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:
41   Cursor(NoneType) {}
42 
43   explicit Cursor(StringRef Str) {
44     Ptr = Str.data();
45     End = Ptr + Str.size();
46   }
47 
48   bool isEOF() const { return Ptr == End; }
49 
50   char peek(int I = 0) const { return End - Ptr <= I ? 0 : Ptr[I]; }
51 
52   void advance(unsigned I = 1) { Ptr += I; }
53 
54   StringRef remaining() const { return StringRef(Ptr, End - Ptr); }
55 
56   StringRef upto(Cursor C) const {
57     assert(C.Ptr >= Ptr && C.Ptr <= End);
58     return StringRef(Ptr, C.Ptr - Ptr);
59   }
60 
61   StringRef::iterator location() const { return Ptr; }
62 
63   operator bool() const { return Ptr != nullptr; }
64 };
65 
66 } // end anonymous namespace
67 
68 MIToken &MIToken::reset(TokenKind Kind, StringRef Range) {
69   this->Kind = Kind;
70   this->Range = Range;
71   return *this;
72 }
73 
74 MIToken &MIToken::setStringValue(StringRef StrVal) {
75   StringValue = StrVal;
76   return *this;
77 }
78 
79 MIToken &MIToken::setOwnedStringValue(std::string StrVal) {
80   StringValueStorage = std::move(StrVal);
81   StringValue = StringValueStorage;
82   return *this;
83 }
84 
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.
91 static Cursor skipWhitespace(Cursor C) {
92   while (isblank(C.peek()))
93     C.advance();
94   return C;
95 }
96 
97 static bool isNewlineChar(char C) { return C == '\n' || C == '\r'; }
98 
99 /// Skip a line comment and return the updated cursor.
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$._0-9]
110 static bool isIdentifierChar(char C) {
111   return isalpha(C) || isdigit(C) || C == '_' || C == '-' || C == '.' ||
112          C == '$';
113 }
114 
115 /// Unescapes the given string value.
116 ///
117 /// Expects the string value to be quoted.
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: \"[^\"]*\"
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 
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 
182 static Cursor maybeLexIntegerOrScalarType(Cursor C, MIToken &Token) {
183   if ((C.peek() != 'i' && C.peek() != 's' && C.peek() != 'p') ||
184       !isdigit(C.peek(1)))
185     return None;
186   char Kind = C.peek();
187   auto Range = C;
188   C.advance(); // Skip 'i', 's', or 'p'
189   while (isdigit(C.peek()))
190     C.advance();
191 
192   Token.reset(Kind == 'i'
193                   ? MIToken::IntegerType
194                   : (Kind == 's' ? MIToken::ScalarType : MIToken::PointerType),
195               Range.upto(C));
196   return C;
197 }
198 
199 static MIToken::TokenKind getIdentifierKind(StringRef Identifier) {
200   return StringSwitch<MIToken::TokenKind>(Identifier)
201       .Case("_", MIToken::underscore)
202       .Case("implicit", MIToken::kw_implicit)
203       .Case("implicit-def", MIToken::kw_implicit_define)
204       .Case("def", MIToken::kw_def)
205       .Case("dead", MIToken::kw_dead)
206       .Case("killed", MIToken::kw_killed)
207       .Case("undef", MIToken::kw_undef)
208       .Case("internal", MIToken::kw_internal)
209       .Case("early-clobber", MIToken::kw_early_clobber)
210       .Case("debug-use", MIToken::kw_debug_use)
211       .Case("renamable", MIToken::kw_renamable)
212       .Case("tied-def", MIToken::kw_tied_def)
213       .Case("frame-setup", MIToken::kw_frame_setup)
214       .Case("frame-destroy", MIToken::kw_frame_destroy)
215       .Case("debug-location", MIToken::kw_debug_location)
216       .Case("same_value", MIToken::kw_cfi_same_value)
217       .Case("offset", MIToken::kw_cfi_offset)
218       .Case("rel_offset", MIToken::kw_cfi_rel_offset)
