1 /* $NetBSD: radixsort.c,v 1.18 2009/08/21 20:49:50 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 1.18 2009/08/21 20:49:50 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 75 __P((const u_char **, int, int, const u_char *, u_int)); 76 static void r_sort_a __P((const u_char **, int, int, const u_char *, u_int)); 77 static void r_sort_b __P((const u_char **, 78 const u_char **, int, int, const u_char *, u_int)); 79 80 #define THRESHOLD 20 /* Divert to simplesort(). */ 81 #define SIZE 512 /* Default stack size. */ 82 83 #define SETUP { \ 84 if (tab == NULL) { \ 85 tr = tr0; \ 86 for (c = 0; c < endch; c++) \ 87 tr0[c] = c + 1; \ 88 tr0[c] = 0; \ 89 for (c++; c < 256; c++) \ 90 tr0[c] = c; \ 91 endch = 0; \ 92 } else { \ 93 endch = tab[endch]; \ 94 tr = tab; \ 95 if (endch != 0 && endch != 255) { \ 96 errno = EINVAL; \ 97 return (-1); \ 98 } \ 99 } \ 100 } 101 102 int 103 radixsort(a, n, tab, endch) 104 const u_char **a, *tab; 105 int n; 106 u_int endch; 107 { 108 const u_char *tr; 109 u_int c; 110 u_char tr0[256]; 111 112 _DIAGASSERT(a != NULL); 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 u_int c; 127 u_char tr0[256]; 128 129 _DIAGASSERT(a != NULL); 130 if (a == NULL) { 131 errno = EFAULT; 132 return (-1); 133 } 134 135 SETUP; 136 if (n < THRESHOLD) 137 simplesort(a, n, 0, tr, endch); 138 else { 139 if ((ta = malloc(n * sizeof(a))) == NULL) 140 return (-1); 141 r_sort_b(a, ta, n, 0, tr, endch); 142 free(ta); 143 } 144 return (0); 145 } 146 147 #define empty(s) (s >= sp) 148 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si 149 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i 150 #define swap(a, b, t) t = a, a = b, b = t 151 152 /* Unstable, in-place sort. */ 153 static void 154 r_sort_a(a, n, i, tr, endch) 155 const u_char **a; 156 int n, i; 157 const u_char *tr; 158 u_int endch; 159 { 160 static u_int count[256], nc, bmin; 161 u_int c; 162 const u_char **ak, *r; 163 stack s[SIZE], *sp, *sp0, *sp1, temp; 164 u_int *cp, bigc; 165 const u_char **an, *t, **aj, **top[256]; 166 167 _DIAGASSERT(a != NULL); 168 _DIAGASSERT(tr != NULL); 169 170 /* Set up stack. */ 171 sp = s; 172 push(a, n, i); 173 while (!empty(s)) { 174 pop(a, n, i); 175 if (n < THRESHOLD) { 176 simplesort(a, n, i, tr, endch); 177 continue; 178 } 179 an = a + n; 180 181 /* Make character histogram. */ 182 if (nc == 0) { 183 bmin = 255; /* First occupied bin, excluding eos. */ 184 for (ak = a; ak < an;) { 185 c = tr[(*ak++)[i]]; 186 if (++count[c] == 1 && c != endch) { 187 if (c < bmin) 188 bmin = c; 189 nc++; 190 } 191 } 192 if (sp + nc > s + SIZE) { /* Get more stack. */ 193 r_sort_a(a, n, i, tr, endch); 194 continue; 195 } 196 } 197 198 /* 199 * Set top[]; push incompletely sorted bins onto stack. 200 * top[] = pointers to last out-of-place element in bins. 201 * count[] = counts of elements in bins. 202 * Before permuting: top[c-1] + count[c] = top[c]; 203 * during deal: top[c] counts down to top[c-1]. 