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/Analysis/Verifier.h"
31 #include "llvm/Bitcode/BitstreamReader.h"
32 #include "llvm/Bitcode/LLVMBitCodes.h"
33 #include "llvm/Support/CommandLine.h"
34 #include "llvm/Support/ManagedStatic.h"
35 #include "llvm/Support/MemoryBuffer.h"
36 #include "llvm/System/Signals.h"
43 static cl::opt<std::string>
44 InputFilename(cl::Positional, cl::desc("<input bitcode>"), cl::init("-"));
46 static cl::opt<std::string>
47 OutputFilename("-o", cl::init("-"), cl::desc("<output file>"));
49 static cl::opt<bool> Dump("dump", cl::desc("Dump low level bitcode trace"));
51 //===----------------------------------------------------------------------===//
52 // Bitcode specific analysis.
53 //===----------------------------------------------------------------------===//
55 static cl::opt<bool> NoHistogram("disable-histogram",
56 cl::desc("Do not print per-code histogram"));
59 NonSymbolic("non-symbolic",
60 cl::desc("Emit numberic info in dump even if"
61 " symbolic info is available"));
63 /// CurStreamType - If we can sniff the flavor of this stream, we can produce
71 /// GetBlockName - Return a symbolic block name if known, otherwise return
73 static const char *GetBlockName(unsigned BlockID) {
74 // Standard blocks for all bitcode files.
75 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) {
76 if (BlockID == bitc::BLOCKINFO_BLOCK_ID)
77 return "BLOCKINFO_BLOCK";
81 if (CurStreamType != LLVMIRBitstream) return 0;
85 case bitc::MODULE_BLOCK_ID: return "MODULE_BLOCK";
86 case bitc::PARAMATTR_BLOCK_ID: return "PARAMATTR_BLOCK";
87 case bitc::TYPE_BLOCK_ID: return "TYPE_BLOCK";
88 case bitc::CONSTANTS_BLOCK_ID: return "CONSTANTS_BLOCK";
89 case bitc::FUNCTION_BLOCK_ID: return "FUNCTION_BLOCK";
90 case bitc::TYPE_SYMTAB_BLOCK_ID: return "TYPE_SYMTAB";
91 case bitc::VALUE_SYMTAB_BLOCK_ID: return "VALUE_SYMTAB";
95 /// GetCodeName - Return a symbolic code name if known, otherwise return
97 static const char *GetCodeName(unsigned CodeID, unsigned BlockID) {
98 // Standard blocks for all bitcode files.
99 if (BlockID < bitc::FIRST_APPLICATION_BLOCKID) {
100 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) {
103 case bitc::MODULE_CODE_VERSION: return "VERSION";
109 if (CurStreamType != LLVMIRBitstream) return 0;
113 case bitc::MODULE_BLOCK_ID:
116 case bitc::MODULE_CODE_VERSION: return "VERSION";
117 case bitc::MODULE_CODE_TRIPLE: return "TRIPLE";
118 case bitc::MODULE_CODE_DATALAYOUT: return "DATALAYOUT";
119 case bitc::MODULE_CODE_ASM: return "ASM";
120 case bitc::MODULE_CODE_SECTIONNAME: return "SECTIONNAME";
121 case bitc::MODULE_CODE_DEPLIB: return "DEPLIB";
122 case bitc::MODULE_CODE_GLOBALVAR: return "GLOBALVAR";
123 case bitc::MODULE_CODE_FUNCTION: return "FUNCTION";
124 case bitc::MODULE_CODE_ALIAS: return "ALIAS";
125 case bitc::MODULE_CODE_PURGEVALS: return "PURGEVALS";
127 case bitc::PARAMATTR_BLOCK_ID:
130 case bitc::PARAMATTR_CODE_ENTRY: return "ENTRY";
132 case bitc::TYPE_BLOCK_ID:
135 case bitc::TYPE_CODE_NUMENTRY: return "NUMENTRY";
136 case bitc::TYPE_CODE_VOID: return "VOID";
137 case bitc::TYPE_CODE_FLOAT: return "FLOAT";
138 case bitc::TYPE_CODE_DOUBLE: return "DOUBLE";
139 case bitc::TYPE_CODE_LABEL: return "LABEL";
