DwarfAccelTable.h revision 234285
1234285Sdim//==-- llvm/CodeGen/DwarfAccelTable.h - Dwarf Accelerator Tables -*- C++ -*-==// 2234285Sdim// 3234285Sdim// The LLVM Compiler Infrastructure 4234285Sdim// 5234285Sdim// This file is distributed under the University of Illinois Open Source 6234285Sdim// License. See LICENSE.TXT for details. 7234285Sdim// 8234285Sdim//===----------------------------------------------------------------------===// 9234285Sdim// 10234285Sdim// This file contains support for writing dwarf accelerator tables. 11234285Sdim// 12234285Sdim//===----------------------------------------------------------------------===// 13234285Sdim 14234285Sdim#ifndef CODEGEN_ASMPRINTER_DWARFACCELTABLE_H__ 15234285Sdim#define CODEGEN_ASMPRINTER_DWARFACCELTABLE_H__ 16234285Sdim 17234285Sdim#include "llvm/ADT/StringMap.h" 18234285Sdim#include "llvm/MC/MCSymbol.h" 19234285Sdim#include "llvm/Support/Dwarf.h" 20234285Sdim#include "llvm/Support/DataTypes.h" 21234285Sdim#include "llvm/Support/Debug.h" 22234285Sdim#include "llvm/Support/ErrorHandling.h" 23234285Sdim#include "llvm/Support/Format.h" 24234285Sdim#include "llvm/Support/FormattedStream.h" 25234285Sdim#include "DIE.h" 26234285Sdim#include <vector> 27234285Sdim#include <map> 28234285Sdim 29234285Sdim// The dwarf accelerator tables are an indirect hash table optimized 30234285Sdim// for null lookup rather than access to known data. They are output into 31234285Sdim// an on-disk format that looks like this: 32234285Sdim// 33234285Sdim// .-------------. 34234285Sdim// | HEADER | 35234285Sdim// |-------------| 36234285Sdim// | BUCKETS | 37234285Sdim// |-------------| 38234285Sdim// | HASHES | 39234285Sdim// |-------------| 40234285Sdim// | OFFSETS | 41234285Sdim// |-------------| 42234285Sdim// | DATA | 43234285Sdim// `-------------' 44234285Sdim// 45234285Sdim// where the header contains a magic number, version, type of hash function, 46234285Sdim// the number of buckets, total number of hashes, and room for a special 47234285Sdim// struct of data and the length of that struct. 48234285Sdim// 49234285Sdim// The buckets contain an index (e.g. 6) into the hashes array. The hashes 50234285Sdim// section contains all of the 32-bit hash values in contiguous memory, and 51234285Sdim// the offsets contain the offset into the data area for the particular 52234285Sdim// hash. 53234285Sdim// 54234285Sdim// For a lookup example, we could hash a function name and take it modulo the 55234285Sdim// number of buckets giving us our bucket. From there we take the bucket value 56234285Sdim// as an index into the hashes table and look at each successive hash as long 57234285Sdim// as the hash value is still the same modulo result (bucket value) as earlier. 58234285Sdim// If we have a match we look at that same entry in the offsets table and 59234285Sdim// grab the offset in the data for our final match. 60234285Sdim 61234285Sdimnamespace llvm { 62234285Sdim 63234285Sdimclass AsmPrinter; 64234285Sdimclass DIE; 65234285Sdimclass DwarfDebug; 66234285Sdim 67234285Sdimclass DwarfAccelTable { 68234285Sdim 69234285Sdim enum HashFunctionType { 70234285Sdim eHashFunctionDJB = 0u 71234285Sdim }; 72234285Sdim 73234285Sdim static uint32_t HashDJB (StringRef Str) { 74234285Sdim uint32_t h = 5381; 75234285Sdim for (unsigned i = 0, e = Str.size(); i != e; ++i) 76234285Sdim h = ((h << 5) + h) + Str[i]; 77234285Sdim return h; 78234285Sdim } 79234285Sdim 80234285Sdim // Helper function to compute the number of buckets needed based on 81234285Sdim // the number of unique hashes. 82234285Sdim void ComputeBucketCount (void); 83234285Sdim 84234285Sdim struct TableHeader { 85234285Sdim uint32_t magic; // 'HASH' magic value to allow endian detection 86234285Sdim uint16_t version; // Version number. 87234285Sdim uint16_t hash_function; // The hash function enumeration that was used. 88234285Sdim uint32_t bucket_count; // The number of buckets in this hash table. 89234285Sdim uint32_t hashes_count; // The total number of unique hash values 90234285Sdim // and hash data offsets in this table. 