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 FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
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 "packet_internal.h"
32 #include "libavutil/internal.h"
33 #include "libavutil/mem.h"
34 #include "libavutil/opt.h"
35
36 #define END_OF_STREAM 0x001
37
38 #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
39 #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
40 #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
41
42 typedef struct XavsContext {
43 AVClass *class;
44 xavs_param_t params;
45 xavs_t *enc;
46 xavs_picture_t pic;
47 uint8_t *sei;
48 int sei_size;
49 int end_of_stream;
50 float crf;
51 int cqp;
52 int b_bias;
53 float cplxblur;
54 int direct_pred;
55 int aud;
56 int fast_pskip;
57 int motion_est;
58 int mbtree;
59 int mixed_refs;
60 int b_frame_strategy;
61 int chroma_offset;
62 int scenechange_threshold;
63 int noise_reduction;
64
65 int64_t *pts_buffer;
66 int out_frame_count;
67 } XavsContext;
68
XAVS_log(void * p,int level,const char * fmt,va_list args)69 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
70 {
71 static const int level_map[] = {
72 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
73 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
74 [XAVS_LOG_INFO] = AV_LOG_INFO,
75 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
76 };
77
78 if (level < 0 || level > XAVS_LOG_DEBUG)
79 return;
80
81 av_vlog(p, level_map[level], fmt, args);
82 }
83
encode_nals(AVCodecContext * ctx,AVPacket * pkt,xavs_nal_t * nals,int nnal)84 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
85 xavs_nal_t *nals, int nnal)
86 {
87 XavsContext *x4 = ctx->priv_data;
88 uint8_t *p;
89 int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
90
91 if (!nnal)
92 return 0;
93
94 for (i = 0; i < nnal; i++)
95 size += nals[i].i_payload;
96
97 if ((ret = ff_alloc_packet2(ctx, pkt, size, 0)) < 0)
98 return ret;
99 p = pkt->data;
100
101 /* Write the SEI as part of the first frame. */
102 if (x4->sei_size > 0 && nnal > 0) {
103 memcpy(p, x4->sei, x4->sei_size);
104 p += x4->sei_size;
105 x4->sei_size = 0;
106 }
107
108 for (i = 0; i < nnal; i++) {
109 s = xavs_nal_encode(p, &size, 1, nals + i);
110 if (s < 0)
111 return -1;
112 p += s;
113 }
114 pkt->size = p - pkt->data;
115
116 return 1;
117 }
118
XAVS_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * frame,int * got_packet)119 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
120 const AVFrame *frame, int *got_packet)
121 {
122 XavsContext *x4 = avctx->priv_data;
123 xavs_nal_t *nal;
124 int nnal, i, ret;
125 xavs_picture_t pic_out;
126 int pict_type;
127
128 x4->pic.img.i_csp = XAVS_CSP_I420;
129 x4->pic.img.i_plane = 3;
130
131 if (frame) {
132 for (i = 0; i < 3; i++) {
133 x4->pic.img.plane[i] = frame->data[i];
134 x4->pic.img.i_stride[i] = frame->linesize[i];
135 }
136
137 x4->pic.i_pts = frame->pts;
138 x4->pic.i_type = XAVS_TYPE_AUTO;
139 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
140 }
141
142 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
143 frame? &x4->pic: NULL, &pic_out) < 0)
144 return -1;
145
146 ret = encode_nals(avctx, pkt, nal, nnal);
147
148 if (ret < 0)
149 return -1;
150
151 if (!ret) {
152 if (!frame && !(x4->end_of_stream)) {
153 if ((ret = ff_alloc_packet2(avctx, pkt, 4, 0)) < 0)
154 return ret;
155
156 pkt->data[0] = 0x0;
157 pkt->data[1] = 0x0;
158 pkt->data[2] = 0x01;
159 pkt->data[3] = 0xb1;
160 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
161 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
162 x4->end_of_stream = END_OF_STREAM;
163 *got_packet = 1;
164 }
165 return 0;
166 }
167
168 #if FF_API_CODED_FRAME
169 FF_DISABLE_DEPRECATION_WARNINGS
170 avctx->coded_frame->pts = pic_out.i_pts;
171 FF_ENABLE_DEPRECATION_WARNINGS
172 #endif
173 pkt->pts = pic_out.i_pts;
174 if (avctx->has_b_frames) {
175 if (!x4->out_frame_count)
176 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
177 else
178 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
179 } else
180 pkt->dts = pkt->pts;
181
182 switch (pic_out.i_type) {
183 case XAVS_TYPE_IDR:
184 case XAVS_TYPE_I:
185 pict_type = AV_PICTURE_TYPE_I;
186 break;
187 case XAVS_TYPE_P:
188 pict_type = AV_PICTURE_TYPE_P;
189 break;
190 case XAVS_TYPE_B:
191 case XAVS_TYPE_BREF:
192 pict_type = AV_PICTURE_TYPE_B;
193 break;
194 default:
195 pict_type = AV_PICTURE_TYPE_NONE;
196 }
197 #if FF_API_CODED_FRAME
198 FF_DISABLE_DEPRECATION_WARNINGS
199 avctx->coded_frame->pict_type = pict_type;
200 FF_ENABLE_DEPRECATION_WARNINGS
201 #endif
202
203 /* There is no IDR frame in AVS JiZhun */
204 /* Sequence header is used as a flag */
205 if (pic_out.i_type == XAVS_TYPE_I) {
206 #if FF_API_CODED_FRAME
207 FF_DISABLE_DEPRECATION_WARNINGS
