Fix linking: emms_c is a macro, #include the correct header file.
[libav.git] / libavformat / grab.c
1 /*
2 * Linux video grab interface
3 * Copyright (c) 2000,2001 Fabrice Bellard.
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21 #include "avformat.h"
22 #include "dsputil.h"
23 #include <unistd.h>
24 #include <fcntl.h>
25 #include <sys/ioctl.h>
26 #include <sys/mman.h>
27 #include <sys/time.h>
28 #define _LINUX_TIME_H 1
29 #include <linux/videodev.h>
30 #include <time.h>
31
32 typedef struct {
33 int fd;
34 int frame_format; /* see VIDEO_PALETTE_xxx */
35 int use_mmap;
36 int width, height;
37 int frame_rate;
38 int frame_rate_base;
39 int64_t time_frame;
40 int frame_size;
41 struct video_capability video_cap;
42 struct video_audio audio_saved;
43 uint8_t *video_buf;
44 struct video_mbuf gb_buffers;
45 struct video_mmap gb_buf;
46 int gb_frame;
47
48 /* ATI All In Wonder specific stuff */
49 /* XXX: remove and merge in libavcodec/imgconvert.c */
50 int aiw_enabled;
51 int deint;
52 int halfw;
53 uint8_t *src_mem;
54 uint8_t *lum_m4_mem;
55 } VideoData;
56
57 struct {
58 int palette;
59 int depth;
60 enum PixelFormat pix_fmt;
61 } video_formats [] = {
62 {.palette = VIDEO_PALETTE_YUV420P, .depth = 12, .pix_fmt = PIX_FMT_YUV420P },
63 {.palette = VIDEO_PALETTE_YUV422, .depth = 16, .pix_fmt = PIX_FMT_YUYV422 },
64 {.palette = VIDEO_PALETTE_UYVY, .depth = 16, .pix_fmt = PIX_FMT_UYVY422 },
65 {.palette = VIDEO_PALETTE_YUYV, .depth = 16, .pix_fmt = PIX_FMT_YUYV422 },
66 /* NOTE: v4l uses BGR24, not RGB24 */
67 {.palette = VIDEO_PALETTE_RGB24, .depth = 24, .pix_fmt = PIX_FMT_BGR24 },
68 {.palette = VIDEO_PALETTE_RGB565, .depth = 16, .pix_fmt = PIX_FMT_BGR565 },
69 {.palette = VIDEO_PALETTE_GREY, .depth = 8, .pix_fmt = PIX_FMT_GRAY8 },
70 };
71
72
73 static int aiw_init(VideoData *s);
74 static int aiw_read_picture(VideoData *s, uint8_t *data);
75 static int aiw_close(VideoData *s);
76
77 static int grab_read_header(AVFormatContext *s1, AVFormatParameters *ap)
78 {
79 VideoData *s = s1->priv_data;
80 AVStream *st;
81 int width, height;
82 int video_fd, frame_size;
83 int ret, frame_rate, frame_rate_base;
84 int desired_palette, desired_depth;
85 struct video_tuner tuner;
86 struct video_audio audio;
87 struct video_picture pict;
88 int j;
89 int vformat_num = sizeof(video_formats) / sizeof(video_formats[0]);
90
91 if (ap->width <= 0 || ap->height <= 0 || ap->time_base.den <= 0) {
92 av_log(s1, AV_LOG_ERROR, "Bad capture size (%dx%d) or wrong time base (%d)\n",
93 ap->width, ap->height, ap->time_base.den);
94
95 return -1;
96 }
97
98 width = ap->width;
99 height = ap->height;
100 frame_rate = ap->time_base.den;
101 frame_rate_base = ap->time_base.num;
102
103 if((unsigned)width > 32767 || (unsigned)height > 32767) {
104 av_log(s1, AV_LOG_ERROR, "Capture size is out of range: %dx%d\n",
105 width, height);
106
107 return -1;
108 }
109
110 st = av_new_stream(s1, 0);
111 if (!st)
112 return AVERROR(ENOMEM);
