1/* adler32.c -- compute the Adler-32 checksum of a data stream
2 * Copyright (C) 1995-2011 Mark Adler
3 * For conditions of distribution and use, see copyright notice in zlib.h
4 */
5
6/* @(#) $Id$ */
7
8#include "zutil.h"
9
10#include <lib/cksum.h>
11
12#define BASE 65521      /* largest prime smaller than 65536 */
13#define NMAX 5552
14/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
15
16#define DO1(buf,i)  {adler += (buf)[i]; sum2 += adler;}
17#define DO2(buf,i)  DO1(buf,i); DO1(buf,i+1);
18#define DO4(buf,i)  DO2(buf,i); DO2(buf,i+2);
19#define DO8(buf,i)  DO4(buf,i); DO4(buf,i+4);
20#define DO16(buf)   DO8(buf,0); DO8(buf,8);
21
22/* use NO_DIVIDE if your processor does not do division in hardware --
23   try it both ways to see which is faster */
24#ifdef NO_DIVIDE
25/* note that this assumes BASE is 65521, where 65536 % 65521 == 15
26   (thank you to John Reiser for pointing this out) */
27#  define CHOP(a) \
28    do { \
29        uint32_t tmp = a >> 16; \
30        a &= 0xffffUL; \
31        a += (tmp << 4) - tmp; \
32    } while (0)
33#  define MOD28(a) \
34    do { \
35        CHOP(a); \
36        if (a >= BASE) a -= BASE; \
37    } while (0)
38#  define MOD(a) \
39    do { \
40        CHOP(a); \
41        MOD28(a); \
42    } while (0)
43#  define MOD63(a) \
44    do { /* this assumes a is not negative */ \
45        z_off64_t tmp = a >> 32; \
46        a &= 0xffffffffL; \
47        a += (tmp << 8) - (tmp << 5) + tmp; \
48        tmp = a >> 16; \
49        a &= 0xffffL; \
50        a += (tmp << 4) - tmp; \
51        tmp = a >> 16; \
52        a &= 0xffffL; \
53        a += (tmp << 4) - tmp; \
54        if (a >= BASE) a -= BASE; \
55    } while (0)
56#else
57#  define MOD(a) a %= BASE
58#  define MOD28(a) a %= BASE
59#  define MOD63(a) a %= BASE
60#endif
61
62/* ========================================================================= */
63uint32_t ZEXPORT adler32(uint32_t adler, const uint8_t* buf, size_t len)
64{
65    uint32_t sum2;
66    uint32_t n;
67
68    /* split Adler-32 into component sums */
69    sum2 = (adler >> 16) & 0xffff;
70    adler &= 0xffff;
71
72    /* in case user likes doing a byte at a time, keep it fast */
73    if (len == 1) {
74        adler += buf[0];
75        if (adler >= BASE)
76            adler -= BASE;
77        sum2 += adler;
78        if (sum2 >= BASE)
79            sum2 -= BASE;
80        return adler | (sum2 << 16);
81    }
82
83    /* initial Adler-32 value (deferred check for len == 1 speed) */
84    if (buf == Z_NULL)
85        return 1L;
86
87    /* in case short lengths are provided, keep it somewhat fast */
88    if (len < 16) {
89        while (len--) {
90            adler += *buf++;
91            sum2 += adler;
92        }
93        if (adler >= BASE)
94            adler -= BASE;
95        MOD28(sum2);            /* only added so many BASE's */
96        return adler | (sum2 << 16);
97    }
98
99    /* do length NMAX blocks -- requires just one modulo operation */
100    while (len >= NMAX) {
101        len -= NMAX;
102        n = NMAX / 16;          /* NMAX is divisible by 16 */
103        do {
104            DO16(buf);          /* 16 sums unrolled */
105            buf += 16;
106        } while (--n);
107        MOD(adler);
108        MOD(sum2);
109    }
110
111    /* do remaining bytes (less than NMAX, still just one modulo) */
112    if (len) {                  /* avoid modulos if none remaining */
113        while (len >= 16) {
114            len -= 16;
115            DO16(buf);
116            buf += 16;
117        }
118        while (len--) {
119            adler += *buf++;
120            sum2 += adler;
121        }
122        MOD(adler);
123        MOD(sum2);
124    }
125
126    /* return recombined sums */
127    return adler | (sum2 << 16);
128}
129
130/* ========================================================================= */
131uint32_t ZEXPORT adler32_combine(uint32_t adler1, uint32_t adler2, size_t len2)
132{
133    uint32_t sum1;
134    uint32_t sum2;
135    uint32_t rem;
136
137    /* the derivation of this formula is left as an exercise for the reader */
138    MOD63(len2);                /* assumes len2 >= 0 */
139    rem = (uint32_t)len2;
140    sum1 = adler1 & 0xffff;
141    sum2 = rem * sum1;
142    MOD(sum2);
143    sum1 += (adler2 & 0xffff) + BASE - 1;
144    sum2 += ((adler1 >> 16) & 0xffff) + ((adler2 >> 16) & 0xffff) + BASE - rem;
145    if (sum1 >= BASE) sum1 -= BASE;
146    if (sum1 >= BASE) sum1 -= BASE;
147    if (sum2 >= (BASE << 1)) sum2 -= (BASE << 1);
148    if (sum2 >= BASE) sum2 -= BASE;
149    return sum1 | (sum2 << 16);
150}
151