1 | /* Copyright (C) 1991, 1993, 1996, 1997 Free Software Foundation, Inc.
|
---|
2 | Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
|
---|
3 | with help from Dan Sahlin (dan@sics.se) and
|
---|
4 | commentary by Jim Blandy (jimb@ai.mit.edu);
|
---|
5 | adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
|
---|
6 | and implemented by Roland McGrath (roland@ai.mit.edu).
|
---|
7 |
|
---|
8 | NOTE: The canonical source of this file is maintained with the GNU C Library.
|
---|
9 | Bugs can be reported to bug-glibc@gnu.org.
|
---|
10 |
|
---|
11 | This program is free software; you can redistribute it and/or modify it
|
---|
12 | under the terms of the GNU General Public License as published by the
|
---|
13 | Free Software Foundation; either version 2, or (at your option) any
|
---|
14 | later version.
|
---|
15 |
|
---|
16 | This program is distributed in the hope that it will be useful,
|
---|
17 | but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
18 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
19 | GNU General Public License for more details.
|
---|
20 |
|
---|
21 | You should have received a copy of the GNU General Public License
|
---|
22 | along with this program; if not, write to the Free Software
|
---|
23 | Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301,
|
---|
24 | USA. */
|
---|
25 |
|
---|
26 | #ifdef HAVE_CONFIG_H
|
---|
27 | #include <config.h>
|
---|
28 | #endif
|
---|
29 |
|
---|
30 | #undef __ptr_t
|
---|
31 | #if defined (__cplusplus) || (defined (__STDC__) && __STDC__)
|
---|
32 | # define __ptr_t void *
|
---|
33 | #else /* Not C++ or ANSI C. */
|
---|
34 | # define __ptr_t char *
|
---|
35 | #endif /* C++ or ANSI C. */
|
---|
36 |
|
---|
37 | #if defined (_LIBC)
|
---|
38 | # include <string.h>
|
---|
39 | #endif
|
---|
40 |
|
---|
41 | #if defined (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 |
|
---|
53 | #undef memchr
|
---|
54 |
|
---|
55 |
|
---|
56 | /* Search no more than N bytes of S for C. */
|
---|
57 | __ptr_t
|
---|
58 | memchr (s, c, n)
|
---|
59 | const __ptr_t s;
|
---|
60 | int c;
|
---|
61 | size_t n;
|
---|
62 | {
|
---|
63 | const unsigned char *char_ptr;
|
---|
64 | const unsigned long int *longword_ptr;
|
---|
65 | unsigned long int longword, magic_bits, charmask;
|
---|
66 |
|
---|
67 | c = (unsigned char) c;
|
---|
68 |
|
---|
69 | /* Handle the first few characters by reading one character at a time.
|
---|
70 | Do this until CHAR_PTR is aligned on a longword boundary. */
|
---|
71 | for (char_ptr = (const unsigned char *) s;
|
---|
72 | n > 0 && ((unsigned long int) char_ptr
|
---|
73 | & (sizeof (longword) - 1)) != 0;
|
---|
74 | --n, ++char_ptr)
|
---|
75 | if (*char_ptr == c)
|
---|
76 | return (__ptr_t) char_ptr;
|
---|
77 |
|
---|
78 | /* All these elucidatory comments refer to 4-byte longwords,
|
---|
79 | but the theory applies equally well to 8-byte longwords. */
|
---|
80 |
|
---|
81 | longword_ptr = (unsigned long int *) char_ptr;
|
---|
82 |
|
---|
83 | /* Bits 31, 24, 16, and 8 of this number are zero. Call these bits
|
---|
84 | the "holes." Note that there is a hole just to the left of
|
---|
85 | each byte, with an extra at the end:
|
---|
86 |
|
---|
87 | bits: 01111110 11111110 11111110 11111111
|
---|
88 | bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
|
---|
89 |
|
---|
90 | The 1-bits make sure that carries propagate to the next 0-bit.
|
---|
91 | The 0-bits provide holes for carries to fall into. */
|
---|
92 |
|
---|
93 | if (sizeof (longword) != 4 && sizeof (longword) != 8)
|
---|
94 | abort ();
|
---|
95 |
|
---|
96 | #if LONG_MAX <= LONG_MAX_32_BITS
|
---|
97 | magic_bits = 0x7efefeff;
|
---|
98 | #else
|
---|
99 | magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
|
---|
100 | #endif
|
---|
101 |
|
---|
102 | /* Set up a longword, each of whose bytes is C. */
|
---|
103 | charmask = c | (c << 8);
|
---|
104 | charmask |= charmask << 16;
|
---|
105 | #if LONG_MAX > LONG_MAX_32_BITS
|
---|
106 | charmask |= charmask << 32;
|
---|
107 | #endif
|
---|
108 |
|
---|
109 | /* Instead of the traditional loop which tests each character,
|
---|
110 | we will test a longword at a time. The tricky part is testing
|
---|
111 | if *any of the four* bytes in the longword in question are zero. */
|
---|
112 | while (n >= sizeof (longword))
|
---|
113 | {
|
---|
114 | /* We tentatively exit the loop if adding MAGIC_BITS to
|
---|
115 | LONGWORD fails to change any of the hole bits of LONGWORD.
