1 //===-- LLParser.h - Parser Class -------------------------------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file defines the parser class for .ll files.
12 //===----------------------------------------------------------------------===//
14 #ifndef LLVM_LIB_ASMPARSER_LLPARSER_H
15 #define LLVM_LIB_ASMPARSER_LLPARSER_H
18 #include "llvm/ADT/DenseMap.h"
19 #include "llvm/ADT/StringMap.h"
20 #include "llvm/IR/Attributes.h"
21 #include "llvm/IR/Instructions.h"
22 #include "llvm/IR/Module.h"
23 #include "llvm/IR/Operator.h"
24 #include "llvm/IR/Type.h"
25 #include "llvm/IR/ValueHandle.h"
43 /// ValID - Represents a reference of a definition of some sort with no type.
44 /// There are several cases where we have to parse the value but where the
45 /// type can depend on later context. This may either be a numeric reference
46 /// or a symbolic (%var) reference. This is just a discriminated union.
49 t_LocalID, t_GlobalID, // ID in UIntVal.
50 t_LocalName, t_GlobalName, // Name in StrVal.
51 t_APSInt, t_APFloat, // Value in APSIntVal/APFloatVal.
52 t_Null, t_Undef, t_Zero, // No value.
53 t_EmptyArray, // No value: []
54 t_Constant, // Value in ConstantVal.
55 t_InlineAsm, // Value in FTy/StrVal/StrVal2/UIntVal.
56 t_ConstantStruct, // Value in ConstantStructElts.
57 t_PackedConstantStruct // Value in ConstantStructElts.
63 std::string StrVal, StrVal2;
65 APFloat APFloatVal{0.0};
66 Constant *ConstantVal;
67 std::unique_ptr<Constant *[]> ConstantStructElts;
70 ValID(const ValID &RHS)
71 : Kind(RHS.Kind), Loc(RHS.Loc), UIntVal(RHS.UIntVal), FTy(RHS.FTy),
72 StrVal(RHS.StrVal), StrVal2(RHS.StrVal2), APSIntVal(RHS.APSIntVal),
73 APFloatVal(RHS.APFloatVal), ConstantVal(RHS.ConstantVal) {
74 assert(!RHS.ConstantStructElts);
77 bool operator<(const ValID &RHS) const {
78 if (Kind == t_LocalID || Kind == t_GlobalID)
79 return UIntVal < RHS.UIntVal;
80 assert((Kind == t_LocalName || Kind == t_GlobalName ||
81 Kind == t_ConstantStruct || Kind == t_PackedConstantStruct) &&
82 "Ordering not defined for this ValID kind yet");
83 return StrVal < RHS.StrVal;
89 typedef LLLexer::LocTy LocTy;
96 // Instruction metadata resolution. Each instruction can have a list of
97 // MDRef info associated with them.
99 // The simpler approach of just creating temporary MDNodes and then calling
100 // RAUW on them when the definition is processed doesn't work because some
101 // instruction metadata kinds, such as dbg, get stored in the IR in an
102 // "optimized" format which doesn't participate in the normal value use
103 // lists. This means that RAUW doesn't work, even on temporary MDNodes
104 // which otherwise support RAUW. Instead, we defer resolving MDNode
105 // references until the definitions have been processed.
108 unsigned MDKind, MDSlot;
111 /// Indicates which operator an operand allows (for the few operands that
112 /// may only reference a certain operator).
113 enum OperatorConstraint {
114 OC_None = 0, // No constraint
115 OC_CatchPad, // Must be CatchPadInst
116 OC_CleanupPad // Must be CleanupPadInst
119 SmallVector<Instruction*, 64> InstsWithTBAATag;
121 // Type resolution handling data structures. The location is set when we
122 // have processed a use of the type but not a definition yet.
123 StringMap<std::pair<Type*, LocTy> > NamedTypes;
124 std::map<unsigned, std::pair<Type*, LocTy> > NumberedTypes;
126 std::map<unsigned, TrackingMDNodeRef> NumberedMetadata;
127 std::map<unsigned, std::pair<TempMDTuple, LocTy>> ForwardRefMDNodes;
129 // Global Value reference information.
130 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
131 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
132 std::vector<GlobalValue*> NumberedVals;
134 // Comdat forward reference information.
135 std::map<std::string, LocTy> ForwardRefComdats;
137 // References to blockaddress. The key is the function ValID, the value is
138 // a list of references to blocks in that function.
139 std::map<ValID, std::map<ValID, GlobalValue *>> ForwardRefBlockAddresses;
140 class PerFunctionState;
141 /// Reference to per-function state to allow basic blocks to be
142 /// forward-referenced by blockaddress instructions within the same
144 PerFunctionState *BlockAddressPFS;
146 // Attribute builder reference information.
