Chris@4: /* adler32.c -- compute the Adler-32 checksum of a data stream
Chris@4:  * Copyright (C) 1995-2011 Mark Adler
Chris@4:  * For conditions of distribution and use, see copyright notice in zlib.h
Chris@4:  */
Chris@4: 
Chris@4: /* @(#) $Id$ */
Chris@4: 
Chris@4: #include "zutil.h"
Chris@4: 
Chris@4: #define local static
Chris@4: 
Chris@4: local uLong adler32_combine_ OF((uLong adler1, uLong adler2, z_off64_t len2));
Chris@4: 
Chris@4: #define BASE 65521      /* largest prime smaller than 65536 */
Chris@4: #define NMAX 5552
Chris@4: /* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
Chris@4: 
Chris@4: #define DO1(buf,i)  {adler += (buf)[i]; sum2 += adler;}
Chris@4: #define DO2(buf,i)  DO1(buf,i); DO1(buf,i+1);
Chris@4: #define DO4(buf,i)  DO2(buf,i); DO2(buf,i+2);
Chris@4: #define DO8(buf,i)  DO4(buf,i); DO4(buf,i+4);
Chris@4: #define DO16(buf)   DO8(buf,0); DO8(buf,8);
Chris@4: 
Chris@4: /* use NO_DIVIDE if your processor does not do division in hardware --
Chris@4:    try it both ways to see which is faster */
Chris@4: #ifdef NO_DIVIDE
Chris@4: /* note that this assumes BASE is 65521, where 65536 % 65521 == 15
Chris@4:    (thank you to John Reiser for pointing this out) */
Chris@4: #  define CHOP(a) \
Chris@4:     do { \
Chris@4:         unsigned long tmp = a >> 16; \
Chris@4:         a &= 0xffffUL; \
Chris@4:         a += (tmp << 4) - tmp; \
Chris@4:     } while (0)
Chris@4: #  define MOD28(a) \
Chris@4:     do { \
Chris@4:         CHOP(a); \
Chris@4:         if (a >= BASE) a -= BASE; \
Chris@4:     } while (0)
Chris@4: #  define MOD(a) \
Chris@4:     do { \
Chris@4:         CHOP(a); \
Chris@4:         MOD28(a); \
Chris@4:     } while (0)
Chris@4: #  define MOD63(a) \
Chris@4:     do { /* this assumes a is not negative */ \
Chris@4:         z_off64_t tmp = a >> 32; \
Chris@4:         a &= 0xffffffffL; \
Chris@4:         a += (tmp << 8) - (tmp << 5) + tmp; \
Chris@4:         tmp = a >> 16; \
Chris@4:         a &= 0xffffL; \
Chris@4:         a += (tmp << 4) - tmp; \
Chris@4:         tmp = a >> 16; \
Chris@4:         a &= 0xffffL; \
Chris@4:         a += (tmp << 4) - tmp; \
Chris@4:         if (a >= BASE) a -= BASE; \
Chris@4:     } while (0)
Chris@4: #else
Chris@4: #  define MOD(a) a %= BASE
Chris@4: #  define MOD28(a) a %= BASE
Chris@4: #  define MOD63(a) a %= BASE
Chris@4: #endif
Chris@4: 
Chris@4: /* ========================================================================= */
Chris@4: uLong ZEXPORT adler32(adler, buf, len)
Chris@4:     uLong adler;
Chris@4:     const Bytef *buf;
Chris@4:     uInt len;
Chris@4: {
Chris@4:     unsigned long sum2;
Chris@4:     unsigned n;
Chris@4: 
Chris@4:     /* split Adler-32 into component sums */
Chris@4:     sum2 = (adler >> 16) & 0xffff;
Chris@4:     adler &= 0xffff;
Chris@4: 
Chris@4:     /* in case user likes doing a byte at a time, keep it fast */
Chris@4:     if (len == 1) {
Chris@4:         adler += buf[0];
Chris@4:         if (adler >= BASE)
Chris@4:             adler -= BASE;
Chris@4:         sum2 += adler;
Chris@4:         if (sum2 >= BASE)
Chris@4:             sum2 -= BASE;
Chris@4:         return adler | (sum2 << 16);
Chris@4:     }
Chris@4: 
Chris@4:     /* initial Adler-32 value (deferred check for len == 1 speed) */
Chris@4:     if (buf == Z_NULL)
