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_MDNode, // Value in MDNodeVal.
56 t_MDString, // Value in MDStringVal.
57 t_ConstantStruct, // Value in ConstantStructElts.
58 t_PackedConstantStruct // Value in ConstantStructElts.
63 std::string StrVal, StrVal2;
66 Constant *ConstantVal;
68 MDString *MDStringVal;
69 Constant **ConstantStructElts;
71 ValID() : Kind(t_LocalID), APFloatVal(0.0) {}
73 if (Kind == t_ConstantStruct || Kind == t_PackedConstantStruct)
74 delete [] 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;
95 // Instruction metadata resolution. Each instruction can have a list of
96 // MDRef info associated with them.
98 // The simpler approach of just creating temporary MDNodes and then calling
99 // RAUW on them when the definition is processed doesn't work because some
100 // instruction metadata kinds, such as dbg, get stored in the IR in an
101 // "optimized" format which doesn't participate in the normal value use
102 // lists. This means that RAUW doesn't work, even on temporary MDNodes
103 // which otherwise support RAUW. Instead, we defer resolving MDNode
104 // references until the definitions have been processed.
107 unsigned MDKind, MDSlot;
109 DenseMap<Instruction*, std::vector<MDRef> > ForwardRefInstMetadata;
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::vector<std::pair<Type*, LocTy> > NumberedTypes;
118 std::vector<TrackingVH<MDNode> > NumberedMetadata;
119 std::map<unsigned, std::pair<TrackingVH<MDNode>, 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 : Context(m->getContext()), Lex(F, SM, Err, m->getContext()), M(m),
145 BlockAddressPFS(nullptr) {}
148 LLVMContext &getContext() { return Context; }
152 bool Error(LocTy L, const Twine &Msg) const {
153 return Lex.Error(L, Msg);
155 bool TokError(const Twine &Msg) const {
156 return Error(Lex.getLoc(), Msg);
159 /// GetGlobalVal - Get a value with the specified name or ID, creating a
160 /// forward reference record if needed. This can return null if the value
161 /// exists but does not have the right type.
162 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
163 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
165 /// Get a Comdat with the specified name, creating a forward reference
166 /// record if needed.
167 Comdat *getComdat(const std::string &N, LocTy Loc);
170 bool ParseToken(lltok::Kind T, const char *ErrMsg);
171 bool EatIfPresent(lltok::Kind T) {
172 if (Lex.getKind() != T) return false;
177 FastMathFlags EatFastMathFlagsIfPresent() {
180 switch (Lex.getKind()) {
181 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
182 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
183 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
184 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
185 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
191 bool ParseOptionalToken(lltok::Kind T, bool &Present,
192 LocTy *Loc = nullptr) {
193 if (Lex.getKind() != T) {
203 bool ParseStringConstant(std::string &Result);
204 bool ParseUInt32(unsigned &Val);
205 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
207 return ParseUInt32(Val);
209 bool ParseUInt64(uint64_t &Val);
210 bool ParseUInt64(uint64_t &Val, LocTy &Loc) {
212 return ParseUInt64(Val);
215 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
216 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
217 bool parseOptionalUnnamedAddr(bool &UnnamedAddr) {
218 return ParseOptionalToken(lltok::kw_unnamed_addr, UnnamedAddr);
220 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
221 bool ParseOptionalParamAttrs(AttrBuilder &B);
222 bool ParseOptionalReturnAttrs(AttrBuilder &B);
223 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
224 bool ParseOptionalLinkage(unsigned &Linkage) {
225 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
227 bool ParseOptionalVisibility(unsigned &Visibility);
228 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
229 bool ParseOptionalCallingConv(CallingConv::ID &CC);
230 bool ParseOptionalAlignment(unsigned &Alignment);
231 bool ParseOptionalDereferenceableBytes(uint64_t &Bytes);
232 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
233 AtomicOrdering &Ordering);
234 bool ParseOrdering(AtomicOrdering &Ordering);
235 bool ParseOptionalStackAlignment(unsigned &Alignment);
236 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
237 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
238 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
239 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
241 if (ParseIndexList(Indices, AteExtraComma)) return true;
243 return TokError("expected index");
247 // Top-Level Entities
248 bool ParseTopLevelEntities();
249 bool ValidateEndOfModule();
250 bool ParseTargetDefinition();
251 bool ParseModuleAsm();
252 bool ParseDepLibs(); // FIXME: Remove in 4.0.
