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