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"
42 /// ValID - Represents a reference of a definition of some sort with no type.
43 /// There are several cases where we have to parse the value but where the
44 /// type can depend on later context. This may either be a numeric reference
45 /// or a symbolic (%var) reference. This is just a discriminated union.
48 t_LocalID, t_GlobalID, // ID in UIntVal.
49 t_LocalName, t_GlobalName, // Name in StrVal.
50 t_APSInt, t_APFloat, // Value in APSIntVal/APFloatVal.
51 t_Null, t_Undef, t_Zero, // No value.
52 t_EmptyArray, // No value: []
53 t_Constant, // Value in ConstantVal.
54 t_InlineAsm, // Value in StrVal/StrVal2/UIntVal.
55 t_ConstantStruct, // Value in ConstantStructElts.
56 t_PackedConstantStruct // Value in ConstantStructElts.
61 std::string StrVal, StrVal2;
64 Constant *ConstantVal;
65 std::unique_ptr<Constant*[]> ConstantStructElts;
67 ValID() : Kind(t_LocalID), APFloatVal(0.0) {}
69 bool operator<(const ValID &RHS) const {
70 if (Kind == t_LocalID || Kind == t_GlobalID)
71 return UIntVal < RHS.UIntVal;
72 assert((Kind == t_LocalName || Kind == t_GlobalName ||
73 Kind == t_ConstantStruct || Kind == t_PackedConstantStruct) &&
74 "Ordering not defined for this ValID kind yet");
75 return StrVal < RHS.StrVal;
81 typedef LLLexer::LocTy LocTy;
87 // Instruction metadata resolution. Each instruction can have a list of
88 // MDRef info associated with them.
90 // The simpler approach of just creating temporary MDNodes and then calling
91 // RAUW on them when the definition is processed doesn't work because some
92 // instruction metadata kinds, such as dbg, get stored in the IR in an
93 // "optimized" format which doesn't participate in the normal value use
94 // lists. This means that RAUW doesn't work, even on temporary MDNodes
95 // which otherwise support RAUW. Instead, we defer resolving MDNode
96 // references until the definitions have been processed.
99 unsigned MDKind, MDSlot;
102 SmallVector<Instruction*, 64> InstsWithTBAATag;
104 // Type resolution handling data structures. The location is set when we
105 // have processed a use of the type but not a definition yet.
106 StringMap<std::pair<Type*, LocTy> > NamedTypes;
107 std::map<unsigned, std::pair<Type*, LocTy> > NumberedTypes;
109 std::map<unsigned, TrackingMDNodeRef> NumberedMetadata;
110 std::map<unsigned, std::pair<TempMDTuple, LocTy>> ForwardRefMDNodes;
112 // Global Value reference information.
113 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
114 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
115 std::vector<GlobalValue*> NumberedVals;
117 // Comdat forward reference information.
118 std::map<std::string, LocTy> ForwardRefComdats;
120 // References to blockaddress. The key is the function ValID, the value is
121 // a list of references to blocks in that function.
122 std::map<ValID, std::map<ValID, GlobalValue *>> ForwardRefBlockAddresses;
123 class PerFunctionState;
124 /// Reference to per-function state to allow basic blocks to be
125 /// forward-referenced by blockaddress instructions within the same
127 PerFunctionState *BlockAddressPFS;
129 // Attribute builder reference information.
130 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
131 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
134 LLParser(StringRef F, SourceMgr &SM, SMDiagnostic &Err, Module *m)
135 : Context(m->getContext()), Lex(F, SM, Err, m->getContext()), M(m),
136 BlockAddressPFS(nullptr) {}
139 LLVMContext &getContext() { return Context; }
143 bool Error(LocTy L, const Twine &Msg) const {
144 return Lex.Error(L, Msg);
146 bool TokError(const Twine &Msg) const {
147 return Error(Lex.getLoc(), Msg);
150 /// GetGlobalVal - Get a value with the specified name or ID, creating a
151 /// forward reference record if needed. This can return null if the value
152 /// exists but does not have the right type.
