1 //===-- llvm-mc.cpp - Machine Code Hacking Driver -------------------------===//
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 utility is a simple driver that allows command line hacking on machine
13 //===----------------------------------------------------------------------===//
15 #include "llvm/MC/MCParser/AsmLexer.h"
16 #include "llvm/MC/MCParser/MCAsmLexer.h"
17 #include "llvm/MC/MCContext.h"
18 #include "llvm/MC/MCCodeEmitter.h"
19 #include "llvm/MC/MCInstPrinter.h"
20 #include "llvm/MC/MCSectionMachO.h"
21 #include "llvm/MC/MCStreamer.h"
22 #include "llvm/Target/TargetAsmBackend.h"
23 #include "llvm/Target/TargetAsmParser.h"
24 #include "llvm/Target/TargetData.h"
25 #include "llvm/Target/TargetRegistry.h"
26 #include "llvm/Target/SubtargetFeature.h" // FIXME.
27 #include "llvm/Target/TargetAsmInfo.h" // FIXME.
28 #include "llvm/Target/TargetLowering.h" // FIXME.
29 #include "llvm/Target/TargetLoweringObjectFile.h" // FIXME.
30 #include "llvm/Target/TargetMachine.h" // FIXME.
31 #include "llvm/Target/TargetSelect.h"
32 #include "llvm/ADT/OwningPtr.h"
33 #include "llvm/Support/CommandLine.h"
34 #include "llvm/Support/FileUtilities.h"
35 #include "llvm/Support/FormattedStream.h"
36 #include "llvm/Support/ManagedStatic.h"
37 #include "llvm/Support/MemoryBuffer.h"
38 #include "llvm/Support/PrettyStackTrace.h"
39 #include "llvm/Support/SourceMgr.h"
40 #include "llvm/Support/ToolOutputFile.h"
41 #include "llvm/Support/Host.h"
42 #include "llvm/Support/Signals.h"
43 #include "llvm/Support/system_error.h"
44 #include "Disassembler.h"
47 static cl::opt<std::string>
48 InputFilename(cl::Positional, cl::desc("<input file>"), cl::init("-"));
50 static cl::opt<std::string>
51 OutputFilename("o", cl::desc("Output filename"),
52 cl::value_desc("filename"));
55 ShowEncoding("show-encoding", cl::desc("Show instruction encodings"));
58 ShowInst("show-inst", cl::desc("Show internal instruction representation"));
61 ShowInstOperands("show-inst-operands",
62 cl::desc("Show instructions operands as parsed"));
64 static cl::opt<unsigned>
65 OutputAsmVariant("output-asm-variant",
66 cl::desc("Syntax variant to use for output printing"));
69 RelaxAll("mc-relax-all", cl::desc("Relax all fixups"));
72 NoExecStack("mc-no-exec-stack", cl::desc("File doesn't need an exec stack"));
75 EnableLogging("enable-api-logging", cl::desc("Enable MC API logging"));
82 static cl::opt<OutputFileType>
83 FileType("filetype", cl::init(OFT_AssemblyFile),
84 cl::desc("Choose an output file type:"),
86 clEnumValN(OFT_AssemblyFile, "asm",
87 "Emit an assembly ('.s') file"),
88 clEnumValN(OFT_Null, "null",
89 "Don't emit anything (for timing purposes)"),
90 clEnumValN(OFT_ObjectFile, "obj",
91 "Emit a native object ('.o') file"),
94 static cl::list<std::string>
95 IncludeDirs("I", cl::desc("Directory of include files"),
96 cl::value_desc("directory"), cl::Prefix);
98 static cl::opt<std::string>
99 ArchName("arch", cl::desc("Target arch to assemble for, "
100 "see -version for available targets"));
102 static cl::opt<std::string>
103 TripleName("triple", cl::desc("Target triple to assemble for, "
104 "see -version for available targets"));
106 static cl::opt<std::string>
108 cl::desc("Target a specific cpu type (-mcpu=help for details)"),
109 cl::value_desc("cpu-name"),
113 NoInitialTextSection("n", cl::desc(
114 "Don't assume assembly file starts in the text section"));
117 SaveTempLabels("L", cl::desc(
118 "Don't discard temporary labels"));
127 static cl::opt<ActionType>
128 Action(cl::desc("Action to perform:"),
129 cl::init(AC_Assemble),
130 cl::values(clEnumValN(AC_AsLex, "as-lex",
131 "Lex tokens from a .s file"),
132 clEnumValN(AC_Assemble, "assemble",
133 "Assemble a .s file (default)"),
134 clEnumValN(AC_Disassemble, "disassemble",
135 "Disassemble strings of hex bytes"),
136 clEnumValN(AC_EDisassemble, "edis",
137 "Enhanced disassembly of strings of hex bytes"),
140 static const Target *GetTarget(const char *ProgName) {
141 // Figure out the target triple.
