1 /* $NetBSD: radixsort.c,v 1.8 1997/07/13 20:16:52 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. 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 1.8 1997/07/13 20:16:52 christos 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 <sys/types.h> 62 #include <stdlib.h> 63 #include <errno.h> 64 65 typedef struct { 66 const u_char **sa; 67 int sn, si; 68 } stack; 69 70 static __inline void simplesort 71 __P((const u_char **, int, int, const u_char *, u_int)); 72 static void r_sort_a __P((const u_char **, int, int, const u_char *, u_int)); 73 static void r_sort_b __P((const u_char **, 74 const u_char **, int, int, const u_char *, u_int)); 75 76 #define THRESHOLD 20 /* Divert to simplesort(). */ 77 #define SIZE 512 /* Default stack size. */ 78 79 #define SETUP { \ 80 if (tab == NULL) { \ 81 tr = tr0; \ 82 for (c = 0; c < endch; c++) \ 83 tr0[c] = c + 1; \ 84 tr0[c] = 0; \ 85 for (c++; c < 256; c++) \ 86 tr0[c] = c; \ 87 endch = 0; \ 88 } else { \ 89 endch = tab[endch]; \ 90 tr = tab; \ 91 if (endch != 0 && endch != 255) { \ 92 errno = EINVAL; \ 93 return (-1); \ 94 } \ 95 } \ 96 } 97 98 int 99 radixsort(a, n, tab, endch) 100 const u_char **a, *tab; 101 int n; 102 u_int endch; 103 { 104 const u_char *tr; 105 int c; 106 u_char tr0[256]; 107 108 SETUP; 109 r_sort_a(a, n, 0, tr, endch); 110 return (0); 111 } 112 113 int 114 sradixsort(a, n, tab, endch) 115 const u_char **a, *tab; 116 int n; 117 u_int endch; 118 { 119 const u_char *tr, **ta; 120 int c; 121 u_char tr0[256]; 122 123 SETUP; 124 if (n < THRESHOLD) 125 simplesort(a, n, 0, tr, endch); 126 else { 127 if ((ta = malloc(n * sizeof(a))) == NULL) 128 return (-1); 129 r_sort_b(a, ta, n, 0, tr, endch); 130 free(ta); 131 } 132 return (0); 133 } 134 135 #define empty(s) (s >= sp) 136 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si 137 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i 138 #define swap(a, b, t) t = a, a = b, b = t 139 140 /* Unstable, in-place sort. */ 141 void 142 r_sort_a(a, n, i, tr, endch) 143 const u_char **a; 144 int n, i; 145 const u_char *tr; 146 u_int endch; 147 { 148 static int count[256], nc, bmin; 149 register int c; 150 register const u_char **ak, *r; 151 stack s[SIZE], *sp, *sp0, *sp1, temp; 152 int *cp, bigc; 153 const u_char **an, *t, **aj, **top[256]; 154 155 /* Set up stack. */ 156 sp = s; 157 push(a, n, i); 158 while (!empty(s)) { 159 pop(a, n, i); 160 if (n < THRESHOLD) { 161 simplesort(a, n, i, tr, endch); 162 continue; 163 } 164 an = a + n; 165 166 /* Make character histogram. */ 167 if (nc == 0) { 168 bmin = 255; /* First occupied bin, excluding eos. */ 169 for (ak = a; ak < an;) { 170 c = tr[(*ak++)[i]]; 171 if (++count[c] == 1 && c != endch) { 172 if (c < bmin) 173 bmin = c; 174 nc++; 175 } 176 } 177 if (sp + nc > s + SIZE) { /* Get more stack. */ 178 r_sort_a(a, n, i, tr, endch); 179 continue; 180 } 181 } 182 183 /* 184 * Set top[]; push incompletely sorted bins onto stack. 185 * top[] = pointers to last out-of-place element in bins. 186 * count[] = counts of elements in bins. 187 * Before permuting: top[c-1] + count[c] = top[c]; 188 * during deal: top[c] counts down to top[c-1]. 