1 //===- MILexer.cpp - Machine instructions lexer implementation ------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the lexing of machine instructions.
11 //
12 //===----------------------------------------------------------------------===//
13
14 #include "MILexer.h"
15 #include "llvm/ADT/APSInt.h"
16 #include "llvm/ADT/None.h"
17 #include "llvm/ADT/STLExtras.h"
18 #include "llvm/ADT/StringExtras.h"
19 #include "llvm/ADT/StringSwitch.h"
20 #include "llvm/ADT/StringRef.h"
21 #include "llvm/ADT/Twine.h"
22 #include <algorithm>
23 #include <cassert>
24 #include <cctype>
25 #include <string>
26
27 using namespace llvm;
28
29 namespace {
30
31 using ErrorCallbackType =
32 function_ref<void(StringRef::iterator Loc, const Twine &)>;
33
34 /// This class provides a way to iterate and get characters from the source
35 /// string.
36 class Cursor {
37 const char *Ptr = nullptr;
38 const char *End = nullptr;
39
40 public:
Cursor(NoneType)41 Cursor(NoneType) {}
42
Cursor(StringRef Str)43 explicit Cursor(StringRef Str) {
44 Ptr = Str.data();
45 End = Ptr + Str.size();
46 }
47
isEOF() const48 bool isEOF() const { return Ptr == End; }
49
peek(int I=0) const50 char peek(int I = 0) const { return End - Ptr <= I ? 0 : Ptr[I]; }
51
advance(unsigned I=1)52 void advance(unsigned I = 1) { Ptr += I; }
53
remaining() const54 StringRef remaining() const { return StringRef(Ptr, End - Ptr); }
55
upto(Cursor C) const56 StringRef upto(Cursor C) const {
57 assert(C.Ptr >= Ptr && C.Ptr <= End);
58 return StringRef(Ptr, C.Ptr - Ptr);
59 }
60
location() const61 StringRef::iterator location() const { return Ptr; }
62
operator bool() const63 operator bool() const { return Ptr != nullptr; }
64 };
65
66 } // end anonymous namespace
67
reset(TokenKind Kind,StringRef Range)68 MIToken &MIToken::reset(TokenKind Kind, StringRef Range) {
69 this->Kind = Kind;
70 this->Range = Range;
71 return *this;
72 }
73
setStringValue(StringRef StrVal)74 MIToken &MIToken::setStringValue(StringRef StrVal) {
75 StringValue = StrVal;
76 return *this;
77 }
78
setOwnedStringValue(std::string StrVal)79 MIToken &MIToken::setOwnedStringValue(std::string StrVal) {
80 StringValueStorage = std::move(StrVal);
81 StringValue = StringValueStorage;
82 return *this;
83 }
84
setIntegerValue(APSInt IntVal)85 MIToken &MIToken::setIntegerValue(APSInt IntVal) {
86 this->IntVal = std::move(IntVal);
87 return *this;
88 }
89
90 /// Skip the leading whitespace characters and return the updated cursor.
skipWhitespace(Cursor C)91 static Cursor skipWhitespace(Cursor C) {
92 while (isblank(C.peek()))
93 C.advance();
94 return C;
95 }
96
isNewlineChar(char C)97 static bool isNewlineChar(char C) { return C == '\n' || C == '\r'; }
98
99 /// Skip a line comment and return the updated cursor.
skipComment(Cursor C)100 static Cursor skipComment(Cursor C) {
101 if (C.peek() != ';')
102 return C;
103 while (!isNewlineChar(C.peek()) && !C.isEOF())
104 C.advance();
105 return C;
106 }
107
108 /// Return true if the given character satisfies the following regular
109 /// expression: [-a-zA-Z$._0-9]
isIdentifierChar(char C)110 static bool isIdentifierChar(char C) {
111 return isalpha(C) || isdigit(C) || C == '_' || C == '-' || C == '.' ||
112 C == '$';
113 }
114
115 /// Unescapes the given string value.
