Normalize triplets in gain_val_tab[][] so gain_exp_tab[] can be just an
[ffmpeg-lucabe.git] / libavcodec / indeo3.c
blob2c310e0c68b686e4023213c9a8b67f13de8994e0
1 /*
2 * Intel Indeo 3 (IV31, IV32, etc.) video decoder for ffmpeg
3 * written, produced, and directed by Alan Smithee
5 * This file is part of FFmpeg.
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.
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.
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
22 #include <stdio.h>
23 #include <stdlib.h>
24 #include <string.h>
25 #include <unistd.h>
27 #include "avcodec.h"
28 #include "dsputil.h"
29 #include "bytestream.h"
31 #include "indeo3data.h"
33 typedef struct
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;
44 typedef struct Indeo3DecodeContext {
45 AVCodecContext *avctx;
46 int width, height;
47 AVFrame frame;
49 YUVBufs iv_frame[2];
50 YUVBufs *cur_frame;
51 YUVBufs *ref_frame;
53 unsigned char *ModPred;
54 unsigned short *corrector_type;
55 } Indeo3DecodeContext;
57 static const 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
63 static const int corrector_type_2[8] = { 9, 7, 6, 8, 5, 4, 3, 2 };
65 static av_cold void build_modpred(Indeo3DecodeContext *s)
67 int i, j;
69 s->ModPred = (unsigned char *) av_malloc (8 * 128);
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));
83 s->corrector_type = (unsigned short *) av_malloc (24 * 256 * sizeof(unsigned short));
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]);
94 static void iv_Decode_Chunk(Indeo3DecodeContext *s, unsigned char *cur,
95 unsigned char *ref, int width, int height, const unsigned char *buf1,
96 long fflags2, const unsigned char *hdr,
97 const unsigned char *buf2, int min_width_160);
99 /* ---------------------------------------------------------------------- */
100 static av_cold void iv_alloc_frames(Indeo3DecodeContext *s)
102 int luma_width, luma_height, luma_pixels, chroma_width, chroma_height,
103 chroma_pixels, i;
104 unsigned int bufsize;
106 luma_width = (s->width + 3) & (~3);
107 luma_height = (s->height + 3) & (~3);
109 s->iv_frame[0].y_w = s->iv_frame[0].y_h =
110 s->iv_frame[0].the_buf_size = 0;
111 s->iv_frame[1].y_w = s->iv_frame[1].y_h =
112 s->iv_frame[1].the_buf_size = 0;
113 s->iv_frame[1].the_buf = NULL;
115 chroma_width = ((luma_width >> 2) + 3) & (~3);
116 chroma_height = ((luma_height>> 2) + 3) & (~3);
117 luma_pixels = luma_width * luma_height;
118 chroma_pixels = chroma_width * chroma_height;
120 bufsize = luma_pixels * 2 + luma_width * 3 +
121 (chroma_pixels + chroma_width) * 4;
123 if((s->iv_frame[0].the_buf =
124 (s->iv_frame[0].the_buf_size == 0 ? av_malloc(bufsize) :
125 av_realloc(s->iv_frame[0].the_buf, bufsize))) == NULL)
126 return;
127 s->iv_frame[0].y_w = s->iv_frame[1].y_w = luma_width;
128 s->iv_frame[0].y_h = s->iv_frame[1].y_h = luma_height;
129 s->iv_frame[0].uv_w = s->iv_frame[1].uv_w = chroma_width;
130 s->iv_frame[0].uv_h = s->iv_frame[1].uv_h = chroma_height;
131 s->iv_frame[0].the_buf_size = bufsize;
133 s->iv_frame[0].Ybuf = s->iv_frame[0].the_buf + luma_width;
134 i = luma_pixels + luma_width * 2;
135 s->iv_frame[1].Ybuf = s->iv_frame[0].the_buf + i;
136 i += (luma_pixels + luma_width);
137 s->iv_frame[0].Ubuf = s->iv_frame[0].the_buf + i;
138 i += (chroma_pixels + chroma_width);
139 s->iv_frame[1].Ubuf = s->iv_frame[0].the_buf + i;
140 i += (chroma_pixels + chroma_width);
141 s->iv_frame[0].Vbuf = s->iv_frame[0].the_buf + i;
142 i += (chroma_pixels + chroma_width);
143 s->iv_frame[1].Vbuf = s->iv_frame[0].the_buf + i;
145 for(i = 1; i <= luma_width; i++)