113 av_set_pts_info(st, 64, 1, 1000000); /* 64 bits pts in us */
114
115 s->width = width;
116 s->height = height;
117 s->frame_rate = frame_rate;
118 s->frame_rate_base = frame_rate_base;
119
120 video_fd = open(s1->filename, O_RDWR);
121 if (video_fd < 0) {
122 perror(s1->filename);
123 goto fail;
124 }
125
126 if (ioctl(video_fd,VIDIOCGCAP, &s->video_cap) < 0) {
127 perror("VIDIOCGCAP");
128 goto fail;
129 }
130
131 if (!(s->video_cap.type & VID_TYPE_CAPTURE)) {
132 av_log(s1, AV_LOG_ERROR, "Fatal: grab device does not handle capture\n");
133 goto fail;
134 }
135
136 desired_palette = -1;
137 desired_depth = -1;
138 for (j = 0; j < vformat_num; j++) {
139 if (ap->pix_fmt == video_formats[j].pix_fmt) {
140 desired_palette = video_formats[j].palette;
141 desired_depth = video_formats[j].depth;
142 break;
143 }
144 }
145
146 /* set tv standard */
147 if (ap->standard && !ioctl(video_fd, VIDIOCGTUNER, &tuner)) {
148 if (!strcasecmp(ap->standard, "pal"))
149 tuner.mode = VIDEO_MODE_PAL;
150 else if (!strcasecmp(ap->standard, "secam"))
151 tuner.mode = VIDEO_MODE_SECAM;
152 else
153 tuner.mode = VIDEO_MODE_NTSC;
154 ioctl(video_fd, VIDIOCSTUNER, &tuner);
155 }
156
157 /* unmute audio */
158 audio.audio = 0;
159 ioctl(video_fd, VIDIOCGAUDIO, &audio);
160 memcpy(&s->audio_saved, &audio, sizeof(audio));
161 audio.flags &= ~VIDEO_AUDIO_MUTE;
162 ioctl(video_fd, VIDIOCSAUDIO, &audio);
163
164 ioctl(video_fd, VIDIOCGPICT, &pict);
165 #if 0
166 printf("v4l: colour=%d hue=%d brightness=%d constrast=%d whiteness=%d\n",
167 pict.colour,
168 pict.hue,
169 pict.brightness,
170 pict.contrast,
171 pict.whiteness);
172 #endif
173 /* try to choose a suitable video format */
174 pict.palette = desired_palette;
175 pict.depth= desired_depth;
176 if (desired_palette == -1 || (ret = ioctl(video_fd, VIDIOCSPICT, &pict)) < 0) {
177 for (j = 0; j < vformat_num; j++) {
178 pict.palette = video_formats[j].palette;
179 pict.depth = video_formats[j].depth;
180 if (-1 != ioctl(video_fd, VIDIOCSPICT, &pict))
181 break;
182 }
183 if (j >= vformat_num)
184 goto fail1;
185 }
186
187 ret = ioctl(video_fd,VIDIOCGMBUF,&s->gb_buffers);
188 if (ret < 0) {
189 /* try to use read based access */
190 struct video_window win;
191 int val;
192
193 win.x = 0;
194 win.y = 0;
195 win.width = width;
196 win.height = height;
197 win.chromakey = -1;
198 win.flags = 0;
199
200 ioctl(video_fd, VIDIOCSWIN, &win);
201
202 s->frame_format = pict.palette;
203
204 val = 1;
205 ioctl(video_fd, VIDIOCCAPTURE, &val);
206
207 s->time_frame = av_gettime() * s->frame_rate / s->frame_rate_base;
208 s->use_mmap = 0;
209
210 /* ATI All In Wonder automatic activation */
211 if (!strcmp(s->video_cap.name, "Km")) {
212 if (aiw_init(s) < 0)
213 goto fail;
214 s->aiw_enabled = 1;
215 /* force 420P format because convertion from YUV422 to YUV420P
216 is done in this driver (ugly) */
217 s->frame_format = VIDEO_PALETTE_YUV420P;
218 }
219 } else {
220 s->video_buf = mmap(0,s->gb_buffers.size,PROT_READ|PROT_WRITE,MAP_SHARED,video_fd,0);
