1 /* Copyright (C) 1991-2013 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, see
21 <http://www.gnu.org/licenses/>. */
28 #define __ptr_t void *
35 #if HAVE_STDLIB_H || defined _LIBC
39 #if HAVE_LIMITS_H || defined _LIBC
43 #define LONG_MAX_32_BITS 2147483647
46 #define LONG_MAX LONG_MAX_32_BITS
49 #include <sys/types.h>
55 # define MEMCHR __memchr
58 /* Search no more than N bytes of S for C. */
65 const unsigned char *char_ptr
;
66 const unsigned long int *longword_ptr
;
67 unsigned long int longword
, magic_bits
, charmask
;
70 c
= (unsigned char) c_in
;
72 /* Handle the first few characters by reading one character at a time.
73 Do this until CHAR_PTR is aligned on a longword boundary. */
74 for (char_ptr
= (const unsigned char *) s
;
75 n
> 0 && ((unsigned long int) char_ptr
76 & (sizeof (longword
) - 1)) != 0;
79 return (__ptr_t
) char_ptr
;
81 /* All these elucidatory comments refer to 4-byte longwords,
82 but the theory applies equally well to 8-byte longwords. */
84 longword_ptr
= (unsigned long int *) char_ptr
;
86 /* Bits 31, 24, 16, and 8 of this number are zero. Call these bits
87 the "holes." Note that there is a hole just to the left of
88 each byte, with an extra at the end:
90 bits: 01111110 11111110 11111110 11111111
91 bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
93 The 1-bits make sure that carries propagate to the next 0-bit.
94 The 0-bits provide holes for carries to fall into. */
96 if (sizeof (longword
) != 4 && sizeof (longword
) != 8)
99 #if LONG_MAX <= LONG_MAX_32_BITS
100 magic_bits
= 0x7efefeff;
102 magic_bits
= ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
105 /* Set up a longword, each of whose bytes is C. */
106 charmask
= c
| (c
<< 8);
107 charmask
|= charmask
<< 16;
108 #if LONG_MAX > LONG_MAX_32_BITS
109 charmask
|= charmask
<< 32;
112 /* Instead of the traditional loop which tests each character,
113 we will test a longword at a time. The tricky part is testing
114 if *any of the four* bytes in the longword in question are zero. */
115 while (n
>= sizeof (longword
))
117 /* We tentatively exit the loop if adding MAGIC_BITS to
118 LONGWORD fails to change any of the hole bits of LONGWORD.
120 1) Is this safe? Will it catch all the zero bytes?
121 Suppose there is a byte with all zeros. Any carry bits
122 propagating from its left will fall into the hole at its
123 least significant bit and stop. Since there will be no
124 carry from its most significant bit, the LSB of the
125 byte to the left will be unchanged, and the zero will be
128 2) Is this worthwhile? Will it ignore everything except
129 zero bytes? Suppose every byte of LONGWORD has a bit set
130 somewhere. There will be a carry into bit 8. If bit 8
131 is set, this will carry into bit 16. If bit 8 is clear,
132 one of bits 9-15 must be set, so there will be a carry
133 into bit 16. Similarly, there will be a carry into bit
134 24. If one of bits 24-30 is set, there will be a carry
135 into bit 31, so all of the hole bits will be changed.
137 The one misfire occurs when bits 24-30 are clear and bit
138 31 is set; in this case, the hole at bit 31 is not
139 changed. If we had access to the processor carry flag,
140 we could close this loophole by putting the fourth hole
143 So it ignores everything except 128's, when they're aligned
146 3) But wait! Aren't we looking for C, not zero?
147 Good point. So what we do is XOR LONGWORD with a longword,
148 each of whose bytes is C. This turns each byte that is C
151 longword
= *longword_ptr
++ ^ charmask
;
153 /* Add MAGIC_BITS to LONGWORD. */
154 if ((((longword
+ magic_bits
)
156 /* Set those bits that were unchanged by the addition. */
159 /* Look at only the hole bits. If any of the hole bits
160 are unchanged, most likely one of the bytes was a
164 /* Which of the bytes was C? If none of them were, it was
165 a misfire; continue the search. */
167 const unsigned char *cp
= (const unsigned char *) (longword_ptr
- 1);
172 return (__ptr_t
) &cp
[1];
174 return (__ptr_t
) &cp
[2];
176 return (__ptr_t
) &cp
[3];
177 #if LONG_MAX > 2147483647
179 return (__ptr_t
) &cp
[4];
181 return (__ptr_t
) &cp
[5];
183 return (__ptr_t
) &cp
[6];
185 return (__ptr_t
) &cp
[7];
189 n
-= sizeof (longword
);
192 char_ptr
= (const unsigned char *) longword_ptr
;
197 return (__ptr_t
) char_ptr
;
205 weak_alias (__memchr
, memchr
)
207 libc_hidden_builtin_def (memchr
)