native Indeo3 decoder implementation
[libav.git] / libavcodec / indeo3.c
1 /*
2 * Intel Indeo 3 (IV31, IV32, etc.) video decoder for ffmpeg
3 * written, produced, and directed by Alan Smithee
4 *
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.
9 *
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
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
18 */
19
20 #include <stdio.h>
21 #include <stdlib.h>
22 #include <string.h>
23 #include <unistd.h>
24
25 #include "common.h"
26 #include "avcodec.h"
27 #include "dsputil.h"
28 #include "mpegvideo.h"
29 #include "bswap.h"
30
31 #include "indeo3data.h"
32
33 typedef struct
34 {
35 unsigned char *Ybuf;
36 unsigned char *Ubuf;
37 unsigned char *Vbuf;
38 unsigned char *the_buf;
39 unsigned int the_buf_size;
40 unsigned short y_w, y_h;
41 unsigned short uv_w, uv_h;
42 } YUVBufs;
43
44 typedef struct Indeo3DecodeContext {
45 AVCodecContext *avctx;
46 int width, height;
47 AVFrame frame;
48
49 YUVBufs iv_frame[2];
50 YUVBufs *cur_frame;
51 YUVBufs *ref_frame;
52
53 unsigned char *ModPred;
54 unsigned short *corrector_type;
55 } Indeo3DecodeContext;
56
57 static int corrector_type_0[24] = {
58 195, 159, 133, 115, 101, 93, 87, 77,
59 195, 159, 133, 115, 101, 93, 87, 77,
60 128, 79, 79, 79, 79, 79, 79, 79
61 };
62
63 static int corrector_type_2[8] = { 9, 7, 6, 8, 5, 4, 3, 2 };
64
65 static void build_modpred(Indeo3DecodeContext *s)
66 {
67 int i, j;
68
69 s->ModPred = (unsigned char *) av_malloc (8 * 128);
70
71 for (i=0; i < 128; ++i) {
72 s->ModPred[i+0*128] = (i > 126) ? 254 : 2*((i + 1) - ((i + 1) % 2));
73 s->ModPred[i+1*128] = (i == 7) ? 20 : ((i == 119 || i == 120)
74 ? 236 : 2*((i + 2) - ((i + 1) % 3)));
75 s->ModPred[i+2*128] = (i > 125) ? 248 : 2*((i + 2) - ((i + 2) % 4));
76 s->ModPred[i+3*128] = 2*((i + 1) - ((i - 3) % 5));
77 s->ModPred[i+4*128] = (i == 8) ? 20 : 2*((i + 1) - ((i - 3) % 6));
78 s->ModPred[i+5*128] = 2*((i + 4) - ((i + 3) % 7));
79 s->ModPred[i+6*128] = (i > 123) ? 240 : 2*((i + 4) - ((i + 4) % 8));
80 s->ModPred[i+7*128] = 2*((i + 5) - ((i + 4) % 9));
81 }
82
83 s->corrector_type = (unsigned short *) av_malloc (24 * 256 * sizeof(unsigned short));
84
85 for (i=0; i < 24; ++i) {
86 for (j=0; j < 256; ++j) {
87 s->corrector_type[i*256+j] = (j < corrector_type_0[i])
88 ? 1 : ((j < 248 || (i == 16 && j == 248))
89 ? 0 : corrector_type_2[j - 248]);
90 }
91 }
92 }
93
94 static void iv_Decode_Chunk(Indeo3DecodeContext *s, unsigned char *cur,
95 unsigned char *ref, int width, int height, unsigned char *buf1,
96 long fflags2, unsigned char *hdr,
97 unsigned char *buf2, int min_width_160);
98
99 #define min(a,b) ((a) < (b) ? (a) : (b))
100
101 /* ---------------------------------------------------------------------- */
102 static void iv_alloc_frames(Indeo3DecodeContext *s)
103 {
104 int luma_width, luma_height, luma_pixels, chroma_width, chroma_height,
105 chroma_pixels, bufsize, i;
106
107 luma_width = (s->width + 15) & -0x10;
108 luma_height = (s->height + 15) & -0x10;
109
110 s->iv_frame[0].y_w = s->iv_frame[0].y_h =
111 s->iv_frame[0].the_buf_size = 0;
112 s->iv_frame[1].y_w = s->iv_frame[1].y_h =
113 s->iv_frame[1].the_buf_size = 0;
114 s->iv_frame[1].the_buf = NULL;
115
116 chroma_width = luma_width >> 2;
117 chroma_height = luma_height >> 2;
118 luma_pixels = luma_width * luma_height;
119 chroma_pixels = chroma_width * chroma_height;
120
121 bufsize = luma_pixels * 2 + luma_width * 3 +
122 (chroma_pixels + chroma_width) * 4;
123
124 if((s->iv_frame[0].the_buf =
125 (s->iv_frame[0].the_buf_size == 0 ? av_malloc(bufsize) :
126 av_realloc(s->iv_frame[0].the_buf, bufsize))) == NULL)
127 return;
128 s->iv_frame[0].y_w = s->iv_frame[1].y_w = luma_width;
129 s->iv_frame[0].y_h = s->iv_frame[1].y_h = luma_height;
130 s->iv_frame[0].uv_w = s->iv_frame[1].uv_w = chroma_width;
