libstdc++-v3/libsupc++/hash_bytes.cc

   1 // Definition of _Hash_bytes. -*- C++ -*-
   2
   3 // Copyright (C) 2010 Free Software Foundation, Inc.
   4 //
   5 // This file is part of the GNU ISO C++ Library.  This library is free
   6 // software; you can redistribute it and/or modify it under the
   7 // terms of the GNU General Public License as published by the
   8 // Free Software Foundation; either version 3, or (at your option)
   9 // any later version.
  10
  11 // This library is distributed in the hope that it will be useful,
  12 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 // GNU General Public License for more details.
  15
  16 // Under Section 7 of GPL version 3, you are granted additional
  17 // permissions described in the GCC Runtime Library Exception, version
  18 // 3.1, as published by the Free Software Foundation.
  19
  20 // You should have received a copy of the GNU General Public License and
  21 // a copy of the GCC Runtime Library Exception along with this program;
  22 // see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
  23 // <http://www.gnu.org/licenses/>.
  24
  25 // This file defines Hash_bytes, a primitive used for defining hash
  26 // functions. Based on public domain MurmurHashUnaligned2, by Austin
  27 // Appleby.  http://murmurhash.googlepages.com/
  28
  29 // This file also defines _Fnv_hash_bytes, another primitive with
  30 // exactly the same interface but using a different hash algorithm,
  31 // Fowler / Noll / Vo (FNV) Hash (type FNV-1a). The Murmur hash
  32 // function apears to be better in both speed and hash quality, and
  33 // FNV is provided primarily for backward compatibility.
  34
  35 #include <bits/c++config.h>
  36
  37 namespace
  38 {
  39   inline std::size_t
  40   unaligned_load(const char* p)
  41   {
  42     std::size_t result;
  43     __builtin_memcpy(&result, p, sizeof(result));
  44     return result;
  45   }
  46
  47 #if __SIZEOF_SIZE_T__ == 8
  48   // Loads n bytes, where 1 <= n < 8.
  49   inline std::size_t
  50   load_bytes(const char* p, int n)
  51   {
  52     std::size_t result = 0;
  53     --n;
  54     do
  55       result = (result << 8) + static_cast<unsigned char>(p[n]);
  56     while (--n >= 0);
  57     return result;
  58   }
  59
  60   inline std::size_t
  61   shift_mix(std::size_t v)
  62   { return v ^ (v >> 47);}
  63 #endif
  64 }
  65
  66 namespace std
  67 {
  68 #if __SIZEOF_SIZE_T__ == 4
  69
  70   // Implementation of Murmur hash for 32-bit size_t.
  71   size_t
  72   _Hash_bytes(const void* ptr, size_t len, size_t seed)
  73   {
  74     const size_t m = 0x5bd1e995;
  75     size_t hash = seed ^ len;
  76     const char* buf = static_cast<const char*>(ptr);
  77
  78     // Mix 4 bytes at a time into the hash.
  79     while(len >= 4)
  80       {
  81         size_t k = unaligned_load(buf);
  82         k *= m;
  83         k ^= k >> 24;
  84         k *= m;
  85         hash *= m;
  86         hash ^= k;
  87         buf += 4;
  88         len -= 4;
  89       }
  90
  91     // Handle the last few bytes of the input array.
  92     switch(len)
  93       {
  94       case 3:
  95         hash ^= static_cast<unsigned char>(buf[2]) << 16;
  96       case 2:
  97         hash ^= static_cast<unsigned char>(buf[1]) << 8;
  98       case 1:
  99         hash ^= static_cast<unsigned char>(buf[0]);
 100         hash *= m;
 101       };
 102
 103     // Do a few final mixes of the hash.
 104     hash ^= hash >> 13;
 105     hash *= m;
 106     hash ^= hash >> 15;
 107     return hash;
 108   }
 109
 110   // Implementation of FNV hash for 32-bit size_t.
 111   size_t
 112   _Fnv_hash_bytes(const void* ptr, size_t len, size_t hash)
 113   {
 114     const char* cptr = static_cast<const char*>(ptr);
 115     for (; len; --len)
 116       {
 117         hash ^= static_cast<size_t>(*cptr++);
 118         hash *= static_cast<size_t>(16777619UL);
 119       }
 120     return hash;
 121   }
 122
 123 #elif __SIZEOF_SIZE_T__ == 8
 124
 125   // Implementation of Murmur hash for 64-bit size_t.
 126   size_t
 127   _Hash_bytes(const void* ptr, size_t len, size_t seed)
 128   {
 129     static const size_t mul = (0xc6a4a793UL << 32UL) + 0x5bd1e995UL;
 130     const char* const buf = static_cast<const char*>(ptr);
 131
 132     // Remove the bytes not divisible by the sizeof(size_t).  This
 133     // allows the main loop to process the data as 64-bit integers.
 134     const int len_aligned = len & ~0x7;
 135     const char* const end = buf + len_aligned;
 136     size_t hash = seed ^ (len * mul);
 137     for (const char* p = buf; p != end; p += 8)
 138       {
 139         const size_t data = shift_mix(unaligned_load(p) * mul) * mul;
 140         hash ^= data;
 141         hash *= mul;
 142       }
 143     if ((len & 0x7) != 0)
 144       {
 145         const size_t data = load_bytes(end, len & 0x7);
 146         hash ^= data;
 147         hash *= mul;
 148       }
 149     hash = shift_mix(hash) * mul;
 150     hash = shift_mix(hash);
 151     return hash;
 152   }
 153
 154   // Implementation of FNV hash for 64-bit size_t.
 155   size_t
 156   _Fnv_hash_bytes(const void* ptr, size_t len, size_t hash)
 157   {
 158     const char* cptr = static_cast<const char*>(ptr);
 159     for (; len; --len)
 160       {
 161         hash ^= static_cast<size_t>(*cptr++);
 162         hash *= static_cast<size_t>(1099511628211ULL);
 163       }
 164     return hash;
 165   }
 166
 167 #else
 168
 169   // Dummy hash implementation for unusual sizeof(size_t).
 170   size_t
 171   _Hash_bytes(const void* ptr, size_t len, size_t seed)
 172   {
 173     size_t hash = seed;
 174     const char* cptr = reinterpret_cast<const char*>(ptr);
 175     for (; len; --len)
 176       hash = (hash * 131) + *cptr++;
 177     return hash;
 178   }
 179
 180   size_t
 181   _Fnv_hash_bytes(const void* ptr, size_t len, size_t seed)
 182   { return _Hash_bytes(ptr, len, seed); }
 183
 184 #endif /* __SIZEOF_SIZE_T__ */
 185 }