2 * AVS encoding using the xavs library
3 * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
5 * This file is part of Libav.
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.
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.
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
31 #include "libavutil/internal.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/opt.h"
35 #define END_OF_STREAM 0x001
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*/
41 typedef struct XavsContext {
64 static void XAVS_log(void *p, int level, const char *fmt, va_list args)
66 static const int level_map[] = {
67 [XAVS_LOG_ERROR] = AV_LOG_ERROR,
68 [XAVS_LOG_WARNING] = AV_LOG_WARNING,
69 [XAVS_LOG_INFO] = AV_LOG_INFO,
70 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
73 if (level < 0 || level > XAVS_LOG_DEBUG)
76 av_vlog(p, level_map[level], fmt, args);
79 static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
80 xavs_nal_t *nals, int nnal)
82 XavsContext *x4 = ctx->priv_data;
84 int i, s, ret, size = x4->sei_size + FF_MIN_BUFFER_SIZE;
89 for (i = 0; i < nnal; i++)
90 size += nals[i].i_payload;
92 if ((ret = ff_alloc_packet(pkt, size)) < 0) {
93 av_log(ctx, AV_LOG_ERROR, "Error getting output packet of size %d.\n", size);
98 /* Write the SEI as part of the first frame. */
99 if (x4->sei_size > 0 && nnal > 0) {
100 memcpy(p, x4->sei, x4->sei_size);
105 for (i = 0; i < nnal; i++) {
106 s = xavs_nal_encode(p, &size, 1, nals + i);
111 pkt->size = p - pkt->data;
116 static int XAVS_frame(AVCodecContext *ctx, AVPacket *pkt,
117 const AVFrame *frame, int *got_packet)
119 XavsContext *x4 = ctx->priv_data;
122 xavs_picture_t pic_out;
124 x4->pic.img.i_csp = XAVS_CSP_I420;
125 x4->pic.img.i_plane = 3;
128 for (i = 0; i < 3; i++) {
129 x4->pic.img.plane[i] = frame->data[i];
130 x4->pic.img.i_stride[i] = frame->linesize[i];
133 x4->pic.i_pts = frame->pts;
134 x4->pic.i_type = XAVS_TYPE_AUTO;
135 x4->pts_buffer[ctx->frame_number % (ctx->max_b_frames+1)] = frame->pts;
138 if (xavs_encoder_encode(x4->enc, &nal, &nnal,
139 frame? &x4->pic: NULL, &pic_out) < 0)
142 ret = encode_nals(ctx, pkt, nal, nnal);
148 if (!frame && !(x4->end_of_stream)) {
149 if ((ret = ff_alloc_packet(pkt, 4)) < 0)
156 pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(ctx->max_b_frames+1)] -
157 x4->pts_buffer[(x4->out_frame_count-2)%(ctx->max_b_frames+1)];
158 x4->end_of_stream = END_OF_STREAM;
164 x4->out_pic.pts = pic_out.i_pts;
165 pkt->pts = pic_out.i_pts;
166 if (ctx->has_b_frames) {
167 if (!x4->out_frame_count)
168 pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
170 pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(ctx->max_b_frames+1)];
174 switch (pic_out.i_type) {
177 x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
180 x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
184 x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
188 /* There is no IDR frame in AVS JiZhun */
189 /* Sequence header is used as a flag */
190 if (pic_out.i_type == XAVS_TYPE_I) {
191 x4->out_pic.key_frame = 1;
192 pkt->flags |= AV_PKT_FLAG_KEY;
195 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
197 x4->out_frame_count++;
202 static av_cold int XAVS_close(AVCodecContext *avctx)
204 XavsContext *x4 = avctx->priv_data;
206 av_freep(&avctx->extradata);
208 av_freep(&x4->pts_buffer);
211 xavs_encoder_close(x4->enc);
216 static av_cold int XAVS_init(AVCodecContext *avctx)
218 XavsContext *x4 = avctx->priv_data;
221 xavs_param_default(&x4->params);
223 x4->params.pf_log = XAVS_log;
224 x4->params.p_log_private = avctx;
225 x4->params.i_keyint_max = avctx->gop_size;
226 if (avctx->bit_rate) {
227 x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
228 x4->params.rc.i_rc_method = XAVS_RC_ABR;
230 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
231 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
232 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
233 if (avctx->flags & CODEC_FLAG_PASS2) {
234 x4->params.rc.b_stat_read = 1;
237 x4->params.rc.i_rc_method = XAVS_RC_CRF;
238 x4->params.rc.f_rf_constant = x4->crf;
239 } else if (x4->cqp >= 0) {
240 x4->params.rc.i_rc_method = XAVS_RC_CQP;
241 x4->params.rc.i_qp_constant = x4->cqp;
246 x4->params.b_aud = x4->aud;
248 x4->params.rc.b_mb_tree = x4->mbtree;
249 if (x4->direct_pred >= 0)
250 x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
251 if (x4->fast_pskip >= 0)
252 x4->params.analyse.b_fast_pskip = x4->fast_pskip;
253 if (x4->mixed_refs >= 0)
254 x4->params.analyse.b_mixed_references = x4->mixed_refs;
255 if (x4->b_bias != INT_MIN)
256 x4->params.i_bframe_bias = x4->b_bias;
257 if (x4->cplxblur >= 0)
258 x4->params.rc.f_complexity_blur = x4->cplxblur;
260 x4->params.i_bframe = avctx->max_b_frames;
261 /* cabac is not included in AVS JiZhun Profile */
262 x4->params.b_cabac = 0;
264 x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
266 avctx->has_b_frames = !!avctx->max_b_frames;
268 /* AVS doesn't allow B picture as reference */
