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 "codec_internal.h"
31 #include "encode.h"
32 #include "packet_internal.h"
33 #include "libavutil/internal.h"
34 #include "libavutil/mem.h"
35 #include "libavutil/opt.h"
36
37 #define END_OF_STREAM 0x001
38
39 #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
40 #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
41 #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
42
43 typedef struct XavsContext {
44 AVClass *class;
45 xavs_param_t params;
46 xavs_t *enc;
47 xavs_picture_t pic;
48 uint8_t *sei;
49 int sei_size;
50 int end_of_stream;
51 float crf;
52 int cqp;
53 int b_bias;
54 float cplxblur;
55 int direct_pred;
56 int aud;
57 int fast_pskip;
58 int motion_est;
59 int mbtree;
60 int mixed_refs;
61 int b_frame_strategy;
62 int chroma_offset;
63 int scenechange_threshold;
64 int noise_reduction;
65
66 int64_t *pts_buffer;
67 int out_frame_count;
68 } XavsContext;
69
XAVS_log(void * p,int level,const char * fmt,va_list args)70 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
71 {
72 static const int level_map[] = {
73 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
74 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
75 [XAVS_LOG_INFO] = AV_LOG_INFO,
76 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
77 };
78
79 if (level < 0 || level > XAVS_LOG_DEBUG)
80 return;
81
82 av_vlog(p, level_map[level], fmt, args);
83 }
84
encode_nals(AVCodecContext * ctx,AVPacket * pkt,xavs_nal_t * nals,int nnal)85 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
86 xavs_nal_t *nals, int nnal)
87 {
88 XavsContext *x4 = ctx->priv_data;
89 int64_t size = x4->sei_size;
90 uint8_t *p, *p_end;
91 int i, s, ret;
92
93 if (!nnal)
94 return 0;
95
96 for (i = 0; i < nnal; i++)
97 size += 3U + nals[i].i_payload;
98
99 if ((ret = ff_get_encode_buffer(ctx, pkt, size, 0)) < 0)
100 return ret;
101 p = pkt->data;
102 p_end = pkt->data + size;
103
104 /* Write the SEI as part of the first frame. */
105 if (x4->sei_size > 0 && nnal > 0) {
106 memcpy(p, x4->sei, x4->sei_size);
107 p += x4->sei_size;
108 x4->sei_size = 0;
109 }
110
111 for (i = 0; i < nnal; i++) {
112 int size = p_end - p;
113 s = xavs_nal_encode(p, &size, 1, nals + i);
114 if (s < 0)
115 return AVERROR_EXTERNAL;
116 if (s != 3U + nals[i].i_payload)
117 return AVERROR_EXTERNAL;
118 p += s;
119 }
120
121 return 1;
122 }
123
XAVS_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * frame,int * got_packet)124 static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
125 const AVFrame *frame, int *got_packet)
126 {
127 XavsContext *x4 = avctx->priv_data;
128 xavs_nal_t *nal;
129 int nnal, i, ret;
130 xavs_picture_t pic_out;
131 int pict_type;
132
133 x4->pic.img.i_csp = XAVS_CSP_I420;
134 x4->pic.img.i_plane = 3;
135
136 if (frame) {
137 for (i = 0; i < 3; i++) {
138 x4->pic.img.plane[i] = frame->data[i];
139 x4->pic.img.i_stride[i] = frame->linesize[i];
140 }
141
142 x4->pic.i_pts = frame->pts;
143 x4->pic.i_type = XAVS_TYPE_AUTO;
144 x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
145 }
146
147 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
148 frame? &x4->pic: NULL, &pic_out) < 0)
149 return AVERROR_EXTERNAL;
150
151 ret = encode_nals(avctx, pkt, nal, nnal);
152
153 if (ret < 0)
154 return ret;
155
156 if (!ret) {
157 if (!frame && !(x4->end_of_stream)) {
158 if ((ret = ff_get_encode_buffer(avctx, pkt, 4, 0)) < 0)
159 return ret;
160
161 pkt->data[0] = 0x0;
162 pkt->data[1] = 0x0;
163 pkt->data[2] = 0x01;
164 pkt->data[3] = 0xb1;
165 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
166 x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
167 x4->end_of_stream = END_OF_STREAM;
168 *got_packet = 1;
169 }
170 return 0;
171 }
172
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
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 pkt->flags |= AV_PKT_FLAG_KEY;
202 }
203
204 ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
205
206 x4->out_frame_count++;
207 *got_packet = ret;
208 return 0;
209 }
210
XAVS_close(AVCodecContext * avctx)211 static av_cold int XAVS_close(AVCodecContext *avctx)
212 {
213 XavsContext *x4 = avctx->priv_data;