221 if ((unsigned char*)-1 == s->video_buf) {
222 s->video_buf = mmap(0,s->gb_buffers.size,PROT_READ|PROT_WRITE,MAP_PRIVATE,video_fd,0);
223 if ((unsigned char*)-1 == s->video_buf) {
224 perror("mmap");
225 goto fail;
226 }
227 }
228 s->gb_frame = 0;
229 s->time_frame = av_gettime() * s->frame_rate / s->frame_rate_base;
230
231 /* start to grab the first frame */
232 s->gb_buf.frame = s->gb_frame % s->gb_buffers.frames;
233 s->gb_buf.height = height;
234 s->gb_buf.width = width;
235 s->gb_buf.format = pict.palette;
236
237 ret = ioctl(video_fd, VIDIOCMCAPTURE, &s->gb_buf);
238 if (ret < 0) {
239 if (errno != EAGAIN) {
240 fail1:
241 av_log(s1, AV_LOG_ERROR, "Fatal: grab device does not support suitable format\n");
242 } else {
243 av_log(s1, AV_LOG_ERROR,"Fatal: grab device does not receive any video signal\n");
244 }
245 goto fail;
246 }
247 for (j = 1; j < s->gb_buffers.frames; j++) {
248 s->gb_buf.frame = j;
249 ioctl(video_fd, VIDIOCMCAPTURE, &s->gb_buf);
250 }
251 s->frame_format = s->gb_buf.format;
252 s->use_mmap = 1;
253 }
254
255 for (j = 0; j < vformat_num; j++) {
256 if (s->frame_format == video_formats[j].palette) {
257 frame_size = width * height * video_formats[j].depth / 8;
258 st->codec->pix_fmt = video_formats[j].pix_fmt;
259 break;
260 }
261 }
262
263 if (j >= vformat_num)
264 goto fail;
265
266 s->fd = video_fd;
267 s->frame_size = frame_size;
268
269 st->codec->codec_type = CODEC_TYPE_VIDEO;
270 st->codec->codec_id = CODEC_ID_RAWVIDEO;
271 st->codec->width = width;
272 st->codec->height = height;
273 st->codec->time_base.den = frame_rate;
274 st->codec->time_base.num = frame_rate_base;
275 st->codec->bit_rate = frame_size * 1/av_q2d(st->codec->time_base) * 8;
276
277 return 0;
278 fail:
279 if (video_fd >= 0)
280 close(video_fd);
281 av_free(st);
282 return AVERROR_IO;
283 }
284
285 static int v4l_mm_read_picture(VideoData *s, uint8_t *buf)
286 {
287 uint8_t *ptr;
288
289 while (ioctl(s->fd, VIDIOCSYNC, &s->gb_frame) < 0 &&
290 (errno == EAGAIN || errno == EINTR));
291
292 ptr = s->video_buf + s->gb_buffers.offsets[s->gb_frame];
293 memcpy(buf, ptr, s->frame_size);
294
295 /* Setup to capture the next frame */
296 s->gb_buf.frame = s->gb_frame;
297 if (ioctl(s->fd, VIDIOCMCAPTURE, &s->gb_buf) < 0) {
298 if (errno == EAGAIN)
299 av_log(NULL, AV_LOG_ERROR, "Cannot Sync\n");
300 else
301 perror("VIDIOCMCAPTURE");
302 return AVERROR_IO;
303 }
304
305 /* This is now the grabbing frame */
306 s->gb_frame = (s->gb_frame + 1) % s->gb_buffers.frames;
307
308 return s->frame_size;
309 }
310
311 static int grab_read_packet(AVFormatContext *s1, AVPacket *pkt)
312 {
313 VideoData *s = s1->priv_data;
314 int64_t curtime, delay;
315 struct timespec ts;
316
317 /* Calculate the time of the next frame */
318 s->time_frame += INT64_C(1000000);
319
320 /* wait based on the frame rate */
321 for(;;) {
322 curtime = av_gettime();
323 delay = s->time_frame * s->frame_rate_base / s->frame_rate - curtime;
