string/rawmemchr.c

   1 /* Copyright (C) 1991-2015 Free Software Foundation, Inc.
   2    This file is part of the GNU C Library.
   3    Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
   4    with help from Dan Sahlin (dan@sics.se) and
   5    commentary by Jim Blandy (jimb@ai.mit.edu);
   6    adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
   7    and implemented by Roland McGrath (roland@ai.mit.edu).
   8
   9    The GNU C Library is free software; you can redistribute it and/or
  10    modify it under the terms of the GNU Lesser General Public
  11    License as published by the Free Software Foundation; either
  12    version 2.1 of the License, or (at your option) any later version.
  13
  14    The GNU C Library is distributed in the hope that it will be useful,
  15    but WITHOUT ANY WARRANTY; without even the implied warranty of
  16    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17    Lesser General Public License for more details.
  18
  19    You should have received a copy of the GNU Lesser General Public
  20    License along with the GNU C Library; if not, see
  21    <http://www.gnu.org/licenses/>.  */
  22
  23 #ifdef HAVE_CONFIG_H
  24 #include <config.h>
  25 #endif
  26
  27 #undef __ptr_t
  28 #define __ptr_t void *
  29
  30 #if defined (_LIBC)
  31 # include <string.h>
  32 # include <memcopy.h>
  33 # include <stdlib.h>
  34 #endif
  35
  36 #if defined (HAVE_LIMITS_H) || defined (_LIBC)
  37 # include <limits.h>
  38 #endif
  39
  40 #define LONG_MAX_32_BITS 2147483647
  41
  42 #ifndef LONG_MAX
  43 #define LONG_MAX LONG_MAX_32_BITS
  44 #endif
  45
  46 #include <sys/types.h>
  47
  48 #undef memchr
  49
  50 #ifndef RAWMEMCHR
  51 # define RAWMEMCHR __rawmemchr
  52 #endif
  53
  54 /* Find the first occurrence of C in S.  */
  55 __ptr_t
  56 RAWMEMCHR (s, c_in)
  57      const __ptr_t s;
  58      int c_in;
  59 {
  60   const unsigned char *char_ptr;
  61   const unsigned long int *longword_ptr;
  62   unsigned long int longword, magic_bits, charmask;
  63   unsigned char c;
  64
  65   c = (unsigned char) c_in;
  66
  67   /* Handle the first few characters by reading one character at a time.
  68      Do this until CHAR_PTR is aligned on a longword boundary.  */
  69   for (char_ptr = (const unsigned char *) s;
  70        ((unsigned long int) char_ptr & (sizeof (longword) - 1)) != 0;
  71        ++char_ptr)
  72     if (*char_ptr == c)
  73       return (__ptr_t) char_ptr;
  74
  75   /* All these elucidatory comments refer to 4-byte longwords,
  76      but the theory applies equally well to 8-byte longwords.  */
  77
  78   longword_ptr = (unsigned long int *) char_ptr;
  79
  80   /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
  81      the "holes."  Note that there is a hole just to the left of
  82      each byte, with an extra at the end:
  83
  84      bits:  01111110 11111110 11111110 11111111
  85      bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
  86
  87      The 1-bits make sure that carries propagate to the next 0-bit.
  88      The 0-bits provide holes for carries to fall into.  */
  89   magic_bits = -1;
  90   magic_bits = magic_bits / 0xff * 0xfe << 1 >> 1 | 1;
  91
  92   /* Set up a longword, each of whose bytes is C.  */
  93   charmask = c | (c << 8);
  94   charmask |= charmask << 16;
  95 #if LONG_MAX > LONG_MAX_32_BITS
  96   charmask |= charmask << 32;
  97 #endif
  98
  99   /* Instead of the traditional loop which tests each character,
 100      we will test a longword at a time.  The tricky part is testing
 101      if *any of the four* bytes in the longword in question are zero.  */
 102   while (1)
 103     {
 104       /* We tentatively exit the loop if adding MAGIC_BITS to
 105          LONGWORD fails to change any of the hole bits of LONGWORD.
 106
 107          1) Is this safe?  Will it catch all the zero bytes?
 108          Suppose there is a byte with all zeros.  Any carry bits
 109          propagating from its left will fall into the hole at its
 110          least significant bit and stop.  Since there will be no
 111          carry from its most significant bit, the LSB of the
 112          byte to the left will be unchanged, and the zero will be
 113          detected.
 114
 115          2) Is this worthwhile?  Will it ignore everything except
 116          zero bytes?  Suppose every byte of LONGWORD has a bit set
 117          somewhere.  There will be a carry into bit 8.  If bit 8
 118          is set, this will carry into bit 16.  If bit 8 is clear,
 119          one of bits 9-15 must be set, so there will be a carry
 120          into bit 16.  Similarly, there will be a carry into bit
 121          24.  If one of bits 24-30 is set, there will be a carry
 122          into bit 31, so all of the hole bits will be changed.
 123
 124          The one misfire occurs when bits 24-30 are clear and bit
 125          31 is set; in this case, the hole at bit 31 is not
 126          changed.  If we had access to the processor carry flag,
 127          we could close this loophole by putting the fourth hole
 128          at bit 32!
 129
 130          So it ignores everything except 128's, when they're aligned
 131          properly.
 132
 133          3) But wait!  Aren't we looking for C, not zero?
 134          Good point.  So what we do is XOR LONGWORD with a longword,
 135          each of whose bytes is C.  This turns each byte that is C
 136          into a zero.  */
 137
 138       longword = *longword_ptr++ ^ charmask;
 139
 140       /* Add MAGIC_BITS to LONGWORD.  */
 141       if ((((longword + magic_bits)
 142
 143             /* Set those bits that were unchanged by the addition.  */
 144             ^ ~longword)
 145
 146            /* Look at only the hole bits.  If any of the hole bits
 147               are unchanged, most likely one of the bytes was a
 148               zero.  */
 149            & ~magic_bits) != 0)
 150         {
 151           /* Which of the bytes was C?  If none of them were, it was
 152              a misfire; continue the search.  */
 153
 154           const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
 155
 156           if (cp[0] == c)
 157             return (__ptr_t) cp;
 158           if (cp[1] == c)
 159             return (__ptr_t) &cp[1];
 160           if (cp[2] == c)
 161             return (__ptr_t) &cp[2];
 162           if (cp[3] == c)
 163             return (__ptr_t) &cp[3];
 164 #if LONG_MAX > 2147483647
 165           if (cp[4] == c)
 166             return (__ptr_t) &cp[4];
 167           if (cp[5] == c)
 168             return (__ptr_t) &cp[5];
 169           if (cp[6] == c)
 170             return (__ptr_t) &cp[6];
 171           if (cp[7] == c)
 172             return (__ptr_t) &cp[7];
 173 #endif
 174         }
 175     }
 176 }
 177 libc_hidden_def (__rawmemchr)
 178 weak_alias (__rawmemchr, rawmemchr)