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/Module.h"
19 #include "llvm/Type.h"
20 #include "llvm/ADT/DenseMap.h"
21 #include "llvm/Support/ValueHandle.h"
36 /// ValID - Represents a reference of a definition of some sort with no type.
37 /// There are several cases where we have to parse the value but where the
38 /// type can depend on later context. This may either be a numeric reference
39 /// or a symbolic (%var) reference. This is just a discriminated union.
42 t_LocalID, t_GlobalID, // ID in UIntVal.
43 t_LocalName, t_GlobalName, // Name in StrVal.
44 t_APSInt, t_APFloat, // Value in APSIntVal/APFloatVal.
45 t_Null, t_Undef, t_Zero, // No value.
46 t_EmptyArray, // No value: []
47 t_Constant, // Value in ConstantVal.
48 t_InlineAsm, // Value in StrVal/StrVal2/UIntVal.
49 t_MDNode, // Value in MDNodeVal.
50 t_MDString // Value in MDStringVal.
55 std::string StrVal, StrVal2;
58 Constant *ConstantVal;
60 MDString *MDStringVal;
61 ValID() : APFloatVal(0.0) {}
63 bool operator<(const ValID &RHS) const {
64 if (Kind == t_LocalID || Kind == t_GlobalID)
65 return UIntVal < RHS.UIntVal;
66 assert((Kind == t_LocalName || Kind == t_GlobalName) &&
67 "Ordering not defined for this ValID kind yet");
68 return StrVal < RHS.StrVal;
74 typedef LLLexer::LocTy LocTy;
80 // Instruction metadata resolution. Each instruction can have a list of
81 // MDRef info associated with them.
83 // The simpler approach of just creating temporary MDNodes and then calling
84 // RAUW on them when the definition is processed doesn't work because some
85 // instruction metadata kinds, such as dbg, get stored in the IR in an
86 // "optimized" format which doesn't participate in the normal value use
87 // lists. This means that RAUW doesn't work, even on temporary MDNodes
88 // which otherwise support RAUW. Instead, we defer resolving MDNode
89 // references until the definitions have been processed.
92 unsigned MDKind, MDSlot;
94 DenseMap<Instruction*, std::vector<MDRef> > ForwardRefInstMetadata;
96 // Type resolution handling data structures.
97 std::map<std::string, std::pair<PATypeHolder, LocTy> > ForwardRefTypes;
98 std::map<unsigned, std::pair<PATypeHolder, LocTy> > ForwardRefTypeIDs;
99 std::vector<PATypeHolder> NumberedTypes;
100 std::vector<TrackingVH<MDNode> > NumberedMetadata;
101 std::map<unsigned, std::pair<TrackingVH<MDNode>, LocTy> > ForwardRefMDNodes;
103 /// Loc - This is the location of the upref.
106 /// NestingLevel - The number of nesting levels that need to be popped
107 /// before this type is resolved.
108 unsigned NestingLevel;
110 /// LastContainedTy - This is the type at the current binding level for
111 /// the type. Every time we reduce the nesting level, this gets updated.
112 const Type *LastContainedTy;
114 /// UpRefTy - This is the actual opaque type that the upreference is
115 /// represented with.
118 UpRefRecord(LocTy L, unsigned NL, OpaqueType *URTy)
119 : Loc(L), NestingLevel(NL), LastContainedTy((Type*)URTy),
122 std::vector<UpRefRecord> UpRefs;
124 // Global Value reference information.
125 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
126 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
127 std::vector<GlobalValue*> NumberedVals;
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::vector<std::pair<ValID, GlobalValue*> > >
132 ForwardRefBlockAddresses;
136 LLParser(MemoryBuffer *F, SourceMgr &SM, SMDiagnostic &Err, Module *m) :
137 Context(m->getContext()), Lex(F, SM, Err, m->getContext()),
138 M(m), MallocF(NULL) {}
141 LLVMContext& getContext() { return Context; }
145 bool Error(LocTy L, const Twine &Msg) const {
146 return Lex.Error(L, Msg);
148 bool TokError(const Twine &Msg) const {
149 return Error(Lex.getLoc(), Msg);
152 /// GetGlobalVal - Get a value with the specified name or ID, creating a
153 /// forward reference record if needed. This can return null if the value
154 /// exists but does not have the right type.
