Mercurial > sdl-ios-xcode

comparison src/stdlib/SDL_qsort.c @ 1330:450721ad5436

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
It's now possible to build SDL without any C runtime at all on Windows, using Visual C++ 2005
author Sam Lantinga <slouken@libsdl.org>
date Mon, 06 Feb 2006 08:28:51 +0000
parents
children 1cbaeee565b1
comparison
equal deleted inserted replaced
1329:bc67bbf87818 1330:450721ad5436
1 /* qsort.c
2 * (c) 1998 Gareth McCaughan
3 *
4 * This is a drop-in replacement for the C library's |qsort()| routine.
5 *
6 * Features:
7 * - Median-of-three pivoting (and more)
8 * - Truncation and final polishing by a single insertion sort
9 * - Early truncation when no swaps needed in pivoting step
10 * - Explicit recursion, guaranteed not to overflow
11 * - A few little wrinkles stolen from the GNU |qsort()|.
12 * - separate code for non-aligned / aligned / word-size objects
13 *
14 * This code may be reproduced freely provided
15 * - this file is retained unaltered apart from minor
16 * changes for portability and efficiency
17 * - no changes are made to this comment
18 * - any changes that *are* made are clearly flagged
19 * - the _ID string below is altered by inserting, after
20 * the date, the string " altered" followed at your option
21 * by other material. (Exceptions: you may change the name
22 * of the exported routine without changing the ID string.
23 * You may change the values of the macros TRUNC_* and
24 * PIVOT_THRESHOLD without changing the ID string, provided
25 * they remain constants with TRUNC_nonaligned, TRUNC_aligned
26 * and TRUNC_words/WORD_BYTES between 8 and 24, and
27 * PIVOT_THRESHOLD between 32 and 200.)
28 *
29 * You may use it in anything you like; you may make money
30 * out of it; you may distribute it in object form or as
31 * part of an executable without including source code;
32 * you don't have to credit me. (But it would be nice if
33 * you did.)
34 *
35 * If you find problems with this code, or find ways of
36 * making it significantly faster, please let me know!
37 * My e-mail address, valid as of early 1998 and certainly
38 * OK for at least the next 18 months, is
39 * gjm11@dpmms.cam.ac.uk
40 * Thanks!
41 *
42 * Gareth McCaughan Peterhouse Cambridge 1998
43 */
44
45 /*
46 #include <assert.h>
47 #include <stdlib.h>
48 #include <string.h>
49 */
50 #define assert(X)
51 #include "SDL_stdlib.h"
52 #include "SDL_string.h"
53
54 #ifndef HAVE_QSORT
55
56 static char _ID[]="<qsort.c gjm 1.12 1998-03-19>";
57
58 /* How many bytes are there per word? (Must be a power of 2,
59 * and must in fact equal sizeof(int).)
60 */
61 #define WORD_BYTES sizeof(int)
62
63 /* How big does our stack need to be? Answer: one entry per
64 * bit in a |size_t|.
65 */
66 #define STACK_SIZE (8*sizeof(size_t))
67
68 /* Different situations have slightly different requirements,
69 * and we make life epsilon easier by using different truncation
70 * points for the three different cases.
71 * So far, I have tuned TRUNC_words and guessed that the same
72 * value might work well for the other two cases. Of course
73 * what works well on my machine might work badly on yours.
74 */
75 #define TRUNC_nonaligned 12
76 #define TRUNC_aligned 12
77 #define TRUNC_words 12*WORD_BYTES /* nb different meaning */
78
79 /* We use a simple pivoting algorithm for shortish sub-arrays
80 * and a more complicated one for larger ones. The threshold
81 * is PIVOT_THRESHOLD.
82 */
83 #define PIVOT_THRESHOLD 40
84
85 typedef struct { char * first; char * last; } stack_entry;
86 #define pushLeft {stack[stacktop].first=ffirst;stack[stacktop++].last=last;}
87 #define pushRight {stack[stacktop].first=first;stack[stacktop++].last=llast;}
88 #define doLeft {first=ffirst;llast=last;continue;}
89 #define doRight {ffirst=first;last=llast;continue;}
90 #define pop {if (--stacktop<0) break;\
91 first=ffirst=stack[stacktop].first;\
92 last=llast=stack[stacktop].last;\
93 continue;}
94
95 /* Some comments on the implementation.
96 * 1. When we finish partitioning the array into "low"
97 * and "high", we forget entirely about short subarrays,
98 * because they'll be done later by insertion sort.