253 bool ParseUnnamedType();
254 bool ParseNamedType();
258 bool ParseGlobalType(bool &IsConstant);
259 bool ParseUnnamedGlobal();
260 bool ParseNamedGlobal();
261 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
262 bool HasLinkage, unsigned Visibility,
263 unsigned DLLStorageClass,
264 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
265 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Linkage,
266 unsigned Visibility, unsigned DLLStorageClass,
267 GlobalVariable::ThreadLocalMode TLM, bool UnnamedAddr);
269 bool ParseStandaloneMetadata();
270 bool ParseNamedMetadata();
271 bool ParseMDString(MDString *&Result);
272 bool ParseMDNodeID(MDNode *&Result);
273 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
274 bool ParseUnnamedAttrGrp();
275 bool ParseFnAttributeValuePairs(AttrBuilder &B,
276 std::vector<unsigned> &FwdRefAttrGrps,
277 bool inAttrGrp, LocTy &BuiltinLoc);
280 bool ParseType(Type *&Result, bool AllowVoid = false);
281 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
283 return ParseType(Result, AllowVoid);
285 bool ParseAnonStructType(Type *&Result, bool Packed);
286 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
287 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
288 std::pair<Type*, LocTy> &Entry,
291 bool ParseArrayVectorType(Type *&Result, bool isVector);
292 bool ParseFunctionType(Type *&Result);
294 // Function Semantic Analysis.
295 class PerFunctionState {
298 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
299 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
300 std::vector<Value*> NumberedVals;
302 /// FunctionNumber - If this is an unnamed function, this is the slot
303 /// number of it, otherwise it is -1.
306 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
309 Function &getFunction() const { return F; }
311 bool FinishFunction();
313 /// GetVal - Get a value with the specified name or ID, creating a
314 /// forward reference record if needed. This can return null if the value
315 /// exists but does not have the right type.
316 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
317 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
319 /// SetInstName - After an instruction is parsed and inserted into its
320 /// basic block, this installs its name.
321 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
324 /// GetBB - Get a basic block with the specified name or ID, creating a
325 /// forward reference record if needed. This can return null if the value
326 /// is not a BasicBlock.
327 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
328 BasicBlock *GetBB(unsigned ID, LocTy Loc);
330 /// DefineBB - Define the specified basic block, which is either named or
331 /// unnamed. If there is an error, this returns null otherwise it returns
332 /// the block being defined.
333 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
335 bool resolveForwardRefBlockAddresses();
338 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
339 PerFunctionState *PFS);
341 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
342 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
343 return ParseValue(Ty, V, &PFS);
345 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
346 PerFunctionState &PFS) {
348 return ParseValue(Ty, V, &PFS);
351 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
352 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
353 return ParseTypeAndValue(V, &PFS);
355 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
357 return ParseTypeAndValue(V, PFS);
359 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
360 PerFunctionState &PFS);
361 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
363 return ParseTypeAndBasicBlock(BB, Loc, PFS);
371 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
372 : Loc(loc), V(v), Attrs(attrs) {}
374 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
375 PerFunctionState &PFS,
376 bool IsMustTailCall = false,
377 bool InVarArgsFunc = false);
380 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
381 bool ParseGlobalValue(Type *Ty, Constant *&V);
382 bool ParseGlobalTypeAndValue(Constant *&V);
383 bool ParseGlobalValueVector(SmallVectorImpl<Constant *> &Elts);
384 bool parseOptionalComdat(Comdat *&C);
385 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
386 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
387 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
388 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
396 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
397 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
399 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
400 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
401 bool ParseFunctionBody(Function &Fn);
402 bool ParseBasicBlock(PerFunctionState &PFS);
404 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
406 // Instruction Parsing. Each instruction parsing routine can return with a
407 // normal result, an error result, or return having eaten an extra comma.
408 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
409 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
410 PerFunctionState &PFS);
411 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
413 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
414 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
415 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
416 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
417 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
418 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
420 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
421 unsigned OperandType);
422 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
423 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
424 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
425 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
426 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
427 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
428 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
429 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
430 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
431 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
432 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
433 CallInst::TailCallKind IsTail);
434 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
435 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
436 int ParseStore(Instruction *&I, PerFunctionState &PFS);
437 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
438 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
439 int ParseFence(Instruction *&I, PerFunctionState &PFS);
440 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
441 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
442 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
444 // Use-list order directives.
445 bool ParseUseListOrder(PerFunctionState *PFS = nullptr);
446 bool ParseUseListOrderBB();
447 bool ParseUseListOrderIndexes(SmallVectorImpl<unsigned> &Indexes);
448 bool sortUseListOrder(Value *V, ArrayRef<unsigned> Indexes, SMLoc Loc);
450 } // End llvm namespace