libavcodec/libx265.c

   1 /*
   2  * libx265 encoder
   3  *
   4  * Copyright (c) 2013-2014 Derek Buitenhuis
   5  *
   6  * This file is part of Libav.
   7  *
   8  * Libav is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * Libav is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with Libav; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 #if defined(_MSC_VER)
  24 #define X265_API_IMPORTS 1
  25 #endif
  26
  27 #include <x265.h>
  28 #include <float.h>
  29
  30 #include "libavutil/internal.h"
  31 #include "libavutil/common.h"
  32 #include "libavutil/opt.h"
  33 #include "libavutil/pixdesc.h"
  34 #include "avcodec.h"
  35 #include "internal.h"
  36
  37 typedef struct libx265Context {
  38     const AVClass *class;
  39
  40     x265_encoder *encoder;
  41     x265_param   *params;
  42     const x265_api *api;
  43
  44     float crf;
  45     int   forced_idr;
  46     char *preset;
  47     char *tune;
  48     char *x265_opts;
  49 } libx265Context;
  50
  51 static int is_keyframe(NalUnitType naltype)
  52 {
  53     switch (naltype) {
  54     case NAL_UNIT_CODED_SLICE_BLA_W_LP:
  55     case NAL_UNIT_CODED_SLICE_BLA_W_RADL:
  56     case NAL_UNIT_CODED_SLICE_BLA_N_LP:
  57     case NAL_UNIT_CODED_SLICE_IDR_W_RADL:
  58     case NAL_UNIT_CODED_SLICE_IDR_N_LP:
  59     case NAL_UNIT_CODED_SLICE_CRA:
  60         return 1;
  61     default:
  62         return 0;
  63     }
  64 }
  65
  66 static av_cold int libx265_encode_close(AVCodecContext *avctx)
  67 {
  68     libx265Context *ctx = avctx->priv_data;
  69
  70     ctx->api->param_free(ctx->params);
  71
  72     if (ctx->encoder)
  73         ctx->api->encoder_close(ctx->encoder);
  74
  75     return 0;
  76 }
  77
  78 static av_cold int libx265_encode_init(AVCodecContext *avctx)
  79 {
  80     libx265Context *ctx = avctx->priv_data;
  81
  82     ctx->api = x265_api_get(av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth);
  83     if (!ctx->api)
  84         ctx->api = x265_api_get(0);
  85
  86     if (avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL &&
  87         !av_pix_fmt_desc_get(avctx->pix_fmt)->log2_chroma_w) {
  88         av_log(avctx, AV_LOG_ERROR,
  89                "4:2:2 and 4:4:4 support is not fully defined for HEVC yet. "
  90                "Set -strict experimental to encode anyway.\n");
  91         return AVERROR(ENOSYS);
  92     }
  93
  94     ctx->params = ctx->api->param_alloc();
  95     if (!ctx->params) {
  96         av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
  97         return AVERROR(ENOMEM);
  98     }
  99
 100     if (ctx->api->param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
 101         int i;
 102
 103         av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
 104         av_log(avctx, AV_LOG_INFO, "Possible presets:");
 105         for (i = 0; x265_preset_names[i]; i++)
 106             av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
 107
 108         av_log(avctx, AV_LOG_INFO, "\n");
 109         av_log(avctx, AV_LOG_INFO, "Possible tunes:");
 110         for (i = 0; x265_tune_names[i]; i++)
 111             av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
 112
 113         av_log(avctx, AV_LOG_INFO, "\n");
 114
 115         return AVERROR(EINVAL);
 116     }
 117
 118     ctx->params->frameNumThreads = avctx->thread_count;
 119     ctx->params->fpsNum          = avctx->time_base.den;
 120     ctx->params->fpsDenom        = avctx->time_base.num * avctx->ticks_per_frame;
 121     ctx->params->sourceWidth     = avctx->width;
 122     ctx->params->sourceHeight    = avctx->height;
 123     ctx->params->bEnablePsnr     = !!(avctx->flags & AV_CODEC_FLAG_PSNR);
 124
 125     /* Tune the CTU size based on input resolution. */
 126     if (ctx->params->sourceWidth < 64 || ctx->params->sourceHeight < 64)
 127         ctx->params->maxCUSize = 32;
 128     if (ctx->params->sourceWidth < 32 || ctx->params->sourceHeight < 32)
 129         ctx->params->maxCUSize = 16;
 130     if (ctx->params->sourceWidth < 16 || ctx->params->sourceHeight < 16) {
 131         av_log(avctx, AV_LOG_ERROR, "Image size is too small (%dx%d).\n",
 132                ctx->params->sourceWidth, ctx->params->sourceHeight);
 133         return AVERROR(EINVAL);
 134     }
 135
 136     if ((avctx->color_primaries <= AVCOL_PRI_BT2020 &&
 137          avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
 138         (avctx->color_trc <= AVCOL_TRC_BT2020_12 &&
 139          avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
 140         (avctx->colorspace <= AVCOL_SPC_BT2020_CL &&
 141          avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
 142
 143         ctx->params->vui.bEnableVideoSignalTypePresentFlag  = 1;
 144         ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
