Mercurial > libavcodec.hg
annotate imgresample.c @ 455:7f2d2be2aee9 libavcodec
dont double check vectors
more correct EPZS implementation
much faster halfpel ME
simplifications / reducing duplicate code
4MV in LQ mode
author | michaelni |
---|---|
date | Sun, 02 Jun 2002 12:18:18 +0000 |
parents | 718a22dc121f |
children | b4ee42142ad1 |
rev | line source |
---|---|
0 | 1 /* |
2 * High quality image resampling with polyphase filters | |
429 | 3 * Copyright (c) 2001 Fabrice Bellard. |
0 | 4 * |
429 | 5 * This library is free software; you can redistribute it and/or |
6 * modify it under the terms of the GNU Lesser General Public | |
7 * License as published by the Free Software Foundation; either | |
8 * version 2 of the License, or (at your option) any later version. | |
0 | 9 * |
429 | 10 * This library is distributed in the hope that it will be useful, |
0 | 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of |
429 | 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
13 * Lesser General Public License for more details. | |
0 | 14 * |
429 | 15 * You should have received a copy of the GNU Lesser General Public |
16 * License along with this library; if not, write to the Free Software | |
17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
0 | 18 */ |
396
fce0a2520551
removed useless header includes - use av memory functions
glantau
parents:
18
diff
changeset
|
19 #include "avcodec.h" |
0 | 20 #include "dsputil.h" |
21 | |
17 | 22 #ifdef USE_FASTMEMCPY |
23 #include "fastmemcpy.h" | |
24 #endif | |
25 | |
26 | |
0 | 27 #define NB_COMPONENTS 3 |
28 | |
29 #define PHASE_BITS 4 | |
30 #define NB_PHASES (1 << PHASE_BITS) | |
31 #define NB_TAPS 4 | |
32 #define FCENTER 1 /* index of the center of the filter */ | |
33 | |
34 #define POS_FRAC_BITS 16 | |
35 #define POS_FRAC (1 << POS_FRAC_BITS) | |
36 /* 6 bits precision is needed for MMX */ | |
37 #define FILTER_BITS 8 | |
38 | |
39 #define LINE_BUF_HEIGHT (NB_TAPS * 4) | |
40 | |
41 struct ImgReSampleContext { | |
42 int iwidth, iheight, owidth, oheight; | |
43 int h_incr, v_incr; | |
44 INT16 h_filters[NB_PHASES][NB_TAPS] __align8; /* horizontal filters */ | |
45 INT16 v_filters[NB_PHASES][NB_TAPS] __align8; /* vertical filters */ | |
46 UINT8 *line_buf; | |
47 }; | |
48 | |
49 static inline int get_phase(int pos) | |
50 { | |
51 return ((pos) >> (POS_FRAC_BITS - PHASE_BITS)) & ((1 << PHASE_BITS) - 1); | |
52 } | |
53 | |
54 /* This function must be optimized */ | |
55 static void h_resample_fast(UINT8 *dst, int dst_width, UINT8 *src, int src_width, | |
56 int src_start, int src_incr, INT16 *filters) | |
57 { | |
58 int src_pos, phase, sum, i; | |
59 UINT8 *s; | |
60 INT16 *filter; | |
61 | |
62 src_pos = src_start; | |
63 for(i=0;i<dst_width;i++) { | |
64 #ifdef TEST | |
65 /* test */ | |
66 if ((src_pos >> POS_FRAC_BITS) < 0 || | |
67 (src_pos >> POS_FRAC_BITS) > (src_width - NB_TAPS)) | |
68 abort(); | |
69 #endif | |
70 s = src + (src_pos >> POS_FRAC_BITS); | |
71 phase = get_phase(src_pos); | |
72 filter = filters + phase * NB_TAPS; | |
73 #if NB_TAPS == 4 | |
