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 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       .Case("pre-instr-symbol", MIToken::kw_pre_instr_symbol)
249       .Case("post-instr-symbol", MIToken::kw_post_instr_symbol)
250       .Case("unknown-size", MIToken::kw_unknown_size)
251       .Default(MIToken::Identifier);
252 }
253 
254 static Cursor maybeLexIdentifier(Cursor C, MIToken &Token) {
255   if (!isalpha(C.peek()) && C.peek() != '_')
256     return None;
257   auto Range = C;
258   while (isIdentifierChar(C.peek()))
259     C.advance();
260   auto Identifier = Range.upto(C);
261   Token.reset(getIdentifierKind(Identifier), Identifier)
262       .setStringValue(Identifier);
263   return C;
264 }
265 
266 static Cursor maybeLexMachineBasicBlock(Cursor C, MIToken &Token,
267                                         ErrorCallbackType ErrorCallback) {
268   bool IsReference = C.remaining().startswith("%bb.");
269   if (!IsReference && !C.remaining().startswith("bb."))
270     return None;
271   auto Range = C;
272   unsigned PrefixLength = IsReference ? 4 : 3;
273   C.advance(PrefixLength); // Skip '%bb.' or 'bb.'
274   if (!isdigit(C.peek())) {
275     Token.reset(MIToken::Error, C.remaining());
276     ErrorCallback(C.location(), "expected a number after '%bb.'");
277     return C;
278   }
279   auto NumberRange = C;
280   while (isdigit(C.peek()))
281     C.advance();
282   StringRef Number = NumberRange.upto(C);
283   unsigned StringOffset = PrefixLength + Number.size(); // Drop '%bb.<id>'
284   // TODO: The format bb.<id>.<irname> is supported only when it's not a
285   // reference. Once we deprecate the format where the irname shows up, we
286   // should only lex forward if it is a reference.
287   if (C.peek() == '.') {
288     C.advance(); // Skip '.'
289     ++StringOffset;
290     while (isIdentifierChar(C.peek()))
291       C.advance();
292   }
293   Token.reset(IsReference ? MIToken::MachineBasicBlock
294                           : MIToken::MachineBasicBlockLabel,
295               Range.upto(C))
296       .setIntegerValue(APSInt(Number))
297       .setStringValue(Range.upto(C).drop_front(StringOffset));
298   return C;
299 }
300 
301 static Cursor maybeLexIndex(Cursor C, MIToken &Token, StringRef Rule,
302                             MIToken::TokenKind Kind) {
303   if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
304     return None;
305   auto Range = C;
306   C.advance(Rule.size());
307   auto NumberRange = C;
308   while (isdigit(C.peek()))
309     C.advance();
310   Token.reset(Kind, Range.upto(C)).setIntegerValue(APSInt(NumberRange.upto(C)));
311   return C;
312 }
313 
314 static Cursor maybeLexIndexAndName(Cursor C, MIToken &Token, StringRef Rule,
315                                    MIToken::TokenKind Kind) {
316   if (!C.remaining().startswith(Rule) || !isdigit(C.peek(Rule.size())))
317     return None;
318   auto Range = C;
319   C.advance(Rule.size());
320   auto NumberRange = C;
321   while (isdigit(C.peek()))
322     C.advance();