146 s->iv_frame[0].Ybuf[-i] = s->iv_frame[1].Ybuf[-i] =
147 s->iv_frame[0].Ubuf[-i] = 0x80;
149 for(i = 1; i <= chroma_width; i++) {
150 s->iv_frame[1].Ubuf[-i] = 0x80;
151 s->iv_frame[0].Vbuf[-i] = 0x80;
152 s->iv_frame[1].Vbuf[-i] = 0x80;
153 s->iv_frame[1].Vbuf[chroma_pixels+i-1] = 0x80;
157 /* ---------------------------------------------------------------------- */
158 static av_cold void iv_free_func(Indeo3DecodeContext *s)
160 int i;
162 for(i = 0 ; i < 2 ; i++) {
163 if(s->iv_frame[i].the_buf != NULL)
164 av_free(s->iv_frame[i].the_buf);
165 s->iv_frame[i].Ybuf = s->iv_frame[i].Ubuf =
166 s->iv_frame[i].Vbuf = NULL;
167 s->iv_frame[i].the_buf = NULL;
168 s->iv_frame[i].the_buf_size = 0;
169 s->iv_frame[i].y_w = s->iv_frame[i].y_h = 0;
170 s->iv_frame[i].uv_w = s->iv_frame[i].uv_h = 0;
173 av_free(s->ModPred);
174 av_free(s->corrector_type);
177 /* ---------------------------------------------------------------------- */
178 static unsigned long iv_decode_frame(Indeo3DecodeContext *s,
179 const unsigned char *buf, int buf_size)
181 unsigned int hdr_width, hdr_height,
182 chroma_width, chroma_height;
183 unsigned long fflags1, fflags2, fflags3, offs1, offs2, offs3, offs;
184 const unsigned char *hdr_pos, *buf_pos;
186 buf_pos = buf;
187 buf_pos += 18;
189 fflags1 = bytestream_get_le16(&buf_pos);
190 fflags3 = bytestream_get_le32(&buf_pos);
191 fflags2 = *buf_pos++;
192 buf_pos += 3;
193 hdr_height = bytestream_get_le16(&buf_pos);
194 hdr_width = bytestream_get_le16(&buf_pos);
196 if(avcodec_check_dimensions(NULL, hdr_width, hdr_height))
197 return -1;
199 chroma_height = ((hdr_height >> 2) + 3) & 0x7ffc;
200 chroma_width = ((hdr_width >> 2) + 3) & 0x7ffc;
201 offs1 = bytestream_get_le32(&buf_pos);
202 offs2 = bytestream_get_le32(&buf_pos);
203 offs3 = bytestream_get_le32(&buf_pos);
204 buf_pos += 4;
205 hdr_pos = buf_pos;
206 if(fflags3 == 0x80) return 4;
208 if(fflags1 & 0x200) {
209 s->cur_frame = s->iv_frame + 1;
210 s->ref_frame = s->iv_frame;
211 } else {
212 s->cur_frame = s->iv_frame;
213 s->ref_frame = s->iv_frame + 1;
216 buf_pos = buf + 16 + offs1;
217 offs = bytestream_get_le32(&buf_pos);
219 iv_Decode_Chunk(s, s->cur_frame->Ybuf, s->ref_frame->Ybuf, hdr_width,
220 hdr_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
221 FFMIN(hdr_width, 160));
223 if (!(s->avctx->flags & CODEC_FLAG_GRAY))
226 buf_pos = buf + 16 + offs2;
227 offs = bytestream_get_le32(&buf_pos);
229 iv_Decode_Chunk(s, s->cur_frame->Vbuf, s->ref_frame->Vbuf, chroma_width,
230 chroma_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
231 FFMIN(chroma_width, 40));
233 buf_pos = buf + 16 + offs3;
234 offs = bytestream_get_le32(&buf_pos);
236 iv_Decode_Chunk(s, s->cur_frame->Ubuf, s->ref_frame->Ubuf, chroma_width,
237 chroma_height, buf_pos + offs * 2, fflags2, hdr_pos, buf_pos,
238 FFMIN(chroma_width, 40));
242 return 8;
245 typedef struct {
246 long xpos;
247 long ypos;
248 long width;
249 long height;
250 long split_flag;
251 long split_direction;
252 long usl7;
253 } ustr_t;
255 /* ---------------------------------------------------------------------- */
257 #define LV1_CHECK(buf1,rle_v3,lv1,lp2) \
258 if((lv1 & 0x80) != 0) { \
259 if(rle_v3 != 0) \
260 rle_v3 = 0; \
261 else { \
262 rle_v3 = 1; \
263 buf1 -= 2; \
266 lp2 = 4;
269 #define RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3) \
270 if(rle_v3 == 0) { \
271 rle_v2 = *buf1; \
272 rle_v1 = 1; \
273 if(rle_v2 > 32) { \
274 rle_v2 -= 32; \
275 rle_v1 = 0; \
277 rle_v3 = 1; \
279 buf1--;
282 #define LP2_CHECK(buf1,rle_v3,lp2) \
