1 //===- IRSymtab.h - data definitions for IR symbol tables -------*- 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 contains data definitions and a reader and builder for a symbol
11 // table for LLVM IR. Its purpose is to allow linkers and other consumers of
12 // bitcode files to efficiently read the symbol table for symbol resolution
13 // purposes without needing to construct a module in memory.
15 // As with most object files the symbol table has two parts: the symbol table
16 // itself and a string table which is referenced by the symbol table.
18 // A symbol table corresponds to a single bitcode file, which may consist of
19 // multiple modules, so symbol tables may likewise contain symbols for multiple
22 //===----------------------------------------------------------------------===//
24 #ifndef LLVM_OBJECT_IRSYMTAB_H
25 #define LLVM_OBJECT_IRSYMTAB_H
27 #include "llvm/ADT/ArrayRef.h"
28 #include "llvm/ADT/StringRef.h"
29 #include "llvm/ADT/iterator_range.h"
30 #include "llvm/IR/GlobalValue.h"
31 #include "llvm/Object/SymbolicFile.h"
32 #include "llvm/Support/Endian.h"
33 #include "llvm/Support/Error.h"
40 struct BitcodeFileContents;
41 class StringTableBuilder;
47 // The data structures in this namespace define the low-level serialization
48 // format. Clients that just want to read a symbol table should use the
49 // irsymtab::Reader class.
51 using Word = support::ulittle32_t;
53 /// A reference to a string in the string table.
57 StringRef get(StringRef Strtab) const {
58 return {Strtab.data() + Offset, Size};
62 /// A reference to a range of objects in the symbol table.
63 template <typename T> struct Range {
66 ArrayRef<T> get(StringRef Symtab) const {
67 return {reinterpret_cast<const T *>(Symtab.data() + Offset), Size};
71 /// Describes the range of a particular module's symbols within the symbol
76 /// The index of the first Uncommon for this Module.
80 /// This is equivalent to an IR comdat.
85 /// Contains the information needed by linkers for symbol resolution, as well as
86 /// by the LTO implementation itself.
88 /// The mangled symbol name.
91 /// The unmangled symbol name, or the empty string if this is not an IR
95 /// The index into Header::Comdats, or -1 if not a comdat member.
100 FB_visibility, // 2 bits
101 FB_has_uncommon = FB_visibility + 2,
116 /// This data structure contains rarely used symbol fields and is optionally
117 /// referenced by a Symbol.
119 Word CommonSize, CommonAlign;
121 /// COFF-specific: the name of the symbol that a weak external resolves to
123 Str COFFWeakExternFallbackName;
125 /// Specified section name, if any.
130 /// Version number of the symtab format. This number should be incremented
131 /// when the format changes, but it does not need to be incremented if a
132 /// change to LLVM would cause it to create a different symbol table.
134 enum { kCurrentVersion = 1 };
136 /// The producer's version string (LLVM_VERSION_STRING " " LLVM_REVISION).
137 /// Consumers should rebuild the symbol table from IR if the producer's
138 /// version does not match the consumer's version due to potential differences
139 /// in symbol table format, symbol enumeration order and so on.
142 Range<Module> Modules;
143 Range<Comdat> Comdats;
144 Range<Symbol> Symbols;
145 Range<Uncommon> Uncommons;
147 Str TargetTriple, SourceFileName;
149 /// COFF-specific: linker directives.
153 } // end namespace storage
155 /// Fills in Symtab and StrtabBuilder with a valid symbol and string table for
157 Error build(ArrayRef<Module *> Mods, SmallVector<char, 0> &Symtab,
158 StringTableBuilder &StrtabBuilder, BumpPtrAllocator &Alloc);
160 /// This represents a symbol that has been read from a storage::Symbol and
161 /// possibly a storage::Uncommon.
163 // Copied from storage::Symbol.
164 StringRef Name, IRName;
168 // Copied from storage::Uncommon.
169 uint32_t CommonSize, CommonAlign;
170 StringRef COFFWeakExternFallbackName;
171 StringRef SectionName;
173 /// Returns the mangled symbol name.
174 StringRef getName() const { return Name; }
176 /// Returns the unmangled symbol name, or the empty string if this is not an
178 StringRef getIRName() const { return IRName; }
180 /// Returns the index into the comdat table (see Reader::getComdatTable()), or
181 /// -1 if not a comdat member.
182 int getComdatIndex() const { return ComdatIndex; }
184 using S = storage::Symbol;
186 GlobalValue::VisibilityTypes getVisibility() const {
187 return GlobalValue::VisibilityTypes((Flags >> S::FB_visibility) & 3);
190 bool isUndefined() const { return (Flags >> S::FB_undefined) & 1; }
191 bool isWeak() const { return (Flags >> S::FB_weak) & 1; }
192 bool isCommon() const { return (Flags >> S::FB_common) & 1; }
193 bool isIndirect() const { return (Flags >> S::FB_indirect) & 1; }
194 bool isUsed() const { return (Flags >> S::FB_used) & 1; }
195 bool isTLS() const { return (Flags >> S::FB_tls) & 1; }
197 bool canBeOmittedFromSymbolTable() const {
198 return (Flags >> S::FB_may_omit) & 1;
201 bool isGlobal() const { return (Flags >> S::FB_global) & 1; }
202 bool isFormatSpecific() const { return (Flags >> S::FB_format_specific) & 1; }
203 bool isUnnamedAddr() const { return (Flags >> S::FB_unnamed_addr) & 1; }
204 bool isExecutable() const { return (Flags >> S::FB_executable) & 1; }
206 uint64_t getCommonSize() const {
211 uint32_t getCommonAlignment() const {
216 /// COFF-specific: for weak externals, returns the name of the symbol that is
217 /// used as a fallback if the weak external remains undefined.