147 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
148 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
151 LLParser(StringRef F, SourceMgr &SM, SMDiagnostic &Err, Module *M,
152 SlotMapping *Slots = nullptr)
153 : Context(M->getContext()), Lex(F, SM, Err, M->getContext()), M(M),
154 Slots(Slots), BlockAddressPFS(nullptr) {}
157 bool parseStandaloneConstantValue(Constant *&C, const SlotMapping *Slots);
159 LLVMContext &getContext() { return Context; }
163 bool Error(LocTy L, const Twine &Msg) const {
164 return Lex.Error(L, Msg);
166 bool TokError(const Twine &Msg) const {
167 return Error(Lex.getLoc(), Msg);
170 /// Restore the internal name and slot mappings using the mappings that
171 /// were created at an earlier parsing stage.
172 void restoreParsingState(const SlotMapping *Slots);
174 /// GetGlobalVal - Get a value with the specified name or ID, creating a
175 /// forward reference record if needed. This can return null if the value
176 /// exists but does not have the right type.
177 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
178 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
180 /// Get a Comdat with the specified name, creating a forward reference
181 /// record if needed.
182 Comdat *getComdat(const std::string &N, LocTy Loc);
185 bool ParseToken(lltok::Kind T, const char *ErrMsg);
186 bool EatIfPresent(lltok::Kind T) {
187 if (Lex.getKind() != T) return false;
192 FastMathFlags EatFastMathFlagsIfPresent() {
195 switch (Lex.getKind()) {
196 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
197 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
198 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
199 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
200 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
206 bool ParseOptionalToken(lltok::Kind T, bool &Present,
207 LocTy *Loc = nullptr) {
208 if (Lex.getKind() != T) {
218 bool ParseStringConstant(std::string &Result);
219 bool ParseUInt32(unsigned &Val);
220 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
222 return ParseUInt32(Val);
224 bool ParseUInt64(uint64_t &Val);
225 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
227 return ParseUInt64(Val);
230 bool ParseStringAttribute(AttrBuilder &B);
232 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
233 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
234 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
235 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
237 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
238 bool ParseOptionalParamAttrs(AttrBuilder &B);
239 bool ParseOptionalReturnAttrs(AttrBuilder &B);
240 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
241 bool ParseOptionalLinkage(unsigned &Linkage) {
242 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
244 bool ParseOptionalVisibility(unsigned &Visibility);
245 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
246 bool ParseOptionalCallingConv(unsigned &CC);
247 bool ParseOptionalAlignment(unsigned &Alignment);
248 bool ParseOptionalDerefAttrBytes(lltok::Kind AttrKind, uint64_t &Bytes);
249 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
250 AtomicOrdering &Ordering);
251 bool ParseOrdering(AtomicOrdering &Ordering);
252 bool ParseOptionalStackAlignment(unsigned &Alignment);
253 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
254 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
255 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
256 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
258 if (ParseIndexList(Indices, AteExtraComma)) return true;
260 return TokError("expected index");
264 // Top-Level Entities
265 bool ParseTopLevelEntities();
266 bool ValidateEndOfModule();
267 bool ParseTargetDefinition();
268 bool ParseModuleAsm();
269 bool ParseDepLibs(); // FIXME: Remove in 4.0.
270 bool ParseUnnamedType();
271 bool ParseNamedType();
275 bool ParseGlobalType(bool &IsConstant);
276 bool ParseUnnamedGlobal();
277 bool ParseNamedGlobal();
278 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
279 bool HasLinkage, unsigned Visibility,
280 unsigned DLLStorageClass,
281 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
282 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
283 unsigned Visibility, unsigned DLLStorageClass,
284 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
286 bool ParseStandaloneMetadata();
287 bool ParseNamedMetadata();
288 bool ParseMDString(MDString *&Result);
289 bool ParseMDNodeID(MDNode *&Result);
290 bool ParseUnnamedAttrGrp();
291 bool ParseFnAttributeValuePairs(AttrBuilder &B,
292 std::vector<unsigned> &FwdRefAttrGrps,
293 bool inAttrGrp, LocTy &BuiltinLoc);
296 bool ParseType(Type *&Result, const Twine &Msg, bool AllowVoid = false);
297 bool ParseType(Type *&Result, bool AllowVoid = false) {
298 return ParseType(Result, "expected type", AllowVoid);
300 bool ParseType(Type *&Result, const Twine &Msg, LocTy &Loc,
301 bool AllowVoid = false) {
303 return ParseType(Result, Msg, AllowVoid);
305 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
307 return ParseType(Result, AllowVoid);
309 bool ParseAnonStructType(Type *&Result, bool Packed);
310 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
311 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
312 std::pair<Type*, LocTy> &Entry,
315 bool ParseArrayVectorType(Type *&Result, bool isVector);
316 bool ParseFunctionType(Type *&Result);
318 // Function Semantic Analysis.
319 class PerFunctionState {
322 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
323 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
324 std::vector<Value*> NumberedVals;
326 /// FunctionNumber - If this is an unnamed function, this is the slot
327 /// number of it, otherwise it is -1.
330 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
333 Function &getFunction() const { return F; }
335 bool FinishFunction();
337 /// GetVal - Get a value with the specified name or ID, creating a
338 /// forward reference record if needed. This can return null if the value
339 /// exists but does not have the right type.
340 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc,
341 OperatorConstraint OC = OC_None);
342 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc,
343 OperatorConstraint OC = OC_None);
345 /// SetInstName - After an instruction is parsed and inserted into its
346 /// basic block, this installs its name.
