string/memchr.c

   1 /* Copyright (C) 1991,93,96,97,99,2000,2003 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, write to the Free
  21    Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
  22    02111-1307 USA.  */
  23
  24 #ifdef HAVE_CONFIG_H
  25 #include <config.h>
  26 #endif
  27
  28 #undef __ptr_t
  29 #if defined (__cplusplus) || (defined (__STDC__) && __STDC__)
  30 # define __ptr_t void *
  31 #else /* Not C++ or ANSI C.  */
  32 # define __ptr_t char *
  33 #endif /* C++ or ANSI C.  */
  34
  35 #if defined _LIBC
  36 # include <string.h>
  37 # include <memcopy.h>
  38 #else
  39 # define reg_char char
  40 #endif
  41
  42 #if HAVE_STDLIB_H || defined _LIBC
  43 # include <stdlib.h>
  44 #endif
  45
  46 #if HAVE_LIMITS_H || defined _LIBC
  47 # include <limits.h>
  48 #endif
  49
  50 #define LONG_MAX_32_BITS 2147483647
  51
  52 #ifndef LONG_MAX
  53 #define LONG_MAX LONG_MAX_32_BITS
  54 #endif
  55
  56 #include <sys/types.h>
  57 #if HAVE_BP_SYM_H || defined _LIBC
  58 #include <bp-sym.h>
  59 #else
  60 # define BP_SYM(sym) sym
  61 #endif
  62
  63 #undef memchr
  64 #undef __memchr
  65
  66 /* Search no more than N bytes of S for C.  */
  67 __ptr_t
  68 __memchr (s, c_in, n)
  69      const __ptr_t s;
  70      int c_in;
  71      size_t n;
  72 {
  73   const unsigned char *char_ptr;
  74   const unsigned long int *longword_ptr;
  75   unsigned long int longword, magic_bits, charmask;
  76   unsigned reg_char c;
  77
  78   c = (unsigned char) c_in;
  79
  80   /* Handle the first few characters by reading one character at a time.
  81      Do this until CHAR_PTR is aligned on a longword boundary.  */
  82   for (char_ptr = (const unsigned char *) s;
  83        n > 0 && ((unsigned long int) char_ptr
  84                  & (sizeof (longword) - 1)) != 0;
  85        --n, ++char_ptr)
  86     if (*char_ptr == c)
  87       return (__ptr_t) char_ptr;
  88
  89   /* All these elucidatory comments refer to 4-byte longwords,
  90      but the theory applies equally well to 8-byte longwords.  */
  91
  92   longword_ptr = (unsigned long int *) char_ptr;
  93
  94   /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
  95      the "holes."  Note that there is a hole just to the left of
  96      each byte, with an extra at the end:
  97
  98      bits:  01111110 11111110 11111110 11111111
  99      bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
 100
 101      The 1-bits make sure that carries propagate to the next 0-bit.
 102      The 0-bits provide holes for carries to fall into.  */
 103
 104   if (sizeof (longword) != 4 && sizeof (longword) != 8)
 105     abort ();
 106
 107 #if LONG_MAX <= LONG_MAX_32_BITS
 108   magic_bits = 0x7efefeff;
 109 #else
 110   magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
 111 #endif
 112
 113   /* Set up a longword, each of whose bytes is C.  */
 114   charmask = c | (c << 8);
 115   charmask |= charmask << 16;
 116 #if LONG_MAX > LONG_MAX_32_BITS
 117   charmask |= charmask << 32;
 118 #endif
 119
 120   /* Instead of the traditional loop which tests each character,
 121      we will test a longword at a time.  The tricky part is testing
 122      if *any of the four* bytes in the longword in question are zero.  */
 123   while (n >= sizeof (longword))
 124     {
 125       /* We tentatively exit the loop if adding MAGIC_BITS to
 126          LONGWORD fails to change any of the hole bits of LONGWORD.
 127
 128          1) Is this safe?  Will it catch all the zero bytes?
 129          Suppose there is a byte with all zeros.  Any carry bits
 130          propagating from its left will fall into the hole at its
 131          least significant bit and stop.  Since there will be no
 132          carry from its most significant bit, the LSB of the
 133          byte to the left will be unchanged, and the zero will be
 134          detected.
 135
 136          2) Is this worthwhile?  Will it ignore everything except
 137          zero bytes?  Suppose every byte of LONGWORD has a bit set
 138          somewhere.  There will be a carry into bit 8.  If bit 8
 139          is set, this will carry into bit 16.  If bit 8 is clear,
 140          one of bits 9-15 must be set, so there will be a carry
 141          into bit 16.  Similarly, there will be a carry into bit
 142          24.  If one of bits 24-30 is set, there will be a carry
 143          into bit 31, so all of the hole bits will be changed.
 144
 145          The one misfire occurs when bits 24-30 are clear and bit
 146          31 is set; in this case, the hole at bit 31 is not
 147          changed.  If we had access to the processor carry flag,
 148          we could close this loophole by putting the fourth hole
 149          at bit 32!
 150
 151          So it ignores everything except 128's, when they're aligned
 152          properly.
 153
 154          3) But wait!  Aren't we looking for C, not zero?
 155          Good point.  So what we do is XOR LONGWORD with a longword,
 156          each of whose bytes is C.  This turns each byte that is C
 157          into a zero.  */
 158
 159       longword = *longword_ptr++ ^ charmask;
 160
 161       /* Add MAGIC_BITS to LONGWORD.  */
 162       if ((((longword + magic_bits)
 163
 164             /* Set those bits that were unchanged by the addition.  */
 165             ^ ~longword)
 166
 167            /* Look at only the hole bits.  If any of the hole bits
 168               are unchanged, most likely one of the bytes was a
 169               zero.  */
 170            & ~magic_bits) != 0)
 171         {
 172           /* Which of the bytes was C?  If none of them were, it was
 173              a misfire; continue the search.  */
 174
 175           const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
 176
 177           if (cp[0] == c)
 178             return (__ptr_t) cp;
 179           if (cp[1] == c)
 180             return (__ptr_t) &cp[1];
 181           if (cp[2] == c)
 182             return (__ptr_t) &cp[2];
 183           if (cp[3] == c)
 184             return (__ptr_t) &cp[3];
 185 #if LONG_MAX > 2147483647
 186           if (cp[4] == c)
 187             return (__ptr_t) &cp[4];
 188           if (cp[5] == c)
 189             return (__ptr_t) &cp[5];
 190           if (cp[6] == c)
 191             return (__ptr_t) &cp[6];
 192           if (cp[7] == c)
 193             return (__ptr_t) &cp[7];
 194 #endif
 195         }
 196
 197       n -= sizeof (longword);
 198     }
 199
 200   char_ptr = (const unsigned char *) longword_ptr;
 201
 202   while (n-- > 0)
 203     {
 204       if (*char_ptr == c)
 205         return (__ptr_t) char_ptr;
 206       else
 207         ++char_ptr;
 208     }
 209
 210   return 0;
 211 }
 212 #ifdef weak_alias
 213 weak_alias (__memchr, BP_SYM (memchr))
 214 #endif
 215 libc_hidden_builtin_def (memchr)