qsvvpp: Perform full init only when needed
[libav.git] / libavcodec / libfdk-aacenc.c
CommitLineData
37eeb5e2
MS
1/*
2 * AAC encoder wrapper
3 * Copyright (c) 2012 Martin Storsjo
4 *
5 * This file is part of Libav.
6 *
2f706519
MS
7 * Permission to use, copy, modify, and/or distribute this software for any
8 * purpose with or without fee is hereby granted, provided that the above
9 * copyright notice and this permission notice appear in all copies.
37eeb5e2 10 *
2f706519
MS
11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
37eeb5e2
MS
18 */
19
20#include <fdk-aac/aacenc_lib.h>
21
a903f8f0
JR
22#include "libavutil/channel_layout.h"
23#include "libavutil/common.h"
24#include "libavutil/opt.h"
37eeb5e2
MS
25#include "avcodec.h"
26#include "audio_frame_queue.h"
27#include "internal.h"
37eeb5e2 28
141c960e
MS
29#define FDKENC_VER_AT_LEAST(vl0, vl1) \
30 (defined(AACENCODER_LIB_VL0) && \
31 ((AACENCODER_LIB_VL0 > vl0) || \
32 (AACENCODER_LIB_VL0 == vl0 && AACENCODER_LIB_VL1 >= vl1)))
33
37eeb5e2
MS
34typedef struct AACContext {
35 const AVClass *class;
36 HANDLE_AACENCODER handle;
37 int afterburner;
38 int eld_sbr;
39 int signaling;
160a27c5
KK
40 int latm;
41 int header_period;
375bff70 42 int vbr;
37eeb5e2
MS
43
44 AudioFrameQueue afq;
45} AACContext;
46
47static const AVOption aac_enc_options[] = {
e6153f17
MS
48 { "afterburner", "Afterburner (improved quality)", offsetof(AACContext, afterburner), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
49 { "eld_sbr", "Enable SBR for ELD (for SBR in other configurations, use the -profile parameter)", offsetof(AACContext, eld_sbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
50 { "signaling", "SBR/PS signaling style", offsetof(AACContext, signaling), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 2, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
124134e4
MS
51 { "default", "Choose signaling implicitly (explicit hierarchical by default, implicit if global header is disabled)", 0, AV_OPT_TYPE_CONST, { .i64 = -1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
52 { "implicit", "Implicit backwards compatible signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
53 { "explicit_sbr", "Explicit SBR, implicit PS signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
54 { "explicit_hierarchical", "Explicit hierarchical signaling", 0, AV_OPT_TYPE_CONST, { .i64 = 2 }, 0, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM, "signaling" },
e6153f17
MS
55 { "latm", "Output LATM/LOAS encapsulated data", offsetof(AACContext, latm), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
56 { "header_period", "StreamMuxConfig and PCE repetition period (in frames)", offsetof(AACContext, header_period), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 0xffff, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
375bff70 57 { "vbr", "VBR mode (1-5)", offsetof(AACContext, vbr), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 5, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
37eeb5e2
MS
58 { NULL }
59};
60
61static const AVClass aac_enc_class = {
97cfe1d8
DB
62 .class_name = "libfdk_aac",
63 .item_name = av_default_item_name,
64 .option = aac_enc_options,