189 */ 190 sp0 = sp1 = sp; /* Stack position of biggest bin. */ 191 bigc = 2; /* Size of biggest bin. */ 192 if (endch == 0) /* Special case: set top[eos]. */ 193 top[0] = ak = a + count[0]; 194 else { 195 ak = a; 196 top[255] = an; 197 } 198 for (cp = count + bmin; nc > 0; cp++) { 199 while (*cp == 0) /* Find next non-empty pile. */ 200 cp++; 201 if (*cp > 1) { 202 if (*cp > bigc) { 203 bigc = *cp; 204 sp1 = sp; 205 } 206 push(ak, *cp, i+1); 207 } 208 top[cp-count] = ak += *cp; 209 nc--; 210 } 211 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */ 212 213 /* 214 * Permute misplacements home. Already home: everything 215 * before aj, and in bin[c], items from top[c] on. 216 * Inner loop: 217 * r = next element to put in place; 218 * ak = top[r[i]] = location to put the next element. 219 * aj = bottom of 1st disordered bin. 220 * Outer loop: 221 * Once the 1st disordered bin is done, ie. aj >= ak, 222 * aj<-aj + count[c] connects the bins in a linked list; 223 * reset count[c]. 224 */ 225 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0) 226 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);) 227 swap(*ak, r, t); 228 } 229 } 230 231 /* Stable sort, requiring additional memory. */ 232 void 233 r_sort_b(a, ta, n, i, tr, endch) 234 const u_char **a, **ta; 235 int n, i; 236 const u_char *tr; 237 u_int endch; 238 { 239 static int count[256], nc, bmin; 240 register int c; 241 register const u_char **ak, **ai; 242 stack s[512], *sp, *sp0, *sp1, temp; 243 const u_char **top[256]; 244 int *cp, bigc; 245 246 sp = s; 247 push(a, n, i); 248 while (!empty(s)) { 249 pop(a, n, i); 250 if (n < THRESHOLD) { 251 simplesort(a, n, i, tr, endch); 252 continue; 253 } 254 255 if (nc == 0) { 256 bmin = 255; 257 for (ak = a + n; --ak >= a;) { 258 c = tr[(*ak)[i]]; 259 if (++count[c] == 1 && c != endch) { 260 if (c < bmin) 261 bmin = c; 262 nc++; 263 } 264 } 265 if (sp + nc > s + SIZE) { 266 r_sort_b(a, ta, n, i, tr, endch); 267 continue; 268 } 269 } 270 271 sp0 = sp1 = sp; 272 bigc = 2; 273 if (endch == 0) { 274 top[0] = ak = a + count[0]; 275 count[0] = 0; 276 } else { 277 ak = a; 278 top[255] = a + n; 279 count[255] = 0; 280 } 281 for (cp = count + bmin; nc > 0; cp++) { 282 while (*cp == 0) 283 cp++; 284 if ((c = *cp) > 1) { 285 if (c > bigc) { 286 bigc = c; 287 sp1 = sp; 288 } 289 push(ak, c, i+1); 290 } 291 top[cp-count] = ak += c; 292 *cp = 0; /* Reset count[]. */ 293 nc--; 294 } 295 swap(*sp0, *sp1, temp); 296 297 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */ 298 *--ak = *--ai; 299 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */ 300 *--top[tr[(*ak)[i]]] = *ak; 301 } 302 } 303 304 static __inline void 305 simplesort(a, n, b, tr, endch) /* insertion sort */ 306 register const u_char **a; 307 int n, b; 308 register const u_char *tr; 309 u_int endch; 310 { 311 register u_char ch; 312 const u_char **ak, **ai, *s, *t; 313 314 for (ak = a+1; --n >= 1; ak++) 315 for (ai = ak; ai > a; ai--) { 316 for (s = ai[0] + b, t = ai[-1] + b; 317 (ch = tr[*s]) != endch; s++, t++) 318 if (ch != tr[*t]) 319 break; 320 if (ch >= tr[*t]) 321 break; 322 swap(ai[0], ai[-1], s); 323 } 324 } 325