283 if(lp2 == 0 && rle_v3 != 0) \
284 rle_v3 = 0; \
285 else { \
286 buf1--; \
287 rle_v3 = 1; \
291 #define RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2) \
292 rle_v2--; \
293 if(rle_v2 == 0) { \
294 rle_v3 = 0; \
295 buf1 += 2; \
297 lp2 = 4;
299 static void iv_Decode_Chunk(Indeo3DecodeContext *s,
300 unsigned char *cur, unsigned char *ref, int width, int height,
301 const unsigned char *buf1, long fflags2, const unsigned char *hdr,
302 const unsigned char *buf2, int min_width_160)
304 unsigned char bit_buf;
305 unsigned long bit_pos, lv, lv1, lv2;
306 long *width_tbl, width_tbl_arr[10];
307 const signed char *ref_vectors;
308 unsigned char *cur_frm_pos, *ref_frm_pos, *cp, *cp2;
309 uint32_t *cur_lp, *ref_lp;
310 const uint32_t *correction_lp[2], *correctionloworder_lp[2], *correctionhighorder_lp[2];
311 unsigned short *correction_type_sp[2];
312 ustr_t strip_tbl[20], *strip;
313 int i, j, k, lp1, lp2, flag1, cmd, blks_width, blks_height, region_160_width,
314 rle_v1, rle_v2, rle_v3;
315 unsigned short res;
317 bit_buf = 0;
318 ref_vectors = NULL;
320 width_tbl = width_tbl_arr + 1;
321 i = (width < 0 ? width + 3 : width)/4;
322 for(j = -1; j < 8; j++)
323 width_tbl[j] = i * j;
325 strip = strip_tbl;
327 for(region_160_width = 0; region_160_width < (width - min_width_160); region_160_width += min_width_160);
329 strip->ypos = strip->xpos = 0;
330 for(strip->width = min_width_160; width > strip->width; strip->width *= 2);
331 strip->height = height;
332 strip->split_direction = 0;
333 strip->split_flag = 0;
334 strip->usl7 = 0;
336 bit_pos = 0;
338 rle_v1 = rle_v2 = rle_v3 = 0;
340 while(strip >= strip_tbl) {
341 if(bit_pos <= 0) {
342 bit_pos = 8;
343 bit_buf = *buf1++;
346 bit_pos -= 2;
347 cmd = (bit_buf >> bit_pos) & 0x03;
349 if(cmd == 0) {
350 strip++;
351 memcpy(strip, strip-1, sizeof(ustr_t));
352 strip->split_flag = 1;
353 strip->split_direction = 0;
354 strip->height = (strip->height > 8 ? ((strip->height+8)>>4)<<3 : 4);
355 continue;
356 } else if(cmd == 1) {
357 strip++;
358 memcpy(strip, strip-1, sizeof(ustr_t));
359 strip->split_flag = 1;
360 strip->split_direction = 1;
361 strip->width = (strip->width > 8 ? ((strip->width+8)>>4)<<3 : 4);
362 continue;
363 } else if(cmd == 2) {
364 if(strip->usl7 == 0) {
365 strip->usl7 = 1;
366 ref_vectors = NULL;
367 continue;
369 } else if(cmd == 3) {
370 if(strip->usl7 == 0) {
371 strip->usl7 = 1;
372 ref_vectors = (const signed char*)buf2 + (*buf1 * 2);
373 buf1++;
374 continue;
378 cur_frm_pos = cur + width * strip->ypos + strip->xpos;
380 if((blks_width = strip->width) < 0)
381 blks_width += 3;
382 blks_width >>= 2;
383 blks_height = strip->height;
385 if(ref_vectors != NULL) {
386 ref_frm_pos = ref + (ref_vectors[0] + strip->ypos) * width +
387 ref_vectors[1] + strip->xpos;
388 } else
389 ref_frm_pos = cur_frm_pos - width_tbl[4];
391 if(cmd == 2) {
392 if(bit_pos <= 0) {
393 bit_pos = 8;
394 bit_buf = *buf1++;
397 bit_pos -= 2;
398 cmd = (bit_buf >> bit_pos) & 0x03;
400 if(cmd == 0 || ref_vectors != NULL) {
401 for(lp1 = 0; lp1 < blks_width; lp1++) {
402 for(i = 0, j = 0; i < blks_height; i++, j += width_tbl[1])
403 ((uint32_t *)cur_frm_pos)[j] = ((uint32_t *)ref_frm_pos)[j];
404 cur_frm_pos += 4;
405 ref_frm_pos += 4;
407 } else if(cmd != 1)
408 return;
409 } else {
410 k = *buf1 >> 4;
411 j = *buf1 & 0x0f;
412 buf1++;
413 lv = j + fflags2;
415 if((lv - 8) <= 7 && (k == 0 || k == 3 || k == 10)) {
416 cp2 = s->ModPred + ((lv - 8) << 7);
417 cp = ref_frm_pos;
418 for(i = 0; i < blks_width << 2; i++) {
419 int v = *cp >> 1;
420 *(cp++) = cp2[v];
