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).
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.
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.
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
25
# define __ptr_t void *
26
# define reg_char char
32
#define LONG_MAX_32_BITS 2147483647
35
#define LONG_MAX LONG_MAX_32_BITS
38
/* Search no more than N bytes of S for C. */
45
const unsigned char *char_ptr;
46
const unsigned long int *longword_ptr;
47
unsigned long int longword, magic_bits, charmask;
50
c = (unsigned char) c_in;
52
/* Handle the first few characters by reading one character at a time.
53
Do this until CHAR_PTR is aligned on a longword boundary. */
54
for (char_ptr = (const unsigned char *) s;
55
n > 0 && ((unsigned long int) char_ptr
56
& (sizeof (longword) - 1)) != 0;
59
return (__ptr_t) char_ptr;
61
/* All these elucidatory comments refer to 4-byte longwords,
62
but the theory applies equally well to 8-byte longwords. */
64
longword_ptr = (unsigned long int *) char_ptr;
66
/* Bits 31, 24, 16, and 8 of this number are zero. Call these bits
67
the "holes." Note that there is a hole just to the left of
68
each byte, with an extra at the end:
70
bits: 01111110 11111110 11111110 11111111
71
bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
73
The 1-bits make sure that carries propagate to the next 0-bit.
74
The 0-bits provide holes for carries to fall into. */
76
if (sizeof (longword) != 4 && sizeof (longword) != 8)
79
#if LONG_MAX <= LONG_MAX_32_BITS
80
magic_bits = 0x7efefeff;
82
magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
85
/* Set up a longword, each of whose bytes is C. */
86
charmask = c | (c << 8);
87
charmask |= charmask << 16;
88
#if LONG_MAX > LONG_MAX_32_BITS
89
charmask |= charmask << 32;
92
/* Instead of the traditional loop which tests each character,
93
we will test a longword at a time. The tricky part is testing
94
if *any of the four* bytes in the longword in question are zero. */
95
while (n >= sizeof (longword))
97
/* We tentatively exit the loop if adding MAGIC_BITS to
98
LONGWORD fails to change any of the hole bits of LONGWORD.
100
1) Is this safe? Will it catch all the zero bytes?
101
Suppose there is a byte with all zeros. Any carry bits
102
propagating from its left will fall into the hole at its
103
least significant bit and stop. Since there will be no
104
carry from its most significant bit, the LSB of the
105
byte to the left will be unchanged, and the zero will be
108
2) Is this worthwhile? Will it ignore everything except
109
zero bytes? Suppose every byte of LONGWORD has a bit set
110
somewhere. There will be a carry into bit 8. If bit 8
111
is set, this will carry into bit 16. If bit 8 is clear,
112
one of bits 9-15 must be set, so there will be a carry
113
into bit 16. Similarly, there will be a carry into bit
114
24. If one of bits 24-30 is set, there will be a carry
115
into bit 31, so all of the hole bits will be changed.
117
The one misfire occurs when bits 24-30 are clear and bit
118
31 is set; in this case, the hole at bit 31 is not
119
changed. If we had access to the processor carry flag,
120
we could close this loophole by putting the fourth hole
123
So it ignores everything except 128's, when they're aligned
126
3) But wait! Aren't we looking for C, not zero?
127
Good point. So what we do is XOR LONGWORD with a longword,
128
each of whose bytes is C. This turns each byte that is C
131
longword = *longword_ptr++ ^ charmask;
133
/* Add MAGIC_BITS to LONGWORD. */
134
if ((((longword + magic_bits)
136
/* Set those bits that were unchanged by the addition. */
139
/* Look at only the hole bits. If any of the hole bits
140
are unchanged, most likely one of the bytes was a
144
/* Which of the bytes was C? If none of them were, it was
145
a misfire; continue the search. */
147
const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
152
return (__ptr_t) &cp[1];
154
return (__ptr_t) &cp[2];
156
return (__ptr_t) &cp[3];
157
#if LONG_MAX > 2147483647
159
return (__ptr_t) &cp[4];
161
return (__ptr_t) &cp[5];
163
return (__ptr_t) &cp[6];
165
return (__ptr_t) &cp[7];
169
n -= sizeof (longword);
172
char_ptr = (const unsigned char *) longword_ptr;
177
return (__ptr_t) char_ptr;