1 /*- 2 * Copyright (c) 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Ronnie Kon at Mindcraft Inc., Kevin Lew and Elmer Yglesias. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include <sys/types.h> 34 #include <errno.h> 35 #include <stdlib.h> 36 37 /* 38 * Swap two areas of size number of bytes. Although qsort(3) permits random 39 * blocks of memory to be sorted, sorting pointers is almost certainly the 40 * common case (and, were it not, could easily be made so). Regardless, it 41 * isn't worth optimizing; the SWAP's get sped up by the cache, and pointer 42 * arithmetic gets lost in the time required for comparison function calls. 43 */ 44 #define SWAP(a, b, count, size, tmp) { \ 45 count = size; \ 46 do { \ 47 tmp = *a; \ 48 *a++ = *b; \ 49 *b++ = tmp; \ 50 } while (--count); \ 51 } 52 53 /* Copy one block of size size to another. */ 54 #define COPY(a, b, count, size, tmp1, tmp2) { \ 55 count = size; \ 56 tmp1 = a; \ 57 tmp2 = b; \ 58 do { \ 59 *tmp1++ = *tmp2++; \ 60 } while (--count); \ 61 } 62 63 /* 64 * Build the list into a heap, where a heap is defined such that for 65 * the records K1 ... KN, Kj/2 >= Kj for 1 <= j/2 <= j <= N. 66 * 67 * There are two cases. If j == nmemb, select largest of Ki and Kj. If 68 * j < nmemb, select largest of Ki, Kj and Kj+1. 69 */ 70 #define CREATE(initval, nmemb, par_i, child_i, par, child, size, count, tmp) { \ 71 for (par_i = initval; (child_i = par_i * 2) <= nmemb; \ 72 par_i = child_i) { \ 73 child = base + child_i * size; \ 74 if (child_i < nmemb && compar(child, child + size) < 0) { \ 75 child += size; \ 76 ++child_i; \ 77 } \ 78 par = base + par_i * size; \ 79 if (compar(child, par) <= 0) \ 80 break; \ 81 SWAP(par, child, count, size, tmp); \ 82 } \ 83 } 84 85 /* 86 * Select the top of the heap and 'heapify'. Since by far the most expensive 87 * action is the call to the compar function, a considerable optimization 88 * in the average case can be achieved due to the fact that k, the displaced 89 * element, is usually quite small, so it would be preferable to first 90 * heapify, always maintaining the invariant that the larger child is copied 91 * over its parent's record. 92 * 93 * Then, starting from the *bottom* of the heap, finding k's correct place, 94 * again maintaining the invariant. As a result of the invariant no element 95 * is 'lost' when k is assigned its correct place in the heap. 96 * 97 * The time savings from this optimization are on the order of 15-20% for the 98 * average case. See Knuth, Vol. 3, page 158, problem 18. 99 * 100 * XXX Don't break the #define SELECT line, below. Reiser cpp gets upset. 101 */ 102 #define SELECT(par_i, child_i, nmemb, par, child, size, k, count, tmp1, tmp2) { \ 103 for (par_i = 1; (child_i = par_i * 2) <= nmemb; par_i = child_i) { \ 104 child = base + child_i * size; \ 105 if (child_i < nmemb && compar(child, child + size) < 0) { \ 106 child += size; \ 107 ++child_i; \ 108 } \ 109 par = base + par_i * size; \ 110 COPY(par, child, count, size, tmp1, tmp2); \ 111 } \ 112 for (;;) { \ 113 child_i = par_i; \ 114 par_i = child_i / 2; \ 115 child = base + child_i * size; \ 116 par = base + par_i * size; \ 117 if (child_i == 1 || compar(k, par) < 0) { \ 118 COPY(child, k, count, size, tmp1, tmp2); \ 119 break; \ 120 } \ 121 COPY(child, par, count, size, tmp1, tmp2); \ 122 } \ 123 } 124 125 /* 126 * Heapsort -- Knuth, Vol. 3, page 145. Runs in O (N lg N), both average 127 * and worst. While heapsort is faster than the worst case of quicksort, 128 * the BSD quicksort does median selection so that the chance of finding 129 * a data set that will trigger the worst case is nonexistent. Heapsort's 130 * only advantage over quicksort is that it requires little additional memory. 131 */ 132 int 133 heapsort(void *vbase, size_t nmemb, size_t size, 134 int (*compar)(const void *, const void *)) 135 { 136 size_t cnt, i, j, l; 137 char tmp, *tmp1, *tmp2; 138 char *base, *k, *p, *t; 139 140 if (nmemb <= 1) 141 return (0); 142 143 if (!size) { 144 errno = EINVAL; 145 return (-1); 146 } 147 148 if ((k = malloc(size)) == NULL) 149 return (-1); 150 151 /* 152 * Items are numbered from 1 to nmemb, so offset from size bytes 153 * below the starting address. 154 */ 155 base = (char *)vbase - size; 156 157 for (l = nmemb / 2 + 1; --l;) 158 CREATE(l, nmemb, i, j, t, p, size, cnt, tmp); 159 160 /* 161 * For each element of the heap, save the largest element into its 162 * final slot, save the displaced element (k), then recreate the 163 * heap. 164 */ 165 while (nmemb > 1) { 166 COPY(k, base + nmemb * size, cnt, size, tmp1, tmp2); 167 COPY(base + nmemb * size, base + size, cnt, size, tmp1, tmp2); 168 --nmemb; 169 SELECT(i, j, nmemb, t, p, size, k, cnt, tmp1, tmp2); 170 } 171 free(k); 172 return (0); 173 } 174