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/Attributes.h"
19 #include "llvm/Instructions.h"
20 #include "llvm/Module.h"
21 #include "llvm/Type.h"
22 #include "llvm/ADT/DenseMap.h"
23 #include "llvm/ADT/StringMap.h"
24 #include "llvm/Support/ValueHandle.h"
40 /// ValID - Represents a reference of a definition of some sort with no type.
41 /// There are several cases where we have to parse the value but where the
42 /// type can depend on later context. This may either be a numeric reference
43 /// or a symbolic (%var) reference. This is just a discriminated union.
46 t_LocalID, t_GlobalID, // ID in UIntVal.
47 t_LocalName, t_GlobalName, // Name in StrVal.
48 t_APSInt, t_APFloat, // Value in APSIntVal/APFloatVal.
49 t_Null, t_Undef, t_Zero, // No value.
50 t_EmptyArray, // No value: []
51 t_Constant, // Value in ConstantVal.
52 t_InlineAsm, // Value in StrVal/StrVal2/UIntVal.
53 t_MDNode, // Value in MDNodeVal.
54 t_MDString, // Value in MDStringVal.
55 t_ConstantStruct, // Value in ConstantStructElts.
56 t_PackedConstantStruct // Value in ConstantStructElts.
61 std::string StrVal, StrVal2;
64 Constant *ConstantVal;
66 MDString *MDStringVal;
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 // 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<TrackingVH<MDNode> > NumberedMetadata;
115 std::map<unsigned, std::pair<TrackingVH<MDNode>, 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 // References to blockaddress. The key is the function ValID, the value is
123 // a list of references to blocks in that function.
124 std::map<ValID, std::vector<std::pair<ValID, GlobalValue*> > >
125 ForwardRefBlockAddresses;
128 LLParser(MemoryBuffer *F, SourceMgr &SM, SMDiagnostic &Err, Module *m) :
129 Context(m->getContext()), Lex(F, SM, Err, m->getContext()),
133 LLVMContext &getContext() { return Context; }
137 bool Error(LocTy L, const Twine &Msg) const {
138 return Lex.Error(L, Msg);
140 bool TokError(const Twine &Msg) const {
141 return Error(Lex.getLoc(), Msg);
144 /// GetGlobalVal - Get a value with the specified name or ID, creating a
145 /// forward reference record if needed. This can return null if the value
146 /// exists but does not have the right type.
147 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
148 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
151 bool ParseToken(lltok::Kind T, const char *ErrMsg);
152 bool EatIfPresent(lltok::Kind T) {
153 if (Lex.getKind() != T) return false;
157 bool ParseOptionalToken(lltok::Kind T, bool &Present, LocTy *Loc = 0) {
158 if (Lex.getKind() != T) {
168 bool ParseStringConstant(std::string &Result);
169 bool ParseUInt32(unsigned &Val);
170 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
172 return ParseUInt32(Val);
175 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
176 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
177 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
178 bool ParseOptionalAttrs(AttrBuilder &Attrs, unsigned AttrKind);
179 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
180 bool ParseOptionalLinkage(unsigned &Linkage) {
181 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
183 bool ParseOptionalVisibility(unsigned &Visibility);
184 bool ParseOptionalCallingConv(CallingConv::ID &CC);
185 bool ParseOptionalAlignment(unsigned &Alignment);
186 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
187 AtomicOrdering &Ordering);
188 bool ParseOptionalStackAlignment(unsigned &Alignment);
189 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
190 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
191 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
193 if (ParseIndexList(Indices, AteExtraComma)) return true;
195 return TokError("expected index");
199 // Top-Level Entities
200 bool ParseTopLevelEntities();
201 bool ValidateEndOfModule();
202 bool ParseTargetDefinition();
204 bool ParseModuleAsm();
205 bool ParseUnnamedType();
206 bool ParseNamedType();
210 bool ParseGlobalType(bool &IsConstant);
211 bool ParseUnnamedGlobal();
212 bool ParseNamedGlobal();
213 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
214 bool HasLinkage, unsigned Visibility);
215 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility);
216 bool ParseStandaloneMetadata();
217 bool ParseNamedMetadata();
218 bool ParseMDString(MDString *&Result);
219 bool ParseMDNodeID(MDNode *&Result);
220 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
223 bool ParseType(Type *&Result, bool AllowVoid = false);
224 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
226 return ParseType(Result, AllowVoid);
228 bool ParseAnonStructType(Type *&Result, bool Packed);
229 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
230 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
231 std::pair<Type*, LocTy> &Entry,
234 bool ParseArrayVectorType(Type *&Result, bool isVector);
235 bool ParseFunctionType(Type *&Result);
237 // Function Semantic Analysis.
