contrib/cvs-1.12/lib/memrchr.c

   1 /* memrchr -- find the last occurrence of a byte in a memory block
   2
   3    Copyright (C) 1991, 1993, 1996, 1997, 1999, 2000, 2003, 2004, 2005
   4    Free Software Foundation, Inc.
   5
   6    Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
   7    with help from Dan Sahlin (dan@sics.se) and
   8    commentary by Jim Blandy (jimb@ai.mit.edu);
   9    adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
  10    and implemented by Roland McGrath (roland@ai.mit.edu).
  11
  12    This program is free software; you can redistribute it and/or modify
  13    it under the terms of the GNU General Public License as published by
  14    the Free Software Foundation; either version 2, or (at your option)
  15    any later version.
  16
  17    This program 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 General Public License for more details.
  21
  22    You should have received a copy of the GNU General Public License along
  23    with this program; if not, write to the Free Software Foundation,
  24    Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */
  25
  26 #ifdef HAVE_CONFIG_H
  27 # include <config.h>
  28 #endif
  29
  30 #if defined _LIBC
  31 # include <string.h>
  32 # include <memcopy.h>
  33 #else
  34 # include "memrchr.h"
  35 # define reg_char char
  36 #endif
  37
  38 #include <limits.h>
  39
  40 #undef __memrchr
  41 #undef memrchr
  42
  43 #ifndef weak_alias
  44 # define __memrchr memrchr
  45 #endif
  46
  47 /* Search no more than N bytes of S for C.  */
  48 void *
  49 __memrchr (void const *s, int c_in, size_t n)
  50 {
  51   const unsigned char *char_ptr;
  52   const unsigned long int *longword_ptr;
  53   unsigned long int longword, magic_bits, charmask;
  54   unsigned reg_char c;
  55   int i;
  56
  57   c = (unsigned char) c_in;
  58
  59   /* Handle the last few characters by reading one character at a time.
  60      Do this until CHAR_PTR is aligned on a longword boundary.  */
  61   for (char_ptr = (const unsigned char *) s + n;
  62        n > 0 && (size_t) char_ptr % sizeof longword != 0;
  63        --n)
  64     if (*--char_ptr == c)
  65       return (void *) char_ptr;
  66
  67   /* All these elucidatory comments refer to 4-byte longwords,
  68      but the theory applies equally well to any size longwords.  */
  69
  70   longword_ptr = (const unsigned long int *) char_ptr;
  71
  72   /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
  73      the "holes."  Note that there is a hole just to the left of
  74      each byte, with an extra at the end:
  75
  76      bits:  01111110 11111110 11111110 11111111
  77      bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
  78
  79      The 1-bits make sure that carries propagate to the next 0-bit.
  80      The 0-bits provide holes for carries to fall into.  */
  81
  82   /* Set MAGIC_BITS to be this pattern of 1 and 0 bits.
  83      Set CHARMASK to be a longword, each of whose bytes is C.  */
  84
  85   magic_bits = 0xfefefefe;
  86   charmask = c | (c << 8);
  87   charmask |= charmask << 16;
  88 #if 0xffffffffU < ULONG_MAX
  89   magic_bits |= magic_bits << 32;
  90   charmask |= charmask << 32;
  91   if (8 < sizeof longword)
  92     for (i = 64; i < sizeof longword * 8; i *= 2)
  93       {
  94         magic_bits |= magic_bits << i;
  95         charmask |= charmask << i;
  96       }
  97 #endif
  98   magic_bits = (ULONG_MAX >> 1) & (magic_bits | 1);
  99
 100   /* Instead of the traditional loop which tests each character,
 101      we will test a longword at a time.  The tricky part is testing
 102      if *any of the four* bytes in the longword in question are zero.  */
 103   while (n >= sizeof longword)
 104     {
 105       /* We tentatively exit the loop if adding MAGIC_BITS to
 106          LONGWORD fails to change any of the hole bits of LONGWORD.
 107
 108          1) Is this safe?  Will it catch all the zero bytes?
 109          Suppose there is a byte with all zeros.  Any carry bits
 110          propagating from its left will fall into the hole at its
 111          least significant bit and stop.  Since there will be no
 112          carry from its most significant bit, the LSB of the
 113          byte to the left will be unchanged, and the zero will be
 114          detected.
 115
 116          2) Is this worthwhile?  Will it ignore everything except
 117          zero bytes?  Suppose every byte of LONGWORD has a bit set
 118          somewhere.  There will be a carry into bit 8.  If bit 8
 119          is set, this will carry into bit 16.  If bit 8 is clear,
 120          one of bits 9-15 must be set, so there will be a carry
 121          into bit 16.  Similarly, there will be a carry into bit
 122          24.  If one of bits 24-30 is set, there will be a carry
 123          into bit 31, so all of the hole bits will be changed.
 124
 125          The one misfire occurs when bits 24-30 are clear and bit
 126          31 is set; in this case, the hole at bit 31 is not
 127          changed.  If we had access to the processor carry flag,
 128          we could close this loophole by putting the fourth hole
 129          at bit 32!
 130
 131          So it ignores everything except 128's, when they're aligned
 132          properly.
 133
 134          3) But wait!  Aren't we looking for C, not zero?
 135          Good point.  So what we do is XOR LONGWORD with a longword,
 136          each of whose bytes is C.  This turns each byte that is C
 137          into a zero.  */
 138
 139       longword = *--longword_ptr ^ charmask;
 140
 141       /* Add MAGIC_BITS to LONGWORD.  */
 142       if ((((longword + magic_bits)
 143
 144             /* Set those bits that were unchanged by the addition.  */
 145             ^ ~longword)
 146
 147            /* Look at only the hole bits.  If any of the hole bits
 148               are unchanged, most likely one of the bytes was a
 149               zero.  */
 150            & ~magic_bits) != 0)
 151         {
 152           /* Which of the bytes was C?  If none of them were, it was
 153              a misfire; continue the search.  */
 154
 155           const unsigned char *cp = (const unsigned char *) longword_ptr;
 156
 157           if (8 < sizeof longword)
 158             for (i = sizeof longword - 1; 8 <= i; i--)
 159               if (cp[i] == c)
 160                 return (void *) &cp[i];
 161           if (7 < sizeof longword && cp[7] == c)
 162             return (void *) &cp[7];
 163           if (6 < sizeof longword && cp[6] == c)
 164             return (void *) &cp[6];
 165           if (5 < sizeof longword && cp[5] == c)
 166             return (void *) &cp[5];
 167           if (4 < sizeof longword && cp[4] == c)
 168             return (void *) &cp[4];
 169           if (cp[3] == c)
 170             return (void *) &cp[3];
 171           if (cp[2] == c)
 172             return (void *) &cp[2];
 173           if (cp[1] == c)
 174             return (void *) &cp[1];
 175           if (cp[0] == c)
 176             return (void *) cp;
 177         }
 178
 179       n -= sizeof longword;
 180     }
 181
 182   char_ptr = (const unsigned char *) longword_ptr;
 183
 184   while (n-- > 0)
 185     {
 186       if (*--char_ptr == c)
 187         return (void *) char_ptr;
 188     }
 189
 190   return 0;
 191 }
 192 #ifdef weak_alias
 193 weak_alias (__memrchr, memrchr)
 194 #endif