214
215 av_freep(&x4->sei);
216 av_freep(&x4->pts_buffer);
217
218 if (x4->enc)
219 xavs_encoder_close(x4->enc);
220
221 return 0;
222 }
223
XAVS_init(AVCodecContext * avctx)224 static av_cold int XAVS_init(AVCodecContext *avctx)
225 {
226 XavsContext *x4 = avctx->priv_data;
227
228 x4->sei_size = 0;
229 xavs_param_default(&x4->params);
230
231 x4->params.pf_log = XAVS_log;
232 x4->params.p_log_private = avctx;
233 x4->params.i_keyint_max = avctx->gop_size;
234 if (avctx->bit_rate) {
235 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
236 x4->params.rc.i_rc_method = XAVS_RC_ABR;
237 }
238 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
239 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
240 x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
241 if (avctx->flags & AV_CODEC_FLAG_PASS2) {
242 x4->params.rc.b_stat_read = 1;
243 } else {
244 if (x4->crf >= 0) {
245 x4->params.rc.i_rc_method = XAVS_RC_CRF;
246 x4->params.rc.f_rf_constant = x4->crf;
247 } else if (x4->cqp >= 0) {
248 x4->params.rc.i_rc_method = XAVS_RC_CQP;
249 x4->params.rc.i_qp_constant = x4->cqp;
250 }
251 }
252
253 if (x4->aud >= 0)
254 x4->params.b_aud = x4->aud;
255 if (x4->mbtree >= 0)
256 x4->params.rc.b_mb_tree = x4->mbtree;
257 if (x4->direct_pred >= 0)
258 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
259 if (x4->fast_pskip >= 0)
260 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
261 if (x4->motion_est >= 0)
262 x4->params.analyse.i_me_method = x4->motion_est;
263 if (x4->mixed_refs >= 0)
264 x4->params.analyse.b_mixed_references = x4->mixed_refs;
265 if (x4->b_bias != INT_MIN)
266 x4->params.i_bframe_bias = x4->b_bias;
267 if (x4->cplxblur >= 0)
268 x4->params.rc.f_complexity_blur = x4->cplxblur;
269
270 x4->params.i_bframe = avctx->max_b_frames;
271 /* cabac is not included in AVS JiZhun Profile */
272 x4->params.b_cabac = 0;
273
274 x4->params.i_bframe_adaptive = x4->b_frame_strategy;
275
276 avctx->has_b_frames = !!avctx->max_b_frames;
277
278 /* AVS doesn't allow B picture as reference */
279 /* The max allowed reference frame number of B is 2 */
280 x4->params.i_keyint_min = avctx->keyint_min;
281 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
282 x4->params.i_keyint_min = x4->params.i_keyint_max;
283
284 x4->params.i_scenecut_threshold = x4->scenechange_threshold;
285
286 // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
287
288 x4->params.rc.i_qp_min = avctx->qmin;
289 x4->params.rc.i_qp_max = avctx->qmax;
290 x4->params.rc.i_qp_step = avctx->max_qdiff;
291
292 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
293 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
294
295 x4->params.i_frame_reference = avctx->refs;
296
297 x4->params.i_width = avctx->width;
298 x4->params.i_height = avctx->height;
299 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
300 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
301 /* This is only used for counting the fps */
302 x4->params.i_fps_num = avctx->time_base.den;
303 x4->params.i_fps_den = avctx->time_base.num;
304 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
305
306 x4->params.analyse.i_me_range = avctx->me_range;
307 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
308
309 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
310 /* AVS P2 only enables 8x8 transform */
311 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
312
313 x4->params.analyse.i_trellis = avctx->trellis;
314
315 x4->params.analyse.i_noise_reduction = x4->noise_reduction;
316
317 if (avctx->level > 0)
318 x4->params.i_level_idc = avctx->level;
319
320 if (avctx->bit_rate > 0)
321 x4->params.rc.f_rate_tolerance =
322 (float)avctx->bit_rate_tolerance / avctx->bit_rate;
323
324 if ((avctx->rc_buffer_size) &&
325 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
326 x4->params.rc.f_vbv_buffer_init =
327 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
328 } else
329 x4->params.rc.f_vbv_buffer_init = 0.9;
330
331 /* TAG:do we have MB tree RC method */
332 /* what is the RC method we are now using? Default NO */
333 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