142 if (TripleName.empty())
143 TripleName = sys::getHostTriple();
144 if (!ArchName.empty()) {
145 llvm::Triple TT(TripleName);
146 TT.setArchName(ArchName);
147 TripleName = TT.str();
150 // Get the target specific parser.
152 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
156 errs() << ProgName << ": error: unable to get target for '" << TripleName
157 << "', see --version and --triple.\n";
161 static tool_output_file *GetOutputStream() {
162 if (OutputFilename == "")
163 OutputFilename = "-";
166 tool_output_file *Out = new tool_output_file(OutputFilename.c_str(), Err,
167 raw_fd_ostream::F_Binary);
169 errs() << Err << '\n';
177 static int AsLexInput(const char *ProgName) {
178 OwningPtr<MemoryBuffer> BufferPtr;
179 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) {
180 errs() << ProgName << ": " << ec.message() << '\n';
183 MemoryBuffer *Buffer = BufferPtr.take();
187 // Tell SrcMgr about this buffer, which is what TGParser will pick up.
188 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc());
190 // Record the location of the include directories so that the lexer can find
192 SrcMgr.setIncludeDirs(IncludeDirs);
194 const Target *TheTarget = GetTarget(ProgName);
198 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createAsmInfo(TripleName));
199 assert(MAI && "Unable to create target asm info!");
201 AsmLexer Lexer(*MAI);
202 Lexer.setBuffer(SrcMgr.getMemoryBuffer(0));
204 OwningPtr<tool_output_file> Out(GetOutputStream());
209 while (Lexer.Lex().isNot(AsmToken::Eof)) {
210 AsmToken Tok = Lexer.getTok();
212 switch (Tok.getKind()) {
214 SrcMgr.PrintMessage(Lexer.getLoc(), "unknown token", "warning");
217 case AsmToken::Error:
218 Error = true; // error already printed.
220 case AsmToken::Identifier:
221 Out->os() << "identifier: " << Lexer.getTok().getString();
223 case AsmToken::Integer:
224 Out->os() << "int: " << Lexer.getTok().getString();
227 Out->os() << "real: " << Lexer.getTok().getString();
229 case AsmToken::Register:
230 Out->os() << "register: " << Lexer.getTok().getRegVal();
232 case AsmToken::String:
233 Out->os() << "string: " << Lexer.getTok().getString();
236 case AsmToken::Amp: Out->os() << "Amp"; break;
237 case AsmToken::AmpAmp: Out->os() << "AmpAmp"; break;
238 case AsmToken::At: Out->os() << "At"; break;
239 case AsmToken::Caret: Out->os() << "Caret"; break;
240 case AsmToken::Colon: Out->os() << "Colon"; break;
241 case AsmToken::Comma: Out->os() << "Comma"; break;
242 case AsmToken::Dollar: Out->os() << "Dollar"; break;
243 case AsmToken::Dot: Out->os() << "Dot"; break;
244 case AsmToken::EndOfStatement: Out->os() << "EndOfStatement"; break;
245 case AsmToken::Eof: Out->os() << "Eof"; break;
246 case AsmToken::Equal: Out->os() << "Equal"; break;
247 case AsmToken::EqualEqual: Out->os() << "EqualEqual"; break;
248 case AsmToken::Exclaim: Out->os() << "Exclaim"; break;
249 case AsmToken::ExclaimEqual: Out->os() << "ExclaimEqual"; break;
250 case AsmToken::Greater: Out->os() << "Greater"; break;
251 case AsmToken::GreaterEqual: Out->os() << "GreaterEqual"; break;