65 .version = LIBAVUTIL_VERSION_INT,
37eeb5e2
MS
66};
67
68static const char *aac_get_error(AACENC_ERROR err)
69{
70 switch (err) {
71 case AACENC_OK:
72 return "No error";
73 case AACENC_INVALID_HANDLE:
74 return "Invalid handle";
75 case AACENC_MEMORY_ERROR:
76 return "Memory allocation error";
77 case AACENC_UNSUPPORTED_PARAMETER:
78 return "Unsupported parameter";
79 case AACENC_INVALID_CONFIG:
80 return "Invalid config";
81 case AACENC_INIT_ERROR:
82 return "Initialization error";
83 case AACENC_INIT_AAC_ERROR:
84 return "AAC library initialization error";
85 case AACENC_INIT_SBR_ERROR:
86 return "SBR library initialization error";
87 case AACENC_INIT_TP_ERROR:
88 return "Transport library initialization error";
89 case AACENC_INIT_META_ERROR:
90 return "Metadata library initialization error";
91 case AACENC_ENCODE_ERROR:
92 return "Encoding error";
93 case AACENC_ENCODE_EOF:
94 return "End of file";
95 default:
96 return "Unknown error";
97 }
98}
99
100static int aac_encode_close(AVCodecContext *avctx)
101{
102 AACContext *s = avctx->priv_data;
103
104 if (s->handle)
105 aacEncClose(&s->handle);
37eeb5e2
MS
106 av_freep(&avctx->extradata);
107 ff_af_queue_close(&s->afq);
108
109 return 0;
110}
111
112static av_cold int aac_encode_init(AVCodecContext *avctx)
113{
114 AACContext *s = avctx->priv_data;
115 int ret = AVERROR(EINVAL);
116 AACENC_InfoStruct info = { 0 };
117 CHANNEL_MODE mode;
118 AACENC_ERROR err;
119 int aot = FF_PROFILE_AAC_LOW + 1;
120 int sce = 0, cpe = 0;
121
122 if ((err = aacEncOpen(&s->handle, 0, avctx->channels)) != AACENC_OK) {
123 av_log(avctx, AV_LOG_ERROR, "Unable to open the encoder: %s\n",
124 aac_get_error(err));
125 goto error;
126 }
127
128 if (avctx->profile != FF_PROFILE_UNKNOWN)
129 aot = avctx->profile + 1;
130
131 if ((err = aacEncoder_SetParam(s->handle, AACENC_AOT, aot)) != AACENC_OK) {
132 av_log(avctx, AV_LOG_ERROR, "Unable to set the AOT %d: %s\n",
133 aot, aac_get_error(err));
134 goto error;
135 }
136
137 if (aot == FF_PROFILE_AAC_ELD + 1 && s->eld_sbr) {
138 if ((err = aacEncoder_SetParam(s->handle, AACENC_SBR_MODE,
139 1)) != AACENC_OK) {
140 av_log(avctx, AV_LOG_ERROR, "Unable to enable SBR for ELD: %s\n",
141 aac_get_error(err));
142 goto error;
143 }
144 }
145
146 if ((err = aacEncoder_SetParam(s->handle, AACENC_SAMPLERATE,
147 avctx->sample_rate)) != AACENC_OK) {
148 av_log(avctx, AV_LOG_ERROR, "Unable to set the sample rate %d: %s\n",
149 avctx->sample_rate, aac_get_error(err));
150 goto error;
151 }
152
153 switch (avctx->channels) {
154 case 1: mode = MODE_1; sce = 1; cpe = 0; break;
155 case 2: mode = MODE_2; sce = 0; cpe = 1; break;
156 case 3: mode = MODE_1_2; sce = 1; cpe = 1; break;
157 case 4: mode = MODE_1_2_1; sce = 2; cpe = 1; break;
158 case 5: mode = MODE_1_2_2; sce = 1; cpe = 2; break;
159 case 6: mode = MODE_1_2_2_1; sce = 2; cpe = 2; break;
15fa856b
JF
160/* The version macro is introduced the same time as the 7.1 support, so this
161 should suffice. */
162#ifdef AACENCODER_LIB_VL0
163 case 8:
164 sce = 2;
165 cpe = 3;
166 if (avctx->channel_layout == AV_CH_LAYOUT_7POINT1) {
167 mode = MODE_7_1_REAR_SURROUND;
168 } else {
169 // MODE_1_2_2_2_1 and MODE_7_1_FRONT_CENTER use the same channel layout