324 if (delay <= 0) {
325 if (delay < INT64_C(-1000000) * s->frame_rate_base / s->frame_rate) {
326 /* printf("grabbing is %d frames late (dropping)\n", (int) -(delay / 16666)); */
327 s->time_frame += INT64_C(1000000);
328 }
329 break;
330 }
331 ts.tv_sec = delay / 1000000;
332 ts.tv_nsec = (delay % 1000000) * 1000;
333 nanosleep(&ts, NULL);
334 }
335
336 if (av_new_packet(pkt, s->frame_size) < 0)
337 return AVERROR_IO;
338
339 pkt->pts = curtime;
340
341 /* read one frame */
342 if (s->aiw_enabled) {
343 return aiw_read_picture(s, pkt->data);
344 } else if (s->use_mmap) {
345 return v4l_mm_read_picture(s, pkt->data);
346 } else {
347 if (read(s->fd, pkt->data, pkt->size) != pkt->size)
348 return AVERROR_IO;
349 return s->frame_size;
350 }
351 }
352
353 static int grab_read_close(AVFormatContext *s1)
354 {
355 VideoData *s = s1->priv_data;
356
357 if (s->aiw_enabled)
358 aiw_close(s);
359
360 if (s->use_mmap)
361 munmap(s->video_buf, s->gb_buffers.size);
362
363 /* mute audio. we must force it because the BTTV driver does not
364 return its state correctly */
365 s->audio_saved.flags |= VIDEO_AUDIO_MUTE;
366 ioctl(s->fd, VIDIOCSAUDIO, &s->audio_saved);
367
368 close(s->fd);
369 return 0;
370 }
371
372 AVInputFormat video_grab_v4l_demuxer = {
373 "video4linux",
374 "video grab",
375 sizeof(VideoData),
376 NULL,
377 grab_read_header,
378 grab_read_packet,
379 grab_read_close,
380 .flags = AVFMT_NOFILE,
381 };
382
383 /* All in Wonder specific stuff */
384 /* XXX: remove and merge in libavcodec/imgconvert.c */
385
386 static int aiw_init(VideoData *s)
387 {
388 int width, height;
389
390 width = s->width;
391 height = s->height;
392
393 if ((width == s->video_cap.maxwidth && height == s->video_cap.maxheight) ||
394 (width == s->video_cap.maxwidth && height == s->video_cap.maxheight*2) ||
395 (width == s->video_cap.maxwidth/2 && height == s->video_cap.maxheight)) {
396
397 s->deint=0;
398 s->halfw=0;
399 if (height == s->video_cap.maxheight*2) s->deint=1;
400 if (width == s->video_cap.maxwidth/2) s->halfw=1;
401 } else {
402 av_log(NULL, AV_LOG_ERROR, "\nIncorrect Grab Size Supplied - Supported Sizes Are:\n");
403 av_log(NULL, AV_LOG_ERROR, " %dx%d %dx%d %dx%d\n\n",
404 s->video_cap.maxwidth,s->video_cap.maxheight,
405 s->video_cap.maxwidth,s->video_cap.maxheight*2,
406 s->video_cap.maxwidth/2,s->video_cap.maxheight);
407 goto fail;
408 }
409
410 if (s->halfw == 0) {
411 s->src_mem = av_malloc(s->width*2);
412 } else {
413 s->src_mem = av_malloc(s->width*4);
414 }
415 if (!s->src_mem) goto fail;
416
417 s->lum_m4_mem = av_malloc(s->width);
418 if (!s->lum_m4_mem)
419 goto fail;
420 return 0;
421 fail:
422 av_freep(&s->src_mem);
423 av_freep(&s->lum_m4_mem);
424 return -1;
425 }
426
427 #ifdef HAVE_MMX
428 #include "i386/mmx.h"
429
430 #define LINE_WITH_UV \
431 movq_m2r(ptr[0],mm0); \
432 movq_m2r(ptr[8],mm1); \
433 movq_r2r(mm0, mm4); \
434 punpcklbw_r2r(mm1,mm0); \
435 punpckhbw_r2r(mm1,mm4); \
