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_ASMPARSER_LLPARSER_H
15 #define LLVM_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"
41 /// ValID - Represents a reference of a definition of some sort with no type.
42 /// There are several cases where we have to parse the value but where the
43 /// type can depend on later context. This may either be a numeric reference
44 /// or a symbolic (%var) reference. This is just a discriminated union.
47 t_LocalID, t_GlobalID, // ID in UIntVal.
48 t_LocalName, t_GlobalName, // Name in StrVal.
49 t_APSInt, t_APFloat, // Value in APSIntVal/APFloatVal.
50 t_Null, t_Undef, t_Zero, // No value.
51 t_EmptyArray, // No value: []
52 t_Constant, // Value in ConstantVal.
53 t_InlineAsm, // Value in StrVal/StrVal2/UIntVal.
54 t_MDNode, // Value in MDNodeVal.
55 t_MDString, // Value in MDStringVal.
56 t_ConstantStruct, // Value in ConstantStructElts.
57 t_PackedConstantStruct // Value in ConstantStructElts.
62 std::string StrVal, StrVal2;
65 Constant *ConstantVal;
67 MDString *MDStringVal;
68 Constant **ConstantStructElts;
70 ValID() : Kind(t_LocalID), APFloatVal(0.0) {}
72 if (Kind == t_ConstantStruct || Kind == t_PackedConstantStruct)
73 delete [] ConstantStructElts;
76 bool operator<(const ValID &RHS) const {
77 if (Kind == t_LocalID || Kind == t_GlobalID)
78 return UIntVal < RHS.UIntVal;
79 assert((Kind == t_LocalName || Kind == t_GlobalName ||
80 Kind == t_ConstantStruct || Kind == t_PackedConstantStruct) &&
81 "Ordering not defined for this ValID kind yet");
82 return StrVal < RHS.StrVal;
88 typedef LLLexer::LocTy LocTy;
94 // Instruction metadata resolution. Each instruction can have a list of
95 // MDRef info associated with them.
97 // The simpler approach of just creating temporary MDNodes and then calling
98 // RAUW on them when the definition is processed doesn't work because some
99 // instruction metadata kinds, such as dbg, get stored in the IR in an
100 // "optimized" format which doesn't participate in the normal value use
101 // lists. This means that RAUW doesn't work, even on temporary MDNodes
102 // which otherwise support RAUW. Instead, we defer resolving MDNode
103 // references until the definitions have been processed.
106 unsigned MDKind, MDSlot;
108 DenseMap<Instruction*, std::vector<MDRef> > ForwardRefInstMetadata;
110 SmallVector<Instruction*, 64> InstsWithTBAATag;
112 // Type resolution handling data structures. The location is set when we
113 // have processed a use of the type but not a definition yet.
114 StringMap<std::pair<Type*, LocTy> > NamedTypes;
115 std::vector<std::pair<Type*, LocTy> > NumberedTypes;
117 std::vector<TrackingVH<MDNode> > NumberedMetadata;
118 std::map<unsigned, std::pair<TrackingVH<MDNode>, LocTy> > ForwardRefMDNodes;
120 // Global Value reference information.
121 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
122 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
123 std::vector<GlobalValue*> NumberedVals;
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::vector<std::pair<ValID, GlobalValue*> > >
128 ForwardRefBlockAddresses;
130 // Attribute builder reference information.
131 std::map<Value*, std::vector<unsigned> > ForwardRefAttrGroups;
132 std::map<unsigned, AttrBuilder> NumberedAttrBuilders;
135 LLParser(MemoryBuffer *F, SourceMgr &SM, SMDiagnostic &Err, Module *m) :
136 Context(m->getContext()), Lex(F, SM, Err, m->getContext()),
140 LLVMContext &getContext() { return Context; }
144 bool Error(LocTy L, const Twine &Msg) const {
145 return Lex.Error(L, Msg);
147 bool TokError(const Twine &Msg) const {
148 return Error(Lex.getLoc(), Msg);
151 /// GetGlobalVal - Get a value with the specified name or ID, creating a
152 /// forward reference record if needed. This can return null if the value
153 /// exists but does not have the right type.
154 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
155 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
158 bool ParseToken(lltok::Kind T, const char *ErrMsg);
159 bool EatIfPresent(lltok::Kind T) {
160 if (Lex.getKind() != T) return false;
165 FastMathFlags EatFastMathFlagsIfPresent() {
168 switch (Lex.getKind()) {
169 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
170 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
171 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
172 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
173 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
179 bool ParseOptionalToken(lltok::Kind T, bool &Present,
180 LocTy *Loc = nullptr) {
181 if (Lex.getKind() != T) {
191 bool ParseStringConstant(std::string &Result);
192 bool ParseUInt32(unsigned &Val);
193 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
195 return ParseUInt32(Val);
198 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
199 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
200 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
201 bool ParseOptionalParamAttrs(AttrBuilder &B);
202 bool ParseOptionalReturnAttrs(AttrBuilder &B);
203 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
204 bool ParseOptionalLinkage(unsigned &Linkage) {
205 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
207 bool ParseOptionalVisibility(unsigned &Visibility);
208 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
209 bool ParseOptionalCallingConv(CallingConv::ID &CC);
210 bool ParseOptionalAlignment(unsigned &Alignment);
211 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
212 AtomicOrdering &Ordering);
213 bool ParseOrdering(AtomicOrdering &Ordering);
214 bool ParseOptionalStackAlignment(unsigned &Alignment);
215 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
216 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
217 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
218 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
220 if (ParseIndexList(Indices, AteExtraComma)) return true;
222 return TokError("expected index");
226 // Top-Level Entities
227 bool ParseTopLevelEntities();
228 bool ValidateEndOfModule();
229 bool ParseTargetDefinition();
230 bool ParseModuleAsm();
231 bool ParseDepLibs(); // FIXME: Remove in 4.0.
