1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
|
//===---- ParseStmtAsm.cpp - Assembly Statement Parser --------------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file implements parsing for GCC and Microsoft inline assembly.
//
//===----------------------------------------------------------------------===//
#include "clang/Parse/Parser.h"
#include "RAIIObjectsForParser.h"
#include "clang/AST/ASTContext.h"
#include "clang/Basic/Diagnostic.h"
#include "clang/Basic/TargetInfo.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCInstPrinter.h"
#include "llvm/MC/MCInstrInfo.h"
#include "llvm/MC/MCObjectFileInfo.h"
#include "llvm/MC/MCParser/MCAsmParser.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/MC/MCStreamer.h"
#include "llvm/MC/MCSubtargetInfo.h"
#include "llvm/MC/MCTargetAsmParser.h"
#include "llvm/MC/MCTargetOptions.h"
#include "llvm/Support/SourceMgr.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/TargetSelect.h"
using namespace clang;
namespace {
class ClangAsmParserCallback : public llvm::MCAsmParserSemaCallback {
Parser &TheParser;
SourceLocation AsmLoc;
StringRef AsmString;
/// The tokens we streamed into AsmString and handed off to MC.
ArrayRef<Token> AsmToks;
/// The offset of each token in AsmToks within AsmString.
ArrayRef<unsigned> AsmTokOffsets;
public:
ClangAsmParserCallback(Parser &P, SourceLocation Loc, StringRef AsmString,
ArrayRef<Token> Toks, ArrayRef<unsigned> Offsets)
: TheParser(P), AsmLoc(Loc), AsmString(AsmString), AsmToks(Toks),
AsmTokOffsets(Offsets) {
assert(AsmToks.size() == AsmTokOffsets.size());
}
void *LookupInlineAsmIdentifier(StringRef &LineBuf,
llvm::InlineAsmIdentifierInfo &Info,
bool IsUnevaluatedContext) override {
// Collect the desired tokens.
SmallVector<Token, 16> LineToks;
const Token *FirstOrigToken = nullptr;
findTokensForString(LineBuf, LineToks, FirstOrigToken);
unsigned NumConsumedToks;
ExprResult Result = TheParser.ParseMSAsmIdentifier(
LineToks, NumConsumedToks, &Info, IsUnevaluatedContext);
// If we consumed the entire line, tell MC that.
// Also do this if we consumed nothing as a way of reporting failure.
if (NumConsumedToks == 0 || NumConsumedToks == LineToks.size()) {
// By not modifying LineBuf, we're implicitly consuming it all.
// Otherwise, consume up to the original tokens.
} else {
assert(FirstOrigToken && "not using original tokens?");
// Since we're using original tokens, apply that offset.
assert(FirstOrigToken[NumConsumedToks].getLocation() ==
LineToks[NumConsumedToks].getLocation());
unsigned FirstIndex = FirstOrigToken - AsmToks.begin();
unsigned LastIndex = FirstIndex + NumConsumedToks - 1;
// The total length we've consumed is the relative offset
// of the last token we consumed plus its length.
unsigned TotalOffset =
(AsmTokOffsets[LastIndex] + AsmToks[LastIndex].getLength() -
AsmTokOffsets[FirstIndex]);
LineBuf = LineBuf.substr(0, TotalOffset);
}
// Initialize the "decl" with the lookup result.
Info.OpDecl = static_cast<void *>(Result.get());
return Info.OpDecl;
}
StringRef LookupInlineAsmLabel(StringRef Identifier, llvm::SourceMgr &LSM,
llvm::SMLoc Location,
bool Create) override {
SourceLocation Loc = translateLocation(LSM, Location);
LabelDecl *Label =
TheParser.getActions().GetOrCreateMSAsmLabel(Identifier, Loc, Create);
return Label->getMSAsmLabel();
}
bool LookupInlineAsmField(StringRef Base, StringRef Member,
unsigned &Offset) override {
return TheParser.getActions().LookupInlineAsmField(Base, Member, Offset,
AsmLoc);
}
static void DiagHandlerCallback(const llvm::SMDiagnostic &D, void *Context) {
((ClangAsmParserCallback *)Context)->handleDiagnostic(D);
}
private:
/// Collect the appropriate tokens for the given string.