116 ///
117 /// Expects the string value to be quoted.
unescapeQuotedString(StringRef Value)118 static std::string unescapeQuotedString(StringRef Value) {
119 assert(Value.front() == '"' && Value.back() == '"');
120 Cursor C = Cursor(Value.substr(1, Value.size() - 2));
121
122 std::string Str;
123 Str.reserve(C.remaining().size());
124 while (!C.isEOF()) {
125 char Char = C.peek();
126 if (Char == '\\') {
127 if (C.peek(1) == '\\') {
128 // Two '\' become one
129 Str += '\\';
130 C.advance(2);
131 continue;
132 }
133 if (isxdigit(C.peek(1)) && isxdigit(C.peek(2))) {
134 Str += hexDigitValue(C.peek(1)) * 16 + hexDigitValue(C.peek(2));
135 C.advance(3);
136 continue;
137 }
138 }
139 Str += Char;
140 C.advance();
141 }
142 return Str;
143 }
144
145 /// Lex a string constant using the following regular expression: \"[^\"]*\"
lexStringConstant(Cursor C,ErrorCallbackType ErrorCallback)146 static Cursor lexStringConstant(Cursor C, ErrorCallbackType ErrorCallback) {
147 assert(C.peek() == '"');
148 for (C.advance(); C.peek() != '"'; C.advance()) {
149 if (C.isEOF() || isNewlineChar(C.peek())) {
150 ErrorCallback(
151 C.location(),
152 "end of machine instruction reached before the closing '\"'");
153 return None;
154 }
155 }
156 C.advance();
157 return C;
158 }
159
lexName(Cursor C,MIToken & Token,MIToken::TokenKind Type,unsigned PrefixLength,ErrorCallbackType ErrorCallback)160 static Cursor lexName(Cursor C, MIToken &Token, MIToken::TokenKind Type,
161 unsigned PrefixLength, ErrorCallbackType ErrorCallback) {
162 auto Range = C;
163 C.advance(PrefixLength);
164 if (C.peek() == '"') {
165 if (Cursor R = lexStringConstant(C, ErrorCallback)) {
166 StringRef String = Range.upto(R);
167 Token.reset(Type, String)
168 .setOwnedStringValue(
169 unescapeQuotedString(String.drop_front(PrefixLength)));
170 return R;
171 }
172 Token.reset(MIToken::Error, Range.remaining());
173 return Range;
174 }
175 while (isIdentifierChar(C.peek()))
176 C.advance();
177 Token.reset(Type, Range.upto(C))
178 .setStringValue(Range.upto(C).drop_front(PrefixLength));
179 return C;
180 }
181
getIdentifierKind(StringRef Identifier)182 static MIToken::TokenKind getIdentifierKind(StringRef Identifier) {
183 return StringSwitch<MIToken::TokenKind>(Identifier)
184 .Case("_", MIToken::underscore)
185 .Case("implicit", MIToken::kw_implicit)
186 .Case("implicit-def", MIToken::kw_implicit_define)
187 .Case("def", MIToken::kw_def)
188 .Case("dead", MIToken::kw_dead)
189 .Case("killed", MIToken::kw_killed)
190 .Case("undef", MIToken::kw_undef)
191 .Case("internal", MIToken::kw_internal)
192 .Case("early-clobber", MIToken::kw_early_clobber)
193 .Case("debug-use", MIToken::kw_debug_use)
194 .Case("renamable", MIToken::kw_renamable)
195 .Case("tied-def", MIToken::kw_tied_def)
196 .Case("frame-setup", MIToken::kw_frame_setup)
197 .Case("frame-destroy", MIToken::kw_frame_destroy)
198 .Case("nnan", MIToken::kw_nnan)
199 .Case("ninf", MIToken::kw_ninf)
200 .Case("nsz", MIToken::kw_nsz)
201 .Case("arcp", MIToken::kw_arcp)
202 .Case("contract", MIToken::kw_contract)
203 .Case("afn", MIToken::kw_afn)
204 .Case("reassoc", MIToken::kw_reassoc)
205 .Case("debug-location", MIToken::kw_debug_location)
206 .Case("same_value", MIToken::kw_cfi_same_value)
207 .Case("offset", MIToken::kw_cfi_offset)
208 .Case("rel_offset", MIToken::kw_cfi_rel_offset)
