fd5452193b9d9ad786cc28da9ad893d79a7a92d8
[libav.git] / 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 if ((avctx->color_primaries <= AVCOL_PRI_BT2020 &&
126 avctx->color_primaries != AVCOL_PRI_UNSPECIFIED) ||
127 (avctx->color_trc <= AVCOL_TRC_BT2020_12 &&
128 avctx->color_trc != AVCOL_TRC_UNSPECIFIED) ||
129 (avctx->colorspace <= AVCOL_SPC_BT2020_CL &&
130 avctx->colorspace != AVCOL_SPC_UNSPECIFIED)) {
131
132 ctx->params->vui.bEnableVideoSignalTypePresentFlag = 1;
133 ctx->params->vui.bEnableColorDescriptionPresentFlag = 1;
134
135 // x265 validates the parameters internally
136 ctx->params->vui.colorPrimaries = avctx->color_primaries;
137 ctx->params->vui.transferCharacteristics = avctx->color_trc;
138 ctx->params->vui.matrixCoeffs = avctx->colorspace;
139 }
140
141 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
142 char sar[12];
143 int sar_num, sar_den;
144
145 av_reduce(&sar_num, &sar_den,
146 avctx->sample_aspect_ratio.num,
147 avctx->sample_aspect_ratio.den, 65535);
148 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
149 if (ctx->api->param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
150 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
151 return AVERROR_INVALIDDATA;
152 }
153 }
154
155 switch (avctx->pix_fmt) {
156 case AV_PIX_FMT_YUV420P:
157 case AV_PIX_FMT_YUV420P10:
158 ctx->params->internalCsp = X265_CSP_I420;
159 break;
160 case AV_PIX_FMT_YUV422P:
161 case AV_PIX_FMT_YUV422P10:
162 ctx->params->internalCsp = X265_CSP_I422;
163 break;
164 case AV_PIX_FMT_YUV444P:
165 case AV_PIX_FMT_YUV444P10:
166 ctx->params->internalCsp = X265_CSP_I444;
167 break;
168 }
169
170 if (ctx->crf >= 0) {
171 char crf[6];
172
173 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
174 if (ctx->api->param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
175 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
176 return AVERROR(EINVAL);
177 }
178 } else if (avctx->bit_rate > 0) {
179 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
180 ctx->params->rc.rateControlMode = X265_RC_ABR;
181 }
182
183 if (!(avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER))
184 ctx->params->bRepeatHeaders = 1;
185
186 if (ctx->x265_opts) {
187 AVDictionary *dict = NULL;
188 AVDictionaryEntry *en = NULL;
189
190 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
191 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
192 int parse_ret = ctx->api->param_parse(ctx->params, en->key, en->value);
193
194 switch (parse_ret) {
195 case X265_PARAM_BAD_NAME:
196 av_log(avctx, AV_LOG_WARNING,
197 "Unknown option: %s.\n", en->key);
198 break;
199 case X265_PARAM_BAD_VALUE:
200 av_log(avctx, AV_LOG_WARNING,
201 "Invalid value for %s: %s.\n", en->key, en->value);
202 break;
203 default:
204 break;
205 }
206 }
207 av_dict_free(&dict);
208 }
209 }
210
211 ctx->encoder = ctx->api->encoder_open(ctx->params);
212 if (!ctx->encoder) {
213 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
214 libx265_encode_close(avctx);
215 return AVERROR_INVALIDDATA;
216 }
217
218 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
219 x265_nal *nal;
220 int nnal;
221
222 avctx->extradata_size = ctx->api->encoder_headers(ctx->encoder, &nal, &nnal);
223 if (avctx->extradata_size <= 0) {
224 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
225 libx265_encode_close(avctx);
226 return AVERROR_INVALIDDATA;
227 }
228
229 avctx->extradata = av_malloc(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
230 if (!avctx->extradata) {
231 av_log(avctx, AV_LOG_ERROR,
232 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
233 libx265_encode_close(avctx);
234 return AVERROR(ENOMEM);
235 }
236
237 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
238 }
239
240 return 0;
241 }
242
243 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
244 const AVFrame *pic, int *got_packet)
245 {
246 libx265Context *ctx = avctx->priv_data;
247 x265_picture x265pic;
248 x265_picture x265pic_out = { 0 };
249 x265_nal *nal;
250 uint8_t *dst;
251 int payload = 0;
252 int nnal;
253 int ret;
254 int i;
255
256 ctx->api->picture_init(ctx->params, &x265pic);
257
258 if (pic) {
259 for (i = 0; i < 3; i++) {
260 x265pic.planes[i] = pic->data[i];
261 x265pic.stride[i] = pic->linesize[i];
262 }
263
264 x265pic.pts = pic->pts;
265 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth;
266
267 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ?
