git.sesse.net Git - casparcg/blob - modules/ffmpeg/producer/audio/audio_decoder.cpp

   1 /*\r
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>\r
   3 *\r
   4 * This file is part of CasparCG (www.casparcg.com).\r
   5 *\r
   6 * CasparCG is free software: you can redistribute it and/or modify\r
   7 * it under the terms of the GNU General Public License as published by\r
   8 * the Free Software Foundation, either version 3 of the License, or\r
   9 * (at your option) any later version.\r
  10 *\r
  11 * CasparCG is distributed in the hope that it will be useful,\r
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of\r
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the\r
  14 * GNU General Public License for more details.\r
  15 *\r
  16 * You should have received a copy of the GNU General Public License\r
  17 * along with CasparCG. If not, see <http://www.gnu.org/licenses/>.\r
  18 *\r
  19 * Author: Robert Nagy, ronag89@gmail.com\r
  20 */\r
  21 \r
  22 #include "../../stdafx.h"\r
  23 \r
  24 #include "audio_decoder.h"\r
  25 \r
  26 #include "audio_resampler.h"\r
  27 \r
  28 #include "../util/util.h"\r
  29 #include "../../ffmpeg_error.h"\r
  30 \r
  31 #include <core/video_format.h>\r
  32 \r
  33 #include <common/log/log.h>\r
  34 \r
  35 #include <tbb/cache_aligned_allocator.h>\r
  36 \r
  37 #include <queue>\r
  38 \r
  39 #if defined(_MSC_VER)\r
  40 #pragma warning (push)\r
  41 #pragma warning (disable : 4244)\r
  42 #endif\r
  43 extern "C" \r
  44 {\r
  45         #include <libavformat/avformat.h>\r
  46         #include <libavcodec/avcodec.h>\r
  47 }\r
  48 #if defined(_MSC_VER)\r
  49 #pragma warning (pop)\r
  50 #endif\r
  51 \r
  52 namespace caspar { namespace ffmpeg {\r
  53         \r
  54 struct audio_decoder::implementation : boost::noncopyable\r
  55 {       \r
  56         int                                                                                                                     index_;\r
  57         const safe_ptr<AVCodecContext>                                                          codec_context_;         \r
  58         const core::video_format_desc                                                           format_desc_;\r
  59 \r
  60         audio_resampler                                                                                         resampler_;\r
  61 \r
  62         std::vector<int8_t,  tbb::cache_aligned_allocator<int8_t>>      buffer1_;\r
  63 \r
  64         std::queue<safe_ptr<AVPacket>>                                                          packets_;\r
  65 \r
  66         const int64_t                                                                                           nb_frames_;\r
  67         tbb::atomic<uint32_t>                                                                           file_frame_number_;\r
  68 public:\r
  69         explicit implementation(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
  70                 : format_desc_(format_desc)     \r
  71                 , codec_context_(open_codec(*context, AVMEDIA_TYPE_AUDIO, index_))\r
  72                 , resampler_(format_desc.audio_channels,        codec_context_->channels,\r
  73                                          format_desc.audio_sample_rate, codec_context_->sample_rate,\r
  74                                          AV_SAMPLE_FMT_S32,                             codec_context_->sample_fmt)\r
  75                 , buffer1_(AVCODEC_MAX_AUDIO_FRAME_SIZE*2)\r
  76                 , nb_frames_(0)//context->streams[index_]->nb_frames)\r
  77         {               \r
  78                 file_frame_number_ = 0;\r
  79                 CASPAR_LOG(debug) << "[audio_decoder] " << context->streams[index_]->codec->codec->long_name;      \r
  80         }\r
  81 \r
  82         void push(const std::shared_ptr<AVPacket>& packet)\r
  83         {                       \r
  84                 if(!packet)\r
  85                         return;\r
  86 \r
  87                 if(packet->stream_index == index_ || packet->data == nullptr)\r
  88                         packets_.push(make_safe_ptr(packet));\r
  89         }       \r
  90         \r
  91         std::shared_ptr<core::audio_buffer> poll()\r
  92         {\r
  93                 if(packets_.empty())\r
  94                         return nullptr;\r
  95                                 \r
  96                 auto packet = packets_.front();\r
  97 \r
  98                 if(packet->data == nullptr)\r
  99                 {\r
 100                         packets_.pop();\r
 101                         file_frame_number_ = static_cast<uint32_t>(packet->pos);\r
 102                         avcodec_flush_buffers(codec_context_.get());\r
 103                         return flush_audio();\r
 104                 }\r
 105 \r
 106                 auto audio = decode(*packet);\r
 107 \r
 108                 if(packet->size == 0)                                   \r
 109                         packets_.pop();\r
 110 \r
 111                 return audio;\r
 112         }\r
 113 \r
 114         std::shared_ptr<core::audio_buffer> decode(AVPacket& pkt)\r
 115         {               \r
 116                 buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
 117                 int written_bytes = static_cast<int>(buffer1_.size()) - FF_INPUT_BUFFER_PADDING_SIZE;\r
 118                 \r
 119                 int ret = THROW_ON_ERROR2(avcodec_decode_audio3(codec_context_.get(), reinterpret_cast<int16_t*>(buffer1_.data()), &written_bytes, &pkt), "[audio_decoder]");\r
 120 \r
 121                 // There might be several frames in one packet.\r
 122                 pkt.size -= ret;\r
 123                 pkt.data += ret;\r
 124                         \r
 125                 buffer1_.resize(written_bytes);\r
 126 \r
 127                 buffer1_ = resampler_.resample(std::move(buffer1_));\r
 128                 \r
 129                 const auto n_samples = buffer1_.size() / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);\r
 130                 const auto samples = reinterpret_cast<int32_t*>(buffer1_.data());\r
 131 \r
 132                 ++file_frame_number_;\r
 133 \r
 134                 return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
 135         }\r
 136 \r
 137         bool ready() const\r
 138         {\r
 139                 return packets_.size() > 10;\r
 140         }\r
 141 \r
 142         uint32_t nb_frames() const\r
 143         {\r
 144                 return 0;//std::max<int64_t>(nb_frames_, file_frame_number_);\r
 145         }\r
 146 };\r
 147 \r
 148 audio_decoder::audio_decoder(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new implementation(context, format_desc)){}\r
 149 void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}\r
 150 bool audio_decoder::ready() const{return impl_->ready();}\r
 151 std::shared_ptr<core::audio_buffer> audio_decoder::poll(){return impl_->poll();}\r
 152 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
 153 uint32_t audio_decoder::file_frame_number() const{return impl_->file_frame_number_;}\r
 154 \r
 155 }}