334 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
335
336 x4->params.analyse.i_chroma_qp_offset = x4->chroma_offset;
337
338 x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
339 x4->params.i_log_level = XAVS_LOG_DEBUG;
340 x4->params.i_threads = avctx->thread_count;
341 x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
342
343 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
344 x4->params.b_repeat_headers = 0;
345
346 x4->enc = xavs_encoder_open(&x4->params);
347 if (!x4->enc)
348 return AVERROR_EXTERNAL;
349
350 if (!FF_ALLOCZ_TYPED_ARRAY(x4->pts_buffer, avctx->max_b_frames + 1))
351 return AVERROR(ENOMEM);
352
353 /* TAG: Do we have GLOBAL HEADER in AVS */
354 /* We Have PPS and SPS in AVS */
355 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
356 xavs_nal_t *nal;
357 int nnal, s, i, size;
358 uint8_t *p;
359
360 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
361
362 avctx->extradata = p = av_malloc(s);
363 for (i = 0; i < nnal; i++) {
364 /* Don't put the SEI in extradata. */
365 if (nal[i].i_type == NAL_SEI) {
366 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
367 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
368 return -1;
369
370 continue;
371 }
372 size = xavs_nal_encode(p, &s, 1, nal + i);
373 if (size < 0)
374 return -1;
375 p += size;
376 }
377 avctx->extradata_size = p - avctx->extradata;
378 }
379 return 0;
380 }
381
382 #define OFFSET(x) offsetof(XavsContext, x)
383 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
384 static const AVOption options[] = {
385 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
386 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
387 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
388 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
389 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
390 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
391 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
392 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
393 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
394 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
395 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
396 { "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 },
397 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_BOOL, {.i64 = -1 }, -1, 1, VE},
398 { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = XAVS_ME_DIA }, -1, XAVS_ME_TESA, VE, "motion-est"},
399 { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
400 { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
401 { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
402 { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
403 { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
404 { "b_strategy", "Strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 2, VE},
405 { "chromaoffset", "QP difference between chroma and luma", OFFSET(chroma_offset), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE},
406 { "sc_threshold", "Scene change threshold", OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
407 { "noise_reduction", "Noise reduction", OFFSET(noise_reduction), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, INT_MAX, VE},
408
409 { NULL },
410 };
411
412 static const AVClass xavs_class = {
413 .class_name = "libxavs",
414 .item_name = av_default_item_name,
415 .option = options,
416 .version = LIBAVUTIL_VERSION_INT,
417 };
418
419 static const FFCodecDefault xavs_defaults[] = {
420 { "b", "0" },
421 { NULL },
422 };
423
424 const FFCodec ff_libxavs_encoder = {
425 .p.name = "libxavs",
426 .p.long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
427 .p.type = AVMEDIA_TYPE_VIDEO,
428 .p.id = AV_CODEC_ID_CAVS,
429 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
430 AV_CODEC_CAP_OTHER_THREADS,
431 .priv_data_size = sizeof(XavsContext),
432 .init = XAVS_init,
433 FF_CODEC_ENCODE_CB(XAVS_frame),
434 .close = XAVS_close,
435 .caps_internal = FF_CODEC_CAP_AUTO_THREADS,
436 .p.pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
437 .p.priv_class = &xavs_class,
438 .defaults = xavs_defaults,
439 .p.wrapper_name = "libxavs",
440 };
441