238 class PerFunctionState {
241 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
242 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
243 std::vector<Value*> NumberedVals;
245 /// FunctionNumber - If this is an unnamed function, this is the slot
246 /// number of it, otherwise it is -1.
249 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
252 Function &getFunction() const { return F; }
254 bool FinishFunction();
256 /// GetVal - Get a value with the specified name or ID, creating a
257 /// forward reference record if needed. This can return null if the value
258 /// exists but does not have the right type.
259 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
260 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
262 /// SetInstName - After an instruction is parsed and inserted into its
263 /// basic block, this installs its name.
264 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
267 /// GetBB - Get a basic block with the specified name or ID, creating a
268 /// forward reference record if needed. This can return null if the value
269 /// is not a BasicBlock.
270 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
271 BasicBlock *GetBB(unsigned ID, LocTy Loc);
273 /// DefineBB - Define the specified basic block, which is either named or
274 /// unnamed. If there is an error, this returns null otherwise it returns
275 /// the block being defined.
276 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
279 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
280 PerFunctionState *PFS);
282 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
283 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
284 return ParseValue(Ty, V, &PFS);
286 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
287 PerFunctionState &PFS) {
289 return ParseValue(Ty, V, &PFS);
292 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
293 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
294 return ParseTypeAndValue(V, &PFS);
296 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
298 return ParseTypeAndValue(V, PFS);
300 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
301 PerFunctionState &PFS);
302 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
304 return ParseTypeAndBasicBlock(BB, Loc, PFS);
312 ParamInfo(LocTy loc, Value *v, Attributes attrs)
313 : Loc(loc), V(v), Attrs(attrs) {}
315 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
316 PerFunctionState &PFS);
319 bool ParseValID(ValID &ID, PerFunctionState *PFS = NULL);
320 bool ParseGlobalValue(Type *Ty, Constant *&V);
321 bool ParseGlobalTypeAndValue(Constant *&V);
322 bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
323 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
324 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
325 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
326 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
334 ArgInfo(LocTy L, Type *ty, Attributes Attr, const std::string &N)
335 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
337 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
338 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
339 bool ParseFunctionBody(Function &Fn);
340 bool ParseBasicBlock(PerFunctionState &PFS);
342 // Instruction Parsing. Each instruction parsing routine can return with a
343 // normal result, an error result, or return having eaten an extra comma.
344 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
345 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
346 PerFunctionState &PFS);
347 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
349 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
350 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
351 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
352 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
353 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
354 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
356 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
357 unsigned OperandType);
358 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
359 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
360 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
361 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
362 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
363 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
364 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
365 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
366 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
367 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
368 bool ParseCall(Instruction *&I, PerFunctionState &PFS, bool isTail);
369 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
370 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
371 int ParseStore(Instruction *&I, PerFunctionState &PFS);
372 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
373 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
374 int ParseFence(Instruction *&I, PerFunctionState &PFS);
375 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
376 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
377 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
379 bool ResolveForwardRefBlockAddresses(Function *TheFn,
380 std::vector<std::pair<ValID, GlobalValue*> > &Refs,
381 PerFunctionState *PFS);
383 } // End llvm namespace