208 avctx->coded_frame->key_frame = 1;
209 FF_ENABLE_DEPRECATION_WARNINGS
210 #endif
211 pkt->flags |= AV_PKT_FLAG_KEY;
212 }
213
214 #if FF_API_CODED_FRAME
215 FF_DISABLE_DEPRECATION_WARNINGS
216 avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
217 FF_ENABLE_DEPRECATION_WARNINGS
218 #endif
219
220 ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
221
222 x4->out_frame_count++;
223 *got_packet = ret;
224 return 0;
225 }
226
XAVS_close(AVCodecContext * avctx)227 static av_cold int XAVS_close(AVCodecContext *avctx)
228 {
229 XavsContext *x4 = avctx->priv_data;
230
231 av_freep(&avctx->extradata);
232 av_freep(&x4->sei);
233 av_freep(&x4->pts_buffer);
234
235 if (x4->enc)
236 xavs_encoder_close(x4->enc);
237
238 return 0;
239 }
240
XAVS_init(AVCodecContext * avctx)241 static av_cold int XAVS_init(AVCodecContext *avctx)
242 {
243 XavsContext *x4 = avctx->priv_data;
244
245 x4->sei_size = 0;
246 xavs_param_default(&x4->params);
247
248 x4->params.pf_log = XAVS_log;
249 x4->params.p_log_private = avctx;
250 x4->params.i_keyint_max = avctx->gop_size;
251 if (avctx->bit_rate) {
252 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
253 x4->params.rc.i_rc_method = XAVS_RC_ABR;
254 }
255 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
256 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
257 x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
258 if (avctx->flags & AV_CODEC_FLAG_PASS2) {
259 x4->params.rc.b_stat_read = 1;
260 } else {
261 if (x4->crf >= 0) {
262 x4->params.rc.i_rc_method = XAVS_RC_CRF;
263 x4->params.rc.f_rf_constant = x4->crf;
264 } else if (x4->cqp >= 0) {
265 x4->params.rc.i_rc_method = XAVS_RC_CQP;
266 x4->params.rc.i_qp_constant = x4->cqp;
267 }
268 }
269
270 if (x4->aud >= 0)
271 x4->params.b_aud = x4->aud;
272 if (x4->mbtree >= 0)
273 x4->params.rc.b_mb_tree = x4->mbtree;
274 if (x4->direct_pred >= 0)
275 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
276 if (x4->fast_pskip >= 0)
277 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
278 if (x4->motion_est >= 0)
279 x4->params.analyse.i_me_method = x4->motion_est;
280 if (x4->mixed_refs >= 0)
281 x4->params.analyse.b_mixed_references = x4->mixed_refs;
282 if (x4->b_bias != INT_MIN)
283 x4->params.i_bframe_bias = x4->b_bias;
284 if (x4->cplxblur >= 0)
285 x4->params.rc.f_complexity_blur = x4->cplxblur;
286
287 x4->params.i_bframe = avctx->max_b_frames;
288 /* cabac is not included in AVS JiZhun Profile */
289 x4->params.b_cabac = 0;
290
291 #if FF_API_PRIVATE_OPT
292 FF_DISABLE_DEPRECATION_WARNINGS
293 if (avctx->b_frame_strategy)
294 x4->b_frame_strategy = avctx->b_frame_strategy;
295 FF_ENABLE_DEPRECATION_WARNINGS
296 #endif
297
298 x4->params.i_bframe_adaptive = x4->b_frame_strategy;
299
300 avctx->has_b_frames = !!avctx->max_b_frames;
301
302 /* AVS doesn't allow B picture as reference */
303 /* The max allowed reference frame number of B is 2 */
304 x4->params.i_keyint_min = avctx->keyint_min;
305 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
306 x4->params.i_keyint_min = x4->params.i_keyint_max;
307
308 #if FF_API_PRIVATE_OPT
309 FF_DISABLE_DEPRECATION_WARNINGS
310 if (avctx->scenechange_threshold)
311 x4->scenechange_threshold = avctx->scenechange_threshold;
312 FF_ENABLE_DEPRECATION_WARNINGS
313 #endif
314
315 x4->params.i_scenecut_threshold = x4->scenechange_threshold;
316
317 // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
318
319 x4->params.rc.i_qp_min = avctx->qmin;
320 x4->params.rc.i_qp_max = avctx->qmax;
321 x4->params.rc.i_qp_step = avctx->max_qdiff;
322
323 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
324 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
325
326 x4->params.i_frame_reference = avctx->refs;
327
328 x4->params.i_width = avctx->width;
329 x4->params.i_height = avctx->height;
330 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
331 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
332 /* This is only used for counting the fps */
333 x4->params.i_fps_num = avctx->time_base.den;
334 x4->params.i_fps_den = avctx->time_base.num;
335 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
336
337 x4->params.analyse.i_me_range = avctx->me_range;
338 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
339
340 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
341 /* AVS P2 only enables 8x8 transform */
342 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
343
344 x4->params.analyse.i_trellis = avctx->trellis;
345
346 #if FF_API_PRIVATE_OPT
347 FF_DISABLE_DEPRECATION_WARNINGS
348 if (avctx->noise_reduction >= 0)
349 x4->noise_reduction = avctx->noise_reduction;
350 FF_ENABLE_DEPRECATION_WARNINGS
351 #endif
352
353 x4->params.analyse.i_noise_reduction = x4->noise_reduction;
354
355 if (avctx->level > 0)
356 x4->params.i_level_idc = avctx->level;
