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;
71 : Kind(RHS.Kind), Loc(RHS.Loc), UIntVal(RHS.UIntVal), FTy(RHS.FTy),
72 StrVal(std::move(RHS.StrVal)), StrVal2(std::move(RHS.StrVal2)),
73 APSIntVal(std::move(RHS.APSIntVal)),
74 APFloatVal(std::move(RHS.APFloatVal)), ConstantVal(RHS.ConstantVal),
75 ConstantStructElts(std::move(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 SmallVector<Instruction*, 64> InstsWithTBAATag;
113 // Type resolution handling data structures. The location is set when we
114 // have processed a use of the type but not a definition yet.
115 StringMap<std::pair<Type*, LocTy> > NamedTypes;
116 std::map<unsigned, std::pair<Type*, LocTy> > NumberedTypes;
118 std::map<unsigned, TrackingMDNodeRef> NumberedMetadata;
119 std::map<unsigned, std::pair<TempMDTuple, LocTy>> ForwardRefMDNodes;
121 // Global Value reference information.
122 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
123 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
124 std::vector<GlobalValue*> NumberedVals;
126 // Comdat forward reference information.
127 std::map<std::string, LocTy> ForwardRefComdats;
129 // References to blockaddress. The key is the function ValID, the value is
130 // a list of references to blocks in that function.
131 std::map<ValID, std::map<ValID, GlobalValue *>> ForwardRefBlockAddresses;
132 class PerFunctionState;
133 /// Reference to per-function state to allow basic blocks to be
134 /// forward-referenced by blockaddress instructions within the same
136 PerFunctionState *BlockAddressPFS;
138 // Attribute builder reference information.
139 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
140 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
143 LLParser(StringRef F, SourceMgr &SM, SMDiagnostic &Err, Module *M,
144 SlotMapping *Slots = nullptr)
145 : Context(M->getContext()), Lex(F, SM, Err, M->getContext()), M(M),
146 Slots(Slots), BlockAddressPFS(nullptr) {}
149 bool parseStandaloneConstantValue(Constant *&C);
151 LLVMContext &getContext() { return Context; }
155 bool Error(LocTy L, const Twine &Msg) const {
156 return Lex.Error(L, Msg);
158 bool TokError(const Twine &Msg) const {
159 return Error(Lex.getLoc(), Msg);
162 /// GetGlobalVal - Get a value with the specified name or ID, creating a
163 /// forward reference record if needed. This can return null if the value
164 /// exists but does not have the right type.
165 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
166 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
168 /// Get a Comdat with the specified name, creating a forward reference
169 /// record if needed.
170 Comdat *getComdat(const std::string &N, LocTy Loc);
173 bool ParseToken(lltok::Kind T, const char *ErrMsg);
174 bool EatIfPresent(lltok::Kind T) {
175 if (Lex.getKind() != T) return false;
180 FastMathFlags EatFastMathFlagsIfPresent() {
183 switch (Lex.getKind()) {
184 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
185 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
186 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
187 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
188 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
194 bool ParseOptionalToken(lltok::Kind T, bool &Present,
195 LocTy *Loc = nullptr) {
196 if (Lex.getKind() != T) {
206 bool ParseStringConstant(std::string &Result);
207 bool ParseUInt32(unsigned &Val);
208 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
210 return ParseUInt32(Val);
212 bool ParseUInt64(uint64_t &Val);
213 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
215 return ParseUInt64(Val);
218 bool ParseStringAttribute(AttrBuilder &B);
220 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
221 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
222 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
223 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
225 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
226 bool ParseOptionalParamAttrs(AttrBuilder &B);
227 bool ParseOptionalReturnAttrs(AttrBuilder &B);
228 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
229 bool ParseOptionalLinkage(unsigned &Linkage) {
230 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
232 bool ParseOptionalVisibility(unsigned &Visibility);
233 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
234 bool ParseOptionalCallingConv(unsigned &CC);
235 bool ParseOptionalAlignment(unsigned &Alignment);
236 bool ParseOptionalDerefAttrBytes(lltok::Kind AttrKind, uint64_t &Bytes);
237 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
238 AtomicOrdering &Ordering);
239 bool ParseOrdering(AtomicOrdering &Ordering);
240 bool ParseOptionalStackAlignment(unsigned &Alignment);
241 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
242 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
243 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
244 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
246 if (ParseIndexList(Indices, AteExtraComma)) return true;
248 return TokError("expected index");
252 // Top-Level Entities
253 bool ParseTopLevelEntities();
254 bool ValidateEndOfModule();
255 bool ParseTargetDefinition();
256 bool ParseModuleAsm();
257 bool ParseDepLibs(); // FIXME: Remove in 4.0.