void findTokensForString(StringRef Str, SmallVectorImpl<Token> &TempToks,
const Token *&FirstOrigToken) const {
// For now, assert that the string we're working with is a substring
// of what we gave to MC. This lets us use the original tokens.
assert(!std::less<const char *>()(Str.begin(), AsmString.begin()) &&
!std::less<const char *>()(AsmString.end(), Str.end()));
// Try to find a token whose offset matches the first token.
unsigned FirstCharOffset = Str.begin() - AsmString.begin();
const unsigned *FirstTokOffset = std::lower_bound(
AsmTokOffsets.begin(), AsmTokOffsets.end(), FirstCharOffset);
// For now, assert that the start of the string exactly
// corresponds to the start of a token.
assert(*FirstTokOffset == FirstCharOffset);
// Use all the original tokens for this line. (We assume the
// end of the line corresponds cleanly to a token break.)
unsigned FirstTokIndex = FirstTokOffset - AsmTokOffsets.begin();
FirstOrigToken = &AsmToks[FirstTokIndex];
unsigned LastCharOffset = Str.end() - AsmString.begin();
for (unsigned i = FirstTokIndex, e = AsmTokOffsets.size(); i != e; ++i) {
if (AsmTokOffsets[i] >= LastCharOffset)
break;
TempToks.push_back(AsmToks[i]);
}
}
SourceLocation translateLocation(const llvm::SourceMgr &LSM, llvm::SMLoc SMLoc) {
// Compute an offset into the inline asm buffer.
// FIXME: This isn't right if .macro is involved (but hopefully, no
// real-world code does that).
const llvm::MemoryBuffer *LBuf =
LSM.getMemoryBuffer(LSM.FindBufferContainingLoc(SMLoc));
unsigned Offset = SMLoc.getPointer() - LBuf->getBufferStart();
// Figure out which token that offset points into.
const unsigned *TokOffsetPtr =
std::lower_bound(AsmTokOffsets.begin(), AsmTokOffsets.end(), Offset);
unsigned TokIndex = TokOffsetPtr - AsmTokOffsets.begin();
unsigned TokOffset = *TokOffsetPtr;
// If we come up with an answer which seems sane, use it; otherwise,
// just point at the __asm keyword.
// FIXME: Assert the answer is sane once we handle .macro correctly.
SourceLocation Loc = AsmLoc;
if (TokIndex < AsmToks.size()) {
const Token &Tok = AsmToks[TokIndex];
Loc = Tok.getLocation();
Loc = Loc.getLocWithOffset(Offset - TokOffset);
}
return Loc;
}
void handleDiagnostic(const llvm::SMDiagnostic &D) {
const llvm::SourceMgr &LSM = *D.getSourceMgr();
SourceLocation Loc = translateLocation(LSM, D.getLoc());
TheParser.Diag(Loc, diag::err_inline_ms_asm_parsing) << D.getMessage();
}
};
}
/// Parse an identifier in an MS-style inline assembly block.
///
/// \param CastInfo - a void* so that we don't have to teach Parser.h
/// about the actual type.
ExprResult Parser::ParseMSAsmIdentifier(llvm::SmallVectorImpl<Token> &LineToks,
unsigned &NumLineToksConsumed,
void *CastInfo,
bool IsUnevaluatedContext) {
llvm::InlineAsmIdentifierInfo &Info =
*(llvm::InlineAsmIdentifierInfo *)CastInfo;
// Push a fake token on the end so that we don't overrun the token
// stream. We use ';' because it expression-parsing should never
// overrun it.
const tok::TokenKind EndOfStream = tok::semi;
Token EndOfStreamTok;
EndOfStreamTok.startToken();
EndOfStreamTok.setKind(EndOfStream);
LineToks.push_back(EndOfStreamTok);
// Also copy the current token over.