91234285Sdim uint32_t header_data_len; // The bytes to skip to get to the hash 92234285Sdim // indexes (buckets) for correct alignment. 93234285Sdim // Also written to disk is the implementation specific header data. 94234285Sdim 95234285Sdim static const uint32_t MagicHash = 0x48415348; 96234285Sdim 97234285Sdim TableHeader (uint32_t data_len) : 98234285Sdim magic (MagicHash), version (1), hash_function (eHashFunctionDJB), 99234285Sdim bucket_count (0), hashes_count (0), header_data_len (data_len) 100234285Sdim {} 101234285Sdim 102234285Sdim#ifndef NDEBUG 103234285Sdim void print(raw_ostream &O) { 104234285Sdim O << "Magic: " << format("0x%x", magic) << "\n" 105234285Sdim << "Version: " << version << "\n" 106234285Sdim << "Hash Function: " << hash_function << "\n" 107234285Sdim << "Bucket Count: " << bucket_count << "\n" 108234285Sdim << "Header Data Length: " << header_data_len << "\n"; 109234285Sdim } 110234285Sdim void dump() { print(dbgs()); } 111234285Sdim#endif 112234285Sdim }; 113234285Sdim 114234285Sdimpublic: 115234285Sdim // The HeaderData describes the form of each set of data. In general this 116234285Sdim // is as a list of atoms (atom_count) where each atom contains a type 117234285Sdim // (AtomType type) of data, and an encoding form (form). In the case of 118234285Sdim // data that is referenced via DW_FORM_ref_* the die_offset_base is 119234285Sdim // used to describe the offset for all forms in the list of atoms. 120234285Sdim // This also serves as a public interface of sorts. 121234285Sdim // When written to disk this will have the form: 122234285Sdim // 123234285Sdim // uint32_t die_offset_base 124234285Sdim // uint32_t atom_count 125234285Sdim // atom_count Atoms 126234285Sdim enum AtomType { 127234285Sdim eAtomTypeNULL = 0u, 128234285Sdim eAtomTypeDIEOffset = 1u, // DIE offset, check form for encoding 129234285Sdim eAtomTypeCUOffset = 2u, // DIE offset of the compiler unit header that 130234285Sdim // contains the item in question 131234285Sdim eAtomTypeTag = 3u, // DW_TAG_xxx value, should be encoded as 132234285Sdim // DW_FORM_data1 (if no tags exceed 255) or 133234285Sdim // DW_FORM_data2. 134234285Sdim eAtomTypeNameFlags = 4u, // Flags from enum NameFlags 135234285Sdim eAtomTypeTypeFlags = 5u // Flags from enum TypeFlags 136234285Sdim }; 137234285Sdim 138234285Sdim enum TypeFlags { 139234285Sdim eTypeFlagClassMask = 0x0000000fu, 140234285Sdim 141234285Sdim // Always set for C++, only set for ObjC if this is the 142234285Sdim // @implementation for a class. 143234285Sdim eTypeFlagClassIsImplementation = ( 1u << 1 ) 144234285Sdim }; 145234285Sdim 146234285Sdim // Make these public so that they can be used as a general interface to 147234285Sdim // the class. 148234285Sdim struct Atom { 149234285Sdim AtomType type; // enum AtomType 150234285Sdim uint16_t form; // DWARF DW_FORM_ defines 151234285Sdim 152234285Sdim Atom(AtomType type, uint16_t form) : type(type), form(form) {} 153234285Sdim static const char * AtomTypeString(enum AtomType); 154234285Sdim#ifndef NDEBUG 155234285Sdim void print(raw_ostream &O) { 156234285Sdim O << "Type: " << AtomTypeString(type) << "\n" 157234285Sdim << "Form: " << dwarf::FormEncodingString(form) << "\n"; 158234285Sdim } 159234285Sdim void dump() { 160234285Sdim print(dbgs()); 161234285Sdim } 162234285Sdim#endif 163234285Sdim }; 164234285Sdim 165234285Sdim private: 166234285Sdim struct TableHeaderData { 167234285Sdim 168234285Sdim uint32_t die_offset_base; 169234285Sdim std::vector<Atom> Atoms; 170234285Sdim 171234285Sdim TableHeaderData(std::vector<DwarfAccelTable::Atom> &AtomList, 172234285Sdim uint32_t offset = 0) : 173234285Sdim die_offset_base(offset) { 174234285Sdim for (size_t i = 0, e = AtomList.size(); i != e; ++i) 175234285Sdim Atoms.push_back(AtomList[i]); 176234285Sdim } 177234285Sdim 178234285Sdim TableHeaderData(DwarfAccelTable::Atom Atom, uint32_t offset = 0) 179234285Sdim : die_offset_base(offset) { 180234285Sdim Atoms.push_back(Atom); 181234285Sdim } 182234285Sdim 183234285Sdim#ifndef NDEBUG 184234285Sdim void print (raw_ostream &O) { 185234285Sdim O << "die_offset_base: " << die_offset_base << "\n"; 186234285Sdim for (size_t i = 0; i < Atoms.size(); i++) 187234285Sdim Atoms[i].print(O); 188234285Sdim } 189234285Sdim void dump() { 190234285Sdim print(dbgs()); 191234285Sdim } 192234285Sdim#endif 193234285Sdim }; 194234285Sdim 195234285Sdim // The data itself consists of a str_offset, a count of the DIEs in the 196234285Sdim // hash and the offsets to the DIEs themselves. 