More correct printf format specifiers
[libav.git] / libavcodec / utvideodec.c
1 /*
2 * Ut Video decoder
3 * Copyright (c) 2011 Konstantin Shishkov
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * Ut Video decoder
25 */
26
27 #include <inttypes.h>
28 #include <stdlib.h>
29
30 #include "libavutil/intreadwrite.h"
31 #include "avcodec.h"
32 #include "bytestream.h"
33 #include "get_bits.h"
34 #include "dsputil.h"
35 #include "thread.h"
36 #include "utvideo.h"
37
38 static int build_huff(const uint8_t *src, VLC *vlc, int *fsym)
39 {
40 int i;
41 HuffEntry he[256];
42 int last;
43 uint32_t codes[256];
44 uint8_t bits[256];
45 uint8_t syms[256];
46 uint32_t code;
47
48 *fsym = -1;
49 for (i = 0; i < 256; i++) {
50 he[i].sym = i;
51 he[i].len = *src++;
52 }
53 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
54
55 if (!he[0].len) {
56 *fsym = he[0].sym;
57 return 0;
58 }
59 if (he[0].len > 32)
60 return -1;
61
62 last = 255;
63 while (he[last].len == 255 && last)
64 last--;
65
66 code = 1;
67 for (i = last; i >= 0; i--) {
68 codes[i] = code >> (32 - he[i].len);
69 bits[i] = he[i].len;
70 syms[i] = he[i].sym;
71 code += 0x80000000u >> (he[i].len - 1);
72 }
73
74 return ff_init_vlc_sparse(vlc, FFMIN(he[last].len, 9), last + 1,
75 bits, sizeof(*bits), sizeof(*bits),
76 codes, sizeof(*codes), sizeof(*codes),
77 syms, sizeof(*syms), sizeof(*syms), 0);
78 }
79
80 static int decode_plane(UtvideoContext *c, int plane_no,
81 uint8_t *dst, int step, int stride,
82 int width, int height,
83 const uint8_t *src, int use_pred)
84 {
85 int i, j, slice, pix;
86 int sstart, send;
87 VLC vlc;
88 GetBitContext gb;
89 int prev, fsym;
90 const int cmask = ~(!plane_no && c->avctx->pix_fmt == AV_PIX_FMT_YUV420P);
91
92 if (build_huff(src, &vlc, &fsym)) {
93 av_log(c->avctx, AV_LOG_ERROR, "Cannot build Huffman codes\n");
94 return AVERROR_INVALIDDATA;
95 }
96 if (fsym >= 0) { // build_huff reported a symbol to fill slices with
97 send = 0;
98 for (slice = 0; slice < c->slices; slice++) {
99 uint8_t *dest;
100
101 sstart = send;
102 send = (height * (slice + 1) / c->slices) & cmask;
103 dest = dst + sstart * stride;
104
105 prev = 0x80;
106 for (j = sstart; j < send; j++) {
107 for (i = 0; i < width * step; i += step) {
108 pix = fsym;
109 if (use_pred) {
110 prev += pix;
111 pix = prev;
112 }
113 dest[i] = pix;
114 }
115 dest += stride;
116 }
117 }
118 return 0;
119 }
120
121 src += 256;
122
123 send = 0;
124 for (slice = 0; slice < c->slices; slice++) {
125 uint8_t *dest;
126 int slice_data_start, slice_data_end, slice_size;
127
128 sstart = send;
129 send = (height * (slice + 1) / c->slices) & cmask;
130 dest = dst + sstart * stride;
131
132 // slice offset and size validation was done earlier
133 slice_data_start = slice ? AV_RL32(src + slice * 4 - 4) : 0;
134 slice_data_end = AV_RL32(src + slice * 4);
135 slice_size = slice_data_end - slice_data_start;
136
137 if (!slice_size) {
138 av_log(c->avctx, AV_LOG_ERROR, "Plane has more than one symbol "
139 "yet a slice has a length of zero.\n");
140 goto fail;
141 }
142
143 memcpy(c->slice_bits, src + slice_data_start + c->slices * 4,
144 slice_size);
145 memset(c->slice_bits + slice_size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
146 c->dsp.bswap_buf((uint32_t *) c->slice_bits, (uint32_t *) c->slice_bits,
147 (slice_data_end - slice_data_start + 3) >> 2);
148 init_get_bits(&gb, c->slice_bits, slice_size * 8);
149
150 prev = 0x80;
151 for (j = sstart; j < send; j++) {
152 for (i = 0; i < width * step; i += step) {
153 if (get_bits_left(&gb) <= 0) {
154 av_log(c->avctx, AV_LOG_ERROR,
155 "Slice decoding ran out of bits\n");
156 goto fail;
157 }
158 pix = get_vlc2(&gb, vlc.table, vlc.bits, 4);
159 if (pix < 0) {
160 av_log(c->avctx, AV_LOG_ERROR, "Decoding error\n");
161 goto fail;
162 }
163 if (use_pred) {