Chris@4:         return 1L;
Chris@4: 
Chris@4:     /* in case short lengths are provided, keep it somewhat fast */
Chris@4:     if (len < 16) {
Chris@4:         while (len--) {
Chris@4:             adler += *buf++;
Chris@4:             sum2 += adler;
Chris@4:         }
Chris@4:         if (adler >= BASE)
Chris@4:             adler -= BASE;
Chris@4:         MOD28(sum2);            /* only added so many BASE's */
Chris@4:         return adler | (sum2 << 16);
Chris@4:     }
Chris@4: 
Chris@4:     /* do length NMAX blocks -- requires just one modulo operation */
Chris@4:     while (len >= NMAX) {
Chris@4:         len -= NMAX;
Chris@4:         n = NMAX / 16;          /* NMAX is divisible by 16 */
Chris@4:         do {
Chris@4:             DO16(buf);          /* 16 sums unrolled */
Chris@4:             buf += 16;
Chris@4:         } while (--n);
Chris@4:         MOD(adler);
Chris@4:         MOD(sum2);
Chris@4:     }
Chris@4: 
Chris@4:     /* do remaining bytes (less than NMAX, still just one modulo) */
Chris@4:     if (len) {                  /* avoid modulos if none remaining */
Chris@4:         while (len >= 16) {
Chris@4:             len -= 16;
Chris@4:             DO16(buf);
Chris@4:             buf += 16;
Chris@4:         }
Chris@4:         while (len--) {
Chris@4:             adler += *buf++;
Chris@4:             sum2 += adler;
Chris@4:         }
Chris@4:         MOD(adler);
Chris@4:         MOD(sum2);
Chris@4:     }
Chris@4: 
Chris@4:     /* return recombined sums */
Chris@4:     return adler | (sum2 << 16);
Chris@4: }
Chris@4: 
Chris@4: /* ========================================================================= */
Chris@4: local uLong adler32_combine_(adler1, adler2, len2)
Chris@4:     uLong adler1;
Chris@4:     uLong adler2;
Chris@4:     z_off64_t len2;
Chris@4: {
Chris@4:     unsigned long sum1;
Chris@4:     unsigned long sum2;
Chris@4:     unsigned rem;
Chris@4: 
Chris@4:     /* for negative len, return invalid adler32 as a clue for debugging */
Chris@4:     if (len2 < 0)
Chris@4:         return 0xffffffffUL;
Chris@4: 
Chris@4:     /* the derivation of this formula is left as an exercise for the reader */
Chris@4:     MOD63(len2);                /* assumes len2 >= 0 */
Chris@4:     rem = (unsigned)len2;
Chris@4:     sum1 = adler1 & 0xffff;
Chris@4:     sum2 = rem * sum1;
Chris@4:     MOD(sum2);
Chris@4:     sum1 += (adler2 & 0xffff) + BASE - 1;
Chris@4:     sum2 += ((adler1 >> 16) & 0xffff) + ((adler2 >> 16) & 0xffff) + BASE - rem;
Chris@4:     if (sum1 >= BASE) sum1 -= BASE;
Chris@4:     if (sum1 >= BASE) sum1 -= BASE;
Chris@4:     if (sum2 >= (BASE << 1)) sum2 -= (BASE << 1);
Chris@4:     if (sum2 >= BASE) sum2 -= BASE;
Chris@4:     return sum1 | (sum2 << 16);
Chris@4: }
Chris@4: 
Chris@4: /* ========================================================================= */
Chris@4: uLong ZEXPORT adler32_combine(adler1, adler2, len2)
Chris@4:     uLong adler1;
Chris@4:     uLong adler2;
Chris@4:     z_off_t len2;
Chris@4: {
Chris@4:     return adler32_combine_(adler1, adler2, len2);
Chris@4: }
Chris@4: 
Chris@4: uLong ZEXPORT adler32_combine64(adler1, adler2, len2)
Chris@4:     uLong adler1;
Chris@4:     uLong adler2;
Chris@4:     z_off64_t len2;
Chris@4: {
Chris@4:     return adler32_combine_(adler1, adler2, len2);
Chris@4: }