xref: /dflybsd-src/lib/libc/stdlib/radixsort.c (revision 07caec20a93f40352fd28eec40bb9b40199edf40)
1 /*-
2  * Copyright (c) 1990, 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  * Peter McIlroy and by Dan Bernstein at New York University,
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. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  * $DragonFly: src/lib/libc/stdlib/radixsort.c,v 1.5 2005/11/20 12:37:49 swildner Exp $
37  *
38  * @(#)radixsort.c	8.2 (Berkeley) 4/28/95
39  */
40 
41 /*
42  * Radixsort routines.
43  *
44  * Program r_sort_a() is unstable but uses O(logN) extra memory for a stack.
45  * Use radixsort(a, n, trace, endchar) for this case.
46  *
47  * For stable sorting (using N extra pointers) use sradixsort(), which calls
48  * r_sort_b().
49  *
50  * For a description of this code, see D. McIlroy, P. McIlroy, K. Bostic,
51  * "Engineering Radix Sort".
52  */
53 
54 #include <sys/types.h>
55 #include <stdlib.h>
56 #include <stddef.h>
57 #include <errno.h>
58 
59 typedef struct {
60 	const u_char **sa;
61 	int sn, si;
62 } stack;
63 
64 static inline void simplesort
65 	    (const u_char **, int, int, const u_char *, u_int);
66 static void r_sort_a (const u_char **, int, int, const u_char *, u_int);
67 static void r_sort_b (const u_char **,
68 	    const u_char **, int, int, const u_char *, u_int);
69 
70 #define	THRESHOLD	20		/* Divert to simplesort(). */
71 #define	SIZE		512		/* Default stack size. */
72 
73 #define SETUP {								\
74 	if (tab == NULL) {						\
75 		tr = tr0;						\
76 		for (c = 0; c < endch; c++)				\
77 			tr0[c] = c + 1;					\
78 		tr0[c] = 0;						\
79 		for (c++; c < 256; c++)					\
80 			tr0[c] = c;					\
81 		endch = 0;						\
82 	} else {							\
83 		endch = tab[endch];					\
84 		tr = tab;						\
85 		if (endch != 0 && endch != 255) {			\
86 			errno = EINVAL;					\
87 			return (-1);					\
88 		}							\
89 	}								\
90 }
91 
92 int
93 radixsort(const u_char **a, int n, const u_char *tab, u_int endch)
94 {
95 	const u_char *tr;
96 	int c;
97 	u_char tr0[256];
98 
99 	SETUP;
100 	r_sort_a(a, n, 0, tr, endch);
101 	return (0);
102 }
103 
104 int
105 sradixsort(const u_char **a, int n, const u_char *tab, u_int endch)
106 {
107 	const u_char *tr, **ta;
108 	int c;
109 	u_char tr0[256];
110 
111 	SETUP;
112 	if (n < THRESHOLD)
113 		simplesort(a, n, 0, tr, endch);
114 	else {
115 		if ((ta = malloc(n * sizeof(a))) == NULL)
116 			return (-1);
117 		r_sort_b(a, ta, n, 0, tr, endch);
118 		free(ta);
119 	}
120 	return (0);
121 }
122 
123 #define empty(s)	(s >= sp)
124 #define pop(a, n, i)	a = (--sp)->sa, n = sp->sn, i = sp->si
125 #define push(a, n, i)	sp->sa = a, sp->sn = n, (sp++)->si = i
126 #define swap(a, b, t)	t = a, a = b, b = t
127 
128 /* Unstable, in-place sort. */
129 static void
130 r_sort_a(const u_char **a, int n, int i, const u_char *tr, u_int endch)
131 {
132 	static int count[256], nc, bmin;
133 	int c;
134 	const u_char **ak, *r;
135 	stack s[SIZE], *sp, *sp0, *sp1, temp;
136 	int *cp, bigc;
137 	const u_char **an, *t, **aj, **top[256];
138 
139 	/* Set up stack. */
140 	sp = s;
141 	push(a, n, i);
142 	while (!empty(s)) {
143 		pop(a, n, i);
144 		if (n < THRESHOLD) {
145 			simplesort(a, n, i, tr, endch);
146 			continue;
147 		}
148 		an = a + n;
149 
150 		/* Make character histogram. */
151 		if (nc == 0) {
152 			bmin = 255;	/* First occupied bin, excluding eos. */
153 			for (ak = a; ak < an;) {
154 				c = tr[(*ak++)[i]];
155 				if (++count[c] == 1 && c != endch) {
156 					if (c < bmin)
157 						bmin = c;
158 					nc++;
159 				}
160 			}
161 			if (sp + nc > s + SIZE) {	/* Get more stack. */
162 				r_sort_a(a, n, i, tr, endch);
163 				continue;
164 			}
165 		}
166 
167 		/*
168 		 * Set top[]; push incompletely sorted bins onto stack.