170 mode = MODE_7_1_FRONT_CENTER;
171 }
172 break;
173#endif
37eeb5e2
MS
174 default:
175 av_log(avctx, AV_LOG_ERROR,
176 "Unsupported number of channels %d\n", avctx->channels);
177 goto error;
178 }
179
180 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELMODE,
181 mode)) != AACENC_OK) {
182 av_log(avctx, AV_LOG_ERROR,
183 "Unable to set channel mode %d: %s\n", mode, aac_get_error(err));
184 goto error;
185 }
186
187 if ((err = aacEncoder_SetParam(s->handle, AACENC_CHANNELORDER,
188 1)) != AACENC_OK) {
189 av_log(avctx, AV_LOG_ERROR,
190 "Unable to set wav channel order %d: %s\n",
191 mode, aac_get_error(err));
192 goto error;
193 }
194
7c6eb0a1 195 if (avctx->flags & AV_CODEC_FLAG_QSCALE || s->vbr) {
375bff70 196 int mode = s->vbr ? s->vbr : avctx->global_quality;
37eeb5e2
MS
197 if (mode < 1 || mode > 5) {
198 av_log(avctx, AV_LOG_WARNING,
199 "VBR quality %d out of range, should be 1-5\n", mode);
200 mode = av_clip(mode, 1, 5);
201 }
a10190dc
MS
202 av_log(avctx, AV_LOG_WARNING,
203 "Note, the VBR setting is unsupported and only works with "
204 "some parameter combinations\n");
37eeb5e2
MS
205 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATEMODE,
206 mode)) != AACENC_OK) {
207 av_log(avctx, AV_LOG_ERROR, "Unable to set the VBR bitrate mode %d: %s\n",
208 mode, aac_get_error(err));
209 goto error;
210 }
211 } else {
212 if (avctx->bit_rate <= 0) {
213 if (avctx->profile == FF_PROFILE_AAC_HE_V2) {
214 sce = 1;
215 cpe = 0;
216 }
217 avctx->bit_rate = (96*sce + 128*cpe) * avctx->sample_rate / 44;
218 if (avctx->profile == FF_PROFILE_AAC_HE ||
219 avctx->profile == FF_PROFILE_AAC_HE_V2 ||
95d52464 220 avctx->profile == FF_PROFILE_MPEG2_AAC_HE ||
37eeb5e2
MS
221 s->eld_sbr)
222 avctx->bit_rate /= 2;
223 }
224 if ((err = aacEncoder_SetParam(s->handle, AACENC_BITRATE,
225 avctx->bit_rate)) != AACENC_OK) {
226 av_log(avctx, AV_LOG_ERROR, "Unable to set the bitrate %d: %s\n",
227 avctx->bit_rate, aac_get_error(err));
228 goto error;
229 }
230 }
231
232 /* Choose bitstream format - if global header is requested, use
233 * raw access units, otherwise use ADTS. */
234 if ((err = aacEncoder_SetParam(s->handle, AACENC_TRANSMUX,
b9302677
MS
235 avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? TT_MP4_RAW :
236 s->latm ? TT_MP4_LOAS : TT_MP4_ADTS)) != AACENC_OK) {
37eeb5e2
MS
237 av_log(avctx, AV_LOG_ERROR, "Unable to set the transmux format: %s\n",
238 aac_get_error(err));
239 goto error;
240 }
241
160a27c5
KK
242 if (s->latm && s->header_period) {
243 if ((err = aacEncoder_SetParam(s->handle, AACENC_HEADER_PERIOD,
244 s->header_period)) != AACENC_OK) {
245 av_log(avctx, AV_LOG_ERROR, "Unable to set header period: %s\n",
246 aac_get_error(err));
247 goto error;
248 }
249 }
250
37eeb5e2
MS
251 /* If no signaling mode is chosen, use explicit hierarchical signaling
252 * if using mp4 mode (raw access units, with global header) and
253 * implicit signaling if using ADTS. */
254 if (s->signaling < 0)
7c6eb0a1 255 s->signaling = avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER ? 2 : 0;
37eeb5e2
MS
256
257 if ((err = aacEncoder_SetParam(s->handle, AACENC_SIGNALING_MODE,
258 s->signaling)) != AACENC_OK) {