74 sum = s[0] * filter[0] + | |
75 s[1] * filter[1] + | |
76 s[2] * filter[2] + | |
77 s[3] * filter[3]; | |
78 #else | |
79 { | |
80 int j; | |
81 sum = 0; | |
82 for(j=0;j<NB_TAPS;j++) | |
83 sum += s[j] * filter[j]; | |
84 } | |
85 #endif | |
86 sum = sum >> FILTER_BITS; | |
87 if (sum < 0) | |
88 sum = 0; | |
89 else if (sum > 255) | |
90 sum = 255; | |
91 dst[0] = sum; | |
92 src_pos += src_incr; | |
93 dst++; | |
94 } | |
95 } | |
96 | |
97 /* This function must be optimized */ | |
98 static void v_resample(UINT8 *dst, int dst_width, UINT8 *src, int wrap, | |
99 INT16 *filter) | |
100 { | |
101 int sum, i; | |
102 UINT8 *s; | |
103 | |
104 s = src; | |
105 for(i=0;i<dst_width;i++) { | |
106 #if NB_TAPS == 4 | |
107 sum = s[0 * wrap] * filter[0] + | |
108 s[1 * wrap] * filter[1] + | |
109 s[2 * wrap] * filter[2] + | |
110 s[3 * wrap] * filter[3]; | |
111 #else | |
112 { | |
113 int j; | |
114 UINT8 *s1 = s; | |
115 | |
116 sum = 0; | |
117 for(j=0;j<NB_TAPS;j++) { | |
118 sum += s1[0] * filter[j]; | |
119 s1 += wrap; | |
120 } | |
121 } | |
122 #endif | |
123 sum = sum >> FILTER_BITS; | |
124 if (sum < 0) | |
125 sum = 0; | |
126 else if (sum > 255) | |
127 sum = 255; | |
128 dst[0] = sum; | |
129 dst++; | |
130 s++; | |
131 } | |
132 } | |
133 | |
2 | 134 #ifdef HAVE_MMX |
0 | 135 |
136 #include "i386/mmx.h" | |
137 | |
138 #define FILTER4(reg) \ | |
139 {\ | |
140 s = src + (src_pos >> POS_FRAC_BITS);\ | |
141 phase = get_phase(src_pos);\ | |
142 filter = filters + phase * NB_TAPS;\ | |
143 movq_m2r(*s, reg);\ | |
144 punpcklbw_r2r(mm7, reg);\ | |
145 movq_m2r(*filter, mm6);\ | |
146 pmaddwd_r2r(reg, mm6);\ | |
147 movq_r2r(mm6, reg);\ | |
148 psrlq_i2r(32, reg);\ | |
149 paddd_r2r(mm6, reg);\ | |
150 psrad_i2r(FILTER_BITS, reg);\ | |
151 src_pos += src_incr;\ | |
152 } | |
153 | |
154 #define DUMP(reg) movq_r2m(reg, tmp); printf(#reg "=%016Lx\n", tmp.uq); | |
155 | |
156 /* XXX: do four pixels at a time */ | |
157 static void h_resample_fast4_mmx(UINT8 *dst, int dst_width, UINT8 *src, int src_width, | |
158 int src_start, int src_incr, INT16 *filters) | |
159 { | |
160 int src_pos, phase; | |
161 UINT8 *s; | |
162 INT16 *filter; | |
163 mmx_t tmp; | |
164 | |
165 src_pos = src_start; | |
166 pxor_r2r(mm7, mm7); | |
167 | |
168 while (dst_width >= 4) { | |
169 | |
170 FILTER4(mm0); | |
171 FILTER4(mm1); | |
172 FILTER4(mm2); | |
173 FILTER4(mm3); | |
174 | |
175 packuswb_r2r(mm7, mm0); | |
176 packuswb_r2r(mm7, mm1); | |
177 packuswb_r2r(mm7, mm3); | |
178 packuswb_r2r(mm7, mm2); | |
179 movq_r2m(mm0, tmp); | |
180 dst[0] = tmp.ub[0]; | |
181 movq_r2m(mm1, tmp); | |
182 dst[1] = tmp.ub[0]; | |
183 movq_r2m(mm2, tmp); | |
184 dst[2] = tmp.ub[0]; | |
185 movq_r2m(mm3, tmp); | |
186 dst[3] = tmp.ub[0]; | |
187 dst += 4; | |
188 dst_width -= 4; | |
189 } | |
190 while (dst_width > 0) { | |
191 FILTER4(mm0); | |
192 packuswb_r2r(mm7, mm0); | |
193 movq_r2m(mm0, tmp); | |
194 dst[0] = tmp.ub[0]; | |
195 dst++; | |
196 dst_width--; | |
197 } | |
198 emms(); | |
199 } | |
200 | |
