/** @file * @author Edouard DUPIN * @copyright 2016, Edouard DUPIN, all right reserved * @license GPL v3 (see license file) */ #include #include #include #include #include #include #include #include #include static void init_ffmpeg() { static bool isInit = false; if (isInit == false) { isInit = true; av_register_all(); } } static void unPlanar(void* _bufferOut, const void* _bufferIn, int32_t _nbSample, audio::format _format, int32_t _channelId, int32_t _nbChannel) { switch(_format) { case audio::format_int8: { const uint8_t* in = reinterpret_cast(_bufferIn); uint8_t* out = reinterpret_cast(_bufferOut); for (int32_t sss=0; sss<_nbSample; ++sss) { out[sss*_nbChannel + _channelId] = in[sss]; } return; } case audio::format_int16: { const int16_t* in = reinterpret_cast(_bufferIn); int16_t* out = reinterpret_cast(_bufferOut); for (int32_t sss=0; sss<_nbSample; ++sss) { out[sss*_nbChannel + _channelId] = in[sss]; } return; } case audio::format_int32: { const int32_t* in = reinterpret_cast(_bufferIn); int32_t* out = reinterpret_cast(_bufferOut); for (int32_t sss=0; sss<_nbSample; ++sss) { out[sss*_nbChannel + _channelId] = in[sss]; } return; } case audio::format_float: { const float* in = reinterpret_cast(_bufferIn); float* out = reinterpret_cast(_bufferOut); for (int32_t sss=0; sss<_nbSample; ++sss) { out[sss*_nbChannel + _channelId] = in[sss]; } return; } case audio::format_double: { const double* in = reinterpret_cast(_bufferIn); double* out = reinterpret_cast(_bufferOut); for (int32_t sss=0; sss<_nbSample; ++sss) { out[sss*_nbChannel + _channelId] = in[sss]; } return; } default: break; } } /** * @brief get the next power 2 if the input * @param[in] value Value that we want the next power of 2 * @return result value */ static int32_t nextP2(int32_t _value) { int32_t val=1; for (int32_t iii=1; iii<31; iii++) { if (_value <= val) { return val; } val *=2; } EWOL_CRITICAL("impossible CASE...."); return val; } void appl::MessageElementVideo::setSize(const ivec2& _size) { if (m_imagerealSize != _size) { // Resize the buffer: m_imagerealSize = _size; m_image.resize(ivec2(nextP2(_size.x()), nextP2(_size.y()))); m_lineSize = m_image.getSize().x() * 3; // 3 is for RGBA //m_image.getSize(); } } void appl::MessageElementAudio::configure(audio::format _format, uint32_t _sampleRate, int32_t _nbChannel, int32_t _nbSample) { // resize the buffer: m_buffer.resize(_nbSample*_nbChannel*audio::getFormatBytes(_format)); m_format = _format; m_sampleRate = _sampleRate; m_map.resize(_nbChannel); switch(_nbChannel) { case 1: m_map[0] = audio::channel_frontCenter; break; case 2: m_map[0] = audio::channel_frontLeft; m_map[1] = audio::channel_frontRight; break; case 3: m_map[0] = audio::channel_frontLeft; m_map[1] = audio::channel_lfe; m_map[2] = audio::channel_frontRight; break; case 4: m_map[0] = audio::channel_frontLeft; m_map[1] = audio::channel_frontRight; m_map[2] = audio::channel_rearLeft; m_map[3] = audio::channel_rearRight; break; default: APPL_ERROR("not supportef nbChannel" << _nbChannel); } } appl::MediaDecoder::MediaDecoder() { init_ffmpeg(); m_formatContext = nullptr; m_videoDecoderContext = nullptr; m_audioDecoderContext = nullptr; m_size = ivec2(0,0); m_videoStream = nullptr; m_audioStream = nullptr; m_videoStream_idx = -1; m_audioStream_idx = -1; m_frame = nullptr; m_videoFrameCount = 0; m_audioFrameCount = 0; m_seek = -1; // output format convertion: m_convertContext = nullptr; m_audioPresent = false; m_audioFormat = audio::format_unknow; m_isInit = false; m_stopRequested = false; } appl::MediaDecoder::~MediaDecoder() { uninit(); } int appl::MediaDecoder::decode_packet(int *_gotFrame, int _cached) { int ret = 0; int decoded = m_packet.size; *_gotFrame = 0; if (m_packet.stream_index == m_videoStream_idx) { // decode video frame ret = avcodec_decode_video2(m_videoDecoderContext, m_frame, _gotFrame, &m_packet); if (ret < 0) { APPL_ERROR("Error decoding video frame (" << av_err2str(ret) << ")"); return ret; } if (*_gotFrame) { if ( m_frame->width != m_size.x() || m_frame->height != m_size.y() || m_frame->format != m_pixelFormat) { // To