LineToks.push_back(Tok);
PP.EnterTokenStream(LineToks.begin(), LineToks.size(),
/*disable macros*/ true,
/*owns tokens*/ false);
// Clear the current token and advance to the first token in LineToks.
ConsumeAnyToken();
// Parse an optional scope-specifier if we're in C++.
CXXScopeSpec SS;
if (getLangOpts().CPlusPlus) {
ParseOptionalCXXScopeSpecifier(SS, nullptr, /*EnteringContext=*/false);
}
// Require an identifier here.
SourceLocation TemplateKWLoc;
UnqualifiedId Id;
bool Invalid = true;
ExprResult Result;
if (Tok.is(tok::kw_this)) {
Result = ParseCXXThis();
Invalid = false;
} else {
Invalid = ParseUnqualifiedId(SS,
/*EnteringContext=*/false,
/*AllowDestructorName=*/false,
/*AllowConstructorName=*/false,
/*ObjectType=*/nullptr, TemplateKWLoc, Id);
// Perform the lookup.
Result = Actions.LookupInlineAsmIdentifier(SS, TemplateKWLoc, Id, Info,
IsUnevaluatedContext);
}
// While the next two tokens are 'period' 'identifier', repeatedly parse it as
// a field access. We have to avoid consuming assembler directives that look
// like '.' 'else'.
while (Result.isUsable() && Tok.is(tok::period)) {
Token IdTok = PP.LookAhead(0);
if (IdTok.isNot(tok::identifier))
break;
ConsumeToken(); // Consume the period.
IdentifierInfo *Id = Tok.getIdentifierInfo();
ConsumeToken(); // Consume the identifier.
Result = Actions.LookupInlineAsmVarDeclField(Result.get(), Id->getName(),
Info, Tok.getLocation());
}
// Figure out how many tokens we are into LineToks.
unsigned LineIndex = 0;
if (Tok.is(EndOfStream)) {
LineIndex = LineToks.size() - 2;
} else {
while (LineToks[LineIndex].getLocation() != Tok.getLocation()) {
LineIndex++;
assert(LineIndex < LineToks.size() - 2); // we added two extra tokens
}
}
// If we've run into the poison token we inserted before, or there
// was a parsing error, then claim the entire line.
if (Invalid || Tok.is(EndOfStream)) {
NumLineToksConsumed = LineToks.size() - 2;
} else {
// Otherwise, claim up to the start of the next token.
NumLineToksConsumed = LineIndex;
}
// Finally, restore the old parsing state by consuming all the tokens we
// staged before, implicitly killing off the token-lexer we pushed.
for (unsigned i = 0, e = LineToks.size() - LineIndex - 2; i != e; ++i) {
ConsumeAnyToken();
}
assert(Tok.is(EndOfStream));
ConsumeToken();
// Leave LineToks in its original state.
LineToks.pop_back();
LineToks.pop_back();
return Result;
}
/// Turn a sequence of our tokens back into a string that we can hand
/// to the MC asm parser.
static bool buildMSAsmString(Preprocessor &PP, SourceLocation AsmLoc,
ArrayRef<Token> AsmToks,
SmallVectorImpl<unsigned> &TokOffsets,
SmallString<512> &Asm) {
assert(!AsmToks.empty() && "Didn't expect an empty AsmToks!");
// Is this the start of a new assembly statement?
bool isNewStatement = true;
for (unsigned i = 0, e = AsmToks.size(); i < e; ++i) {
const Token &Tok = AsmToks[i];
// Start each new statement with a newline and a tab.
if (!isNewStatement && (Tok.is(tok::kw_asm) || Tok.isAtStartOfLine())) {
Asm += "\n\t";
isNewStatement = true;
}
// Preserve the existence of leading whitespace except at the
// start of a statement.
if (!isNewStatement && Tok.hasLeadingSpace())
Asm += ' ';
// Remember the offset of this token.
