1 /* $NetBSD: radixsort.c,v 1.9 1997/07/21 14:09:00 jtc 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. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 */ 38 39 #include <sys/cdefs.h> 40 #if defined(LIBC_SCCS) && !defined(lint) 41 #if 0 42 static char sccsid[] = "from: @(#)radixsort.c 8.1 (Berkeley) 6/4/93"; 43 #else 44 __RCSID("$NetBSD: radixsort.c,v 1.9 1997/07/21 14:09:00 jtc Exp $"); 45 #endif 46 #endif /* LIBC_SCCS and not lint */ 47 48 /* 49 * Radixsort routines. 50 * 51 * Program r_sort_a() is unstable but uses O(logN) extra memory for a stack. 52 * Use radixsort(a, n, trace, endchar) for this case. 53 * 54 * For stable sorting (using N extra pointers) use sradixsort(), which calls 55 * r_sort_b(). 56 * 57 * For a description of this code, see D. McIlroy, P. McIlroy, K. Bostic, 58 * "Engineering Radix Sort". 59 */ 60 61 #include "namespace.h" 62 #include <sys/types.h> 63 #include <stdlib.h> 64 #include <errno.h> 65 66 #ifdef __weak_alias 67 __weak_alias(radixsort,_radixsort); 68 __weak_alias(sradixsort,_sradixsort); 69 #endif 70 71 typedef struct { 72 const u_char **sa; 73 int sn, si; 74 } stack; 75 76 static __inline void simplesort 77 __P((const u_char **, int, int, const u_char *, u_int)); 78 static void r_sort_a __P((const u_char **, int, int, const u_char *, u_int)); 79 static void r_sort_b __P((const u_char **, 80 const u_char **, int, int, const u_char *, u_int)); 81 82 #define THRESHOLD 20 /* Divert to simplesort(). */ 83 #define SIZE 512 /* Default stack size. */ 84 85 #define SETUP { \ 86 if (tab == NULL) { \ 87 tr = tr0; \ 88 for (c = 0; c < endch; c++) \ 89 tr0[c] = c + 1; \ 90 tr0[c] = 0; \ 91 for (c++; c < 256; c++) \ 92 tr0[c] = c; \ 93 endch = 0; \ 94 } else { \ 95 endch = tab[endch]; \ 96 tr = tab; \ 97 if (endch != 0 && endch != 255) { \ 98 errno = EINVAL; \ 99 return (-1); \ 100 } \ 101 } \ 102 } 103 104 int 105 radixsort(a, n, tab, endch) 106 const u_char **a, *tab; 107 int n; 108 u_int endch; 109 { 110 const u_char *tr; 111 int c; 112 u_char tr0[256]; 113 114 SETUP; 115 r_sort_a(a, n, 0, tr, endch); 116 return (0); 117 } 118 119 int 120 sradixsort(a, n, tab, endch) 121 const u_char **a, *tab; 122 int n; 123 u_int endch; 124 { 125 const u_char *tr, **ta; 126 int c; 127 u_char tr0[256]; 128 129 SETUP; 130 if (n < THRESHOLD) 131 simplesort(a, n, 0, tr, endch); 132 else { 133 if ((ta = malloc(n * sizeof(a))) == NULL) 134 return (-1); 135 r_sort_b(a, ta, n, 0, tr, endch); 136 free(ta); 137 } 138 return (0); 139 } 140 141 #define empty(s) (s >= sp) 142 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si 143 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i 144 #define swap(a, b, t) t = a, a = b, b = t 145 146 /* Unstable, in-place sort. */ 147 void 148 r_sort_a(a, n, i, tr, endch) 149 const u_char **a; 150 int n, i; 151 const u_char *tr; 152 u_int endch; 153 { 154 static int count[256], nc, bmin; 155 register int c; 156 register const u_char **ak, *r; 157 stack s[SIZE], *sp, *sp0, *sp1, temp; 158 int *cp, bigc; 159 const u_char **an, *t, **aj, **top[256]; 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 void 239 r_sort_b(a, ta, n, i, tr, endch) 240 const u_char **a, **ta; 241 int n, i; 242 const u_char *tr; 243 u_int endch; 244 { 245 static int count[256], nc, bmin; 246 register int c; 247 register const u_char **ak, **ai; 248 stack s[512], *sp, *sp0, *sp1, temp; 249 const u_char **top[256]; 250 int *cp, bigc; 251 252 sp = s; 253 push(a, n, i); 254 while (!empty(s)) { 255 pop(a, n, i); 256 if (n < THRESHOLD) { 257 simplesort(a, n, i, tr, endch); 258 continue; 259 } 260 261 if (nc == 0) { 262 bmin = 255; 263 for (ak = a + n; --ak >= a;) { 264 c = tr[(*ak)[i]]; 265 if (++count[c] == 1 && c != endch) { 266 if (c < bmin) 267 bmin = c; 268 nc++; 269 } 270 } 271 if (sp + nc > s + SIZE) { 272 r_sort_b(a, ta, n, i, tr, endch); 273 continue; 274 } 275 } 276 277 sp0 = sp1 = sp; 278 bigc = 2; 279 if (endch == 0) { 280 top[0] = ak = a + count[0]; 281 count[0] = 0; 282 } else { 283 ak = a; 284 top[255] = a + n; 285 count[255] = 0; 286 } 287 for (cp = count + bmin; nc > 0; cp++) { 288 while (*cp == 0) 289 cp++; 290 if ((c = *cp) > 1) { 291 if (c > bigc) { 292 bigc = c; 293 sp1 = sp; 294 } 295 push(ak, c, i+1); 296 } 297 top[cp-count] = ak += c; 298 *cp = 0; /* Reset count[]. */ 299 nc--; 300 } 301 swap(*sp0, *sp1, temp); 302 303 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */ 304 *--ak = *--ai; 305 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */ 306 *--top[tr[(*ak)[i]]] = *ak; 307 } 308 } 309 310 static __inline void 311 simplesort(a, n, b, tr, endch) /* insertion sort */ 312 register const u_char **a; 313 int n, b; 314 register const u_char *tr; 315 u_int endch; 316 { 317 register u_char ch; 318 const u_char **ak, **ai, *s, *t; 319 320 for (ak = a+1; --n >= 1; ak++) 321 for (ai = ak; ai > a; ai--) { 322 for (s = ai[0] + b, t = ai[-1] + b; 323 (ch = tr[*s]) != endch; s++, t++) 324 if (ch != tr[*t]) 325 break; 326 if (ch >= tr[*t]) 327 break; 328 swap(ai[0], ai[-1], s); 329 } 330 } 331