140 case bitc::TYPE_CODE_OPAQUE: return "OPAQUE";
141 case bitc::TYPE_CODE_INTEGER: return "INTEGER";
142 case bitc::TYPE_CODE_POINTER: return "POINTER";
143 case bitc::TYPE_CODE_FUNCTION: return "FUNCTION";
144 case bitc::TYPE_CODE_STRUCT: return "STRUCT";
145 case bitc::TYPE_CODE_ARRAY: return "ARRAY";
146 case bitc::TYPE_CODE_VECTOR: return "VECTOR";
149 case bitc::CONSTANTS_BLOCK_ID:
152 case bitc::CST_CODE_SETTYPE: return "SETTYPE";
153 case bitc::CST_CODE_NULL: return "NULL";
154 case bitc::CST_CODE_UNDEF: return "UNDEF";
155 case bitc::CST_CODE_INTEGER: return "INTEGER";
156 case bitc::CST_CODE_WIDE_INTEGER: return "WIDE_INTEGER";
157 case bitc::CST_CODE_FLOAT: return "FLOAT";
158 case bitc::CST_CODE_AGGREGATE: return "AGGREGATE";
159 case bitc::CST_CODE_STRING: return "STRING";
160 case bitc::CST_CODE_CSTRING: return "CSTRING";
161 case bitc::CST_CODE_CE_BINOP: return "CE_BINOP";
162 case bitc::CST_CODE_CE_CAST: return "CE_CAST";
163 case bitc::CST_CODE_CE_GEP: return "CE_GEP";
164 case bitc::CST_CODE_CE_SELECT: return "CE_SELECT";
165 case bitc::CST_CODE_CE_EXTRACTELT: return "CE_EXTRACTELT";
166 case bitc::CST_CODE_CE_INSERTELT: return "CE_INSERTELT";
167 case bitc::CST_CODE_CE_SHUFFLEVEC: return "CE_SHUFFLEVEC";
168 case bitc::CST_CODE_CE_CMP: return "CE_CMP";
169 case bitc::CST_CODE_INLINEASM: return "INLINEASM";
171 case bitc::FUNCTION_BLOCK_ID:
174 case bitc::FUNC_CODE_DECLAREBLOCKS: return "DECLAREBLOCKS";
175 case bitc::FUNC_CODE_INST_BB_UNWINDDEST: return "UNWINDDEST";
177 case bitc::FUNC_CODE_INST_BINOP: return "INST_BINOP";
178 case bitc::FUNC_CODE_INST_CAST: return "INST_CAST";
179 case bitc::FUNC_CODE_INST_GEP: return "INST_GEP";
180 case bitc::FUNC_CODE_INST_SELECT: return "INST_SELECT";
181 case bitc::FUNC_CODE_INST_EXTRACTELT: return "INST_EXTRACTELT";
182 case bitc::FUNC_CODE_INST_INSERTELT: return "INST_INSERTELT";
183 case bitc::FUNC_CODE_INST_SHUFFLEVEC: return "INST_SHUFFLEVEC";
184 case bitc::FUNC_CODE_INST_CMP: return "INST_CMP";
186 case bitc::FUNC_CODE_INST_RET: return "INST_RET";
187 case bitc::FUNC_CODE_INST_BR: return "INST_BR";
188 case bitc::FUNC_CODE_INST_SWITCH: return "INST_SWITCH";
189 case bitc::FUNC_CODE_INST_INVOKE: return "INST_INVOKE";
190 case bitc::FUNC_CODE_INST_UNWIND: return "INST_UNWIND";
191 case bitc::FUNC_CODE_INST_UNREACHABLE: return "INST_UNREACHABLE";
193 case bitc::FUNC_CODE_INST_PHI: return "INST_PHI";
194 case bitc::FUNC_CODE_INST_MALLOC: return "INST_MALLOC";
195 case bitc::FUNC_CODE_INST_FREE: return "INST_FREE";
196 case bitc::FUNC_CODE_INST_ALLOCA: return "INST_ALLOCA";
197 case bitc::FUNC_CODE_INST_LOAD: return "INST_LOAD";
198 case bitc::FUNC_CODE_INST_STORE: return "INST_STORE";
199 case bitc::FUNC_CODE_INST_CALL: return "INST_CALL";
200 case bitc::FUNC_CODE_INST_VAARG: return "INST_VAARG";
201 case bitc::FUNC_CODE_INST_STORE2: return "INST_STORE2";
202 case bitc::FUNC_CODE_INST_GETRESULT: return "INST_GETRESULT";
204 case bitc::TYPE_SYMTAB_BLOCK_ID:
207 case bitc::TST_CODE_ENTRY: return "ENTRY";
209 case bitc::VALUE_SYMTAB_BLOCK_ID:
212 case bitc::VST_CODE_ENTRY: return "ENTRY";
213 case bitc::VST_CODE_BBENTRY: return "BBENTRY";
219 struct PerBlockIDStats {
220 /// NumInstances - This the number of times this block ID has been seen.