TokOffsets.push_back(Asm.size());
// Don't actually write '__asm' into the assembly stream.
if (Tok.is(tok::kw_asm)) {
// Complain about __asm at the end of the stream.
if (i + 1 == e) {
PP.Diag(AsmLoc, diag::err_asm_empty);
return true;
}
continue;
}
// Append the spelling of the token.
SmallString<32> SpellingBuffer;
bool SpellingInvalid = false;
Asm += PP.getSpelling(Tok, SpellingBuffer, &SpellingInvalid);
assert(!SpellingInvalid && "spelling was invalid after correct parse?");
// We are no longer at the start of a statement.
isNewStatement = false;
}
// Ensure that the buffer is null-terminated.
Asm.push_back('\0');
Asm.pop_back();
assert(TokOffsets.size() == AsmToks.size());
return false;
}
/// ParseMicrosoftAsmStatement. When -fms-extensions/-fasm-blocks is enabled,
/// this routine is called to collect the tokens for an MS asm statement.
///
/// [MS] ms-asm-statement:
/// ms-asm-block
/// ms-asm-block ms-asm-statement
///
/// [MS] ms-asm-block:
/// '__asm' ms-asm-line '\n'
/// '__asm' '{' ms-asm-instruction-block[opt] '}' ';'[opt]
///
/// [MS] ms-asm-instruction-block
/// ms-asm-line
/// ms-asm-line '\n' ms-asm-instruction-block
///
StmtResult Parser::ParseMicrosoftAsmStatement(SourceLocation AsmLoc) {
SourceManager &SrcMgr = PP.getSourceManager();
SourceLocation EndLoc = AsmLoc;
SmallVector<Token, 4> AsmToks;
bool SingleLineMode = true;
unsigned BraceNesting = 0;
unsigned short savedBraceCount = BraceCount;
bool InAsmComment = false;
FileID FID;
unsigned LineNo = 0;
unsigned NumTokensRead = 0;
SmallVector<SourceLocation, 4> LBraceLocs;
bool SkippedStartOfLine = false;
if (Tok.is(tok::l_brace)) {
// Braced inline asm: consume the opening brace.
SingleLineMode = false;
BraceNesting = 1;
EndLoc = ConsumeBrace();
LBraceLocs.push_back(EndLoc);
++NumTokensRead;
} else {
// Single-line inline asm; compute which line it is on.
std::pair<FileID, unsigned> ExpAsmLoc =
SrcMgr.getDecomposedExpansionLoc(EndLoc);
FID = ExpAsmLoc.first;
LineNo = SrcMgr.getLineNumber(FID, ExpAsmLoc.second);
LBraceLocs.push_back(SourceLocation());
}
SourceLocation TokLoc = Tok.getLocation();
do {
// If we hit EOF, we're done, period.
if (isEofOrEom())
break;
if (!InAsmComment && Tok.is(tok::l_brace)) {
// Consume the opening brace.
SkippedStartOfLine = Tok.isAtStartOfLine();
EndLoc = ConsumeBrace();
BraceNesting++;
LBraceLocs.push_back(EndLoc);
TokLoc = Tok.getLocation();
++NumTokensRead;
continue;
} else if (!InAsmComment && Tok.is(tok::semi)) {
// A semicolon in an asm is the start of a comment.
InAsmComment = true;
if (!SingleLineMode) {
// Compute which line the comment is on.
std::pair<FileID, unsigned> ExpSemiLoc =
SrcMgr.getDecomposedExpansionLoc(TokLoc);
FID = ExpSemiLoc.first;
LineNo = SrcMgr.getLineNumber(FID, ExpSemiLoc.second);
}
} else if (SingleLineMode || InAsmComment) {
// If end-of-line is significant, check whether this token is on a
// new line.
std::pair<FileID, unsigned> ExpLoc =
SrcMgr.getDecomposedExpansionLoc(TokLoc);
if (ExpLoc.first != FID ||
SrcMgr.getLineNumber(ExpLoc.first, ExpLoc.second) != LineNo) {
// If this is a single-line __asm, we're done, except if the next
// line begins with an __asm too, in which case we finish a comment
// if needed and then keep processing the next line as a single
// line __asm.
bool isAsm = Tok.is(tok::kw_asm);
if (SingleLineMode && !isAsm)
break;
// We're no longer in a comment.