 145
 146         // x265 validates the parameters internally
 147         ctx->params->vui.colorPrimaries          = avctx->color_primaries;
 148         ctx->params->vui.transferCharacteristics = avctx->color_trc;
 149         ctx->params->vui.matrixCoeffs            = avctx->colorspace;
 150     }
 151
 152     if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
 153         char sar[12];
 154         int sar_num, sar_den;
 155
 156         av_reduce(&sar_num, &sar_den,
 157                   avctx->sample_aspect_ratio.num,
 158                   avctx->sample_aspect_ratio.den, 65535);
 159         snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
 160         if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
 161             av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
 162             return AVERROR_INVALIDDATA;
 163         }
 164     }
 165
 166     switch (avctx->pix_fmt) {
 167     case AV_PIX_FMT_YUV420P:
 168     case AV_PIX_FMT_YUV420P10:
 169         ctx->params->internalCsp = X265_CSP_I420;
 170         break;
 171     case AV_PIX_FMT_YUV422P:
 172     case AV_PIX_FMT_YUV422P10:
 173         ctx->params->internalCsp = X265_CSP_I422;
 174         break;
 175     case AV_PIX_FMT_YUV444P:
 176     case AV_PIX_FMT_YUV444P10:
 177         ctx->params->internalCsp = X265_CSP_I444;
 178         break;
 179     }
 180
 181     if (ctx->crf >= 0) {
 182         char crf[6];
 183
 184         snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
 185         if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
 186             av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
 187             return AVERROR(EINVAL);
 188         }
 189     } else if (avctx->bit_rate > 0) {
 190         ctx->params->rc.bitrate         = avctx->bit_rate / 1000;
 191         ctx->params->rc.rateControlMode = X265_RC_ABR;
 192     }
 193
 194     if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
 195         ctx->params->bRepeatHeaders = 1;
 196
 197     if (ctx->x265_opts) {
 198         AVDictionary *dict    = NULL;
 199         AVDictionaryEntry *en = NULL;
 200
 201         if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
 202             while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
 203                 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
 204
 205                 switch (parse_ret) {
 206                 case X265_PARAM_BAD_NAME:
 207                     av_log(avctx, AV_LOG_WARNING,
 208                           "Unknown option: %s.\n", en->key);
 209                     break;
 210                 case X265_PARAM_BAD_VALUE:
 211                     av_log(avctx, AV_LOG_WARNING,
 212                           "Invalid value for %s: %s.\n", en->key, en->value);
 213                     break;
 214                 default:
 215                     break;
 216                 }
 217             }
 218             av_dict_free(&dict);
 219         }
 220     }
 221
 222     ctx->encoder = ctx->api->encoder_open(ctx->params);
 223     if (!ctx->encoder) {
 224         av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
 225         libx265_encode_close(avctx);
 226         return AVERROR_INVALIDDATA;
 227     }
 228
 229     if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
 230         x265_nal *nal;
 231         int nnal;
 232
 233         avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
 234         if (avctx->extradata_size <= 0) {
 235             av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
 236             libx265_encode_close(avctx);
 237             return AVERROR_INVALIDDATA;
 238         }
 239
 240         avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
 241         if (!avctx->extradata) {
 242             av_log(avctx, AV_LOG_ERROR,
 243                    "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
 244             libx265_encode_close(avctx);
 245             return AVERROR(ENOMEM);
 246         }
 247
 248         memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
 249     }
 250
 251     return 0;
 252 }
 253
 254 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
 255                                 const AVFrame *pic, int *got_packet)
 256 {
 257     libx265Context *ctx = avctx->priv_data;
 258     x265_picture x265pic;
 259     x265_picture x265pic_out = { 0 };
 260     x265_nal *nal;
 261     uint8_t *dst;
 262     int payload = 0;
 263     int nnal;
 264     int ret;
 265     int i;
 266
 267     ctx->api->picture_init(ctx->params, &x265pic);
 268
 269     if (pic) {
 270         for (i = 0; i < 3; i++) {
 271            x265pic.planes[i] = pic->data[i];
 272            x265pic.stride[i] = pic->linesize[i];
 273         }
 274
 275         x265pic.pts      = pic->pts;