259 av_log(avctx, AV_LOG_ERROR, "Unable to set signaling mode %d: %s\n",
260 s->signaling, aac_get_error(err));
261 goto error;
262 }
263
264 if ((err = aacEncoder_SetParam(s->handle, AACENC_AFTERBURNER,
265 s->afterburner)) != AACENC_OK) {
266 av_log(avctx, AV_LOG_ERROR, "Unable to set afterburner to %d: %s\n",
267 s->afterburner, aac_get_error(err));
268 goto error;
269 }
270
7548831c 271 if (avctx->cutoff > 0) {
375ef652 272 if (avctx->cutoff < (avctx->sample_rate + 255) >> 8 || avctx->cutoff > 20000) {
c37c383e
MA
273 av_log(avctx, AV_LOG_ERROR, "cutoff valid range is %d-20000\n",
274 (avctx->sample_rate + 255) >> 8);
275 goto error;
276 }
7548831c
MS
277 if ((err = aacEncoder_SetParam(s->handle, AACENC_BANDWIDTH,
278 avctx->cutoff)) != AACENC_OK) {
511cf612 279 av_log(avctx, AV_LOG_ERROR, "Unable to set the encoder bandwidth to %d: %s\n",
7548831c
MS
280 avctx->cutoff, aac_get_error(err));
281 goto error;
282 }
283 }
284
37eeb5e2
MS
285 if ((err = aacEncEncode(s->handle, NULL, NULL, NULL, NULL)) != AACENC_OK) {
286 av_log(avctx, AV_LOG_ERROR, "Unable to initialize the encoder: %s\n",
287 aac_get_error(err));
288 return AVERROR(EINVAL);
289 }
290
291 if ((err = aacEncInfo(s->handle, &info)) != AACENC_OK) {
292 av_log(avctx, AV_LOG_ERROR, "Unable to get encoder info: %s\n",
293 aac_get_error(err));
294 goto error;
295 }
296
37eeb5e2 297 avctx->frame_size = info.frameLength;
141c960e
MS
298#if FDKENC_VER_AT_LEAST(4, 0)
299 avctx->initial_padding = info.nDelay;
300#else
2df0c32e 301 avctx->initial_padding = info.encoderDelay;
141c960e 302#endif
37eeb5e2
MS
303 ff_af_queue_init(avctx, &s->afq);
304
7c6eb0a1 305 if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
37eeb5e2
MS
306 avctx->extradata_size = info.confSize;
307 avctx->extradata = av_mallocz(avctx->extradata_size +
059a9348 308 AV_INPUT_BUFFER_PADDING_SIZE);
37eeb5e2
MS
309 if (!avctx->extradata) {
310 ret = AVERROR(ENOMEM);
311 goto error;
312 }
313
314 memcpy(avctx->extradata, info.confBuf, info.confSize);
315 }
316 return 0;
317error:
318 aac_encode_close(avctx);
319 return ret;
320}
321
322static int aac_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
323 const AVFrame *frame, int *got_packet_ptr)
324{
325 AACContext *s = avctx->priv_data;
326 AACENC_BufDesc in_buf = { 0 }, out_buf = { 0 };
327 AACENC_InArgs in_args = { 0 };
328 AACENC_OutArgs out_args = { 0 };
329 int in_buffer_identifier = IN_AUDIO_DATA;
330 int in_buffer_size, in_buffer_element_size;
331 int out_buffer_identifier = OUT_BITSTREAM_DATA;
332 int out_buffer_size, out_buffer_element_size;
333 void *in_ptr, *out_ptr;
334 int ret;
141c960e 335 uint8_t dummy_buf[1];
37eeb5e2
MS
336 AACENC_ERROR err;
337
338 /* handle end-of-stream small frame and flushing */
339 if (!frame) {
141c960e
MS
340 /* Must be a non-null pointer, even if it's a dummy. We could use
341 * the address of anything else on the stack as well. */
342 in_ptr = dummy_buf;
343 in_buffer_size = 0;
344
37eeb5e2
MS
345 in_args.numInSamples = -1;
346 } else {
141c960e
MS
347 in_ptr = frame->data[0];
348 in_buffer_size = 2 * avctx->channels * frame->nb_samples;
37eeb5e2 349
141c960e 350 in_args.numInSamples = avctx->channels * frame->nb_samples;