258 bool ParseUnnamedType();
259 bool ParseNamedType();
263 bool ParseGlobalType(bool &IsConstant);
264 bool ParseUnnamedGlobal();
265 bool ParseNamedGlobal();
266 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
267 bool HasLinkage, unsigned Visibility,
268 unsigned DLLStorageClass,
269 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
270 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
271 unsigned Visibility, unsigned DLLStorageClass,
272 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
274 bool ParseStandaloneMetadata();
275 bool ParseNamedMetadata();
276 bool ParseMDString(MDString *&Result);
277 bool ParseMDNodeID(MDNode *&Result);
278 bool ParseUnnamedAttrGrp();
279 bool ParseFnAttributeValuePairs(AttrBuilder &B,
280 std::vector<unsigned> &FwdRefAttrGrps,
281 bool inAttrGrp, LocTy &BuiltinLoc);
284 bool ParseType(Type *&Result, const Twine &Msg, bool AllowVoid = false);
285 bool ParseType(Type *&Result, bool AllowVoid = false) {
286 return ParseType(Result, "expected type", AllowVoid);
288 bool ParseType(Type *&Result, const Twine &Msg, LocTy &Loc,
289 bool AllowVoid = false) {
291 return ParseType(Result, Msg, AllowVoid);
293 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
295 return ParseType(Result, AllowVoid);
297 bool ParseAnonStructType(Type *&Result, bool Packed);
298 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
299 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
300 std::pair<Type*, LocTy> &Entry,
303 bool ParseArrayVectorType(Type *&Result, bool isVector);
304 bool ParseFunctionType(Type *&Result);
306 // Function Semantic Analysis.
307 class PerFunctionState {
310 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
311 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
312 std::vector<Value*> NumberedVals;
314 /// FunctionNumber - If this is an unnamed function, this is the slot
315 /// number of it, otherwise it is -1.
318 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
321 Function &getFunction() const { return F; }
323 bool FinishFunction();
325 /// GetVal - Get a value with the specified name or ID, creating a
326 /// forward reference record if needed. This can return null if the value
327 /// exists but does not have the right type.
328 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
329 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
331 /// SetInstName - After an instruction is parsed and inserted into its
332 /// basic block, this installs its name.
333 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
336 /// GetBB - Get a basic block with the specified name or ID, creating a
337 /// forward reference record if needed. This can return null if the value
338 /// is not a BasicBlock.
339 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
340 BasicBlock *GetBB(unsigned ID, LocTy Loc);
342 /// DefineBB - Define the specified basic block, which is either named or
343 /// unnamed. If there is an error, this returns null otherwise it returns
344 /// the block being defined.
