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

   1 /*\r
   2 * copyright (c) 2010 Sveriges Television AB <info@casparcg.com>\r
   3 *\r
   4 *  This file is part of CasparCG.\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 */\r
  20 #include "../../stdafx.h"\r
  21 \r
  22 #include "audio_decoder.h"\r
  23 \r
  24 #include <tbb/task_group.h>\r
  25 \r
  26 #if defined(_MSC_VER)\r
  27 #pragma warning (push)\r
  28 #pragma warning (disable : 4244)\r
  29 #endif\r
  30 extern "C" \r
  31 {\r
  32         #define __STDC_CONSTANT_MACROS\r
  33         #define __STDC_LIMIT_MACROS\r
  34         #include <libavformat/avformat.h>\r
  35         #include <libavcodec/avcodec.h>\r
  36 }\r
  37 #if defined(_MSC_VER)\r
  38 #pragma warning (pop)\r
  39 #endif\r
  40 \r
  41 namespace caspar {\r
  42         \r
  43 struct audio_decoder::implementation : boost::noncopyable\r
  44 {       \r
  45         std::shared_ptr<AVCodecContext>                                                         codec_context_;         \r
  46         const core::video_format_desc                                                           format_desc_;\r
  47         int                                                                                                                     index_;\r
  48         std::vector<int16_t, tbb::cache_aligned_allocator<int16_t>>     buffer_;                // avcodec_decode_audio3 needs 4 byte alignment\r
  49         std::vector<int16_t, tbb::cache_aligned_allocator<int16_t>>     audio_samples_;         // avcodec_decode_audio3 needs 4 byte alignment\r
  50         std::queue<std::shared_ptr<AVPacket>>                                           packets_;\r
  51 public:\r
  52         explicit implementation(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
  53                 : format_desc_(format_desc)     \r
  54         {\r
  55                 AVCodec* dec;\r
  56                 index_ = av_find_best_stream(context.get(), AVMEDIA_TYPE_AUDIO, -1, -1, &dec, 0);\r
  57 \r
  58                 int errn = avcodec_open(context->streams[index_]->codec, dec);\r
  59                 if(errn < 0)\r
  60                         return;\r
  61                                 \r
  62                 codec_context_.reset(context->streams[index_]->codec, avcodec_close);\r
  63 \r
  64                 if(codec_context_ &&\r
  65                    (codec_context_->sample_rate != static_cast<int>(format_desc_.audio_sample_rate) || \r
  66                    codec_context_->channels != static_cast<int>(format_desc_.audio_channels)))\r
  67                 {       \r
  68                         BOOST_THROW_EXCEPTION(\r
  69                                 file_read_error()  <<\r
  70                                 msg_info("Invalid sample-rate or number of channels.") <<\r
  71                                 arg_value_info(boost::lexical_cast<std::string>(codec_context_->sample_rate)) << \r
  72                                 arg_name_info("codec_context"));\r
  73                 }               \r
  74         }\r
  75 \r
  76         void push(const std::shared_ptr<AVPacket>& packet)\r
  77         {                       \r
  78                 if(!codec_context_)\r
  79                         return;\r
  80 \r
  81                 if(packet && packet->stream_index != index_)\r
  82                         return;\r
  83 \r
  84                 packets_.push(packet);\r
  85         }       \r
  86         \r
  87         std::vector<std::vector<int16_t>> poll()\r
  88         {\r
  89                 std::vector<std::vector<int16_t>> result;\r
  90 \r
  91                 if(!codec_context_)\r
  92                         result.push_back(std::vector<int16_t>(format_desc_.audio_samples_per_frame, 0));\r
  93                 else if(!packets_.empty())\r
  94                 {\r
  95                         decode(packets_.front());\r
  96                         packets_.pop();\r
  97 \r
  98                         while(audio_samples_.size() > format_desc_.audio_samples_per_frame)\r
  99                         {\r
 100                                 const auto begin = audio_samples_.begin();\r
 101                                 const auto end   = audio_samples_.begin() + format_desc_.audio_samples_per_frame;\r
 102 \r
 103                                 result.push_back(std::vector<int16_t>(begin, end));\r
 104                                 audio_samples_.erase(begin, end);\r
 105                         }\r
 106                 }\r
 107 \r
 108                 return result;\r
 109         }\r
 110 \r
 111         void decode(const std::shared_ptr<AVPacket>& packet)\r
 112         {                                                                                       \r
 113                 if(!packet) // eof\r
 114                 {\r
 115                         auto truncate = audio_samples_.size() % format_desc_.audio_samples_per_frame;\r
 116                         if(truncate > 0)\r
 117                         {\r
 118                                 audio_samples_.resize(audio_samples_.size() - truncate); \r
 119                                 CASPAR_LOG(info) << L"Truncating " << truncate << L" audio-samples."; \r
 120                         }\r
 121                         avcodec_flush_buffers(codec_context_.get());\r
 122                 }\r
 123                 else\r
 124                 {\r
 125                         buffer_.resize(4*format_desc_.audio_sample_rate*2+FF_INPUT_BUFFER_PADDING_SIZE/2, 0);\r
 126 \r
 127                         int written_bytes = buffer_.size() - FF_INPUT_BUFFER_PADDING_SIZE/2;\r
 128                         const int errn = avcodec_decode_audio3(codec_context_.get(), buffer_.data(), &written_bytes, packet.get());\r
 129                         if(errn < 0)\r
 130                         {       \r
 131                                 BOOST_THROW_EXCEPTION(\r
 132                                         invalid_operation() <<\r
 133                                         boost::errinfo_api_function("avcodec_decode_audio2") <<\r
 134                                         boost::errinfo_errno(AVUNERROR(errn)));\r
 135                         }\r
 136 \r
 137                         buffer_.resize(written_bytes/2);\r
 138                         audio_samples_.insert(audio_samples_.end(), buffer_.begin(), buffer_.end());\r
 139                         buffer_.clear();        \r
 140                 }\r
 141         }\r
 142 \r
 143         bool ready() const\r
 144         {\r
 145                 return !codec_context_ || !packets_.empty();\r
 146         }\r
 147 };\r
 148 \r
 149 audio_decoder::audio_decoder(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new implementation(context, format_desc)){}\r
 150 void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}\r
 151 bool audio_decoder::ready() const{return impl_->ready();}\r
 152 std::vector<std::vector<int16_t>> audio_decoder::poll(){return impl_->poll();}\r
 153 }