153 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
154 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
156 /// Get a Comdat with the specified name, creating a forward reference
157 /// record if needed.
158 Comdat *getComdat(const std::string &N, LocTy Loc);
161 bool ParseToken(lltok::Kind T, const char *ErrMsg);
162 bool EatIfPresent(lltok::Kind T) {
163 if (Lex.getKind() != T) return false;
168 FastMathFlags EatFastMathFlagsIfPresent() {
171 switch (Lex.getKind()) {
172 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
173 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
174 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
175 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
176 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
182 bool ParseOptionalToken(lltok::Kind T, bool &Present,
183 LocTy *Loc = nullptr) {
184 if (Lex.getKind() != T) {
194 bool ParseStringConstant(std::string &Result);
195 bool ParseUInt32(unsigned &Val);
196 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
198 return ParseUInt32(Val);
200 bool ParseUInt64(uint64_t &Val);
201 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
203 return ParseUInt64(Val);
206 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
207 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
208 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
209 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
211 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
212 bool ParseOptionalParamAttrs(AttrBuilder &B);
213 bool ParseOptionalReturnAttrs(AttrBuilder &B);
214 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
215 bool ParseOptionalLinkage(unsigned &Linkage) {
216 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
218 bool ParseOptionalVisibility(unsigned &Visibility);
219 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
220 bool ParseOptionalCallingConv(unsigned &CC);
221 bool ParseOptionalAlignment(unsigned &Alignment);
222 bool ParseOptionalDereferenceableBytes(uint64_t &Bytes);
223 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
224 AtomicOrdering &Ordering);
225 bool ParseOrdering(AtomicOrdering &Ordering);
226 bool ParseOptionalStackAlignment(unsigned &Alignment);
227 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
228 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
229 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
230 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
232 if (ParseIndexList(Indices, AteExtraComma)) return true;
234 return TokError("expected index");
238 // Top-Level Entities
239 bool ParseTopLevelEntities();
240 bool ValidateEndOfModule();
241 bool ParseTargetDefinition();
242 bool ParseModuleAsm();
243 bool ParseDepLibs(); // FIXME: Remove in 4.0.
244 bool ParseUnnamedType();
245 bool ParseNamedType();
249 bool ParseGlobalType(bool &IsConstant);
250 bool ParseUnnamedGlobal();
251 bool ParseNamedGlobal();
252 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
253 bool HasLinkage, unsigned Visibility,
254 unsigned DLLStorageClass,
255 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
256 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
257 unsigned Visibility, unsigned DLLStorageClass,
258 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
260 bool ParseStandaloneMetadata();
261 bool ParseNamedMetadata();
262 bool ParseMDString(MDString *&Result);
263 bool ParseMDNodeID(MDNode *&Result);
264 bool ParseUnnamedAttrGrp();
265 bool ParseFnAttributeValuePairs(AttrBuilder &B,
266 std::vector<unsigned> &FwdRefAttrGrps,
267 bool inAttrGrp, LocTy &BuiltinLoc);
270 bool ParseType(Type *&Result, const Twine &Msg, bool AllowVoid = false);
271 bool ParseType(Type *&Result, bool AllowVoid = false) {
272 return ParseType(Result, "expected type", AllowVoid);
274 bool ParseType(Type *&Result, const Twine &Msg, LocTy &Loc,
275 bool AllowVoid = false) {
277 return ParseType(Result, Msg, AllowVoid);
279 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
281 return ParseType(Result, AllowVoid);
283 bool ParseAnonStructType(Type *&Result, bool Packed);
284 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
285 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
286 std::pair<Type*, LocTy> &Entry,
289 bool ParseArrayVectorType(Type *&Result, bool isVector);
290 bool ParseFunctionType(Type *&Result);
292 // Function Semantic Analysis.
293 class PerFunctionState {
296 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
297 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
298 std::vector<Value*> NumberedVals;
300 /// FunctionNumber - If this is an unnamed function, this is the slot
301 /// number of it, otherwise it is -1.
304 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
307 Function &getFunction() const { return F; }
309 bool FinishFunction();
311 /// GetVal - Get a value with the specified name or ID, creating a
312 /// forward reference record if needed. This can return null if the value
313 /// exists but does not have the right type.