252 case AsmToken::GreaterGreater: Out->os() << "GreaterGreater"; break;
253 case AsmToken::Hash: Out->os() << "Hash"; break;
254 case AsmToken::LBrac: Out->os() << "LBrac"; break;
255 case AsmToken::LCurly: Out->os() << "LCurly"; break;
256 case AsmToken::LParen: Out->os() << "LParen"; break;
257 case AsmToken::Less: Out->os() << "Less"; break;
258 case AsmToken::LessEqual: Out->os() << "LessEqual"; break;
259 case AsmToken::LessGreater: Out->os() << "LessGreater"; break;
260 case AsmToken::LessLess: Out->os() << "LessLess"; break;
261 case AsmToken::Minus: Out->os() << "Minus"; break;
262 case AsmToken::Percent: Out->os() << "Percent"; break;
263 case AsmToken::Pipe: Out->os() << "Pipe"; break;
264 case AsmToken::PipePipe: Out->os() << "PipePipe"; break;
265 case AsmToken::Plus: Out->os() << "Plus"; break;
266 case AsmToken::RBrac: Out->os() << "RBrac"; break;
267 case AsmToken::RCurly: Out->os() << "RCurly"; break;
268 case AsmToken::RParen: Out->os() << "RParen"; break;
269 case AsmToken::Slash: Out->os() << "Slash"; break;
270 case AsmToken::Star: Out->os() << "Star"; break;
271 case AsmToken::Tilde: Out->os() << "Tilde"; break;
274 // Print the token string.
276 Out->os().write_escaped(Tok.getString());
277 Out->os() << "\")\n";
280 // Keep output if no errors.
281 if (Error == 0) Out->keep();
286 static int AssembleInput(const char *ProgName) {
287 const Target *TheTarget = GetTarget(ProgName);
291 OwningPtr<MemoryBuffer> BufferPtr;
292 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, BufferPtr)) {
293 errs() << ProgName << ": " << ec.message() << '\n';
296 MemoryBuffer *Buffer = BufferPtr.take();
300 // Tell SrcMgr about this buffer, which is what the parser will pick up.
301 SrcMgr.AddNewSourceBuffer(Buffer, SMLoc());
303 // Record the location of the include directories so that the lexer can find
305 SrcMgr.setIncludeDirs(IncludeDirs);
308 llvm::OwningPtr<MCAsmInfo> MAI(TheTarget->createAsmInfo(TripleName));
309 assert(MAI && "Unable to create target asm info!");
311 // Package up features to be passed to target/subtarget
312 std::string FeaturesStr;
314 SubtargetFeatures Features;
315 Features.setCPU(MCPU);
316 FeaturesStr = Features.getString();
319 // FIXME: We shouldn't need to do this (and link in codegen).
320 // When we split this out, we should do it in a way that makes
321 // it straightforward to switch subtargets on the fly (.e.g,
322 // the .cpu and .code16 directives).
323 OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName,
327 errs() << ProgName << ": error: could not create target for triple '"
328 << TripleName << "'.\n";
332 const TargetAsmInfo *tai = new TargetAsmInfo(*TM);
333 MCContext Ctx(*MAI, tai);
335 Ctx.setAllowTemporaryLabels(false);
337 OwningPtr<tool_output_file> Out(GetOutputStream());
341 formatted_raw_ostream FOS(Out->os());
342 OwningPtr<MCStreamer> Str;
344 const TargetLoweringObjectFile &TLOF =
345 TM->getTargetLowering()->getObjFileLowering();
346 const_cast<TargetLoweringObjectFile&>(TLOF).Initialize(Ctx, *TM);
348 // FIXME: There is a bit of code duplication with addPassesToEmitFile.
