lavc: Move b_frame_strategy and b_sensitivity to codec private options
[libav.git] / libavcodec / libxavs.c
1 /*
2 * AVS encoding using the xavs library
3 * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
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 #include <stdio.h>
23 #include <stdlib.h>
24 #include <string.h>
25 #include <math.h>
26 #include <stdint.h>
27 #include <float.h>
28 #include <xavs.h>
29 #include "avcodec.h"
30 #include "internal.h"
31 #include "libavutil/internal.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/opt.h"
34
35 #define END_OF_STREAM 0x001
36
37 #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
38 #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
39 #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
40
41 typedef struct XavsContext {
42 AVClass *class;
43 xavs_param_t params;
44 xavs_t *enc;
45 xavs_picture_t pic;
46 uint8_t *sei;
47 int sei_size;
48 int end_of_stream;
49 float crf;
50 int cqp;
51 int b_bias;
52 float cplxblur;
53 int direct_pred;
54 int aud;
55 int fast_pskip;
56 int motion_est;
57 int mbtree;
58 int mixed_refs;
59 int b_frame_strategy;
60
61 int64_t *pts_buffer;
62 int out_frame_count;
63 } XavsContext;
64
65 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
66 {
67 static const int level_map[] = {
68 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
69 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
70 [XAVS_LOG_INFO] = AV_LOG_INFO,
71 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
72 };
73
74 if (level < 0 || level > XAVS_LOG_DEBUG)
75 return;
76
77 av_vlog(p, level_map[level], fmt, args);
78 }
79
80 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
81 xavs_nal_t *nals, int nnal)
82 {
83 XavsContext *x4 = ctx->priv_data;
84 uint8_t *p;
85 int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
86
87 if (!nnal)
88 return 0;
89
90 for (i = 0; i < nnal; i++)
91 size += nals[i].i_payload;
92
93 if ((ret = ff_alloc_packet(pkt, size)) < 0) {
94 av_log(ctx, AV_LOG_ERROR, "Error getting output packet of size %d.\n", size);
95 return ret;
96 }
97 p = pkt->data;
98
99 /* Write the SEI as part of the first frame. */
100 if (x4->sei_size > 0 && nnal > 0) {
101 memcpy(p, x4->sei, x4->sei_size);
102 p += x4->sei_size;
103 x4->sei_size = 0;
104 }
105
106 for (i = 0; i < nnal; i++) {
107 s = xavs_nal_encode(p, &size, 1, nals + i);
108 if (s < 0)
109 return -1;
110 p += s;
111 }
112 pkt->size = p - pkt->data;
113
114 return 1;
115 }
116
117 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
118 const AVFrame *frame, int *got_packet)
119 {
120 XavsContext *x4 = avctx->priv_data;
121 xavs_nal_t *nal;
122 int nnal, i, ret;
123 xavs_picture_t pic_out;
124 uint8_t *sd;
125
126 x4->pic.img.i_csp = XAVS_CSP_I420;
127 x4->pic.img.i_plane = 3;
128
129 if (frame) {
130 for (i = 0; i < 3; i++) {
131 x4->pic.img.plane[i] = frame->data[i];
132 x4->pic.img.i_stride[i] = frame->linesize[i];
133 }
134
135 x4->pic.i_pts = frame->pts;
136 x4->pic.i_type = XAVS_TYPE_AUTO;
137 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
138 }
139
140 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
141 frame? &x4->pic: NULL, &pic_out) < 0)
142 return -1;
143
144 ret = encode_nals(avctx, pkt, nal, nnal);
145
146 if (ret < 0)
147 return -1;
148
149 if (!ret) {
150 if (!frame && !(x4->end_of_stream)) {
151 if ((ret = ff_alloc_packet(pkt, 4)) < 0)
152 return ret;
153
154 pkt->data[0] = 0x0;
155 pkt->data[1] = 0x0;
156 pkt->data[2] = 0x01;
157 pkt->data[3] = 0xb1;
158 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
159 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
160 x4->end_of_stream = END_OF_STREAM;
161 *got_packet = 1;
162 }
163 return 0;
164 }
165
166 #if FF_API_CODED_FRAME
167 FF_DISABLE_DEPRECATION_WARNINGS
168 avctx->coded_frame->pts = pic_out.i_pts;
169 FF_ENABLE_DEPRECATION_WARNINGS
170 #endif
171 pkt->pts = pic_out.i_pts;
172 if (avctx->has_b_frames) {
173 if (!x4->out_frame_count)
174 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
175 else
176 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
177 } else
178 pkt->dts = pkt->pts;
179
180 #if FF_API_CODED_FRAME
181 FF_DISABLE_DEPRECATION_WARNINGS
182 switch (pic_out.i_type) {
183 case XAVS_TYPE_IDR:
184 case XAVS_TYPE_I:
185 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
186 break;
187 case XAVS_TYPE_P:
188 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
189 break;
190 case XAVS_TYPE_B:
191 case XAVS_TYPE_BREF:
192 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
193 break;
194 }
195 FF_ENABLE_DEPRECATION_WARNINGS
196 #endif
197
198 /* There is no IDR frame in AVS JiZhun */