424 if(k == 1 || k == 4) {
425 lv = (hdr[j] & 0xf) + fflags2;
426 correction_type_sp[0] = s->corrector_type + (lv << 8);
427 correction_lp[0] = correction + (lv << 8);
428 lv = (hdr[j] >> 4) + fflags2;
429 correction_lp[1] = correction + (lv << 8);
430 correction_type_sp[1] = s->corrector_type + (lv << 8);
431 } else {
432 correctionloworder_lp[0] = correctionloworder_lp[1] = correctionloworder + (lv << 8);
433 correctionhighorder_lp[0] = correctionhighorder_lp[1] = correctionhighorder + (lv << 8);
434 correction_type_sp[0] = correction_type_sp[1] = s->corrector_type + (lv << 8);
435 correction_lp[0] = correction_lp[1] = correction + (lv << 8);
438 switch(k) {
439 case 1:
440 case 0: /********** CASE 0 **********/
441 for( ; blks_height > 0; blks_height -= 4) {
442 for(lp1 = 0; lp1 < blks_width; lp1++) {
443 for(lp2 = 0; lp2 < 4; ) {
444 k = *buf1++;
445 cur_lp = ((uint32_t *)cur_frm_pos) + width_tbl[lp2];
446 ref_lp = ((uint32_t *)ref_frm_pos) + width_tbl[lp2];
448 switch(correction_type_sp[0][k]) {
449 case 0:
450 *cur_lp = le2me_32(((le2me_32(*ref_lp) >> 1) + correction_lp[lp2 & 0x01][k]) << 1);
451 lp2++;
452 break;
453 case 1:
454 res = ((le2me_16(((unsigned short *)(ref_lp))[0]) >> 1) + correction_lp[lp2 & 0x01][*buf1]) << 1;
455 ((unsigned short *)cur_lp)[0] = le2me_16(res);
456 res = ((le2me_16(((unsigned short *)(ref_lp))[1]) >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
457 ((unsigned short *)cur_lp)[1] = le2me_16(res);
458 buf1++;
459 lp2++;
460 break;
461 case 2:
462 if(lp2 == 0) {
463 for(i = 0, j = 0; i < 2; i++, j += width_tbl[1])
464 cur_lp[j] = ref_lp[j];
465 lp2 += 2;
467 break;
468 case 3:
469 if(lp2 < 2) {
470 for(i = 0, j = 0; i < (3 - lp2); i++, j += width_tbl[1])
471 cur_lp[j] = ref_lp[j];
472 lp2 = 3;
474 break;
475 case 8:
476 if(lp2 == 0) {
477 RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3)
479 if(rle_v1 == 1 || ref_vectors != NULL) {
480 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
481 cur_lp[j] = ref_lp[j];
484 RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2)
485 break;
486 } else {
487 rle_v1 = 1;
488 rle_v2 = *buf1 - 1;
490 case 5:
491 LP2_CHECK(buf1,rle_v3,lp2)
492 case 4:
493 for(i = 0, j = 0; i < (4 - lp2); i++, j += width_tbl[1])
494 cur_lp[j] = ref_lp[j];
495 lp2 = 4;
496 break;
498 case 7:
499 if(rle_v3 != 0)
500 rle_v3 = 0;
501 else {
502 buf1--;
503 rle_v3 = 1;
505 case 6:
506 if(ref_vectors != NULL) {
507 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
508 cur_lp[j] = ref_lp[j];
510 lp2 = 4;
511 break;
513 case 9:
514 lv1 = *buf1++;
515 lv = (lv1 & 0x7F) << 1;
516 lv += (lv << 8);
517 lv += (lv << 16);
518 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
519 cur_lp[j] = lv;
521 LV1_CHECK(buf1,rle_v3,lv1,lp2)
522 break;
523 default:
524 return;
528 cur_frm_pos += 4;
529 ref_frm_pos += 4;
532 cur_frm_pos += ((width - blks_width) * 4);
533 ref_frm_pos += ((width - blks_width) * 4);
535 break;
537 case 4:
538 case 3: /********** CASE 3 **********/
539 if(ref_vectors != NULL)
540 return;
541 flag1 = 1;
543 for( ; blks_height > 0; blks_height -= 8) {
544 for(lp1 = 0; lp1 < blks_width; lp1++) {
545 for(lp2 = 0; lp2 < 4; ) {
546 k = *buf1++;
548 cur_lp = ((uint32_t *)cur_frm_pos) + width_tbl[lp2 * 2];
549 ref_lp = ((uint32_t *)cur_frm_pos) + width_tbl[(lp2 * 2) - 1];
551 switch(correction_type_sp[lp2 & 0x01][k]) {
552 case 0:
553 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(*ref_lp) >> 1) + correction_lp[lp2 & 0x01][k]) << 1);
554 if(lp2 > 0 || flag1 == 0 || strip->ypos != 0)
555 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