323   StringRef Number = NumberRange.upto(C);
324   unsigned StringOffset = Rule.size() + Number.size();
325   if (C.peek() == '.') {
326     C.advance();
327     ++StringOffset;
328     while (isIdentifierChar(C.peek()))
329       C.advance();
330   }
331   Token.reset(Kind, Range.upto(C))
332       .setIntegerValue(APSInt(Number))
333       .setStringValue(Range.upto(C).drop_front(StringOffset));
334   return C;
335 }
336 
337 static Cursor maybeLexJumpTableIndex(Cursor C, MIToken &Token) {
338   return maybeLexIndex(C, Token, "%jump-table.", MIToken::JumpTableIndex);
339 }
340 
341 static Cursor maybeLexStackObject(Cursor C, MIToken &Token) {
342   return maybeLexIndexAndName(C, Token, "%stack.", MIToken::StackObject);
343 }
344 
345 static Cursor maybeLexFixedStackObject(Cursor C, MIToken &Token) {
346   return maybeLexIndex(C, Token, "%fixed-stack.", MIToken::FixedStackObject);
347 }
348 
349 static Cursor maybeLexConstantPoolItem(Cursor C, MIToken &Token) {
350   return maybeLexIndex(C, Token, "%const.", MIToken::ConstantPoolItem);
351 }
352 
353 static Cursor maybeLexSubRegisterIndex(Cursor C, MIToken &Token,
354                                        ErrorCallbackType ErrorCallback) {
355   const StringRef Rule = "%subreg.";
356   if (!C.remaining().startswith(Rule))
357     return None;
358   return lexName(C, Token, MIToken::SubRegisterIndex, Rule.size(),
359                  ErrorCallback);
360 }
361 
362 static Cursor maybeLexIRBlock(Cursor C, MIToken &Token,
363                               ErrorCallbackType ErrorCallback) {
364   const StringRef Rule = "%ir-block.";
365   if (!C.remaining().startswith(Rule))
366     return None;
367   if (isdigit(C.peek(Rule.size())))
368     return maybeLexIndex(C, Token, Rule, MIToken::IRBlock);
369   return lexName(C, Token, MIToken::NamedIRBlock, Rule.size(), ErrorCallback);
370 }
371 
372 static Cursor maybeLexIRValue(Cursor C, MIToken &Token,
373                               ErrorCallbackType ErrorCallback) {
374   const StringRef Rule = "%ir.";
375   if (!C.remaining().startswith(Rule))
376     return None;
377   if (isdigit(C.peek(Rule.size())))
378     return maybeLexIndex(C, Token, Rule, MIToken::IRValue);
379   return lexName(C, Token, MIToken::NamedIRValue, Rule.size(), ErrorCallback);
380 }
381 
382 static Cursor maybeLexStringConstant(Cursor C, MIToken &Token,
383                                      ErrorCallbackType ErrorCallback) {
384   if (C.peek() != '"')
385     return None;
386   return lexName(C, Token, MIToken::StringConstant, /*PrefixLength=*/0,
387                  ErrorCallback);
388 }
389 
390 static Cursor lexVirtualRegister(Cursor C, MIToken &Token) {
391   auto Range = C;
392   C.advance(); // Skip '%'
393   auto NumberRange = C;
394   while (isdigit(C.peek()))
395     C.advance();
396   Token.reset(MIToken::VirtualRegister, Range.upto(C))
397       .setIntegerValue(APSInt(NumberRange.upto(C)));
398   return C;
399 }
400 
401 /// Returns true for a character allowed in a register name.
