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 Constant **ConstantStructElts;
67 ValID() : Kind(t_LocalID), APFloatVal(0.0) {}
69 if (Kind == t_ConstantStruct || Kind == t_PackedConstantStruct)
70 delete [] ConstantStructElts;
73 bool operator<(const ValID &RHS) const {
74 if (Kind == t_LocalID || Kind == t_GlobalID)
75 return UIntVal < RHS.UIntVal;
76 assert((Kind == t_LocalName || Kind == t_GlobalName ||
77 Kind == t_ConstantStruct || Kind == t_PackedConstantStruct) &&
78 "Ordering not defined for this ValID kind yet");
79 return StrVal < RHS.StrVal;
85 typedef LLLexer::LocTy LocTy;
91 // Instruction metadata resolution. Each instruction can have a list of
92 // MDRef info associated with them.
94 // The simpler approach of just creating temporary MDNodes and then calling
95 // RAUW on them when the definition is processed doesn't work because some
96 // instruction metadata kinds, such as dbg, get stored in the IR in an
97 // "optimized" format which doesn't participate in the normal value use
98 // lists. This means that RAUW doesn't work, even on temporary MDNodes
99 // which otherwise support RAUW. Instead, we defer resolving MDNode
100 // references until the definitions have been processed.
103 unsigned MDKind, MDSlot;
105 DenseMap<Instruction*, std::vector<MDRef> > ForwardRefInstMetadata;
107 SmallVector<Instruction*, 64> InstsWithTBAATag;
109 // Type resolution handling data structures. The location is set when we
110 // have processed a use of the type but not a definition yet.
111 StringMap<std::pair<Type*, LocTy> > NamedTypes;
112 std::vector<std::pair<Type*, LocTy> > NumberedTypes;
114 std::vector<TrackingMDNodeRef> NumberedMetadata;
115 std::map<unsigned, std::pair<MDNodeFwdDecl *, LocTy>> ForwardRefMDNodes;
117 // Global Value reference information.
118 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
119 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
120 std::vector<GlobalValue*> NumberedVals;
122 // Comdat forward reference information.
123 std::map<std::string, LocTy> ForwardRefComdats;
125 // References to blockaddress. The key is the function ValID, the value is
126 // a list of references to blocks in that function.
127 std::map<ValID, std::map<ValID, GlobalValue *>> ForwardRefBlockAddresses;
128 class PerFunctionState;
129 /// Reference to per-function state to allow basic blocks to be
130 /// forward-referenced by blockaddress instructions within the same
132 PerFunctionState *BlockAddressPFS;
134 // Attribute builder reference information.
135 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
136 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
139 LLParser(StringRef F, SourceMgr &SM, SMDiagnostic &Err, Module *m)
140 : Context(m->getContext()), Lex(F, SM, Err, m->getContext()), M(m),
141 BlockAddressPFS(nullptr) {}
144 LLVMContext &getContext() { return Context; }
148 bool Error(LocTy L, const Twine &Msg) const {
149 return Lex.Error(L, Msg);
151 bool TokError(const Twine &Msg) const {
152 return Error(Lex.getLoc(), Msg);
155 /// GetGlobalVal - Get a value with the specified name or ID, creating a
156 /// forward reference record if needed. This can return null if the value
157 /// exists but does not have the right type.
158 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
159 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
161 /// Get a Comdat with the specified name, creating a forward reference
162 /// record if needed.
163 Comdat *getComdat(const std::string &N, LocTy Loc);
166 bool ParseToken(lltok::Kind T, const char *ErrMsg);
167 bool EatIfPresent(lltok::Kind T) {
168 if (Lex.getKind() != T) return false;
173 FastMathFlags EatFastMathFlagsIfPresent() {
176 switch (Lex.getKind()) {
177 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
178 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
179 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
180 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
181 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
187 bool ParseOptionalToken(lltok::Kind T, bool &Present,
188 LocTy *Loc = nullptr) {
189 if (Lex.getKind() != T) {
199 bool ParseStringConstant(std::string &Result);
200 bool ParseUInt32(unsigned &Val);
201 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
203 return ParseUInt32(Val);
205 bool ParseUInt64(uint64_t &Val);
206 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
208 return ParseUInt64(Val);
211 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
212 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
213 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
214 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
216 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
217 bool ParseOptionalParamAttrs(AttrBuilder &B);
218 bool ParseOptionalReturnAttrs(AttrBuilder &B);
219 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
220 bool ParseOptionalLinkage(unsigned &Linkage) {
221 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
223 bool ParseOptionalVisibility(unsigned &Visibility);
224 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
225 bool ParseOptionalCallingConv(unsigned &CC);
226 bool ParseOptionalAlignment(unsigned &Alignment);
227 bool ParseOptionalDereferenceableBytes(uint64_t &Bytes);
228 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
229 AtomicOrdering &Ordering);
230 bool ParseOrdering(AtomicOrdering &Ordering);
231 bool ParseOptionalStackAlignment(unsigned &Alignment);
232 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
233 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
234 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
235 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
237 if (ParseIndexList(Indices, AteExtraComma)) return true;
239 return TokError("expected index");
243 // Top-Level Entities
244 bool ParseTopLevelEntities();
245 bool ValidateEndOfModule();
246 bool ParseTargetDefinition();
247 bool ParseModuleAsm();
248 bool ParseDepLibs(); // FIXME: Remove in 4.0.