199 /* Sequence header is used as a flag */
200 if (pic_out.i_type == XAVS_TYPE_I) {
201 #if FF_API_CODED_FRAME
202 FF_DISABLE_DEPRECATION_WARNINGS
203 avctx->coded_frame->key_frame = 1;
204 FF_ENABLE_DEPRECATION_WARNINGS
205 #endif
206 pkt->flags |= AV_PKT_FLAG_KEY;
207 }
208
209 #if FF_API_CODED_FRAME
210 FF_DISABLE_DEPRECATION_WARNINGS
211 avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
212 FF_ENABLE_DEPRECATION_WARNINGS
213 #endif
214
215 sd = av_packet_new_side_data(pkt, AV_PKT_DATA_QUALITY_FACTOR, sizeof(int));
216 if (!sd)
217 return AVERROR(ENOMEM);
218 *(int *)sd = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
219
220 x4->out_frame_count++;
221 *got_packet = ret;
222 return 0;
223 }
224
225 static av_cold int XAVS_close(AVCodecContext *avctx)
226 {
227 XavsContext *x4 = avctx->priv_data;
228
229 av_freep(&avctx->extradata);
230 av_free(x4->sei);
231 av_freep(&x4->pts_buffer);
232
233 if (x4->enc)
234 xavs_encoder_close(x4->enc);
235
236 return 0;
237 }
238
239 static av_cold int XAVS_init(AVCodecContext *avctx)
240 {
241 XavsContext *x4 = avctx->priv_data;
242
243 x4->sei_size = 0;
244 xavs_param_default(&x4->params);
245
246 x4->params.pf_log = XAVS_log;
247 x4->params.p_log_private = avctx;
248 x4->params.i_keyint_max = avctx->gop_size;
249 if (avctx->bit_rate) {
250 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
251 x4->params.rc.i_rc_method = XAVS_RC_ABR;
252 }
253 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
254 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
255 x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
256 if (avctx->flags & AV_CODEC_FLAG_PASS2) {
257 x4->params.rc.b_stat_read = 1;
258 } else {
259 if (x4->crf >= 0) {
260 x4->params.rc.i_rc_method = XAVS_RC_CRF;
261 x4->params.rc.f_rf_constant = x4->crf;
262 } else if (x4->cqp >= 0) {
263 x4->params.rc.i_rc_method = XAVS_RC_CQP;
264 x4->params.rc.i_qp_constant = x4->cqp;
265 }
266 }
267
268 if (x4->aud >= 0)
269 x4->params.b_aud = x4->aud;
270 if (x4->mbtree >= 0)
271 x4->params.rc.b_mb_tree = x4->mbtree;
272 if (x4->direct_pred >= 0)
273 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
274 if (x4->fast_pskip >= 0)
275 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
276 if (x4->motion_est >= 0)
277 x4->params.analyse.i_me_method = x4->motion_est;
278 if (x4->mixed_refs >= 0)
279 x4->params.analyse.b_mixed_references = x4->mixed_refs;
280 if (x4->b_bias != INT_MIN)
281 x4->params.i_bframe_bias = x4->b_bias;
282 if (x4->cplxblur >= 0)
283 x4->params.rc.f_complexity_blur = x4->cplxblur;
284
285 #if FF_API_MOTION_EST
286 FF_DISABLE_DEPRECATION_WARNINGS
287 if (x4->motion_est < 0) {
288 switch (avctx->me_method) {
289 case ME_EPZS:
290 x4->params.analyse.i_me_method = XAVS_ME_DIA;
291 break;
292 case ME_HEX:
293 x4->params.analyse.i_me_method = XAVS_ME_HEX;
294 break;
295 case ME_UMH:
296 x4->params.analyse.i_me_method = XAVS_ME_UMH;
297 break;
298 case ME_FULL:
299 x4->params.analyse.i_me_method = XAVS_ME_ESA;
300 break;
301 case ME_TESA:
302 x4->params.analyse.i_me_method = XAVS_ME_TESA;
303 break;
304 default:
305 x4->params.analyse.i_me_method = XAVS_ME_HEX;
306 }
307 }
308 FF_ENABLE_DEPRECATION_WARNINGS
309 #endif
310
311 x4->params.i_bframe = avctx->max_b_frames;
312 /* cabac is not included in AVS JiZhun Profile */
313 x4->params.b_cabac = 0;
314
315 #if FF_API_PRIVATE_OPT
316 FF_DISABLE_DEPRECATION_WARNINGS
317 if (avctx->b_frame_strategy)
318 x4->b_frame_strategy = avctx->b_frame_strategy;
319 FF_ENABLE_DEPRECATION_WARNINGS
320 #endif
321
322 x4->params.i_bframe_adaptive = x4->b_frame_strategy;
323
324 avctx->has_b_frames = !!avctx->max_b_frames;
325
326 /* AVS doesn't allow B picture as reference */
327 /* The max allowed reference frame number of B is 2 */
328 x4->params.i_keyint_min = avctx->keyint_min;
329 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
330 x4->params.i_keyint_min = x4->params.i_keyint_max;
331
332 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
333
334 // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
335
336 x4->params.rc.i_qp_min = avctx->qmin;
337 x4->params.rc.i_qp_max = avctx->qmax;
338 x4->params.rc.i_qp_step = avctx->max_qdiff;
339
340 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
341 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
342
343 x4->params.i_frame_reference = avctx->refs;
344
345 x4->params.i_width = avctx->width;
346 x4->params.i_height = avctx->height;
347 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
348 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
349 /* This is only used for counting the fps */
350 x4->params.i_fps_num = avctx->time_base.den;