556 else
557 cur_lp[0] = le2me_32(((le2me_32(*ref_lp) >> 1) + correction_lp[lp2 & 0x01][k]) << 1);
558 lp2++;
559 break;
561 case 1:
562 res = ((le2me_16(((unsigned short *)ref_lp)[0]) >> 1) + correction_lp[lp2 & 0x01][*buf1]) << 1;
563 ((unsigned short *)cur_lp)[width_tbl[2]] = le2me_16(res);
564 res = ((le2me_16(((unsigned short *)ref_lp)[1]) >> 1) + correction_lp[lp2 & 0x01][k]) << 1;
565 ((unsigned short *)cur_lp)[width_tbl[2]+1] = le2me_16(res);
567 if(lp2 > 0 || flag1 == 0 || strip->ypos != 0)
568 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
569 else
570 cur_lp[0] = cur_lp[width_tbl[1]];
571 buf1++;
572 lp2++;
573 break;
575 case 2:
576 if(lp2 == 0) {
577 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
578 cur_lp[j] = *ref_lp;
579 lp2 += 2;
581 break;
583 case 3:
584 if(lp2 < 2) {
585 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1])
586 cur_lp[j] = *ref_lp;
587 lp2 = 3;
589 break;
591 case 6:
592 lp2 = 4;
593 break;
595 case 7:
596 if(rle_v3 != 0)
597 rle_v3 = 0;
598 else {
599 buf1--;
600 rle_v3 = 1;
602 lp2 = 4;
603 break;
605 case 8:
606 if(lp2 == 0) {
607 RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3)
609 if(rle_v1 == 1) {
610 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
611 cur_lp[j] = ref_lp[j];
614 RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2)
615 break;
616 } else {
617 rle_v2 = (*buf1) - 1;
618 rle_v1 = 1;
620 case 5:
621 LP2_CHECK(buf1,rle_v3,lp2)
622 case 4:
623 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1])
624 cur_lp[j] = *ref_lp;
625 lp2 = 4;
626 break;
628 case 9:
629 av_log(s->avctx, AV_LOG_ERROR, "UNTESTED.\n");
630 lv1 = *buf1++;
631 lv = (lv1 & 0x7F) << 1;
632 lv += (lv << 8);
633 lv += (lv << 16);
635 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
636 cur_lp[j] = lv;
638 LV1_CHECK(buf1,rle_v3,lv1,lp2)
639 break;
641 default:
642 return;
646 cur_frm_pos += 4;
649 cur_frm_pos += (((width * 2) - blks_width) * 4);
650 flag1 = 0;
652 break;
654 case 10: /********** CASE 10 **********/
655 if(ref_vectors == NULL) {
656 flag1 = 1;
658 for( ; blks_height > 0; blks_height -= 8) {
659 for(lp1 = 0; lp1 < blks_width; lp1 += 2) {
660 for(lp2 = 0; lp2 < 4; ) {
661 k = *buf1++;
662 cur_lp = ((uint32_t *)cur_frm_pos) + width_tbl[lp2 * 2];
663 ref_lp = ((uint32_t *)cur_frm_pos) + width_tbl[(lp2 * 2) - 1];
664 lv1 = ref_lp[0];
665 lv2 = ref_lp[1];
666 if(lp2 == 0 && flag1 != 0) {
667 #ifdef WORDS_BIGENDIAN
668 lv1 = lv1 & 0xFF00FF00;
669 lv1 = (lv1 >> 8) | lv1;
670 lv2 = lv2 & 0xFF00FF00;
671 lv2 = (lv2 >> 8) | lv2;
672 #else
673 lv1 = lv1 & 0x00FF00FF;
674 lv1 = (lv1 << 8) | lv1;
675 lv2 = lv2 & 0x00FF00FF;
676 lv2 = (lv2 << 8) | lv2;
677 #endif
680 switch(correction_type_sp[lp2 & 0x01][k]) {
681 case 0:
682 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(lv1) >> 1) + correctionloworder_lp[lp2 & 0x01][k]) << 1);
683 cur_lp[width_tbl[1]+1] = le2me_32(((le2me_32(lv2) >> 1) + correctionhighorder_lp[lp2 & 0x01][k]) << 1);
684 if(lp2 > 0 || strip->ypos != 0 || flag1 == 0) {
685 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
686 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
687 } else {
688 cur_lp[0] = cur_lp[width_tbl[1]];
689 cur_lp[1] = cur_lp[width_tbl[1]+1];
691 lp2++;
692 break;
694 case 1:
695 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(lv1) >> 1) + correctionloworder_lp[lp2 & 0x01][*buf1]) << 1);
696 cur_lp[width_tbl[1]+1] = le2me_32(((le2me_32(lv2) >> 1) + correctionloworder_lp[lp2 & 0x01][k]) << 1);
697 if(lp2 > 0 || strip->ypos != 0 || flag1 == 0) {