201 static void v_resample4_mmx(UINT8 *dst, int dst_width, UINT8 *src, int wrap, | |
202 INT16 *filter) | |
203 { | |
204 int sum, i, v; | |
205 UINT8 *s; | |
206 mmx_t tmp; | |
207 mmx_t coefs[4]; | |
208 | |
209 for(i=0;i<4;i++) { | |
210 v = filter[i]; | |
211 coefs[i].uw[0] = v; | |
212 coefs[i].uw[1] = v; | |
213 coefs[i].uw[2] = v; | |
214 coefs[i].uw[3] = v; | |
215 } | |
216 | |
217 pxor_r2r(mm7, mm7); | |
218 s = src; | |
219 while (dst_width >= 4) { | |
220 movq_m2r(s[0 * wrap], mm0); | |
221 punpcklbw_r2r(mm7, mm0); | |
222 movq_m2r(s[1 * wrap], mm1); | |
223 punpcklbw_r2r(mm7, mm1); | |
224 movq_m2r(s[2 * wrap], mm2); | |
225 punpcklbw_r2r(mm7, mm2); | |
226 movq_m2r(s[3 * wrap], mm3); | |
227 punpcklbw_r2r(mm7, mm3); | |
228 | |
229 pmullw_m2r(coefs[0], mm0); | |
230 pmullw_m2r(coefs[1], mm1); | |
231 pmullw_m2r(coefs[2], mm2); | |
232 pmullw_m2r(coefs[3], mm3); | |
233 | |
234 paddw_r2r(mm1, mm0); | |
235 paddw_r2r(mm3, mm2); | |
236 paddw_r2r(mm2, mm0); | |
237 psraw_i2r(FILTER_BITS, mm0); | |
238 | |
239 packuswb_r2r(mm7, mm0); | |
240 movq_r2m(mm0, tmp); | |
241 | |
242 *(UINT32 *)dst = tmp.ud[0]; | |
243 dst += 4; | |
244 s += 4; | |
245 dst_width -= 4; | |
246 } | |
247 while (dst_width > 0) { | |
248 sum = s[0 * wrap] * filter[0] + | |
249 s[1 * wrap] * filter[1] + | |
250 s[2 * wrap] * filter[2] + | |
251 s[3 * wrap] * filter[3]; | |
252 sum = sum >> FILTER_BITS; | |
253 if (sum < 0) | |
254 sum = 0; | |
255 else if (sum > 255) | |
256 sum = 255; | |
257 dst[0] = sum; | |
258 dst++; | |
259 s++; | |
260 dst_width--; | |
261 } | |
262 emms(); | |
263 } | |
264 #endif | |
265 | |
266 /* slow version to handle limit cases. Does not need optimisation */ | |
267 static void h_resample_slow(UINT8 *dst, int dst_width, UINT8 *src, int src_width, | |
268 int src_start, int src_incr, INT16 *filters) | |
269 { | |
270 int src_pos, phase, sum, j, v, i; | |
271 UINT8 *s, *src_end; | |
272 INT16 *filter; | |
273 | |
274 src_end = src + src_width; | |
275 src_pos = src_start; | |
276 for(i=0;i<dst_width;i++) { | |
277 s = src + (src_pos >> POS_FRAC_BITS); | |
278 phase = get_phase(src_pos); | |
279 filter = filters + phase * NB_TAPS; | |
280 sum = 0; | |
281 for(j=0;j<NB_TAPS;j++) { | |
282 if (s < src) | |
283 v = src[0]; | |
284 else if (s >= src_end) | |
285 v = src_end[-1]; | |
286 else | |
287 v = s[0]; | |
288 sum += v * filter[j]; | |
289 s++; | |
290 } | |
291 sum = sum >> FILTER_BITS; | |
292 if (sum < 0) | |
293 sum = 0; | |
294 else if (sum > 255) | |
295 sum = 255; | |
296 dst[0] = sum; | |
297 src_pos += src_incr; | |
298 dst++; | |
299 } | |
300 } | |
301 | |
302 static void h_resample(UINT8 *dst, int dst_width, UINT8 *src, int src_width, | |
303 int src_start, int src_incr, INT16 *filters) | |
304 { | |
305 int n, src_end; | |
306 | |
307 if (src_start < 0) { | |
308 n = (0 - src_start + src_incr - 1) / src_incr; | |
309 h_resample_slow(dst, n, src, src_width, src_start, src_incr, filters); | |
310 dst += n; | |
311 dst_width -= n; | |
312 src_start += n * src_incr; | |
313 } | |
314 src_end = src_start + dst_width * src_incr; | |
