string/memchr.c

   1 /* Copyright (C) 1991,93,96,97,99,2000,2003,2012 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 #else
  34 # define reg_char char
  35 #endif
  36
  37 #if HAVE_STDLIB_H || defined _LIBC
  38 # include <stdlib.h>
  39 #endif
  40
  41 #if HAVE_LIMITS_H || defined _LIBC
  42 # include <limits.h>
  43 #endif
  44
  45 #define LONG_MAX_32_BITS 2147483647
  46
  47 #ifndef LONG_MAX
  48 #define LONG_MAX LONG_MAX_32_BITS
  49 #endif
  50
  51 #include <sys/types.h>
  52 #if HAVE_BP_SYM_H || defined _LIBC
  53 #include <bp-sym.h>
  54 #else
  55 # define BP_SYM(sym) sym
  56 #endif
  57
  58 #undef memchr
  59 #undef __memchr
  60
  61 /* Search no more than N bytes of S for C.  */
  62 __ptr_t
  63 __memchr (s, c_in, n)
  64      const __ptr_t s;
  65      int c_in;
  66      size_t n;
  67 {
  68   const unsigned char *char_ptr;
  69   const unsigned long int *longword_ptr;
  70   unsigned long int longword, magic_bits, charmask;
  71   unsigned reg_char c;
  72
  73   c = (unsigned char) c_in;
  74
  75   /* Handle the first few characters by reading one character at a time.
  76      Do this until CHAR_PTR is aligned on a longword boundary.  */
  77   for (char_ptr = (const unsigned char *) s;
  78        n > 0 && ((unsigned long int) char_ptr
  79                  & (sizeof (longword) - 1)) != 0;
  80        --n, ++char_ptr)
  81     if (*char_ptr == c)
  82       return (__ptr_t) char_ptr;
  83
  84   /* All these elucidatory comments refer to 4-byte longwords,
  85      but the theory applies equally well to 8-byte longwords.  */
  86
  87   longword_ptr = (unsigned long int *) char_ptr;
  88
  89   /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
  90      the "holes."  Note that there is a hole just to the left of
  91      each byte, with an extra at the end:
  92
  93      bits:  01111110 11111110 11111110 11111111
  94      bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
  95
  96      The 1-bits make sure that carries propagate to the next 0-bit.
  97      The 0-bits provide holes for carries to fall into.  */
  98
  99   if (sizeof (longword) != 4 && sizeof (longword) != 8)
 100     abort ();
 101
 102 #if LONG_MAX <= LONG_MAX_32_BITS
 103   magic_bits = 0x7efefeff;
 104 #else
 105   magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
 106 #endif
 107
 108   /* Set up a longword, each of whose bytes is C.  */
 109   charmask = c | (c << 8);
 110   charmask |= charmask << 16;
 111 #if LONG_MAX > LONG_MAX_32_BITS
 112   charmask |= charmask << 32;
 113 #endif
 114
 115   /* Instead of the traditional loop which tests each character,
 116      we will test a longword at a time.  The tricky part is testing
 117      if *any of the four* bytes in the longword in question are zero.  */
 118   while (n >= sizeof (longword))
 119     {
 120       /* We tentatively exit the loop if adding MAGIC_BITS to
 121          LONGWORD fails to change any of the hole bits of LONGWORD.
 122
 123          1) Is this safe?  Will it catch all the zero bytes?
 124          Suppose there is a byte with all zeros.  Any carry bits
 125          propagating from its left will fall into the hole at its
 126          least significant bit and stop.  Since there will be no
 127          carry from its most significant bit, the LSB of the
 128          byte to the left will be unchanged, and the zero will be
 129          detected.
 130
 131          2) Is this worthwhile?  Will it ignore everything except
 132          zero bytes?  Suppose every byte of LONGWORD has a bit set
 133          somewhere.  There will be a carry into bit 8.  If bit 8
 134          is set, this will carry into bit 16.  If bit 8 is clear,
 135          one of bits 9-15 must be set, so there will be a carry
 136          into bit 16.  Similarly, there will be a carry into bit
 137          24.  If one of bits 24-30 is set, there will be a carry
 138          into bit 31, so all of the hole bits will be changed.
 139
 140          The one misfire occurs when bits 24-30 are clear and bit
 141          31 is set; in this case, the hole at bit 31 is not
 142          changed.  If we had access to the processor carry flag,
 143          we could close this loophole by putting the fourth hole
 144          at bit 32!
 145
 146          So it ignores everything except 128's, when they're aligned
 147          properly.
 148
 149          3) But wait!  Aren't we looking for C, not zero?
 150          Good point.  So what we do is XOR LONGWORD with a longword,
 151          each of whose bytes is C.  This turns each byte that is C
 152          into a zero.  */
 153
 154       longword = *longword_ptr++ ^ charmask;
 155
 156       /* Add MAGIC_BITS to LONGWORD.  */
 157       if ((((longword + magic_bits)
 158
 159             /* Set those bits that were unchanged by the addition.  */
 160             ^ ~longword)
 161
 162            /* Look at only the hole bits.  If any of the hole bits
 163               are unchanged, most likely one of the bytes was a
 164               zero.  */
 165            & ~magic_bits) != 0)
 166         {
 167           /* Which of the bytes was C?  If none of them were, it was
 168              a misfire; continue the search.  */
 169
 170           const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
 171
 172           if (cp[0] == c)
 173             return (__ptr_t) cp;
 174           if (cp[1] == c)
 175             return (__ptr_t) &cp[1];
 176           if (cp[2] == c)
 177             return (__ptr_t) &cp[2];
 178           if (cp[3] == c)
 179             return (__ptr_t) &cp[3];
 180 #if LONG_MAX > 2147483647
 181           if (cp[4] == c)
 182             return (__ptr_t) &cp[4];
 183           if (cp[5] == c)
 184             return (__ptr_t) &cp[5];
 185           if (cp[6] == c)
 186             return (__ptr_t) &cp[6];
 187           if (cp[7] == c)
 188             return (__ptr_t) &cp[7];
 189 #endif
 190         }
 191
 192       n -= sizeof (longword);
 193     }
 194
 195   char_ptr = (const unsigned char *) longword_ptr;
 196
 197   while (n-- > 0)
 198     {
 199       if (*char_ptr == c)
 200         return (__ptr_t) char_ptr;
 201       else
 202         ++char_ptr;
 203     }
 204
 205   return 0;
 206 }
 207 #ifdef weak_alias
 208 weak_alias (__memchr, BP_SYM (memchr))
 209 #endif
 210 libc_hidden_builtin_def (memchr)