436 movq_r2r(mm0,mm5); \
437 punpcklbw_r2r(mm4,mm0); \
438 punpckhbw_r2r(mm4,mm5); \
439 movq_r2r(mm0,mm1); \
440 punpcklbw_r2r(mm5,mm1); \
441 movq_r2m(mm1,lum[0]); \
442 movq_m2r(ptr[16],mm2); \
443 movq_m2r(ptr[24],mm1); \
444 movq_r2r(mm2,mm4); \
445 punpcklbw_r2r(mm1,mm2); \
446 punpckhbw_r2r(mm1,mm4); \
447 movq_r2r(mm2,mm3); \
448 punpcklbw_r2r(mm4,mm2); \
449 punpckhbw_r2r(mm4,mm3); \
450 movq_r2r(mm2,mm1); \
451 punpcklbw_r2r(mm3,mm1); \
452 movq_r2m(mm1,lum[8]); \
453 punpckhdq_r2r(mm2,mm0); \
454 punpckhdq_r2r(mm3,mm5); \
455 movq_r2m(mm0,cb[0]); \
456 movq_r2m(mm5,cr[0]);
457
458 #define LINE_NO_UV \
459 movq_m2r(ptr[0],mm0);\
460 movq_m2r(ptr[8],mm1);\
461 movq_r2r(mm0, mm4);\
462 punpcklbw_r2r(mm1,mm0); \
463 punpckhbw_r2r(mm1,mm4);\
464 movq_r2r(mm0,mm5);\
465 punpcklbw_r2r(mm4,mm0);\
466 punpckhbw_r2r(mm4,mm5);\
467 movq_r2r(mm0,mm1);\
468 punpcklbw_r2r(mm5,mm1);\
469 movq_r2m(mm1,lum[0]);\
470 movq_m2r(ptr[16],mm2);\
471 movq_m2r(ptr[24],mm1);\
472 movq_r2r(mm2,mm4);\
473 punpcklbw_r2r(mm1,mm2);\
474 punpckhbw_r2r(mm1,mm4);\
475 movq_r2r(mm2,mm3);\
476 punpcklbw_r2r(mm4,mm2);\
477 punpckhbw_r2r(mm4,mm3);\
478 movq_r2r(mm2,mm1);\
479 punpcklbw_r2r(mm3,mm1);\
480 movq_r2m(mm1,lum[8]);
481
482 #define LINE_WITHUV_AVG \
483 movq_m2r(ptr[0], mm0);\
484 movq_m2r(ptr[8], mm1);\
485 movq_r2r(mm0, mm4);\
486 punpcklbw_r2r(mm1,mm0);\
487 punpckhbw_r2r(mm1,mm4);\
488 movq_r2r(mm0,mm5);\
489 punpcklbw_r2r(mm4,mm0);\
490 punpckhbw_r2r(mm4,mm5);\
491 movq_r2r(mm0,mm1);\
492 movq_r2r(mm5,mm2);\
493 punpcklbw_r2r(mm7,mm1);\
494 punpcklbw_r2r(mm7,mm2);\
495 paddw_r2r(mm6,mm1);\
496 paddw_r2r(mm2,mm1);\
497 psraw_i2r(1,mm1);\
498 packuswb_r2r(mm7,mm1);\
499 movd_r2m(mm1,lum[0]);\
500 movq_m2r(ptr[16],mm2);\
501 movq_m2r(ptr[24],mm1);\
502 movq_r2r(mm2,mm4);\
503 punpcklbw_r2r(mm1,mm2);\
504 punpckhbw_r2r(mm1,mm4);\
505 movq_r2r(mm2,mm3);\
506 punpcklbw_r2r(mm4,mm2);\
507 punpckhbw_r2r(mm4,mm3);\
508 movq_r2r(mm2,mm1);\
509 movq_r2r(mm3,mm4);\
510 punpcklbw_r2r(mm7,mm1);\
511 punpcklbw_r2r(mm7,mm4);\
512 paddw_r2r(mm6,mm1);\
513 paddw_r2r(mm4,mm1);\
514 psraw_i2r(1,mm1);\
515 packuswb_r2r(mm7,mm1);\
516 movd_r2m(mm1,lum[4]);\
517 punpckhbw_r2r(mm7,mm0);\
518 punpckhbw_r2r(mm7,mm2);\
519 paddw_r2r(mm6,mm0);\
520 paddw_r2r(mm2,mm0);\
521 psraw_i2r(1,mm0);\
522 packuswb_r2r(mm7,mm0);\
523 punpckhbw_r2r(mm7,mm5);\
524 punpckhbw_r2r(mm7,mm3);\
525 paddw_r2r(mm6,mm5);\
526 paddw_r2r(mm3,mm5);\
527 psraw_i2r(1,mm5);\
528 packuswb_r2r(mm7,mm5);\
529 movd_r2m(mm0,cb[0]);\
530 movd_r2m(mm5,cr[0]);
531
532 #define LINE_NOUV_AVG \
533 movq_m2r(ptr[0],mm0);\
534 movq_m2r(ptr[8],mm1);\
535 pand_r2r(mm5,mm0);\
536 pand_r2r(mm5,mm1);\
537 pmaddwd_r2r(mm6,mm0);\
538 pmaddwd_r2r(mm6,mm1);\
539 packssdw_r2r(mm1,mm0);\
540 paddw_r2r(mm6,mm0);\
541 psraw_i2r(1,mm0);\
542 movq_m2r(ptr[16],mm2);\
543 movq_m2r(ptr[24],mm3);\
544 pand_r2r(mm5,mm2);\
545 pand_r2r(mm5,mm3);\
546 pmaddwd_r2r(mm6,mm2);\
547 pmaddwd_r2r(mm6,mm3);\