131 s->iv_frame[0].uv_h = s->iv_frame[1].uv_h = chroma_height;
132 s->iv_frame[0].the_buf_size = bufsize;
133
134 s->iv_frame[0].Ybuf = s->iv_frame[0].the_buf + luma_width;
135 i = luma_pixels + luma_width * 2;
136 s->iv_frame[1].Ybuf = s->iv_frame[0].the_buf + i;
137 i += (luma_pixels + luma_width);
138 s->iv_frame[0].Ubuf = s->iv_frame[0].the_buf + i;
139 i += (chroma_pixels + chroma_width);
140 s->iv_frame[1].Ubuf = s->iv_frame[0].the_buf + i;
141 i += (chroma_pixels + chroma_width);
142 s->iv_frame[0].Vbuf = s->iv_frame[0].the_buf + i;
143 i += (chroma_pixels + chroma_width);
144 s->iv_frame[1].Vbuf = s->iv_frame[0].the_buf + i;
145
146 for(i = 1; i <= luma_width; i++)
147 s->iv_frame[0].Ybuf[-i] = s->iv_frame[1].Ybuf[-i] =
148 s->iv_frame[0].Ubuf[-i] = 0x80;
149
150 for(i = 1; i <= chroma_width; i++) {
151 s->iv_frame[1].Ubuf[-i] = 0x80;
152 s->iv_frame[0].Vbuf[-i] = 0x80;
153 s->iv_frame[1].Vbuf[-i] = 0x80;
154 s->iv_frame[1].Vbuf[chroma_pixels+i-1] = 0x80;
155 }
156 }
157
158 /* ---------------------------------------------------------------------- */
159 static void iv_free_func(Indeo3DecodeContext *s)
160 {
161 int i;
162
163 for(i = 0 ; i < 2 ; i++) {
164 if(s->iv_frame[i].the_buf != NULL)
165 av_free(s->iv_frame[i].the_buf);
166 s->iv_frame[i].Ybuf = s->iv_frame[i].Ubuf =
167 s->iv_frame[i].Vbuf = NULL;
168 s->iv_frame[i].the_buf = NULL;
169 s->iv_frame[i].the_buf_size = 0;
170 s->iv_frame[i].y_w = s->iv_frame[i].y_h = 0;
171 s->iv_frame[i].uv_w = s->iv_frame[i].uv_h = 0;
172 }
173
174 av_free(s->ModPred);
175 av_free(s->corrector_type);
176 }
177
178 /* ---------------------------------------------------------------------- */
179 static unsigned long iv_decode_frame(Indeo3DecodeContext *s,
180 unsigned char *buf, int buf_size)
181 {
182 unsigned int hdr_width, hdr_height,
183 chroma_width, chroma_height;
184 unsigned long fflags1, fflags2, fflags3, offs1, offs2, offs3, offs;
185 unsigned char *hdr_pos, *buf_pos;
186
187 buf_pos = buf;
188 buf_pos += 18;
189
190 fflags1 = le2me_16(*(uint16_t *)buf_pos);
191 buf_pos += 2;
192 fflags3 = le2me_32(*(uint32_t *)buf_pos);
193 buf_pos += 4;
194 fflags2 = *buf_pos++;
195 buf_pos += 3;
196 hdr_height = le2me_16(*(uint16_t *)buf_pos);
197 buf_pos += 2;
198 hdr_width = le2me_16(*(uint16_t *)buf_pos);
199 buf_pos += 2;
200 chroma_height = ((hdr_height >> 2) + 3) & 0x7ffc;
201 chroma_width = ((hdr_width >> 2) + 3) & 0x7ffc;
202 offs1 = le2me_32(*(uint32_t *)buf_pos);
203 buf_pos += 4;
204 offs2 = le2me_32(*(uint32_t *)buf_pos);
205 buf_pos += 4;
206 offs3 = le2me_32(*(uint32_t *)buf_pos);
207 buf_pos += 8;
208 hdr_pos = buf_pos;
209 if(fflags3 == 0x80) return 4;
210
211 if(fflags1 & 0x200) {
212 s->cur_frame = s->iv_frame + 1;
213 s->ref_frame = s->iv_frame;
214 } else {
215 s->cur_frame = s->iv_frame;
216 s->ref_frame = s->iv_frame + 1;
217 }
218
219 buf_pos = buf + 16 + offs1;
220 offs = le2me_32(*(uint32_t *)buf_pos);
221 buf_pos += 4;
222
223 iv_Decode_Chunk(s, s->cur_frame->Ybuf, s->ref_frame->Ybuf, hdr_width,
224 hdr_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
225 min(hdr_width, 160));
226
227 buf_pos = buf + 16 + offs2;
228 offs = le2me_32(*(uint32_t *)buf_pos);
229 buf_pos += 4;
230
231 iv_Decode_Chunk(s, s->cur_frame->Vbuf, s->ref_frame->Vbuf, chroma_width,
232 chroma_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
233 min(chroma_width, 40));
234
235 buf_pos = buf + 16 + offs3;
236 offs = le2me_32(*(uint32_t *)buf_pos);
237 buf_pos += 4;
238
239 iv_Decode_Chunk(s, s->cur_frame->Ubuf, s->ref_frame->Ubuf, chroma_width,
240 chroma_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
241 min(chroma_width, 40));
242
243 return 8;
244 }
245
246 typedef struct {
247 long xpos;
248 long ypos;
249 long width;
250 long height;
251 long split_flag;