219       .Case("def_cfa_register", MIToken::kw_cfi_def_cfa_register)
220       .Case("def_cfa_offset", MIToken::kw_cfi_def_cfa_offset)
221       .Case("adjust_cfa_offset", MIToken::kw_cfi_adjust_cfa_offset)
222       .Case("escape", MIToken::kw_cfi_escape)
223       .Case("def_cfa", MIToken::kw_cfi_def_cfa)
224       .Case("remember_state", MIToken::kw_cfi_remember_state)
225       .Case("restore", MIToken::kw_cfi_restore)
226       .Case("restore_state", MIToken::kw_cfi_restore_state)
227       .Case("undefined", MIToken::kw_cfi_undefined)
228       .Case("register", MIToken::kw_cfi_register)
229       .Case("window_save", MIToken::kw_cfi_window_save)
230       .Case("blockaddress", MIToken::kw_blockaddress)
231       .Case("intrinsic", MIToken::kw_intrinsic)
232       .Case("target-index", MIToken::kw_target_index)
233       .Case("half", MIToken::kw_half)
234       .Case("float", MIToken::kw_float)
235       .Case("double", MIToken::kw_double)
236       .Case("x86_fp80", MIToken::kw_x86_fp80)
237       .Case("fp128", MIToken::kw_fp128)
238       .Case("ppc_fp128", MIToken::kw_ppc_fp128)
239       .Case("target-flags", MIToken::kw_target_flags)
240       .Case("volatile", MIToken::kw_volatile)
241       .Case("non-temporal", MIToken::kw_non_temporal)
242       .Case("dereferenceable", MIToken::kw_dereferenceable)
243       .Case("invariant", MIToken::kw_invariant)
244       .Case("align", MIToken::kw_align)
245       .Case("addrspace", MIToken::kw_addrspace)
246       .Case("stack", MIToken::kw_stack)
247       .Case("got", MIToken::kw_got)
248       .Case("jump-table", MIToken::kw_jump_table)
249       .Case("constant-pool", MIToken::kw_constant_pool)
250       .Case("call-entry", MIToken::kw_call_entry)
251       .Case("liveout", MIToken::kw_liveout)
252       .Case("address-taken", MIToken::kw_address_taken)
253       .Case("landing-pad", MIToken::kw_landing_pad)
254       .Case("liveins", MIToken::kw_liveins)
255       .Case("successors", MIToken::kw_successors)
256       .Case("floatpred", MIToken::kw_floatpred)
257       .Case("intpred", MIToken::kw_intpred)
258       .Default(MIToken::Identifier);
259 }
260 
261 static Cursor maybeLexIdentifier(Cursor C, MIToken &Token) {
262   if (!isalpha(C.peek()) && C.peek() != '_')
263     return None;
264   auto Range = C;
265   while (isIdentifierChar(C.peek()))
266     C.advance();
267   auto Identifier = Range.upto(C);
268   Token.reset(getIdentifierKind(Identifier), Identifier)
269       .setStringValue(Identifier);
270   return C;
271 }
272 
273 static Cursor maybeLexMachineBasicBlock(Cursor C, MIToken &Token,
274                                         ErrorCallbackType ErrorCallback) {
275   bool IsReference = C.remaining().startswith("%bb.");
276   if (!IsReference && !C.remaining().startswith("bb."))
277     return None;
278   auto Range = C;
279   unsigned PrefixLength = IsReference ? 4 : 3;
280   C.advance(PrefixLength); // Skip '%bb.' or 'bb.'
281   if (!isdigit(C.peek())) {
282     Token.reset(MIToken::Error, C.remaining());
283     ErrorCallback(C.location(), "expected a number after '%bb.'");
284     return C;
285   }
286   auto NumberRange = C;
287   while (isdigit(C.peek()))
288     C.advance();
289   StringRef Number = NumberRange.upto(C);
290   unsigned StringOffset = PrefixLength + Number.size(); // Drop '%bb.<id>'
291   // TODO: The format bb.<id>.<irname> is supported only when it's not a
292   // reference. Once we deprecate the format where the irname shows up, we
293   // should only lex forward if it is a reference.
