1 /*
2 * libkvazaar encoder
3 *
4 * Copyright (c) 2015 Tampere University of Technology
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #include <kvazaar.h>
24 #include <stdint.h>
25 #include <string.h>
26
27 #include "libavutil/attributes.h"
28 #include "libavutil/avassert.h"
29 #include "libavutil/dict.h"
30 #include "libavutil/error.h"
31 #include "libavutil/imgutils.h"
32 #include "libavutil/internal.h"
33 #include "libavutil/log.h"
34 #include "libavutil/mem.h"
35 #include "libavutil/pixdesc.h"
36 #include "libavutil/opt.h"
37
38 #include "avcodec.h"
39 #include "internal.h"
40
41 typedef struct LibkvazaarContext {
42 const AVClass *class;
43
44 const kvz_api *api;
45 kvz_encoder *encoder;
46 kvz_config *config;
47
48 char *kvz_params;
49 } LibkvazaarContext;
50
libkvazaar_init(AVCodecContext * avctx)51 static av_cold int libkvazaar_init(AVCodecContext *avctx)
52 {
53 LibkvazaarContext *const ctx = avctx->priv_data;
54 const kvz_api *const api = ctx->api = kvz_api_get(8);
55 kvz_config *cfg = NULL;
56 kvz_encoder *enc = NULL;
57
58 /* Kvazaar requires width and height to be multiples of eight. */
59 if (avctx->width % 8 || avctx->height % 8) {
60 av_log(avctx, AV_LOG_ERROR,
61 "Video dimensions are not a multiple of 8 (%dx%d).\n",
62 avctx->width, avctx->height);
63 return AVERROR(ENOSYS);
64 }
65
66 ctx->config = cfg = api->config_alloc();
67 if (!cfg) {
68 av_log(avctx, AV_LOG_ERROR,
69 "Could not allocate kvazaar config structure.\n");
70 return AVERROR(ENOMEM);
71 }
72
73 if (!api->config_init(cfg)) {
74 av_log(avctx, AV_LOG_ERROR,
75 "Could not initialize kvazaar config structure.\n");
76 return AVERROR_BUG;
77 }
78
79 cfg->width = avctx->width;
80 cfg->height = avctx->height;
81
82 if (avctx->framerate.num > 0 && avctx->framerate.den > 0) {
83 if (avctx->ticks_per_frame > INT_MAX / avctx->framerate.den) {
84 av_log(avctx, AV_LOG_ERROR,
85 "Could not set framerate for kvazaar: integer overflow\n");
86 return AVERROR(EINVAL);
87 }
88 cfg->framerate_num = avctx->framerate.num;
89 cfg->framerate_denom = avctx->time_base.den * avctx->ticks_per_frame;
90 } else {
91 if (avctx->ticks_per_frame > INT_MAX / avctx->time_base.num) {
92 av_log(avctx, AV_LOG_ERROR,
93 "Could not set framerate for kvazaar: integer overflow\n");
94 return AVERROR(EINVAL);
95 }
96 cfg->framerate_num = avctx->time_base.den;
97 cfg->framerate_denom = avctx->time_base.num * avctx->ticks_per_frame;
98 }
99 cfg->target_bitrate = avctx->bit_rate;
100 cfg->vui.sar_width = avctx->sample_aspect_ratio.num;
101 cfg->vui.sar_height = avctx->sample_aspect_ratio.den;
102
103 if (ctx->kvz_params) {
104 AVDictionary *dict = NULL;
105 if (!av_dict_parse_string(&dict, ctx->kvz_params, "=", ",", 0)) {
106 AVDictionaryEntry *entry = NULL;
107 while ((entry = av_dict_get(dict, "", entry, AV_DICT_IGNORE_SUFFIX))) {
108 if (!api->config_parse(cfg, entry->key, entry->value)) {
109 av_log(avctx, AV_LOG_WARNING, "Invalid option: %s=%s.\n",
110 entry->key, entry->value);
111 }
112 }
113 }
114 av_dict_free(&dict);
115 }
116
117 ctx->encoder = enc = api->encoder_open(cfg);
118 if (!enc) {
119 av_log(avctx, AV_LOG_ERROR, "Could not open kvazaar encoder.\n");
120 return AVERROR_BUG;
121 }
122
123 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
124 kvz_data_chunk *data_out = NULL;
125 kvz_data_chunk *chunk = NULL;
126 uint32_t len_out;
127 uint8_t *p;
128
129 if (!api->encoder_headers(enc, &data_out, &len_out))
130 return AVERROR(ENOMEM);
131
132 avctx->extradata = p = av_mallocz(len_out + AV_INPUT_BUFFER_PADDING_SIZE);
133 if (!p) {
134 ctx->api->chunk_free(data_out);
135 return AVERROR(ENOMEM);
136 }
137
138 avctx->extradata_size = len_out;
139
140 for (chunk = data_out; chunk != NULL; chunk = chunk->next) {
141 memcpy(p, chunk->data, chunk->len);
142 p += chunk->len;
143 }
144
145 ctx->api->chunk_free(data_out);
146 }
147
148 return 0;
149 }
150
libkvazaar_close(AVCodecContext * avctx)151 static av_cold int libkvazaar_close(AVCodecContext *avctx)
152 {
153 LibkvazaarContext *ctx = avctx->priv_data;
154
155 if (ctx->api) {
156 ctx->api->encoder_close(ctx->encoder);
157 ctx->api->config_destroy(ctx->config);
158 }
159
160 if (avctx->extradata)
161 av_freep(&avctx->extradata);
162
163 return 0;
164 }
165
libkvazaar_encode(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * frame,int * got_packet_ptr)166 static int libkvazaar_encode(AVCodecContext *avctx,
167 AVPacket *avpkt,
168 const AVFrame *frame,
169 int *got_packet_ptr)
170 {
171 LibkvazaarContext *ctx = avctx->priv_data;
172 kvz_picture *input_pic = NULL;
173 kvz_picture *recon_pic = NULL;
174 kvz_frame_info frame_info;
175 kvz_data_chunk *data_out = NULL;
176 uint32_t len_out = 0;
177 int retval = 0;
178
179 *got_packet_ptr = 0;
180
181 if (frame) {
182 if (frame->width != ctx->config->width ||
183 frame->height != ctx->config->height) {
184 av_log(avctx, AV_LOG_ERROR,
185 "Changing video dimensions during encoding is not supported. "
186 "(changed from %dx%d to %dx%d)\n",
187 ctx->config->width, ctx->config->height,
188 frame->width, frame->height);
189 retval = AVERROR_INVALIDDATA;
190 goto done;
191 }
192
193 if (frame->format != avctx->pix_fmt) {
194 av_log(avctx, AV_LOG_ERROR,
195 "Changing pixel format during encoding is not supported. "
196 "(changed from %s to %s)\n",
197 av_get_pix_fmt_name(avctx->pix_fmt),
198 av_get_pix_fmt_name(frame->format));
199 retval = AVERROR_INVALIDDATA;
200 goto done;
201 }
202
203 // Allocate input picture for kvazaar.