209 .Case("def_cfa_register", MIToken::kw_cfi_def_cfa_register)
210 .Case("def_cfa_offset", MIToken::kw_cfi_def_cfa_offset)
211 .Case("adjust_cfa_offset", MIToken::kw_cfi_adjust_cfa_offset)
212 .Case("escape", MIToken::kw_cfi_escape)
213 .Case("def_cfa", MIToken::kw_cfi_def_cfa)
214 .Case("remember_state", MIToken::kw_cfi_remember_state)
215 .Case("restore", MIToken::kw_cfi_restore)
216 .Case("restore_state", MIToken::kw_cfi_restore_state)
217 .Case("undefined", MIToken::kw_cfi_undefined)
218 .Case("register", MIToken::kw_cfi_register)
219 .Case("window_save", MIToken::kw_cfi_window_save)
220 .Case("blockaddress", MIToken::kw_blockaddress)
221 .Case("intrinsic", MIToken::kw_intrinsic)
222 .Case("target-index", MIToken::kw_target_index)
223 .Case("half", MIToken::kw_half)
224 .Case("float", MIToken::kw_float)
225 .Case("double", MIToken::kw_double)
226 .Case("x86_fp80", MIToken::kw_x86_fp80)
227 .Case("fp128", MIToken::kw_fp128)
228 .Case("ppc_fp128", MIToken::kw_ppc_fp128)
229 .Case("target-flags", MIToken::kw_target_flags)
230 .Case("volatile", MIToken::kw_volatile)
231 .Case("non-temporal", MIToken::kw_non_temporal)
232 .Case("dereferenceable", MIToken::kw_dereferenceable)
233 .Case("invariant", MIToken::kw_invariant)
234 .Case("align", MIToken::kw_align)
235 .Case("addrspace", MIToken::kw_addrspace)
236 .Case("stack", MIToken::kw_stack)
237 .Case("got", MIToken::kw_got)
238 .Case("jump-table", MIToken::kw_jump_table)
239 .Case("constant-pool", MIToken::kw_constant_pool)
240 .Case("call-entry", MIToken::kw_call_entry)
241 .Case("liveout", MIToken::kw_liveout)
242 .Case("address-taken", MIToken::kw_address_taken)
243 .Case("landing-pad", MIToken::kw_landing_pad)
244 .Case("liveins", MIToken::kw_liveins)
245 .Case("successors", MIToken::kw_successors)
246 .Case("floatpred", MIToken::kw_floatpred)
247 .Case("intpred", MIToken::kw_intpred)
248 .Default(MIToken::Identifier);
249 }
250
maybeLexIdentifier(Cursor C,MIToken & Token)251 static Cursor maybeLexIdentifier(Cursor C, MIToken &Token) {
252 if (!isalpha(C.peek()) && C.peek() != '_')
253 return None;
254 auto Range = C;
255 while (isIdentifierChar(C.peek()))
256 C.advance();
257 auto Identifier = Range.upto(C);
258 Token.reset(getIdentifierKind(Identifier), Identifier)
259 .setStringValue(Identifier);
260 return C;
261 }
262
maybeLexMachineBasicBlock(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)263 static Cursor maybeLexMachineBasicBlock(Cursor C, MIToken &Token,
264 ErrorCallbackType ErrorCallback) {
265 bool IsReference = C.remaining().startswith("%bb.");
266 if (!IsReference && !C.remaining().startswith("bb."))
267 return None;
268 auto Range = C;
269 unsigned PrefixLength = IsReference ? 4 : 3;
270 C.advance(PrefixLength); // Skip '%bb.' or 'bb.'
271 if (!isdigit(C.peek())) {
272 Token.reset(MIToken::Error, C.remaining());
273 ErrorCallback(C.location(), "expected a number after '%bb.'");
274 return C;
275 }
276 auto NumberRange = C;
277 while (isdigit(C.peek()))
278 C.advance();
279 StringRef Number = NumberRange.upto(C);
280 unsigned StringOffset = PrefixLength + Number.size(); // Drop '%bb.<id>'
281 // TODO: The format bb.<id>.<irname> is supported only when it's not a
282 // reference. Once we deprecate the format where the irname shows up, we
283 // should only lex forward if it is a reference.