37eeb5e2
MS
351
352 /* add current frame to the queue */
1d7ffd06 353 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
37eeb5e2
MS
354 return ret;
355 }
356
141c960e
MS
357 in_buffer_element_size = 2;
358 in_buf.numBufs = 1;
359 in_buf.bufs = &in_ptr;
360 in_buf.bufferIdentifiers = &in_buffer_identifier;
361 in_buf.bufSizes = &in_buffer_size;
362 in_buf.bufElSizes = &in_buffer_element_size;
363
37eeb5e2
MS
364 /* The maximum packet size is 6144 bits aka 768 bytes per channel. */
365 if ((ret = ff_alloc_packet(avpkt, FFMAX(8192, 768 * avctx->channels)))) {
366 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
367 return ret;
368 }
369
370 out_ptr = avpkt->data;
371 out_buffer_size = avpkt->size;
372 out_buffer_element_size = 1;
373 out_buf.numBufs = 1;
374 out_buf.bufs = &out_ptr;
375 out_buf.bufferIdentifiers = &out_buffer_identifier;
376 out_buf.bufSizes = &out_buffer_size;
377 out_buf.bufElSizes = &out_buffer_element_size;
378
379 if ((err = aacEncEncode(s->handle, &in_buf, &out_buf, &in_args,
380 &out_args)) != AACENC_OK) {
381 if (!frame && err == AACENC_ENCODE_EOF)
382 return 0;
383 av_log(avctx, AV_LOG_ERROR, "Unable to encode frame: %s\n",
384 aac_get_error(err));
385 return AVERROR(EINVAL);
386 }
387
388 if (!out_args.numOutBytes)
389 return 0;
390
391 /* Get the next frame pts & duration */
392 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
393 &avpkt->duration);
394
395 avpkt->size = out_args.numOutBytes;
396 *got_packet_ptr = 1;
397 return 0;
398}
399
400static const AVProfile profiles[] = {
401 { FF_PROFILE_AAC_LOW, "LC" },
402 { FF_PROFILE_AAC_HE, "HE-AAC" },
403 { FF_PROFILE_AAC_HE_V2, "HE-AACv2" },
404 { FF_PROFILE_AAC_LD, "LD" },
405 { FF_PROFILE_AAC_ELD, "ELD" },
406 { FF_PROFILE_UNKNOWN },
407};
408
409static const AVCodecDefault aac_encode_defaults[] = {
410 { "b", "0" },
411 { NULL }
412};
413
414static const uint64_t aac_channel_layout[] = {
415 AV_CH_LAYOUT_MONO,
416 AV_CH_LAYOUT_STEREO,
417 AV_CH_LAYOUT_SURROUND,
418 AV_CH_LAYOUT_4POINT0,
419 AV_CH_LAYOUT_5POINT0_BACK,
420 AV_CH_LAYOUT_5POINT1_BACK,
15fa856b
JF
421#ifdef AACENCODER_LIB_VL0
422 AV_CH_LAYOUT_7POINT1_WIDE_BACK,
423 AV_CH_LAYOUT_7POINT1,
424#endif
37eeb5e2
MS
425 0,
426};
427
94a69dee
NC
428static const int aac_sample_rates[] = {
429 96000, 88200, 64000, 48000, 44100, 32000,
430 24000, 22050, 16000, 12000, 11025, 8000, 0
431};
432
37eeb5e2 433AVCodec ff_libfdk_aac_encoder = {
c8ba8be2 434 .name = "libfdk_aac",
b2bed932 435 .long_name = NULL_IF_CONFIG_SMALL("Fraunhofer FDK AAC"),
c8ba8be2
NC
436 .type = AVMEDIA_TYPE_AUDIO,
437 .id = AV_CODEC_ID_AAC,
438 .priv_data_size = sizeof(AACContext),
439 .init = aac_encode_init,
440 .encode2 = aac_encode_frame,
441 .close = aac_encode_close,
def97856 442 .capabilities = AV_CODEC_CAP_SMALL_LAST_FRAME | AV_CODEC_CAP_DELAY,
c8ba8be2
NC
443 .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16,
444 AV_SAMPLE_FMT_NONE },
c8ba8be2
NC
445 .priv_class = &aac_enc_class,
446 .defaults = aac_encode_defaults,
447 .profiles = profiles,
94a69dee 448 .supported_samplerates = aac_sample_rates,
c8ba8be2 449 .channel_layouts = aac_channel_layout,
47687a2f 450 .wrapper_name = "libfdk",
37eeb5e2 451};