1 /* $NetBSD: radixsort.c,v 1.19 2009/09/05 08:53:06 dsl Exp $ */
2
3 /*-
4 * Copyright (c) 1990, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Peter McIlroy and by Dan Bernstein at New York University,
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 * may be used to endorse or promote products derived from this software
20 * without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35 #include <sys/cdefs.h>
36 #if defined(LIBC_SCCS) && !defined(lint)
37 #if 0
38 static char sccsid[] = "@(#)radixsort.c 8.2 (Berkeley) 4/28/95";
39 #else
40 __RCSID("$NetBSD: radixsort.c,v 1.19 2009/09/05 08:53:06 dsl Exp $");
41 #endif
42 #endif /* LIBC_SCCS and not lint */
43
44 /*
45 * Radixsort routines.
46 *
47 * Program r_sort_a() is unstable but uses O(logN) extra memory for a stack.
48 * Use radixsort(a, n, trace, endchar) for this case.
49 *
50 * For stable sorting (using N extra pointers) use sradixsort(), which calls
51 * r_sort_b().
52 *
53 * For a description of this code, see D. McIlroy, P. McIlroy, K. Bostic,
54 * "Engineering Radix Sort".
55 */
56
57 #include "namespace.h"
58 #include <sys/types.h>
59
60 #include <assert.h>
61 #include <errno.h>
62 #include <stdlib.h>
63
64 #ifdef __weak_alias
65 __weak_alias(radixsort,_radixsort)
66 __weak_alias(sradixsort,_sradixsort)
67 #endif
68
69 typedef struct {
70 const u_char **sa;
71 int sn, si;
72 } stack;
73
74 static inline void simplesort(const u_char **, int, int, const u_char *, u_int);
75 static void r_sort_a(const u_char **, int, int, const u_char *, u_int);
76 static void r_sort_b(const u_char **,
77 const u_char **, int, int, const u_char *, u_int);
78
79 #define THRESHOLD 20 /* Divert to simplesort(). */
80 #define SIZE 512 /* Default stack size. */
81
82 #define SETUP { \
83 if (tab == NULL) { \
84 tr = tr0; \
85 for (c = 0; c < endch; c++) \
86 tr0[c] = c + 1; \
87 tr0[c] = 0; \
88 for (c++; c < 256; c++) \
89 tr0[c] = c; \
90 endch = 0; \
91 } else { \
92 endch = tab[endch]; \
93 tr = tab; \
94 if (endch != 0 && endch != 255) { \
95 errno = EINVAL; \
96 return (-1); \
97 } \
98 } \
99 }
100
101 int
radixsort(const u_char ** a,int n,const u_char * tab,u_int endch)102 radixsort(const u_char **a, int n, const u_char *tab, u_int endch)
103 {
104 const u_char *tr;
105 u_int c;
106 u_char tr0[256];
107
108 _DIAGASSERT(a != NULL);
109
110 SETUP;
111 r_sort_a(a, n, 0, tr, endch);
112 return (0);
113 }
114
115 int
sradixsort(const u_char ** a,int n,const u_char * tab,u_int endch)116 sradixsort(const u_char **a, int n, const u_char *tab, u_int endch)
117 {
118 const u_char *tr, **ta;
119 u_int c;
120 u_char tr0[256];
121
122 _DIAGASSERT(a != NULL);
123 if (a == NULL) {
124 errno = EFAULT;
125 return (-1);
126 }
127
128 SETUP;
129 if (n < THRESHOLD)
130 simplesort(a, n, 0, tr, endch);
131 else {
132 if ((ta = malloc(n * sizeof(a))) == NULL)
133 return (-1);
134 r_sort_b(a, ta, n, 0, tr, endch);
135 free(ta);
136 }
137 return (0);
138 }
139
140 #define empty(s) (s >= sp)
141 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si
142 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i
143 #define swap(a, b, t) t = a, a = b, b = t
144
145 /* Unstable, in-place sort. */
146 static void
r_sort_a(const u_char ** a,int n,int i,const u_char * tr,u_int endch)147 r_sort_a(const u_char **a, int n, int i, const u_char *tr, u_int endch)
148 {
149 static u_int count[256], nc, bmin;
150 u_int c;
151 const u_char **ak, *r;
152 stack s[SIZE], *sp, *sp0, *sp1, temp;
153 u_int *cp, bigc;
154 const u_char **an, *t, **aj, **top[256];
155
156 _DIAGASSERT(a != NULL);
157 _DIAGASSERT(tr != NULL);
158
159 /* Set up stack. */
160 sp = s;
161 push(a, n, i);
162 while (!empty(s)) {
163 pop(a, n, i);
164 if (n < THRESHOLD) {
165 simplesort(a, n, i, tr, endch);
166 continue;
167 }
168 an = a + n;
169
170 /* Make character histogram. */
171 if (nc == 0) {
172 bmin = 255; /* First occupied bin, excluding eos. */
173 for (ak = a; ak < an;) {
174 c = tr[(*ak++)[i]];
175 if (++count[c] == 1 && c != endch) {
176 if (c < bmin)
177 bmin = c;
178 nc++;
179 }
180 }
181 if (sp + nc > s + SIZE) { /* Get more stack. */
182 r_sort_a(a, n, i, tr, endch);
183 continue;
184 }
185 }
186
187 /*
188 * Set top[]; push incompletely sorted bins onto stack.
