X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~mdw/git/mLib/blobdiff_plain/dd3c57bc8cac59e0d657ee665ce462988d27d714..18c831dcd0ae4d660c70ccac69d27ed2a97851be:/unihash.c?ds=sidebyside diff --git a/unihash.c b/unihash.c deleted file mode 100644 index 2997e4f..0000000 --- a/unihash.c +++ /dev/null @@ -1,217 +0,0 @@ -/* -*-c-*- - * - * Simple and efficient universal hashing for hashtables - * - * (c) 2003 Straylight/Edgeware - */ - -/*----- Licensing notice --------------------------------------------------* - * - * This file is part of the mLib utilities library. - * - * mLib is free software; you can redistribute it and/or modify - * it under the terms of the GNU Library General Public License as - * published by the Free Software Foundation; either version 2 of the - * License, or (at your option) any later version. - * - * mLib is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU Library General Public License for more details. - * - * You should have received a copy of the GNU Library General Public - * License along with mLib; if not, write to the Free - * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, - * MA 02111-1307, USA. - */ - -/*----- Header files ------------------------------------------------------*/ - -#include -#include - -#include "unihash.h" - -/*----- Main code ---------------------------------------------------------*/ - -/* --- @unihash_setkey@ --- * - * - * Arguments: @unihash_info *i@ = where to store the precomputed tables - * @uint32 k@ = the key to set, randomly chosen - * - * Returns: --- - * - * Use: Calculates the tables required for efficient hashing. - */ - -static uint32 mul(uint32 x, uint32 y) -{ - uint32 z = 0; - while (y) { - if (y & 1) z ^= x; - if (x & (1 << 31)) - x = U32(x << 1) ^ UNIHASH_POLY; - else - x = U32(x << 1); - y = U32(y >> 1); - } - return (z); -} - -void unihash_setkey(unihash_info *i, uint32 k) -{ - size_t a; - size_t b; - uint32 x = 1; - - for (a = 0; a < UNIHASH_NBATCH; a++) { - x = mul(x, k); - for (b = 0; b < 256; b++) { - i->s[a][0][b] = mul(x, b << 0); - i->s[a][1][b] = mul(x, b << 8); - i->s[a][2][b] = mul(x, b << 16); - i->s[a][3][b] = mul(x, b << 24); - } - } -} - -/* --- @unihash_hash@ --- * - * - * Arguments: @const unihash_info *i@ = pointer to precomputed table - * @uint32 a@ = @i->[0][0][1]@ or value from previous call - * @const void *p@ = pointer to data to hash - * @size_t sz@ = size of the data - * - * Returns: Hash of data so far. - * - * Use: Hashes data. Call this as many times as needed. - */ - -uint32 unihash_hash(const unihash_info *i, uint32 a, - const void *p, size_t sz) -{ - const octet *pp = p; - - assert(UNIHASH_NBATCH == 4); - -#define FULLMULT(u, x) \ - (i->s[u][0][U8((x) >> 0)] ^ i->s[u][1][U8((x) >> 8)] ^ \ - i->s[u][2][U8((x) >> 16)] ^ i->s[u][3][U8((x) >> 24)]); - -#define BYTEMULT(u, x) i->s[u][0][x] - - /* --- Do the main bulk in batches of %$n$% bytes --- * - * - * We have %$a$% and %$m_{n-1}, \ldots, m_1, m_0$%; we want - * - * %$a' = (a + m_{n-1}) k^n + m_{n-2} k^{n-1} + \cdots + m_1 k^2 + m_0 k$% - */ - - while (sz >= UNIHASH_NBATCH) { - a ^= *pp++; - a = FULLMULT(3, a); - a ^= BYTEMULT(2, *pp++); - a ^= BYTEMULT(1, *pp++); - a ^= BYTEMULT(0, *pp++); - sz -= UNIHASH_NBATCH; - } - - /* --- The tail end is a smaller batch --- */ - - switch (sz) { - case 3: a ^= *pp++; a = FULLMULT(2, a); goto batch_2; - case 2: a ^= *pp++; a = FULLMULT(1, a); goto batch_1; - case 1: a ^= *pp++; a = FULLMULT(0, a); goto batch_0; - batch_2: a ^= BYTEMULT(1, *pp++); - batch_1: a ^= BYTEMULT(0, *pp++); - batch_0: break; - } - - return (a); -} - -/* --- @unihash@ --- * - * - * Arguments: @const unihash_info *i@ = precomputed tables - * @const void *p@ = pointer to data to hash - * @size_t sz@ = size of the data - * - * Returns: The hash value computed. - * - * Use: All-in-one hashing function. No faster than using the - * separate calls, but more convenient. - */ - -uint32 unihash(const unihash_info *i, const void *p, size_t sz) -{ - return (UNIHASH(i, p, sz)); -} - -/*----- Test rig ----------------------------------------------------------*/ - -#ifdef TEST_RIG - -#include "testrig.h" - -static int verify(dstr *v) -{ - unihash_info ui; - uint32 k; - uint32 h, hh; - size_t n; - int i, c; - const char *p; - int ok = 1; - - static const int step[] = { 0, 1, 5, 6, 7, 8, 23, -1 }; - - /* --- Set up for using this key --- */ - - k = *(uint32 *)v[0].buf; - h = *(uint32 *)v[2].buf; - unihash_setkey(&ui, k); - - /* --- Hash the data a lot --- */ - - for (i = 0; step[i] >= 0; i++) { - c = step[i]; - if (!c) - hh = unihash(&ui, v[1].buf, v[1].len); - else { - hh = UNIHASH_INIT(&ui); - p = v[1].buf; - n = v[1].len; - while (n) { - if (c > n) c = n; - hh = unihash_hash(&ui, hh, p, c); - p += c; - n -= c; - } - } - if (h != hh) { - ok = 0; - fprintf(stderr, "\nunihash failed\n"); - fprintf(stderr, " key = %08lx\n", (unsigned long)k); - fprintf(stderr, " data = %s\n", v[1].buf); - fprintf(stderr, " step = %d\n", step[i]); - fprintf(stderr, " expected = %08lx\n", (unsigned long)h); - fprintf(stderr, " computed = %08lx\n", (unsigned long)hh); - } - } - return (ok); -} - -static const test_chunk tests[] = { - { "hash", verify, { &type_uint32, &type_string, &type_uint32 } }, - { 0, 0, { 0 } } -}; - -int main(int argc, char *argv[]) -{ - test_run(argc, argv, tests, "unihash.in"); - return (0); -} - -#endif - -/*----- That's all, folks -------------------------------------------------*/