1 /* $NetBSD: radixsort.c,v 1.6 1995/12/28 08:52:38 thorpej 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 #if defined(LIBC_SCCS) && !defined(lint) 40 #if 0 41 static char sccsid[] = "from: @(#)radixsort.c 8.1 (Berkeley) 6/4/93"; 42 #else 43 static char *rcsid = "$NetBSD: radixsort.c,v 1.6 1995/12/28 08:52:38 thorpej Exp $"; 44 #endif 45 #endif /* LIBC_SCCS and not lint */ 46 47 /* 48 * Radixsort routines. 49 * 50 * Program r_sort_a() is unstable but uses O(logN) extra memory for a stack. 51 * Use radixsort(a, n, trace, endchar) for this case. 52 * 53 * For stable sorting (using N extra pointers) use sradixsort(), which calls 54 * r_sort_b(). 55 * 56 * For a description of this code, see D. McIlroy, P. McIlroy, K. Bostic, 57 * "Engineering Radix Sort". 58 */ 59 60 #include <sys/types.h> 61 #include <stdlib.h> 62 #include <errno.h> 63 64 typedef struct { 65 const u_char **sa; 66 int sn, si; 67 } stack; 68 69 static inline void simplesort 70 __P((const u_char **, int, int, const u_char *, u_int)); 71 static void r_sort_a __P((const u_char **, int, int, const u_char *, u_int)); 72 static void r_sort_b __P((const u_char **, 73 const u_char **, int, int, const u_char *, u_int)); 74 75 #define THRESHOLD 20 /* Divert to simplesort(). */ 76 #define SIZE 512 /* Default stack size. */ 77 78 #define SETUP { \ 79 if (tab == NULL) { \ 80 tr = tr0; \ 81 for (c = 0; c < endch; c++) \ 82 tr0[c] = c + 1; \ 83 tr0[c] = 0; \ 84 for (c++; c < 256; c++) \ 85 tr0[c] = c; \ 86 endch = 0; \ 87 } else { \ 88 endch = tab[endch]; \ 89 tr = tab; \ 90 if (endch != 0 && endch != 255) { \ 91 errno = EINVAL; \ 92 return (-1); \ 93 } \ 94 } \ 95 } 96 97 int 98 radixsort(a, n, tab, endch) 99 const u_char **a, *tab; 100 int n; 101 u_int endch; 102 { 103 const u_char *tr; 104 int c; 105 u_char tr0[256]; 106 107 SETUP; 108 r_sort_a(a, n, 0, tr, endch); 109 return (0); 110 } 111 112 int 113 sradixsort(a, n, tab, endch) 114 const u_char **a, *tab; 115 int n; 116 u_int endch; 117 { 118 const u_char *tr, **ta; 119 int c; 120 u_char tr0[256]; 121 122 SETUP; 123 if (n < THRESHOLD) 124 simplesort(a, n, 0, tr, endch); 125 else { 126 if ((ta = malloc(n * sizeof(a))) == NULL) 127 return (-1); 128 r_sort_b(a, ta, n, 0, tr, endch); 129 free(ta); 130 } 131 return (0); 132 } 133 134 #define empty(s) (s >= sp) 135 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si 136 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i 137 #define swap(a, b, t) t = a, a = b, b = t 138 139 /* Unstable, in-place sort. */ 140 void 141 r_sort_a(a, n, i, tr, endch) 142 const u_char **a; 143 int n, i; 144 const u_char *tr; 145 u_int endch; 146 { 147 static int count[256], nc, bmin; 148 register int c; 149 register const u_char **ak, *r; 150 stack s[SIZE], *sp, *sp0, *sp1, temp; 151 int *cp, bigc; 152 const u_char **an, *t, **aj, **top[256]; 153 154 /* Set up stack. */ 155 sp = s; 156 push(a, n, i); 157 while (!empty(s)) { 158 pop(a, n, i); 159 if (n < THRESHOLD) { 160 simplesort(a, n, i, tr, endch); 161 continue; 162 } 163 an = a + n; 164 165 /* Make character histogram. */ 166 if (nc == 0) { 167 bmin = 255; /* First occupied bin, excluding eos. */ 168 for (ak = a; ak < an;) { 169 c = tr[(*ak++)[i]]; 170 if (++count[c] == 1 && c != endch) { 171 if (c < bmin) 172 bmin = c; 173 nc++; 174 } 175 } 176 if (sp + nc > s + SIZE) { /* Get more stack. */ 177 r_sort_a(a, n, i, tr, endch); 178 continue; 179 } 180 } 181 182 /* 183 * Set top[]; push incompletely sorted bins onto stack. 