libx265: Add crf private option
[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 av_log(avctx, AV_LOG_ERROR, "Invalid preset or tune.\n");
104 return AVERROR(EINVAL);
105 }
106
107 ctx->params->frameNumThreads = avctx->thread_count;
108 ctx->params->fpsNum = avctx->time_base.den;
109 ctx->params->fpsDenom = avctx->time_base.num * avctx->ticks_per_frame;
110 ctx->params->sourceWidth = avctx->width;
111 ctx->params->sourceHeight = avctx->height;
112 ctx->params->bEnablePsnr = !!(avctx->flags & CODEC_FLAG_PSNR);
113
114 if (avctx->sample_aspect_ratio.num > 0 && avctx->sample_aspect_ratio.den > 0) {
115 char sar[12];
116 int sar_num, sar_den;
117
118 av_reduce(&sar_num, &sar_den,
119 avctx->sample_aspect_ratio.num,
120 avctx->sample_aspect_ratio.den, 65535);
121 snprintf(sar, sizeof(sar), "%d:%d", sar_num, sar_den);
122 if (x265_param_parse(ctx->params, "sar", sar) == X265_PARAM_BAD_VALUE) {
123 av_log(avctx, AV_LOG_ERROR, "Invalid SAR: %d:%d.\n", sar_num, sar_den);
124 return AVERROR_INVALIDDATA;
125 }
126 }
127
128 switch (avctx->pix_fmt) {
129 case AV_PIX_FMT_YUV420P:
130 case AV_PIX_FMT_YUV420P10:
131 ctx->params->internalCsp = X265_CSP_I420;
132 break;
133 case AV_PIX_FMT_YUV422P:
134 case AV_PIX_FMT_YUV422P10:
135 ctx->params->internalCsp = X265_CSP_I422;
136 break;
137 case AV_PIX_FMT_YUV444P:
138 case AV_PIX_FMT_YUV444P10:
139 ctx->params->internalCsp = X265_CSP_I444;
140 break;
141 }
142
143 if (ctx->crf >= 0) {
144 char crf[6];
145
146 snprintf(crf, sizeof(crf), "%2.2f", ctx->crf);
147 if (x265_param_parse(ctx->params, "crf", crf) == X265_PARAM_BAD_VALUE) {
148 av_log(avctx, AV_LOG_ERROR, "Invalid crf: %2.2f.\n", ctx->crf);
149 return AVERROR(EINVAL);
150 }
151 } else if (avctx->bit_rate > 0) {
152 ctx->params->rc.bitrate = avctx->bit_rate / 1000;
153 ctx->params->rc.rateControlMode = X265_RC_ABR;
154 }
155
156 if (!(avctx->flags & CODEC_FLAG_GLOBAL_HEADER))
157 ctx->params->bRepeatHeaders = 1;
158
159 if (ctx->x265_opts) {
160 AVDictionary *dict = NULL;
161 AVDictionaryEntry *en = NULL;
162
163 if (!av_dict_parse_string(&dict, ctx->x265_opts, "=", ":", 0)) {
164 while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
165 int parse_ret = x265_param_parse(ctx->params, en->key, en->value);
166
167 switch (parse_ret) {
168 case X265_PARAM_BAD_NAME:
169 av_log(avctx, AV_LOG_WARNING,
170 "Unknown option: %s.\n", en->key);
171 break;
172 case X265_PARAM_BAD_VALUE:
173 av_log(avctx, AV_LOG_WARNING,
174 "Invalid value for %s: %s.\n", en->key, en->value);
175 break;
176 default:
177 break;
178 }
179 }
180 av_dict_free(&dict);
181 }
182 }
183
184 ctx->encoder = x265_encoder_open(ctx->params);
185 if (!ctx->encoder) {
186 av_log(avctx, AV_LOG_ERROR, "Cannot open libx265 encoder.\n");
187 libx265_encode_close(avctx);
188 return AVERROR_INVALIDDATA;
189 }
190
191 if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
192 x265_nal *nal;
193 int nnal;
194
195 avctx->extradata_size = x265_encoder_headers(ctx->encoder, &nal, &nnal);
196 if (avctx->extradata_size <= 0) {
197 av_log(avctx, AV_LOG_ERROR, "Cannot encode headers.\n");
198 libx265_encode_close(avctx);
199 return AVERROR_INVALIDDATA;
200 }
201
202 avctx->extradata = av_malloc(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
203 if (!avctx->extradata) {
204 av_log(avctx, AV_LOG_ERROR,
205 "Cannot allocate HEVC header of size %d.\n", avctx->extradata_size);
206 libx265_encode_close(avctx);
207 return AVERROR(ENOMEM);
208 }
209
210 memcpy(avctx->extradata, nal[0].payload, avctx->extradata_size);
211 }
212
213 return 0;
214 }
