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/opt.h" 32 33#define END_OF_STREAM 0x001 34 35#define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */ 36#define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */ 37#define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/ 38 39typedef struct XavsContext { 40 xavs_param_t params; 41 xavs_t *enc; 42 xavs_picture_t pic; 43 uint8_t *sei; 44 int sei_size; 45 AVFrame out_pic; 46 int end_of_stream; 47 float crf; 48 int cqp; 49 int b_bias; 50 float cplxblur; 51 int direct_pred; 52 int aud; 53 int fast_pskip; 54 int mbtree; 55 int mixed_refs; 56} XavsContext; 57 58static void XAVS_log(void *p, int level, const char *fmt, va_list args) 59{ 60 static const int level_map[] = { 61 [XAVS_LOG_ERROR] = AV_LOG_ERROR, 62 [XAVS_LOG_WARNING] = AV_LOG_WARNING, 63 [XAVS_LOG_INFO] = AV_LOG_INFO, 64 [XAVS_LOG_DEBUG] = AV_LOG_DEBUG 65 }; 66 67 if (level < 0 || level > XAVS_LOG_DEBUG) 68 return; 69 70 av_vlog(p, level_map[level], fmt, args); 71} 72 73static int encode_nals(AVCodecContext *ctx, uint8_t *buf, 74 int size, xavs_nal_t *nals, 75 int nnal, int skip_sei) 76{ 77 XavsContext *x4 = ctx->priv_data; 78 uint8_t *p = buf; 79 int i, s; 80 81 /* Write the SEI as part of the first frame. */ 82 if (x4->sei_size > 0 && nnal > 0) { 83 memcpy(p, x4->sei, x4->sei_size); 84 p += x4->sei_size; 85 x4->sei_size = 0; 86 } 87 88 for (i = 0; i < nnal; i++) { 89 /* Don't put the SEI in extradata. */ 90 if (skip_sei && nals[i].i_type == NAL_SEI) { 91 x4->sei = av_malloc( 5 + nals[i].i_payload * 4 / 3 ); 92 if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nals + i) < 0) 93 return -1; 94 95 continue; 96 } 97 s = xavs_nal_encode(p, &size, 1, nals + i); 98 if (s < 0) 99 return -1; 100 p += s; 101 } 102 103 return p - buf; 104} 105 106static int XAVS_frame(AVCodecContext *ctx, uint8_t *buf, 107 int bufsize, void *data) 108{ 109 XavsContext *x4 = ctx->priv_data; 110 AVFrame *frame = data; 111 xavs_nal_t *nal; 112 int nnal, i; 113 xavs_picture_t pic_out; 114 115 x4->pic.img.i_csp = XAVS_CSP_I420; 116 x4->pic.img.i_plane = 3; 117 118 if (frame) { 119 for (i = 0; i < 3; i++) { 120 x4->pic.img.plane[i] = frame->data[i]; 121 x4->pic.img.i_stride[i] = frame->linesize[i]; 122 } 123 124 x4->pic.i_pts = frame->pts; 125 x4->pic.i_type = XAVS_TYPE_AUTO; 126 } 127 128 if (xavs_encoder_encode(x4->enc, &nal, &nnal, 129 frame? &x4->pic: NULL, &pic_out) < 0) 130 return -1; 131 132 bufsize = encode_nals(ctx, buf, bufsize, nal, nnal, 0); 133 134 if (bufsize < 0) 135 return -1; 136 137 if (!bufsize && !frame && !(x4->end_of_stream)){ 138 buf[bufsize] = 0x0; 139 buf[bufsize+1] = 0x0; 140 buf[bufsize+2] = 0x01; 141 buf[bufsize+3] = 0xb1; 142 bufsize += 4; 143 x4->end_of_stream = END_OF_STREAM; 144 return bufsize; 145 } 146 /* FIXME: libxavs now provides DTS */ 147 /* but AVFrame doesn't have a field for it. */ 148 x4->out_pic.pts = pic_out.i_pts; 149 150 switch (pic_out.i_type) { 151 case XAVS_TYPE_IDR: 152 case XAVS_TYPE_I: 153 x4->out_pic.pict_type = AV_PICTURE_TYPE_I; 154 break; 155 case XAVS_TYPE_P: 156 x4->out_pic.pict_type = AV_PICTURE_TYPE_P; 157 break; 158 case XAVS_TYPE_B: 159 case XAVS_TYPE_BREF: 160 x4->out_pic.pict_type = AV_PICTURE_TYPE_B; 161 break; 162 } 163 164 /* There is no IDR frame in AVS JiZhun */ 165 /* Sequence header is used as a flag */ 166 x4->out_pic.key_frame = pic_out.i_type == XAVS_TYPE_I; 167 168 x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA; 169 170 return bufsize; 171} 172 173static av_cold int XAVS_close(AVCodecContext *avctx) 174{ 175 XavsContext *x4 = avctx->priv_data; 176 177 av_freep(&avctx->extradata); 178 av_free(x4->sei); 179 180 if (x4->enc) 181 xavs_encoder_close(x4->enc); 182 183 return 0; 184} 185 186static av_cold int XAVS_init(AVCodecContext *avctx) 187{ 188 XavsContext *x4 = avctx->priv_data; 189 190 x4->sei_size = 0; 191 xavs_param_default(&x4->params); 192 193 x4->params.pf_log = XAVS_log; 194 x4->params.p_log_private = avctx; 195 x4->params.i_keyint_max = avctx->gop_size; 196 if (avctx->bit_rate) { 197 x4->params.rc.i_bitrate = avctx->bit_rate / 1000; 198 x4->params.rc.i_rc_method = XAVS_RC_ABR; 199 } 200 x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000; 201 x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000; 202 x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1; 203 if (avctx->flags & CODEC_FLAG_PASS2) { 204 x4->params.rc.b_stat_read = 1; 205 } else { 206#if FF_API_X264_GLOBAL_OPTS 207 if (avctx->crf) { 208 x4->params.rc.i_rc_method = XAVS_RC_CRF; 209 x4->params.rc.f_rf_constant = avctx->crf; 210 } else if (avctx->cqp > -1) { 211 x4->params.rc.i_rc_method = XAVS_RC_CQP; 212 x4->params.rc.i_qp_constant = avctx->cqp; 213 } 214#endif 215 216 if (x4->crf >= 0) { 217 x4->params.rc.i_rc_method = XAVS_RC_CRF; 218 x4->params.rc.f_rf_constant = x4->crf; 219 } else if (x4->cqp >= 0) { 220 x4->params.rc.i_rc_method = XAVS_RC_CQP; 221 x4->params.rc.i_qp_constant = x4->cqp; 222 } 223 } 224 225#if FF_API_X264_GLOBAL_OPTS 226 if (avctx->bframebias) 227 x4->params.i_bframe_bias = avctx->bframebias; 228 if (avctx->deblockalpha) 229 x4->params.i_deblocking_filter_alphac0 = avctx->deblockalpha; 230 if (avctx->deblockbeta) 231 x4->params.i_deblocking_filter_beta = avctx->deblockbeta; 232 if (avctx->complexityblur >= 0) 233 x4->params.rc.f_complexity_blur = avctx->complexityblur; 234 if (avctx->directpred >= 0) 235 x4->params.analyse.i_direct_mv_pred = avctx->directpred; 236 if (avctx->partitions) { 237 if (avctx->partitions & XAVS_PART_I8X8) 238 x4->params.analyse.inter |= XAVS_ANALYSE_I8x8; 239 if (avctx->partitions & XAVS_PART_P8X8) 240 x4->params.analyse.inter |= XAVS_ANALYSE_PSUB16x16; 241 if (avctx->partitions & XAVS_PART_B8X8) 242 x4->params.analyse.inter |= XAVS_ANALYSE_BSUB16x16; 243 } 244 x4->params.rc.b_mb_tree = !!(avctx->flags2 & CODEC_FLAG2_MBTREE); 245 x4->params.b_aud = avctx->flags2 & CODEC_FLAG2_AUD; 246 x4->params.analyse.b_mixed_references = avctx->flags2 & CODEC_FLAG2_MIXED_REFS; 247 x4->params.analyse.b_fast_pskip = avctx->flags2 & CODEC_FLAG2_FASTPSKIP; 248 x4->params.analyse.b_weighted_bipred = avctx->flags2 & CODEC_FLAG2_WPRED; 249#endif 250 251 if (x4->aud >= 0) 252 x4->params.b_aud = x4->aud; 253 if (x4->mbtree >= 0) 254 x4->params.rc.b_mb_tree = x4->mbtree; 255 if (x4->direct_pred >= 0) 256 x4->params.analyse.i_direct_mv_pred = x4->direct_pred; 257 if (x4->fast_pskip >= 0) 258 x4->params.analyse.b_fast_pskip = x4->fast_pskip; 259 if (x4->mixed_refs >= 0) 260 x4->params.analyse.b_mixed_references = x4->mixed_refs; 261 if (x4->b_bias != INT_MIN) 262 x4->params.i_bframe_bias = x4->b_bias; 263 if (x4->cplxblur >= 0) 264 x4->params.rc.f_complexity_blur = x4->cplxblur; 265 266 x4->params.i_bframe = avctx->max_b_frames; 267 /* cabac is not included in AVS JiZhun Profile */ 268 x4->params.b_cabac = 0; 269 270 x4->params.i_bframe_adaptive = avctx->b_frame_strategy; 271 272 avctx->has_b_frames = !!avctx->max_b_frames; 273 274 /* AVS doesn't allow B picture as reference */ 275 /* The max allowed reference frame number of B is 2 */ 276 x4->params.i_keyint_min = avctx->keyint_min; 277 if (x4->params.i_keyint_min > x4->params.i_keyint_max) 278 x4->params.i_keyint_min = x4->params.i_keyint_max; 279 280 x4->params.i_scenecut_threshold = avctx->scenechange_threshold; 281 282 // x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER; 283 284 x4->params.rc.i_qp_min = avctx->qmin; 285 x4->params.rc.i_qp_max = avctx->qmax; 286 x4->params.rc.i_qp_step = avctx->max_qdiff; 287 288 x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */ 289 x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */ 290 291 x4->params.i_frame_reference = avctx->refs; 292 293 x4->params.i_width = avctx->width; 294 x4->params.i_height = avctx->height; 295 x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num; 296 x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den; 297 /* This is only used for counting the fps */ 298 x4->params.i_fps_num = avctx->time_base.den; 299 x4->params.i_fps_den = avctx->time_base.num; 300 x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16; 301 302 switch (avctx->me_method) { 303 case ME_EPZS: 304 x4->params.analyse.i_me_method = XAVS_ME_DIA; 305 break; 306 case ME_HEX: 307 x4->params.analyse.i_me_method = XAVS_ME_HEX; 308 break; 309 case ME_UMH: 310 x4->params.analyse.i_me_method = XAVS_ME_UMH; 311 break; 312 case ME_FULL: 313 x4->params.analyse.i_me_method = XAVS_ME_ESA; 314 break; 315 case ME_TESA: 316 x4->params.analyse.i_me_method = XAVS_ME_TESA; 317 break; 318 default: 319 x4->params.analyse.i_me_method = XAVS_ME_HEX; 320 } 321 322 x4->params.analyse.i_me_range = avctx->me_range; 323 x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality; 324 325 x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA; 326 /* AVS P2 only enables 8x8 transform */ 327 x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & CODEC_FLAG2_8X8DCT; 328 329 x4->params.analyse.i_trellis = avctx->trellis; 330 x4->params.analyse.i_noise_reduction = avctx->noise_reduction; 331 332 if (avctx->level > 0) 333 x4->params.i_level_idc = avctx->level; 334 335 x4->params.rc.f_rate_tolerance = 336 (float)avctx->bit_rate_tolerance/avctx->bit_rate; 337 338 if ((avctx->rc_buffer_size) && 339 (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) { 340 x4->params.rc.f_vbv_buffer_init = 341 (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size; 342 } else 343 x4->params.rc.f_vbv_buffer_init = 0.9; 344 345 /* TAG:do we have MB tree RC method */ 346 /* what is the RC method we are now using? Default NO */ 347 x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor); 348 x4->params.rc.f_pb_factor = avctx->b_quant_factor; 349 x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset; 350 351 x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR; 352 x4->params.i_log_level = XAVS_LOG_DEBUG; 353 x4->params.i_threads = avctx->thread_count; 354 x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT; 355 356 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) 357 x4->params.b_repeat_headers = 0; 358 359 x4->enc = xavs_encoder_open(&x4->params); 360 if (!x4->enc) 361 return -1; 362 363 avctx->coded_frame = &x4->out_pic; 364 /* TAG: Do we have GLOBAL HEADER in AVS */ 365 /* We Have PPS and SPS in AVS */ 366 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) { 367 xavs_nal_t *nal; 368 int nnal, s; 369 370 s = xavs_encoder_headers(x4->enc, &nal, &nnal); 371 372 avctx->extradata = av_malloc(s); 373 avctx->extradata_size = encode_nals(avctx, avctx->extradata, s, nal, nnal, 1); 374 } 375 return 0; 376} 377 378#define OFFSET(x) offsetof(XavsContext, x) 379#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM 380static const AVOption options[] = { 381 { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE }, 382 { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE }, 383 { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE }, 384 { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE}, 385 { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "direct-pred" }, 386 { "none", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" }, 387 { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" }, 388 { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" }, 389 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" }, 390 { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE}, 391 { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE}, 392 { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {-1}, -1, 1, VE }, 393 { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE}, 394 { NULL }, 395}; 396 397static const AVClass class = { 398 .class_name = "libxavs", 399 .item_name = av_default_item_name, 400 .option = options, 401 .version = LIBAVUTIL_VERSION_INT, 402}; 403 404static const AVCodecDefault xavs_defaults[] = { 405 { "b", "0" }, 406 { NULL }, 407}; 408 409AVCodec ff_libxavs_encoder = { 410 .name = "libxavs", 411 .type = AVMEDIA_TYPE_VIDEO, 412 .id = CODEC_ID_CAVS, 413 .priv_data_size = sizeof(XavsContext), 414 .init = XAVS_init, 415 .encode = XAVS_frame, 416 .close = XAVS_close, 417 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS, 418 .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_NONE }, 419 .long_name = NULL_IF_CONFIG_SMALL("libxavs - the Chinese Audio Video Standard Encoder"), 420 .priv_class = &class, 421 .defaults = xavs_defaults, 422}; 423 424