294   if (C.peek() == '.') {
295     C.advance(); // Skip '.'
296     ++StringOffset;
297     while (isIdentifierChar(C.peek()))
298       C.advance();
299   }
300   Token.reset(IsReference ? MIToken::MachineBasicBlock
301                           : MIToken::MachineBasicBlockLabel,
302               Range.upto(C))
303       .setIntegerValue(APSInt(Number))
304       .setStringValue(Range.upto(C).drop_front(StringOffset));
305   return C;
306 }
307 
308 static Cursor maybeLexIndex(Cursor C, MIToken &Token, StringRef Rule,
309                             MIToken::TokenKind Kind) {
310   if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
311     return None;
312   auto Range = C;
313   C.advance(Rule.size());
314   auto NumberRange = C;
315   while (isdigit(C.peek()))
316     C.advance();
317   Token.reset(Kind, Range.upto(C)).setIntegerValue(APSInt(NumberRange.upto(C)));
318   return C;
319 }
320 
321 static Cursor maybeLexIndexAndName(Cursor C, MIToken &Token, StringRef Rule,
322                                    MIToken::TokenKind Kind) {
323   if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
324     return None;
325   auto Range = C;
326   C.advance(Rule.size());
327   auto NumberRange = C;
328   while (isdigit(C.peek()))
329     C.advance();
330   StringRef Number = NumberRange.upto(C);
331   unsigned StringOffset = Rule.size() + Number.size();
332   if (C.peek() == '.') {
333     C.advance();
334     ++StringOffset;
335     while (isIdentifierChar(C.peek()))
336       C.advance();
337   }
338   Token.reset(Kind, Range.upto(C))
339       .setIntegerValue(APSInt(Number))
340       .setStringValue(Range.upto(C).drop_front(StringOffset));
341   return C;
342 }
343 
344 static Cursor maybeLexJumpTableIndex(Cursor C, MIToken &Token) {
345   return maybeLexIndex(C, Token, "%jump-table.", MIToken::JumpTableIndex);
346 }
347 
348 static Cursor maybeLexStackObject(Cursor C, MIToken &Token) {
349   return maybeLexIndexAndName(C, Token, "%stack.", MIToken::StackObject);
350 }
351 
352 static Cursor maybeLexFixedStackObject(Cursor C, MIToken &Token) {
353   return maybeLexIndex(C, Token, "%fixed-stack.", MIToken::FixedStackObject);
354 }
355 
356 static Cursor maybeLexConstantPoolItem(Cursor C, MIToken &Token) {
357   return maybeLexIndex(C, Token, "%const.", MIToken::ConstantPoolItem);
358 }
359 
360 static Cursor maybeLexSubRegisterIndex(Cursor C, MIToken &Token,
361                                        ErrorCallbackType ErrorCallback) {
362   const StringRef Rule = "%subreg.";
363   if (!C.remaining().startswith(Rule))
364     return None;
365   return lexName(C, Token, MIToken::SubRegisterIndex, Rule.size(),
366                  ErrorCallback);
367 }
368 
369 static Cursor maybeLexIRBlock(Cursor C, MIToken &Token,
370                               ErrorCallbackType ErrorCallback) {
371   const StringRef Rule = "%ir-block.";
372   if (!C.remaining().startswith(Rule))
373     return None;
374   if (isdigit(C.peek(Rule.size())))
375     return maybeLexIndex(C, Token, Rule, MIToken::IRBlock);
376   return lexName(C, Token, MIToken::NamedIRBlock, Rule.size(), ErrorCallback);
377 }
378 
379 static Cursor maybeLexIRValue(Cursor C, MIToken &Token,
380                               ErrorCallbackType ErrorCallback) {
381   const StringRef Rule = "%ir.";
382   if (!C.remaining().startswith(Rule))
383     return None;
384   if (isdigit(C.peek(Rule.size())))
385     return maybeLexIndex(C, Token, Rule, MIToken::IRValue);
386   return lexName(C, Token, MIToken::NamedIRValue, Rule.size(), ErrorCallback);
387 }
388 
389 static Cursor maybeLexStringConstant(Cursor C, MIToken &Token,
390                                      ErrorCallbackType ErrorCallback) {
391   if (C.peek() != '"')
392     return None;
393   return lexName(C, Token, MIToken::StringConstant, /*PrefixLength=*/0,
394                  ErrorCallback);
395 }
396 
397 static Cursor lexVirtualRegister(Cursor C, MIToken &Token) {
398   auto Range = C;
399   C.advance(); // Skip '%'
400   auto NumberRange = C;
401   while (isdigit(C.peek()))
402     C.advance();
403   Token.reset(MIToken::VirtualRegister, Range.upto(C))
404       .setIntegerValue(APSInt(NumberRange.upto(C)));
405   return C;
406 }
407 
408 /// Returns true for a character allowed in a register name.