164 prev += pix;
165 pix = prev;
166 }
167 dest[i] = pix;
168 }
169 dest += stride;
170 }
171 if (get_bits_left(&gb) > 32)
172 av_log(c->avctx, AV_LOG_WARNING,
173 "%d bits left after decoding slice\n", get_bits_left(&gb));
174 }
175
176 ff_free_vlc(&vlc);
177
178 return 0;
179 fail:
180 ff_free_vlc(&vlc);
181 return AVERROR_INVALIDDATA;
182 }
183
184 static void restore_rgb_planes(uint8_t *src, int step, int stride, int width,
185 int height)
186 {
187 int i, j;
188 uint8_t r, g, b;
189
190 for (j = 0; j < height; j++) {
191 for (i = 0; i < width * step; i += step) {
192 r = src[i];
193 g = src[i + 1];
194 b = src[i + 2];
195 src[i] = r + g - 0x80;
196 src[i + 2] = b + g - 0x80;
197 }
198 src += stride;
199 }
200 }
201
202 static void restore_median(uint8_t *src, int step, int stride,
203 int width, int height, int slices, int rmode)
204 {
205 int i, j, slice;
206 int A, B, C;
207 uint8_t *bsrc;
208 int slice_start, slice_height;
209 const int cmask = ~rmode;
210
211 for (slice = 0; slice < slices; slice++) {
212 slice_start = ((slice * height) / slices) & cmask;
213 slice_height = ((((slice + 1) * height) / slices) & cmask) -
214 slice_start;
215
216 bsrc = src + slice_start * stride;
217
218 // first line - left neighbour prediction
219 bsrc[0] += 0x80;
220 A = bsrc[0];
221 for (i = step; i < width * step; i += step) {
222 bsrc[i] += A;
223 A = bsrc[i];
224 }
225 bsrc += stride;
226 if (slice_height == 1)
227 continue;
228 // second line - first element has top prediction, the rest uses median
229 C = bsrc[-stride];
230 bsrc[0] += C;
231 A = bsrc[0];
232 for (i = step; i < width * step; i += step) {
233 B = bsrc[i - stride];
234 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
235 C = B;
236 A = bsrc[i];
237 }
238 bsrc += stride;
239 // the rest of lines use continuous median prediction
240 for (j = 2; j < slice_height; j++) {
241 for (i = 0; i < width * step; i += step) {
242 B = bsrc[i - stride];
243 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
244 C = B;
245 A = bsrc[i];
246 }
247 bsrc += stride;
248 }
249 }
250 }
251
252 /* UtVideo interlaced mode treats every two lines as a single one,
253 * so restoring function should take care of possible padding between
254 * two parts of the same "line".
255 */
256 static void restore_median_il(uint8_t *src, int step, int stride,
257 int width, int height, int slices, int rmode)
258 {
259 int i, j, slice;
260 int A, B, C;
261 uint8_t *bsrc;
262 int slice_start, slice_height;
263 const int cmask = ~(rmode ? 3 : 1);
264 const int stride2 = stride << 1;
265
266 for (slice = 0; slice < slices; slice++) {
267 slice_start = ((slice * height) / slices) & cmask;
268 slice_height = ((((slice + 1) * height) / slices) & cmask) -
269 slice_start;
270 slice_height >>= 1;
271
272 bsrc = src + slice_start * stride;
273
274 // first line - left neighbour prediction
275 bsrc[0] += 0x80;
276 A = bsrc[0];
277 for (i = step; i < width * step; i += step) {
278 bsrc[i] += A;
279 A = bsrc[i];
280 }
281 for (i = 0; i < width * step; i += step) {
282 bsrc[stride + i] += A;
283 A = bsrc[stride + i];
284 }
285 bsrc += stride2;
286 if (slice_height == 1)
287 continue;
288 // second line - first element has top prediction, the rest uses median
289 C = bsrc[-stride2];
290 bsrc[0] += C;
291 A = bsrc[0];
292 for (i = step; i < width * step; i += step) {
293 B = bsrc[i - stride2];
294 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
295 C = B;
296 A = bsrc[i];
297 }
298 for (i = 0; i < width * step; i += step) {
299 B = bsrc[i - stride];
300 bsrc[stride + i] += mid_pred(A, B, (uint8_t)(A + B - C));
301 C = B;
302 A = bsrc[stride + i];
303 }
304 bsrc += stride2;
305 // the rest of lines use continuous median prediction
306 for (j = 2; j < slice_height; j++) {
307 for (i = 0; i < width * step; i += step) {