252 long split_direction;
253 long usl7;
254 } ustr_t;
255
256 /* ---------------------------------------------------------------------- */
257
258 static void iv_Decode_Chunk(Indeo3DecodeContext *s,
259 unsigned char *cur, unsigned char *ref, int width, int height,
260 unsigned char *buf1, long fflags2, unsigned char *hdr,
261 unsigned char *buf2, int min_width_160)
262 {
263 unsigned char bit_buf;
264 unsigned long bit_pos, lv, lv1, lv2;
265 long *width_tbl, width_tbl_arr[10];
266 char *ref_vectors;
267 unsigned char *cur_frm_pos, *ref_frm_pos, *cp, *cp2;
268 unsigned long *cur_lp, *ref_lp, *correction_lp[2], *correctionloworder_lp[2],
269 *correctionhighorder_lp[2];
270 unsigned short *correction_type_sp[2];
271 ustr_t xustr[20], *ptr_ustr;
272 int i, j, k, lp1, lp2, flag1, cmd, blks_width, blks_height, region_160_width,
273 rle_v1, rle_v2, rle_v3;
274
275 bit_buf = 0;
276 ref_vectors = NULL;
277
278 width_tbl = width_tbl_arr + 1;
279 i = (width < 0 ? width + 3 : width)/4;
280 for(j = -1; j < 8; j++)
281 width_tbl[j] = i * j;
282
283 ptr_ustr = xustr;
284
285 for(region_160_width = 0; region_160_width < (width - min_width_160); region_160_width += min_width_160);
286
287 ptr_ustr->ypos = ptr_ustr->xpos = 0;
288 for(ptr_ustr->width = min_width_160; width > ptr_ustr->width; ptr_ustr->width *= 2);
289 ptr_ustr->height = height;
290 ptr_ustr->split_direction = 0;
291 ptr_ustr->split_flag = 0;
292 ptr_ustr->usl7 = 0;
293
294 bit_pos = 0;
295
296 rle_v1 = rle_v2 = rle_v3 = 0;
297
298 while(ptr_ustr >= xustr) {
299 if(bit_pos <= 0) {
300 bit_pos = 8;
301 bit_buf = *buf1++;
302 }
303
304 bit_pos -= 2;
305 cmd = (bit_buf >> bit_pos) & 0x03;
306
307 if(cmd == 0) {
308 ptr_ustr++;
309 memcpy(ptr_ustr, ptr_ustr-1, sizeof(ustr_t));
310 ptr_ustr->split_flag = 1;
311 ptr_ustr->split_direction = 0;
312 ptr_ustr->height = (ptr_ustr->height > 8 ? ((ptr_ustr->height+8)>>4)<<3 : 4);
313 continue;
314 } else if(cmd == 1) {
315 ptr_ustr++;
316 memcpy(ptr_ustr, ptr_ustr-1, sizeof(ustr_t));
317 ptr_ustr->split_flag = 1;
318 ptr_ustr->split_direction = 1;
319 ptr_ustr->width = (ptr_ustr->width > 8 ? ((ptr_ustr->width+8)>>4)<<3 : 4);
320 continue;
321 } else if(cmd == 2) {
322 if(ptr_ustr->usl7 == 0) {
323 ptr_ustr->usl7 = 1;
324 ref_vectors = NULL;
325 continue;
326 }
327 } else if(cmd == 3) {
328 if(ptr_ustr->usl7 == 0) {
329 ptr_ustr->usl7 = 1;
330 ref_vectors = buf2 + (*buf1 * 2);
331 buf1++;
332 continue;
333 }
334 }
335
336 cur_frm_pos = cur + width * ptr_ustr->ypos + ptr_ustr->xpos;
337
338 if((blks_width = ptr_ustr->width) < 0)
339 blks_width += 3;
340 blks_width >>= 2;
341 blks_height = ptr_ustr->height;
342
343 if(ref_vectors != NULL) {
344 ref_frm_pos = ref + (ref_vectors[0] + ptr_ustr->ypos) * width +
345 ref_vectors[1] + ptr_ustr->xpos;
346 } else
347 ref_frm_pos = cur_frm_pos - width_tbl[4];
348
349 if(cmd == 2) {
350 if(bit_pos <= 0) {
351 bit_pos = 8;
352 bit_buf = *buf1++;
353 }
354
355 bit_pos -= 2;
356 cmd = (bit_buf >> bit_pos) & 0x03;
357
358 if(cmd == 0 || ref_vectors != NULL) {
359 for(lp1 = 0; lp1 < blks_width; lp1++) {
360 for(i = 0, j = 0; i < blks_height; i++, j += width_tbl[1])
361 ((unsigned long *)cur_frm_pos)[j] = ((unsigned long *)ref_frm_pos)[j];
362 cur_frm_pos += 4;
363 ref_frm_pos += 4;
364 }
365 } else if(cmd != 1)
366 return;
367 } else {
368 k = *buf1 >> 4;
369 j = *buf1 & 0x0f;
370 buf1++;
371 lv = j + fflags2;
372
373 if((lv - 8) <= 7 && (k == 0 || k == 3 || k == 10)) {
374 cp2 = s->ModPred + ((lv - 8) << 7);
375 cp = ref_frm_pos;
376 for(i = 0; i < blks_width << 2; i++) { *(cp++) = cp2[*cp >> 1]; }
377 }
378
379 if(k == 1 || k == 4) {
380 lv = (hdr[j] & 0xf) + fflags2;
381 correction_type_sp[0] = s->corrector_type + (lv << 8);
382 correction_lp[0] = correction + (lv << 8);
383 lv = (hdr[j] >> 4) + fflags2;
