c04299aaf358ceb754b07b9b8fc8f16f2d40ad99
[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
43 float crf;
44 char *preset;
45 char *tune;
46 char *x265_opts;
47 } libx265Context;
48
49 static int is_keyframe(NalUnitType naltype)
50 {
51 switch (naltype) {
52 case NAL_UNIT_CODED_SLICE_BLA_W_LP:
53 case NAL_UNIT_CODED_SLICE_BLA_W_RADL:
54 case NAL_UNIT_CODED_SLICE_BLA_N_LP:
55 case NAL_UNIT_CODED_SLICE_IDR_W_RADL:
56 case NAL_UNIT_CODED_SLICE_IDR_N_LP:
57 case NAL_UNIT_CODED_SLICE_CRA:
58 return 1;
59 default:
60 return 0;
61 }
62 }
63
64 static av_cold int libx265_encode_close(AVCodecContext *avctx)
65 {
66 libx265Context *ctx = avctx->priv_data;
67
68 av_frame_free(&avctx->coded_frame);
69
70 x265_param_free(ctx->params);
71
72 if (ctx->encoder)
73 x265_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 if (avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL &&
83 !av_pix_fmt_desc_get(avctx->pix_fmt)->log2_chroma_w) {
84 av_log(avctx, AV_LOG_ERROR,
85 "4:2:2 and 4:4:4 support is not fully defined for HEVC yet. "
86 "Set -strict experimental to encode anyway.\n");
87 return AVERROR(ENOSYS);
88 }
89
90 avctx->coded_frame = av_frame_alloc();
91 if (!avctx->coded_frame) {
92 av_log(avctx, AV_LOG_ERROR, "Could not allocate frame.\n");
93 return AVERROR(ENOMEM);
94 }
95
96 ctx->params = x265_param_alloc();
97 if (!ctx->params) {
98 av_log(avctx, AV_LOG_ERROR, "Could not allocate x265 param structure.\n");
99 return AVERROR(ENOMEM);
100 }
101
102 if (x265_param_default_preset(ctx->params, ctx->preset, ctx->tune) < 0) {
103 int i;
104
105 av_log(avctx, AV_LOG_ERROR, "Error setting preset/tune %s/%s.\n", ctx->preset, ctx->tune);
106 av_log(avctx, AV_LOG_INFO, "Possible presets:");
107 for (i = 0; x265_preset_names[i]; i++)
108 av_log(avctx, AV_LOG_INFO, " %s", x265_preset_names[i]);
109
110 av_log(avctx, AV_LOG_INFO, "\n");
111 av_log(avctx, AV_LOG_INFO, "Possible tunes:");
112 for (i = 0; x265_tune_names[i]; i++)
113 av_log(avctx, AV_LOG_INFO, " %s", x265_tune_names[i]);
114
115 av_log(avctx, AV_LOG_INFO, "\n");
116
117 return AVERROR(EINVAL);
118 }
119
120 ctx->params->frameNumThreads = avctx->thread_count;
121 ctx->params->fpsNum = avctx->time_base.den;
122 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
123 ctx->params->sourceWidth = avctx->width;
124 ctx->params->sourceHeight = avctx->height;
125 ctx->params->bEnablePsnr = !!(avctx->flags & CODEC_FLAG_PSNR);
126
127 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
128 char sar[12];
129 int sar_num, sar_den;
130
131 av_reduce(&sar_num, &sar_den,
132 avctx->sample_aspect_ratio.num,
133 avctx->sample_aspect_ratio.den, 65535);
134 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
135 if (x265_param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
136 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
137 return AVERROR_INVALIDDATA;
138 }
139 }
140
141 switch (avctx->pix_fmt) {
142 case AV_PIX_FMT_YUV420P:
143 case AV_PIX_FMT_YUV420P10:
144 ctx->params->internalCsp = X265_CSP_I420;
145 break;
146 case AV_PIX_FMT_YUV422P:
147 case AV_PIX_FMT_YUV422P10:
148 ctx->params->internalCsp = X265_CSP_I422;
149 break;
150 case AV_PIX_FMT_YUV444P:
151 case AV_PIX_FMT_YUV444P10:
152 ctx->params->internalCsp = X265_CSP_I444;
153 break;
154 }
155
156 if (ctx->crf >= 0) {
157 char crf[6];
158
159 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
160 if (x265_param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
161 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
162 return AVERROR(EINVAL);
163 }
164 } else if (avctx->bit_rate > 0) {
165 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
166 ctx->params->rc.rateControlMode = X265_RC_ABR;
167 }
168
169 if (!(avctx->flags & CODEC_FLAG_GLOBAL_HEADER))
170 ctx->params->bRepeatHeaders = 1;
171
172 if (ctx->x265_opts) {
173 AVDictionary *dict = NULL;
174 AVDictionaryEntry *en = NULL;
175
176 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
177 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
178 int parse_ret = x265_param_parse(ctx->params, en->key, en->value);
179
180 switch (parse_ret) {
181 case X265_PARAM_BAD_NAME:
182 av_log(avctx, AV_LOG_WARNING,
183 "Unknown option: %s.\n", en->key);
184 break;
185 case X265_PARAM_BAD_VALUE:
186 av_log(avctx, AV_LOG_WARNING,
187 "Invalid value for %s: %s.\n", en->key, en->value);
188 break;
189 default:
190 break;
191 }
192 }
193 av_dict_free(&dict);
194 }
195 }
196
197 ctx->encoder = x265_encoder_open(ctx->params);
198 if (!ctx->encoder) {