|
---|
116 |
|
---|
117 | 1) Is this safe? Will it catch all the zero bytes?
|
---|
118 | Suppose there is a byte with all zeros. Any carry bits
|
---|
119 | propagating from its left will fall into the hole at its
|
---|
120 | least significant bit and stop. Since there will be no
|
---|
121 | carry from its most significant bit, the LSB of the
|
---|
122 | byte to the left will be unchanged, and the zero will be
|
---|
123 | detected.
|
---|
124 |
|
---|
125 | 2) Is this worthwhile? Will it ignore everything except
|
---|
126 | zero bytes? Suppose every byte of LONGWORD has a bit set
|
---|
127 | somewhere. There will be a carry into bit 8. If bit 8
|
---|
128 | is set, this will carry into bit 16. If bit 8 is clear,
|
---|
129 | one of bits 9-15 must be set, so there will be a carry
|
---|
130 | into bit 16. Similarly, there will be a carry into bit
|
---|
131 | 24. If one of bits 24-30 is set, there will be a carry
|
---|
132 | into bit 31, so all of the hole bits will be changed.
|
---|
133 |
|
---|
134 | The one misfire occurs when bits 24-30 are clear and bit
|
---|
135 | 31 is set; in this case, the hole at bit 31 is not
|
---|
136 | changed. If we had access to the processor carry flag,
|
---|
137 | we could close this loophole by putting the fourth hole
|
---|
138 | at bit 32!
|
---|
139 |
|
---|
140 | So it ignores everything except 128's, when they're aligned
|
---|
141 | properly.
|
---|
142 |
|
---|
143 | 3) But wait! Aren't we looking for C, not zero?
|
---|
144 | Good point. So what we do is XOR LONGWORD with a longword,
|
---|
145 | each of whose bytes is C. This turns each byte that is C
|
---|
146 | into a zero. */
|
---|
147 |
|
---|
148 | longword = *longword_ptr++ ^ charmask;
|
---|
149 |
|
---|
150 | /* Add MAGIC_BITS to LONGWORD. */
|
---|
151 | if ((((longword + magic_bits)
|
---|
152 |
|
---|
153 | /* Set those bits that were unchanged by the addition. */
|
---|
154 | ^ ~longword)
|
---|
155 |
|
---|
156 | /* Look at only the hole bits. If any of the hole bits
|
---|
157 | are unchanged, most likely one of the bytes was a
|
---|
158 | zero. */
|
---|
159 | & ~magic_bits) != 0)
|
---|
160 | {
|
---|
161 | /* Which of the bytes was C? If none of them were, it was
|
---|
162 | a misfire; continue the search. */
|
---|
163 |
|
---|
164 | const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
|
---|
165 |
|
---|
166 | if (cp[0] == c)
|
---|
167 | return (__ptr_t) cp;
|
---|
168 | if (cp[1] == c)
|
---|
169 | return (__ptr_t) &cp[1];
|
---|
170 | if (cp[2] == c)
|
---|
171 | return (__ptr_t) &cp[2];
|
---|
172 | if (cp[3] == c)
|
---|
173 | return (__ptr_t) &cp[3];
|
---|
174 | #if LONG_MAX > 2147483647
|
---|
175 | if (cp[4] == c)
|
---|
176 | return (__ptr_t) &cp[4];
|
---|
177 | if (cp[5] == c)
|
---|
178 | return (__ptr_t) &cp[5];
|
---|
179 | if (cp[6] == c)
|
---|
180 | return (__ptr_t) &cp[6];
|
---|
181 | if (cp[7] == c)
|
---|
182 | return (__ptr_t) &cp[7];
|
---|
183 | #endif
|
---|
184 | }
|
---|
185 |
|
---|
186 | n -= sizeof (longword);
|
---|
187 | }
|
---|
188 |
|
---|
189 | char_ptr = (const unsigned char *) longword_ptr;
|
---|
190 |
|
---|
191 | while (n-- > 0)
|
---|
192 | {
|
---|
193 | if (*char_ptr == c)
|
---|
194 | return (__ptr_t) char_ptr;
|
---|
195 | else
|
---|
196 | ++char_ptr;
|
---|
197 | }
|
---|
198 |
|
---|
199 | return 0;
|
---|
200 | }
|
---|