284 if (C.peek() == '.') {
285 C.advance(); // Skip '.'
286 ++StringOffset;
287 while (isIdentifierChar(C.peek()))
288 C.advance();
289 }
290 Token.reset(IsReference ? MIToken::MachineBasicBlock
291 : MIToken::MachineBasicBlockLabel,
292 Range.upto(C))
293 .setIntegerValue(APSInt(Number))
294 .setStringValue(Range.upto(C).drop_front(StringOffset));
295 return C;
296 }
297
maybeLexIndex(Cursor C,MIToken & Token,StringRef Rule,MIToken::TokenKind Kind)298 static Cursor maybeLexIndex(Cursor C, MIToken &Token, StringRef Rule,
299 MIToken::TokenKind Kind) {
300 if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
301 return None;
302 auto Range = C;
303 C.advance(Rule.size());
304 auto NumberRange = C;
305 while (isdigit(C.peek()))
306 C.advance();
307 Token.reset(Kind, Range.upto(C)).setIntegerValue(APSInt(NumberRange.upto(C)));
308 return C;
309 }
310
maybeLexIndexAndName(Cursor C,MIToken & Token,StringRef Rule,MIToken::TokenKind Kind)311 static Cursor maybeLexIndexAndName(Cursor C, MIToken &Token, StringRef Rule,
312 MIToken::TokenKind Kind) {
313 if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
314 return None;
315 auto Range = C;
316 C.advance(Rule.size());
317 auto NumberRange = C;
318 while (isdigit(C.peek()))
319 C.advance();
320 StringRef Number = NumberRange.upto(C);
321 unsigned StringOffset = Rule.size() + Number.size();
322 if (C.peek() == '.') {
323 C.advance();
324 ++StringOffset;
325 while (isIdentifierChar(C.peek()))
326 C.advance();
327 }
328 Token.reset(Kind, Range.upto(C))
329 .setIntegerValue(APSInt(Number))
330 .setStringValue(Range.upto(C).drop_front(StringOffset));
331 return C;
332 }
333
maybeLexJumpTableIndex(Cursor C,MIToken & Token)334 static Cursor maybeLexJumpTableIndex(Cursor C, MIToken &Token) {
335 return maybeLexIndex(C, Token, "%jump-table.", MIToken::JumpTableIndex);
336 }
337
maybeLexStackObject(Cursor C,MIToken & Token)338 static Cursor maybeLexStackObject(Cursor C, MIToken &Token) {
339 return maybeLexIndexAndName(C, Token, "%stack.", MIToken::StackObject);
340 }
341
maybeLexFixedStackObject(Cursor C,MIToken & Token)342 static Cursor maybeLexFixedStackObject(Cursor C, MIToken &Token) {
343 return maybeLexIndex(C, Token, "%fixed-stack.", MIToken::FixedStackObject);
344 }
345
maybeLexConstantPoolItem(Cursor C,MIToken & Token)346 static Cursor maybeLexConstantPoolItem(Cursor C, MIToken &Token) {
347 return maybeLexIndex(C, Token, "%const.", MIToken::ConstantPoolItem);
348 }
349
maybeLexSubRegisterIndex(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)350 static Cursor maybeLexSubRegisterIndex(Cursor C, MIToken &Token,
351 ErrorCallbackType ErrorCallback) {
352 const StringRef Rule = "%subreg.";
353 if (!C.remaining().startswith(Rule))
354 return None;
355 return lexName(C, Token, MIToken::SubRegisterIndex, Rule.size(),
356 ErrorCallback);
357 }
358
maybeLexIRBlock(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)359 static Cursor maybeLexIRBlock(Cursor C, MIToken &Token,
360 ErrorCallbackType ErrorCallback) {
361 const StringRef Rule = "%ir-block.";
362 if (!C.remaining().startswith(Rule))
363 return None;
364 if (isdigit(C.peek(Rule.size())))
365 return maybeLexIndex(C, Token, Rule, MIToken::IRBlock);
366 return lexName(C, Token, MIToken::NamedIRBlock, Rule.size(), ErrorCallback);
367 }
368
maybeLexIRValue(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)369 static Cursor maybeLexIRValue(Cursor C, MIToken &Token,
370 ErrorCallbackType ErrorCallback) {
371 const StringRef Rule = "%ir.";
372 if (!C.remaining().startswith(Rule))
373 return None;
374 if (isdigit(C.peek(Rule.size())))
375 return maybeLexIndex(C, Token, Rule, MIToken::IRValue);
376 return lexName(C, Token, MIToken::NamedIRValue, Rule.size(), ErrorCallback);
377 }
378
maybeLexStringConstant(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)379 static Cursor maybeLexStringConstant(Cursor C, MIToken &Token,
380 ErrorCallbackType ErrorCallback) {
381 if (C.peek() != '"')
382 return None;
383 return lexName(C, Token, MIToken::StringConstant, /*PrefixLength=*/0,
384 ErrorCallback);
385 }
386
lexVirtualRegister(Cursor C,MIToken & Token)387 static Cursor lexVirtualRegister(Cursor C, MIToken &Token) {
388 auto Range = C;
389 C.advance(); // Skip '%'
390 auto NumberRange = C;
391 while (isdigit(C.peek()))
392 C.advance();
393 Token.reset(MIToken::VirtualRegister, Range.upto(C))
394 .setIntegerValue(APSInt(NumberRange.upto(C)));
395 return C;
396 }
397
398 /// Returns true for a character allowed in a register name.