232 bool ParseUnnamedType();
233 bool ParseNamedType();
237 bool ParseGlobalType(bool &IsConstant);
238 bool ParseUnnamedGlobal();
239 bool ParseNamedGlobal();
240 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
241 bool HasLinkage, unsigned Visibility,
242 unsigned DLLStorageClass);
243 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility,
244 unsigned DLLStorageClass);
245 bool ParseStandaloneMetadata();
246 bool ParseNamedMetadata();
247 bool ParseMDString(MDString *&Result);
248 bool ParseMDNodeID(MDNode *&Result);
249 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
250 bool ParseUnnamedAttrGrp();
251 bool ParseFnAttributeValuePairs(AttrBuilder &B,
252 std::vector<unsigned> &FwdRefAttrGrps,
253 bool inAttrGrp, LocTy &BuiltinLoc);
256 bool ParseType(Type *&Result, bool AllowVoid = false);
257 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
259 return ParseType(Result, AllowVoid);
261 bool ParseAnonStructType(Type *&Result, bool Packed);
262 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
263 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
264 std::pair<Type*, LocTy> &Entry,
267 bool ParseArrayVectorType(Type *&Result, bool isVector);
268 bool ParseFunctionType(Type *&Result);
270 // Function Semantic Analysis.
271 class PerFunctionState {
274 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
275 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
276 std::vector<Value*> NumberedVals;
278 /// FunctionNumber - If this is an unnamed function, this is the slot
279 /// number of it, otherwise it is -1.
282 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
285 Function &getFunction() const { return F; }
287 bool FinishFunction();
289 /// GetVal - Get a value with the specified name or ID, creating a
290 /// forward reference record if needed. This can return null if the value
291 /// exists but does not have the right type.
292 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
293 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
295 /// SetInstName - After an instruction is parsed and inserted into its
296 /// basic block, this installs its name.
297 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
300 /// GetBB - Get a basic block with the specified name or ID, creating a
301 /// forward reference record if needed. This can return null if the value
302 /// is not a BasicBlock.
303 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
304 BasicBlock *GetBB(unsigned ID, LocTy Loc);
306 /// DefineBB - Define the specified basic block, which is either named or
307 /// unnamed. If there is an error, this returns null otherwise it returns
308 /// the block being defined.
309 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
312 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
313 PerFunctionState *PFS);
315 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
316 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
317 return ParseValue(Ty, V, &PFS);
319 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
320 PerFunctionState &PFS) {
322 return ParseValue(Ty, V, &PFS);
325 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
326 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
327 return ParseTypeAndValue(V, &PFS);
329 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
331 return ParseTypeAndValue(V, PFS);
333 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
334 PerFunctionState &PFS);
335 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
337 return ParseTypeAndBasicBlock(BB, Loc, PFS);
345 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
346 : Loc(loc), V(v), Attrs(attrs) {}
348 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
349 PerFunctionState &PFS);
352 bool ParseValID(ValID &ID, PerFunctionState *PFS = nullptr);
353 bool ParseGlobalValue(Type *Ty, Constant *&V);
354 bool ParseGlobalTypeAndValue(Constant *&V);
355 bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
356 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
357 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
358 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
359 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
367 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
368 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
370 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
371 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
372 bool ParseFunctionBody(Function &Fn);
373 bool ParseBasicBlock(PerFunctionState &PFS);
375 enum TailCallType { TCT_None, TCT_Tail, TCT_MustTail };
377 // Instruction Parsing. Each instruction parsing routine can return with a
378 // normal result, an error result, or return having eaten an extra comma.
379 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
380 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
381 PerFunctionState &PFS);
382 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
384 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
385 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
386 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
387 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
388 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
389 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
391 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
392 unsigned OperandType);
393 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
394 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
395 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
396 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
397 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
398 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
399 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
400 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
401 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
402 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
403 bool ParseCall(Instruction *&I, PerFunctionState &PFS,
404 CallInst::TailCallKind IsTail);
405 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
406 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
407 int ParseStore(Instruction *&I, PerFunctionState &PFS);
408 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
409 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
410 int ParseFence(Instruction *&I, PerFunctionState &PFS);
411 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
412 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
413 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
415 bool ResolveForwardRefBlockAddresses(Function *TheFn,
416 std::vector<std::pair<ValID, GlobalValue*> > &Refs,
417 PerFunctionState *PFS);
419 } // End llvm namespace