InAsmComment = false;
if (isAsm) {
LineNo = SrcMgr.getLineNumber(ExpLoc.first, ExpLoc.second);
SkippedStartOfLine = Tok.isAtStartOfLine();
}
} else if (!InAsmComment && Tok.is(tok::r_brace)) {
// In MSVC mode, braces only participate in brace matching and
// separating the asm statements. This is an intentional
// departure from the Apple gcc behavior.
if (!BraceNesting)
break;
}
}
if (!InAsmComment && BraceNesting && Tok.is(tok::r_brace) &&
BraceCount == (savedBraceCount + BraceNesting)) {
// Consume the closing brace.
SkippedStartOfLine = Tok.isAtStartOfLine();
EndLoc = ConsumeBrace();
BraceNesting--;
// Finish if all of the opened braces in the inline asm section were
// consumed.
if (BraceNesting == 0 && !SingleLineMode)
break;
else {
LBraceLocs.pop_back();
TokLoc = Tok.getLocation();
++NumTokensRead;
continue;
}
}
// Consume the next token; make sure we don't modify the brace count etc.
// if we are in a comment.
EndLoc = TokLoc;
if (InAsmComment)
PP.Lex(Tok);
else {
// Set the token as the start of line if we skipped the original start
// of line token in case it was a nested brace.
if (SkippedStartOfLine)
Tok.setFlag(Token::StartOfLine);
AsmToks.push_back(Tok);
ConsumeAnyToken();
}
TokLoc = Tok.getLocation();
++NumTokensRead;
SkippedStartOfLine = false;
} while (1);
if (BraceNesting && BraceCount != savedBraceCount) {
// __asm without closing brace (this can happen at EOF).
for (unsigned i = 0; i < BraceNesting; ++i) {
Diag(Tok, diag::err_expected) << tok::r_brace;
Diag(LBraceLocs.back(), diag::note_matching) << tok::l_brace;
LBraceLocs.pop_back();
}
return StmtError();
} else if (NumTokensRead == 0) {
// Empty __asm.
Diag(Tok, diag::err_expected) << tok::l_brace;
return StmtError();
}
// Okay, prepare to use MC to parse the assembly.
SmallVector<StringRef, 4> ConstraintRefs;
SmallVector<Expr *, 4> Exprs;
SmallVector<StringRef, 4> ClobberRefs;
// We need an actual supported target.
const llvm::Triple &TheTriple = Actions.Context.getTargetInfo().getTriple();
llvm::Triple::ArchType ArchTy = TheTriple.getArch();
const std::string &TT = TheTriple.getTriple();
const llvm::Target *TheTarget = nullptr;
bool UnsupportedArch =
(ArchTy != llvm::Triple::x86 && ArchTy != llvm::Triple::x86_64);
if (UnsupportedArch) {
Diag(AsmLoc, diag::err_msasm_unsupported_arch) << TheTriple.getArchName();
} else {
std::string Error;
TheTarget = llvm::TargetRegistry::lookupTarget(TT, Error);
if (!TheTarget)
Diag(AsmLoc, diag::err_msasm_unable_to_create_target) << Error;
}
assert(!LBraceLocs.empty() && "Should have at least one location here");
// If we don't support assembly, or the assembly is empty, we don't
// need to instantiate the AsmParser, etc.
if (!TheTarget || AsmToks.empty()) {
return Actions.ActOnMSAsmStmt(AsmLoc, LBraceLocs[0], AsmToks, StringRef(),
/*NumOutputs*/ 0, /*NumInputs*/ 0,
ConstraintRefs, ClobberRefs, Exprs, EndLoc);
}
// Expand the tokens into a string buffer.
