string/rawmemchr.c

   1 /* Copyright (C) 1991,93,96,97,99,2000,2002 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 # include <stdlib.h>
  39 #else
  40 # define reg_char char
  41 #endif
  42
  43 #if defined (HAVE_LIMITS_H) || defined (_LIBC)
  44 # include <limits.h>
  45 #endif
  46
  47 #define LONG_MAX_32_BITS 2147483647
  48
  49 #ifndef LONG_MAX
  50 #define LONG_MAX LONG_MAX_32_BITS
  51 #endif
  52
  53 #include <sys/types.h>
  54
  55 #undef memchr
  56
  57
  58 /* Find the first occurrence of C in S.  */
  59 __ptr_t
  60 __rawmemchr (s, c_in)
  61      const __ptr_t s;
  62      int c_in;
  63 {
  64   const unsigned char *char_ptr;
  65   const unsigned long int *longword_ptr;
  66   unsigned long int longword, magic_bits, charmask;
  67   unsigned reg_char c;
  68
  69   c = (unsigned char) c_in;
  70
  71   /* Handle the first few characters by reading one character at a time.
  72      Do this until CHAR_PTR is aligned on a longword boundary.  */
  73   for (char_ptr = (const unsigned char *) s;
  74        ((unsigned long int) char_ptr & (sizeof (longword) - 1)) != 0;
  75        ++char_ptr)
  76     if (*char_ptr == c)
  77       return (__ptr_t) char_ptr;
  78
  79   /* All these elucidatory comments refer to 4-byte longwords,
  80      but the theory applies equally well to 8-byte longwords.  */
  81
  82   longword_ptr = (unsigned long int *) char_ptr;
  83
  84   /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
  85      the "holes."  Note that there is a hole just to the left of
  86      each byte, with an extra at the end:
  87
  88      bits:  01111110 11111110 11111110 11111111
  89      bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
  90
  91      The 1-bits make sure that carries propagate to the next 0-bit.
  92      The 0-bits provide holes for carries to fall into.  */
  93
  94   if (sizeof (longword) != 4 && sizeof (longword) != 8)
  95     abort ();
  96
  97 #if LONG_MAX <= LONG_MAX_32_BITS
  98   magic_bits = 0x7efefeff;
  99 #else
 100   magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
 101 #endif
 102
 103   /* Set up a longword, each of whose bytes is C.  */
 104   charmask = c | (c << 8);
 105   charmask |= charmask << 16;
 106 #if LONG_MAX > LONG_MAX_32_BITS
 107   charmask |= charmask << 32;
 108 #endif
 109
 110   /* Instead of the traditional loop which tests each character,
 111      we will test a longword at a time.  The tricky part is testing
 112      if *any of the four* bytes in the longword in question are zero.  */
 113   while (1)
 114     {
 115       /* We tentatively exit the loop if adding MAGIC_BITS to
 116          LONGWORD fails to change any of the hole bits of LONGWORD.
 117
 118          1) Is this safe?  Will it catch all the zero bytes?
 119          Suppose there is a byte with all zeros.  Any carry bits
 120          propagating from its left will fall into the hole at its
 121          least significant bit and stop.  Since there will be no
 122          carry from its most significant bit, the LSB of the
 123          byte to the left will be unchanged, and the zero will be
 124          detected.
 125
 126          2) Is this worthwhile?  Will it ignore everything except
 127          zero bytes?  Suppose every byte of LONGWORD has a bit set
 128          somewhere.  There will be a carry into bit 8.  If bit 8
 129          is set, this will carry into bit 16.  If bit 8 is clear,
 130          one of bits 9-15 must be set, so there will be a carry
 131          into bit 16.  Similarly, there will be a carry into bit
 132          24.  If one of bits 24-30 is set, there will be a carry
 133          into bit 31, so all of the hole bits will be changed.
 134
 135          The one misfire occurs when bits 24-30 are clear and bit
 136          31 is set; in this case, the hole at bit 31 is not
 137          changed.  If we had access to the processor carry flag,
 138          we could close this loophole by putting the fourth hole
 139          at bit 32!
 140
 141          So it ignores everything except 128's, when they're aligned
 142          properly.
 143
 144          3) But wait!  Aren't we looking for C, not zero?
 145          Good point.  So what we do is XOR LONGWORD with a longword,
 146          each of whose bytes is C.  This turns each byte that is C
 147          into a zero.  */
 148
 149       longword = *longword_ptr++ ^ charmask;
 150
 151       /* Add MAGIC_BITS to LONGWORD.  */
 152       if ((((longword + magic_bits)
 153
 154             /* Set those bits that were unchanged by the addition.  */
 155             ^ ~longword)
 156
 157            /* Look at only the hole bits.  If any of the hole bits
 158               are unchanged, most likely one of the bytes was a
 159               zero.  */
 160            & ~magic_bits) != 0)
 161         {
 162           /* Which of the bytes was C?  If none of them were, it was
 163              a misfire; continue the search.  */
 164
 165           const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
 166
 167           if (cp[0] == c)
 168             return (__ptr_t) cp;
 169           if (cp[1] == c)
 170             return (__ptr_t) &cp[1];
 171           if (cp[2] == c)
 172             return (__ptr_t) &cp[2];
 173           if (cp[3] == c)
 174             return (__ptr_t) &cp[3];
 175 #if LONG_MAX > 2147483647
 176           if (cp[4] == c)
 177             return (__ptr_t) &cp[4];
 178           if (cp[5] == c)
 179             return (__ptr_t) &cp[5];
 180           if (cp[6] == c)
 181             return (__ptr_t) &cp[6];
 182           if (cp[7] == c)
 183             return (__ptr_t) &cp[7];
 184 #endif
 185         }
 186     }
 187 }
 188 libc_hidden_def (__rawmemchr)
 189 weak_alias (__rawmemchr, rawmemchr)