1*c54f35caSApple OSS Distributions /*-
2*c54f35caSApple OSS Distributions * Copyright (c) 2009 Xin LI <[email protected]>
3*c54f35caSApple OSS Distributions * All rights reserved.
4*c54f35caSApple OSS Distributions *
5*c54f35caSApple OSS Distributions * Redistribution and use in source and binary forms, with or without
6*c54f35caSApple OSS Distributions * modification, are permitted provided that the following conditions
7*c54f35caSApple OSS Distributions * are met:
8*c54f35caSApple OSS Distributions * 1. Redistributions of source code must retain the above copyright
9*c54f35caSApple OSS Distributions * notice, this list of conditions and the following disclaimer.
10*c54f35caSApple OSS Distributions * 2. Redistributions in binary form must reproduce the above copyright
11*c54f35caSApple OSS Distributions * notice, this list of conditions and the following disclaimer in the
12*c54f35caSApple OSS Distributions * documentation and/or other materials provided with the distribution.
13*c54f35caSApple OSS Distributions *
14*c54f35caSApple OSS Distributions * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15*c54f35caSApple OSS Distributions * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16*c54f35caSApple OSS Distributions * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17*c54f35caSApple OSS Distributions * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18*c54f35caSApple OSS Distributions * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19*c54f35caSApple OSS Distributions * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20*c54f35caSApple OSS Distributions * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21*c54f35caSApple OSS Distributions * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22*c54f35caSApple OSS Distributions * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23*c54f35caSApple OSS Distributions * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24*c54f35caSApple OSS Distributions * SUCH DAMAGE.
25*c54f35caSApple OSS Distributions */
26*c54f35caSApple OSS Distributions
27*c54f35caSApple OSS Distributions #include "string.h"
28*c54f35caSApple OSS Distributions #include <limits.h>
29*c54f35caSApple OSS Distributions
30*c54f35caSApple OSS Distributions /*
31*c54f35caSApple OSS Distributions * Portable strlen() for 32-bit and 64-bit systems.
32*c54f35caSApple OSS Distributions *
33*c54f35caSApple OSS Distributions * Rationale: it is generally much more efficient to do word length
34*c54f35caSApple OSS Distributions * operations and avoid branches on modern computer systems, as
35*c54f35caSApple OSS Distributions * compared to byte-length operations with a lot of branches.
36*c54f35caSApple OSS Distributions *
37*c54f35caSApple OSS Distributions * The expression:
38*c54f35caSApple OSS Distributions *
39*c54f35caSApple OSS Distributions * ((x - 0x01....01) & ~x & 0x80....80)
40*c54f35caSApple OSS Distributions *
41*c54f35caSApple OSS Distributions * would evaluate to a non-zero value iff any of the bytes in the
42*c54f35caSApple OSS Distributions * original word is zero. However, we can further reduce ~1/3 of
43*c54f35caSApple OSS Distributions * time if we consider that strlen() usually operate on 7-bit ASCII
44*c54f35caSApple OSS Distributions * by employing the following expression, which allows false positive
45*c54f35caSApple OSS Distributions * when high bit of 1 and use the tail case to catch these case:
46*c54f35caSApple OSS Distributions *
47*c54f35caSApple OSS Distributions * ((x - 0x01....01) & 0x80....80)
48*c54f35caSApple OSS Distributions *
49*c54f35caSApple OSS Distributions * This is more than 5.2 times as fast as the raw implementation on
50*c54f35caSApple OSS Distributions * Intel T7300 under long mode for strings longer than word length.
51*c54f35caSApple OSS Distributions */
52*c54f35caSApple OSS Distributions
53*c54f35caSApple OSS Distributions /* Magic numbers for the algorithm */
54*c54f35caSApple OSS Distributions #if LONG_BIT == 32
55*c54f35caSApple OSS Distributions static const unsigned long mask01 = 0x01010101;
56*c54f35caSApple OSS Distributions static const unsigned long mask80 = 0x80808080;
57*c54f35caSApple OSS Distributions #elif LONG_BIT == 64
58*c54f35caSApple OSS Distributions static const unsigned long mask01 = 0x0101010101010101;
59*c54f35caSApple OSS Distributions static const unsigned long mask80 = 0x8080808080808080;
60*c54f35caSApple OSS Distributions #else
61*c54f35caSApple OSS Distributions #error Unsupported word size
62*c54f35caSApple OSS Distributions #endif
63*c54f35caSApple OSS Distributions
64*c54f35caSApple OSS Distributions #define LONGPTR_MASK (sizeof(long) - 1)
65*c54f35caSApple OSS Distributions
66*c54f35caSApple OSS Distributions /*
67*c54f35caSApple OSS Distributions * Helper macro to return string length if we caught the zero
68*c54f35caSApple OSS Distributions * byte.
69*c54f35caSApple OSS Distributions */
70*c54f35caSApple OSS Distributions #define testbyte(x) \
71*c54f35caSApple OSS Distributions do { \
72*c54f35caSApple OSS Distributions if (p[x] == '\0') \
73*c54f35caSApple OSS Distributions return (p - str + x); \
74*c54f35caSApple OSS Distributions } while (0)
75*c54f35caSApple OSS Distributions
76*c54f35caSApple OSS Distributions __attribute__((visibility("hidden")))
77*c54f35caSApple OSS Distributions size_t
_libkernel_strlen(const char * str)78*c54f35caSApple OSS Distributions _libkernel_strlen(const char *str)
79*c54f35caSApple OSS Distributions {
80*c54f35caSApple OSS Distributions const char *p;
81*c54f35caSApple OSS Distributions const unsigned long *lp;
82*c54f35caSApple OSS Distributions
83*c54f35caSApple OSS Distributions /* Skip the first few bytes until we have an aligned p */
84*c54f35caSApple OSS Distributions for (p = str; (uintptr_t)p & LONGPTR_MASK; p++) {
85*c54f35caSApple OSS Distributions if (*p == '\0') {
86*c54f35caSApple OSS Distributions return p - str;
87*c54f35caSApple OSS Distributions }
88*c54f35caSApple OSS Distributions }
89*c54f35caSApple OSS Distributions
90*c54f35caSApple OSS Distributions /* Scan the rest of the string using word sized operation */
91*c54f35caSApple OSS Distributions for (lp = (const unsigned long *)p;; lp++) {
92*c54f35caSApple OSS Distributions if ((*lp - mask01) & mask80) {
93*c54f35caSApple OSS Distributions p = (const char *)(lp);
94*c54f35caSApple OSS Distributions testbyte(0);
95*c54f35caSApple OSS Distributions testbyte(1);
96*c54f35caSApple OSS Distributions testbyte(2);
97*c54f35caSApple OSS Distributions testbyte(3);
98*c54f35caSApple OSS Distributions #if (LONG_BIT >= 64)
99*c54f35caSApple OSS Distributions testbyte(4);
100*c54f35caSApple OSS Distributions testbyte(5);
101*c54f35caSApple OSS Distributions testbyte(6);
102*c54f35caSApple OSS Distributions testbyte(7);
103*c54f35caSApple OSS Distributions #endif
104*c54f35caSApple OSS Distributions }
105*c54f35caSApple OSS Distributions }
106*c54f35caSApple OSS Distributions
107*c54f35caSApple OSS Distributions /* NOTREACHED */
108*c54f35caSApple OSS Distributions return 0;
109*c54f35caSApple OSS Distributions }
110