215
216 static int libx265_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
217 const AVFrame *pic, int *got_packet)
218 {
219 libx265Context *ctx = avctx->priv_data;
220 x265_picture x265pic;
221 x265_picture x265pic_out = { { 0 } };
222 x265_nal *nal;
223 uint8_t *dst;
224 int payload = 0;
225 int nnal;
226 int ret;
227 int i;
228
229 x265_picture_init(ctx->params, &x265pic);
230
231 if (pic) {
232 for (i = 0; i < 3; i++) {
233 x265pic.planes[i] = pic->data[i];
234 x265pic.stride[i] = pic->linesize[i];
235 }
236
237 x265pic.pts = pic->pts;
238 x265pic.bitDepth = av_pix_fmt_desc_get(avctx->pix_fmt)->comp[0].depth_minus1 + 1;
239
240 x265pic.sliceType = pic->pict_type == AV_PICTURE_TYPE_I ? X265_TYPE_I :
241 pic->pict_type == AV_PICTURE_TYPE_P ? X265_TYPE_P :
242 pic->pict_type == AV_PICTURE_TYPE_B ? X265_TYPE_B :
243 X265_TYPE_AUTO;
244 }
245
246 ret = x265_encoder_encode(ctx->encoder, &nal, &nnal,
247 pic ? &x265pic : NULL, &x265pic_out);
248 if (ret < 0)
249 return AVERROR_UNKNOWN;
250
251 if (!nnal)
252 return 0;
253
254 for (i = 0; i < nnal; i++)
255 payload += nal[i].sizeBytes;
256
257 ret = ff_alloc_packet(pkt, payload);
258 if (ret < 0) {
259 av_log(avctx, AV_LOG_ERROR, "Error getting output packet.\n");
260 return ret;
261 }
262 dst = pkt->data;
263
264 for (i = 0; i < nnal; i++) {
265 memcpy(dst, nal[i].payload, nal[i].sizeBytes);
266 dst += nal[i].sizeBytes;
267
268 if (is_keyframe(nal[i].type))
269 pkt->flags |= AV_PKT_FLAG_KEY;
270 }
271
272 pkt->pts = x265pic_out.pts;
273 pkt->dts = x265pic_out.dts;
274
275 *got_packet = 1;
276 return 0;
277 }
278
279 static const enum AVPixelFormat x265_csp_eight[] = {
280 AV_PIX_FMT_YUV420P,
281 AV_PIX_FMT_YUV422P,
282 AV_PIX_FMT_YUV444P,
283 AV_PIX_FMT_NONE
284 };
285
286 static const enum AVPixelFormat x265_csp_twelve[] = {
287 AV_PIX_FMT_YUV420P,
288 AV_PIX_FMT_YUV422P,
289 AV_PIX_FMT_YUV444P,
290 AV_PIX_FMT_YUV420P10,
291 AV_PIX_FMT_YUV422P10,
292 AV_PIX_FMT_YUV444P10,
293 AV_PIX_FMT_NONE
294 };
295
296 static av_cold void libx265_encode_init_csp(AVCodec *codec)
297 {
298 if (x265_max_bit_depth == 8)
299 codec->pix_fmts = x265_csp_eight;
300 else if (x265_max_bit_depth == 12)
301 codec->pix_fmts = x265_csp_twelve;
302 }
303
304 #define OFFSET(x) offsetof(libx265Context, x)
305 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
306 static const AVOption options[] = {
307 { "crf", "set the x265 crf", OFFSET(crf), AV_OPT_TYPE_FLOAT, { .dbl = -1 }, -1, FLT_MAX, VE },
308 { "preset", "set the x265 preset", OFFSET(preset), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
309 { "tune", "set the x265 tune parameter", OFFSET(tune), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
310 { "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 },
311 { NULL }
312 };
313
314 static const AVClass class = {
315 .class_name = "libx265",
316 .item_name = av_default_item_name,
317 .option = options,
318 .version = LIBAVUTIL_VERSION_INT,
319 };
320
321 static const AVCodecDefault x265_defaults[] = {
322 { "b", "0" },
323 { NULL },
324 };
325
326 AVCodec ff_libx265_encoder = {
327 .name = "libx265",
328 .long_name = NULL_IF_CONFIG_SMALL("libx265 H.265 / HEVC"),
329 .type = AVMEDIA_TYPE_VIDEO,
330 .id = AV_CODEC_ID_HEVC,
331 .init = libx265_encode_init,
332 .init_static_data = libx265_encode_init_csp,
333 .encode2 = libx265_encode_frame,
334 .close = libx265_encode_close,
335 .priv_data_size = sizeof(libx265Context),
336 .priv_class = &class,
337 .defaults = x265_defaults,
338 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
339 };