isRegisterChar(char C)399 static bool isRegisterChar(char C) {
400 return isIdentifierChar(C) && C != '.';
401 }
402
lexNamedVirtualRegister(Cursor C,MIToken & Token)403 static Cursor lexNamedVirtualRegister(Cursor C, MIToken &Token) {
404 Cursor Range = C;
405 C.advance(); // Skip '%'
406 while (isRegisterChar(C.peek()))
407 C.advance();
408 Token.reset(MIToken::NamedVirtualRegister, Range.upto(C))
409 .setStringValue(Range.upto(C).drop_front(1)); // Drop the '%'
410 return C;
411 }
412
maybeLexRegister(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)413 static Cursor maybeLexRegister(Cursor C, MIToken &Token,
414 ErrorCallbackType ErrorCallback) {
415 if (C.peek() != '%' && C.peek() != '$')
416 return None;
417
418 if (C.peek() == '%') {
419 if (isdigit(C.peek(1)))
420 return lexVirtualRegister(C, Token);
421
422 if (isRegisterChar(C.peek(1)))
423 return lexNamedVirtualRegister(C, Token);
424
425 return None;
426 }
427
428 assert(C.peek() == '$');
429 auto Range = C;
430 C.advance(); // Skip '$'
431 while (isRegisterChar(C.peek()))
432 C.advance();
433 Token.reset(MIToken::NamedRegister, Range.upto(C))
434 .setStringValue(Range.upto(C).drop_front(1)); // Drop the '$'
435 return C;
436 }
437
maybeLexGlobalValue(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)438 static Cursor maybeLexGlobalValue(Cursor C, MIToken &Token,
439 ErrorCallbackType ErrorCallback) {
440 if (C.peek() != '@')
441 return None;
442 if (!isdigit(C.peek(1)))
443 return lexName(C, Token, MIToken::NamedGlobalValue, /*PrefixLength=*/1,
444 ErrorCallback);
445 auto Range = C;
446 C.advance(1); // Skip the '@'
447 auto NumberRange = C;
448 while (isdigit(C.peek()))
449 C.advance();
450 Token.reset(MIToken::GlobalValue, Range.upto(C))
451 .setIntegerValue(APSInt(NumberRange.upto(C)));
452 return C;
453 }
454
maybeLexExternalSymbol(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)455 static Cursor maybeLexExternalSymbol(Cursor C, MIToken &Token,
456 ErrorCallbackType ErrorCallback) {
457 if (C.peek() != '&')
458 return None;
459 return lexName(C, Token, MIToken::ExternalSymbol, /*PrefixLength=*/1,
460 ErrorCallback);
461 }
462
isValidHexFloatingPointPrefix(char C)463 static bool isValidHexFloatingPointPrefix(char C) {
464 return C == 'H' || C == 'K' || C == 'L' || C == 'M';
465 }
466
lexFloatingPointLiteral(Cursor Range,Cursor C,MIToken & Token)467 static Cursor lexFloatingPointLiteral(Cursor Range, Cursor C, MIToken &Token) {
468 C.advance();
469 // Skip over [0-9]*([eE][-+]?[0-9]+)?
