1 /* memrchr -- find the last occurrence of a byte in a memory block 2 3 Copyright (C) 1991, 1993, 1996, 1997, 1999, 2000, 2003, 2004, 2005 4 Free Software Foundation, Inc. 5 6 Based on strlen implementation by Torbjorn Granlund (tege@sics.se), 7 with help from Dan Sahlin (dan@sics.se) and 8 commentary by Jim Blandy (jimb@ai.mit.edu); 9 adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu), 10 and implemented by Roland McGrath (roland@ai.mit.edu). 11 12 This program is free software; you can redistribute it and/or modify 13 it under the terms of the GNU General Public License as published by 14 the Free Software Foundation; either version 2, or (at your option) 15 any later version. 16 17 This program is distributed in the hope that it will be useful, 18 but WITHOUT ANY WARRANTY; without even the implied warranty of 19 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 20 GNU General Public License for more details. 21 22 You should have received a copy of the GNU General Public License along 23 with this program; if not, write to the Free Software Foundation, 24 Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ 25 #include <sys/cdefs.h> 26 __RCSID("$NetBSD: memrchr.c,v 1.2 2016/05/17 14:00:09 christos Exp $"); 27 28 29 #ifdef HAVE_CONFIG_H 30 # include <config.h> 31 #endif 32 33 #if defined _LIBC 34 # include <string.h> 35 # include <memcopy.h> 36 #else 37 # include "memrchr.h" 38 # define reg_char char 39 #endif 40 41 #include <limits.h> 42 43 #undef __memrchr 44 #undef memrchr 45 46 #ifndef weak_alias 47 # define __memrchr memrchr 48 #endif 49 50 /* Search no more than N bytes of S for C. */ 51 void * 52 __memrchr (void const *s, int c_in, size_t n) 53 { 54 const unsigned char *char_ptr; 55 const unsigned long int *longword_ptr; 56 unsigned long int longword, magic_bits, charmask; 57 unsigned reg_char c; 58 int i; 59 60 c = (unsigned char) c_in; 61 62 /* Handle the last few characters by reading one character at a time. 63 Do this until CHAR_PTR is aligned on a longword boundary. */ 64 for (char_ptr = (const unsigned char *) s + n; 65 n > 0 && (size_t) char_ptr % sizeof longword != 0; 66 --n) 67 if (*--char_ptr == c) 68 return (void *) char_ptr; 69 70 /* All these elucidatory comments refer to 4-byte longwords, 71 but the theory applies equally well to any size longwords. */ 72 73 longword_ptr = (const unsigned long int *) char_ptr; 74 75 /* Bits 31, 24, 16, and 8 of this number are zero. Call these bits 76 the "holes." Note that there is a hole just to the left of 77 each byte, with an extra at the end: 78 79 bits: 01111110 11111110 11111110 11111111 80 bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD 81 82 The 1-bits make sure that carries propagate to the next 0-bit. 83 The 0-bits provide holes for carries to fall into. */ 84 85 /* Set MAGIC_BITS to be this pattern of 1 and 0 bits. 86 Set CHARMASK to be a longword, each of whose bytes is C. */ 87 88 magic_bits = 0xfefefefe; 89 charmask = c | (c << 8); 90 charmask |= charmask << 16; 91 #if 0xffffffffU < ULONG_MAX 92 magic_bits |= magic_bits << 32; 93 charmask |= charmask << 32; 94 if (8 < sizeof longword) 95 for (i = 64; i < sizeof longword * 8; i *= 2) 96 { 97 magic_bits |= magic_bits << i; 98 charmask |= charmask << i; 99 } 100 #endif 101 magic_bits = (ULONG_MAX >> 1) & (magic_bits | 1); 102 103 /* Instead of the traditional loop which tests each character, 104 we will test a longword at a time. The tricky part is testing 105 if *any of the four* bytes in the longword in question are zero. */ 106 while (n >= sizeof longword) 107 { 108 /* We tentatively exit the loop if adding MAGIC_BITS to 109 LONGWORD fails to change any of the hole bits of LONGWORD. 110 111 1) Is this safe? Will it catch all the zero bytes? 112 Suppose there is a byte with all zeros. Any carry bits 113 propagating from its left will fall into the hole at its 114 least significant bit and stop. Since there will be no 115 carry from its most significant bit, the LSB of the 116 byte to the left will be unchanged, and the zero will be 117 detected. 118 119 2) Is this worthwhile? Will it ignore everything except 120 zero bytes? Suppose every byte of LONGWORD has a bit set 121 somewhere. There will be a carry into bit 8. If bit 8 122 is set, this will carry into bit 16. If bit 8 is clear, 123 one of bits 9-15 must be set, so there will be a carry 124 into bit 16. Similarly, there will be a carry into bit 125 24. If one of bits 24-30 is set, there will be a carry 126 into bit 31, so all of the hole bits will be changed. 127 128 The one misfire occurs when bits 24-30 are clear and bit 129 31 is set; in this case, the hole at bit 31 is not 130 changed. If we had access to the processor carry flag, 131 we could close this loophole by putting the fourth hole 132 at bit 32! 133 134 So it ignores everything except 128's, when they're aligned 135 properly. 136 137 3) But wait! Aren't we looking for C, not zero? 138 Good point. So what we do is XOR LONGWORD with a longword, 139 each of whose bytes is C. This turns each byte that is C 140 into a zero. */ 141 142 longword = *--longword_ptr ^ charmask; 143 144 /* Add MAGIC_BITS to LONGWORD. */ 145 if ((((longword + magic_bits) 146 147 /* Set those bits that were unchanged by the addition. */ 148 ^ ~longword) 149 150 /* Look at only the hole bits. If any of the hole bits 151 are unchanged, most likely one of the bytes was a 152 zero. */ 153 & ~magic_bits) != 0) 154 { 155 /* Which of the bytes was C? If none of them were, it was 156 a misfire; continue the search. */ 157 158 const unsigned char *cp = (const unsigned char *) longword_ptr; 159 160 if (8 < sizeof longword) 161 for (i = sizeof longword - 1; 8 <= i; i--) 162 if (cp[i] == c) 163 return (void *) &cp[i]; 164 if (7 < sizeof longword && cp[7] == c) 165 return (void *) &cp[7]; 166 if (6 < sizeof longword && cp[6] == c) 167 return (void *) &cp[6]; 168 if (5 < sizeof longword && cp[5] == c) 169 return (void *) &cp[5]; 170 if (4 < sizeof longword && cp[4] == c) 171 return (void *) &cp[4]; 172 if (cp[3] == c) 173 return (void *) &cp[3]; 174 if (cp[2] == c) 175 return (void *) &cp[2]; 176 if (cp[1] == c) 177 return (void *) &cp[1]; 178 if (cp[0] == c) 179 return (void *) cp; 180 } 181 182 n -= sizeof longword; 183 } 184 185 char_ptr = (const unsigned char *) longword_ptr; 186 187 while (n-- > 0) 188 { 189 if (*--char_ptr == c) 190 return (void *) char_ptr; 191 } 192 193 return 0; 194 } 195 #ifdef weak_alias 196 weak_alias (__memrchr, memrchr) 197 #endif 198