bool AudioEncoderFFmpeg::encode(const AudioFrame &frame)
{
DPTR_D(AudioEncoderFFmpeg);
AVFrame *f = NULL;
if (frame.isValid()) {
f = av_frame_alloc();
const AudioFormat fmt(frame.format());
f->format = fmt.sampleFormatFFmpeg();
f->channel_layout = fmt.channelLayoutFFmpeg();
// f->channels = fmt.channels(); //remove? not availale in libav9
// must be (not the last frame) exactly frame_size unless CODEC_CAP_VARIABLE_FRAME_SIZE is set (frame_size==0)
// TODO: mpv use pcmhack for avctx.frame_size==0. can we use input frame.samplesPerChannel?
f->nb_samples = d.frame_size;
/// f->quality = d.avctx->global_quality; //TODO
// TODO: record last pts. mpv compute pts internally and also use playback time
f->pts = int64_t(frame.timestamp()*fmt.sampleRate()); // TODO
// pts is set in muxer
const int nb_planes = frame.planeCount();
// bytes between 2 samples on a plane. TODO: add to AudioFormat? what about bytesPerFrame?
const int sample_stride = fmt.isPlanar() ? fmt.bytesPerSample() : fmt.bytesPerSample()*fmt.channels();
for (int i = 0; i < nb_planes; ++i) {
f->linesize[i] = f->nb_samples * sample_stride;// frame.bytesPerLine(i); //
f->extended_data[i] = (uint8_t*)frame.constBits(i);
}
}
AVPacket pkt;
av_init_packet(&pkt);
pkt.data = (uint8_t*)d.buffer.constData();
pkt.size = d.buffer.size();
int got_packet = 0;
int ret = avcodec_encode_audio2(d.avctx, &pkt, f, &got_packet);
av_frame_free(&f);
if (ret < 0) {
//qWarning("error avcodec_encode_audio2: %s" ,av_err2str(ret));
return false; //false
}
if (!got_packet) {
qWarning("no packet got");
return false; //false
}
// qDebug("enc avpkt.pts: %lld, dts: %lld.", pkt.pts, pkt.dts);
d.packet = Packet::fromAVPacket(&pkt, av_q2d(d.avctx->time_base));
// qDebug("enc packet.pts: %.3f, dts: %.3f.", d.packet.pts, d.packet.dts);
return true;
}
static SLDataFormat_PCM audioFormatToSL(const AudioFormat &format)
{
SLDataFormat_PCM format_pcm;
format_pcm.formatType = SL_DATAFORMAT_PCM;
format_pcm.numChannels = format.channels();
format_pcm.samplesPerSec = format.sampleRate() * 1000;
format_pcm.bitsPerSample = format.bytesPerSample()*8;
format_pcm.containerSize = format_pcm.bitsPerSample;
// TODO: more layouts
format_pcm.channelMask = format.channels() == 1 ? SL_SPEAKER_FRONT_CENTER : SL_SPEAKER_FRONT_LEFT | SL_SPEAKER_FRONT_RIGHT;
#ifdef SL_BYTEORDER_NATIVE
format_pcm.endianness = SL_BYTEORDER_NATIVE;
#else
union { unsigned short num; char buf[sizeof(unsigned short)]; } endianness;
endianness.num = 1;
format_pcm.endianness = endianness.buf[0] ? SL_BYTEORDER_LITTLEENDIAN : SL_BYTEORDER_BIGENDIAN;
#endif
return format_pcm;
}
static FormatCode getFormatCode(const AudioFormat &format)
{
if (format.sampleFormat == AF_SAMPFMT_FLOAT)
return kFloat;
if (format.sampleFormat == AF_SAMPFMT_DOUBLE)
return kDouble;
if (format.isInteger())
{
switch (format.bytesPerSample(false))
{
case 1: return kInt8;
case 2: return kInt16;
case 3: return kInt24;
case 4: return kInt32;
}
}
/* NOTREACHED */
assert(false);
return kUndefined;
}