347 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
350 /// GetBB - Get a basic block with the specified name or ID, creating a
351 /// forward reference record if needed. This can return null if the value
352 /// is not a BasicBlock.
353 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
354 BasicBlock *GetBB(unsigned ID, LocTy Loc);
356 /// DefineBB - Define the specified basic block, which is either named or
357 /// unnamed. If there is an error, this returns null otherwise it returns
358 /// the block being defined.
359 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
361 bool resolveForwardRefBlockAddresses();
364 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
365 PerFunctionState *PFS,
366 OperatorConstraint OC = OC_None);
368 bool parseConstantValue(Type *Ty, Constant *&C);
369 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS,
370 OperatorConstraint OC = OC_None);
371 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS,
372 OperatorConstraint OC = OC_None) {
373 return ParseValue(Ty, V, &PFS, OC);
375 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
376 PerFunctionState &PFS) {
378 return ParseValue(Ty, V, &PFS);
381 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
382 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
383 return ParseTypeAndValue(V, &PFS);
385 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
387 return ParseTypeAndValue(V, PFS);
389 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
390 PerFunctionState &PFS);
391 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
393 return ParseTypeAndBasicBlock(BB, Loc, PFS);
401 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
402 : Loc(loc), V(v), Attrs(attrs) {}
404 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
405 PerFunctionState &PFS,
406 bool IsMustTailCall = false,
407 bool InVarArgsFunc = false);
409 bool ParseExceptionArgs(SmallVectorImpl<Value *> &Args,
410 PerFunctionState &PFS);
413 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
414 bool ParseGlobalValue(Type *Ty, Constant *&V);
415 bool ParseGlobalTypeAndValue(Constant *&V);
416 bool ParseGlobalValueVector(SmallVectorImpl<Constant *> &Elts);
417 bool parseOptionalComdat(StringRef GlobalName, Comdat *&C);
418 bool ParseMetadataAsValue(Value *&V, PerFunctionState &PFS);
419 bool ParseValueAsMetadata(Metadata *&MD, const Twine &TypeMsg,
420 PerFunctionState *PFS);
421 bool ParseMetadata(Metadata *&MD, PerFunctionState *PFS);
422 bool ParseMDTuple(MDNode *&MD, bool IsDistinct = false);
423 bool ParseMDNode(MDNode *&MD);
424 bool ParseMDNodeTail(MDNode *&MD);
425 bool ParseMDNodeVector(SmallVectorImpl<Metadata *> &MDs);
426 bool ParseMetadataAttachment(unsigned &Kind, MDNode *&MD);
427 bool ParseInstructionMetadata(Instruction &Inst);
428 bool ParseOptionalFunctionMetadata(Function &F);
430 template <class FieldTy>
431 bool ParseMDField(LocTy Loc, StringRef Name, FieldTy &Result);
432 template <class FieldTy> bool ParseMDField(StringRef Name, FieldTy &Result);
433 template <class ParserTy>
434 bool ParseMDFieldsImplBody(ParserTy parseField);
435 template <class ParserTy>
436 bool ParseMDFieldsImpl(ParserTy parseField, LocTy &ClosingLoc);
437 bool ParseSpecializedMDNode(MDNode *&N, bool IsDistinct = false);
439 #define HANDLE_SPECIALIZED_MDNODE_LEAF(CLASS) \
440 bool Parse##CLASS(MDNode *&Result, bool IsDistinct);
441 #include "llvm/IR/Metadata.def"
449 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
450 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
452 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
453 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
454 bool ParseFunctionBody(Function &Fn);
455 bool ParseBasicBlock(PerFunctionState &PFS);
457 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
459 // Instruction Parsing. Each instruction parsing routine can return with a
460 // normal result, an error result, or return having eaten an extra comma.
461 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
462 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
463 PerFunctionState &PFS);
464 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
466 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
467 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
468 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
469 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
470 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
471 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
472 bool ParseCleanupRet(Instruction *&Inst, PerFunctionState &PFS);
473 bool ParseCatchRet(Instruction *&Inst, PerFunctionState &PFS);
474 bool ParseCatchPad(Instruction *&Inst, PerFunctionState &PFS);
475 bool ParseTerminatePad(Instruction *&Inst, PerFunctionState &PFS);
476 bool ParseCleanupPad(Instruction *&Inst, PerFunctionState &PFS);
477 bool ParseCatchEndPad(Instruction *&Inst, PerFunctionState &PFS);
478 bool ParseCleanupEndPad(Instruction *&Inst, PerFunctionState &PFS);
480 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
481 unsigned OperandType);
482 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
483 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
484 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
485 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
486 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
487 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
488 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
489 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
490 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
491 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
492 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
493 CallInst::TailCallKind IsTail);
494 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
495 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
496 int ParseStore(Instruction *&I, PerFunctionState &PFS);
497 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
498 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
499 int ParseFence(Instruction *&I, PerFunctionState &PFS);
500 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
501 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
502 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
504 // Use-list order directives.
505 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
506 bool ParseUseListOrderBB();
507 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
508 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
510 } // End llvm namespace