470 while (isdigit(C.peek()))
471 C.advance();
472 if ((C.peek() == 'e' || C.peek() == 'E') &&
473 (isdigit(C.peek(1)) ||
474 ((C.peek(1) == '-' || C.peek(1) == '+') && isdigit(C.peek(2))))) {
475 C.advance(2);
476 while (isdigit(C.peek()))
477 C.advance();
478 }
479 Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
480 return C;
481 }
482
maybeLexHexadecimalLiteral(Cursor C,MIToken & Token)483 static Cursor maybeLexHexadecimalLiteral(Cursor C, MIToken &Token) {
484 if (C.peek() != '0' || (C.peek(1) != 'x' && C.peek(1) != 'X'))
485 return None;
486 Cursor Range = C;
487 C.advance(2);
488 unsigned PrefLen = 2;
489 if (isValidHexFloatingPointPrefix(C.peek())) {
490 C.advance();
491 PrefLen++;
492 }
493 while (isxdigit(C.peek()))
494 C.advance();
495 StringRef StrVal = Range.upto(C);
496 if (StrVal.size() <= PrefLen)
497 return None;
498 if (PrefLen == 2)
499 Token.reset(MIToken::HexLiteral, Range.upto(C));
500 else // It must be 3, which means that there was a floating-point prefix.
501 Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
502 return C;
503 }
504
maybeLexNumericalLiteral(Cursor C,MIToken & Token)505 static Cursor maybeLexNumericalLiteral(Cursor C, MIToken &Token) {
506 if (!isdigit(C.peek()) && (C.peek() != '-' || !isdigit(C.peek(1))))
507 return None;
508 auto Range = C;
509 C.advance();
510 while (isdigit(C.peek()))
511 C.advance();
512 if (C.peek() == '.')
513 return lexFloatingPointLiteral(Range, C, Token);
514 StringRef StrVal = Range.upto(C);
515 Token.reset(MIToken::IntegerLiteral, StrVal).setIntegerValue(APSInt(StrVal));
516 return C;
517 }
518
getMetadataKeywordKind(StringRef Identifier)519 static MIToken::TokenKind getMetadataKeywordKind(StringRef Identifier) {
520 return StringSwitch<MIToken::TokenKind>(Identifier)
521 .Case("!tbaa", MIToken::md_tbaa)
522 .Case("!alias.scope", MIToken::md_alias_scope)
523 .Case("!noalias", MIToken::md_noalias)
524 .Case("!range", MIToken::md_range)
525 .Case("!DIExpression", MIToken::md_diexpr)
526 .Default(MIToken::Error);
527 }
528
maybeLexExlaim(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)529 static Cursor maybeLexExlaim(Cursor C, MIToken &Token,
530 ErrorCallbackType ErrorCallback) {
531 if (C.peek() != '!')
532 return None;
533 auto Range = C;
534 C.advance(1);
535 if (isdigit(C.peek()) || !isIdentifierChar(C.peek())) {
536 Token.reset(MIToken::exclaim, Range.upto(C));
537 return C;
538 }
539 while (isIdentifierChar(C.peek()))
540 C.advance();
541 StringRef StrVal = Range.upto(C);
542 Token.reset(getMetadataKeywordKind(StrVal), StrVal);
543 if (Token.isError())
544 ErrorCallback(Token.location(),
545 "use of unknown metadata keyword '" + StrVal + "'");
546 return C;
547 }
548
symbolToken(char C)549 static MIToken::TokenKind symbolToken(char C) {
550 switch (C) {
551 case ',':
552 return MIToken::comma;
553 case '.':
554 return MIToken::dot;
555 case '=':
556 return MIToken::equal;
557 case ':':
558 return MIToken::colon;
559 case '(':
560 return MIToken::lparen;
561 case ')':
562 return MIToken::rparen;
563 case '{':
564 return MIToken::lbrace;
565 case '}':
566 return MIToken::rbrace;
567 case '+':
568 return MIToken::plus;
569 case '-':
570 return MIToken::minus;
571 case '<':
572 return MIToken::less;
573 case '>':
574 return MIToken::greater;
575 default:
576 return MIToken::Error;
577 }
578 }
579
maybeLexSymbol(Cursor C,MIToken & Token)580 static Cursor maybeLexSymbol(Cursor C, MIToken &Token) {
581 MIToken::TokenKind Kind;