308 B = bsrc[i - stride2];
309 bsrc[i] += mid_pred(A, B, (uint8_t)(A + B - C));
310 C = B;
311 A = bsrc[i];
312 }
313 for (i = 0; i < width * step; i += step) {
314 B = bsrc[i - stride];
315 bsrc[i + stride] += mid_pred(A, B, (uint8_t)(A + B - C));
316 C = B;
317 A = bsrc[i + stride];
318 }
319 bsrc += stride2;
320 }
321 }
322 }
323
324 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
325 AVPacket *avpkt)
326 {
327 const uint8_t *buf = avpkt->data;
328 int buf_size = avpkt->size;
329 UtvideoContext *c = avctx->priv_data;
330 int i, j;
331 const uint8_t *plane_start[5];
332 int plane_size, max_slice_size = 0, slice_start, slice_end, slice_size;
333 int ret;
334 GetByteContext gb;
335 ThreadFrame frame = { .f = data };
336
337 if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) {
338 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
339 return ret;
340 }
341
342 ff_thread_finish_setup(avctx);
343
344 /* parse plane structure to get frame flags and validate slice offsets */
345 bytestream2_init(&gb, buf, buf_size);
346 for (i = 0; i < c->planes; i++) {
347 plane_start[i] = gb.buffer;
348 if (bytestream2_get_bytes_left(&gb) < 256 + 4 * c->slices) {
349 av_log(avctx, AV_LOG_ERROR, "Insufficient data for a plane\n");
350 return AVERROR_INVALIDDATA;
351 }
352 bytestream2_skipu(&gb, 256);
353 slice_start = 0;
354 slice_end = 0;
355 for (j = 0; j < c->slices; j++) {
356 slice_end = bytestream2_get_le32u(&gb);
357 slice_size = slice_end - slice_start;
358 if (slice_end < 0 || slice_size < 0 ||
359 bytestream2_get_bytes_left(&gb) < slice_end) {
360 av_log(avctx, AV_LOG_ERROR, "Incorrect slice size\n");
361 return AVERROR_INVALIDDATA;
362 }
363 slice_start = slice_end;
364 max_slice_size = FFMAX(max_slice_size, slice_size);
365 }
366 plane_size = slice_end;
367 bytestream2_skipu(&gb, plane_size);
368 }
369 plane_start[c->planes] = gb.buffer;
370 if (bytestream2_get_bytes_left(&gb) < c->frame_info_size) {
371 av_log(avctx, AV_LOG_ERROR, "Not enough data for frame information\n");
372 return AVERROR_INVALIDDATA;
373 }
374 c->frame_info = bytestream2_get_le32u(&gb);
375 av_log(avctx, AV_LOG_DEBUG, "frame information flags %"PRIX32"\n",
376 c->frame_info);
377
378 c->frame_pred = (c->frame_info >> 8) & 3;
379
380 if (c->frame_pred == PRED_GRADIENT) {
381 avpriv_request_sample(avctx, "Frame with gradient prediction");
382 return AVERROR_PATCHWELCOME;
383 }
384
385 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
386 max_slice_size + FF_INPUT_BUFFER_PADDING_SIZE);
387
388 if (!c->slice_bits) {
389 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer\n");
390 return AVERROR(ENOMEM);
391 }
392
393 switch (c->avctx->pix_fmt) {
394 case AV_PIX_FMT_RGB24:
395 case AV_PIX_FMT_RGBA:
396 for (i = 0; i < c->planes; i++) {
397 ret = decode_plane(c, i, frame.f->data[0] + ff_ut_rgb_order[i],
398 c->planes, frame.f->linesize[0], avctx->width,
399 avctx->height, plane_start[i],
400 c->frame_pred == PRED_LEFT);
401 if (ret)
402 return ret;
403 if (c->frame_pred == PRED_MEDIAN) {
404 if (!c->interlaced) {
405 restore_median(frame.f->data[0] + ff_ut_rgb_order[i],
406 c->planes, frame.f->linesize[0], avctx->width,
407 avctx->height, c->slices, 0);
408 } else {
409 restore_median_il(frame.f->data[0] + ff_ut_rgb_order[i],
410 c->planes, frame.f->linesize[0],
411 avctx->width, avctx->height, c->slices,
412 0);
413 }
414 }
415 }
416 restore_rgb_planes(frame.f->data[0], c->planes, frame.f->linesize[0],
417 avctx->width, avctx->height);
418 break;
419 case AV_PIX_FMT_YUV420P:
420 for (i = 0; i < 3; i++) {
421 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
422 avctx->width >> !!i, avctx->height >> !!i,
423 plane_start[i], c->frame_pred == PRED_LEFT);
424 if (ret)
425 return ret;
426 if (c->frame_pred == PRED_MEDIAN) {
427 if (!c->interlaced) {