SmallString<512> AsmString;
SmallVector<unsigned, 8> TokOffsets;
if (buildMSAsmString(PP, AsmLoc, AsmToks, TokOffsets, AsmString))
return StmtError();
std::unique_ptr<llvm::MCRegisterInfo> MRI(TheTarget->createMCRegInfo(TT));
std::unique_ptr<llvm::MCAsmInfo> MAI(TheTarget->createMCAsmInfo(*MRI, TT));
// Get the instruction descriptor.
std::unique_ptr<llvm::MCInstrInfo> MII(TheTarget->createMCInstrInfo());
std::unique_ptr<llvm::MCObjectFileInfo> MOFI(new llvm::MCObjectFileInfo());
std::unique_ptr<llvm::MCSubtargetInfo> STI(
TheTarget->createMCSubtargetInfo(TT, "", ""));
llvm::SourceMgr TempSrcMgr;
llvm::MCContext Ctx(MAI.get(), MRI.get(), MOFI.get(), &TempSrcMgr);
MOFI->InitMCObjectFileInfo(TheTriple, llvm::Reloc::Default,
llvm::CodeModel::Default, Ctx);
std::unique_ptr<llvm::MemoryBuffer> Buffer =
llvm::MemoryBuffer::getMemBuffer(AsmString, "<MS inline asm>");
// Tell SrcMgr about this buffer, which is what the parser will pick up.
TempSrcMgr.AddNewSourceBuffer(std::move(Buffer), llvm::SMLoc());
std::unique_ptr<llvm::MCStreamer> Str(createNullStreamer(Ctx));
std::unique_ptr<llvm::MCAsmParser> Parser(
createMCAsmParser(TempSrcMgr, Ctx, *Str.get(), *MAI));
// FIXME: init MCOptions from sanitizer flags here.
llvm::MCTargetOptions MCOptions;
std::unique_ptr<llvm::MCTargetAsmParser> TargetParser(
TheTarget->createMCAsmParser(*STI, *Parser, *MII, MCOptions));
std::unique_ptr<llvm::MCInstPrinter> IP(
TheTarget->createMCInstPrinter(llvm::Triple(TT), 1, *MAI, *MII, *MRI));
// Change to the Intel dialect.
Parser->setAssemblerDialect(1);
Parser->setTargetParser(*TargetParser.get());
Parser->setParsingInlineAsm(true);
TargetParser->setParsingInlineAsm(true);
ClangAsmParserCallback Callback(*this, AsmLoc, AsmString, AsmToks,
TokOffsets);
TargetParser->setSemaCallback(&Callback);
TempSrcMgr.setDiagHandler(ClangAsmParserCallback::DiagHandlerCallback,
&Callback);
unsigned NumOutputs;
unsigned NumInputs;
std::string AsmStringIR;
SmallVector<std::pair<void *, bool>, 4> OpExprs;
SmallVector<std::string, 4> Constraints;
SmallVector<std::string, 4> Clobbers;
if (Parser->parseMSInlineAsm(AsmLoc.getPtrEncoding(), AsmStringIR, NumOutputs,
NumInputs, OpExprs, Constraints, Clobbers,
MII.get(), IP.get(), Callback))
return StmtError();
// Filter out "fpsw". Clang doesn't accept it, and it always lists flags and
// fpsr as clobbers.
auto End = std::remove(Clobbers.begin(), Clobbers.end(), "fpsw");
Clobbers.erase(End, Clobbers.end());
// Build the vector of clobber StringRefs.
ClobberRefs.insert(ClobberRefs.end(), Clobbers.begin(), Clobbers.end());
// Recast the void pointers and build the vector of constraint StringRefs.
unsigned NumExprs = NumOutputs + NumInputs;
ConstraintRefs.resize(NumExprs);
Exprs.resize(NumExprs);
for (unsigned i = 0, e = NumExprs; i != e; ++i) {
Expr *OpExpr = static_cast<Expr *>(OpExprs[i].first);
if (!OpExpr)
return StmtError();
// Need address of variable.
if (OpExprs[i].second)
OpExpr =
Actions.BuildUnaryOp(getCurScope(), AsmLoc, UO_AddrOf, OpExpr).get();
ConstraintRefs[i] = StringRef(Constraints[i]);
Exprs[i] = OpExpr;
}
// FIXME: We should be passing source locations for better diagnostics.
return Actions.ActOnMSAsmStmt(AsmLoc, LBraceLocs[0], AsmToks, AsmStringIR,
NumOutputs, NumInputs, ConstraintRefs,
ClobberRefs, Exprs, EndLoc);
}
/// ParseAsmStatement - Parse a GNU extended asm statement.