582 unsigned Length = 1;
583 if (C.peek() == ':' && C.peek(1) == ':') {
584 Kind = MIToken::coloncolon;
585 Length = 2;
586 } else
587 Kind = symbolToken(C.peek());
588 if (Kind == MIToken::Error)
589 return None;
590 auto Range = C;
591 C.advance(Length);
592 Token.reset(Kind, Range.upto(C));
593 return C;
594 }
595
maybeLexNewline(Cursor C,MIToken & Token)596 static Cursor maybeLexNewline(Cursor C, MIToken &Token) {
597 if (!isNewlineChar(C.peek()))
598 return None;
599 auto Range = C;
600 C.advance();
601 Token.reset(MIToken::Newline, Range.upto(C));
602 return C;
603 }
604
maybeLexEscapedIRValue(Cursor C,MIToken & Token,ErrorCallbackType ErrorCallback)605 static Cursor maybeLexEscapedIRValue(Cursor C, MIToken &Token,
606 ErrorCallbackType ErrorCallback) {
607 if (C.peek() != '`')
608 return None;
609 auto Range = C;
610 C.advance();
611 auto StrRange = C;
612 while (C.peek() != '`') {
613 if (C.isEOF() || isNewlineChar(C.peek())) {
614 ErrorCallback(
615 C.location(),
616 "end of machine instruction reached before the closing '`'");
617 Token.reset(MIToken::Error, Range.remaining());
618 return C;
619 }
620 C.advance();
621 }
622 StringRef Value = StrRange.upto(C);
623 C.advance();
624 Token.reset(MIToken::QuotedIRValue, Range.upto(C)).setStringValue(Value);
625 return C;
626 }
627
lexMIToken(StringRef Source,MIToken & Token,ErrorCallbackType ErrorCallback)628 StringRef llvm::lexMIToken(StringRef Source, MIToken &Token,
629 ErrorCallbackType ErrorCallback) {
630 auto C = skipComment(skipWhitespace(Cursor(Source)));
631 if (C.isEOF()) {
632 Token.reset(MIToken::Eof, C.remaining());
633 return C.remaining();
634 }
635
636 if (Cursor R = maybeLexMachineBasicBlock(C, Token, ErrorCallback))
637 return R.remaining();
638 if (Cursor R = maybeLexIdentifier(C, Token))
639 return R.remaining();
640 if (Cursor R = maybeLexJumpTableIndex(C, Token))
641 return R.remaining();
642 if (Cursor R = maybeLexStackObject(C, Token))
643 return R.remaining();
644 if (Cursor R = maybeLexFixedStackObject(C, Token))
645 return R.remaining();
646 if (Cursor R = maybeLexConstantPoolItem(C, Token))
647 return R.remaining();
648 if (Cursor R = maybeLexSubRegisterIndex(C, Token, ErrorCallback))
649 return R.remaining();
650 if (Cursor R = maybeLexIRBlock(C, Token, ErrorCallback))
651 return R.remaining();
652 if (Cursor R = maybeLexIRValue(C, Token, ErrorCallback))
653 return R.remaining();
654 if (Cursor R = maybeLexRegister(C, Token, ErrorCallback))
655 return R.remaining();
656 if (Cursor R = maybeLexGlobalValue(C, Token, ErrorCallback))
657 return R.remaining();
658 if (Cursor R = maybeLexExternalSymbol(C, Token, ErrorCallback))
659 return R.remaining();
660 if (Cursor R = maybeLexHexadecimalLiteral(C, Token))
661 return R.remaining();
662 if (Cursor R = maybeLexNumericalLiteral(C, Token))
663 return R.remaining();
664 if (Cursor R = maybeLexExlaim(C, Token, ErrorCallback))
665 return R.remaining();
666 if (Cursor R = maybeLexSymbol(C, Token))
667 return R.remaining();
668 if (Cursor R = maybeLexNewline(C, Token))
669 return R.remaining();
670 if (Cursor R = maybeLexEscapedIRValue(C, Token, ErrorCallback))
671 return R.remaining();
672 if (Cursor R = maybeLexStringConstant(C, Token, ErrorCallback))
673 return R.remaining();
674
675 Token.reset(MIToken::Error, C.remaining());
676 ErrorCallback(C.location(),
677 Twine("unexpected character '") + Twine(C.peek()) + "'");
678 return C.remaining();
679 }
680