351 x4->params.i_fps_den = avctx->time_base.num;
352 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
353
354 x4->params.analyse.i_me_range = avctx->me_range;
355 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
356
357 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
358 /* AVS P2 only enables 8x8 transform */
359 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
360
361 x4->params.analyse.i_trellis = avctx->trellis;
362 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
363
364 if (avctx->level > 0)
365 x4->params.i_level_idc = avctx->level;
366
367 if (avctx->bit_rate > 0)
368 x4->params.rc.f_rate_tolerance =
369 (float)avctx->bit_rate_tolerance / avctx->bit_rate;
370
371 if ((avctx->rc_buffer_size) &&
372 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
373 x4->params.rc.f_vbv_buffer_init =
374 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
375 } else
376 x4->params.rc.f_vbv_buffer_init = 0.9;
377
378 /* TAG:do we have MB tree RC method */
379 /* what is the RC method we are now using? Default NO */
380 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
381 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
382 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
383
384 x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
385 x4->params.i_log_level = XAVS_LOG_DEBUG;
386 x4->params.i_threads = avctx->thread_count;
387 x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
388
389 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
390 x4->params.b_repeat_headers = 0;
391
392 x4->enc = xavs_encoder_open(&x4->params);
393 if (!x4->enc)
394 return -1;
395
396 if (!(x4->pts_buffer = av_mallocz((avctx->max_b_frames+1) * sizeof(*x4->pts_buffer))))
397 return AVERROR(ENOMEM);
398
399 /* TAG: Do we have GLOBAL HEADER in AVS */
400 /* We Have PPS and SPS in AVS */
401 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
402 xavs_nal_t *nal;
403 int nnal, s, i, size;
404 uint8_t *p;
405
406 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
407
408 avctx->extradata = p = av_malloc(s);
409 for (i = 0; i < nnal; i++) {
410 /* Don't put the SEI in extradata. */
411 if (nal[i].i_type == NAL_SEI) {
412 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
413 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
414 return -1;
415
416 continue;
417 }
418 size = xavs_nal_encode(p, &s, 1, nal + i);
419 if (size < 0)
420 return -1;
421 p += size;
422 }
423 avctx->extradata_size = p - avctx->extradata;
424 }
425 return 0;
426 }
427
428 #define OFFSET(x) offsetof(XavsContext, x)
429 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
430 static const AVOption options[] = {
431 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
432 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
433 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
434 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
435 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
436 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
437 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
438 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
439 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
440 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
441 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
442 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE },
443 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
444 { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, XAVS_ME_TESA, VE, "motion-est"},
445 { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
446 { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
447 { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
448 { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
449 { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
450 { "b_strategy", "Strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, VE},
451
452 { NULL },
453 };
454
455 static const AVClass class = {
456 .class_name = "libxavs",
457 .item_name = av_default_item_name,
458 .option = options,
459 .version = LIBAVUTIL_VERSION_INT,
460 };
461
462 static const AVCodecDefault xavs_defaults[] = {
463 { "b", "0" },
464 { NULL },
465 };
466
467 AVCodec ff_libxavs_encoder = {
468 .name = "libxavs",
469 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
470 .type = AVMEDIA_TYPE_VIDEO,
471 .id = AV_CODEC_ID_CAVS,
472 .priv_data_size = sizeof(XavsContext),
473 .init = XAVS_init,
474 .encode2 = XAVS_frame,
475 .close = XAVS_close,
476 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
477 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
478 .priv_class = &class,
479 .defaults = xavs_defaults,
480 };