698 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
699 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
700 } else {
701 cur_lp[0] = cur_lp[width_tbl[1]];
702 cur_lp[1] = cur_lp[width_tbl[1]+1];
704 buf1++;
705 lp2++;
706 break;
708 case 2:
709 if(lp2 == 0) {
710 if(flag1 != 0) {
711 for(i = 0, j = width_tbl[1]; i < 3; i++, j += width_tbl[1]) {
712 cur_lp[j] = lv1;
713 cur_lp[j+1] = lv2;
715 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
716 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
717 } else {
718 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1]) {
719 cur_lp[j] = lv1;
720 cur_lp[j+1] = lv2;
723 lp2 += 2;
725 break;
727 case 3:
728 if(lp2 < 2) {
729 if(lp2 == 0 && flag1 != 0) {
730 for(i = 0, j = width_tbl[1]; i < 5; i++, j += width_tbl[1]) {
731 cur_lp[j] = lv1;
732 cur_lp[j+1] = lv2;
734 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
735 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
736 } else {
737 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1]) {
738 cur_lp[j] = lv1;
739 cur_lp[j+1] = lv2;
742 lp2 = 3;
744 break;
746 case 8:
747 if(lp2 == 0) {
748 RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3)
749 if(rle_v1 == 1) {
750 if(flag1 != 0) {
751 for(i = 0, j = width_tbl[1]; i < 7; i++, j += width_tbl[1]) {
752 cur_lp[j] = lv1;
753 cur_lp[j+1] = lv2;
755 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
756 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
757 } else {
758 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1]) {
759 cur_lp[j] = lv1;
760 cur_lp[j+1] = lv2;
764 RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2)
765 break;
766 } else {
767 rle_v1 = 1;
768 rle_v2 = (*buf1) - 1;
770 case 5:
771 LP2_CHECK(buf1,rle_v3,lp2)
772 case 4:
773 if(lp2 == 0 && flag1 != 0) {
774 for(i = 0, j = width_tbl[1]; i < 7; i++, j += width_tbl[1]) {
775 cur_lp[j] = lv1;
776 cur_lp[j+1] = lv2;
778 cur_lp[0] = ((cur_lp[-width_tbl[1]] >> 1) + (cur_lp[width_tbl[1]] >> 1)) & 0xFEFEFEFE;
779 cur_lp[1] = ((cur_lp[-width_tbl[1]+1] >> 1) + (cur_lp[width_tbl[1]+1] >> 1)) & 0xFEFEFEFE;
780 } else {
781 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1]) {
782 cur_lp[j] = lv1;
783 cur_lp[j+1] = lv2;
786 lp2 = 4;
787 break;
789 case 6:
790 lp2 = 4;
791 break;
793 case 7:
794 if(lp2 == 0) {
795 if(rle_v3 != 0)
796 rle_v3 = 0;
797 else {
798 buf1--;
799 rle_v3 = 1;
801 lp2 = 4;
803 break;
805 case 9:
806 av_log(s->avctx, AV_LOG_ERROR, "UNTESTED.\n");
807 lv1 = *buf1;
808 lv = (lv1 & 0x7F) << 1;
809 lv += (lv << 8);
810 lv += (lv << 16);
811 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
812 cur_lp[j] = lv;
813 LV1_CHECK(buf1,rle_v3,lv1,lp2)
814 break;
816 default:
817 return;
821 cur_frm_pos += 8;
824 cur_frm_pos += (((width * 2) - blks_width) * 4);
825 flag1 = 0;
827 } else {
828 for( ; blks_height > 0; blks_height -= 8) {
829 for(lp1 = 0; lp1 < blks_width; lp1 += 2) {
830 for(lp2 = 0; lp2 < 4; ) {
831 k = *buf1++;
832 cur_lp = ((uint32_t *)cur_frm_pos) + width_tbl[lp2 * 2];
833 ref_lp = ((uint32_t *)ref_frm_pos) + width_tbl[lp2 * 2];
835 switch(correction_type_sp[lp2 & 0x01][k]) {
836 case 0:
837 lv1 = correctionloworder_lp[lp2 & 0x01][k];
838 lv2 = correctionhighorder_lp[lp2 & 0x01][k];
839 cur_lp[0] = le2me_32(((le2me_32(ref_lp[0]) >> 1) + lv1) << 1);
840 cur_lp[1] = le2me_32(((le2me_32(ref_lp[1]) >> 1) + lv2) << 1);
841 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(ref_lp[width_tbl[1]]) >> 1) + lv1) << 1);
842 cur_lp[width_tbl[1]+1] = le2me_32(((le2me_32(ref_lp[width_tbl[1]+1]) >> 1) + lv2) << 1);
843 lp2++;
844 break;
846 case 1:
847 lv1 = correctionloworder_lp[lp2 & 0x01][*buf1++];
848 lv2 = correctionloworder_lp[lp2 & 0x01][k];
849 cur_lp[0] = le2me_32(((le2me_32(ref_lp[0]) >> 1) + lv1) << 1);
850 cur_lp[1] = le2me_32(((le2me_32(ref_lp[1]) >> 1) + lv2) << 1);
851 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(ref_lp[width_tbl[1]]) >> 1) + lv1) << 1);
852 cur_lp[width_tbl[1]+1] = le2me_32(((le2me_32(ref_lp[width_tbl[1]+1]) >> 1) + lv2) << 1);
853 lp2++;
854 break;
856 case 2:
857 if(lp2 == 0) {
858 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1]) {
859 cur_lp[j] = ref_lp[j];
860 cur_lp[j+1] = ref_lp[j+1];
862 lp2 += 2;
864 break;
866 case 3:
867 if(lp2 < 2) {
868 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1]) {
869 cur_lp[j] = ref_lp[j];
870 cur_lp[j+1] = ref_lp[j+1];
872 lp2 = 3;
874 break;
876 case 8:
877 if(lp2 == 0) {
878 RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3)
879 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1]) {
880 ((uint32_t *)cur_frm_pos)[j] = ((uint32_t *)ref_frm_pos)[j];
881 ((uint32_t *)cur_frm_pos)[j+1] = ((uint32_t *)ref_frm_pos)[j+1];
883 RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2)
884 break;
885 } else {
886 rle_v1 = 1;
887 rle_v2 = (*buf1) - 1;
889 case 5:
890 case 7:
891 LP2_CHECK(buf1,rle_v3,lp2)
892 case 6:
893 case 4:
894 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1]) {
895 cur_lp[j] = ref_lp[j];
896 cur_lp[j+1] = ref_lp[j+1];
898 lp2 = 4;
899 break;
901 case 9:
902 av_log(s->avctx, AV_LOG_ERROR, "UNTESTED.\n");
903 lv1 = *buf1;
904 lv = (lv1 & 0x7F) << 1;
905 lv += (lv << 8);
906 lv += (lv << 16);
907 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
908 ((uint32_t *)cur_frm_pos)[j] = ((uint32_t *)cur_frm_pos)[j+1] = lv;
909 LV1_CHECK(buf1,rle_v3,lv1,lp2)
910 break;
912 default:
913 return;
917 cur_frm_pos += 8;
918 ref_frm_pos += 8;
921 cur_frm_pos += (((width * 2) - blks_width) * 4);
922 ref_frm_pos += (((width * 2) - blks_width) * 4);
925 break;
927 case 11: /********** CASE 11 **********/
928 if(ref_vectors == NULL)
929 return;
931 for( ; blks_height > 0; blks_height -= 8) {
932 for(lp1 = 0; lp1 < blks_width; lp1++) {
933 for(lp2 = 0; lp2 < 4; ) {
934 k = *buf1++;
935 cur_lp = ((uint32_t *)cur_frm_pos) + width_tbl[lp2 * 2];
936 ref_lp = ((uint32_t *)ref_frm_pos) + width_tbl[lp2 * 2];
938 switch(correction_type_sp[lp2 & 0x01][k]) {
939 case 0:
940 cur_lp[0] = le2me_32(((le2me_32(*ref_lp) >> 1) + correction_lp[lp2 & 0x01][k]) << 1);
941 cur_lp[width_tbl[1]] = le2me_32(((le2me_32(ref_lp[width_tbl[1]]) >> 1) + correction_lp[lp2 & 0x01][k]) << 1);
942 lp2++;
943 break;
945 case 1:
946 lv1 = (unsigned short)(correction_lp[lp2 & 0x01][*buf1++]);
947 lv2 = (unsigned short)(correction_lp[lp2 & 0x01][k]);
948 res = (unsigned short)(((le2me_16(((unsigned short *)ref_lp)[0]) >> 1) + lv1) << 1);
949 ((unsigned short *)cur_lp)[0] = le2me_16(res);
950 res = (unsigned short)(((le2me_16(((unsigned short *)ref_lp)[1]) >> 1) + lv2) << 1);
951 ((unsigned short *)cur_lp)[1] = le2me_16(res);
952 res = (unsigned short)(((le2me_16(((unsigned short *)ref_lp)[width_tbl[2]]) >> 1) + lv1) << 1);
953 ((unsigned short *)cur_lp)[width_tbl[2]] = le2me_16(res);
954 res = (unsigned short)(((le2me_16(((unsigned short *)ref_lp)[width_tbl[2]+1]) >> 1) + lv2) << 1);
955 ((unsigned short *)cur_lp)[width_tbl[2]+1] = le2me_16(res);
956 lp2++;
957 break;
959 case 2:
960 if(lp2 == 0) {
961 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
962 cur_lp[j] = ref_lp[j];
963 lp2 += 2;
965 break;
967 case 3:
968 if(lp2 < 2) {
969 for(i = 0, j = 0; i < 6 - (lp2 * 2); i++, j += width_tbl[1])