204 input_pic = ctx->api->picture_alloc(frame->width, frame->height);
205 if (!input_pic) {
206 av_log(avctx, AV_LOG_ERROR, "Failed to allocate picture.\n");
207 retval = AVERROR(ENOMEM);
208 goto done;
209 }
210
211 // Copy pixels from frame to input_pic.
212 {
213 int dst_linesizes[4] = {
214 frame->width,
215 frame->width / 2,
216 frame->width / 2,
217 0
218 };
219 av_image_copy(input_pic->data, dst_linesizes,
220 (const uint8_t **)frame->data, frame->linesize,
221 frame->format, frame->width, frame->height);
222 }
223
224 input_pic->pts = frame->pts;
225 }
226
227 retval = ctx->api->encoder_encode(ctx->encoder,
228 input_pic,
229 &data_out, &len_out,
230 &recon_pic, NULL,
231 &frame_info);
232 if (!retval) {
233 av_log(avctx, AV_LOG_ERROR, "Failed to encode frame.\n");
234 retval = AVERROR_INVALIDDATA;
235 goto done;
236 } else
237 retval = 0; /* kvazaar returns 1 on success */
238
239 if (data_out) {
240 kvz_data_chunk *chunk = NULL;
241 uint64_t written = 0;
242
243 retval = ff_alloc_packet2(avctx, avpkt, len_out, len_out);
244 if (retval < 0) {
245 av_log(avctx, AV_LOG_ERROR, "Failed to allocate output packet.\n");
246 goto done;
247 }
248
249 for (chunk = data_out; chunk != NULL; chunk = chunk->next) {
250 av_assert0(written + chunk->len <= len_out);
251 memcpy(avpkt->data + written, chunk->data, chunk->len);
252 written += chunk->len;
253 }
254
255 avpkt->pts = recon_pic->pts;
256 avpkt->dts = recon_pic->dts;
257 avpkt->flags = 0;
258 // IRAP VCL NAL unit types span the range
259 // [BLA_W_LP (16), RSV_IRAP_VCL23 (23)].
260 if (frame_info.nal_unit_type >= KVZ_NAL_BLA_W_LP &&
261 frame_info.nal_unit_type <= KVZ_NAL_RSV_IRAP_VCL23) {
262 avpkt->flags |= AV_PKT_FLAG_KEY;
263 }
264
265 *got_packet_ptr = 1;
266 }
267
268 done:
269 ctx->api->picture_free(input_pic);
270 ctx->api->picture_free(recon_pic);
271 ctx->api->chunk_free(data_out);
272 return retval;
273 }
274
275 static const enum AVPixelFormat pix_fmts[] = {
276 AV_PIX_FMT_YUV420P,
277 AV_PIX_FMT_NONE
278 };
279
280 #define OFFSET(x) offsetof(LibkvazaarContext, x)
281 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
282 static const AVOption options[] = {
283 { "kvazaar-params", "Set kvazaar parameters as a comma-separated list of key=value pairs.",
284 OFFSET(kvz_params), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
285 { NULL },
286 };
287
288 static const AVClass class = {
289 .class_name = "libkvazaar",
290 .item_name = av_default_item_name,
291 .option = options,
292 .version = LIBAVUTIL_VERSION_INT,
293 };
294
295 static const AVCodecDefault defaults[] = {
296 { "b", "0" },
297 { NULL },
298 };
299
300 AVCodec ff_libkvazaar_encoder = {
301 .name = "libkvazaar",
302 .long_name = NULL_IF_CONFIG_SMALL("libkvazaar H.265 / HEVC"),
303 .type = AVMEDIA_TYPE_VIDEO,
304 .id = AV_CODEC_ID_HEVC,
305 .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
306 .pix_fmts = pix_fmts,
307
308 .priv_class = &class,
309 .priv_data_size = sizeof(LibkvazaarContext),
310 .defaults = defaults,
311
312 .init = libkvazaar_init,
313 .encode2 = libkvazaar_encode,
314 .close = libkvazaar_close,
315
316 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
317
318 .wrapper_name = "libkvazaar",
319 };
320