155 GlobalValue *GetGlobalVal(const std::string &N, const Type *Ty, LocTy Loc);
156 GlobalValue *GetGlobalVal(unsigned ID, const Type *Ty, LocTy Loc);
159 bool ParseToken(lltok::Kind T, const char *ErrMsg);
160 bool EatIfPresent(lltok::Kind T) {
161 if (Lex.getKind() != T) return false;
165 bool ParseOptionalToken(lltok::Kind T, bool &Present, LocTy *Loc = 0) {
166 if (Lex.getKind() != T) {
176 bool ParseStringConstant(std::string &Result);
177 bool ParseUInt32(unsigned &Val);
178 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
180 return ParseUInt32(Val);
182 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
183 bool ParseOptionalAttrs(unsigned &Attrs, unsigned AttrKind);
184 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
185 bool ParseOptionalLinkage(unsigned &Linkage) {
186 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
188 bool ParseOptionalVisibility(unsigned &Visibility);
189 bool ParseOptionalCallingConv(CallingConv::ID &CC);
190 bool ParseOptionalAlignment(unsigned &Alignment);
191 bool ParseOptionalStackAlignment(unsigned &Alignment);
192 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
193 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
194 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
196 if (ParseIndexList(Indices, AteExtraComma)) return true;
198 return TokError("expected index");
202 // Top-Level Entities
203 bool ParseTopLevelEntities();
204 bool ValidateEndOfModule();
205 bool ParseTargetDefinition();
207 bool ParseModuleAsm();
208 bool ParseUnnamedType();
209 bool ParseNamedType();
213 bool ParseGlobalType(bool &IsConstant);
214 bool ParseUnnamedGlobal();
215 bool ParseNamedGlobal();
216 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
217 bool HasLinkage, unsigned Visibility);
218 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility);
219 bool ParseStandaloneMetadata();
220 bool ParseNamedMetadata();
221 bool ParseMDString(MDString *&Result);
222 bool ParseMDNodeID(MDNode *&Result);
223 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
226 bool ParseType(PATypeHolder &Result, bool AllowVoid = false);
227 bool ParseType(PATypeHolder &Result, LocTy &Loc, bool AllowVoid = false) {
229 return ParseType(Result, AllowVoid);
231 bool ParseTypeRec(PATypeHolder &H);
232 bool ParseStructType(PATypeHolder &H, bool Packed);
233 bool ParseArrayVectorType(PATypeHolder &H, bool isVector);
234 bool ParseFunctionType(PATypeHolder &Result);
235 PATypeHolder HandleUpRefs(const Type *Ty);
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, const Type *Ty, LocTy Loc);
260 Value *GetVal(unsigned ID, const 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(const Type *Ty, ValID &ID, Value *&V,
280 PerFunctionState *PFS);
282 bool ParseValue(const Type *Ty, Value *&V, PerFunctionState &PFS);
283 bool ParseValue(const Type *Ty, Value *&V, LocTy &Loc,
284 PerFunctionState &PFS) {
286 return ParseValue(Ty, V, PFS);
289 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS);
290 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
292 return ParseTypeAndValue(V, PFS);
294 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
295 PerFunctionState &PFS);
296 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
298 return ParseTypeAndBasicBlock(BB, Loc, PFS);
306 ParamInfo(LocTy loc, Value *v, unsigned attrs)
307 : Loc(loc), V(v), Attrs(attrs) {}
309 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
310 PerFunctionState &PFS);
313 bool ParseValID(ValID &ID, PerFunctionState *PFS = NULL);
314 bool ParseGlobalValue(const Type *Ty, Constant *&V);
315 bool ParseGlobalTypeAndValue(Constant *&V);
316 bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
317 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
318 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
319 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
320 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
328 ArgInfo(LocTy L, PATypeHolder Ty, unsigned Attr, const std::string &N)
329 : Loc(L), Type(Ty), Attrs(Attr), Name(N) {}
331 bool ParseArgumentList(std::vector<ArgInfo> &ArgList,
332 bool &isVarArg, bool inType);
333 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
334 bool ParseFunctionBody(Function &Fn);
335 bool ParseBasicBlock(PerFunctionState &PFS);
337 // Instruction Parsing. Each instruction parsing routine can return with a
338 // normal result, an error result, or return having eaten an extra comma.
339 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
340 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
341 PerFunctionState &PFS);
342 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
344 int ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
345 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
346 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
347 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
348 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
350 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
351 unsigned OperandType);
352 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
353 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
354 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
355 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
356 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
357 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
358 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
359 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
360 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
361 bool ParseCall(Instruction *&I, PerFunctionState &PFS, bool isTail);
362 int ParseAlloc(Instruction *&I, PerFunctionState &PFS,
363 BasicBlock *BB = 0, bool isAlloca = true);
364 bool ParseFree(Instruction *&I, PerFunctionState &PFS, BasicBlock *BB);
365 int ParseLoad(Instruction *&I, PerFunctionState &PFS, bool isVolatile);
366 int ParseStore(Instruction *&I, PerFunctionState &PFS, bool isVolatile);
367 bool ParseGetResult(Instruction *&I, PerFunctionState &PFS);
368 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
369 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
370 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
372 bool ResolveForwardRefBlockAddresses(Function *TheFn,
373 std::vector<std::pair<ValID, GlobalValue*> > &Refs,
374 PerFunctionState *PFS);
376 } // End llvm namespace