548 packssdw_r2r(mm3,mm2);\
549 paddw_r2r(mm6,mm2);\
550 psraw_i2r(1,mm2);\
551 packuswb_r2r(mm2,mm0);\
552 movq_r2m(mm0,lum[0]);
553
554 #define DEINT_LINE_LUM(ptroff) \
555 movd_m2r(lum_m4[(ptroff)],mm0);\
556 movd_m2r(lum_m3[(ptroff)],mm1);\
557 movd_m2r(lum_m2[(ptroff)],mm2);\
558 movd_m2r(lum_m1[(ptroff)],mm3);\
559 movd_m2r(lum[(ptroff)],mm4);\
560 punpcklbw_r2r(mm7,mm0);\
561 movd_r2m(mm2,lum_m4[(ptroff)]);\
562 punpcklbw_r2r(mm7,mm1);\
563 punpcklbw_r2r(mm7,mm2);\
564 punpcklbw_r2r(mm7,mm3);\
565 punpcklbw_r2r(mm7,mm4);\
566 psllw_i2r(2,mm1);\
567 psllw_i2r(1,mm2);\
568 paddw_r2r(mm6,mm1);\
569 psllw_i2r(2,mm3);\
570 paddw_r2r(mm2,mm1);\
571 paddw_r2r(mm4,mm0);\
572 paddw_r2r(mm3,mm1);\
573 psubusw_r2r(mm0,mm1);\
574 psrlw_i2r(3,mm1);\
575 packuswb_r2r(mm7,mm1);\
576 movd_r2m(mm1,lum_m2[(ptroff)]);
577
578 #else
579 #include "dsputil.h"
580
581 #define LINE_WITH_UV \
582 lum[0]=ptr[0];lum[1]=ptr[2];lum[2]=ptr[4];lum[3]=ptr[6];\
583 cb[0]=ptr[1];cb[1]=ptr[5];\
584 cr[0]=ptr[3];cr[1]=ptr[7];\
585 lum[4]=ptr[8];lum[5]=ptr[10];lum[6]=ptr[12];lum[7]=ptr[14];\
586 cb[2]=ptr[9];cb[3]=ptr[13];\
587 cr[2]=ptr[11];cr[3]=ptr[15];\
588 lum[8]=ptr[16];lum[9]=ptr[18];lum[10]=ptr[20];lum[11]=ptr[22];\
589 cb[4]=ptr[17];cb[5]=ptr[21];\
590 cr[4]=ptr[19];cr[5]=ptr[23];\
591 lum[12]=ptr[24];lum[13]=ptr[26];lum[14]=ptr[28];lum[15]=ptr[30];\
592 cb[6]=ptr[25];cb[7]=ptr[29];\
593 cr[6]=ptr[27];cr[7]=ptr[31];
594
595 #define LINE_NO_UV \
596 lum[0]=ptr[0];lum[1]=ptr[2];lum[2]=ptr[4];lum[3]=ptr[6];\
597 lum[4]=ptr[8];lum[5]=ptr[10];lum[6]=ptr[12];lum[7]=ptr[14];\
598 lum[8]=ptr[16];lum[9]=ptr[18];lum[10]=ptr[20];lum[11]=ptr[22];\
599 lum[12]=ptr[24];lum[13]=ptr[26];lum[14]=ptr[28];lum[15]=ptr[30];
600
601 #define LINE_WITHUV_AVG \
602 sum=(ptr[0]+ptr[2]+1) >> 1;lum[0]=sum; \
603 sum=(ptr[4]+ptr[6]+1) >> 1;lum[1]=sum; \
604 sum=(ptr[1]+ptr[5]+1) >> 1;cb[0]=sum; \
605 sum=(ptr[3]+ptr[7]+1) >> 1;cr[0]=sum; \
606 sum=(ptr[8]+ptr[10]+1) >> 1;lum[2]=sum; \
607 sum=(ptr[12]+ptr[14]+1) >> 1;lum[3]=sum; \
608 sum=(ptr[9]+ptr[13]+1) >> 1;cb[1]=sum; \
609 sum=(ptr[11]+ptr[15]+1) >> 1;cr[1]=sum; \
610 sum=(ptr[16]+ptr[18]+1) >> 1;lum[4]=sum; \
611 sum=(ptr[20]+ptr[22]+1) >> 1;lum[5]=sum; \
612 sum=(ptr[17]+ptr[21]+1) >> 1;cb[2]=sum; \
613 sum=(ptr[19]+ptr[23]+1) >> 1;cr[2]=sum; \
614 sum=(ptr[24]+ptr[26]+1) >> 1;lum[6]=sum; \
615 sum=(ptr[28]+ptr[30]+1) >> 1;lum[7]=sum; \
616 sum=(ptr[25]+ptr[29]+1) >> 1;cb[3]=sum; \
617 sum=(ptr[27]+ptr[31]+1) >> 1;cr[3]=sum;
618
619 #define LINE_NOUV_AVG \
620 sum=(ptr[0]+ptr[2]+1) >> 1;lum[0]=sum; \
621 sum=(ptr[4]+ptr[6]+1) >> 1;lum[1]=sum; \
622 sum=(ptr[8]+ptr[10]+1) >> 1;lum[2]=sum; \
623 sum=(ptr[12]+ptr[14]+1) >> 1;lum[3]=sum; \
624 sum=(ptr[16]+ptr[18]+1) >> 1;lum[4]=sum; \
625 sum=(ptr[20]+ptr[22]+1) >> 1;lum[5]=sum; \
626 sum=(ptr[24]+ptr[26]+1) >> 1;lum[6]=sum; \
627 sum=(ptr[28]+ptr[30]+1) >> 1;lum[7]=sum;
628
629 #define DEINT_LINE_LUM(ptroff) \
630 sum=(-lum_m4[(ptroff)]+(lum_m3[(ptroff)]<<2)+(lum_m2[(ptroff)]<<1)+(lum_m1[(ptroff)]<<2)-lum[(ptroff)]); \
