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