269 /* The max allowed reference frame number of B is 2 */
270 x4->params.i_keyint_min = avctx->keyint_min;
271 if (x4->params.i_keyint_min > x4->params.i_keyint_max)
272 x4->params.i_keyint_min = x4->params.i_keyint_max;
274 x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
276 // x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
278 x4->params.rc.i_qp_min = avctx->qmin;
279 x4->params.rc.i_qp_max = avctx->qmax;
280 x4->params.rc.i_qp_step = avctx->max_qdiff;
282 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
283 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
285 x4->params.i_frame_reference = avctx->refs;
287 x4->params.i_width = avctx->width;
288 x4->params.i_height = avctx->height;
289 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
290 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
291 /* This is only used for counting the fps */
292 x4->params.i_fps_num = avctx->time_base.den;
293 x4->params.i_fps_den = avctx->time_base.num;
294 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
296 switch (avctx->me_method) {
298 x4->params.analyse.i_me_method = XAVS_ME_DIA;
301 x4->params.analyse.i_me_method = XAVS_ME_HEX;
304 x4->params.analyse.i_me_method = XAVS_ME_UMH;
307 x4->params.analyse.i_me_method = XAVS_ME_ESA;
310 x4->params.analyse.i_me_method = XAVS_ME_TESA;
313 x4->params.analyse.i_me_method = XAVS_ME_HEX;
316 x4->params.analyse.i_me_range = avctx->me_range;
317 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
319 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
320 /* AVS P2 only enables 8x8 transform */
321 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & CODEC_FLAG2_8X8DCT;
323 x4->params.analyse.i_trellis = avctx->trellis;
324 x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
326 if (avctx->level > 0)
327 x4->params.i_level_idc = avctx->level;
329 x4->params.rc.f_rate_tolerance =
330 (float)avctx->bit_rate_tolerance/avctx->bit_rate;
332 if ((avctx->rc_buffer_size) &&
333 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
334 x4->params.rc.f_vbv_buffer_init =
335 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
337 x4->params.rc.f_vbv_buffer_init = 0.9;
339 /* TAG:do we have MB tree RC method */
340 /* what is the RC method we are now using? Default NO */
341 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
342 x4->params.rc.f_pb_factor = avctx->b_quant_factor;
343 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
345 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
346 x4->params.i_log_level = XAVS_LOG_DEBUG;
347 x4->params.i_threads = avctx->thread_count;
348 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
350 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
351 x4->params.b_repeat_headers = 0;
353 x4->enc = xavs_encoder_open(&x4->params);
357 if (!(x4->pts_buffer = av_mallocz((avctx->max_b_frames+1) * sizeof(*x4->pts_buffer))))
358 return AVERROR(ENOMEM);
360 avctx->coded_frame = &x4->out_pic;
361 /* TAG: Do we have GLOBAL HEADER in AVS */
362 /* We Have PPS and SPS in AVS */
363 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
365 int nnal, s, i, size;
368 s = xavs_encoder_headers(x4->enc, &nal, &nnal);
370 avctx->extradata = p = av_malloc(s);
371 for (i = 0; i < nnal; i++) {
372 /* Don't put the SEI in extradata. */
373 if (nal[i].i_type == NAL_SEI) {
374 x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
375 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
380 size = xavs_nal_encode(p, &s, 1, nal + i);
385 avctx->extradata_size = p - avctx->extradata;
390 #define OFFSET(x) offsetof(XavsContext, x)
391 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
392 static const AVOption options[] = {
393 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
394 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
395 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
396 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
397 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
398 { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
399 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
400 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
401 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
402 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
403 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
404 { "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 },
405 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
409 static const AVClass class = {
410 .class_name = "libxavs",
411 .item_name = av_default_item_name,
413 .version = LIBAVUTIL_VERSION_INT,
416 static const AVCodecDefault xavs_defaults[] = {
421 AVCodec ff_libxavs_encoder = {
423 .type = AVMEDIA_TYPE_VIDEO,
424 .id = AV_CODEC_ID_CAVS,
425 .priv_data_size = sizeof(XavsContext),
427 .encode2 = XAVS_frame,
429 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
430 .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
431 .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
432 .priv_class = &class,
433 .defaults = xavs_defaults,