631 lum_m4[(ptroff)]=lum_m2[(ptroff)];\
632 lum_m2[(ptroff)]=cm[(sum+4)>>3];\
633 sum=(-lum_m4[(ptroff)+1]+(lum_m3[(ptroff)+1]<<2)+(lum_m2[(ptroff)+1]<<1)+(lum_m1[(ptroff)+1]<<2)-lum[(ptroff)+1]); \
634 lum_m4[(ptroff)+1]=lum_m2[(ptroff)+1];\
635 lum_m2[(ptroff)+1]=cm[(sum+4)>>3];\
636 sum=(-lum_m4[(ptroff)+2]+(lum_m3[(ptroff)+2]<<2)+(lum_m2[(ptroff)+2]<<1)+(lum_m1[(ptroff)+2]<<2)-lum[(ptroff)+2]); \
637 lum_m4[(ptroff)+2]=lum_m2[(ptroff)+2];\
638 lum_m2[(ptroff)+2]=cm[(sum+4)>>3];\
639 sum=(-lum_m4[(ptroff)+3]+(lum_m3[(ptroff)+3]<<2)+(lum_m2[(ptroff)+3]<<1)+(lum_m1[(ptroff)+3]<<2)-lum[(ptroff)+3]); \
640 lum_m4[(ptroff)+3]=lum_m2[(ptroff)+3];\
641 lum_m2[(ptroff)+3]=cm[(sum+4)>>3];
642
643 #endif
644
645
646 /* Read two fields separately. */
647 static int aiw_read_picture(VideoData *s, uint8_t *data)
648 {
649 uint8_t *ptr, *lum, *cb, *cr;
650 int h;
651 #ifndef HAVE_MMX
652 int sum;
653 #endif
654 uint8_t* src = s->src_mem;
655 uint8_t *ptrend = &src[s->width*2];
656 lum=data;
657 cb=&lum[s->width*s->height];
658 cr=&cb[(s->width*s->height)/4];
659 if (s->deint == 0 && s->halfw == 0) {
660 while (read(s->fd,src,s->width*2) < 0) {
661 usleep(100);
662 }
663 for (h = 0; h < s->height-2; h+=2) {
664 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
665 LINE_WITH_UV
666 }
667 read(s->fd,src,s->width*2);
668 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16) {
669 LINE_NO_UV
670 }
671 read(s->fd,src,s->width*2);
672 }
673 /*
674 * Do last two lines
675 */
676 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
677 LINE_WITH_UV
678 }
679 read(s->fd,src,s->width*2);
680 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16) {
681 LINE_NO_UV
682 }
683 /* drop second field */
684 while (read(s->fd,src,s->width*2) < 0) {
685 usleep(100);
686 }
687 for (h = 0; h < s->height - 1; h++) {
688 read(s->fd,src,s->width*2);
689 }
690 } else if (s->halfw == 1) {
691 #ifdef HAVE_MMX
692 mmx_t rounder;
693 mmx_t masker;
694 rounder.uw[0]=1;
695 rounder.uw[1]=1;
696 rounder.uw[2]=1;
697 rounder.uw[3]=1;
698 masker.ub[0]=0xff;
699 masker.ub[1]=0;
700 masker.ub[2]=0xff;
701 masker.ub[3]=0;
702 masker.ub[4]=0xff;
703 masker.ub[5]=0;
704 masker.ub[6]=0xff;
705 masker.ub[7]=0;
706 pxor_r2r(mm7,mm7);
707 movq_m2r(rounder,mm6);
708 #endif
709 while (read(s->fd,src,s->width*4) < 0) {
710 usleep(100);
711 }
712 ptrend = &src[s->width*4];
713 for (h = 0; h < s->height-2; h+=2) {
714 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=8, cb+=4, cr+=4) {
715 LINE_WITHUV_AVG
716 }
717 read(s->fd,src,s->width*4);
718 #ifdef HAVE_MMX
719 movq_m2r(masker,mm5);
720 #endif
721 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=8) {
722 LINE_NOUV_AVG
723 }
724 read(s->fd,src,s->width*4);
725 }
726 /*
727 * Do last two lines
728 */
729 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=8, cb+=4, cr+=4) {
730 LINE_WITHUV_AVG
731 }
732 read(s->fd,src,s->width*4);
733 #ifdef HAVE_MMX
