOperandInfo(i);
527 if (!Op.isEncoding() || Op.getEncoding() != BitCodeAbbrevOp::Array)
529 assert(i + 2 == e && "Array op not second to last");
531 bool ArrayIsPrintable = true;
532 for (unsigned j = i - 1, je = Record.size(); j != je; ++j) {
533 if (!isprint(static_cast<unsigned char>(Record[j]))) {
534 ArrayIsPrintable = false;
537 Str += (char)Record[j];
539 if (ArrayIsPrintable)
540 outs() << " record string = '" << Str << "'";
546 outs() << " blob data = ";
547 if (ShowBinaryBlobs) {
549 outs().write_escaped(Blob, /*hex=*/true) << "'";
551 bool BlobIsPrintable = true;
552 for (unsigned i = 0, e = Blob.size(); i != e; ++i)
553 if (!isprint(static_cast<unsigned char>(Blob[i]))) {
554 BlobIsPrintable = false;
559 outs() << "'" << Blob << "'";
561 outs() << "unprintable, " << Blob.size() << " bytes.";
570 static void PrintSize(double Bits) {
571 outs() << format("%.2f/%.2fB/%luW", Bits, Bits/8,(unsigned long)(Bits/32));
573 static void PrintSize(uint64_t Bits) {
574 outs() << format("%lub/%.2fB/%luW", (unsigned long)Bits,
575 (double)Bits/8, (unsigned long)(Bits/32));
578 static bool openBitcodeFile(StringRef Path,
579 std::unique_ptr<MemoryBuffer> &MemBuf,
580 BitstreamReader &StreamFile,
581 BitstreamCursor &Stream,
582 CurStreamTypeType &CurStreamType) {
583 // Read the input file.
584 ErrorOr<std::unique_ptr<MemoryBuffer>> MemBufOrErr =
585 MemoryBuffer::getFileOrSTDIN(Path);
586 if (std::error_code EC = MemBufOrErr.getError())
587 return Error(Twine("Error reading '") + Path + "': " + EC.message());
588 MemBuf = std::move(MemBufOrErr.get());
590 if (MemBuf->getBufferSize() & 3)
591 return Error("Bitcode stream should be a multiple of 4 bytes in length");
593 const unsigned char *BufPtr = (const unsigned char *)MemBuf->getBufferStart();
594 const unsigned char *EndBufPtr = BufPtr + MemBuf->getBufferSize();
596 // If we have a wrapper header, parse it and ignore the non-bc file contents.
597 // The magic number is 0x0B17C0DE stored in little endian.
598 if (isBitcodeWrapper(BufPtr, EndBufPtr))
599 if (SkipBitcodeWrapperHeader(BufPtr, EndBufPtr, true))
600 return Error("Invalid bitcode wrapper header");
602 StreamFile = BitstreamReader(BufPtr, EndBufPtr);
603 Stream = BitstreamCursor(StreamFile);
604 StreamFile.CollectBlockInfoNames();
606 // Read the stream signature.
608 Signature[0] = Stream.Read(8);
609 Signature[1] = Stream.Read(8);
610 Signature[2] = Stream.Read(4);
611 Signature[3] = Stream.Read(4);
612 Signature[4] = Stream.Read(4);
613 Signature[5] = Stream.Read(4);
615 // Autodetect the file contents, if it is one we know.
616 CurStreamType = UnknownBitstream;
617 if (Signature[0] == 'B' && Signature[1] == 'C' &&
618 Signature[2] == 0x0 && Signature[3] == 0xC &&
619 Signature[4] == 0xE && Signature[5] == 0xD)
620 CurStreamType = LLVMIRBitstream;
625 /// AnalyzeBitcode - Analyze the bitcode file specified by InputFilename.
626 static int AnalyzeBitcode() {
627 std::unique_ptr<MemoryBuffer> StreamBuffer;
628 BitstreamReader StreamFile;
629 BitstreamCursor Stream;
630 CurStreamTypeType CurStreamType;
631 if (openBitcodeFile(InputFilename, StreamBuffer, StreamFile, Stream,
635 // Read block info from BlockInfoFilename, if specified.
636 // The block info must be a top-level block.
637 if (!BlockInfoFilename.empty()) {
638 std::unique_ptr<MemoryBuffer> BlockInfoBuffer;
639 BitstreamReader BlockInfoFile;
640 BitstreamCursor BlockInfoCursor;
641 CurStreamTypeType BlockInfoStreamType;
642 if (openBitcodeFile(BlockInfoFilename, BlockInfoBuffer, BlockInfoFile,
643 BlockInfoCursor, BlockInfoStreamType))
646 while (!BlockInfoCursor.AtEndOfStream()) {
647 unsigned Code = BlockInfoCursor.ReadCode();
648 if (Code != bitc::ENTER_SUBBLOCK)
649 return Error("Invalid record at top-level in block info file");
651 unsigned BlockID = BlockInfoCursor.ReadSubBlockID();
652 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) {
653 if (BlockInfoCursor.ReadBlockInfoBlock())
654 return Error("Malformed BlockInfoBlock in block info file");
658 BlockInfoCursor.SkipBlock();
661 StreamFile.takeBlockInfo(std::move(BlockInfoFile));
664 unsigned NumTopBlocks = 0;
666 // Parse the top-level structure. We only allow blocks at the top-level.