 276         x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
 277
 278         x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
 279                                               (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
 280                             pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
 281                             pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
 282                             X265_TYPE_AUTO;
 283     }
 284
 285     ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
 286                                    pic ? &x265pic : NULL, &x265pic_out);
 287     if (ret < 0)
 288         return AVERROR_UNKNOWN;
 289
 290     if (!nnal)
 291         return 0;
 292
 293     for (i = 0; i < nnal; i++)
 294         payload += nal[i].sizeBytes;
 295
 296     ret = ff_alloc_packet(pkt, payload);
 297     if (ret < 0) {
 298         av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
 299         return ret;
 300     }
 301     dst = pkt->data;
 302
 303     for (i = 0; i < nnal; i++) {
 304         memcpy(dst, nal[i].payload, nal[i].sizeBytes);
 305         dst += nal[i].sizeBytes;
 306
 307         if (is_keyframe(nal[i].type))
 308             pkt->flags |= AV_PKT_FLAG_KEY;
 309     }
 310
 311     pkt->pts = x265pic_out.pts;
 312     pkt->dts = x265pic_out.dts;
 313
 314 #if FF_API_CODED_FRAME
 315 FF_DISABLE_DEPRECATION_WARNINGS
 316     switch (x265pic_out.sliceType) {
 317     case X265_TYPE_IDR:
 318     case X265_TYPE_I:
 319         avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
 320         break;
 321     case X265_TYPE_P:
 322         avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
 323         break;
 324     case X265_TYPE_B:
 325         avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
 326         break;
 327     }
 328 FF_ENABLE_DEPRECATION_WARNINGS
 329 #endif
 330
 331     *got_packet = 1;
 332     return 0;
 333 }
 334
 335 static const enum AVPixelFormat x265_csp_eight[] = {
 336     AV_PIX_FMT_YUV420P,
 337     AV_PIX_FMT_YUV422P,
 338     AV_PIX_FMT_YUV444P,
 339     AV_PIX_FMT_NONE
 340 };
 341
 342 static const enum AVPixelFormat x265_csp_twelve[] = {
 343     AV_PIX_FMT_YUV420P,
 344     AV_PIX_FMT_YUV422P,
 345     AV_PIX_FMT_YUV444P,
 346     AV_PIX_FMT_YUV420P10,
 347     AV_PIX_FMT_YUV422P10,
 348     AV_PIX_FMT_YUV444P10,
 349     AV_PIX_FMT_NONE
 350 };
 351
 352 static av_cold void libx265_encode_init_csp(AVCodec *codec)
 353 {
 354     if (x265_max_bit_depth == 8)
 355         codec->pix_fmts = x265_csp_eight;
 356     else if (x265_max_bit_depth == 12)
 357         codec->pix_fmts = x265_csp_twelve;
 358 }
 359
 360 #define OFFSET(x) offsetof(libx265Context, x)
 361 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
 362 static const AVOption options[] = {
 363     { "crf",         "set the x265 crf",                                                            OFFSET(crf),       AV_OPT_TYPE_FLOAT,  { .dbl = -1 }, -1, FLT_MAX, VE },
 364     { "forced-idr",  "if forcing keyframes, force them as IDR frames",                              OFFSET(forced_idr),AV_OPT_TYPE_INT,    { .i64 =  0 },  0,       1, VE },
 365     { "preset",      "set the x265 preset",                                                         OFFSET(preset),    AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
 366     { "tune",        "set the x265 tune parameter",                                                 OFFSET(tune),      AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
 367     { "x265-params", "set the x265 configuration using a :-separated list of key=value parameters", OFFSET(x265_opts), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
 368     { NULL }
 369 };
 370
 371 static const AVClass class = {
 372     .class_name = "libx265",
 373     .item_name  = av_default_item_name,
 374     .option     = options,
 375     .version    = LIBAVUTIL_VERSION_INT,
 376 };
 377
 378 static const AVCodecDefault x265_defaults[] = {
 379     { "b", "0" },
 380     { NULL },
 381 };
 382
 383 AVCodec ff_libx265_encoder = {
 384     .name             = "libx265",
 385     .long_name        = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
 386     .type             = AVMEDIA_TYPE_VIDEO,
 387     .id               = AV_CODEC_ID_HEVC,
 388     .init             = libx265_encode_init,
 389     .init_static_data = libx265_encode_init_csp,
 390     .encode2          = libx265_encode_frame,
 391     .close            = libx265_encode_close,
 392     .priv_data_size   = sizeof(libx265Context),
 393     .priv_class       = &class,
 394     .defaults         = x265_defaults,
 395     .capabilities     = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
 396     .wrapper_name     = "libx265",
 397 };