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

   1 #include "../../../stdafx.h"\r
   2 \r
   3 #include "audio_decoder.h"\r
   4 \r
   5 #if defined(_MSC_VER)\r
   6 #pragma warning (push)\r
   7 #pragma warning (disable : 4244)\r
   8 #endif\r
   9 extern "C" \r
  10 {\r
  11         #define __STDC_CONSTANT_MACROS\r
  12         #define __STDC_LIMIT_MACROS\r
  13         #include <libavformat/avformat.h>\r
  14         #include <libavcodec/avcodec.h>\r
  15 }\r
  16 #if defined(_MSC_VER)\r
  17 #pragma warning (pop)\r
  18 #endif\r
  19 \r
  20 namespace caspar { namespace core { namespace ffmpeg{\r
  21 \r
  22 struct audio_decoder::implementation : boost::noncopyable\r
  23 {\r
  24         typedef std::vector<short, tbb::cache_aligned_allocator<short>> buffer;\r
  25         \r
  26         AVCodecContext* codec_context_;\r
  27 \r
  28         buffer audio_buffer_;   \r
  29         buffer current_chunk_;\r
  30 \r
  31         size_t audio_frame_size_;\r
  32 \r
  33         static const size_t SAMPLE_RATE = 48000;\r
  34         static const size_t N_CHANNELS = 2;\r
  35 \r
  36 public:\r
  37         explicit implementation(AVCodecContext* codec_context, double fps) \r
  38                 : codec_context_(codec_context)\r
  39                 , audio_buffer_(4*SAMPLE_RATE*2+FF_INPUT_BUFFER_PADDING_SIZE/2)\r
  40                 , audio_frame_size_(static_cast<size_t>(static_cast<double>(SAMPLE_RATE) / fps) * N_CHANNELS){}\r
  41                 \r
  42         std::vector<std::vector<short>> execute(const aligned_buffer& audio_packet)\r
  43         {                       \r
  44                 int written_bytes = audio_buffer_.size()*2 - FF_INPUT_BUFFER_PADDING_SIZE;\r
  45                 const int result = avcodec_decode_audio2(codec_context_, audio_buffer_.data(), &written_bytes, audio_packet.data(), audio_packet.size());\r
  46 \r
  47                 if(result <= 0 || codec_context_->sample_rate != SAMPLE_RATE || codec_context_->channels != 2)\r
  48                         return std::vector<std::vector<short>>();\r
  49                                                 \r
  50                 current_chunk_.insert(current_chunk_.end(), audio_buffer_.data(), audio_buffer_.data() + written_bytes/2);\r
  51 \r
  52                 std::vector<std::vector<short>> chunks_;\r
  53                                 \r
  54                 const auto last = current_chunk_.end() - current_chunk_.size() % audio_frame_size_;\r
  55 \r
  56                 for(auto it = current_chunk_.begin(); it != last; it += audio_frame_size_)              \r
  57                         chunks_.push_back(std::vector<short>(it, it + audio_frame_size_));              \r
  58 \r
  59                 current_chunk_.erase(current_chunk_.begin(), last);\r
  60                 \r
  61                 return chunks_;\r
  62         }\r
  63 };\r
  64 \r
  65 audio_decoder::audio_decoder(AVCodecContext* codec_context, double fps) : impl_(new implementation(codec_context, fps)){}\r
  66 std::vector<std::vector<short>> audio_decoder::execute(const aligned_buffer& audio_packet){return impl_->execute(audio_packet);}\r
  67 }}}