169 		 * top[] = pointers to last out-of-place element in bins.
170 		 * count[] = counts of elements in bins.
171 		 * Before permuting: top[c-1] + count[c] = top[c];
172 		 * during deal: top[c] counts down to top[c-1].
173 		 */
174 		sp0 = sp1 = sp;		/* Stack position of biggest bin. */
175 		bigc = 2;		/* Size of biggest bin. */
176 		if (endch == 0)		/* Special case: set top[eos]. */
177 			top[0] = ak = a + count[0];
178 		else {
179 			ak = a;
180 			top[255] = an;
181 		}
182 		for (cp = count + bmin; nc > 0; cp++) {
183 			while (*cp == 0)	/* Find next non-empty pile. */
184 				cp++;
185 			if (*cp > 1) {
186 				if (*cp > bigc) {
187 					bigc = *cp;
188 					sp1 = sp;
189 				}
190 				push(ak, *cp, i+1);
191 			}
192 			top[cp-count] = ak += *cp;
193 			nc--;
194 		}
195 		swap(*sp0, *sp1, temp);	/* Play it safe -- biggest bin last. */
196 
197 		/*
198 		 * Permute misplacements home.  Already home: everything
199 		 * before aj, and in bin[c], items from top[c] on.
200 		 * Inner loop:
201 		 *	r = next element to put in place;
202 		 *	ak = top[r[i]] = location to put the next element.
203 		 *	aj = bottom of 1st disordered bin.
204 		 * Outer loop:
205 		 *	Once the 1st disordered bin is done, ie. aj >= ak,
206 		 *	aj<-aj + count[c] connects the bins in a linked list;
207 		 *	reset count[c].
208 		 */
209 		for (aj = a; aj < an;  *aj = r, aj += count[c], count[c] = 0)
210 			for (r = *aj;  aj < (ak = --top[c = tr[r[i]]]);)
211 				swap(*ak, r, t);
212 	}
213 }
214 
215 /* Stable sort, requiring additional memory. */
216 static void
217 r_sort_b(const u_char **a, const u_char **ta, int n, int i, const u_char *tr,
218 	 u_int endch)
219 {
220 	static int count[256], nc, bmin;
221 	int c;
222 	const u_char **ak, **ai;
223 	stack s[512], *sp, *sp0, *sp1, temp;
224 	const u_char **top[256];
225 	int *cp, bigc;
226 
227 	sp = s;
228 	push(a, n, i);
229 	while (!empty(s)) {
230 		pop(a, n, i);
231 		if (n < THRESHOLD) {
232 			simplesort(a, n, i, tr, endch);
233 			continue;
234 		}
235 
236 		if (nc == 0) {
237 			bmin = 255;
238 			for (ak = a + n; --ak >= a;) {
239 				c = tr[(*ak)[i]];
240 				if (++count[c] == 1 && c != endch) {
241 					if (c < bmin)
242 						bmin = c;
243 					nc++;
244 				}
245 			}
246 			if (sp + nc > s + SIZE) {
247 				r_sort_b(a, ta, n, i, tr, endch);
248 				continue;
249 			}
250 		}
251 
252 		sp0 = sp1 = sp;
253 		bigc = 2;
254 		if (endch == 0) {
255 			top[0] = ak = a + count[0];
256 			count[0] = 0;
257 		} else {
258 			ak = a;
259 			top[255] = a + n;
260 			count[255] = 0;
261 		}
262 		for (cp = count + bmin; nc > 0; cp++) {
263 			while (*cp == 0)
264 				cp++;
265 			if ((c = *cp) > 1) {
266 				if (c > bigc) {
267 					bigc = c;
268 					sp1 = sp;
269 				}
270 				push(ak, c, i+1);
271 			}
272 			top[cp-count] = ak += c;
273 			*cp = 0;			/* Reset count[]. */
274 			nc--;
275 		}
276 		swap(*sp0, *sp1, temp);
277 
278 		for (ak = ta + n, ai = a+n; ak > ta;)	/* Copy to temp. */
279 			*--ak = *--ai;
280 		for (ak = ta+n; --ak >= ta;)		/* Deal to piles. */
281 			*--top[tr[(*ak)[i]]] = *ak;
282 	}
283 }
284 
285 /* insertion sort */
286 static inline void
287 simplesort(const u_char **a, int n, int b, const u_char *tr, u_int endch)
288 {
289 	u_char ch;
290 	const u_char  **ak, **ai, *s, *t;
291 
292 	for (ak = a+1; --n >= 1; ak++)
293 		for (ai = ak; ai > a; ai--) {
294 			for (s = ai[0] + b, t = ai[-1] + b;
295 			    (ch = tr[*s]) != endch; s++, t++)
296 				if (ch != tr[*t])
297 					break;
298 			if (ch >= tr[*t])
299 				break;
300 			swap(ai[0], ai[-1], s);
301 		}
302 }
303