Debian Codegraph @

src/strchrnul.c

raw
1 /* Searching in a string.
2 Copyright (C) 2003, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
3
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 3 of the License, or
7 (at your option) any later version.
8
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
13
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
16
17 #include <config.h>
18
19 #ifndef HAVE_STRCHRNUL
20
21
22 /* Specification. */
23 #include <string.h>
24 #include "rawmemchr.h"
25 #include "libfakechroot.h"
26
27 /* Find the first occurrence of C in S or the final NUL byte. */
28 LOCAL char *
29 strchrnul (const char *s, int c_in)
30 {
31 /* On 32-bit hardware, choosing longword to be a 32-bit unsigned
32 long instead of a 64-bit uintmax_t tends to give better
33 performance. On 64-bit hardware, unsigned long is generally 64
34 bits already. Change this typedef to experiment with
35 performance. */
36 typedef unsigned long int longword;
37
38 const unsigned char *char_ptr;
39 const longword *longword_ptr;
40 longword repeated_one;
41 longword repeated_c;
42 unsigned char c;
43
44 c = (unsigned char) c_in;
45 if (!c)
46 return rawmemchr (s, 0);
47
48 /* Handle the first few bytes by reading one byte at a time.
49 Do this until CHAR_PTR is aligned on a longword boundary. */
50 for (char_ptr = (const unsigned char *) s;
51 (size_t) char_ptr % sizeof (longword) != 0;
52 ++char_ptr)
53 if (!*char_ptr || *char_ptr == c)
54 return (char *) char_ptr;
55
56 longword_ptr = (const longword *) char_ptr;
57
58 /* All these elucidatory comments refer to 4-byte longwords,
59 but the theory applies equally well to any size longwords. */
60
61 /* Compute auxiliary longword values:
62 repeated_one is a value which has a 1 in every byte.
63 repeated_c has c in every byte. */
64 repeated_one = 0x01010101;
65 repeated_c = c | (c << 8);
66 repeated_c |= repeated_c << 16;
67 if (0xffffffffU < (longword) -1)
68 {
69 repeated_one |= repeated_one << 31 << 1;
70 repeated_c |= repeated_c << 31 << 1;
71 if (8 < sizeof (longword))
72 {
73 size_t i;
74
75 for (i = 64; i < sizeof (longword) * 8; i *= 2)
76 {
77 repeated_one |= repeated_one << i;
78 repeated_c |= repeated_c << i;
79 }
80 }
81 }
82
83 /* Instead of the traditional loop which tests each byte, we will
84 test a longword at a time. The tricky part is testing if *any of
85 the four* bytes in the longword in question are equal to NUL or
86 c. We first use an xor with repeated_c. This reduces the task
87 to testing whether *any of the four* bytes in longword1 or
88 longword2 is zero.
89
90 Let's consider longword1. We compute tmp =
91 ((longword1 - repeated_one) & ~longword1) & (repeated_one << 7).
92 That is, we perform the following operations:
93 1. Subtract repeated_one.
94 2. & ~longword1.
95 3. & a mask consisting of 0x80 in every byte.
96 Consider what happens in each byte:
97 - If a byte of longword1 is zero, step 1 and 2 transform it into 0xff,
98 and step 3 transforms it into 0x80. A carry can also be propagated
99 to more significant bytes.
100 - If a byte of longword1 is nonzero, let its lowest 1 bit be at
101 position k (0 <= k <= 7); so the lowest k bits are 0. After step 1,
102 the byte ends in a single bit of value 0 and k bits of value 1.
103 After step 2, the result is just k bits of value 1: 2^k - 1. After
104 step 3, the result is 0. And no carry is produced.
105 So, if longword1 has only non-zero bytes, tmp is zero.
106 Whereas if longword1 has a zero byte, call j the position of the least
107 significant zero byte. Then the result has a zero at positions 0, ...,
108 j-1 and a 0x80 at position j. We cannot predict the result at the more
109 significant bytes (positions j+1..3), but it does not matter since we
110 already have a non-zero bit at position 8*j+7.
111
112 The test whether any byte in longword1 or longword2 is zero is equivalent
113 to testing whether tmp1 is nonzero or tmp2 is nonzero. We can combine
114 this into a single test, whether (tmp1 | tmp2) is nonzero.
115
116 This test can read more than one byte beyond the end of a string,
117 depending on where the terminating NUL is encountered. However,
118 this is considered safe since the initialization phase ensured
119 that the read will be aligned, therefore, the read will not cross
120 page boundaries and will not cause a fault. */
121
122 while (1)
123 {
124 longword longword1 = *longword_ptr ^ repeated_c;
125 longword longword2 = *longword_ptr;
126
127 if (((((longword1 - repeated_one) & ~longword1)
128 | ((longword2 - repeated_one) & ~longword2))
129 & (repeated_one << 7)) != 0)
130 break;
131 longword_ptr++;
132 }
133
134 char_ptr = (const unsigned char *) longword_ptr;
135
136 /* At this point, we know that one of the sizeof (longword) bytes
137 starting at char_ptr is == 0 or == c. On little-endian machines,
138 we could determine the first such byte without any further memory
139 accesses, just by looking at the tmp result from the last loop
140 iteration. But this does not work on big-endian machines.
141 Choose code that works in both cases. */
142
143 char_ptr = (unsigned char *) longword_ptr;
144 while (*char_ptr && (*char_ptr != c))
145 char_ptr++;
146 return (char *) char_ptr;
147 }
148
149 #else
150 typedef int empty_translation_unit;
151 #endif