1 /*- 2 * Copyright (c) 2009, 2010 Xin LI <delphij@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/libkern.h> 31 #include <sys/limits.h> 32 33 /* 34 * Portable strlen() for 32-bit and 64-bit systems. 35 * 36 * Rationale: it is generally much more efficient to do word length 37 * operations and avoid branches on modern computer systems, as 38 * compared to byte-length operations with a lot of branches. 39 * 40 * The expression: 41 * 42 * ((x - 0x01....01) & ~x & 0x80....80) 43 * 44 * would evaluate to a non-zero value iff any of the bytes in the 45 * original word is zero. 46 * 47 * On multi-issue processors, we can divide the above expression into: 48 * a) (x - 0x01....01) 49 * b) (~x & 0x80....80) 50 * c) a & b 51 * 52 * Where, a) and b) can be partially computed in parallel. 53 * 54 * The algorithm above is found on "Hacker's Delight" by 55 * Henry S. Warren, Jr. 56 */ 57 58 /* Magic numbers for the algorithm */ 59 #if LONG_BIT == 32 60 static const unsigned long mask01 = 0x01010101; 61 static const unsigned long mask80 = 0x80808080; 62 #elif LONG_BIT == 64 63 static const unsigned long mask01 = 0x0101010101010101; 64 static const unsigned long mask80 = 0x8080808080808080; 65 #else 66 #error Unsupported word size 67 #endif 68 69 #define LONGPTR_MASK (sizeof(long) - 1) 70 71 /* 72 * Helper macro to return string length if we caught the zero 73 * byte. 74 */ 75 #define testbyte(x) \ 76 do { \ 77 if (p[x] == '\0') \ 78 return (p - str + x); \ 79 } while (0) 80 81 size_t 82 strlen(const char *str) 83 { 84 const char *p; 85 const unsigned long *lp; 86 long va, vb; 87 88 /* 89 * Before trying the hard (unaligned byte-by-byte access) way 90 * to figure out whether there is a nul character, try to see 91 * if there is a nul character is within this accessible word 92 * first. 93 * 94 * p and (p & ~LONGPTR_MASK) must be equally accessible since 95 * they always fall in the same memory page, as long as page 96 * boundaries is integral multiple of word size. 97 */ 98 lp = (const unsigned long *)((uintptr_t)str & ~LONGPTR_MASK); 99 va = (*lp - mask01); 100 vb = ((~*lp) & mask80); 101 lp++; 102 if (va & vb) 103 /* Check if we have \0 in the first part */ 104 for (p = str; p < (const char *)lp; p++) 105 if (*p == '\0') 106 return (p - str); 107 108 /* Scan the rest of the string using word sized operation */ 109 for (; ; lp++) { 110 va = (*lp - mask01); 111 vb = ((~*lp) & mask80); 112 if (va & vb) { 113 p = (const char *)(lp); 114 testbyte(0); 115 testbyte(1); 116 testbyte(2); 117 testbyte(3); 118 #if (LONG_BIT >= 64) 119 testbyte(4); 120 testbyte(5); 121 testbyte(6); 122 testbyte(7); 123 #endif 124 } 125 } 126 127 /* NOTREACHED */ 128 return (0); 129 } 130