handle this change, one could call av_image_alloc again and decode the following frames into another rawvideo file. APPL_ERROR("Width, height and pixel format have to be constant in a rawvideo file, but the width, height or pixel format of the input video changed:"); APPL_ERROR("old: size=" << m_size << " format=" << av_get_pix_fmt_name(m_pixelFormat)); APPL_ERROR("new: size=" << ivec2(m_frame->width,m_frame->height) << " format=" << av_get_pix_fmt_name((enum AVPixelFormat)m_frame->format)); return -1; } if (m_updateVideoTimeStampAfterSeek == true) { m_currentVideoTime = m_currentAudioTime; m_updateVideoTimeStampAfterSeek = false; m_seekApply = m_currentVideoTime; // => ready to display } echrono::Duration packetTime(double(m_frame->pkt_pts) * double(m_videoDecoderContext->time_base.num) / double(m_videoDecoderContext->time_base.den)); APPL_INFO("video_frame " << (_cached?"(cached)":"") << " n=" << m_videoFrameCount << " coded_n=" << m_frame->coded_picture_number << " pts=" << av_ts2timestr(m_frame->pkt_pts, &m_videoDecoderContext->time_base) << " " << packetTime); m_videoFrameCount++; int32_t slotId = videoGetEmptySlot(); if (slotId == -1) { APPL_ERROR("an error occured when getting an empty slot for video"); } else { m_videoPool[slotId].setSize(ivec2(m_frame->width, m_frame->height)); uint8_t* dataPointer = (uint8_t*)(m_videoPool[slotId].m_image.getTextureDataPointer()); // Convert Image in RGB: sws_scale(m_convertContext, (const uint8_t **)(m_frame->data), m_frame->linesize, 0, m_frame->height, &dataPointer, &m_videoPool[slotId].m_lineSize); m_videoPool[slotId].m_id = m_videoFrameCount; m_videoPool[slotId].m_time = m_currentVideoTime; m_videoPool[slotId].m_duration = echrono::Duration(0, 1000000000.0/float(getFps(m_videoDecoderContext))); m_currentVideoTime += m_videoPool[slotId].m_duration; m_videoPool[slotId].m_isUsed = true; } } } else if (m_packet.stream_index == m_audioStream_idx) { // decode audio frame ret = avcodec_decode_audio4(m_audioDecoderContext, m_frame, _gotFrame, &m_packet); if (ret < 0) { APPL_ERROR("Error decoding audio frame (" << av_err2str(ret) << ")"); return ret; } // Some audio decoders decode only part of the packet, and have to be called again with the remainder of the packet data. decoded = FFMIN(ret, m_packet.size); if (*_gotFrame) { echrono::Duration packetTime(double(m_frame->pkt_pts) * double(m_audioDecoderContext->time_base.num) / double(m_audioDecoderContext->time_base.den)); if (m_updateVideoTimeStampAfterSeek == true) { // seek specific usecase ==> drop frame to have fast display m_currentAudioTime = packetTime; } else { APPL_INFO("audio_frame " << (_cached?"(cached)":"") << " n=" << m_audioFrameCount << " nb_samples=" << m_frame->nb_samples << " pts=" << packetTime); m_audioFrameCount++; int32_t slotId = audioGetEmptySlot(); if (slotId == -1) { APPL_ERROR("an error occured when getting an empty slot for audio"); } else { //m_frame->channel_layout audio::format format = audio::format_unknow; switch(av_get_packed_sample_fmt((enum AVSampleFormat)m_frame->format)) { case AV_SAMPLE_FMT_U8: format = audio::format_int8; break; case AV_SAMPLE_FMT_S16: format = audio::format_int16; break; case AV_SAMPLE_FMT_S32: format = audio::format_int32; break; case AV_SAMPLE_FMT_FLT: format = audio::format_float; break; case AV_SAMPLE_FMT_DBL: format = audio::format_double; break; default: break; } if (format == audio::format_unknow) { APPL_ERROR("Unsupported audio format :" << m_frame->format << " ..."); } else { // configure Message: m_audioPool[slotId].configure(format, m_frame->sample_rate, m_frame->channels, m_frame->nb_samples); if (av_sample_fmt_is_planar((enum AVSampleFormat)m_frame->format) == 1) { for (int32_t ccc=0; cccchannels; ++ccc) { unPlanar(&m_audioPool[slotId].m_buffer[0], m_frame->extended_data[ccc], m_frame->nb_samples, m_audioPool[slotId].m_format, ccc, m_frame->channels); } } else { // inject data in the buffer: memcpy(&m_audioPool[slotId].m_buffer[0], m_frame->extended_data[0], m_audioPool[slotId].m_buffer.size()); } // We use the Time of the packet ==> better synchronisation