3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 #if defined(LIBC_SCCS) && !defined(lint)
39 static char sccsid[] = "@(#)merge.c 8.2 (Berkeley) 2/14/94";
40 #endif /* LIBC_SCCS and not lint */
43 * Hybrid exponential search/linear search merge sort with hybrid
44 * natural/pairwise first pass. Requires about .3% more comparisons
45 * for random data than LSMS with pairwise first pass alone.
46 * It works for objects as small as two bytes.
50 #define THRESHOLD 16 /* Best choice for natural merge cut-off. */
52 /* #define NATURAL to get hybrid natural merge.
53 * (The default is pairwise merging.)
60 #include "lib/rpmdb_internal.h" /* XXX for mergesort */
62 #define ISIZE sizeof(int)
63 #define PSIZE sizeof(unsigned char *)
64 #define ICOPY_LIST(src, dst, last) \
66 *(int*)dst = *(int*)src, src += ISIZE, dst += ISIZE; \
68 #define ICOPY_ELT(src, dst, i) \
70 *(int*) dst = *(int*) src, src += ISIZE, dst += ISIZE; \
73 #define CCOPY_LIST(src, dst, last) \
77 #define CCOPY_ELT(src, dst, i) \
83 * Find the next possible pointer head. (Trickery for forcing an array
84 * to do double duty as a linked list when objects do not align with word
87 /* Assumption: PSIZE is a power of 2. */
88 #define EVAL(p) (unsigned char **) \
89 ((unsigned char *)0 + \
90 (((unsigned char *)p + PSIZE - 1 - (unsigned char *) 0) & ~(PSIZE - 1)))
92 #define swap(a, b) { \
96 tmp = *a; *a++ = *s; *s++ = tmp; \
100 #define reverse(bot, top) { \
105 tmp = *bot; *bot++ = *s; *s++ = tmp; \
112 * This is to avoid out-of-bounds addresses in sorting the
116 insertionsort(unsigned char *a, size_t n, size_t size,
117 int (*cmp) (const void *, const void *))
119 unsigned char *ai, *s, *t, *u, tmp;
122 for (ai = a+size; --n >= 1; ai += size)
123 for (t = ai; t > a; t -= size) {
132 * Optional hybrid natural/pairwise first pass. Eats up list1 in runs of
133 * increasing order, list2 in a corresponding linked list. Checks for runs
134 * when THRESHOLD/2 pairs compare with same sense. (Only used when NATURAL
135 * is defined. Otherwise simple pairwise merging is used.)
138 setup(unsigned char *list1, unsigned char *list2,
139 size_t n, size_t size, int (*cmp) (const void *, const void *))
141 int i, length, size2, tmp, sense;
142 unsigned char *f1, *f2, *s, *l2, *last, *p2;
146 insertionsort(list1, n, size, cmp);
147 *EVAL(list2) = (unsigned char*) list2 + n*size;
151 * Avoid running pointers out of bounds; limit n to evens
155 insertionsort(list1 + (n - i) * size, i, size, cmp);
156 last = list1 + size * (n - i);
157 *EVAL(list2 + (last - list1)) = list2 + n * size;
162 sense = (cmp(f1, f1 + size) > 0);
163 for (; f1 < last; sense = !sense) {
165 /* Find pairs with same sense. */
166 for (f2 = f1 + size2; f2 < last; f2 += size2) {
167 if ((cmp(f2, f2+ size) > 0) != sense)
171 if (length < THRESHOLD) { /* Pairwise merge */
173 p2 = *EVAL(p2) = f1 + size2 - list1 + list2;
175 swap (f1, f1 + size);
176 } while ((f1 += size2) < f2);
177 } else { /* Natural merge */
179 for (f2 = f1 + size2; f2 < l2; f2 += size2) {
180 if ((cmp(f2-size, f2) > 0) != sense) {
181 p2 = *EVAL(p2) = f2 - list1 + list2;
183 reverse(f1, f2-size);
188 reverse (f1, f2-size);
190 if (f2 < last || cmp(f2 - size, f2) > 0)
191 p2 = *EVAL(p2) = f2 - list1 + list2;
193 p2 = *EVAL(p2) = list2 + n*size;
196 #else /* pairwise merge only. */
197 for (f1 = list1, p2 = list2; f1 < last; f1 += size2) {
198 p2 = *EVAL(p2) = p2 + size2;
199 if (cmp (f1, f1 + size) > 0)
206 * Arguments are as for qsort.
209 mergesort(void *base, size_t nmemb, size_t size,
210 int (*cmp) (const void *, const void *))
212 register int i, sense;
214 register unsigned char *f1, *f2, *t, *b, *q, *l1, *l2;
215 register unsigned char *tp2;
216 unsigned char *list2;
217 unsigned char *list1;
218 unsigned char *p2, *p, *last, **p1;
220 if (size < PSIZE / 2) { /* Pointers must fit into 2 * size. */
230 * Stupid subtraction for the Cray.
233 if (!(size % ISIZE) && !(((char *)base - (char *)0) % ISIZE))
236 if ((list2 = malloc(nmemb * size + PSIZE)) == NULL)
240 setup(list1, list2, nmemb, size, cmp);
241 last = list2 + nmemb * size;
243 while (*EVAL(list2) != last) {
246 for (tp2 = p2 = list2; p2 != last; p1 = EVAL(l2)) {
249 f2 = l1 = list1 + (p2 - list2);
252 l2 = list1 + (p2 - list2);
253 while (f1 < l1 && f2 < l2) {
254 if ((*cmp)(f1, f2) <= 0) {
263 if (!big) { /* here i = 0 */
264 while ((b += size) < t && cmp(q, b) >sense)
270 EXPONENTIAL: for (i = size; ; i <<= 1)
271 if ((p = (b + i)) >= t) {
272 if ((p = t - size) > b &&
273 (*cmp)(q, p) <= sense)
278 } else if ((*cmp)(q, p) <= sense) {
286 i = (((t - b) / size) >> 1) * size;
287 if ((*cmp)(q, p = b + i) <= sense)
293 FASTCASE: while (i > size)
295 p = b + (i >>= 1)) <= sense)
304 ICOPY_LIST(f2, tp2, b);
305 ICOPY_ELT(f1, tp2, i);
307 CCOPY_LIST(f2, tp2, b);
308 CCOPY_ELT(f1, tp2, i);
312 ICOPY_LIST(f1, tp2, b);
313 ICOPY_ELT(f2, tp2, i);
315 CCOPY_LIST(f1, tp2, b);
316 CCOPY_ELT(f2, tp2, i);
322 ICOPY_LIST(f2, tp2, l2);
324 CCOPY_LIST(f2, tp2, l2);
325 } else if (f1 < l1) {
327 ICOPY_LIST(f1, tp2, l1);
329 CCOPY_LIST(f1, tp2, l1);
333 tp2 = list1; /* swap list1, list2 */
336 last = list2 + nmemb*size;
339 memmove(list2, list1, nmemb*size);
346 /* mergesort is implemented in System on Mac OS X */
347 #endif /* __APPLE__ */