197234285Sdim // On disk each data section is ended with a 0 KeyType as the end of the 198234285Sdim // hash chain. 199234285Sdim // On output this looks like: 200234285Sdim // uint32_t str_offset 201234285Sdim // uint32_t hash_data_count 202234285Sdim // HashData[hash_data_count] 203234285Sdimpublic: 204234285Sdim struct HashDataContents { 205234285Sdim DIE *Die; // Offsets 206234285Sdim char Flags; // Specific flags to output 207234285Sdim 208234285Sdim HashDataContents(DIE *D, char Flags) : 209234285Sdim Die(D), 210234285Sdim Flags(Flags) { } 211234285Sdim #ifndef NDEBUG 212234285Sdim void print(raw_ostream &O) const { 213234285Sdim O << " Offset: " << Die->getOffset() << "\n"; 214234285Sdim O << " Tag: " << dwarf::TagString(Die->getTag()) << "\n"; 215234285Sdim O << " Flags: " << Flags << "\n"; 216234285Sdim } 217234285Sdim #endif 218234285Sdim }; 219234285Sdimprivate: 220234285Sdim struct HashData { 221234285Sdim StringRef Str; 222234285Sdim uint32_t HashValue; 223234285Sdim MCSymbol *Sym; 224234285Sdim std::vector<struct HashDataContents*> Data; // offsets 225234285Sdim HashData(StringRef S) : Str(S) { 226234285Sdim HashValue = DwarfAccelTable::HashDJB(S); 227234285Sdim } 228234285Sdim void addData(struct HashDataContents *Datum) { Data.push_back(Datum); } 229234285Sdim #ifndef NDEBUG 230234285Sdim void print(raw_ostream &O) { 231234285Sdim O << "Name: " << Str << "\n"; 232234285Sdim O << " Hash Value: " << format("0x%x", HashValue) << "\n"; 233234285Sdim O << " Symbol: " ; 234234285Sdim if (Sym) Sym->print(O); 235234285Sdim else O << "<none>"; 236234285Sdim O << "\n"; 237234285Sdim for (size_t i = 0; i < Data.size(); i++) { 238234285Sdim O << " Offset: " << Data[i]->Die->getOffset() << "\n"; 239234285Sdim O << " Tag: " << dwarf::TagString(Data[i]->Die->getTag()) << "\n"; 240234285Sdim O << " Flags: " << Data[i]->Flags << "\n"; 241234285Sdim } 242234285Sdim } 243234285Sdim void dump() { 244234285Sdim print(dbgs()); 245234285Sdim } 246234285Sdim #endif 247234285Sdim }; 248234285Sdim 249234285Sdim DwarfAccelTable(const DwarfAccelTable&); // DO NOT IMPLEMENT 250234285Sdim void operator=(const DwarfAccelTable&); // DO NOT IMPLEMENT 251234285Sdim 252234285Sdim // Internal Functions 253234285Sdim void EmitHeader(AsmPrinter *); 254234285Sdim void EmitBuckets(AsmPrinter *); 255234285Sdim void EmitHashes(AsmPrinter *); 256234285Sdim void EmitOffsets(AsmPrinter *, MCSymbol *); 257234285Sdim void EmitData(AsmPrinter *, DwarfDebug *D); 258234285Sdim 259234285Sdim // Output Variables 260234285Sdim TableHeader Header; 261234285Sdim TableHeaderData HeaderData; 262234285Sdim std::vector<HashData*> Data; 263234285Sdim 264234285Sdim // String Data 265234285Sdim typedef std::vector<struct HashDataContents*> DataArray; 266234285Sdim typedef StringMap<DataArray> StringEntries; 267234285Sdim StringEntries Entries; 268234285Sdim 269234285Sdim // Buckets/Hashes/Offsets 270234285Sdim typedef std::vector<HashData*> HashList; 271234285Sdim typedef std::vector<HashList> BucketList; 272234285Sdim BucketList Buckets; 273234285Sdim HashList Hashes; 274234285Sdim 275234285Sdim // Public Implementation 276234285Sdim public: 277234285Sdim DwarfAccelTable(DwarfAccelTable::Atom); 278234285Sdim DwarfAccelTable(std::vector<DwarfAccelTable::Atom> &); 279234285Sdim ~DwarfAccelTable(); 280234285Sdim void AddName(StringRef, DIE*, char = 0); 281234285Sdim void FinalizeTable(AsmPrinter *, const char *); 282234285Sdim void Emit(AsmPrinter *, MCSymbol *, DwarfDebug *); 283234285Sdim#ifndef NDEBUG 284234285Sdim void print(raw_ostream &O); 285234285Sdim void dump() { print(dbgs()); } 286234285Sdim#endif 287234285Sdim}; 288234285Sdim 289234285Sdim} 290234285Sdim#endif 291