249 bool ParseUnnamedType();
250 bool ParseNamedType();
254 bool ParseGlobalType(bool &IsConstant);
255 bool ParseUnnamedGlobal();
256 bool ParseNamedGlobal();
257 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
258 bool HasLinkage, unsigned Visibility,
259 unsigned DLLStorageClass,
260 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
261 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
262 unsigned Visibility, unsigned DLLStorageClass,
263 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
265 bool ParseStandaloneMetadata();
266 bool ParseNamedMetadata();
267 bool ParseMDString(MDString *&Result);
268 bool ParseMDNodeID(MDNode *&Result);
269 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
270 bool ParseUnnamedAttrGrp();
271 bool ParseFnAttributeValuePairs(AttrBuilder &B,
272 std::vector<unsigned> &FwdRefAttrGrps,
273 bool inAttrGrp, LocTy &BuiltinLoc);
276 bool ParseType(Type *&Result, const Twine &Msg, bool AllowVoid = false);
277 bool ParseType(Type *&Result, bool AllowVoid = false) {
278 return ParseType(Result, "expected type", AllowVoid);
280 bool ParseType(Type *&Result, const Twine &Msg, LocTy &Loc,
281 bool AllowVoid = false) {
283 return ParseType(Result, Msg, AllowVoid);
285 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
287 return ParseType(Result, AllowVoid);
289 bool ParseAnonStructType(Type *&Result, bool Packed);
290 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
291 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
292 std::pair<Type*, LocTy> &Entry,
295 bool ParseArrayVectorType(Type *&Result, bool isVector);
296 bool ParseFunctionType(Type *&Result);
298 // Function Semantic Analysis.
299 class PerFunctionState {
302 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
303 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
304 std::vector<Value*> NumberedVals;
306 /// FunctionNumber - If this is an unnamed function, this is the slot
307 /// number of it, otherwise it is -1.
310 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
313 Function &getFunction() const { return F; }
315 bool FinishFunction();
317 /// GetVal - Get a value with the specified name or ID, creating a
318 /// forward reference record if needed. This can return null if the value
319 /// exists but does not have the right type.
320 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
321 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
323 /// SetInstName - After an instruction is parsed and inserted into its
324 /// basic block, this installs its name.
325 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
328 /// GetBB - Get a basic block with the specified name or ID, creating a
329 /// forward reference record if needed. This can return null if the value
330 /// is not a BasicBlock.
331 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
332 BasicBlock *GetBB(unsigned ID, LocTy Loc);
334 /// DefineBB - Define the specified basic block, which is either named or
335 /// unnamed. If there is an error, this returns null otherwise it returns
336 /// the block being defined.
337 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
339 bool resolveForwardRefBlockAddresses();
342 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
343 PerFunctionState *PFS);
345 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
346 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
347 return ParseValue(Ty, V, &PFS);
349 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
350 PerFunctionState &PFS) {
352 return ParseValue(Ty, V, &PFS);
355 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
356 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
357 return ParseTypeAndValue(V, &PFS);
359 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
361 return ParseTypeAndValue(V, PFS);
363 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
364 PerFunctionState &PFS);
365 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
367 return ParseTypeAndBasicBlock(BB, Loc, PFS);
375 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
376 : Loc(loc), V(v), Attrs(attrs) {}
378 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
379 PerFunctionState &PFS,
380 bool IsMustTailCall = false,
381 bool InVarArgsFunc = false);
384 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
385 bool ParseGlobalValue(Type *Ty, Constant *&V);
386 bool ParseGlobalTypeAndValue(Constant *&V);
387 bool ParseGlobalValueVector(SmallVectorImpl<Constant *> &Elts);
388 bool parseOptionalComdat(Comdat *&C);
389 bool ParseMetadataAsValue(Value *&V, PerFunctionState &PFS);
390 bool ParseValueAsMetadata(Metadata *&MD, PerFunctionState *PFS);
391 bool ParseMetadata(Metadata *&MD, PerFunctionState *PFS);
392 bool ParseMDNode(MDNode *&MD);
393 bool ParseMDNodeVector(SmallVectorImpl<Metadata *> &MDs);
394 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
402 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
403 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
405 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
406 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
407 bool ParseFunctionBody(Function &Fn);
408 bool ParseBasicBlock(PerFunctionState &PFS);
410 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
412 // Instruction Parsing. Each instruction parsing routine can return with a
413 // normal result, an error result, or return having eaten an extra comma.
414 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
415 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
416 PerFunctionState &PFS);
417 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
419 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
420 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
421 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
422 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
423 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
424 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
426 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
427 unsigned OperandType);
428 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
429 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
430 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
431 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
432 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
433 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
434 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
435 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
436 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
437 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
438 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
439 CallInst::TailCallKind IsTail);
440 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
441 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
442 int ParseStore(Instruction *&I, PerFunctionState &PFS);
443 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
444 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
445 int ParseFence(Instruction *&I, PerFunctionState &PFS);
446 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
447 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
448 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
450 // Use-list order directives.
451 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
452 bool ParseUseListOrderBB();
453 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
454 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
456 } // End llvm namespace