221 unsigned NumInstances;
223 /// NumBits - The total size in bits of all of these blocks.
226 /// NumSubBlocks - The total number of blocks these blocks contain.
227 unsigned NumSubBlocks;
229 /// NumAbbrevs - The total number of abbreviations.
232 /// NumRecords - The total number of records these blocks contain, and the
233 /// number that are abbreviated.
234 unsigned NumRecords, NumAbbreviatedRecords;
236 /// CodeFreq - Keep track of the number of times we see each code.
237 std::vector<unsigned> CodeFreq;
240 : NumInstances(0), NumBits(0),
241 NumSubBlocks(0), NumAbbrevs(0), NumRecords(0), NumAbbreviatedRecords(0) {}
244 static std::map<unsigned, PerBlockIDStats> BlockIDStats;
248 /// Error - All bitcode analysis errors go through this function, making this a
249 /// good place to breakpoint if debugging.
250 static bool Error(const std::string &Err) {
251 std::cerr << Err << "\n";
255 /// ParseBlock - Read a block, updating statistics, etc.
256 static bool ParseBlock(BitstreamReader &Stream, unsigned IndentLevel) {
257 std::string Indent(IndentLevel*2, ' ');
258 uint64_t BlockBitStart = Stream.GetCurrentBitNo();
259 unsigned BlockID = Stream.ReadSubBlockID();
261 // Get the statistics for this BlockID.
262 PerBlockIDStats &BlockStats = BlockIDStats[BlockID];
264 BlockStats.NumInstances++;
266 // BLOCKINFO is a special part of the stream.
267 if (BlockID == bitc::BLOCKINFO_BLOCK_ID) {
268 if (Dump) std::cerr << Indent << "<BLOCKINFO_BLOCK/>\n";
269 if (Stream.ReadBlockInfoBlock())
270 return Error("Malformed BlockInfoBlock");
271 uint64_t BlockBitEnd = Stream.GetCurrentBitNo();
272 BlockStats.NumBits += BlockBitEnd-BlockBitStart;
276 unsigned NumWords = 0;
277 if (Stream.EnterSubBlock(BlockID, &NumWords))
278 return Error("Malformed block record");
280 const char *BlockName = 0;
282 std::cerr << Indent << "<";
283 if ((BlockName = GetBlockName(BlockID)))
284 std::cerr << BlockName;
286 std::cerr << "UnknownBlock" << BlockID;
288 if (NonSymbolic && BlockName)
289 std::cerr << " BlockID=" << BlockID;
291 std::cerr << " NumWords=" << NumWords
292 << " BlockCodeSize=" << Stream.GetAbbrevIDWidth() << ">\n";
295 SmallVector<uint64_t, 64> Record;
297 // Read all the records for this block.
299 if (Stream.AtEndOfStream())
300 return Error("Premature end of bitstream");
302 // Read the code for this record.