402 static bool isRegisterChar(char C) {
403   return isIdentifierChar(C) && C != '.';
404 }
405 
406 static Cursor lexNamedVirtualRegister(Cursor C, MIToken &Token) {
407   Cursor Range = C;
408   C.advance(); // Skip '%'
409   while (isRegisterChar(C.peek()))
410     C.advance();
411   Token.reset(MIToken::NamedVirtualRegister, Range.upto(C))
412       .setStringValue(Range.upto(C).drop_front(1)); // Drop the '%'
413   return C;
414 }
415 
416 static Cursor maybeLexRegister(Cursor C, MIToken &Token,
417                                ErrorCallbackType ErrorCallback) {
418   if (C.peek() != '%' && C.peek() != '$')
419     return None;
420 
421   if (C.peek() == '%') {
422     if (isdigit(C.peek(1)))
423       return lexVirtualRegister(C, Token);
424 
425     if (isRegisterChar(C.peek(1)))
426       return lexNamedVirtualRegister(C, Token);
427 
428     return None;
429   }
430 
431   assert(C.peek() == '$');
432   auto Range = C;
433   C.advance(); // Skip '$'
434   while (isRegisterChar(C.peek()))
435     C.advance();
436   Token.reset(MIToken::NamedRegister, Range.upto(C))
437       .setStringValue(Range.upto(C).drop_front(1)); // Drop the '$'
438   return C;
439 }
440 
441 static Cursor maybeLexGlobalValue(Cursor C, MIToken &Token,
442                                   ErrorCallbackType ErrorCallback) {
443   if (C.peek() != '@')
444     return None;
445   if (!isdigit(C.peek(1)))
446     return lexName(C, Token, MIToken::NamedGlobalValue, /*PrefixLength=*/1,
447                    ErrorCallback);
448   auto Range = C;
449   C.advance(1); // Skip the '@'
450   auto NumberRange = C;
451   while (isdigit(C.peek()))
452     C.advance();
453   Token.reset(MIToken::GlobalValue, Range.upto(C))
454       .setIntegerValue(APSInt(NumberRange.upto(C)));
455   return C;
456 }
457 
458 static Cursor maybeLexExternalSymbol(Cursor C, MIToken &Token,
459                                      ErrorCallbackType ErrorCallback) {
460   if (C.peek() != '&')
461     return None;
462   return lexName(C, Token, MIToken::ExternalSymbol, /*PrefixLength=*/1,
463                  ErrorCallback);
464 }
465 
466 static Cursor maybeLexMCSymbol(Cursor C, MIToken &Token,
467                                ErrorCallbackType ErrorCallback) {
468   const StringRef Rule = "<mcsymbol ";
469   if (!C.remaining().startswith(Rule))
470     return None;
471   auto Start = C;
472   C.advance(Rule.size());
473 
474   // Try a simple unquoted name.
475   if (C.peek() != '"') {
476     while (isIdentifierChar(C.peek()))
477       C.advance();
478     StringRef String = Start.upto(C).drop_front(Rule.size());
479     if (C.peek() != '>') {
480       ErrorCallback(C.location(),
481                     "expected the '<mcsymbol ...' to be closed by a '>'");
482       Token.reset(MIToken::Error, Start.remaining());
483       return Start;
484     }
485     C.advance();
486 
487     Token.reset(MIToken::MCSymbol, Start.upto(C)).setStringValue(String);
488     return C;
489   }
490 
491   // Otherwise lex out a quoted name.
492   Cursor R = lexStringConstant(C, ErrorCallback);
493   if (!R) {
494     ErrorCallback(C.location(),
495                   "unable to parse quoted string from opening quote");
496     Token.reset(MIToken::Error, Start.remaining());
497     return Start;
498   }
499   StringRef String = Start.upto(R).drop_front(Rule.size());
500   if (R.peek() != '>') {
501     ErrorCallback(R.location(),
502                   "expected the '<mcsymbol ...' to be closed by a '>'");
503     Token.reset(MIToken::Error, Start.remaining());
504     return Start;
505   }
506   R.advance();
507 
508   Token.reset(MIToken::MCSymbol, Start.upto(R))
509       .setOwnedStringValue(unescapeQuotedString(String));
510   return R;
511 }
512 
513 static bool isValidHexFloatingPointPrefix(char C) {
514   return C == 'H' || C == 'K' || C == 'L' || C == 'M';
515 }
516 
517 static Cursor lexFloatingPointLiteral(Cursor Range, Cursor C, MIToken &Token) {
518   C.advance();
519   // Skip over [0-9]*([eE][-+]?[0-9]+)?