99 * Doing lots of little insertion sorts might be a win
100 * on large datasets for locality-of-reference reasons,
101 * but it makes the code much nastier and increases
102 * bookkeeping overhead.
103 * 2. We always save the shorter and get to work on the
104 * longer. This guarantees that every time we push
105 * an item onto the stack its size is <= 1/2 of that
106 * of its parent; so the stack can't need more than
107 * log_2(max-array-size) entries.
108 * 3. We choose a pivot by looking at the first, last
109 * and middle elements. We arrange them into order
110 * because it's easy to do that in conjunction with
111 * choosing the pivot, and it makes things a little
112 * easier in the partitioning step. Anyway, the pivot
113 * is the middle of these three. It's still possible
114 * to construct datasets where the algorithm takes
115 * time of order n^2, but it simply never happens in
116 * practice.
117 * 3' Newsflash: On further investigation I find that
118 * it's easy to construct datasets where median-of-3
119 * simply isn't good enough. So on large-ish subarrays
120 * we do a more sophisticated pivoting: we take three
121 * sets of 3 elements, find their medians, and then
122 * take the median of those.
123 * 4. We copy the pivot element to a separate place
124 * because that way we can always do our comparisons
125 * directly against a pointer to that separate place,
126 * and don't have to wonder "did we move the pivot
127 * element?". This makes the inner loop better.
128 * 5. It's possible to make the pivoting even more
129 * reliable by looking at more candidates when n
130 * is larger. (Taking this to its logical conclusion
131 * results in a variant of quicksort that doesn't
132 * have that n^2 worst case.) However, the overhead
133 * from the extra bookkeeping means that it's just
134 * not worth while.
135 * 6. This is pretty clean and portable code. Here are
136 * all the potential portability pitfalls and problems
137 * I know of:
138 * - In one place (the insertion sort) I construct
139 * a pointer that points just past the end of the
140 * supplied array, and assume that (a) it won't
141 * compare equal to any pointer within the array,
142 * and (b) it will compare equal to a pointer
143 * obtained by stepping off the end of the array.
144 * These might fail on some segmented architectures.
145 * - I assume that there are 8 bits in a |char| when
146 * computing the size of stack needed. This would
147 * fail on machines with 9-bit or 16-bit bytes.
148 * - I assume that if |((int)base&(sizeof(int)-1))==0|
149 * and |(size&(sizeof(int)-1))==0| then it's safe to
150 * get at array elements via |int*|s, and that if
151 * actually |size==sizeof(int)| as well then it's
152 * safe to treat the elements as |int|s. This might
153 * fail on systems that convert pointers to integers
154 * in non-standard ways.
155 * - I assume that |8*sizeof(size_t)<=INT_MAX|. This
156 * would be false on a machine with 8-bit |char|s,
157 * 16-bit |int|s and 4096-bit |size_t|s. :-)
158 */
159
160 /* The recursion logic is the same in each case: */
161 #define Recurse(Trunc) \
162 { size_t l=last-ffirst,r=llast-first; \
163 if (l<Trunc) { \
164 if (r>=Trunc) doRight \
165 else pop \
166 } \
167 else if (l<=r) { pushLeft; doRight } \
168 else if (r>=Trunc) { pushRight; doLeft }\
169 else doLeft \
170 }
171
172 /* and so is the pivoting logic: */
173 #define Pivot(swapper,sz) \
174 if ((size_t)(last-first)>PIVOT_THRESHOLD*sz) mid=pivot_big(first,mid,last,sz,compare);\
175 else { \
176 if (compare(first,mid)<0) { \
177 if (compare(mid,last)>0) { \
178 swapper(mid,last); \
179 if (compare(first,mid)>0) swapper(first,mid);\
180 } \
181 } \
182 else { \
183 if (compare(mid,last)>0) swapper(first,last)\
184 else { \
185 swapper(first,mid); \
186 if (compare(mid,last)>0) swapper(mid,last);\
187 } \
188 } \
189 first+=sz; last-=sz; \
190 }
191
192 #ifdef DEBUG_QSORT
193 #include <stdio.h>
194 #endif
195
196 /* and so is the partitioning logic: */
197 #define Partition(swapper,sz) { \
198 int swapped=0; \
199 do { \
200 while (compare(first,pivot)<0) first+=sz; \
201 while (compare(pivot,last)<0) last-=sz; \
202 if (first<last) { \
203 swapper(first,last); swapped=1; \
204 first+=sz; last-=sz; } \
205 else if (first==last) { first+=sz; last-=sz; break; }\
206 } while (first<=last); \
207 if (!swapped) pop \
208 }
209
210 /* and so is the pre-insertion-sort operation of putting
211 * the smallest element into place as a sentinel.