184 * top[] = pointers to last out-of-place element in bins. 185 * count[] = counts of elements in bins. 186 * Before permuting: top[c-1] + count[c] = top[c]; 187 * during deal: top[c] counts down to top[c-1]. 188 */ 189 sp0 = sp1 = sp; /* Stack position of biggest bin. */ 190 bigc = 2; /* Size of biggest bin. */ 191 if (endch == 0) /* Special case: set top[eos]. */ 192 top[0] = ak = a + count[0]; 193 else { 194 ak = a; 195 top[255] = an; 196 } 197 for (cp = count + bmin; nc > 0; cp++) { 198 while (*cp == 0) /* Find next non-empty pile. */ 199 cp++; 200 if (*cp > 1) { 201 if (*cp > bigc) { 202 bigc = *cp; 203 sp1 = sp; 204 } 205 push(ak, *cp, i+1); 206 } 207 top[cp-count] = ak += *cp; 208 nc--; 209 } 210 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */ 211 212 /* 213 * Permute misplacements home. Already home: everything 214 * before aj, and in bin[c], items from top[c] on. 215 * Inner loop: 216 * r = next element to put in place; 217 * ak = top[r[i]] = location to put the next element. 218 * aj = bottom of 1st disordered bin. 219 * Outer loop: 220 * Once the 1st disordered bin is done, ie. aj >= ak, 221 * aj<-aj + count[c] connects the bins in a linked list; 222 * reset count[c]. 223 */ 224 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0) 225 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);) 226 swap(*ak, r, t); 227 } 228 } 229 230 /* Stable sort, requiring additional memory. */ 231 void 232 r_sort_b(a, ta, n, i, tr, endch) 233 const u_char **a, **ta; 234 int n, i; 235 const u_char *tr; 236 u_int endch; 237 { 238 static int count[256], nc, bmin; 239 register int c; 240 register const u_char **ak, **ai; 241 stack s[512], *sp, *sp0, *sp1, temp; 242 const u_char **top[256]; 243 int *cp, bigc; 244 245 sp = s; 246 push(a, n, i); 247 while (!empty(s)) { 248 pop(a, n, i); 249 if (n < THRESHOLD) { 250 simplesort(a, n, i, tr, endch); 251 continue; 252 } 253 254 if (nc == 0) { 255 bmin = 255; 256 for (ak = a + n; --ak >= a;) { 257 c = tr[(*ak)[i]]; 258 if (++count[c] == 1 && c != endch) { 259 if (c < bmin) 260 bmin = c; 261 nc++; 262 } 263 } 264 if (sp + nc > s + SIZE) { 265 r_sort_b(a, ta, n, i, tr, endch); 266 continue; 267 } 268 } 269 270 sp0 = sp1 = sp; 271 bigc = 2; 272 if (endch == 0) { 273 top[0] = ak = a + count[0]; 274 count[0] = 0; 275 } else { 276 ak = a; 277 top[255] = a + n; 278 count[255] = 0; 279 } 280 for (cp = count + bmin; nc > 0; cp++) { 281 while (*cp == 0) 282 cp++; 283 if ((c = *cp) > 1) { 284 if (c > bigc) { 285 bigc = c; 286 sp1 = sp; 287 } 288 push(ak, c, i+1); 289 } 290 top[cp-count] = ak += c; 291 *cp = 0; /* Reset count[]. */ 292 nc--; 293 } 294 swap(*sp0, *sp1, temp); 295 296 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */ 297 *--ak = *--ai; 298 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */ 299 *--top[tr[(*ak)[i]]] = *ak; 300 } 301 } 302 303 static inline void 304 simplesort(a, n, b, tr, endch) /* insertion sort */ 305 register const u_char **a; 306 int n, b; 307 register const u_char *tr; 308 u_int endch; 309 { 310 register u_char ch; 311 const u_char **ak, **ai, *s, *t; 312 313 for (ak = a+1; --n >= 1; ak++) 314 for (ai = ak; ai > a; ai--) { 315 for (s = ai[0] + b, t = ai[-1] + b; 316 (ch = tr[*s]) != endch; s++, t++) 317 if (ch != tr[*t]) 318 break; 319 if (ch >= tr[*t]) 320 break; 321 swap(ai[0], ai[-1], s); 322 } 323 } 324