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, LocTy *Loc = 0) {
180 if (Lex.getKind() != T) {
190 bool ParseStringConstant(std::string &Result);
191 bool ParseUInt32(unsigned &Val);
192 bool ParseUInt32(unsigned &Val, LocTy &Loc) {
194 return ParseUInt32(Val);
197 bool ParseTLSModel(GlobalVariable::ThreadLocalMode &TLM);
198 bool ParseOptionalThreadLocal(GlobalVariable::ThreadLocalMode &TLM);
199 bool ParseOptionalAddrSpace(unsigned &AddrSpace);
200 bool ParseOptionalParamAttrs(AttrBuilder &B);
201 bool ParseOptionalReturnAttrs(AttrBuilder &B);
202 bool ParseOptionalLinkage(unsigned &Linkage, bool &HasLinkage);
203 bool ParseOptionalLinkage(unsigned &Linkage) {
204 bool HasLinkage; return ParseOptionalLinkage(Linkage, HasLinkage);
206 bool ParseOptionalVisibility(unsigned &Visibility);
207 bool ParseOptionalDLLStorageClass(unsigned &DLLStorageClass);
208 bool ParseOptionalCallingConv(CallingConv::ID &CC);
209 bool ParseOptionalAlignment(unsigned &Alignment);
210 bool ParseScopeAndOrdering(bool isAtomic, SynchronizationScope &Scope,
211 AtomicOrdering &Ordering);
212 bool ParseOrdering(AtomicOrdering &Ordering);
213 bool ParseOptionalStackAlignment(unsigned &Alignment);
214 bool ParseOptionalCommaAlign(unsigned &Alignment, bool &AteExtraComma);
215 bool ParseOptionalCommaInAlloca(bool &IsInAlloca);
216 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices,bool &AteExtraComma);
217 bool ParseIndexList(SmallVectorImpl<unsigned> &Indices) {
219 if (ParseIndexList(Indices, AteExtraComma)) return true;
221 return TokError("expected index");
225 // Top-Level Entities
226 bool ParseTopLevelEntities();
227 bool ValidateEndOfModule();
228 bool ParseTargetDefinition();
229 bool ParseModuleAsm();
230 bool ParseDepLibs(); // FIXME: Remove in 4.0.
231 bool ParseUnnamedType();
232 bool ParseNamedType();
236 bool ParseGlobalType(bool &IsConstant);
237 bool ParseUnnamedGlobal();
238 bool ParseNamedGlobal();
239 bool ParseGlobal(const std::string &Name, LocTy Loc, unsigned Linkage,
240 bool HasLinkage, unsigned Visibility,
241 unsigned DLLStorageClass);
242 bool ParseAlias(const std::string &Name, LocTy Loc, unsigned Visibility,
243 unsigned DLLStorageClass);
244 bool ParseStandaloneMetadata();
245 bool ParseNamedMetadata();
246 bool ParseMDString(MDString *&Result);
247 bool ParseMDNodeID(MDNode *&Result);
248 bool ParseMDNodeID(MDNode *&Result, unsigned &SlotNo);
249 bool ParseUnnamedAttrGrp();
250 bool ParseFnAttributeValuePairs(AttrBuilder &B,
251 std::vector<unsigned> &FwdRefAttrGrps,
252 bool inAttrGrp, LocTy &BuiltinLoc);
255 bool ParseType(Type *&Result, bool AllowVoid = false);
256 bool ParseType(Type *&Result, LocTy &Loc, bool AllowVoid = false) {
258 return ParseType(Result, AllowVoid);
260 bool ParseAnonStructType(Type *&Result, bool Packed);
261 bool ParseStructBody(SmallVectorImpl<Type*> &Body);
262 bool ParseStructDefinition(SMLoc TypeLoc, StringRef Name,
263 std::pair<Type*, LocTy> &Entry,
266 bool ParseArrayVectorType(Type *&Result, bool isVector);
267 bool ParseFunctionType(Type *&Result);
269 // Function Semantic Analysis.
270 class PerFunctionState {
273 std::map<std::string, std::pair<Value*, LocTy> > ForwardRefVals;
274 std::map<unsigned, std::pair<Value*, LocTy> > ForwardRefValIDs;
275 std::vector<Value*> NumberedVals;
277 /// FunctionNumber - If this is an unnamed function, this is the slot
278 /// number of it, otherwise it is -1.
281 PerFunctionState(LLParser &p, Function &f, int FunctionNumber);
284 Function &getFunction() const { return F; }
286 bool FinishFunction();
288 /// GetVal - Get a value with the specified name or ID, creating a
289 /// forward reference record if needed. This can return null if the value
290 /// exists but does not have the right type.
291 Value *GetVal(const std::string &Name, Type *Ty, LocTy Loc);
292 Value *GetVal(unsigned ID, Type *Ty, LocTy Loc);
294 /// SetInstName - After an instruction is parsed and inserted into its
295 /// basic block, this installs its name.
296 bool SetInstName(int NameID, const std::string &NameStr, LocTy NameLoc,
299 /// GetBB - Get a basic block with the specified name or ID, creating a
300 /// forward reference record if needed. This can return null if the value
301 /// is not a BasicBlock.
302 BasicBlock *GetBB(const std::string &Name, LocTy Loc);
303 BasicBlock *GetBB(unsigned ID, LocTy Loc);
305 /// DefineBB - Define the specified basic block, which is either named or
306 /// unnamed. If there is an error, this returns null otherwise it returns
307 /// the block being defined.