428 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
429 avctx->width >> !!i, avctx->height >> !!i,
430 c->slices, !i);
431 } else {
432 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
433 avctx->width >> !!i,
434 avctx->height >> !!i,
435 c->slices, !i);
436 }
437 }
438 }
439 break;
440 case AV_PIX_FMT_YUV422P:
441 for (i = 0; i < 3; i++) {
442 ret = decode_plane(c, i, frame.f->data[i], 1, frame.f->linesize[i],
443 avctx->width >> !!i, avctx->height,
444 plane_start[i], c->frame_pred == PRED_LEFT);
445 if (ret)
446 return ret;
447 if (c->frame_pred == PRED_MEDIAN) {
448 if (!c->interlaced) {
449 restore_median(frame.f->data[i], 1, frame.f->linesize[i],
450 avctx->width >> !!i, avctx->height,
451 c->slices, 0);
452 } else {
453 restore_median_il(frame.f->data[i], 1, frame.f->linesize[i],
454 avctx->width >> !!i, avctx->height,
455 c->slices, 0);
456 }
457 }
458 }
459 break;
460 }
461
462 frame.f->key_frame = 1;
463 frame.f->pict_type = AV_PICTURE_TYPE_I;
464 frame.f->interlaced_frame = !!c->interlaced;
465
466 *got_frame = 1;
467
468 /* always report that the buffer was completely consumed */
469 return buf_size;
470 }
471
472 static av_cold int decode_init(AVCodecContext *avctx)
473 {
474 UtvideoContext * const c = avctx->priv_data;
475
476 c->avctx = avctx;
477
478 ff_dsputil_init(&c->dsp, avctx);
479
480 if (avctx->extradata_size < 16) {
481 av_log(avctx, AV_LOG_ERROR,
482 "Insufficient extradata size %d, should be at least 16\n",
483 avctx->extradata_size);
484 return AVERROR_INVALIDDATA;
485 }
486
487 av_log(avctx, AV_LOG_DEBUG, "Encoder version %d.%d.%d.%d\n",
488 avctx->extradata[3], avctx->extradata[2],
489 avctx->extradata[1], avctx->extradata[0]);
490 av_log(avctx, AV_LOG_DEBUG, "Original format %"PRIX32"\n",
491 AV_RB32(avctx->extradata + 4));
492 c->frame_info_size = AV_RL32(avctx->extradata + 8);
493 c->flags = AV_RL32(avctx->extradata + 12);
494
495 if (c->frame_info_size != 4)
496 avpriv_request_sample(avctx, "Frame info not 4 bytes");
497 av_log(avctx, AV_LOG_DEBUG, "Encoding parameters %08"PRIX32"\n", c->flags);
498 c->slices = (c->flags >> 24) + 1;
499 c->compression = c->flags & 1;
500 c->interlaced = c->flags & 0x800;
501
502 c->slice_bits_size = 0;
503
504 switch (avctx->codec_tag) {
505 case MKTAG('U', 'L', 'R', 'G'):
506 c->planes = 3;
507 avctx->pix_fmt = AV_PIX_FMT_RGB24;
508 break;
509 case MKTAG('U', 'L', 'R', 'A'):
510 c->planes = 4;
511 avctx->pix_fmt = AV_PIX_FMT_RGBA;
512 break;
513 case MKTAG('U', 'L', 'Y', '0'):
514 c->planes = 3;
515 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
516 avctx->colorspace = AVCOL_SPC_BT470BG;
517 break;
518 case MKTAG('U', 'L', 'Y', '2'):
519 c->planes = 3;
520 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
521 avctx->colorspace = AVCOL_SPC_BT470BG;
522 break;
523 case MKTAG('U', 'L', 'H', '0'):
524 c->planes = 3;
525 avctx->pix_fmt = AV_PIX_FMT_YUV420P;
526 avctx->colorspace = AVCOL_SPC_BT709;
527 break;
528 case MKTAG('U', 'L', 'H', '2'):
529 c->planes = 3;
530 avctx->pix_fmt = AV_PIX_FMT_YUV422P;
531 avctx->colorspace = AVCOL_SPC_BT709;
532 break;
533 default:
534 av_log(avctx, AV_LOG_ERROR, "Unknown Ut Video FOURCC provided (%08X)\n",
535 avctx->codec_tag);
536 return AVERROR_INVALIDDATA;
537 }
538
539 return 0;
540 }
541
542 static av_cold int decode_end(AVCodecContext *avctx)
543 {
544 UtvideoContext * const c = avctx->priv_data;
545
546 av_freep(&c->slice_bits);
547
548 return 0;
549 }
550
551 AVCodec ff_utvideo_decoder = {
552 .name = "utvideo",
553 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
554 .type = AVMEDIA_TYPE_VIDEO,
555 .id = AV_CODEC_ID_UTVIDEO,
556 .priv_data_size = sizeof(UtvideoContext),
557 .init = decode_init,
558 .close = decode_end,
559 .decode = decode_frame,
560 .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS,
561 };