268 (ctx->forced_idr ? X265_TYPE_IDR : X265_TYPE_I) :
269 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
270 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
271 X265_TYPE_AUTO;
272 }
273
274 ret = ctx->api->encoder_encode(ctx->encoder, &nal, &nnal,
275 pic ? &x265pic : NULL, &x265pic_out);
276 if (ret < 0)
277 return AVERROR_UNKNOWN;
278
279 if (!nnal)
280 return 0;
281
282 for (i = 0; i < nnal; i++)
283 payload += nal[i].sizeBytes;
284
285 ret = ff_alloc_packet(pkt, payload);
286 if (ret < 0) {
287 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
288 return ret;
289 }
290 dst = pkt->data;
291
292 for (i = 0; i < nnal; i++) {
293 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
294 dst += nal[i].sizeBytes;
295
296 if (is_keyframe(nal[i].type))
297 pkt->flags |= AV_PKT_FLAG_KEY;
298 }
299
300 pkt->pts = x265pic_out.pts;
301 pkt->dts = x265pic_out.dts;
302
303 #if FF_API_CODED_FRAME
304 FF_DISABLE_DEPRECATION_WARNINGS
305 switch (x265pic_out.sliceType) {
306 case X265_TYPE_IDR:
307 case X265_TYPE_I:
308 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
309 break;
310 case X265_TYPE_P:
311 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
312 break;
313 case X265_TYPE_B:
314 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
315 break;
316 }
317 FF_ENABLE_DEPRECATION_WARNINGS
318 #endif
319
320 *got_packet = 1;
321 return 0;
322 }
323
324 static const enum AVPixelFormat x265_csp_eight[] = {
325 AV_PIX_FMT_YUV420P,
326 AV_PIX_FMT_YUV422P,
327 AV_PIX_FMT_YUV444P,
328 AV_PIX_FMT_NONE
329 };
330
331 static const enum AVPixelFormat x265_csp_twelve[] = {
332 AV_PIX_FMT_YUV420P,
333 AV_PIX_FMT_YUV422P,
334 AV_PIX_FMT_YUV444P,
335 AV_PIX_FMT_YUV420P10,
336 AV_PIX_FMT_YUV422P10,
337 AV_PIX_FMT_YUV444P10,
338 AV_PIX_FMT_NONE
339 };
340
341 static av_cold void libx265_encode_init_csp(AVCodec *codec)
342 {
343 if (x265_max_bit_depth == 8)
344 codec->pix_fmts = x265_csp_eight;
345 else if (x265_max_bit_depth == 12)
346 codec->pix_fmts = x265_csp_twelve;
347 }
348
349 #define OFFSET(x) offsetof(libx265Context, x)
350 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
351 static const AVOption options[] = {
352 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
353 { "forced-idr", "if forcing keyframes, force them as IDR frames", OFFSET(forced_idr),AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
354 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
355 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
356 { "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 },
357 { NULL }
358 };
359
360 static const AVClass class = {
361 .class_name = "libx265",
362 .item_name = av_default_item_name,
363 .option = options,
364 .version = LIBAVUTIL_VERSION_INT,
365 };
366
367 static const AVCodecDefault x265_defaults[] = {
368 { "b", "0" },
369 { NULL },
370 };
371
372 AVCodec ff_libx265_encoder = {
373 .name = "libx265",
374 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
375 .type = AVMEDIA_TYPE_VIDEO,
376 .id = AV_CODEC_ID_HEVC,
377 .init = libx265_encode_init,
378 .init_static_data = libx265_encode_init_csp,
379 .encode2 = libx265_encode_frame,
380 .close = libx265_encode_close,
381 .priv_data_size = sizeof(libx265Context),
382 .priv_class = &class,
383 .defaults = x265_defaults,
384 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
385 .wrapper_name = "libx265",
386 };