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/Support/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 // Type resolution handling data structures. The location is set when we
111 // have processed a use of the type but not a definition yet.
112 StringMap<std::pair<Type*, LocTy> > NamedTypes;
113 std::vector<std::pair<Type*, LocTy> > NumberedTypes;
115 std::vector<TrackingVH<MDNode> > NumberedMetadata;
116 std::map<unsigned, std::pair<TrackingVH<MDNode>, LocTy> > ForwardRefMDNodes;
118 // Global Value reference information.
119 std::map<std::string, std::pair<GlobalValue*, LocTy> > ForwardRefVals;
120 std::map<unsigned, std::pair<GlobalValue*, LocTy> > ForwardRefValIDs;
121 std::vector<GlobalValue*> NumberedVals;
123 // References to blockaddress. The key is the function ValID, the value is
124 // a list of references to blocks in that function.
125 std::map<ValID, std::vector<std::pair<ValID, GlobalValue*> > >
126 ForwardRefBlockAddresses;
128 // Attribute builder reference information.
129 std::map<unsigned, AttrBuilder> ForwardRefAttrBuilder;
132 LLParser(MemoryBuffer *F, SourceMgr &SM, SMDiagnostic &Err, Module *m) :
133 Context(m->getContext()), Lex(F, SM, Err, m->getContext()),
137 LLVMContext &getContext() { return Context; }
141 bool Error(LocTy L, const Twine &Msg) const {
142 return Lex.Error(L, Msg);
144 bool TokError(const Twine &Msg) const {
145 return Error(Lex.getLoc(), Msg);
148 /// GetGlobalVal - Get a value with the specified name or ID, creating a
149 /// forward reference record if needed. This can return null if the value
150 /// exists but does not have the right type.
151 GlobalValue *GetGlobalVal(const std::string &N, Type *Ty, LocTy Loc);
152 GlobalValue *GetGlobalVal(unsigned ID, Type *Ty, LocTy Loc);
155 bool ParseToken(lltok::Kind T, const char *ErrMsg);
156 bool EatIfPresent(lltok::Kind T) {
157 if (Lex.getKind() != T) return false;
162 FastMathFlags EatFastMathFlagsIfPresent() {
165 switch (Lex.getKind()) {
166 case lltok::kw_fast: FMF.setUnsafeAlgebra(); Lex.Lex(); continue;
167 case lltok::kw_nnan: FMF.setNoNaNs(); Lex.Lex(); continue;
168 case lltok::kw_ninf: FMF.setNoInfs(); Lex.Lex(); continue;
169 case lltok::kw_nsz: FMF.setNoSignedZeros(); Lex.Lex(); continue;
170 case lltok::kw_arcp: FMF.setAllowReciprocal(); Lex.Lex(); continue;
176 bool ParseOptionalToken(lltok::Kind T, bool &Present, LocTy *Loc = 0) {
177 if (Lex.getKind() != T) {
187 bool ParseStringConstant(std::string &Result);
188 bool ParseUInt32(unsigned &Val);
189 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
191 return ParseUInt32(Val);
194 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
195 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
196 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
197 bool ParseOptionalFuncAttrs(AttrBuilder &B);
198 bool ParseOptionalParamAttrs(AttrBuilder &B);
199 bool ParseOptionalReturnAttrs(AttrBuilder &B);
200 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
201 bool ParseOptionalLinkage(unsigned &Linkage) {
202 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
204 bool ParseOptionalVisibility(unsigned &Visibility);
205 bool ParseOptionalCallingConv(CallingConv::ID &CC);
206 bool ParseOptionalAlignment(unsigned &Alignment);
207 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
208 AtomicOrdering &Ordering);
209 bool ParseOptionalStackAlignment(unsigned &Alignment);
210 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
211 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
212 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
214 if (ParseIndexList(Indices, AteExtraComma)) return true;
216 return TokError("expected index");
220 // Top-Level Entities
221 bool ParseTopLevelEntities();
222 bool ValidateEndOfModule();
223 bool ParseTargetDefinition();
224 bool ParseModuleAsm();
225 bool ParseDepLibs(); // FIXME: Remove in 4.0.