384 correction_lp[1] = correction + (lv << 8);
385 correction_type_sp[1] = s->corrector_type + (lv << 8);
386 } else {
387 correctionloworder_lp[0] = correctionloworder_lp[1] = correctionloworder + (lv << 8);
388 correctionhighorder_lp[0] = correctionhighorder_lp[1] = correctionhighorder + (lv << 8);
389 correction_type_sp[0] = correction_type_sp[1] = s->corrector_type + (lv << 8);
390 correction_lp[0] = correction_lp[1] = correction + (lv << 8);
391 }
392
393 switch(k) {
394 case 1:
395 case 0: /********** CASE 0 **********/
396 for( ; blks_height > 0; blks_height -= 4) {
397 for(lp1 = 0; lp1 < blks_width; lp1++) {
398 for(lp2 = 0; lp2 < 4; ) {
399 k = *buf1++;
400 cur_lp = ((unsigned long *)cur_frm_pos) + width_tbl[lp2];
401 ref_lp = ((unsigned long *)ref_frm_pos) + width_tbl[lp2];
402
403 switch(correction_type_sp[0][k]) {
404 case 0:
405 *cur_lp = ((*ref_lp >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
406 lp2++;
407 break;
408 case 1:
409 ((unsigned short *)cur_lp)[0] = ((((unsigned short *)(ref_lp))[0] >> 1)
410 + correction_lp[lp2 & 0x01][*buf1++]) << 1;
411 ((unsigned short *)cur_lp)[1] = ((((unsigned short *)(ref_lp))[1] >> 1)
412 + correction_lp[lp2 & 0x01][k]) << 1;
413 lp2++;
414 break;
415 case 2:
416 if(lp2 == 0) {
417 for(i = 0, j = 0; i < 2; i++, j += width_tbl[1])
418 cur_lp[j] = ref_lp[j];
419 lp2 += 2;
420 }
421 break;
422 case 3:
423 if(lp2 < 2) {
424 for(i = 0, j = 0; i < (3 - lp2); i++, j += width_tbl[1])
425 cur_lp[j] = ref_lp[j];
426 lp2 = 3;
427 }
428 break;
429 case 8:
430 if(lp2 == 0) {
431 if(rle_v3 == 0) {
432 rle_v2 = *buf1;
433 rle_v1 = 1;
434 if(rle_v2 > 32) {
435 rle_v2 -= 32;
436 rle_v1 = 0;
437 }
438 rle_v3 = 1;
439 }
440 buf1--;
441
442 if(rle_v1 == 1 || ref_vectors != NULL) {
443 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
444 cur_lp[j] = ref_lp[j];
445 }
446
447 rle_v2--;
448 if(rle_v2 == 0) {
449 rle_v3 = 0;
450 buf1 += 2;
451 }
452 lp2 = 4;
453 break;
454 } else {
455 rle_v1 = 1;
456 rle_v2 = *buf1 - 1;
457 }
458 case 5:
459 if(lp2 == 0 && rle_v3 != 0)
460 rle_v3 = 0;
461 else {
462 buf1--;
463 rle_v3 = 1;
464 }
465 case 4:
466 for(i = 0, j = 0; i < (4 - lp2); i++, j += width_tbl[1])
467 cur_lp[j] = ref_lp[j];
468 lp2 = 4;
469 break;
470
471 case 7:
472 if(rle_v3 != 0)
473 rle_v3 = 0;
474 else {
475 buf1--;
476 rle_v3 = 1;
477 }
478 case 6:
479 if(ref_vectors != NULL) {
480 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
481 cur_lp[j] = ref_lp[j];
482 }
483 lp2 = 4;
484 break;
485
486 case 9:
487 lv1 = *buf1++;
488 lv = (lv1 & 0x7F) << 1;
489 lv += (lv << 8);
490 lv += (lv << 16);
491 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
492 cur_lp[j] = lv;
493
494 if((lv1 & 0x80) != 0) {
495 if(rle_v3 != 0)
496 rle_v3 = 0;
497 else {
498 rle_v3 = 1;
499 buf1 -= 2;
500 }
501 }
502 lp2 = 4;
503 break;
504 default:
505 return;
506 }
507 }
508
509 cur_frm_pos += 4;
510 ref_frm_pos += 4;
511 }
512
513 cur_frm_pos += ((width - blks_width) * 4);
514 ref_frm_pos += ((width - blks_width) * 4);
515 }
516 break;
517
518 case 4:
519 case 3: /********** CASE 3 **********/
520 if(ref_vectors != NULL)
521 return;
522 flag1 = 1;
523
524 for( ; blks_height > 0; blks_height -= 8) {
525 for(lp1 = 0; lp1 < blks_width; lp1++) {
526 for(lp2 = 0; lp2 < 4; ) {
527 k = *buf1++;
528
529 cur_lp = ((unsigned long *)cur_frm_pos) + width_tbl[lp2 * 2];
530 ref_lp = ((unsigned long *)cur_frm_pos) + width_tbl[(lp2 * 2) - 1];
531
532 switch(correction_type_sp[lp2 & 0x01][k]) {
533 case 0:
534 cur_lp[width_tbl[1]] = ((*ref_lp >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
535 if(lp2 > 0 || flag1 == 0 || ptr_ustr->ypos != 0)
536 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
537 else
538 cur_lp[0] = ((*ref_lp >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
539 lp2++;
540 break;
541
542 case 1:
543 ((unsigned short *)cur_lp)[width_tbl[2]] =
544 ((((unsigned short *)ref_lp)[0] >> 1) + correction_lp[lp2 & 0x01][*buf1++]) << 1;
545 ((unsigned short *)cur_lp)[width_tbl[2]+1] =
546 ((((unsigned short *)ref_lp)[1] >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
547 if(lp2 > 0 || flag1 == 0 || ptr_ustr->ypos != 0)
548 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
549 else
550 cur_lp[0] = cur_lp[width_tbl[1]];
551 lp2++;
552 break;
553
554 case 2:
555 if(lp2 == 0) {
556 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
557 cur_lp[j] = *ref_lp;
558 lp2 += 2;
559 }
560 break;
561
562 case 3:
563 if(lp2 < 2) {
564 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1])
565 cur_lp[j] = *ref_lp;
566 lp2 = 3;
567 }
568 break;
569
570 case 6:
571 lp2 = 4;
572 break;
573
574 case 7:
575 if(rle_v3 != 0)
576 rle_v3 = 0;
577 else {
578 buf1--;
579 rle_v3 = 1;
580 }
581 lp2 = 4;
582 break;
583
584 case 8:
585 if(lp2 == 0) {
586 if(rle_v3 == 0) {
587 rle_v2 = *buf1;
588 rle_v1 = 1;
589 if(rle_v2 > 32) {
590 rle_v2 -= 32;
591 rle_v1 = 0;
592 }
593 rle_v3 = 1;
594 }
595 buf1--;
596
597 if(rle_v1 == 1) {
598 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
599 cur_lp[j] = ref_lp[j];
600 }
601
602 rle_v2--;
603 if(rle_v2 == 0) {
604 rle_v3 = 0;
605 buf1 += 2;
606 }
607 lp2 = 4;
608 break;
609 } else {
610 rle_v2 = (*buf1) - 1;
611 rle_v1 = 1;
612 }
613 case 5:
614 if(lp2 == 0 && rle_v3 != 0)
615 rle_v3 = 0;
616 else {
617 buf1--;
618 rle_v3 = 1;
619 }
620 case 4:
621 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1])
622 cur_lp[j] = *ref_lp;
623 lp2 = 4;
624 break;
625
626 case 9:
627 fprintf(stderr, "UNTESTED.\n");
628 lv1 = *buf1++;
629 lv = (lv1 & 0x7F) << 1;
630 lv += (lv << 8);
631 lv += (lv << 16);
632
633 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
634 cur_lp[j] = lv;
635
636 if((lv1 & 0x80) != 0) {
637 if(rle_v3 != 0)
638 rle_v3 = 0;
639 else {
640 rle_v3 = 1;
641 buf1 -= 2;
642 }
643 }
644 lp2 = 4;
645 break;
646
647 default:
648 return;
649 }
650 }
651
652 cur_frm_pos += 4;
653 }
654
655 cur_frm_pos += (((width * 2) - blks_width) * 4);
656 flag1 = 0;
657 }
658 break;
659
660 case 10: /********** CASE 10 **********/
661 if(ref_vectors == NULL) {
662 flag1 = 1;
663
664 for( ; blks_height > 0; blks_height -= 8) {
665 for(lp1 = 0; lp1 < blks_width; lp1 += 2) {
666 for(lp2 = 0; lp2 < 4; ) {
667 k = *buf1++;
668 cur_lp = ((unsigned long *)cur_frm_pos) + width_tbl[lp2 * 2];
669 ref_lp = ((unsigned long *)cur_frm_pos) + width_tbl[(lp2 * 2) - 1];
670 lv1 = ref_lp[0];
671 lv2 = ref_lp[1];
672 if(lp2 == 0 && flag1 != 0) {
673 lv1 = lv1 & 0x00FF00FF;
674 lv1 = (lv1 << 8) | lv1;
675 lv2 = lv2 & 0x00FF00FF;
676 lv2 = (lv2 << 8) | lv2;
677 }
678
679 switch(correction_type_sp[lp2 & 0x01][k]) {
680 case 0:
681 cur_lp[width_tbl[1]] = ((lv1 >> 1) + correctionloworder_lp[lp2 & 0x01][k]) << 1;
682 cur_lp[width_tbl[1]+1] = ((lv2 >> 1) + correctionhighorder_lp[lp2 & 0x01][k]) << 1;
683 if(lp2 > 0 || ptr_ustr->ypos != 0 || flag1 == 0) {
684 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
685 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
686 } else {
687 cur_lp[0] = cur_lp[width_tbl[1]];
688 cur_lp[1] = cur_lp[width_tbl[1]+1];
689 }
690 lp2++;
691 break;
692
693 case 1:
694 cur_lp[width_tbl[1]] = ((lv1 >> 1) + correctionloworder_lp[lp2 & 0x01][*buf1++]) << 1;
695 cur_lp[width_tbl[1]+1] = ((lv2 >> 1) + correctionloworder_lp[lp2 & 0x01][k]) << 1;
696 if(lp2 > 0 || ptr_ustr->ypos != 0 || flag1 == 0) {
697 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
698 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
699 } else {
700 cur_lp[0] = cur_lp[width_tbl[1]];
701 cur_lp[1] = cur_lp[width_tbl[1]+1];
702 }
703 lp2++;
704 break;
705
706 case 2:
707 if(lp2 == 0) {
708 if(flag1 != 0) {
