Debian Codegraph @

lib/strchrnul.c

raw
1 /* Searching in a string.
2 Copyright (C) 2003, 2007-2026 Free Software Foundation, Inc.
3
4 This file is free software: you can redistribute it and/or modify
5 it under the terms of the GNU Lesser General Public License as
6 published by the Free Software Foundation; either version 2.1 of the
7 License, or (at your option) any later version.
8
9 This file 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 Lesser General Public License for more details.
13
14 You should have received a copy of the GNU Lesser General Public License
15 along with this program. If not, see <https://www.gnu.org/licenses/>. */
16
17 #include <config.h>
18
19 /* Specification. */
20 #include <string.h>
21
22 /* Find the first occurrence of C in S or the final NUL byte. */
23 char *
24 strchrnul (const char *s, int c_in)
25 {
26 /* On 32-bit hardware, choosing longword to be a 32-bit unsigned
27 long instead of a 64-bit uintmax_t tends to give better
28 performance. On 64-bit hardware, unsigned long is generally 64
29 bits already. Change this typedef to experiment with
30 performance. */
31 typedef unsigned long int longword;
32
33 unsigned char c = (unsigned char) c_in;
34 if (!c)
35 return rawmemchr (s, 0);
36
37 const longword *longword_ptr;
38
39 {
40 const unsigned char *char_ptr;
41
42 /* Handle the first few bytes by reading one byte at a time.
43 Do this until CHAR_PTR is aligned on a longword boundary. */
44 for (char_ptr = (const unsigned char *) s;
45 (size_t) char_ptr % sizeof (longword) != 0;
46 ++char_ptr)
47 if (!*char_ptr || *char_ptr == c)
48 return (char *) char_ptr;
49
50 longword_ptr = (const longword *) char_ptr;
51 }
52
53 /* All these elucidatory comments refer to 4-byte longwords,
54 but the theory applies equally well to any size longwords. */
55 {
56 /* Compute auxiliary longword values:
57 repeated_one is a value which has a 1 in every byte.
58 repeated_c has c in every byte. */
59 longword repeated_one = 0x01010101;
60 longword repeated_c = c | (c << 8);
61 repeated_c |= repeated_c << 16;
62 if (0xffffffffU < (longword) -1)
63 {
64 repeated_one |= repeated_one << 31 << 1;
65 repeated_c |= repeated_c << 31 << 1;
66 if (8 < sizeof (longword))
67 for (size_t i = 64; i < sizeof (longword) * 8; i *= 2)
68 {
69 repeated_one |= repeated_one << i;
70 repeated_c |= repeated_c << i;
71 }
72 }
73
74 /* Instead of the traditional loop which tests each byte, we will
75 test a longword at a time. The tricky part is testing if *any of
76 the four* bytes in the longword in question are equal to NUL or
77 c. We first use an xor with repeated_c. This reduces the task
78 to testing whether *any of the four* bytes in longword1 or
79 longword2 is zero.
80
81 Let's consider longword1. We compute tmp =
82 ((longword1 - repeated_one) & ~longword1) & (repeated_one << 7).
83 That is, we perform the following operations:
84 1. Subtract repeated_one.
85 2. & ~longword1.
86 3. & a mask consisting of 0x80 in every byte.
87 Consider what happens in each byte:
88 - If a byte of longword1 is zero, step 1 and 2 transform it into 0xff,
89 and step 3 transforms it into 0x80. A carry can also be propagated
90 to more significant bytes.
91 - If a byte of longword1 is nonzero, let its lowest 1 bit be at
92 position k (0 <= k <= 7); so the lowest k bits are 0. After step 1,
93 the byte ends in a single bit of value 0 and k bits of value 1.
94 After step 2, the result is just k bits of value 1: 2^k - 1. After
95 step 3, the result is 0. And no carry is produced.
96 So, if longword1 has only non-zero bytes, tmp is zero.
97 Whereas if longword1 has a zero byte, call j the position of the least
98 significant zero byte. Then the result has a zero at positions 0, ...,
99 j-1 and a 0x80 at position j. We cannot predict the result at the more
100 significant bytes (positions j+1..3), but it does not matter since we
101 already have a non-zero bit at position 8*j+7.
102
103 The test whether any byte in longword1 or longword2 is zero is equivalent
104 to testing whether tmp1 is nonzero or tmp2 is nonzero. We can combine
105 this into a single test, whether (tmp1 | tmp2) is nonzero.
106
107 This test can read more than one byte beyond the end of a string,
108 depending on where the terminating NUL is encountered. However,
109 this is considered safe since the initialization phase ensured
110 that the read will be aligned, therefore, the read will not cross
111 page boundaries and will not cause a fault. */
112
113 while (1)
114 {
115 longword longword1 = *longword_ptr ^ repeated_c;
116 longword longword2 = *longword_ptr;
117
118 if (((((longword1 - repeated_one) & ~longword1)
119 | ((longword2 - repeated_one) & ~longword2))
120 & (repeated_one << 7)) != 0)
121 break;
122 longword_ptr++;
123 }
124 }
125
126 {
127 const unsigned char *char_ptr = (const unsigned char *) longword_ptr;
128
129 /* At this point, we know that one of the sizeof (longword) bytes
130 starting at char_ptr is == 0 or == c. On little-endian machines,
131 we could determine the first such byte without any further memory
132 accesses, just by looking at the tmp result from the last loop
133 iteration. But this does not work on big-endian machines.
134 Choose code that works in both cases. */
135
136 while (*char_ptr && (*char_ptr != c))
137 char_ptr++;
138 return (char *) char_ptr;
139 }
140 }