303 unsigned AbbrevID = Stream.ReadCode();
305 case bitc::END_BLOCK: {
306 if (Stream.ReadBlockEnd())
307 return Error("Error at end of block");
308 uint64_t BlockBitEnd = Stream.GetCurrentBitNo();
309 BlockStats.NumBits += BlockBitEnd-BlockBitStart;
311 std::cerr << Indent << "</";
313 std::cerr << BlockName << ">\n";
315 std::cerr << "UnknownBlock" << BlockID << ">\n";
319 case bitc::ENTER_SUBBLOCK: {
320 uint64_t SubBlockBitStart = Stream.GetCurrentBitNo();
321 if (ParseBlock(Stream, IndentLevel+1))
323 ++BlockStats.NumSubBlocks;
324 uint64_t SubBlockBitEnd = Stream.GetCurrentBitNo();
326 // Don't include subblock sizes in the size of this block.
327 BlockBitStart += SubBlockBitEnd-SubBlockBitStart;
330 case bitc::DEFINE_ABBREV:
331 Stream.ReadAbbrevRecord();
332 ++BlockStats.NumAbbrevs;
335 ++BlockStats.NumRecords;
336 if (AbbrevID != bitc::UNABBREV_RECORD)
337 ++BlockStats.NumAbbreviatedRecords;
340 unsigned Code = Stream.ReadRecord(AbbrevID, Record);
342 // Increment the # occurrences of this code.
343 if (BlockStats.CodeFreq.size() <= Code)
344 BlockStats.CodeFreq.resize(Code+1);
345 BlockStats.CodeFreq[Code]++;
348 std::cerr << Indent << " <";
349 if (const char *CodeName = GetCodeName(Code, BlockID))
350 std::cerr << CodeName;
352 std::cerr << "UnknownCode" << Code;
353 if (NonSymbolic && GetCodeName(Code, BlockID))
354 std::cerr << " codeid=" << Code;
355 if (AbbrevID != bitc::UNABBREV_RECORD)
356 std::cerr << " abbrevid=" << AbbrevID;
358 for (unsigned i = 0, e = Record.size(); i != e; ++i)
359 std::cerr << " op" << i << "=" << (int64_t)Record[i];
369 static void PrintSize(double Bits) {
370 std::cerr << Bits << "b/" << Bits/8 << "B/" << Bits/32 << "W";
374 /// AnalyzeBitcode - Analyze the bitcode file specified by InputFilename.
375 static int AnalyzeBitcode() {
376 // Read the input file.
377 MemoryBuffer *Buffer = MemoryBuffer::getFileOrSTDIN(InputFilename.c_str());
380 return Error("Error reading '" + InputFilename + "'.");
382 if (Buffer->getBufferSize() & 3)
383 return Error("Bitcode stream should be a multiple of 4 bytes in length");
385 unsigned char *BufPtr = (unsigned char *)Buffer->getBufferStart();
386 BitstreamReader Stream(BufPtr, BufPtr+Buffer->getBufferSize());
389 // Read the stream signature.
391 Signature[0] = Stream.Read(8);
392 Signature[1] = Stream.Read(8);
393 Signature[2] = Stream.Read(4);
394 Signature[3] = Stream.Read(4);
395 Signature[4] = Stream.Read(4);
396 Signature[5] = Stream.Read(4);
398 // Autodetect the file contents, if it is one we know.
399 CurStreamType = UnknownBitstream;
400 if (Signature[0] == 'B' && Signature[1] == 'C' &&
401 Signature[2] == 0x0 && Signature[3] == 0xC &&
402 Signature[4] == 0xE && Signature[5] == 0xD)
403 CurStreamType = LLVMIRBitstream;
405 unsigned NumTopBlocks = 0;
407 // Parse the top-level structure. We only allow blocks at the top-level.
408 while (!Stream.AtEndOfStream()) {
409 unsigned Code = Stream.ReadCode();
410 if (Code != bitc::ENTER_SUBBLOCK)
411 return Error("Invalid record at top-level");
413 if (ParseBlock(Stream, 0))
418 if (Dump) std::cerr << "\n\n";
420 uint64_t BufferSizeBits = Buffer->getBufferSize()*8;
421 // Print a summary of the read file.