709 for(i = 0, j = width_tbl[1]; i < 3; i++, j += width_tbl[1]) {
710 cur_lp[j] = lv1;
711 cur_lp[j+1] = lv2;
712 }
713 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
714 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
715 } else {
716 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1]) {
717 cur_lp[j] = lv1;
718 cur_lp[j+1] = lv2;
719 }
720 }
721 lp2 += 2;
722 }
723 break;
724
725 case 3:
726 if(lp2 < 2) {
727 if(lp2 == 0 && flag1 != 0) {
728 for(i = 0, j = width_tbl[1]; i < 5; i++, j += width_tbl[1]) {
729 cur_lp[j] = lv1;
730 cur_lp[j+1] = lv2;
731 }
732 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
733 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
734 } else {
735 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1]) {
736 cur_lp[j] = lv1;
737 cur_lp[j+1] = lv2;
738 }
739 }
740 lp2 = 3;
741 }
742 break;
743
744 case 8:
745 if(lp2 == 0) {
746 if(rle_v3 == 0) {
747 rle_v2 = *buf1;
748 rle_v1 = 1;
749 if(rle_v2 > 32) {
750 rle_v2 -= 32;
751 rle_v1 = 0;
752 }
753 rle_v3 = 1;
754 }
755 buf1--;
756 if(rle_v1 == 1) {
757 if(flag1 != 0) {
758 for(i = 0, j = width_tbl[1]; i < 7; i++, j += width_tbl[1]) {
759 cur_lp[j] = lv1;
760 cur_lp[j+1] = lv2;
761 }
762 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
763 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
764 } else {
765 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1]) {
766 cur_lp[j] = lv1;
767 cur_lp[j+1] = lv2;
768 }
769 }
770 }
771 rle_v2--;
772 if(rle_v2 == 0) {
773 rle_v3 = 0;
774 buf1 += 2;
775 }
776 lp2 = 4;
777 break;
778 } else {
779 rle_v1 = 1;
780 rle_v2 = (*buf1) - 1;
781 }
782 case 5:
783 if(lp2 == 0 && rle_v3 != 0)
784 rle_v3 = 0;
785 else {
786 buf1--;
787 rle_v3 = 1;
788 }
789 case 4:
790 if(lp2 == 0 && flag1 != 0) {
791 for(i = 0, j = width_tbl[1]; i < 7; i++, j += width_tbl[1]) {
792 cur_lp[j] = lv1;
793 cur_lp[j+1] = lv2;
794 }
795 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
796 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
797 } else {
798 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1]) {
799 cur_lp[j] = lv1;
800 cur_lp[j+1] = lv2;
801 }
802 }
803 lp2 = 4;
804 break;
805
806 case 6:
807 lp2 = 4;
808 break;
809
810 case 7:
811 if(lp2 == 0) {
812 if(rle_v3 != 0)
813 rle_v3 = 0;
814 else {
815 buf1--;
816 rle_v3 = 1;
817 }
818 lp2 = 4;
819 }
820 break;
821
822 case 9:
823 fprintf(stderr, "UNTESTED.\n");
824 lv1 = *buf1;
825 lv = (lv1 & 0x7F) << 1;
826 lv += (lv << 8);
827 lv += (lv << 16);
828 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
829 cur_lp[j] = lv;
830 if((lv1 & 0x80) != 0) {
831 if(rle_v3 != 0)
832 rle_v3 = 0;
833 else {
834 rle_v3 = 1;
835 buf1 -= 2;
836 }
837 }
838 lp2 = 4;
839 break;
840
841 default:
842 return;
843 }
844 }
845
846 cur_frm_pos += 8;
847 }
848
849 cur_frm_pos += (((width * 2) - blks_width) * 4);
850 flag1 = 0;
851 }
852 } else {
853 for( ; blks_height > 0; blks_height -= 8) {
854 for(lp1 = 0; lp1 < blks_width; lp1 += 2) {
855 for(lp2 = 0; lp2 < 4; ) {
856 k = *buf1++;
857 cur_lp = ((unsigned long *)cur_frm_pos) + width_tbl[lp2 * 2];
858 ref_lp = ((unsigned long *)ref_frm_pos) + width_tbl[lp2 * 2];
859
860 switch(correction_type_sp[lp2 & 0x01][k]) {
861 case 0:
862 lv1 = correctionloworder_lp[lp2 & 0x01][k];
863 lv2 = correctionhighorder_lp[lp2 & 0x01][k];
864 cur_lp[0] = ((ref_lp[0] >> 1) + lv1) << 1;
865 cur_lp[1] = ((ref_lp[1] >> 1) + lv2) << 1;
866 cur_lp[width_tbl[1]] = ((ref_lp[width_tbl[1]] >> 1) + lv1) << 1;
867 cur_lp[width_tbl[1]+1] = ((ref_lp[width_tbl[1]+1] >> 1) + lv2) << 1;
868 lp2++;
869 break;
870
871 case 1:
872 lv1 = correctionloworder_lp[lp2 & 0x01][*buf1++];
873 lv2 = correctionloworder_lp[lp2 & 0x01][k];
874 cur_lp[0] = ((ref_lp[0] >> 1) + lv1) << 1;
875 cur_lp[1] = ((ref_lp[1] >> 1) + lv2) << 1;
876 cur_lp[width_tbl[1]] = ((ref_lp[width_tbl[1]] >> 1) + lv1) << 1;