315 if (src_end > ((src_width - NB_TAPS) << POS_FRAC_BITS)) { | |
316 n = (((src_width - NB_TAPS + 1) << POS_FRAC_BITS) - 1 - src_start) / | |
317 src_incr; | |
318 } else { | |
319 n = dst_width; | |
320 } | |
2 | 321 #ifdef HAVE_MMX |
0 | 322 if ((mm_flags & MM_MMX) && NB_TAPS == 4) |
323 h_resample_fast4_mmx(dst, n, | |
324 src, src_width, src_start, src_incr, filters); | |
325 else | |
326 #endif | |
327 h_resample_fast(dst, n, | |
328 src, src_width, src_start, src_incr, filters); | |
329 if (n < dst_width) { | |
330 dst += n; | |
331 dst_width -= n; | |
332 src_start += n * src_incr; | |
333 h_resample_slow(dst, dst_width, | |
334 src, src_width, src_start, src_incr, filters); | |
335 } | |
336 } | |
337 | |
338 static void component_resample(ImgReSampleContext *s, | |
339 UINT8 *output, int owrap, int owidth, int oheight, | |
340 UINT8 *input, int iwrap, int iwidth, int iheight) | |
341 { | |
342 int src_y, src_y1, last_src_y, ring_y, phase_y, y1, y; | |
343 UINT8 *new_line, *src_line; | |
344 | |
345 last_src_y = - FCENTER - 1; | |
346 /* position of the bottom of the filter in the source image */ | |
347 src_y = (last_src_y + NB_TAPS) * POS_FRAC; | |
348 ring_y = NB_TAPS; /* position in ring buffer */ | |
349 for(y=0;y<oheight;y++) { | |
350 /* apply horizontal filter on new lines from input if needed */ | |
351 src_y1 = src_y >> POS_FRAC_BITS; | |
352 while (last_src_y < src_y1) { | |
353 if (++ring_y >= LINE_BUF_HEIGHT + NB_TAPS) | |
354 ring_y = NB_TAPS; | |
355 last_src_y++; | |
356 /* handle limit conditions : replicate line (slighly | |
357 inefficient because we filter multiple times */ | |
358 y1 = last_src_y; | |
359 if (y1 < 0) { | |
360 y1 = 0; | |
361 } else if (y1 >= iheight) { | |
362 y1 = iheight - 1; | |
363 } | |
364 src_line = input + y1 * iwrap; | |
365 new_line = s->line_buf + ring_y * owidth; | |
366 /* apply filter and handle limit cases correctly */ | |
367 h_resample(new_line, owidth, | |
368 src_line, iwidth, - FCENTER * POS_FRAC, s->h_incr, | |
369 &s->h_filters[0][0]); | |
370 /* handle ring buffer wraping */ | |
371 if (ring_y >= LINE_BUF_HEIGHT) { | |
372 memcpy(s->line_buf + (ring_y - LINE_BUF_HEIGHT) * owidth, | |
373 new_line, owidth); | |
374 } | |
375 } | |
376 /* apply vertical filter */ | |
377 phase_y = get_phase(src_y); | |
2 | 378 #ifdef HAVE_MMX |
0 | 379 /* desactivated MMX because loss of precision */ |
380 if ((mm_flags & MM_MMX) && NB_TAPS == 4 && 0) | |
381 v_resample4_mmx(output, owidth, | |
382 s->line_buf + (ring_y - NB_TAPS + 1) * owidth, owidth, | |
383 &s->v_filters[phase_y][0]); | |
384 else | |
385 #endif | |
386 v_resample(output, owidth, | |
387 s->line_buf + (ring_y - NB_TAPS + 1) * owidth, owidth, | |
388 &s->v_filters[phase_y][0]); | |
389 | |
390 src_y += s->v_incr; | |
391 output += owrap; | |
392 } | |
393 } | |
394 | |
395 /* XXX: the following filter is quite naive, but it seems to suffice | |
396 for 4 taps */ | |
397 static void build_filter(INT16 *filter, float factor) | |
398 { | |
399 int ph, i, v; | |
400 float x, y, tab[NB_TAPS], norm, mult; | |