409 static bool isRegisterChar(char C) {
410   return isIdentifierChar(C) && C != '.';
411 }
412 
413 static Cursor lexNamedVirtualRegister(Cursor C, MIToken &Token) {
414   Cursor Range = C;
415   C.advance(); // Skip '%'
416   while (isRegisterChar(C.peek()))
417     C.advance();
418   Token.reset(MIToken::NamedVirtualRegister, Range.upto(C))
419       .setStringValue(Range.upto(C).drop_front(1)); // Drop the '%'
420   return C;
421 }
422 
423 static Cursor maybeLexRegister(Cursor C, MIToken &Token,
424                                ErrorCallbackType ErrorCallback) {
425   if (C.peek() != '%' && C.peek() != '$')
426     return None;
427 
428   if (C.peek() == '%') {
429     if (isdigit(C.peek(1)))
430       return lexVirtualRegister(C, Token);
431 
432     if (isRegisterChar(C.peek(1)))
433       return lexNamedVirtualRegister(C, Token);
434 
435     return None;
436   }
437 
438   assert(C.peek() == '$');
439   auto Range = C;
440   C.advance(); // Skip '$'
441   while (isRegisterChar(C.peek()))
442     C.advance();
443   Token.reset(MIToken::NamedRegister, Range.upto(C))
444       .setStringValue(Range.upto(C).drop_front(1)); // Drop the '$'
445   return C;
446 }
447 
448 static Cursor maybeLexGlobalValue(Cursor C, MIToken &Token,
449                                   ErrorCallbackType ErrorCallback) {
450   if (C.peek() != '@')
451     return None;
452   if (!isdigit(C.peek(1)))
453     return lexName(C, Token, MIToken::NamedGlobalValue, /*PrefixLength=*/1,
454                    ErrorCallback);
455   auto Range = C;
456   C.advance(1); // Skip the '@'
457   auto NumberRange = C;
458   while (isdigit(C.peek()))
459     C.advance();
460   Token.reset(MIToken::GlobalValue, Range.upto(C))
461       .setIntegerValue(APSInt(NumberRange.upto(C)));
462   return C;
463 }
464 
465 static Cursor maybeLexExternalSymbol(Cursor C, MIToken &Token,
466                                      ErrorCallbackType ErrorCallback) {
467   if (C.peek() != '&')
468     return None;
469   return lexName(C, Token, MIToken::ExternalSymbol, /*PrefixLength=*/1,
470                  ErrorCallback);
471 }
472 
473 static bool isValidHexFloatingPointPrefix(char C) {
474   return C == 'H' || C == 'K' || C == 'L' || C == 'M';
475 }
476 
477 static Cursor lexFloatingPointLiteral(Cursor Range, Cursor C, MIToken &Token) {
478   C.advance();
479   // Skip over [0-9]*([eE][-+]?[0-9]+)?