877 cur_lp[width_tbl[1]+1] = ((ref_lp[width_tbl[1]+1] >> 1) + lv2) << 1;
878 lp2++;
879 break;
880
881 case 2:
882 if(lp2 == 0) {
883 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1]) {
884 cur_lp[j] = ref_lp[j];
885 cur_lp[j+1] = ref_lp[j+1];
886 }
887 lp2 += 2;
888 }
889 break;
890
891 case 3:
892 if(lp2 < 2) {
893 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1]) {
894 cur_lp[j] = ref_lp[j];
895 cur_lp[j+1] = ref_lp[j+1];
896 }
897 lp2 = 3;
898 }
899 break;
900
901 case 8:
902 if(lp2 == 0) {
903 if(rle_v3 == 0) {
904 rle_v2 = *buf1;
905 rle_v1 = 1;
906 if(rle_v2 > 32) {
907 rle_v2 -= 32;
908 rle_v1 = 0;
909 }
910 rle_v3 = 1;
911 }
912 buf1--;
913 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1]) {
914 ((unsigned long *)cur_frm_pos)[j] = ((unsigned long *)ref_frm_pos)[j];
915 ((unsigned long *)cur_frm_pos)[j+1] = ((unsigned long *)ref_frm_pos)[j+1];
916 }
917 rle_v2--;
918 if(rle_v2 == 0) {
919 rle_v3 = 0;
920 buf1 += 2;
921 }
922 lp2 = 4;
923 break;
924 } else {
925 rle_v1 = 1;
926 rle_v2 = (*buf1) - 1;
927 }
928 case 5:
929 case 7:
930 if(lp2 == 0 && rle_v3 != 0)
931 rle_v3 = 0;
932 else {
933 buf1--;
934 rle_v3 = 1;
935 }
936 case 6:
937 case 4:
938 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1]) {
939 cur_lp[j] = ref_lp[j];
940 cur_lp[j+1] = ref_lp[j+1];
941 }
942 lp2 = 4;
943 break;
944
945 case 9:
946 fprintf(stderr, "UNTESTED.\n");
947 lv1 = *buf1;
948 lv = (lv1 & 0x7F) << 1;
949 lv += (lv << 8);
950 lv += (lv << 16);
951 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
952 ((unsigned long *)cur_frm_pos)[j] = ((unsigned long *)cur_frm_pos)[j+1] = lv;
953 if((lv1 & 0x80) != 0) {
954 if(rle_v3 != 0)
955 rle_v3 = 0;
956 else {
957 rle_v3 = 1;
958 buf1 -= 2;
959 }
960 }
961 lp2 = 4;
962 break;
963
964 default:
965 return;
966 }
967 }
968
969 cur_frm_pos += 8;
970 ref_frm_pos += 8;
971 }
972
973 cur_frm_pos += (((width * 2) - blks_width) * 4);
974 ref_frm_pos += (((width * 2) - blks_width) * 4);
975 }
976 }
977 break;
978
979 case 11: /********** CASE 11 **********/
980 if(ref_vectors == NULL)
981 return;
982
983 for( ; blks_height > 0; blks_height -= 8) {
984 for(lp1 = 0; lp1 < blks_width; lp1++) {
985 for(lp2 = 0; lp2 < 4; ) {
986 k = *buf1++;
987 cur_lp = ((unsigned long *)cur_frm_pos) + width_tbl[lp2 * 2];
988 ref_lp = ((unsigned long *)ref_frm_pos) + width_tbl[lp2 * 2];
989
990 switch(correction_type_sp[lp2 & 0x01][k]) {
991 case 0:
992 cur_lp[0] = ((*ref_lp >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
993 cur_lp[width_tbl[1]] = ((ref_lp[width_tbl[1]] >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
994 lp2++;
995 break;
996
997 case 1:
998 lv1 = (unsigned short)(correction_lp[lp2 & 0x01][*buf1++]);
999 lv2 = (unsigned short)(correction_lp[lp2 & 0x01][k]);
1000 ((unsigned short *)cur_lp)[0] = ((((unsigned short *)ref_lp)[0] >> 1) + lv1) << 1;
1001 ((unsigned short *)cur_lp)[1] = ((((unsigned short *)ref_lp)[1] >> 1) + lv2) << 1;
1002 ((unsigned short *)cur_lp)[width_tbl[2]] = ((((unsigned short *)ref_lp)[width_tbl[2]] >> 1) + lv1) << 1;
1003 ((unsigned short *)cur_lp)[width_tbl[2]+1] = ((((unsigned short *)ref_lp)[width_tbl[2]+1] >> 1) + lv2) << 1;
1004 lp2++;
1005 break;
1006
1007 case 2:
1008 if(lp2 == 0) {
1009 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
1010 cur_lp[j] = ref_lp[j];
1011 lp2 += 2;
1012 }
1013 break;
1014
1015 case 3:
1016 if(lp2 < 2) {
1017 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1])
1018 cur_lp[j] = ref_lp[j];
1019 lp2 = 3;
1020 }
1021 break;
1022
1023 case 8:
1024 if(lp2 == 0) {
1025 if(rle_v3 == 0) {
1026 rle_v2 = *buf1;
1027 rle_v1 = 1;
1028 if(rle_v2 > 32) {
1029 rle_v2 -= 32;
1030 rle_v1 = 0;
1031 }
1032 rle_v3 = 1;
1033 }
1034 buf1--;
1035
1036 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
1037 cur_lp[j] = ref_lp[j];