/// asm-statement:
/// gnu-asm-statement
/// ms-asm-statement
///
/// [GNU] gnu-asm-statement:
/// 'asm' type-qualifier[opt] '(' asm-argument ')' ';'
///
/// [GNU] asm-argument:
/// asm-string-literal
/// asm-string-literal ':' asm-operands[opt]
/// asm-string-literal ':' asm-operands[opt] ':' asm-operands[opt]
/// asm-string-literal ':' asm-operands[opt] ':' asm-operands[opt]
/// ':' asm-clobbers
///
/// [GNU] asm-clobbers:
/// asm-string-literal
/// asm-clobbers ',' asm-string-literal
///
StmtResult Parser::ParseAsmStatement(bool &msAsm) {
assert(Tok.is(tok::kw_asm) && "Not an asm stmt");
SourceLocation AsmLoc = ConsumeToken();
if (getLangOpts().AsmBlocks && Tok.isNot(tok::l_paren) &&
!isTypeQualifier()) {
msAsm = true;
return ParseMicrosoftAsmStatement(AsmLoc);
}
DeclSpec DS(AttrFactory);
SourceLocation Loc = Tok.getLocation();
ParseTypeQualifierListOpt(DS, AR_VendorAttributesParsed);
// GNU asms accept, but warn, about type-qualifiers other than volatile.
if (DS.getTypeQualifiers() & DeclSpec::TQ_const)
Diag(Loc, diag::w_asm_qualifier_ignored) << "const";
if (DS.getTypeQualifiers() & DeclSpec::TQ_restrict)
Diag(Loc, diag::w_asm_qualifier_ignored) << "restrict";
// FIXME: Once GCC supports _Atomic, check whether it permits it here.
if (DS.getTypeQualifiers() & DeclSpec::TQ_atomic)
Diag(Loc, diag::w_asm_qualifier_ignored) << "_Atomic";
// Remember if this was a volatile asm.
bool isVolatile = DS.getTypeQualifiers() & DeclSpec::TQ_volatile;
if (Tok.isNot(tok::l_paren)) {
Diag(Tok, diag::err_expected_lparen_after) << "asm";
SkipUntil(tok::r_paren, StopAtSemi);
return StmtError();
}
BalancedDelimiterTracker T(*this, tok::l_paren);
T.consumeOpen();
ExprResult AsmString(ParseAsmStringLiteral());
// Check if GNU-style InlineAsm is disabled.
// Error on anything other than empty string.
if (!(getLangOpts().GNUAsm || AsmString.isInvalid())) {
const auto *SL = cast<StringLiteral>(AsmString.get());
if (!SL->getString().trim().empty())
Diag(Loc, diag::err_gnu_inline_asm_disabled);
}
if (AsmString.isInvalid()) {
// Consume up to and including the closing paren.
T.skipToEnd();
return StmtError();
}
SmallVector<IdentifierInfo *, 4> Names;
ExprVector Constraints;
ExprVector Exprs;
ExprVector Clobbers;
if (Tok.is(tok::r_paren)) {
// We have a simple asm expression like 'asm("foo")'.
T.consumeClose();
return Actions.ActOnGCCAsmStmt(AsmLoc, /*isSimple*/ true, isVolatile,
/*NumOutputs*/ 0, /*NumInputs*/ 0, nullptr,
Constraints, Exprs, AsmString.get(),
Clobbers, T.getCloseLocation());
}
// Parse Outputs, if present.
bool AteExtraColon = false;
if (Tok.is(tok::colon) || Tok.is(tok::coloncolon)) {
// In C++ mode, parse "::" like ": :".