480   while (isdigit(C.peek()))
481     C.advance();
482   if ((C.peek() == 'e' || C.peek() == 'E') &&
483       (isdigit(C.peek(1)) ||
484        ((C.peek(1) == '-' || C.peek(1) == '+') && isdigit(C.peek(2))))) {
485     C.advance(2);
486     while (isdigit(C.peek()))
487       C.advance();
488   }
489   Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
490   return C;
491 }
492 
493 static Cursor maybeLexHexadecimalLiteral(Cursor C, MIToken &Token) {
494   if (C.peek() != '0' || (C.peek(1) != 'x' && C.peek(1) != 'X'))
495     return None;
496   Cursor Range = C;
497   C.advance(2);
498   unsigned PrefLen = 2;
499   if (isValidHexFloatingPointPrefix(C.peek())) {
500     C.advance();
501     PrefLen++;
502   }
503   while (isxdigit(C.peek()))
504     C.advance();
505   StringRef StrVal = Range.upto(C);
506   if (StrVal.size() <= PrefLen)
507     return None;
508   if (PrefLen == 2)
509     Token.reset(MIToken::HexLiteral, Range.upto(C));
510   else // It must be 3, which means that there was a floating-point prefix.
511     Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
512   return C;
513 }
514 
515 static Cursor maybeLexNumericalLiteral(Cursor C, MIToken &Token) {
516   if (!isdigit(C.peek()) && (C.peek() != '-' || !isdigit(C.peek(1))))
517     return None;
518   auto Range = C;
519   C.advance();
520   while (isdigit(C.peek()))
521     C.advance();
522   if (C.peek() == '.')
523     return lexFloatingPointLiteral(Range, C, Token);
524   StringRef StrVal = Range.upto(C);
525   Token.reset(MIToken::IntegerLiteral, StrVal).setIntegerValue(APSInt(StrVal));
526   return C;
527 }
528 
529 static MIToken::TokenKind getMetadataKeywordKind(StringRef Identifier) {
530   return StringSwitch<MIToken::TokenKind>(Identifier)
531       .Case("!tbaa", MIToken::md_tbaa)
532       .Case("!alias.scope", MIToken::md_alias_scope)
533       .Case("!noalias", MIToken::md_noalias)
534       .Case("!range", MIToken::md_range)
535       .Case("!DIExpression", MIToken::md_diexpr)
536       .Default(MIToken::Error);
537 }
538 
539 static Cursor maybeLexExlaim(Cursor C, MIToken &Token,
540                              ErrorCallbackType ErrorCallback) {
541   if (C.peek() != '!')
542     return None;
543   auto Range = C;
544   C.advance(1);
545   if (isdigit(C.peek()) || !isIdentifierChar(C.peek())) {
546     Token.reset(MIToken::exclaim, Range.upto(C));
547     return C;
548   }
549   while (isIdentifierChar(C.peek()))
550     C.advance();
551   StringRef StrVal = Range.upto(C);
552   Token.reset(getMetadataKeywordKind(StrVal), StrVal);
553   if (Token.isError())
554     ErrorCallback(Token.location(),
555                   "use of unknown metadata keyword '" + StrVal + "'");
556   return C;
557 }
558 
559 static MIToken::TokenKind symbolToken(char C) {
560   switch (C) {
561   case ',':
562     return MIToken::comma;
563   case '.':
564     return MIToken::dot;
565   case '=':
566     return MIToken::equal;
567   case ':':
568     return MIToken::colon;
569   case '(':
570     return MIToken::lparen;
571   case ')':
572     return MIToken::rparen;
573   case '{':
574     return MIToken::lbrace;
575   case '}':
576     return MIToken::rbrace;
577   case '+':
578     return MIToken::plus;
579   case '-':
580     return MIToken::minus;
581   case '<':
582     return MIToken::less;
583   case '>':
584     return MIToken::greater;
585   default:
586     return MIToken::Error;
587   }
588 }
589 
590 static Cursor maybeLexSymbol(Cursor C, MIToken &Token) {
591   MIToken::TokenKind Kind;
592   unsigned Length = 1;
593   if (C.peek() == ':' && C.peek(1) == ':') {