1038
1039 rle_v2--;
1040 if(rle_v2 == 0) {
1041 rle_v3 = 0;
1042 buf1 += 2;
1043 }
1044 lp2 = 4;
1045 break;
1046 } else {
1047 rle_v1 = 1;
1048 rle_v2 = (*buf1) - 1;
1049 }
1050 case 5:
1051 case 7:
1052 if(lp2 == 0 && rle_v3 != 0)
1053 rle_v3 = 0;
1054 else {
1055 buf1--;
1056 rle_v3 = 1;
1057 }
1058 case 4:
1059 case 6:
1060 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1])
1061 cur_lp[j] = ref_lp[j];
1062 lp2 = 4;
1063 break;
1064
1065 case 9:
1066 fprintf(stderr, "UNTESTED.\n");
1067 lv1 = *buf1++;
1068 lv = (lv1 & 0x7F) << 1;
1069 lv += (lv << 8);
1070 lv += (lv << 16);
1071 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
1072 cur_lp[j] = lv;
1073 if((lv1 & 0x80) != 0) {
1074 if(rle_v3 != 0)
1075 rle_v3 = 0;
1076 else {
1077 rle_v3 = 1;
1078 buf1 -= 2;
1079 }
1080 }
1081 lp2 = 4;
1082 break;
1083
1084 default:
1085 return;
1086 }
1087 }
1088
1089 cur_frm_pos += 4;
1090 ref_frm_pos += 4;
1091 }
1092
1093 cur_frm_pos += (((width * 2) - blks_width) * 4);
1094 ref_frm_pos += (((width * 2) - blks_width) * 4);
1095 }
1096 break;
1097
1098 default:
1099 return;
1100 }
1101 }
1102
1103 if(ptr_ustr < xustr)
1104 return;
1105
1106 for( ; ptr_ustr >= xustr; ptr_ustr--) {
1107 if(ptr_ustr->split_flag != 0) {
1108 ptr_ustr->split_flag = 0;
1109 ptr_ustr->usl7 = (ptr_ustr-1)->usl7;
1110
1111 if(ptr_ustr->split_direction) {
1112 ptr_ustr->xpos += ptr_ustr->width;
1113 ptr_ustr->width = (ptr_ustr-1)->width - ptr_ustr->width;
1114 if(region_160_width <= ptr_ustr->xpos && width < ptr_ustr->width + ptr_ustr->xpos)
1115 ptr_ustr->width = width - ptr_ustr->xpos;
1116 } else {
1117 ptr_ustr->ypos += ptr_ustr->height;
1118 ptr_ustr->height = (ptr_ustr-1)->height - ptr_ustr->height;
1119 }
1120 break;
1121 }
1122 }
1123 }
1124 }
1125
1126 static int indeo3_decode_init(AVCodecContext *avctx)
1127 {
1128 Indeo3DecodeContext *s = avctx->priv_data;
1129
1130 s->avctx = avctx;
1131 s->width = avctx->width;
1132 s->height = avctx->height;
1133 avctx->pix_fmt = PIX_FMT_YUV410P;
1134 avctx->has_b_frames = 0;
1135
1136 build_modpred(s);
1137 iv_alloc_frames(s);
1138
1139 return 0;
1140 }
1141
1142 static int indeo3_decode_frame(AVCodecContext *avctx,
1143 void *data, int *data_size,
1144 unsigned char *buf, int buf_size)
1145 {
1146 Indeo3DecodeContext *s=avctx->priv_data;
1147 unsigned char *src, *dest;
1148 int y;
1149
1150 iv_decode_frame(s, buf, buf_size);
1151
1152 s->frame.reference = 0;
1153 if(avctx->get_buffer(avctx, &s->frame) < 0) {
1154 fprintf(stderr, "get_buffer() failed\n");
1155 return -1;
1156 }
1157
1158 src = s->cur_frame->Ybuf;
1159 dest = s->frame.data[0];
1160 for (y = 0; y < s->height; y++) {
1161 memcpy(dest, src, s->cur_frame->y_w);
1162 src += s->cur_frame->y_w;
1163 dest += s->frame.linesize[0];
1164 }
1165
1166 src = s->cur_frame->Ubuf;
1167 dest = s->frame.data[1];
1168 for (y = 0; y < s->height / 4; y++) {
1169 memcpy(dest, src, s->cur_frame->uv_w);
1170 src += s->cur_frame->uv_w;
1171 dest += s->frame.linesize[1];
1172 }
1173
1174 src = s->cur_frame->Vbuf;
1175 dest = s->frame.data[2];
1176 for (y = 0; y < s->height / 4; y++) {
1177 memcpy(dest, src, s->cur_frame->uv_w);
1178 src += s->cur_frame->uv_w;
1179 dest += s->frame.linesize[2];
1180 }
1181
1182 *data_size=sizeof(AVFrame);
1183 *(AVFrame*)data= s->frame;
1184
1185 avctx->release_buffer(avctx, &s->frame);
1186
1187 return buf_size;
1188 }
1189
1190 static int indeo3_decode_end(AVCodecContext *avctx)
1191 {
1192 Indeo3DecodeContext *s = avctx->priv_data;
1193
1194 iv_free_func(s);
1195
1196 return 0;
1197 }
1198
1199 AVCodec indeo3_decoder = {
1200 "indeo3",
1201 CODEC_TYPE_VIDEO,
1202 CODEC_ID_INDEO3,
1203 sizeof(Indeo3DecodeContext),
1204 indeo3_decode_init,
1205 NULL,
1206 indeo3_decode_end,
1207 indeo3_decode_frame,
1208 0,
1209 NULL
1210 };