357
358 if (avctx->bit_rate > 0)
359 x4->params.rc.f_rate_tolerance =
360 (float)avctx->bit_rate_tolerance / avctx->bit_rate;
361
362 if ((avctx->rc_buffer_size) &&
363 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
364 x4->params.rc.f_vbv_buffer_init =
365 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
366 } else
367 x4->params.rc.f_vbv_buffer_init = 0.9;
368
369 /* TAG:do we have MB tree RC method */
370 /* what is the RC method we are now using? Default NO */
371 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
372 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
373
374 #if FF_API_PRIVATE_OPT
375 FF_DISABLE_DEPRECATION_WARNINGS
376 if (avctx->chromaoffset)
377 x4->chroma_offset = avctx->chromaoffset;
378 FF_ENABLE_DEPRECATION_WARNINGS
379 #endif
380
381 x4->params.analyse.i_chroma_qp_offset = x4->chroma_offset;
382
383 x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
384 x4->params.i_log_level = XAVS_LOG_DEBUG;
385 x4->params.i_threads = avctx->thread_count;
386 x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
387
388 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
389 x4->params.b_repeat_headers = 0;
390
391 x4->enc = xavs_encoder_open(&x4->params);
392 if (!x4->enc)
393 return -1;
394
395 if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
396 return AVERROR(ENOMEM);
397
398 /* TAG: Do we have GLOBAL HEADER in AVS */
399 /* We Have PPS and SPS in AVS */
400 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
401 xavs_nal_t *nal;
402 int nnal, s, i, size;
403 uint8_t *p;
404
405 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
406
407 avctx->extradata = p = av_malloc(s);
408 for (i = 0; i < nnal; i++) {
409 /* Don't put the SEI in extradata. */
410 if (nal[i].i_type == NAL_SEI) {
411 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
412 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
413 return -1;
414
415 continue;
416 }
417 size = xavs_nal_encode(p, &s, 1, nal + i);
418 if (size < 0)
419 return -1;
420 p += size;
421 }
422 avctx->extradata_size = p - avctx->extradata;
423 }
424 return 0;
425 }
426
427 #define OFFSET(x) offsetof(XavsContext, x)
428 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
429 static const AVOption options[] = {
430 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
431 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
432 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
433 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
434 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
435 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
436 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
437 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
438 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
439 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
440 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
441 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_BOOL, {.i64 = -1}, -1, 1, VE },
442 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
443 { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = XAVS_ME_DIA }, -1, XAVS_ME_TESA, VE, "motion-est"},
444 { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
445 { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
446 { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
447 { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
448 { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
449 { "b_strategy", "Strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, VE},
450 { "chromaoffset", "QP difference between chroma and luma", OFFSET(chroma_offset), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE},
451 { "sc_threshold", "Scene change threshold", OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
452 { "noise_reduction", "Noise reduction", OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
453
454 { NULL },
455 };
456
457 static const AVClass xavs_class = {
458 .class_name = "libxavs",
459 .item_name = av_default_item_name,
460 .option = options,
461 .version = LIBAVUTIL_VERSION_INT,
462 };
463
464 static const AVCodecDefault xavs_defaults[] = {
465 { "b", "0" },
466 { NULL },
467 };
468
469 AVCodec ff_libxavs_encoder = {
470 .name = "libxavs",
471 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
472 .type = AVMEDIA_TYPE_VIDEO,
473 .id = AV_CODEC_ID_CAVS,
474 .priv_data_size = sizeof(XavsContext),
475 .init = XAVS_init,
476 .encode2 = XAVS_frame,
477 .close = XAVS_close,
478 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_OTHER_THREADS,
479 .caps_internal = FF_CODEC_CAP_AUTO_THREADS,
480 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
481 .priv_class = &xavs_class,
482 .defaults = xavs_defaults,
483 .wrapper_name = "libxavs",
484 };
485