when seeking m_currentAudioTime = packetTime; m_audioPool[slotId].m_id = m_audioFrameCount; m_audioPool[slotId].m_time = m_currentAudioTime; m_audioPool[slotId].m_duration = echrono::Duration(0,(1000000000.0*m_frame->nb_samples)/float(m_frame->sample_rate)); m_currentAudioTime += m_audioPool[slotId].m_duration; m_audioPool[slotId].m_isUsed = true; } } } } } return decoded; } int appl::MediaDecoder::open_codec_context(int *_streamId, AVFormatContext *_formatContext, enum AVMediaType _type) { int ret = 0; int stream_index = 0; AVStream *st = nullptr; AVCodecContext *dec_ctx = nullptr; AVCodec *dec = nullptr; AVDictionary *opts = nullptr; ret = av_find_best_stream(_formatContext, _type, -1, -1, nullptr, 0); if (ret < 0) { APPL_ERROR("Could not find " << av_get_media_type_string(_type) << " stream in input file '" << m_sourceFilename << "'"); return ret; } else { stream_index = ret; st = _formatContext->streams[stream_index]; // find decoder for the stream dec_ctx = st->codec; dec = avcodec_find_decoder(dec_ctx->codec_id); if (!dec) { APPL_ERROR("Failed to find " << av_get_media_type_string(_type) << " codec"); return AVERROR(EINVAL); } // Init the decoders, with or without reference counting av_dict_set(&opts, "refcounted_frames", "0", 0); av_dict_set(&opts, "threads", "auto", 0); if ((ret = avcodec_open2(dec_ctx, dec, &opts)) < 0) { APPL_ERROR("Failed to open " << av_get_media_type_string(_type) << " codec"); return ret; } *_streamId = stream_index; } return 0; } double appl::MediaDecoder::getFps(AVCodecContext *_avctx) { return 1.0 / av_q2d(_avctx->time_base) / FFMAX(_avctx->ticks_per_frame, 1); } void appl::MediaDecoder::init(ememory::SharedPtr _property, uint32_t _mediaId) { // TODO : Correct this later ... We first download the media and after we play it // TODO : We need to down load only a small part ... // get the requested node: if (_property == nullptr) { APPL_ERROR("Request play of not handle property ==> nullptr"); return; } if (_property->connection.isAlive() == false) { APPL_ERROR("Request play of not connected handle ==> 'not alive'"); return; } zeus::service::ProxyVideo remoteServiceVideo = _property->connection.getService("video"); // remove all media (for test) if (remoteServiceVideo.exist() == false) { APPL_ERROR("Vide servie is ==> 'not alive'"); return; } // TODO : Do it better ... zeus::ProxyFile dataFile = remoteServiceVideo.mediaGet(_mediaId).wait().get(); std::string mimeType = dataFile.getMineType().wait().get(); // create temporary file: etk::FSNode tmpFile("CACHE:videoPlayer." + zeus::getExtention(mimeType)); APPL_WARNING("Store in tmpFile : " << tmpFile << " ==> " << tmpFile.getName()); std::string sha512String = zeus::storeInFile(dataFile, tmpFile.getName()); APPL_WARNING("Store in tmpFile : " << tmpFile << " ==> " << tmpFile.getFileSystemName() << " DONE"); init(tmpFile.getFileSystemName()); // TODO : init(tmpFile); } void appl::MediaDecoder::init(const std::string& _filename) { m_updateVideoTimeStampAfterSeek = false; m_sourceFilename = _filename; ethread::setName("ffmpegThread"); // open input file, and allocate format context if (avformat_open_input(&m_formatContext, m_sourceFilename.c_str(), nullptr, nullptr) < 0) { APPL_ERROR("Could not open source file " << m_sourceFilename); exit(1); } // retrieve stream information if (avformat_find_stream_info(m_formatContext, nullptr) < 0) { APPL_ERROR("Could not find stream information"); // TODO : check this, this will create a memeory leak return;; } m_duration = echrono::Duration(double(m_formatContext->duration)/double(AV_TIME_BASE)); APPL_INFO("Stream duration : " << m_duration); // Open Video decoder: if (open_codec_context(&m_videoStream_idx, m_formatContext, AVMEDIA_TYPE_VIDEO) >= 0) { m_videoStream = m_formatContext->streams[m_videoStream_idx]; m_videoDecoderContext = m_videoStream->codec; // allocate image where the decoded image will be put m_size.setValue(m_videoDecoderContext->width, m_videoDecoderContext->height); m_pixelFormat = m_videoDecoderContext->pix_fmt; m_videoPool.resize(10); APPL_INFO("Open video stream with property: size=" << m_size << " pixel format=" << av_get_pix_fmt_name(m_pixelFormat) << " fps=" << getFps(m_videoDecoderContext) << " tick/frame=" << m_videoDecoderContext->ticks_per_frame); // convert the image format: m_convertContext = sws_getContext(m_size.x(), m_size.y(), m_pixelFormat, m_size.x(), m_size.y(), AV_PIX_FMT_RGB24, 0, 0, 0, 0); } // Open Audio Decoder: if (open_codec_context(&m_audioStream_idx, m_formatContext, AVMEDIA_TYPE_AUDIO) >= 0) { m_audioPresent = true; m_audioStream = m_formatContext->streams[m_audioStream_idx]; m_audioDecoderContext = m_audioStream->codec; m_audioPool.resize(80); // Number of channels: m_audioDecoderContext->channels // Framerate: m_audioDecoderContext->sample_rate APPL_INFO("Open audio stream with audio property: " << int32_t(m_audioDecoderContext->channels) << " channel(s) & samplerate=" << m_audioDecoderContext->sample_rate << " Hz"); //m_frame->channel_layout m_audioSampleRate = m_audioDecoderContext->sample_rate; m_audioFormat = audio::format_unknow; switch(av_get_packed_sample_fmt(m_audioDecoderContext->sample_fmt)) { case AV_SAMPLE_FMT_U8: m_audioFormat = audio::format_int8; break; case AV_SAMPLE_FMT_S16: m_audioFormat = audio::format_int16; break; case AV_SAMPLE_FMT_S32: m_audioFormat = audio::format_int32; break; case AV_SAMPLE_FMT_FLT: m_audioFormat = audio::format_float; break; case AV_SAMPLE_FMT_DBL: m_audioFormat = audio::format_double; break; default: m_audioFormat = audio::format_unknow; break; } // TODO : Do it better: m_audioMap.resize(m_audioDecoderContext->channels); switch(m_audioDecoderContext->channels) { case 1: m_audioMap[0] = audio::channel_frontCenter; break; case 2: m_audioMap[0] = audio::channel_frontLeft; m_audioMap[1] = audio::channel_frontRight; break; case 3: m_audioMap[0] = audio::channel_frontLeft; m_audioMap[1] = audio::channel_lfe; m_audioMap[2] = audio::channel_frontRight; break; case 4: m_audioMap[0] = audio::channel_frontLeft; m_audioMap[1] = audio::channel_frontRight; m_audioMap[2] = audio::channel_rearLeft; m_audioMap[3] = audio::channel_rearRight; break; default: APPL_ERROR("not supportef nbChannel " << m_audioDecoderContext->channels); } APPL_PRINT("Audio configuration : " << m_audioMap << " " << m_audioFormat << " sampleRate=" <= 0) { AVPacket orig_pkt = m_packet; do { ret = decode_packet(&gotFrame, 0); if (ret < 0) { break; } m_packet.data += ret; m_packet.size -= ret; } while (m_packet.size > 0); av_packet_unref(&orig_pkt); } return (ret < 0); } void appl::MediaDecoder::uninit() { if (m_isInit == false) { return; } APPL_PRINT("Demuxing & Decoding succeeded..."); avcodec_close(m_videoDecoderContext); avcodec_close(m_audioDecoderContext); avformat_close_input(&m_formatContext); av_frame_free(&m_frame); m_isInit = false; } void appl::MediaDecoder::stop() { m_stopRequested = true; } void appl::MediaDecoder::flushMessage() { // flush all decoders ... avcodec_flush_buffers(m_audioStream->codec); avcodec_flush_buffers(m_videoStream->codec); // TODO : Protect this ... // Disable use of all buffer for (int32_t iii=0; iii=0) { id = m_audioStream_idx; tmpPos = m_currentAudioTime; } else if (m_videoStream_idx>=0) { id = m_videoStream_idx; tmpPos = m_currentVideoTime; } int64_t seekTarget = av_rescale_q(seekPos, AV_TIME_BASE_Q, m_formatContext->streams[id]->time_base); APPL_INFO("request seek at: " << seekPos << " and get position: " << seekTarget); int flags = _time < tmpPos ? AVSEEK_FLAG_BACKWARD : 0; // AVSEEK_FLAG_ANY if (av_seek_frame(m_formatContext, id, seekTarget, flags) < 0) { APPL_ERROR(" Unable to seek"); return; } m_currentVideoTime = _time; m_currentAudioTime = _time; m_updateVideoTimeStampAfterSeek = true; APPL_INFO("Request seeking : " << _time << " done"); } int32_t appl::MediaDecoder::videoGetEmptySlot() { for (int32_t iii=0; iii m_videoPool[iii].m_id) { smallerId = m_videoPool[iii].m_id; findId = iii; } } return findId; } int32_t appl::MediaDecoder::audioGetOlderSlot() { int32_t smallerId = 0x7FFFFFFF; int32_t findId = -1; for (int32_t iii=0; iii m_audioPool[iii].m_id) { smallerId = m_audioPool[iii].m_id; findId = iii; } } return findId; }