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.)
57 #include "rpmdb/rpmdb_internal.h" /* XXX for mergesort */
59 #define ISIZE sizeof(int)
60 #define PSIZE sizeof(unsigned char *)
61 #define ICOPY_LIST(src, dst, last) \
63 *(int*)dst = *(int*)src, src += ISIZE, dst += ISIZE; \
65 #define ICOPY_ELT(src, dst, i) \
67 *(int*) dst = *(int*) src, src += ISIZE, dst += ISIZE; \
70 #define CCOPY_LIST(src, dst, last) \
74 #define CCOPY_ELT(src, dst, i) \
80 * Find the next possible pointer head. (Trickery for forcing an array
81 * to do double duty as a linked list when objects do not align with word
84 /* Assumption: PSIZE is a power of 2. */
85 #define EVAL(p) (unsigned char **) \
86 ((unsigned char *)0 + \
87 (((unsigned char *)p + PSIZE - 1 - (unsigned char *) 0) & ~(PSIZE - 1)))
89 #define swap(a, b) { \
93 tmp = *a; *a++ = *s; *s++ = tmp; \
97 #define reverse(bot, top) { \
102 tmp = *bot; *bot++ = *s; *s++ = tmp; \
109 * This is to avoid out-of-bounds addresses in sorting the
113 insertionsort(unsigned char *a, size_t n, size_t size,
114 int (*cmp) (const void *, const void *))
116 unsigned char *ai, *s, *t, *u, tmp;
119 for (ai = a+size; --n >= 1; ai += size)
120 for (t = ai; t > a; t -= size) {
129 * Optional hybrid natural/pairwise first pass. Eats up list1 in runs of
130 * increasing order, list2 in a corresponding linked list. Checks for runs
131 * when THRESHOLD/2 pairs compare with same sense. (Only used when NATURAL
132 * is defined. Otherwise simple pairwise merging is used.)
135 setup(unsigned char *list1, unsigned char *list2,
136 size_t n, size_t size, int (*cmp) (const void *, const void *))
138 int i, length, size2, tmp, sense;
139 unsigned char *f1, *f2, *s, *l2, *last, *p2;
143 insertionsort(list1, n, size, cmp);
144 *EVAL(list2) = (unsigned char*) list2 + n*size;
148 * Avoid running pointers out of bounds; limit n to evens
152 insertionsort(list1 + (n - i) * size, i, size, cmp);
153 last = list1 + size * (n - i);
154 *EVAL(list2 + (last - list1)) = list2 + n * size;
159 sense = (cmp(f1, f1 + size) > 0);
160 for (; f1 < last; sense = !sense) {
162 /* Find pairs with same sense. */
163 for (f2 = f1 + size2; f2 < last; f2 += size2) {
164 if ((cmp(f2, f2+ size) > 0) != sense)
168 if (length < THRESHOLD) { /* Pairwise merge */
170 p2 = *EVAL(p2) = f1 + size2 - list1 + list2;
172 swap (f1, f1 + size);
173 } while ((f1 += size2) < f2);
174 } else { /* Natural merge */
176 for (f2 = f1 + size2; f2 < l2; f2 += size2) {
177 if ((cmp(f2-size, f2) > 0) != sense) {
178 p2 = *EVAL(p2) = f2 - list1 + list2;
180 reverse(f1, f2-size);
185 reverse (f1, f2-size);
187 if (f2 < last || cmp(f2 - size, f2) > 0)
188 p2 = *EVAL(p2) = f2 - list1 + list2;
190 p2 = *EVAL(p2) = list2 + n*size;
193 #else /* pairwise merge only. */
194 for (f1 = list1, p2 = list2; f1 < last; f1 += size2) {
195 p2 = *EVAL(p2) = p2 + size2;
196 if (cmp (f1, f1 + size) > 0)
203 * Arguments are as for qsort.
206 mergesort(void *base, size_t nmemb, size_t size,
207 int (*cmp) (const void *, const void *))
209 register int i, sense;
211 register unsigned char *f1, *f2, *t, *b, *q, *l1, *l2;
212 register unsigned char *tp2;
213 unsigned char *list2;
214 unsigned char *list1;
215 unsigned char *p2, *p, *last, **p1;
217 if (size < PSIZE / 2) { /* Pointers must fit into 2 * size. */
227 * Stupid subtraction for the Cray.
230 if (!(size % ISIZE) && !(((char *)base - (char *)0) % ISIZE))
233 if ((list2 = malloc(nmemb * size + PSIZE)) == NULL)
237 setup(list1, list2, nmemb, size, cmp);
238 last = list2 + nmemb * size;
240 while (*EVAL(list2) != last) {
243 for (tp2 = p2 = list2; p2 != last; p1 = EVAL(l2)) {
246 f2 = l1 = list1 + (p2 - list2);
249 l2 = list1 + (p2 - list2);
250 while (f1 < l1 && f2 < l2) {
251 if ((*cmp)(f1, f2) <= 0) {
260 if (!big) { /* here i = 0 */
261 while ((b += size) < t && cmp(q, b) >sense)
267 EXPONENTIAL: for (i = size; ; i <<= 1)
268 if ((p = (b + i)) >= t) {
269 if ((p = t - size) > b &&
270 (*cmp)(q, p) <= sense)
275 } else if ((*cmp)(q, p) <= sense) {
283 i = (((t - b) / size) >> 1) * size;
284 if ((*cmp)(q, p = b + i) <= sense)
290 FASTCASE: while (i > size)
292 p = b + (i >>= 1)) <= sense)
301 ICOPY_LIST(f2, tp2, b);
302 ICOPY_ELT(f1, tp2, i);
304 CCOPY_LIST(f2, tp2, b);
305 CCOPY_ELT(f1, tp2, i);
309 ICOPY_LIST(f1, tp2, b);
310 ICOPY_ELT(f2, tp2, i);
312 CCOPY_LIST(f1, tp2, b);
313 CCOPY_ELT(f2, tp2, i);
319 ICOPY_LIST(f2, tp2, l2);
321 CCOPY_LIST(f2, tp2, l2);
322 } else if (f1 < l1) {
324 ICOPY_LIST(f1, tp2, l1);
326 CCOPY_LIST(f1, tp2, l1);
330 tp2 = list1; /* swap list1, list2 */
333 last = list2 + nmemb*size;
336 memmove(list2, list1, nmemb*size);
343 /* mergesort is implemented in System on Mac OS X */
344 #endif /* __APPLE__ */