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_Metadata, // Value in MetadataVal.
56 t_ConstantStruct, // Value in ConstantStructElts.
57 t_PackedConstantStruct // Value in ConstantStructElts.
62 std::string StrVal, StrVal2;
65 Constant *ConstantVal;
66 MetadataAsValue *MetadataVal;
67 Constant **ConstantStructElts;
69 ValID() : Kind(t_LocalID), APFloatVal(0.0) {}
71 if (Kind == t_ConstantStruct || Kind == t_PackedConstantStruct)
72 delete [] ConstantStructElts;
75 bool operator<(const ValID &RHS) const {
76 if (Kind == t_LocalID || Kind == t_GlobalID)
77 return UIntVal < RHS.UIntVal;
78 assert((Kind == t_LocalName || Kind == t_GlobalName ||
79 Kind == t_ConstantStruct || Kind == t_PackedConstantStruct) &&
80 "Ordering not defined for this ValID kind yet");
81 return StrVal < RHS.StrVal;
87 typedef LLLexer::LocTy LocTy;
93 // Instruction metadata resolution. Each instruction can have a list of
94 // MDRef info associated with them.
96 // The simpler approach of just creating temporary MDNodes and then calling
97 // RAUW on them when the definition is processed doesn't work because some
98 // instruction metadata kinds, such as dbg, get stored in the IR in an
99 // "optimized" format which doesn't participate in the normal value use
100 // lists. This means that RAUW doesn't work, even on temporary MDNodes
101 // which otherwise support RAUW. Instead, we defer resolving MDNode
102 // references until the definitions have been processed.
105 unsigned MDKind, MDSlot;
107 DenseMap<Instruction*, std::vector<MDRef> > ForwardRefInstMetadata;
109 SmallVector<Instruction*, 64> InstsWithTBAATag;
111 // Type resolution handling data structures. The location is set when we
112 // have processed a use of the type but not a definition yet.
113 StringMap<std::pair<Type*, LocTy> > NamedTypes;
114 std::vector<std::pair<Type*, LocTy> > NumberedTypes;
116 std::vector<TrackingMDNodeRef> NumberedMetadata;
117 std::map<unsigned, std::pair<MDNodeFwdDecl *, LocTy>> ForwardRefMDNodes;
119 // Global Value reference information.
120 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
121 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
122 std::vector<GlobalValue*> NumberedVals;
124 // Comdat forward reference information.
125 std::map<std::string, LocTy> ForwardRefComdats;
127 // References to blockaddress. The key is the function ValID, the value is
128 // a list of references to blocks in that function.
129 std::map<ValID, std::map<ValID, GlobalValue *>> ForwardRefBlockAddresses;
130 class PerFunctionState;
131 /// Reference to per-function state to allow basic blocks to be
132 /// forward-referenced by blockaddress instructions within the same
134 PerFunctionState *BlockAddressPFS;
136 // Attribute builder reference information.
137 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
138 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
141 LLParser(StringRef F, SourceMgr &SM, SMDiagnostic &Err, Module *m)
142 : Context(m->getContext()), Lex(F, SM, Err, m->getContext()), M(m),
143 BlockAddressPFS(nullptr) {}
146 LLVMContext &getContext() { return Context; }
150 bool Error(LocTy L, const Twine &Msg) const {
151 return Lex.Error(L, Msg);
153 bool TokError(const Twine &Msg) const {
154 return Error(Lex.getLoc(), Msg);
157 /// GetGlobalVal - Get a value with the specified name or ID, creating a
158 /// forward reference record if needed. This can return null if the value
159 /// exists but does not have the right type.
160 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
161 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
163 /// Get a Comdat with the specified name, creating a forward reference
164 /// record if needed.