212 * Doing this makes the inner loop nicer. I got this
213 * idea from the GNU implementation of qsort().
214 */
215 #define PreInsertion(swapper,limit,sz) \
216 first=base; \
217 last=first + (nmemb>limit ? limit : nmemb-1)*sz;\
218 while (last!=base) { \
219 if (compare(first,last)>0) first=last; \
220 last-=sz; } \
221 if (first!=base) swapper(first,(char*)base);
222
223 /* and so is the insertion sort, in the first two cases: */
224 #define Insertion(swapper) \
225 last=((char*)base)+nmemb*size; \
226 for (first=((char*)base)+size;first!=last;first+=size) { \
227 char *test; \
228 /* Find the right place for |first|. \
229 * My apologies for var reuse. */ \
230 for (test=first-size;compare(test,first)>0;test-=size) ; \
231 test+=size; \
232 if (test!=first) { \
233 /* Shift everything in [test,first) \
234 * up by one, and place |first| \
235 * where |test| is. */ \
236 memcpy(pivot,first,size); \
237 memmove(test+size,test,first-test); \
238 memcpy(test,pivot,size); \
239 } \
240 }
241
242 #define SWAP_nonaligned(a,b) { \
243 register char *aa=(a),*bb=(b); \
244 register size_t sz=size; \
245 do { register char t=*aa; *aa++=*bb; *bb++=t; } while (--sz); }
246
247 #define SWAP_aligned(a,b) { \
248 register int *aa=(int*)(a),*bb=(int*)(b); \
249 register size_t sz=size; \
250 do { register int t=*aa;*aa++=*bb; *bb++=t; } while (sz-=WORD_BYTES); }
251
252 #define SWAP_words(a,b) { \
253 register int t=*((int*)a); *((int*)a)=*((int*)b); *((int*)b)=t; }
254
255 /* ---------------------------------------------------------------------- */
256
257 static char * pivot_big(char *first, char *mid, char *last, size_t size,
258 int compare(const void *, const void *)) {
259 int d=(((last-first)/size)>>3)*size;
260 char *m1,*m2,*m3;
261 { char *a=first, *b=first+d, *c=first+2*d;
262 #ifdef DEBUG_QSORT
263 fprintf(stderr,"< %d %d %d\n",*(int*)a,*(int*)b,*(int*)c);
264 #endif
265 m1 = compare(a,b)<0 ?
266 (compare(b,c)<0 ? b : (compare(a,c)<0 ? c : a))
267 : (compare(a,c)<0 ? a : (compare(b,c)<0 ? c : b));
268 }
269 { char *a=mid-d, *b=mid, *c=mid+d;
270 #ifdef DEBUG_QSORT
271 fprintf(stderr,". %d %d %d\n",*(int*)a,*(int*)b,*(int*)c);
272 #endif
273 m2 = compare(a,b)<0 ?
274 (compare(b,c)<0 ? b : (compare(a,c)<0 ? c : a))
275 : (compare(a,c)<0 ? a : (compare(b,c)<0 ? c : b));
276 }
277 { char *a=last-2*d, *b=last-d, *c=last;
278 #ifdef DEBUG_QSORT
279 fprintf(stderr,"> %d %d %d\n",*(int*)a,*(int*)b,*(int*)c);
280 #endif
281 m3 = compare(a,b)<0 ?
282 (compare(b,c)<0 ? b : (compare(a,c)<0 ? c : a))
283 : (compare(a,c)<0 ? a : (compare(b,c)<0 ? c : b));
284 }
285 #ifdef DEBUG_QSORT
286 fprintf(stderr,"-> %d %d %d\n",*(int*)m1,*(int*)m2,*(int*)m3);
287 #endif
288 return compare(m1,m2)<0 ?
