1 //==-- llvm/CodeGen/DwarfAccelTable.h - Dwarf Accelerator 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 support for writing dwarf accelerator tables.
12 //===----------------------------------------------------------------------===//
14 #ifndef CODEGEN_ASMPRINTER_DWARFACCELTABLE_H__
15 #define CODEGEN_ASMPRINTER_DWARFACCELTABLE_H__
18 #include "llvm/ADT/ArrayRef.h"
19 #include "llvm/ADT/StringMap.h"
20 #include "llvm/MC/MCSymbol.h"
21 #include "llvm/Support/DataTypes.h"
22 #include "llvm/Support/Debug.h"
23 #include "llvm/Support/Dwarf.h"
24 #include "llvm/Support/ErrorHandling.h"
25 #include "llvm/Support/Format.h"
26 #include "llvm/Support/FormattedStream.h"
30 // The dwarf accelerator tables are an indirect hash table optimized
31 // for null lookup rather than access to known data. They are output into
32 // an on-disk format that looks like this:
46 // where the header contains a magic number, version, type of hash function,
47 // the number of buckets, total number of hashes, and room for a special
48 // struct of data and the length of that struct.
50 // The buckets contain an index (e.g. 6) into the hashes array. The hashes
51 // section contains all of the 32-bit hash values in contiguous memory, and
52 // the offsets contain the offset into the data area for the particular
55 // For a lookup example, we could hash a function name and take it modulo the
56 // number of buckets giving us our bucket. From there we take the bucket value
57 // as an index into the hashes table and look at each successive hash as long
58 // as the hash value is still the same modulo result (bucket value) as earlier.
59 // If we have a match we look at that same entry in the offsets table and
60 // grab the offset in the data for our final match.
68 class DwarfAccelTable {
70 static uint32_t HashDJB(StringRef Str) {
72 for (unsigned i = 0, e = Str.size(); i != e; ++i)
73 h = ((h << 5) + h) + Str[i];
77 // Helper function to compute the number of buckets needed based on
78 // the number of unique hashes.
79 void ComputeBucketCount(void);
82 uint32_t magic; // 'HASH' magic value to allow endian detection
83 uint16_t version; // Version number.
84 uint16_t hash_function; // The hash function enumeration that was used.
85 uint32_t bucket_count; // The number of buckets in this hash table.
86 uint32_t hashes_count; // The total number of unique hash values
87 // and hash data offsets in this table.
88 uint32_t header_data_len; // The bytes to skip to get to the hash
89 // indexes (buckets) for correct alignment.
90 // Also written to disk is the implementation specific header data.
92 static const uint32_t MagicHash = 0x48415348;
94 TableHeader(uint32_t data_len)
95 : magic(MagicHash), version(1),
96 hash_function(dwarf::DW_hash_function_djb), bucket_count(0),
97 hashes_count(0), header_data_len(data_len) {}
100 void print(raw_ostream &O) {
101 O << "Magic: " << format("0x%x", magic) << "\n"
102 << "Version: " << version << "\n"
103 << "Hash Function: " << hash_function << "\n"
104 << "Bucket Count: " << bucket_count << "\n"
105 << "Header Data Length: " << header_data_len << "\n";
107 void dump() { print(dbgs()); }
112 // The HeaderData describes the form of each set of data. In general this
113 // is as a list of atoms (atom_count) where each atom contains a type
114 // (AtomType type) of data, and an encoding form (form). In the case of
115 // data that is referenced via DW_FORM_ref_* the die_offset_base is
116 // used to describe the offset for all forms in the list of atoms.
117 // This also serves as a public interface of sorts.
118 // When written to disk this will have the form:
120 // uint32_t die_offset_base
121 // uint32_t atom_count
124 // Make these public so that they can be used as a general interface to
127 uint16_t type; // enum AtomType
128 uint16_t form; // DWARF DW_FORM_ defines
130 Atom(uint16_t type, uint16_t form) : type(type), form(form) {}
132 void print(raw_ostream &O) {
133 O << "Type: " << dwarf::AtomTypeString(type) << "\n"
134 << "Form: " << dwarf::FormEncodingString(form) << "\n";
136 void dump() { print(dbgs()); }
141 struct TableHeaderData {
142 uint32_t die_offset_base;
143 SmallVector<Atom, 1> Atoms;
145 TableHeaderData(ArrayRef<Atom> AtomList, uint32_t offset = 0)
146 : die_offset_base(offset), Atoms(AtomList.begin(), AtomList.end()) {}
149 void print(raw_ostream &O) {
150 O << "die_offset_base: " << die_offset_base << "\n";
151 for (size_t i = 0; i < Atoms.size(); i++)
154 void dump() { print(dbgs()); }
158 // The data itself consists of a str_offset, a count of the DIEs in the
159 // hash and the offsets to the DIEs themselves.
160 // On disk each data section is ended with a 0 KeyType as the end of the
162 // On output this looks like:
163 // uint32_t str_offset
164 // uint32_t hash_data_count
165 // HashData[hash_data_count]
167 struct HashDataContents {
168 const DIE *Die; // Offsets
169 char Flags; // Specific flags to output
171 HashDataContents(const DIE *D, char Flags) : Die(D), Flags(Flags) {}
173 void print(raw_ostream &O) const {
174 O << " Offset: " << Die->getOffset() << "\n";
175 O << " Tag: " << dwarf::TagString(Die->getTag()) << "\n";
176 O << " Flags: " << Flags << "\n";
186 ArrayRef<HashDataContents *> Data; // offsets
187 HashData(StringRef S, ArrayRef<HashDataContents *> Data)
188 : Str(S), Data(Data) {
189 HashValue = DwarfAccelTable::HashDJB(S);
192 void print(raw_ostream &O) {
193 O << "Name: " << Str << "\n";
194 O << " Hash Value: " << format("0x%x", HashValue) << "\n";
201 for (size_t i = 0; i < Data.size(); i++) {
202 O << " Offset: " << Data[i]->Die->getOffset() << "\n";
203 O << " Tag: " << dwarf::TagString(Data[i]->Die->getTag()) << "\n";
204 O << " Flags: " << Data[i]->Flags << "\n";
207 void dump() { print(dbgs()); }
211 DwarfAccelTable(const DwarfAccelTable &) LLVM_DELETED_FUNCTION;
212 void operator=(const DwarfAccelTable &) LLVM_DELETED_FUNCTION;
214 // Internal Functions
215 void EmitHeader(AsmPrinter *);
216 void EmitBuckets(AsmPrinter *);
217 void EmitHashes(AsmPrinter *);
218 void EmitOffsets(AsmPrinter *, MCSymbol *);
219 void EmitData(AsmPrinter *, DwarfFile *D);
221 // Allocator for HashData and HashDataContents.
222 BumpPtrAllocator Allocator;
226 TableHeaderData HeaderData;
227 std::vector<HashData *> Data;
230 typedef std::vector<HashDataContents *> DataArray;
231 typedef StringMap<DataArray, BumpPtrAllocator &> StringEntries;
232 StringEntries Entries;
234 // Buckets/Hashes/Offsets
235 typedef std::vector<HashData *> HashList;
236 typedef std::vector<HashList> BucketList;
240 // Public Implementation
242 DwarfAccelTable(ArrayRef<DwarfAccelTable::Atom>);
244 void AddName(StringRef, const DIE *, char = 0);
245 void FinalizeTable(AsmPrinter *, StringRef);
246 void Emit(AsmPrinter *, MCSymbol *, DwarfFile *);
248 void print(raw_ostream &O);
249 void dump() { print(dbgs()); }