667 while (!Stream.AtEndOfStream()) {
668 unsigned Code = Stream.ReadCode();
669 if (Code != bitc::ENTER_SUBBLOCK)
670 return Error("Invalid record at top-level");
672 unsigned BlockID = Stream.ReadSubBlockID();
674 if (ParseBlock(Stream, BlockID, 0, CurStreamType))
679 if (Dump) outs() << "\n\n";
681 uint64_t BufferSizeBits = StreamFile.getBitcodeBytes().getExtent() * CHAR_BIT;
682 // Print a summary of the read file.
683 outs() << "Summary of " << InputFilename << ":\n";
684 outs() << " Total size: ";
685 PrintSize(BufferSizeBits);
687 outs() << " Stream type: ";
688 switch (CurStreamType) {
689 case UnknownBitstream: outs() << "unknown\n"; break;
690 case LLVMIRBitstream: outs() << "LLVM IR\n"; break;
692 outs() << " # Toplevel Blocks: " << NumTopBlocks << "\n";
695 // Emit per-block stats.
696 outs() << "Per-block Summary:\n";
697 for (std::map<unsigned, PerBlockIDStats>::iterator I = BlockIDStats.begin(),
698 E = BlockIDStats.end(); I != E; ++I) {
699 outs() << " Block ID #" << I->first;
700 if (const char *BlockName = GetBlockName(I->first, StreamFile,
702 outs() << " (" << BlockName << ")";
705 const PerBlockIDStats &Stats = I->second;
706 outs() << " Num Instances: " << Stats.NumInstances << "\n";
707 outs() << " Total Size: ";
708 PrintSize(Stats.NumBits);
710 double pct = (Stats.NumBits * 100.0) / BufferSizeBits;
711 outs() << " Percent of file: " << format("%2.4f%%", pct) << "\n";
712 if (Stats.NumInstances > 1) {
713 outs() << " Average Size: ";
714 PrintSize(Stats.NumBits/(double)Stats.NumInstances);
716 outs() << " Tot/Avg SubBlocks: " << Stats.NumSubBlocks << "/"
717 << Stats.NumSubBlocks/(double)Stats.NumInstances << "\n";
718 outs() << " Tot/Avg Abbrevs: " << Stats.NumAbbrevs << "/"
719 << Stats.NumAbbrevs/(double)Stats.NumInstances << "\n";
720 outs() << " Tot/Avg Records: " << Stats.NumRecords << "/"
721 << Stats.NumRecords/(double)Stats.NumInstances << "\n";
723 outs() << " Num SubBlocks: " << Stats.NumSubBlocks << "\n";
724 outs() << " Num Abbrevs: " << Stats.NumAbbrevs << "\n";
725 outs() << " Num Records: " << Stats.NumRecords << "\n";
727 if (Stats.NumRecords) {
728 double pct = (Stats.NumAbbreviatedRecords * 100.0) / Stats.NumRecords;
729 outs() << " Percent Abbrevs: " << format("%2.4f%%", pct) << "\n";
733 // Print a histogram of the codes we see.
734 if (!NoHistogram && !Stats.CodeFreq.empty()) {
735 std::vector<std::pair<unsigned, unsigned> > FreqPairs; // <freq,code>
736 for (unsigned i = 0, e = Stats.CodeFreq.size(); i != e; ++i)
737 if (unsigned Freq = Stats.CodeFreq[i].NumInstances)
738 FreqPairs.push_back(std::make_pair(Freq, i));
739 std::stable_sort(FreqPairs.begin(), FreqPairs.end());
740 std::reverse(FreqPairs.begin(), FreqPairs.end());
742 outs() << "\tRecord Histogram:\n";
743 outs() << "\t\t Count # Bits %% Abv Record Kind\n";
744 for (unsigned i = 0, e = FreqPairs.size(); i != e; ++i) {
745 const PerRecordStats &RecStats = Stats.CodeFreq[FreqPairs[i].second];
747 outs() << format("\t\t%7d %9lu",
748 RecStats.NumInstances,
749 (unsigned long)RecStats.TotalBits);
751 if (RecStats.NumAbbrev)
754 (double)RecStats.NumAbbrev/RecStats.NumInstances*100);
758 if (const char *CodeName =
759 GetCodeName(FreqPairs[i].second, I->first, StreamFile,
761 outs() << CodeName << "\n";
763 outs() << "UnknownCode" << FreqPairs[i].second << "\n";
773 int main(int argc, char **argv) {
774 // Print a stack trace if we signal out.
775 sys::PrintStackTraceOnErrorSignal();
776 PrettyStackTraceProgram X(argc, argv);
777 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit.
778 cl::ParseCommandLineOptions(argc, argv, "llvm-bcanalyzer file analyzer\n");
780 return AnalyzeBitcode();