1 /* $NetBSD: radixsort.c,v 1.22 2022/03/12 17:31:39 christos 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 1.22 2022/03/12 17:31:39 christos 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 ta = NULL;
133 errno = reallocarr(&ta, n, sizeof(*ta));
134 if (errno)
135 return -1;
136 r_sort_b(a, ta, n, 0, tr, endch);
137 free(ta);
138 }
139 return (0);
140 }
141
142 #define empty(s) (s >= sp)
143 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si
144 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i
145 #define swap(a, b, t) t = a, a = b, b = t
146
147 /* Unstable, in-place sort. */
148 static void
r_sort_a(const u_char ** a,int n,int i,const u_char * tr,u_int endch)149 r_sort_a(const u_char **a, int n, int i, const u_char *tr, u_int endch)
150 {
151 static u_int count[256], nc, bmin;
152 u_int c;
153 const u_char **ak, *r;
154 stack s[SIZE], *sp, *sp0, *sp1, temp;
155 u_int *cp, bigc;
156 const u_char **an, *t, **aj, **top[256];
157
158 _DIAGASSERT(a != NULL);
159 _DIAGASSERT(tr != NULL);
160
161 /* Set up stack. */
162 sp = s;
163 push(a, n, i);
164 while (!empty(s)) {
165 pop(a, n, i);
166 if (n < THRESHOLD) {
167 simplesort(a, n, i, tr, endch);
168 continue;
169 }
170 an = a + n;
171
172 /* Make character histogram. */
173 if (nc == 0) {
174 bmin = 255; /* First occupied bin, excluding eos. */
175 for (ak = a; ak < an;) {
176 c = tr[(*ak++)[i]];
177 if (++count[c] == 1 && c != endch) {
178 if (c < bmin)
179 bmin = c;
180 nc++;
181 }
182 }
183 if (sp + nc > s + SIZE) { /* Get more stack. */
184 r_sort_a(a, n, i, tr, endch);
185 continue;
186 }
187 }
188
189 /*
190 * Set top[]; push incompletely sorted bins onto stack.
191 * top[] = pointers to last out-of-place element in bins.
192 * count[] = counts of elements in bins.
193 * Before permuting: top[c-1] + count[c] = top[c];
194 * during deal: top[c] counts down to top[c-1].
195 */
196 sp0 = sp1 = sp; /* Stack position of biggest bin. */
197 bigc = 2; /* Size of biggest bin. */
198 if (endch == 0) /* Special case: set top[eos]. */
199 top[0] = ak = a + count[0];
200 else {
201 ak = a;
202 top[255] = an;
203 }
204 for (cp = count + bmin; nc > 0; cp++) {
205 while (*cp == 0) /* Find next non-empty pile. */
206 cp++;
207 if (*cp > 1) {
208 if (*cp > bigc) {
209 bigc = *cp;
210 sp1 = sp;
211 }
212 push(ak, *cp, i+1);
213 }
214 top[cp-count] = ak += *cp;
215 nc--;
216 }
217 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */
218
219 /*
220 * Permute misplacements home. Already home: everything
221 * before aj, and in bin[c], items from top[c] on.
222 * Inner loop:
223 * r = next element to put in place;
224 * ak = top[r[i]] = location to put the next element.
225 * aj = bottom of 1st disordered bin.
226 * Outer loop:
227 * Once the 1st disordered bin is done, ie. aj >= ak,
228 * aj<-aj + count[c] connects the bins in a linked list;
229 * reset count[c].
230 */
231 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0)
232 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);)
233 swap(*ak, r, t);
234 }
235 }
236
237 /* Stable sort, requiring additional memory. */
238 static void
r_sort_b(const u_char ** a,const u_char ** ta,int n,int i,const u_char * tr,u_int endch)239 r_sort_b(const u_char **a, const u_char **ta, int n, int i, const u_char *tr,
240 u_int endch)
241 {
242 static u_int count[256], nc, bmin;
243 u_int c;
244 const u_char **ak, **ai;
245 stack s[512], *sp, *sp0, *sp1, temp;
246 const u_char **top[256];
247 u_int *cp, bigc;
248
249 _DIAGASSERT(a != NULL);
250 _DIAGASSERT(ta != NULL);
251 _DIAGASSERT(tr != NULL);
252
253 sp = s;
254 push(a, n, i);
255 while (!empty(s)) {
256 pop(a, n, i);
257 if (n < THRESHOLD) {
258 simplesort(a, n, i, tr, endch);
259 continue;
260 }
261
262 if (nc == 0) {
263 bmin = 255;
264 for (ak = a + n; --ak >= a;) {
265 c = tr[(*ak)[i]];
266 if (++count[c] == 1 && c != endch) {
267 if (c < bmin)
268 bmin = c;
269 nc++;
270 }
271 }
272 if (sp + nc > s + SIZE) {
273 r_sort_b(a, ta, n, i, tr, endch);
274 continue;
275 }
276 }
277
278 sp0 = sp1 = sp;
279 bigc = 2;
280 if (endch == 0) {
281 top[0] = ak = a + count[0];
282 count[0] = 0;
283 } else {
284 ak = a;
285 top[255] = a + n;
286 count[255] = 0;
287 }
288 for (cp = count + bmin; nc > 0; cp++) {
289 while (*cp == 0)
290 cp++;
291 if ((c = *cp) > 1) {
292 if (c > bigc) {
293 bigc = c;
294 sp1 = sp;
295 }
296 push(ak, c, i+1);
297 }
298 top[cp-count] = ak += c;
299 *cp = 0; /* Reset count[]. */
300 nc--;
301 }
302 swap(*sp0, *sp1, temp);
303
304 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */
305 *--ak = *--ai;
306 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */
307 *--top[tr[(*ak)[i]]] = *ak;
308 }
309 }
310
311 /* insertion sort */
312 static inline void
simplesort(const u_char ** a,int n,int b,const u_char * tr,u_int endch)313 simplesort(const u_char **a, int n, int b, const u_char *tr, u_int endch)
314 {
315 u_char ch;
316 const u_char **ak, **ai, *s, *t;
317
318 _DIAGASSERT(a != NULL);
319 _DIAGASSERT(tr != NULL);
320
321 for (ak = a+1; --n >= 1; ak++)
322 for (ai = ak; ai > a; ai--) {
323 for (s = ai[0] + b, t = ai[-1] + b;
324 (ch = tr[*s]) != endch; s++, t++)
325 if (ch != tr[*t])
326 break;
327 if (ch >= tr[*t])
328 break;
329 swap(ai[0], ai[-1], s);
330 }
331 }
332