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