594     Kind = MIToken::coloncolon;
595     Length = 2;
596   } else
597     Kind = symbolToken(C.peek());
598   if (Kind == MIToken::Error)
599     return None;
600   auto Range = C;
601   C.advance(Length);
602   Token.reset(Kind, Range.upto(C));
603   return C;
604 }
605 
606 static Cursor maybeLexNewline(Cursor C, MIToken &Token) {
607   if (!isNewlineChar(C.peek()))
608     return None;
609   auto Range = C;
610   C.advance();
611   Token.reset(MIToken::Newline, Range.upto(C));
612   return C;
613 }
614 
615 static Cursor maybeLexEscapedIRValue(Cursor C, MIToken &Token,
616                                      ErrorCallbackType ErrorCallback) {
617   if (C.peek() != '`')
618     return None;
619   auto Range = C;
620   C.advance();
621   auto StrRange = C;
622   while (C.peek() != '`') {
623     if (C.isEOF() || isNewlineChar(C.peek())) {
624       ErrorCallback(
625           C.location(),
626           "end of machine instruction reached before the closing '`'");
627       Token.reset(MIToken::Error, Range.remaining());
628       return C;
629     }
630     C.advance();
631   }
632   StringRef Value = StrRange.upto(C);
633   C.advance();
634   Token.reset(MIToken::QuotedIRValue, Range.upto(C)).setStringValue(Value);
635   return C;
636 }
637 
638 StringRef llvm::lexMIToken(StringRef Source, MIToken &Token,
639                            ErrorCallbackType ErrorCallback) {
640   auto C = skipComment(skipWhitespace(Cursor(Source)));
641   if (C.isEOF()) {
642     Token.reset(MIToken::Eof, C.remaining());
643     return C.remaining();
644   }
645 
646   if (Cursor R = maybeLexIntegerOrScalarType(C, Token))
647     return R.remaining();
648   if (Cursor R = maybeLexMachineBasicBlock(C, Token, ErrorCallback))
649     return R.remaining();
650   if (Cursor R = maybeLexIdentifier(C, Token))
651     return R.remaining();
652   if (Cursor R = maybeLexJumpTableIndex(C, Token))
653     return R.remaining();
654   if (Cursor R = maybeLexStackObject(C, Token))
655     return R.remaining();
656   if (Cursor R = maybeLexFixedStackObject(C, Token))
657     return R.remaining();
658   if (Cursor R = maybeLexConstantPoolItem(C, Token))
659     return R.remaining();
660   if (Cursor R = maybeLexSubRegisterIndex(C, Token, ErrorCallback))
661     return R.remaining();
662   if (Cursor R = maybeLexIRBlock(C, Token, ErrorCallback))
663     return R.remaining();
664   if (Cursor R = maybeLexIRValue(C, Token, ErrorCallback))
665     return R.remaining();
666   if (Cursor R = maybeLexRegister(C, Token, ErrorCallback))
667     return R.remaining();
668   if (Cursor R = maybeLexGlobalValue(C, Token, ErrorCallback))
669     return R.remaining();
670   if (Cursor R = maybeLexExternalSymbol(C, Token, ErrorCallback))
671     return R.remaining();
672   if (Cursor R = maybeLexHexadecimalLiteral(C, Token))
673     return R.remaining();
674   if (Cursor R = maybeLexNumericalLiteral(C, Token))
675     return R.remaining();
676   if (Cursor R = maybeLexExlaim(C, Token, ErrorCallback))
677     return R.remaining();
678   if (Cursor R = maybeLexSymbol(C, Token))
679     return R.remaining();
680   if (Cursor R = maybeLexNewline(C, Token))
681     return R.remaining();
682   if (Cursor R = maybeLexEscapedIRValue(C, Token, ErrorCallback))
683     return R.remaining();
684   if (Cursor R = maybeLexStringConstant(C, Token, ErrorCallback))
685     return R.remaining();
686 
687   Token.reset(MIToken::Error, C.remaining());
688   ErrorCallback(C.location(),
689                 Twine("unexpected character '") + Twine(C.peek()) + "'");
690   return C.remaining();
691 }
692