520   while (isdigit(C.peek()))
521     C.advance();
522   if ((C.peek() == 'e' || C.peek() == 'E') &&
523       (isdigit(C.peek(1)) ||
524        ((C.peek(1) == '-' || C.peek(1) == '+') && isdigit(C.peek(2))))) {
525     C.advance(2);
526     while (isdigit(C.peek()))
527       C.advance();
528   }
529   Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
530   return C;
531 }
532 
533 static Cursor maybeLexHexadecimalLiteral(Cursor C, MIToken &Token) {
534   if (C.peek() != '0' || (C.peek(1) != 'x' && C.peek(1) != 'X'))
535     return None;
536   Cursor Range = C;
537   C.advance(2);
538   unsigned PrefLen = 2;
539   if (isValidHexFloatingPointPrefix(C.peek())) {
540     C.advance();
541     PrefLen++;
542   }
543   while (isxdigit(C.peek()))
544     C.advance();
545   StringRef StrVal = Range.upto(C);
546   if (StrVal.size() <= PrefLen)
547     return None;
548   if (PrefLen == 2)
549     Token.reset(MIToken::HexLiteral, Range.upto(C));
550   else // It must be 3, which means that there was a floating-point prefix.
551     Token.reset(MIToken::FloatingPointLiteral, Range.upto(C));
552   return C;
553 }
554 
555 static Cursor maybeLexNumericalLiteral(Cursor C, MIToken &Token) {
556   if (!isdigit(C.peek()) && (C.peek() != '-' || !isdigit(C.peek(1))))
557     return None;
558   auto Range = C;
559   C.advance();
560   while (isdigit(C.peek()))
561     C.advance();
562   if (C.peek() == '.')
563     return lexFloatingPointLiteral(Range, C, Token);
564   StringRef StrVal = Range.upto(C);
565   Token.reset(MIToken::IntegerLiteral, StrVal).setIntegerValue(APSInt(StrVal));
566   return C;
567 }
568 
569 static MIToken::TokenKind getMetadataKeywordKind(StringRef Identifier) {
570   return StringSwitch<MIToken::TokenKind>(Identifier)
571       .Case("!tbaa", MIToken::md_tbaa)
572       .Case("!alias.scope", MIToken::md_alias_scope)
573       .Case("!noalias", MIToken::md_noalias)
574       .Case("!range", MIToken::md_range)
575       .Case("!DIExpression", MIToken::md_diexpr)
576       .Default(MIToken::Error);
577 }
578 
579 static Cursor maybeLexExlaim(Cursor C, MIToken &Token,
580                              ErrorCallbackType ErrorCallback) {
581   if (C.peek() != '!')
582     return None;
583   auto Range = C;
584   C.advance(1);
585   if (isdigit(C.peek()) || !isIdentifierChar(C.peek())) {
586     Token.reset(MIToken::exclaim, Range.upto(C));
587     return C;
588   }
589   while (isIdentifierChar(C.peek()))
590     C.advance();
591   StringRef StrVal = Range.upto(C);
592   Token.reset(getMetadataKeywordKind(StrVal), StrVal);
593   if (Token.isError())
594     ErrorCallback(Token.location(),
595                   "use of unknown metadata keyword '" + StrVal + "'");
596   return C;
597 }
598 
599 static MIToken::TokenKind symbolToken(char C) {
600   switch (C) {
601   case ',':
602     return MIToken::comma;
603   case '.':
604     return MIToken::dot;
605   case '=':
606     return MIToken::equal;
607   case ':':
608     return MIToken::colon;
609   case '(':
610     return MIToken::lparen;
611   case ')':
612     return MIToken::rparen;
613   case '{':
614     return MIToken::lbrace;
615   case '}':
616     return MIToken::rbrace;
617   case '+':
618     return MIToken::plus;
619   case '-':
620     return MIToken::minus;
621   case '<':
622     return MIToken::less;
623   case '>':
624     return MIToken::greater;
625   default:
626     return MIToken::Error;
627   }
628 }
629 
630 static Cursor maybeLexSymbol(Cursor C, MIToken &Token) {
631   MIToken::TokenKind Kind;
632   unsigned Length = 1;
633   if (C.peek() == ':' && C.peek(1) == ':') {