345 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
347 bool resolveForwardRefBlockAddresses();
350 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
351 PerFunctionState *PFS);
353 bool parseConstantValue(Type *Ty, Constant *&C);
354 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
355 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
356 return ParseValue(Ty, V, &PFS);
358 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
359 PerFunctionState &PFS) {
361 return ParseValue(Ty, V, &PFS);
364 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
365 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
366 return ParseTypeAndValue(V, &PFS);
368 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
370 return ParseTypeAndValue(V, PFS);
372 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
373 PerFunctionState &PFS);
374 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
376 return ParseTypeAndBasicBlock(BB, Loc, PFS);
384 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
385 : Loc(loc), V(v), Attrs(attrs) {}
387 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
388 PerFunctionState &PFS,
389 bool IsMustTailCall = false,
390 bool InVarArgsFunc = false);
392 bool ParseExceptionArgs(SmallVectorImpl<Value *> &Args,
393 PerFunctionState &PFS);
396 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
397 bool ParseGlobalValue(Type *Ty, Constant *&V);
398 bool ParseGlobalTypeAndValue(Constant *&V);
399 bool ParseGlobalValueVector(SmallVectorImpl<Constant *> &Elts);
400 bool parseOptionalComdat(StringRef GlobalName, Comdat *&C);
401 bool ParseMetadataAsValue(Value *&V, PerFunctionState &PFS);
402 bool ParseValueAsMetadata(Metadata *&MD, const Twine &TypeMsg,
403 PerFunctionState *PFS);
404 bool ParseMetadata(Metadata *&MD, PerFunctionState *PFS);
405 bool ParseMDTuple(MDNode *&MD, bool IsDistinct = false);
406 bool ParseMDNode(MDNode *&MD);
407 bool ParseMDNodeTail(MDNode *&MD);
408 bool ParseMDNodeVector(SmallVectorImpl<Metadata *> &MDs);
409 bool ParseMetadataAttachment(unsigned &Kind, MDNode *&MD);
410 bool ParseInstructionMetadata(Instruction &Inst);
411 bool ParseOptionalFunctionMetadata(Function &F);
413 template <class FieldTy>
414 bool ParseMDField(LocTy Loc, StringRef Name, FieldTy &Result);
415 template <class FieldTy> bool ParseMDField(StringRef Name, FieldTy &Result);
416 template <class ParserTy>
417 bool ParseMDFieldsImplBody(ParserTy parseField);
418 template <class ParserTy>
419 bool ParseMDFieldsImpl(ParserTy parseField, LocTy &ClosingLoc);
420 bool ParseSpecializedMDNode(MDNode *&N, bool IsDistinct = false);
422 #define HANDLE_SPECIALIZED_MDNODE_LEAF(CLASS) \
423 bool Parse##CLASS(MDNode *&Result, bool IsDistinct);
424 #include "llvm/IR/Metadata.def"
432 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
433 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
435 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
436 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
437 bool ParseFunctionBody(Function &Fn);
438 bool ParseBasicBlock(PerFunctionState &PFS);
440 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
442 // Instruction Parsing. Each instruction parsing routine can return with a
443 // normal result, an error result, or return having eaten an extra comma.
444 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
445 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
446 PerFunctionState &PFS);
447 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
449 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
450 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
451 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
452 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
453 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
454 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
455 bool ParseCleanupRet(Instruction *&Inst, PerFunctionState &PFS);
456 bool ParseCatchRet(Instruction *&Inst, PerFunctionState &PFS);
457 bool ParseCatchPad(Instruction *&Inst, PerFunctionState &PFS);
458 bool ParseTerminatePad(Instruction *&Inst, PerFunctionState &PFS);
459 bool ParseCleanupPad(Instruction *&Inst, PerFunctionState &PFS);
460 bool ParseCatchEndPad(Instruction *&Inst, PerFunctionState &PFS);
462 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
463 unsigned OperandType);
464 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
465 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
466 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
467 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
468 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
469 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
470 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
471 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
472 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
473 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
474 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
475 CallInst::TailCallKind IsTail);
476 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
477 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
478 int ParseStore(Instruction *&I, PerFunctionState &PFS);
479 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
480 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
481 int ParseFence(Instruction *&I, PerFunctionState &PFS);
482 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
483 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
484 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
486 // Use-list order directives.
487 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
488 bool ParseUseListOrderBB();
489 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
490 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
492 } // End llvm namespace