289 (compare(m2,m3)<0 ? m2 : (compare(m1,m3)<0 ? m3 : m1))
290 : (compare(m1,m3)<0 ? m1 : (compare(m2,m3)<0 ? m3 : m2));
291 }
292
293 /* ---------------------------------------------------------------------- */
294
295 static void qsort_nonaligned(void *base, size_t nmemb, size_t size,
296 int (*compare)(const void *, const void *)) {
297
298 stack_entry stack[STACK_SIZE];
299 int stacktop=0;
300 char *first,*last;
301 char *pivot=malloc(size);
302 size_t trunc=TRUNC_nonaligned*size;
303 assert(pivot!=0);
304
305 first=(char*)base; last=first+(nmemb-1)*size;
306
307 if ((size_t)(last-first)>trunc) {
308 char *ffirst=first, *llast=last;
309 while (1) {
310 /* Select pivot */
311 { char * mid=first+size*((last-first)/size >> 1);
312 Pivot(SWAP_nonaligned,size);
313 memcpy(pivot,mid,size);
314 }
315 /* Partition. */
316 Partition(SWAP_nonaligned,size);
317 /* Prepare to recurse/iterate. */
318 Recurse(trunc)
319 }
320 }
321 PreInsertion(SWAP_nonaligned,TRUNC_nonaligned,size);
322 Insertion(SWAP_nonaligned);
323 free(pivot);
324 }
325
326 static void qsort_aligned(void *base, size_t nmemb, size_t size,
327 int (*compare)(const void *, const void *)) {
328
329 stack_entry stack[STACK_SIZE];
330 int stacktop=0;
331 char *first,*last;
332 char *pivot=malloc(size);
333 size_t trunc=TRUNC_aligned*size;
334 assert(pivot!=0);
335
336 first=(char*)base; last=first+(nmemb-1)*size;
337
338 if ((size_t)(last-first)>trunc) {
339 char *ffirst=first,*llast=last;
340 while (1) {
341 /* Select pivot */
342 { char * mid=first+size*((last-first)/size >> 1);
343 Pivot(SWAP_aligned,size);
344 memcpy(pivot,mid,size);
345 }
346 /* Partition. */
347 Partition(SWAP_aligned,size);
348 /* Prepare to recurse/iterate. */
349 Recurse(trunc)
350 }
351 }
352 PreInsertion(SWAP_aligned,TRUNC_aligned,size);
353 Insertion(SWAP_aligned);
354 free(pivot);
355 }
356
357 static void qsort_words(void *base, size_t nmemb,
358 int (*compare)(const void *, const void *)) {
359
360 stack_entry stack[STACK_SIZE];
361 int stacktop=0;
362 char *first,*last;
363 char *pivot=malloc(WORD_BYTES);
364 assert(pivot!=0);
365
366 first=(char*)base; last=first+(nmemb-1)*WORD_BYTES;
367
368 if (last-first>TRUNC_words) {
369 char *ffirst=first, *llast=last;
370 while (1) {
371 #ifdef DEBUG_QSORT
372 fprintf(stderr,"Doing %d:%d: ",
373 (first-(char*)base)/WORD_BYTES,
374 (last-(char*)base)/WORD_BYTES);
375 #endif
376 /* Select pivot */
377 { char * mid=first+WORD_BYTES*((last-first) / (2*WORD_BYTES));
378 Pivot(SWAP_words,WORD_BYTES);
379 *(int*)pivot=*(int*)mid;
380 }
381 #ifdef DEBUG_QSORT
382 fprintf(stderr,"pivot=%d\n",*(int*)pivot);
383 #endif
384 /* Partition. */
385 Partition(SWAP_words,WORD_BYTES);
386 /* Prepare to recurse/iterate. */
387 Recurse(TRUNC_words)
388 }
389 }
390 PreInsertion(SWAP_words,(TRUNC_words/WORD_BYTES),WORD_BYTES);
391 /* Now do insertion sort. */
392 last=((char*)base)+nmemb*WORD_BYTES;
393 for (first=((char*)base)+WORD_BYTES;first!=last;first+=WORD_BYTES) {
394 /* Find the right place for |first|. My apologies for var reuse */
395 int *pl=(int*)(first-WORD_BYTES),*pr=(int*)first;
396 *(int*)pivot=*(int*)first;
397 for (;compare(pl,pivot)>0;pr=pl,--pl) {
398 *pr=*pl; }
399 if (pr!=(int*)first) *pr=*(int*)pivot;
400 }
401 free(pivot);
402 }
403
404 /* ---------------------------------------------------------------------- */
405
406 void SDL_qsort(void *base, size_t nmemb, size_t size,
407 int (*compare)(const void *, const void *)) {
408
409 if (nmemb<=1) return;
410 if (((int)base|size)&(WORD_BYTES-1))
411 qsort_nonaligned(base,nmemb,size,compare);
412 else if (size!=WORD_BYTES)
413 qsort_aligned(base,nmemb,size,compare);
414 else
415 qsort_words(base,nmemb,compare);
416 }
417
418 #endif /* !HAVE_QSORT */