634     Kind = MIToken::coloncolon;
635     Length = 2;
636   } else
637     Kind = symbolToken(C.peek());
638   if (Kind == MIToken::Error)
639     return None;
640   auto Range = C;
641   C.advance(Length);
642   Token.reset(Kind, Range.upto(C));
643   return C;
644 }
645 
646 static Cursor maybeLexNewline(Cursor C, MIToken &Token) {
647   if (!isNewlineChar(C.peek()))
648     return None;
649   auto Range = C;
650   C.advance();
651   Token.reset(MIToken::Newline, Range.upto(C));
652   return C;
653 }
654 
655 static Cursor maybeLexEscapedIRValue(Cursor C, MIToken &Token,
656                                      ErrorCallbackType ErrorCallback) {
657   if (C.peek() != '`')
658     return None;
659   auto Range = C;
660   C.advance();
661   auto StrRange = C;
662   while (C.peek() != '`') {
663     if (C.isEOF() || isNewlineChar(C.peek())) {
664       ErrorCallback(
665           C.location(),
666           "end of machine instruction reached before the closing '`'");
667       Token.reset(MIToken::Error, Range.remaining());
668       return C;
669     }
670     C.advance();
671   }
672   StringRef Value = StrRange.upto(C);
673   C.advance();
674   Token.reset(MIToken::QuotedIRValue, Range.upto(C)).setStringValue(Value);
675   return C;
676 }
677 
678 StringRef llvm::lexMIToken(StringRef Source, MIToken &Token,
679                            ErrorCallbackType ErrorCallback) {
680   auto C = skipComment(skipWhitespace(Cursor(Source)));
681   if (C.isEOF()) {
682     Token.reset(MIToken::Eof, C.remaining());
683     return C.remaining();
684   }
685 
686   if (Cursor R = maybeLexMachineBasicBlock(C, Token, ErrorCallback))
687     return R.remaining();
688   if (Cursor R = maybeLexIdentifier(C, Token))
689     return R.remaining();
690   if (Cursor R = maybeLexJumpTableIndex(C, Token))
691     return R.remaining();
692   if (Cursor R = maybeLexStackObject(C, Token))
693     return R.remaining();
694   if (Cursor R = maybeLexFixedStackObject(C, Token))
695     return R.remaining();
696   if (Cursor R = maybeLexConstantPoolItem(C, Token))
697     return R.remaining();
698   if (Cursor R = maybeLexSubRegisterIndex(C, Token, ErrorCallback))
699     return R.remaining();
700   if (Cursor R = maybeLexIRBlock(C, Token, ErrorCallback))
701     return R.remaining();
702   if (Cursor R = maybeLexIRValue(C, Token, ErrorCallback))
703     return R.remaining();
704   if (Cursor R = maybeLexRegister(C, Token, ErrorCallback))
705     return R.remaining();
706   if (Cursor R = maybeLexGlobalValue(C, Token, ErrorCallback))
707     return R.remaining();
708   if (Cursor R = maybeLexExternalSymbol(C, Token, ErrorCallback))
709     return R.remaining();
710   if (Cursor R = maybeLexMCSymbol(C, Token, ErrorCallback))
711     return R.remaining();
712   if (Cursor R = maybeLexHexadecimalLiteral(C, Token))
713     return R.remaining();
714   if (Cursor R = maybeLexNumericalLiteral(C, Token))
715     return R.remaining();
716   if (Cursor R = maybeLexExlaim(C, Token, ErrorCallback))
717     return R.remaining();
718   if (Cursor R = maybeLexSymbol(C, Token))
719     return R.remaining();
720   if (Cursor R = maybeLexNewline(C, Token))
721     return R.remaining();
722   if (Cursor R = maybeLexEscapedIRValue(C, Token, ErrorCallback))
723     return R.remaining();
724   if (Cursor R = maybeLexStringConstant(C, Token, ErrorCallback))
725     return R.remaining();
726 
727   Token.reset(MIToken::Error, C.remaining());
728   ErrorCallback(C.location(),
729                 Twine("unexpected character '") + Twine(C.peek()) + "'");
730   return C.remaining();
731 }
732