204 */ 205 sp0 = sp1 = sp; /* Stack position of biggest bin. */ 206 bigc = 2; /* Size of biggest bin. */ 207 if (endch == 0) /* Special case: set top[eos]. */ 208 top[0] = ak = a + count[0]; 209 else { 210 ak = a; 211 top[255] = an; 212 } 213 for (cp = count + bmin; nc > 0; cp++) { 214 while (*cp == 0) /* Find next non-empty pile. */ 215 cp++; 216 if (*cp > 1) { 217 if (*cp > bigc) { 218 bigc = *cp; 219 sp1 = sp; 220 } 221 push(ak, *cp, i+1); 222 } 223 top[cp-count] = ak += *cp; 224 nc--; 225 } 226 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */ 227 228 /* 229 * Permute misplacements home. Already home: everything 230 * before aj, and in bin[c], items from top[c] on. 231 * Inner loop: 232 * r = next element to put in place; 233 * ak = top[r[i]] = location to put the next element. 234 * aj = bottom of 1st disordered bin. 235 * Outer loop: 236 * Once the 1st disordered bin is done, ie. aj >= ak, 237 * aj<-aj + count[c] connects the bins in a linked list; 238 * reset count[c]. 239 */ 240 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0) 241 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);) 242 swap(*ak, r, t); 243 } 244 } 245 246 /* Stable sort, requiring additional memory. */ 247 static void 248 r_sort_b(a, ta, n, i, tr, endch) 249 const u_char **a, **ta; 250 int n, i; 251 const u_char *tr; 252 u_int endch; 253 { 254 static u_int count[256], nc, bmin; 255 u_int c; 256 const u_char **ak, **ai; 257 stack s[512], *sp, *sp0, *sp1, temp; 258 const u_char **top[256]; 259 u_int *cp, bigc; 260 261 _DIAGASSERT(a != NULL); 262 _DIAGASSERT(ta != NULL); 263 _DIAGASSERT(tr != NULL); 264 265 sp = s; 266 push(a, n, i); 267 while (!empty(s)) { 268 pop(a, n, i); 269 if (n < THRESHOLD) { 270 simplesort(a, n, i, tr, endch); 271 continue; 272 } 273 274 if (nc == 0) { 275 bmin = 255; 276 for (ak = a + n; --ak >= a;) { 277 c = tr[(*ak)[i]]; 278 if (++count[c] == 1 && c != endch) { 279 if (c < bmin) 280 bmin = c; 281 nc++; 282 } 283 } 284 if (sp + nc > s + SIZE) { 285 r_sort_b(a, ta, n, i, tr, endch); 286 continue; 287 } 288 } 289 290 sp0 = sp1 = sp; 291 bigc = 2; 292 if (endch == 0) { 293 top[0] = ak = a + count[0]; 294 count[0] = 0; 295 } else { 296 ak = a; 297 top[255] = a + n; 298 count[255] = 0; 299 } 300 for (cp = count + bmin; nc > 0; cp++) { 301 while (*cp == 0) 302 cp++; 303 if ((c = *cp) > 1) { 304 if (c > bigc) { 305 bigc = c; 306 sp1 = sp; 307 } 308 push(ak, c, i+1); 309 } 310 top[cp-count] = ak += c; 311 *cp = 0; /* Reset count[]. */ 312 nc--; 313 } 314 swap(*sp0, *sp1, temp); 315 316 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */ 317 *--ak = *--ai; 318 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */ 319 *--top[tr[(*ak)[i]]] = *ak; 320 } 321 } 322 323 static inline void 324 simplesort(a, n, b, tr, endch) /* insertion sort */ 325 const u_char **a; 326 int n, b; 327 const u_char *tr; 328 u_int endch; 329 { 330 u_char ch; 331 const u_char **ak, **ai, *s, *t; 332 333 _DIAGASSERT(a != NULL); 334 _DIAGASSERT(tr != NULL); 335 336 for (ak = a+1; --n >= 1; ak++) 337 for (ai = ak; ai > a; ai--) { 338 for (s = ai[0] + b, t = ai[-1] + b; 339 (ch = tr[*s]) != endch; s++, t++) 340 if (ch != tr[*t]) 341 break; 342 if (ch >= tr[*t]) 343 break; 344 swap(ai[0], ai[-1], s); 345 } 346 } 347