401 | |
402 /* if upsampling, only need to interpolate, no filter */ | |
403 if (factor > 1.0) | |
404 factor = 1.0; | |
405 | |
406 for(ph=0;ph<NB_PHASES;ph++) { | |
407 norm = 0; | |
408 for(i=0;i<NB_TAPS;i++) { | |
409 | |
410 x = M_PI * ((float)(i - FCENTER) - (float)ph / NB_PHASES) * factor; | |
411 if (x == 0) | |
412 y = 1.0; | |
413 else | |
414 y = sin(x) / x; | |
415 tab[i] = y; | |
416 norm += y; | |
417 } | |
418 | |
419 /* normalize so that an uniform color remains the same */ | |
420 mult = (float)(1 << FILTER_BITS) / norm; | |
421 for(i=0;i<NB_TAPS;i++) { | |
422 v = (int)(tab[i] * mult); | |
423 filter[ph * NB_TAPS + i] = v; | |
424 } | |
425 } | |
426 } | |
427 | |
428 ImgReSampleContext *img_resample_init(int owidth, int oheight, | |
429 int iwidth, int iheight) | |
430 { | |
431 ImgReSampleContext *s; | |
432 | |
433 s = av_mallocz(sizeof(ImgReSampleContext)); | |
434 if (!s) | |
435 return NULL; | |
436 s->line_buf = av_mallocz(owidth * (LINE_BUF_HEIGHT + NB_TAPS)); | |
437 if (!s->line_buf) | |
438 goto fail; | |
439 | |
440 s->owidth = owidth; | |
441 s->oheight = oheight; | |
442 s->iwidth = iwidth; | |
443 s->iheight = iheight; | |
444 | |
445 s->h_incr = (iwidth * POS_FRAC) / owidth; | |
446 s->v_incr = (iheight * POS_FRAC) / oheight; | |
447 | |
448 build_filter(&s->h_filters[0][0], (float)owidth / (float)iwidth); | |
449 build_filter(&s->v_filters[0][0], (float)oheight / (float)iheight); | |
450 | |
451 return s; | |
452 fail: | |
396
fce0a2520551
removed useless header includes - use av memory functions
glantau
parents:
18
diff
changeset
|
453 av_free(s); |
0 | 454 return NULL; |
455 } | |
456 | |
457 void img_resample(ImgReSampleContext *s, | |
458 AVPicture *output, AVPicture *input) | |
459 { | |
460 int i, shift; | |
461 | |
462 for(i=0;i<3;i++) { | |
463 shift = (i == 0) ? 0 : 1; | |
464 component_resample(s, output->data[i], output->linesize[i], | |
465 s->owidth >> shift, s->oheight >> shift, | |
466 input->data[i], input->linesize[i], | |
467 s->iwidth >> shift, s->iheight >> shift); | |
468 } | |
469 } | |
470 | |
471 void img_resample_close(ImgReSampleContext *s) | |
472 { | |
396
fce0a2520551
removed useless header includes - use av memory functions
glantau
parents:
18
diff
changeset
|
473 av_free(s->line_buf); |
fce0a2520551
removed useless header includes - use av memory functions
glantau
parents:
18
diff
changeset
|
474 av_free(s); |
0 | 475 } |
476 | |
477 #ifdef TEST | |
478 | |
479 void *av_mallocz(int size) | |
480 { | |
481 void *ptr; | |
482 ptr = malloc(size); | |
483 memset(ptr, 0, size); | |
484 return ptr; | |
485 } | |
486 | |
487 /* input */ | |
488 #define XSIZE 256 | |
489 #define YSIZE 256 | |
490 UINT8 img[XSIZE * YSIZE]; | |
491 | |
492 /* output */ | |
493 #define XSIZE1 512 | |
494 #define YSIZE1 512 | |
495 UINT8 img1[XSIZE1 * YSIZE1]; | |
496 UINT8 img2[XSIZE1 * YSIZE1]; | |
497 | |
498 void save_pgm(const char *filename, UINT8 *img, int xsize, int ysize) | |
499 { | |
500 FILE *f; | |
501 f=fopen(filename,"w"); | |
502 fprintf(f,"P5\n%d %d\n%d\n", xsize, ysize, 255); | |