218 StringRef getCOFFWeakExternalFallback() const {
219 assert(isWeak() && isIndirect());
220 return COFFWeakExternFallbackName;
223 StringRef getSectionName() const { return SectionName; }
226 /// This class can be used to read a Symtab and Strtab produced by
229 StringRef Symtab, Strtab;
231 ArrayRef<storage::Module> Modules;
232 ArrayRef<storage::Comdat> Comdats;
233 ArrayRef<storage::Symbol> Symbols;
234 ArrayRef<storage::Uncommon> Uncommons;
236 StringRef str(storage::Str S) const { return S.get(Strtab); }
238 template <typename T> ArrayRef<T> range(storage::Range<T> R) const {
239 return R.get(Symtab);
242 const storage::Header &header() const {
243 return *reinterpret_cast<const storage::Header *>(Symtab.data());
250 Reader(StringRef Symtab, StringRef Strtab) : Symtab(Symtab), Strtab(Strtab) {
251 Modules = range(header().Modules);
252 Comdats = range(header().Comdats);
253 Symbols = range(header().Symbols);
254 Uncommons = range(header().Uncommons);
257 using symbol_range = iterator_range<object::content_iterator<SymbolRef>>;
259 /// Returns the symbol table for the entire bitcode file.
260 /// The symbols enumerated by this method are ephemeral, but they can be
261 /// copied into an irsymtab::Symbol object.
262 symbol_range symbols() const;
264 size_t getNumModules() const { return Modules.size(); }
266 /// Returns a slice of the symbol table for the I'th module in the file.
267 /// The symbols enumerated by this method are ephemeral, but they can be
268 /// copied into an irsymtab::Symbol object.
269 symbol_range module_symbols(unsigned I) const;
271 StringRef getTargetTriple() const { return str(header().TargetTriple); }
273 /// Returns the source file path specified at compile time.
274 StringRef getSourceFileName() const { return str(header().SourceFileName); }
276 /// Returns a table with all the comdats used by this file.
277 std::vector<StringRef> getComdatTable() const {
278 std::vector<StringRef> ComdatTable;
279 ComdatTable.reserve(Comdats.size());
280 for (auto C : Comdats)
281 ComdatTable.push_back(str(C.Name));
285 /// COFF-specific: returns linker options specified in the input file.
286 StringRef getCOFFLinkerOpts() const { return str(header().COFFLinkerOpts); }
289 /// Ephemeral symbols produced by Reader::symbols() and
290 /// Reader::module_symbols().
291 class Reader::SymbolRef : public Symbol {
292 const storage::Symbol *SymI, *SymE;
293 const storage::Uncommon *UncI;
300 Name = R->str(SymI->Name);
301 IRName = R->str(SymI->IRName);
302 ComdatIndex = SymI->ComdatIndex;
305 if (Flags & (1 << storage::Symbol::FB_has_uncommon)) {
306 CommonSize = UncI->CommonSize;
307 CommonAlign = UncI->CommonAlign;
308 COFFWeakExternFallbackName = R->str(UncI->COFFWeakExternFallbackName);
309 SectionName = R->str(UncI->SectionName);
311 // Reset this field so it can be queried unconditionally for all symbols.
316 SymbolRef(const storage::Symbol *SymI, const storage::Symbol *SymE,
317 const storage::Uncommon *UncI, const Reader *R)
318 : SymI(SymI), SymE(SymE), UncI(UncI), R(R) {
324 if (Flags & (1 << storage::Symbol::FB_has_uncommon))
329 bool operator==(const SymbolRef &Other) const { return SymI == Other.SymI; }
332 inline Reader::symbol_range Reader::symbols() const {
333 return {SymbolRef(Symbols.begin(), Symbols.end(), Uncommons.begin(), this),
334 SymbolRef(Symbols.end(), Symbols.end(), nullptr, this)};
337 inline Reader::symbol_range Reader::module_symbols(unsigned I) const {
338 const storage::Module &M = Modules[I];
339 const storage::Symbol *MBegin = Symbols.begin() + M.Begin,
340 *MEnd = Symbols.begin() + M.End;
341 return {SymbolRef(MBegin, MEnd, Uncommons.begin() + M.UncBegin, this),
342 SymbolRef(MEnd, MEnd, nullptr, this)};
345 /// The contents of the irsymtab in a bitcode file. Any underlying data for the
346 /// irsymtab are owned by Symtab and Strtab.
347 struct FileContents {
348 SmallVector<char, 0> Symtab, Strtab;
352 /// Reads the contents of a bitcode file, creating its irsymtab if necessary.
353 Expected<FileContents> readBitcode(const BitcodeFileContents &BFC);
355 } // end namespace irsymtab
356 } // end namespace llvm
358 #endif // LLVM_OBJECT_IRSYMTAB_H