734 movq_m2r(masker,mm5);
735 #endif
736 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=8) {
737 LINE_NOUV_AVG
738 }
739 /* drop second field */
740 while (read(s->fd,src,s->width*4) < 0) {
741 usleep(100);
742 }
743 for (h = 0; h < s->height - 1; h++) {
744 read(s->fd,src,s->width*4);
745 }
746 } else {
747 uint8_t *lum_m1, *lum_m2, *lum_m3, *lum_m4;
748 #ifdef HAVE_MMX
749 mmx_t rounder;
750 rounder.uw[0]=4;
751 rounder.uw[1]=4;
752 rounder.uw[2]=4;
753 rounder.uw[3]=4;
754 movq_m2r(rounder,mm6);
755 pxor_r2r(mm7,mm7);
756 #else
757 uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
758 #endif
759
760 /* read two fields and deinterlace them */
761 while (read(s->fd,src,s->width*2) < 0) {
762 usleep(100);
763 }
764 for (h = 0; h < (s->height/2)-2; h+=2) {
765 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
766 LINE_WITH_UV
767 }
768 read(s->fd,src,s->width*2);
769 /* skip a luminance line - will be filled in later */
770 lum += s->width;
771 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
772 LINE_WITH_UV
773 }
774 /* skip a luminance line - will be filled in later */
775 lum += s->width;
776 read(s->fd,src,s->width*2);
777 }
778 /*
779 * Do last two lines
780 */
781 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
782 LINE_WITH_UV
783 }
784 /* skip a luminance line - will be filled in later */
785 lum += s->width;
786 read(s->fd,src,s->width*2);
787 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, cb+=8, cr+=8) {
788 LINE_WITH_UV
789 }
790 /*
791 *
792 * SECOND FIELD
793 *
794 */
795 lum=&data[s->width];
796 while (read(s->fd,src,s->width*2) < 0) {
797 usleep(10);
798 }
799 /* First (and last) two lines not interlaced */
800 for (h = 0; h < 2; h++) {
801 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16) {
802 LINE_NO_UV
803 }
804 read(s->fd,src,s->width*2);
805 /* skip a luminance line */
806 lum += s->width;
807 }
808 lum_m1=&lum[-s->width];
809 lum_m2=&lum_m1[-s->width];
810 lum_m3=&lum_m2[-s->width];
811 memmove(s->lum_m4_mem,&lum_m3[-s->width],s->width);
812 for (; h < (s->height/2)-1; h++) {
813 lum_m4=s->lum_m4_mem;
814 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16,lum_m1+=16,lum_m2+=16,lum_m3+=16,lum_m4+=16) {
815 LINE_NO_UV
816
817 DEINT_LINE_LUM(0)
818 DEINT_LINE_LUM(4)
819 DEINT_LINE_LUM(8)
820 DEINT_LINE_LUM(12)
821 }
822 read(s->fd,src,s->width*2);
823 /* skip a luminance line */
824 lum += s->width;
825 lum_m1 += s->width;
826 lum_m2 += s->width;
827 lum_m3 += s->width;
828 // lum_m4 += s->width;
829 }
830 /*
831 * Do last line
832 */
833 lum_m4=s->lum_m4_mem;
834 for (ptr = &src[0]; ptr < ptrend; ptr+=32, lum+=16, lum_m1+=16, lum_m2+=16, lum_m3+=16, lum_m4+=16) {
835 LINE_NO_UV
836
837 DEINT_LINE_LUM(0)
838 DEINT_LINE_LUM(4)
839 DEINT_LINE_LUM(8)
840 DEINT_LINE_LUM(12)
841 }
842 }
843 emms_c();
844 return s->frame_size;
845 }
846
847 static int aiw_close(VideoData *s)
848 {
849 av_freep(&s->lum_m4_mem);
850 av_freep(&s->src_mem);
851 return 0;
852 }