226 bool ParseUnnamedType();
227 bool ParseNamedType();
231 bool ParseGlobalType(bool &IsConstant);
232 bool ParseUnnamedGlobal();
233 bool ParseNamedGlobal();
234 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
235 bool HasLinkage, unsigned Visibility);
236 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility);
237 bool ParseStandaloneMetadata();
238 bool ParseNamedMetadata();
239 bool ParseMDString(MDString *&Result);
240 bool ParseMDNodeID(MDNode *&Result);
241 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
242 bool ParseUnnamedAttrGrp();
243 bool ParseAttributeValuePairs(AttrBuilder &B);
246 bool ParseType(Type *&Result, bool AllowVoid = false);
247 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
249 return ParseType(Result, AllowVoid);
251 bool ParseAnonStructType(Type *&Result, bool Packed);
252 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
253 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
254 std::pair<Type*, LocTy> &Entry,
257 bool ParseArrayVectorType(Type *&Result, bool isVector);
258 bool ParseFunctionType(Type *&Result);
260 // Function Semantic Analysis.
261 class PerFunctionState {
264 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
265 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
266 std::vector<Value*> NumberedVals;
268 /// FunctionNumber - If this is an unnamed function, this is the slot
269 /// number of it, otherwise it is -1.
272 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
275 Function &getFunction() const { return F; }
277 bool FinishFunction();
279 /// GetVal - Get a value with the specified name or ID, creating a
280 /// forward reference record if needed. This can return null if the value
281 /// exists but does not have the right type.
282 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
283 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
285 /// SetInstName - After an instruction is parsed and inserted into its
286 /// basic block, this installs its name.
287 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
290 /// GetBB - Get a basic block with the specified name or ID, creating a
291 /// forward reference record if needed. This can return null if the value
292 /// is not a BasicBlock.
293 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
294 BasicBlock *GetBB(unsigned ID, LocTy Loc);
296 /// DefineBB - Define the specified basic block, which is either named or
297 /// unnamed. If there is an error, this returns null otherwise it returns
298 /// the block being defined.
299 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
302 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
303 PerFunctionState *PFS);
305 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
306 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
307 return ParseValue(Ty, V, &PFS);
309 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
310 PerFunctionState &PFS) {
312 return ParseValue(Ty, V, &PFS);
315 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
316 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
317 return ParseTypeAndValue(V, &PFS);
319 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
321 return ParseTypeAndValue(V, PFS);
323 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
324 PerFunctionState &PFS);
325 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
327 return ParseTypeAndBasicBlock(BB, Loc, PFS);
335 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
336 : Loc(loc), V(v), Attrs(attrs) {}
338 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
339 PerFunctionState &PFS);
342 bool ParseValID(ValID &ID, PerFunctionState *PFS = NULL);
343 bool ParseGlobalValue(Type *Ty, Constant *&V);
344 bool ParseGlobalTypeAndValue(Constant *&V);
345 bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
346 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
347 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
348 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
349 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
357 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
358 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
360 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
361 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
362 bool ParseFunctionBody(Function &Fn);
363 bool ParseBasicBlock(PerFunctionState &PFS);
365 // Instruction Parsing. Each instruction parsing routine can return with a
366 // normal result, an error result, or return having eaten an extra comma.
367 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
368 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
369 PerFunctionState &PFS);
370 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
372 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
373 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
374 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
375 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
376 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
377 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
379 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
380 unsigned OperandType);
381 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
382 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
383 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
384 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
385 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
386 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
387 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
388 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
389 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
390 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
391 bool ParseCall(Instruction *&I, PerFunctionState &PFS, bool isTail);
392 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
393 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
394 int ParseStore(Instruction *&I, PerFunctionState &PFS);
395 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
396 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
397 int ParseFence(Instruction *&I, PerFunctionState &PFS);
398 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
399 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
400 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
402 bool ResolveForwardRefBlockAddresses(Function *TheFn,
403 std::vector<std::pair<ValID, GlobalValue*> > &Refs,
404 PerFunctionState *PFS);
406 } // End llvm namespace