189 * top[] = pointers to last out-of-place element in bins.
190 * count[] = counts of elements in bins.
191 * Before permuting: top[c-1] + count[c] = top[c];
192 * during deal: top[c] counts down to top[c-1].
193 */
194 sp0 = sp1 = sp; /* Stack position of biggest bin. */
195 bigc = 2; /* Size of biggest bin. */
196 if (endch == 0) /* Special case: set top[eos]. */
197 top[0] = ak = a + count[0];
198 else {
199 ak = a;
200 top[255] = an;
201 }
202 for (cp = count + bmin; nc > 0; cp++) {
203 while (*cp == 0) /* Find next non-empty pile. */
204 cp++;
205 if (*cp > 1) {
206 if (*cp > bigc) {
207 bigc = *cp;
208 sp1 = sp;
209 }
210 push(ak, *cp, i+1);
211 }
212 top[cp-count] = ak += *cp;
213 nc--;
214 }
215 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */
216
217 /*
218 * Permute misplacements home. Already home: everything
219 * before aj, and in bin[c], items from top[c] on.
220 * Inner loop:
221 * r = next element to put in place;
222 * ak = top[r[i]] = location to put the next element.
223 * aj = bottom of 1st disordered bin.
224 * Outer loop:
225 * Once the 1st disordered bin is done, ie. aj >= ak,
226 * aj<-aj + count[c] connects the bins in a linked list;
227 * reset count[c].
228 */
229 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0)
230 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);)
231 swap(*ak, r, t);
232 }
233 }
234
235 /* Stable sort, requiring additional memory. */
236 static void
r_sort_b(const u_char ** a,const u_char ** ta,int n,int i,const u_char * tr,u_int endch)237 r_sort_b(const u_char **a, const u_char **ta, int n, int i, const u_char *tr,
238 u_int endch)
239 {
240 static u_int count[256], nc, bmin;
241 u_int c;
242 const u_char **ak, **ai;
243 stack s[512], *sp, *sp0, *sp1, temp;
244 const u_char **top[256];
245 u_int *cp, bigc;
246
247 _DIAGASSERT(a != NULL);
248 _DIAGASSERT(ta != NULL);
249 _DIAGASSERT(tr != NULL);
250
251 sp = s;
252 push(a, n, i);
253 while (!empty(s)) {
254 pop(a, n, i);
255 if (n < THRESHOLD) {
256 simplesort(a, n, i, tr, endch);
257 continue;
258 }
259
260 if (nc == 0) {
261 bmin = 255;
262 for (ak = a + n; --ak >= a;) {
263 c = tr[(*ak)[i]];
264 if (++count[c] == 1 && c != endch) {
265 if (c < bmin)
266 bmin = c;
267 nc++;
268 }
269 }
270 if (sp + nc > s + SIZE) {
271 r_sort_b(a, ta, n, i, tr, endch);
272 continue;
273 }
274 }
275
276 sp0 = sp1 = sp;
277 bigc = 2;
278 if (endch == 0) {
279 top[0] = ak = a + count[0];
280 count[0] = 0;
281 } else {
282 ak = a;
283 top[255] = a + n;
284 count[255] = 0;
285 }
286 for (cp = count + bmin; nc > 0; cp++) {
287 while (*cp == 0)
288 cp++;
289 if ((c = *cp) > 1) {
290 if (c > bigc) {
291 bigc = c;
292 sp1 = sp;
293 }
294 push(ak, c, i+1);
295 }
296 top[cp-count] = ak += c;
297 *cp = 0; /* Reset count[]. */
298 nc--;
299 }
300 swap(*sp0, *sp1, temp);
301
302 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */
303 *--ak = *--ai;
304 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */
305 *--top[tr[(*ak)[i]]] = *ak;
306 }
307 }
308
309 /* insertion sort */
310 static inline void
simplesort(const u_char ** a,int n,int b,const u_char * tr,u_int endch)311 simplesort(const u_char **a, int n, int b, const u_char *tr, u_int endch)
312 {
313 u_char ch;
314 const u_char **ak, **ai, *s, *t;
315
316 _DIAGASSERT(a != NULL);
317 _DIAGASSERT(tr != NULL);
318
319 for (ak = a+1; --n >= 1; ak++)
320 for (ai = ak; ai > a; ai--) {
321 for (s = ai[0] + b, t = ai[-1] + b;
322 (ch = tr[*s]) != endch; s++, t++)
323 if (ch != tr[*t])
324 break;
325 if (ch >= tr[*t])
326 break;
327 swap(ai[0], ai[-1], s);
328 }
329 }
330