503 fwrite(img,1, xsize * ysize,f); | |
504 fclose(f); | |
505 } | |
506 | |
507 static void dump_filter(INT16 *filter) | |
508 { | |
509 int i, ph; | |
510 | |
511 for(ph=0;ph<NB_PHASES;ph++) { | |
512 printf("%2d: ", ph); | |
513 for(i=0;i<NB_TAPS;i++) { | |
514 printf(" %5.2f", filter[ph * NB_TAPS + i] / 256.0); | |
515 } | |
516 printf("\n"); | |
517 } | |
518 } | |
519 | |
2 | 520 #ifdef HAVE_MMX |
412 | 521 extern int mm_flags; |
0 | 522 #endif |
523 | |
524 int main(int argc, char **argv) | |
525 { | |
526 int x, y, v, i, xsize, ysize; | |
527 ImgReSampleContext *s; | |
528 float fact, factors[] = { 1/2.0, 3.0/4.0, 1.0, 4.0/3.0, 16.0/9.0, 2.0 }; | |
529 char buf[256]; | |
530 | |
531 /* build test image */ | |
532 for(y=0;y<YSIZE;y++) { | |
533 for(x=0;x<XSIZE;x++) { | |
534 if (x < XSIZE/2 && y < YSIZE/2) { | |
535 if (x < XSIZE/4 && y < YSIZE/4) { | |
536 if ((x % 10) <= 6 && | |
537 (y % 10) <= 6) | |
538 v = 0xff; | |
539 else | |
540 v = 0x00; | |
541 } else if (x < XSIZE/4) { | |
542 if (x & 1) | |
543 v = 0xff; | |
544 else | |
545 v = 0; | |
546 } else if (y < XSIZE/4) { | |
547 if (y & 1) | |
548 v = 0xff; | |
549 else | |
550 v = 0; | |
551 } else { | |
552 if (y < YSIZE*3/8) { | |
553 if ((y+x) & 1) | |
554 v = 0xff; | |
555 else | |
556 v = 0; | |
557 } else { | |
558 if (((x+3) % 4) <= 1 && | |
559 ((y+3) % 4) <= 1) | |
560 v = 0xff; | |
561 else | |
562 v = 0x00; | |
563 } | |
564 } | |
565 } else if (x < XSIZE/2) { | |
566 v = ((x - (XSIZE/2)) * 255) / (XSIZE/2); | |
567 } else if (y < XSIZE/2) { | |
568 v = ((y - (XSIZE/2)) * 255) / (XSIZE/2); | |
569 } else { | |
570 v = ((x + y - XSIZE) * 255) / XSIZE; | |
571 } | |
572 img[y * XSIZE + x] = v; | |
573 } | |
574 } | |
575 save_pgm("/tmp/in.pgm", img, XSIZE, YSIZE); | |
576 for(i=0;i<sizeof(factors)/sizeof(float);i++) { | |
577 fact = factors[i]; | |
578 xsize = (int)(XSIZE * fact); | |
579 ysize = (int)(YSIZE * fact); | |
580 s = img_resample_init(xsize, ysize, XSIZE, YSIZE); | |
581 printf("Factor=%0.2f\n", fact); | |
582 dump_filter(&s->h_filters[0][0]); | |
583 component_resample(s, img1, xsize, xsize, ysize, | |
584 img, XSIZE, XSIZE, YSIZE); | |
585 img_resample_close(s); | |
586 | |
587 sprintf(buf, "/tmp/out%d.pgm", i); | |
588 save_pgm(buf, img1, xsize, ysize); | |
589 } | |
590 | |
591 /* mmx test */ | |
2 | 592 #ifdef HAVE_MMX |
0 | 593 printf("MMX test\n"); |
594 fact = 0.72; | |
595 xsize = (int)(XSIZE * fact); | |
596 ysize = (int)(YSIZE * fact); | |
597 mm_flags = MM_MMX; | |
598 s = img_resample_init(xsize, ysize, XSIZE, YSIZE); | |
599 component_resample(s, img1, xsize, xsize, ysize, | |
600 img, XSIZE, XSIZE, YSIZE); | |
601 | |
602 mm_flags = 0; | |
603 s = img_resample_init(xsize, ysize, XSIZE, YSIZE); | |
604 component_resample(s, img2, xsize, xsize, ysize, | |
605 img, XSIZE, XSIZE, YSIZE); | |
606 if (memcmp(img1, img2, xsize * ysize) != 0) { | |
607 fprintf(stderr, "mmx error\n"); | |
608 exit(1); | |
609 } | |
610 printf("MMX OK\n"); | |
611 #endif | |
612 return 0; | |
613 } | |
614 | |
615 #endif |