199 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
200 libx265_encode_close(avctx);
201 return AVERROR_INVALIDDATA;
202 }
203
204 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
205 x265_nal *nal;
206 int nnal;
207
208 avctx->extradata_size = x265_encoder_headers(ctx->encoder, &nal, &nnal);
209 if (avctx->extradata_size <= 0) {
210 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
211 libx265_encode_close(avctx);
212 return AVERROR_INVALIDDATA;
213 }
214
215 avctx->extradata = av_malloc(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
216 if (!avctx->extradata) {
217 av_log(avctx, AV_LOG_ERROR,
218 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
219 libx265_encode_close(avctx);
220 return AVERROR(ENOMEM);
221 }
222
223 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
224 }
225
226 return 0;
227 }
228
229 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
230 const AVFrame *pic, int *got_packet)
231 {
232 libx265Context *ctx = avctx->priv_data;
233 x265_picture x265pic;
234 x265_picture x265pic_out = { { 0 } };
235 x265_nal *nal;
236 uint8_t *dst;
237 int payload = 0;
238 int nnal;
239 int ret;
240 int i;
241
242 x265_picture_init(ctx->params, &x265pic);
243
244 if (pic) {
245 for (i = 0; i < 3; i++) {
246 x265pic.planes[i] = pic->data[i];
247 x265pic.stride[i] = pic->linesize[i];
248 }
249
250 x265pic.pts = pic->pts;
251 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth_minus1 + 1;
252
253 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ? X265_TYPE_I :
254 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
255 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
256 X265_TYPE_AUTO;
257 }
258
259 ret = x265_encoder_encode(ctx->encoder, &nal, &nnal,
260 pic ? &x265pic : NULL, &x265pic_out);
261 if (ret < 0)
262 return AVERROR_UNKNOWN;
263
264 if (!nnal)
265 return 0;
266
267 for (i = 0; i < nnal; i++)
268 payload += nal[i].sizeBytes;
269
270 ret = ff_alloc_packet(pkt, payload);
271 if (ret < 0) {
272 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
273 return ret;
274 }
275 dst = pkt->data;
276
277 for (i = 0; i < nnal; i++) {
278 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
279 dst += nal[i].sizeBytes;
280
281 if (is_keyframe(nal[i].type))
282 pkt->flags |= AV_PKT_FLAG_KEY;
283 }
284
285 pkt->pts = x265pic_out.pts;
286 pkt->dts = x265pic_out.dts;
287
288 switch (x265pic_out.sliceType) {
289 case X265_TYPE_IDR:
290 case X265_TYPE_I:
291 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
292 break;
293 case X265_TYPE_P:
294 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
295 break;
296 case X265_TYPE_B:
297 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
298 break;
299 }
300
301 *got_packet = 1;
302 return 0;
303 }
304
305 static const enum AVPixelFormat x265_csp_eight[] = {
306 AV_PIX_FMT_YUV420P,
307 AV_PIX_FMT_YUV422P,
308 AV_PIX_FMT_YUV444P,
309 AV_PIX_FMT_NONE
310 };
311
312 static const enum AVPixelFormat x265_csp_twelve[] = {
313 AV_PIX_FMT_YUV420P,
314 AV_PIX_FMT_YUV422P,
315 AV_PIX_FMT_YUV444P,
316 AV_PIX_FMT_YUV420P10,
317 AV_PIX_FMT_YUV422P10,
318 AV_PIX_FMT_YUV444P10,
319 AV_PIX_FMT_NONE
320 };
321
322 static av_cold void libx265_encode_init_csp(AVCodec *codec)
323 {
324 if (x265_max_bit_depth == 8)
325 codec->pix_fmts = x265_csp_eight;
326 else if (x265_max_bit_depth == 12)
327 codec->pix_fmts = x265_csp_twelve;
328 }
329
330 #define OFFSET(x) offsetof(libx265Context, x)
331 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
332 static const AVOption options[] = {
333 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
334 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
335 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
336 { "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 },
337 { NULL }
338 };
339
340 static const AVClass class = {
341 .class_name = "libx265",
342 .item_name = av_default_item_name,
343 .option = options,
344 .version = LIBAVUTIL_VERSION_INT,
345 };
346
347 static const AVCodecDefault x265_defaults[] = {
348 { "b", "0" },
349 { NULL },
350 };
351
352 AVCodec ff_libx265_encoder = {
353 .name = "libx265",
354 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
355 .type = AVMEDIA_TYPE_VIDEO,
356 .id = AV_CODEC_ID_HEVC,
357 .init = libx265_encode_init,
358 .init_static_data = libx265_encode_init_csp,
359 .encode2 = libx265_encode_frame,
360 .close = libx265_encode_close,
361 .priv_data_size = sizeof(libx265Context),
362 .priv_class = &class,
363 .defaults = x265_defaults,
364 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
365 };