314 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
315 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
317 /// SetInstName - After an instruction is parsed and inserted into its
318 /// basic block, this installs its name.
319 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
322 /// GetBB - Get a basic block with the specified name or ID, creating a
323 /// forward reference record if needed. This can return null if the value
324 /// is not a BasicBlock.
325 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
326 BasicBlock *GetBB(unsigned ID, LocTy Loc);
328 /// DefineBB - Define the specified basic block, which is either named or
329 /// unnamed. If there is an error, this returns null otherwise it returns
330 /// the block being defined.
331 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
333 bool resolveForwardRefBlockAddresses();
336 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
337 PerFunctionState *PFS);
339 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
340 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
341 return ParseValue(Ty, V, &PFS);
343 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
344 PerFunctionState &PFS) {
346 return ParseValue(Ty, V, &PFS);
349 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
350 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
351 return ParseTypeAndValue(V, &PFS);
353 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
355 return ParseTypeAndValue(V, PFS);
357 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
358 PerFunctionState &PFS);
359 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
361 return ParseTypeAndBasicBlock(BB, Loc, PFS);
369 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
370 : Loc(loc), V(v), Attrs(attrs) {}
372 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
373 PerFunctionState &PFS,
374 bool IsMustTailCall = false,
375 bool InVarArgsFunc = false);
378 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
379 bool ParseGlobalValue(Type *Ty, Constant *&V);
380 bool ParseGlobalTypeAndValue(Constant *&V);
381 bool ParseGlobalValueVector(SmallVectorImpl<Constant *> &Elts);
382 bool parseOptionalComdat(StringRef GlobalName, Comdat *&C);
383 bool ParseMetadataAsValue(Value *&V, PerFunctionState &PFS);
384 bool ParseValueAsMetadata(Metadata *&MD, const Twine &TypeMsg,
385 PerFunctionState *PFS);
386 bool ParseMetadata(Metadata *&MD, PerFunctionState *PFS);
387 bool ParseMDTuple(MDNode *&MD, bool IsDistinct = false);
388 bool ParseMDNode(MDNode *&MD);
389 bool ParseMDNodeTail(MDNode *&MD);
390 bool ParseMDNodeVector(SmallVectorImpl<Metadata *> &MDs);
391 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
393 template <class FieldTy>
394 bool ParseMDField(LocTy Loc, StringRef Name, FieldTy &Result);
395 template <class FieldTy> bool ParseMDField(StringRef Name, FieldTy &Result);
396 template <class ParserTy>
397 bool ParseMDFieldsImplBody(ParserTy parseField);
398 template <class ParserTy>
399 bool ParseMDFieldsImpl(ParserTy parseField, LocTy &ClosingLoc);
400 bool ParseSpecializedMDNode(MDNode *&N, bool IsDistinct = false);
402 #define HANDLE_SPECIALIZED_MDNODE_LEAF(CLASS) \
403 bool Parse##CLASS(MDNode *&Result, bool IsDistinct);
404 #include "llvm/IR/Metadata.def"
412 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
413 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
415 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
416 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
417 bool ParseFunctionBody(Function &Fn);
418 bool ParseBasicBlock(PerFunctionState &PFS);
420 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
422 // Instruction Parsing. Each instruction parsing routine can return with a
423 // normal result, an error result, or return having eaten an extra comma.
424 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
425 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
426 PerFunctionState &PFS);
427 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
429 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
430 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
431 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
432 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
433 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
434 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
436 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
437 unsigned OperandType);
438 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
439 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
440 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
441 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
442 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
443 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
444 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
445 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
446 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
447 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
448 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
449 CallInst::TailCallKind IsTail);
450 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
451 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
452 int ParseStore(Instruction *&I, PerFunctionState &PFS);
453 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
454 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
455 int ParseFence(Instruction *&I, PerFunctionState &PFS);
456 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
457 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
458 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
460 // Use-list order directives.
461 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
462 bool ParseUseListOrderBB();
463 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
464 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
466 } // End llvm namespace