970 cur_lp[j] = ref_lp[j];
971 lp2 = 3;
973 break;
975 case 8:
976 if(lp2 == 0) {
977 RLE_V3_CHECK(buf1,rle_v1,rle_v2,rle_v3)
979 for(i = 0, j = 0; i < 8; i++, j += width_tbl[1])
980 cur_lp[j] = ref_lp[j];
982 RLE_V2_CHECK(buf1,rle_v2, rle_v3,lp2)
983 break;
984 } else {
985 rle_v1 = 1;
986 rle_v2 = (*buf1) - 1;
988 case 5:
989 case 7:
990 LP2_CHECK(buf1,rle_v3,lp2)
991 case 4:
992 case 6:
993 for(i = 0, j = 0; i < 8 - (lp2 * 2); i++, j += width_tbl[1])
994 cur_lp[j] = ref_lp[j];
995 lp2 = 4;
996 break;
998 case 9:
999 av_log(s->avctx, AV_LOG_ERROR, "UNTESTED.\n");
1000 lv1 = *buf1++;
1001 lv = (lv1 & 0x7F) << 1;
1002 lv += (lv << 8);
1003 lv += (lv << 16);
1004 for(i = 0, j = 0; i < 4; i++, j += width_tbl[1])
1005 cur_lp[j] = lv;
1006 LV1_CHECK(buf1,rle_v3,lv1,lp2)
1007 break;
1009 default:
1010 return;
1014 cur_frm_pos += 4;
1015 ref_frm_pos += 4;
1018 cur_frm_pos += (((width * 2) - blks_width) * 4);
1019 ref_frm_pos += (((width * 2) - blks_width) * 4);
1021 break;
1023 default:
1024 return;
1028 if(strip < strip_tbl)
1029 return;
1031 for( ; strip >= strip_tbl; strip--) {
1032 if(strip->split_flag != 0) {
1033 strip->split_flag = 0;
1034 strip->usl7 = (strip-1)->usl7;
1036 if(strip->split_direction) {
1037 strip->xpos += strip->width;
1038 strip->width = (strip-1)->width - strip->width;
1039 if(region_160_width <= strip->xpos && width < strip->width + strip->xpos)
1040 strip->width = width - strip->xpos;
1041 } else {
1042 strip->ypos += strip->height;
1043 strip->height = (strip-1)->height - strip->height;
1045 break;
1051 static av_cold int indeo3_decode_init(AVCodecContext *avctx)
1053 Indeo3DecodeContext *s = avctx->priv_data;
1055 s->avctx = avctx;
1056 s->width = avctx->width;
1057 s->height = avctx->height;
1058 avctx->pix_fmt = PIX_FMT_YUV410P;
1060 build_modpred(s);
1061 iv_alloc_frames(s);
1063 return 0;
1066 static int indeo3_decode_frame(AVCodecContext *avctx,
1067 void *data, int *data_size,
1068 const unsigned char *buf, int buf_size)
1070 Indeo3DecodeContext *s=avctx->priv_data;
1071 unsigned char *src, *dest;
1072 int y;
1074 iv_decode_frame(s, buf, buf_size);
1076 if(s->frame.data[0])
1077 avctx->release_buffer(avctx, &s->frame);
1079 s->frame.reference = 0;
1080 if(avctx->get_buffer(avctx, &s->frame) < 0) {
1081 av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1082 return -1;
1085 src = s->cur_frame->Ybuf;
1086 dest = s->frame.data[0];
1087 for (y = 0; y < s->height; y++) {
1088 memcpy(dest, src, s->cur_frame->y_w);
1089 src += s->cur_frame->y_w;
1090 dest += s->frame.linesize[0];
1093 if (!(s->avctx->flags & CODEC_FLAG_GRAY))
1095 src = s->cur_frame->Ubuf;
1096 dest = s->frame.data[1];
1097 for (y = 0; y < s->height / 4; y++) {
1098 memcpy(dest, src, s->cur_frame->uv_w);
1099 src += s->cur_frame->uv_w;
1100 dest += s->frame.linesize[1];
1103 src = s->cur_frame->Vbuf;
1104 dest = s->frame.data[2];
1105 for (y = 0; y < s->height / 4; y++) {
1106 memcpy(dest, src, s->cur_frame->uv_w);
1107 src += s->cur_frame->uv_w;
1108 dest += s->frame.linesize[2];
1112 *data_size=sizeof(AVFrame);
1113 *(AVFrame*)data= s->frame;
1115 return buf_size;
1118 static av_cold int indeo3_decode_end(AVCodecContext *avctx)
1120 Indeo3DecodeContext *s = avctx->priv_data;
1122 iv_free_func(s);
1124 return 0;
1127 AVCodec indeo3_decoder = {
1128 "indeo3",
1129 CODEC_TYPE_VIDEO,
1130 CODEC_ID_INDEO3,
1131 sizeof(Indeo3DecodeContext),
1132 indeo3_decode_init,
1133 NULL,
1134 indeo3_decode_end,
1135 indeo3_decode_frame,
1137 NULL,
1138 .long_name = NULL_IF_CONFIG_SMALL("Intel Indeo 3"),