AteExtraColon = Tok.is(tok::coloncolon);
ConsumeToken();
if (!AteExtraColon && ParseAsmOperandsOpt(Names, Constraints, Exprs))
return StmtError();
}
unsigned NumOutputs = Names.size();
// Parse Inputs, if present.
if (AteExtraColon || Tok.is(tok::colon) || Tok.is(tok::coloncolon)) {
// In C++ mode, parse "::" like ": :".
if (AteExtraColon)
AteExtraColon = false;
else {
AteExtraColon = Tok.is(tok::coloncolon);
ConsumeToken();
}
if (!AteExtraColon && ParseAsmOperandsOpt(Names, Constraints, Exprs))
return StmtError();
}
assert(Names.size() == Constraints.size() &&
Constraints.size() == Exprs.size() && "Input operand size mismatch!");
unsigned NumInputs = Names.size() - NumOutputs;
// Parse the clobbers, if present.
if (AteExtraColon || Tok.is(tok::colon)) {
if (!AteExtraColon)
ConsumeToken();
// Parse the asm-string list for clobbers if present.
if (Tok.isNot(tok::r_paren)) {
while (1) {
ExprResult Clobber(ParseAsmStringLiteral());
if (Clobber.isInvalid())
break;
Clobbers.push_back(Clobber.get());
if (!TryConsumeToken(tok::comma))
break;
}
}
}
T.consumeClose();
return Actions.ActOnGCCAsmStmt(
AsmLoc, false, isVolatile, NumOutputs, NumInputs, Names.data(),
Constraints, Exprs, AsmString.get(), Clobbers, T.getCloseLocation());
}
/// ParseAsmOperands - Parse the asm-operands production as used by
/// asm-statement, assuming the leading ':' token was eaten.
///
/// [GNU] asm-operands:
/// asm-operand
/// asm-operands ',' asm-operand
///
/// [GNU] asm-operand:
/// asm-string-literal '(' expression ')'
/// '[' identifier ']' asm-string-literal '(' expression ')'
///
//
// FIXME: Avoid unnecessary std::string trashing.
bool Parser::ParseAsmOperandsOpt(SmallVectorImpl<IdentifierInfo *> &Names,
SmallVectorImpl<Expr *> &Constraints,
SmallVectorImpl<Expr *> &Exprs) {
// 'asm-operands' isn't present?
if (!isTokenStringLiteral() && Tok.isNot(tok::l_square))
return false;
while (1) {
// Read the [id] if present.
if (Tok.is(tok::l_square)) {
BalancedDelimiterTracker T(*this, tok::l_square);
T.consumeOpen();
if (Tok.isNot(tok::identifier)) {
Diag(Tok, diag::err_expected) << tok::identifier;
SkipUntil(tok::r_paren, StopAtSemi);
return true;
}
IdentifierInfo *II = Tok.getIdentifierInfo();
ConsumeToken();
Names.push_back(II);
T.consumeClose();
} else
Names.push_back(nullptr);
ExprResult Constraint(ParseAsmStringLiteral());
if (Constraint.isInvalid()) {
SkipUntil(tok::r_paren, StopAtSemi);
return true;
}
Constraints.push_back(Constraint.get());
if (Tok.isNot(tok::l_paren)) {
Diag(Tok, diag::err_expected_lparen_after) << "asm operand";
SkipUntil(tok::r_paren, StopAtSemi);
return true;
}
// Read the parenthesized expression.
BalancedDelimiterTracker T(*this, tok::l_paren);
T.consumeOpen();
ExprResult Res = Actions.CorrectDelayedTyposInExpr(ParseExpression());
T.consumeClose();
if (Res.isInvalid()) {
SkipUntil(tok::r_paren, StopAtSemi);
return true;
}
Exprs.push_back(Res.get());
// Eat the comma and continue parsing if it exists.
if (!TryConsumeToken(tok::comma))
return false;
}
}
|