349 if (FileType == OFT_AssemblyFile) {
351 TheTarget->createMCInstPrinter(*TM, OutputAsmVariant, *MAI);
352 MCCodeEmitter *CE = 0;
353 TargetAsmBackend *TAB = 0;
355 CE = TheTarget->createCodeEmitter(*TM, Ctx);
356 TAB = TheTarget->createAsmBackend(TripleName);
358 Str.reset(TheTarget->createAsmStreamer(Ctx, FOS, /*asmverbose*/true,
360 /*useCFI*/ true, IP, CE, TAB,
362 } else if (FileType == OFT_Null) {
363 Str.reset(createNullStreamer(Ctx));
365 assert(FileType == OFT_ObjectFile && "Invalid file type!");
366 MCCodeEmitter *CE = TheTarget->createCodeEmitter(*TM, Ctx);
367 TargetAsmBackend *TAB = TheTarget->createAsmBackend(TripleName);
368 Str.reset(TheTarget->createObjectStreamer(TripleName, Ctx, *TAB,
374 Str.reset(createLoggingStreamer(Str.take(), errs()));
377 OwningPtr<MCAsmParser> Parser(createMCAsmParser(*TheTarget, SrcMgr, Ctx,
379 OwningPtr<TargetAsmParser> TAP(TheTarget->createAsmParser(*Parser, *TM));
382 << ": error: this target does not support assembly parsing.\n";
386 Parser->setShowParsedOperands(ShowInstOperands);
387 Parser->setTargetParser(*TAP.get());
389 int Res = Parser->Run(NoInitialTextSection);
391 // Keep output if no errors.
392 if (Res == 0) Out->keep();
397 static int DisassembleInput(const char *ProgName, bool Enhanced) {
398 const Target *TheTarget = GetTarget(ProgName);
402 OwningPtr<MemoryBuffer> Buffer;
403 if (error_code ec = MemoryBuffer::getFileOrSTDIN(InputFilename, Buffer)) {
404 errs() << ProgName << ": " << ec.message() << '\n';
408 OwningPtr<tool_output_file> Out(GetOutputStream());
415 Disassembler::disassembleEnhanced(TripleName, *Buffer.take(), Out->os());
417 // Package up features to be passed to target/subtarget
418 std::string FeaturesStr;
420 SubtargetFeatures Features;
421 Features.setCPU(MCPU);
422 FeaturesStr = Features.getString();
425 // FIXME: We shouldn't need to do this (and link in codegen).
426 // When we split this out, we should do it in a way that makes
427 // it straightforward to switch subtargets on the fly (.e.g,
428 // the .cpu and .code16 directives).
429 OwningPtr<TargetMachine> TM(TheTarget->createTargetMachine(TripleName,
433 errs() << ProgName << ": error: could not create target for triple '"
434 << TripleName << "'.\n";
438 Res = Disassembler::disassemble(*TheTarget, *TM, TripleName,
439 *Buffer.take(), Out->os());
442 // Keep output if no errors.
443 if (Res == 0) Out->keep();
449 int main(int argc, char **argv) {
450 // Print a stack trace if we signal out.
451 sys::PrintStackTraceOnErrorSignal();
452 PrettyStackTraceProgram X(argc, argv);
453 llvm_shutdown_obj Y; // Call llvm_shutdown() on exit.
455 // Initialize targets and assembly printers/parsers.
456 llvm::InitializeAllTargetInfos();
457 // FIXME: We shouldn't need to initialize the Target(Machine)s.
458 llvm::InitializeAllTargets();
459 llvm::InitializeAllAsmPrinters();
460 llvm::InitializeAllAsmParsers();
461 llvm::InitializeAllDisassemblers();
463 cl::ParseCommandLineOptions(argc, argv, "llvm machine code playground\n");
464 TripleName = Triple::normalize(TripleName);
469 return AsLexInput(argv[0]);
471 return AssembleInput(argv[0]);
473 return DisassembleInput(argv[0], false);
474 case AC_EDisassemble:
475 return DisassembleInput(argv[0], true);