165 Comdat *getComdat(const std::string &N, LocTy Loc);
168 bool ParseToken(lltok::Kind T, const char *ErrMsg);
169 bool EatIfPresent(lltok::Kind T) {
170 if (Lex.getKind() != T) return false;
175 FastMathFlags EatFastMathFlagsIfPresent() {
178 switch (Lex.getKind()) {
179 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
180 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
181 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
182 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
183 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
189 bool ParseOptionalToken(lltok::Kind T, bool &Present,
190 LocTy *Loc = nullptr) {
191 if (Lex.getKind() != T) {
201 bool ParseStringConstant(std::string &Result);
202 bool ParseUInt32(unsigned &Val);
203 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
205 return ParseUInt32(Val);
207 bool ParseUInt64(uint64_t &Val);
208 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
210 return ParseUInt64(Val);
213 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
214 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
215 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
216 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
218 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
219 bool ParseOptionalParamAttrs(AttrBuilder &B);
220 bool ParseOptionalReturnAttrs(AttrBuilder &B);
221 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
222 bool ParseOptionalLinkage(unsigned &Linkage) {
223 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
225 bool ParseOptionalVisibility(unsigned &Visibility);
226 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
227 bool ParseOptionalCallingConv(unsigned &CC);
228 bool ParseOptionalAlignment(unsigned &Alignment);
229 bool ParseOptionalDereferenceableBytes(uint64_t &Bytes);
230 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
231 AtomicOrdering &Ordering);
232 bool ParseOrdering(AtomicOrdering &Ordering);
233 bool ParseOptionalStackAlignment(unsigned &Alignment);
234 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
235 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
236 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
237 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
239 if (ParseIndexList(Indices, AteExtraComma)) return true;
241 return TokError("expected index");
245 // Top-Level Entities
246 bool ParseTopLevelEntities();
247 bool ValidateEndOfModule();
248 bool ParseTargetDefinition();
249 bool ParseModuleAsm();
250 bool ParseDepLibs(); // FIXME: Remove in 4.0.
251 bool ParseUnnamedType();
252 bool ParseNamedType();
256 bool ParseGlobalType(bool &IsConstant);
257 bool ParseUnnamedGlobal();
258 bool ParseNamedGlobal();
259 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
260 bool HasLinkage, unsigned Visibility,
261 unsigned DLLStorageClass,
262 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
263 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
264 unsigned Visibility, unsigned DLLStorageClass,
265 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
267 bool ParseStandaloneMetadata();
268 bool ParseNamedMetadata();
269 bool ParseMDString(MDString *&Result);
270 bool ParseMDNodeID(MDNode *&Result);
271 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
272 bool ParseUnnamedAttrGrp();
273 bool ParseFnAttributeValuePairs(AttrBuilder &B,
274 std::vector<unsigned> &FwdRefAttrGrps,
275 bool inAttrGrp, LocTy &BuiltinLoc);
278 bool ParseType(Type *&Result, bool AllowVoid = false);
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(Comdat *&C);
383 bool ParseMetadataAsValue(ValID &ID, PerFunctionState *PFS);
384 bool ParseMetadata(Metadata *&MD, PerFunctionState *PFS);
385 bool ParseMDNode(MDNode *&MD);
386 bool ParseMDNodeOrLocal(Metadata *&MD, PerFunctionState *PFS);
387 bool ParseMDNodeVector(SmallVectorImpl<Metadata *> &,
388 PerFunctionState *PFS);
389 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
397 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
398 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
400 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
401 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
402 bool ParseFunctionBody(Function &Fn);
403 bool ParseBasicBlock(PerFunctionState &PFS);
405 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
407 // Instruction Parsing. Each instruction parsing routine can return with a
408 // normal result, an error result, or return having eaten an extra comma.
409 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
410 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
411 PerFunctionState &PFS);
412 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
414 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
415 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
416 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
417 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
418 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
419 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
421 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
422 unsigned OperandType);
423 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
424 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
425 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
426 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
427 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
428 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
429 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
430 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
431 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
432 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
433 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
434 CallInst::TailCallKind IsTail);
435 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
436 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
437 int ParseStore(Instruction *&I, PerFunctionState &PFS);
438 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
439 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
440 int ParseFence(Instruction *&I, PerFunctionState &PFS);
441 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
442 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
443 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
445 // Use-list order directives.
446 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
447 bool ParseUseListOrderBB();
448 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
449 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
451 } // End llvm namespace