422 std::cerr << "Summary of " << InputFilename << ":\n";
423 std::cerr << " Total size: ";
424 PrintSize(BufferSizeBits);
426 std::cerr << " Stream type: ";
427 switch (CurStreamType) {
428 default: assert(0 && "Unknown bitstream type");
429 case UnknownBitstream: std::cerr << "unknown\n"; break;
430 case LLVMIRBitstream: std::cerr << "LLVM IR\n"; break;
432 std::cerr << " # Toplevel Blocks: " << NumTopBlocks << "\n";
435 // Emit per-block stats.
436 std::cerr << "Per-block Summary:\n";
437 for (std::map<unsigned, PerBlockIDStats>::iterator I = BlockIDStats.begin(),
438 E = BlockIDStats.end(); I != E; ++I) {
439 std::cerr << " Block ID #" << I->first;
440 if (const char *BlockName = GetBlockName(I->first))
441 std::cerr << " (" << BlockName << ")";
444 const PerBlockIDStats &Stats = I->second;
445 std::cerr << " Num Instances: " << Stats.NumInstances << "\n";
446 std::cerr << " Total Size: ";
447 PrintSize(Stats.NumBits);
449 std::cerr << " % of file: "
450 << Stats.NumBits/(double)BufferSizeBits*100 << "\n";
451 if (Stats.NumInstances > 1) {
452 std::cerr << " Average Size: ";
453 PrintSize(Stats.NumBits/(double)Stats.NumInstances);
455 std::cerr << " Tot/Avg SubBlocks: " << Stats.NumSubBlocks << "/"
456 << Stats.NumSubBlocks/(double)Stats.NumInstances << "\n";
457 std::cerr << " Tot/Avg Abbrevs: " << Stats.NumAbbrevs << "/"
458 << Stats.NumAbbrevs/(double)Stats.NumInstances << "\n";
459 std::cerr << " Tot/Avg Records: " << Stats.NumRecords << "/"
460 << Stats.NumRecords/(double)Stats.NumInstances << "\n";
462 std::cerr << " Num SubBlocks: " << Stats.NumSubBlocks << "\n";
463 std::cerr << " Num Abbrevs: " << Stats.NumAbbrevs << "\n";
464 std::cerr << " Num Records: " << Stats.NumRecords << "\n";
466 if (Stats.NumRecords)
467 std::cerr << " % Abbrev Recs: " << (Stats.NumAbbreviatedRecords/
468 (double)Stats.NumRecords)*100 << "\n";
471 // Print a histogram of the codes we see.
472 if (!NoHistogram && !Stats.CodeFreq.empty()) {
473 std::vector<std::pair<unsigned, unsigned> > FreqPairs; // <freq,code>
474 for (unsigned i = 0, e = Stats.CodeFreq.size(); i != e; ++i)
475 if (unsigned Freq = Stats.CodeFreq[i])
476 FreqPairs.push_back(std::make_pair(Freq, i));
477 std::stable_sort(FreqPairs.begin(), FreqPairs.end());
478 std::reverse(FreqPairs.begin(), FreqPairs.end());
480 std::cerr << "\tCode Histogram:\n";
481 for (unsigned i = 0, e = FreqPairs.size(); i != e; ++i) {
482 std::cerr << "\t\t" << FreqPairs[i].first << "\t";
483 if (const char *CodeName = GetCodeName(FreqPairs[i].second, I->first))
484 std::cerr << CodeName << "\n";
486 std::cerr << "UnknownCode" << FreqPairs[i].second << "\n";
496 int main(int argc, char **argv) {
497 llvm_shutdown_obj X; // Call llvm_shutdown() on exit.
498 cl::ParseCommandLineOptions(argc, argv, "llvm-bcanalyzer file analyzer\n");
500 sys::PrintStackTraceOnErrorSignal();
502 return AnalyzeBitcode();