mdw@git.distorted.org.uk Git - mLib/blob - hash/unihash.c

   1 /* -*-c-*-
   2  *
   3  * Simple and efficient universal hashing for hashtables
   4  *
   5  * (c) 2003 Straylight/Edgeware
   6  */
   7
   8 /*----- Licensing notice --------------------------------------------------*
   9  *
  10  * This file is part of the mLib utilities library.
  11  *
  12  * mLib is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU Library General Public License as
  14  * published by the Free Software Foundation; either version 2 of the
  15  * License, or (at your option) any later version.
  16  *
  17  * mLib is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  * GNU Library General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Library General Public
  23  * License along with mLib; if not, write to the Free
  24  * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
  25  * MA 02111-1307, USA.
  26  */
  27
  28 /*----- Header files ------------------------------------------------------*/
  29
  30 #include <assert.h>
  31 #include <stdlib.h>
  32
  33 #include "unihash.h"
  34
  35 /*----- Main code ---------------------------------------------------------*/
  36
  37 /* --- @unihash_setkey@ --- *
  38  *
  39  * Arguments:   @unihash_info *i@ = where to store the precomputed tables
  40  *              @uint32 k@ = the key to set, randomly chosen
  41  *
  42  * Returns:     ---
  43  *
  44  * Use:         Calculates the tables required for efficient hashing.
  45  */
  46
  47 static uint32 mul(uint32 x, uint32 y)
  48 {
  49   uint32 z = 0;
  50   while (y) {
  51     if (y & 1) z ^= x;
  52     if (x & (1 << 31))
  53       x = U32(x << 1) ^ UNIHASH_POLY;
  54     else
  55       x = U32(x << 1);
  56     y = U32(y >> 1);
  57   }
  58   return (z);
  59 }
  60
  61 void unihash_setkey(unihash_info *i, uint32 k)
  62 {
  63   size_t a;
  64   size_t b;
  65   uint32 x = 1;
  66
  67   for (a = 0; a < UNIHASH_NBATCH; a++) {
  68     x = mul(x, k);
  69     for (b = 0; b < 256; b++) {
  70       i->s[a][0][b] = mul(x, b <<  0);
  71       i->s[a][1][b] = mul(x, b <<  8);
  72       i->s[a][2][b] = mul(x, b << 16);
  73       i->s[a][3][b] = mul(x, b << 24);
  74     }
  75   }
  76 }
  77
  78 /* --- @unihash_hash@ --- *
  79  *
  80  * Arguments:   @const unihash_info *i@ = pointer to precomputed table
  81  *              @uint32 a@ = @i->[0][0][1]@ or value from previous call
  82  *              @const void *p@ = pointer to data to hash
  83  *              @size_t sz@ = size of the data
  84  *
  85  * Returns:     Hash of data so far.
  86  *
  87  * Use:         Hashes data.  Call this as many times as needed.
  88  */
  89
  90 uint32 unihash_hash(const unihash_info *i, uint32 a,
  91                     const void *p, size_t sz)
  92 {
  93   const octet *pp = p;
  94
  95   assert(UNIHASH_NBATCH == 4);
  96
  97 #define FULLMULT(u, x)                                                  \
  98   (i->s[u][0][U8((x) >>  0)] ^ i->s[u][1][U8((x) >>  8)] ^              \
  99    i->s[u][2][U8((x) >> 16)] ^ i->s[u][3][U8((x) >> 24)]);
 100
 101 #define BYTEMULT(u, x) i->s[u][0][x]
 102
 103   /* --- Do the main bulk in batches of %$n$% bytes --- *
 104    *
 105    * We have %$a$% and %$m_{n-1}, \ldots, m_1, m_0$%; we want
 106    *
 107    * %$a' = (a + m_{n-1}) k^n + m_{n-2} k^{n-1} + \cdots + m_1 k^2 + m_0 k$%
 108    */
 109
 110   while (sz >= UNIHASH_NBATCH) {
 111     a ^= *pp++;
 112     a = FULLMULT(3, a);
 113     a ^= BYTEMULT(2, *pp++);
 114     a ^= BYTEMULT(1, *pp++);
 115     a ^= BYTEMULT(0, *pp++);
 116     sz -= UNIHASH_NBATCH;
 117   }
 118
 119   /* --- The tail end is a smaller batch --- */
 120
 121   switch (sz) {
 122     case  3: a ^= *pp++; a = FULLMULT(2, a); goto batch_2;
 123     case  2: a ^= *pp++; a = FULLMULT(1, a); goto batch_1;
 124     case  1: a ^= *pp++; a = FULLMULT(0, a); goto batch_0;
 125     batch_2: a ^= BYTEMULT(1, *pp++);
 126     batch_1: a ^= BYTEMULT(0, *pp++);
 127     batch_0: break;
 128   }
 129
 130   return (a);
 131 }
 132
 133 /* --- @unihash@ --- *
 134  *
 135  * Arguments:   @const unihash_info *i@ = precomputed tables
 136  *              @const void *p@ = pointer to data to hash
 137  *              @size_t sz@ = size of the data
 138  *
 139  * Returns:     The hash value computed.
 140  *
 141  * Use:         All-in-one hashing function.  No faster than using the
 142  *              separate calls, but more convenient.
 143  */
 144
 145 uint32 unihash(const unihash_info *i, const void *p, size_t sz)
 146   { return (UNIHASH(i, p, sz)); }
 147
 148 /*----- That's all, folks -------------------------------------------------*/