308 BasicBlock *DefineBB(const std::string &Name, LocTy Loc);
311 bool ConvertValIDToValue(Type *Ty, ValID &ID, Value *&V,
312 PerFunctionState *PFS);
314 bool ParseValue(Type *Ty, Value *&V, PerFunctionState *PFS);
315 bool ParseValue(Type *Ty, Value *&V, PerFunctionState &PFS) {
316 return ParseValue(Ty, V, &PFS);
318 bool ParseValue(Type *Ty, Value *&V, LocTy &Loc,
319 PerFunctionState &PFS) {
321 return ParseValue(Ty, V, &PFS);
324 bool ParseTypeAndValue(Value *&V, PerFunctionState *PFS);
325 bool ParseTypeAndValue(Value *&V, PerFunctionState &PFS) {
326 return ParseTypeAndValue(V, &PFS);
328 bool ParseTypeAndValue(Value *&V, LocTy &Loc, PerFunctionState &PFS) {
330 return ParseTypeAndValue(V, PFS);
332 bool ParseTypeAndBasicBlock(BasicBlock *&BB, LocTy &Loc,
333 PerFunctionState &PFS);
334 bool ParseTypeAndBasicBlock(BasicBlock *&BB, PerFunctionState &PFS) {
336 return ParseTypeAndBasicBlock(BB, Loc, PFS);
344 ParamInfo(LocTy loc, Value *v, AttributeSet attrs)
345 : Loc(loc), V(v), Attrs(attrs) {}
347 bool ParseParameterList(SmallVectorImpl<ParamInfo> &ArgList,
348 PerFunctionState &PFS);
351 bool ParseValID(ValID &ID, PerFunctionState *PFS = NULL);
352 bool ParseGlobalValue(Type *Ty, Constant *&V);
353 bool ParseGlobalTypeAndValue(Constant *&V);
354 bool ParseGlobalValueVector(SmallVectorImpl<Constant*> &Elts);
355 bool ParseMetadataListValue(ValID &ID, PerFunctionState *PFS);
356 bool ParseMetadataValue(ValID &ID, PerFunctionState *PFS);
357 bool ParseMDNodeVector(SmallVectorImpl<Value*> &, PerFunctionState *PFS);
358 bool ParseInstructionMetadata(Instruction *Inst, PerFunctionState *PFS);
366 ArgInfo(LocTy L, Type *ty, AttributeSet Attr, const std::string &N)
367 : Loc(L), Ty(ty), Attrs(Attr), Name(N) {}
369 bool ParseArgumentList(SmallVectorImpl<ArgInfo> &ArgList, bool &isVarArg);
370 bool ParseFunctionHeader(Function *&Fn, bool isDefine);
371 bool ParseFunctionBody(Function &Fn);
372 bool ParseBasicBlock(PerFunctionState &PFS);
374 // Instruction Parsing. Each instruction parsing routine can return with a
375 // normal result, an error result, or return having eaten an extra comma.
376 enum InstResult { InstNormal = 0, InstError = 1, InstExtraComma = 2 };
377 int ParseInstruction(Instruction *&Inst, BasicBlock *BB,
378 PerFunctionState &PFS);
379 bool ParseCmpPredicate(unsigned &Pred, unsigned Opc);
381 bool ParseRet(Instruction *&Inst, BasicBlock *BB, PerFunctionState &PFS);
382 bool ParseBr(Instruction *&Inst, PerFunctionState &PFS);
383 bool ParseSwitch(Instruction *&Inst, PerFunctionState &PFS);
384 bool ParseIndirectBr(Instruction *&Inst, PerFunctionState &PFS);
385 bool ParseInvoke(Instruction *&Inst, PerFunctionState &PFS);
386 bool ParseResume(Instruction *&Inst, PerFunctionState &PFS);
388 bool ParseArithmetic(Instruction *&I, PerFunctionState &PFS, unsigned Opc,
389 unsigned OperandType);
390 bool ParseLogical(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
391 bool ParseCompare(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
392 bool ParseCast(Instruction *&I, PerFunctionState &PFS, unsigned Opc);
393 bool ParseSelect(Instruction *&I, PerFunctionState &PFS);
394 bool ParseVA_Arg(Instruction *&I, PerFunctionState &PFS);
395 bool ParseExtractElement(Instruction *&I, PerFunctionState &PFS);
396 bool ParseInsertElement(Instruction *&I, PerFunctionState &PFS);
397 bool ParseShuffleVector(Instruction *&I, PerFunctionState &PFS);
398 int ParsePHI(Instruction *&I, PerFunctionState &PFS);
399 bool ParseLandingPad(Instruction *&I, PerFunctionState &PFS);
400 bool ParseCall(Instruction *&I, PerFunctionState &PFS, bool isTail);
401 int ParseAlloc(Instruction *&I, PerFunctionState &PFS);
402 int ParseLoad(Instruction *&I, PerFunctionState &PFS);
403 int ParseStore(Instruction *&I, PerFunctionState &PFS);
404 int ParseCmpXchg(Instruction *&I, PerFunctionState &PFS);
405 int ParseAtomicRMW(Instruction *&I, PerFunctionState &PFS);
406 int ParseFence(Instruction *&I, PerFunctionState &PFS);
407 int ParseGetElementPtr(Instruction *&I, PerFunctionState &PFS);
408 int ParseExtractValue(Instruction *&I, PerFunctionState &PFS);
409 int ParseInsertValue(Instruction *&I, PerFunctionState &PFS);
411 bool ResolveForwardRefBlockAddresses(Function *TheFn,
412 std::vector<std::pair<ValID, GlobalValue*> > &Refs,
413 PerFunctionState *PFS);
415 } // End llvm namespace