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.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::impl : boost::noncopyable\r
  55 {       \r
  56         monitor::basic_subject                                                                          event_subject_;\r
  57         int                                                                                                                     index_;\r
  58         const std::shared_ptr<AVCodecContext>                                           codec_context_;         \r
  59         const core::video_format_desc                                                           format_desc_;\r
  60 \r
  61         boost::optional<audio_resampler>                                                        resampler_;\r
  62 \r
  63         std::vector<int8_t,  tbb::cache_aligned_allocator<int8_t>>      buffer1_;\r
  64 \r
  65         std::queue<spl::shared_ptr<AVPacket>>                                           packets_;\r
  66 \r
  67         const int64_t                                                                                           nb_frames_;\r
  68         tbb::atomic<uint32_t>                                                                           file_frame_number_;\r
  69 public:\r
  70         explicit impl() \r
  71                 : nb_frames_(0)//context->streams[index_]->nb_frames)\r
  72         {               \r
  73                 file_frame_number_ = 0;   \r
  74         }\r
  75 \r
  76         explicit impl(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
  77                 : format_desc_(format_desc)     \r
  78                 , codec_context_(open_codec(*context, AVMEDIA_TYPE_AUDIO, index_))\r
  79                 , resampler_(audio_resampler(format_desc.audio_channels,        codec_context_->channels,\r
  80                                          format_desc.audio_sample_rate, codec_context_->sample_rate,\r
  81                                          AV_SAMPLE_FMT_S32,                             codec_context_->sample_fmt))\r
  82                 , buffer1_(AVCODEC_MAX_AUDIO_FRAME_SIZE*2)\r
  83                 , nb_frames_(0)//context->streams[index_]->nb_frames)\r
  84         {               \r
  85                 file_frame_number_ = 0;   \r
  86         }\r
  87 \r
  88         void push(const std::shared_ptr<AVPacket>& packet)\r
  89         {                       \r
  90                 if(!packet)\r
  91                         return;\r
  92 \r
  93                 if(packet->stream_index == index_ || packet->data == nullptr)\r
  94                         packets_.push(spl::make_shared_ptr(packet));\r
  95         }       \r
  96         \r
  97         std::shared_ptr<core::audio_buffer> poll()\r
  98         {\r
  99                 if(packets_.empty())\r
 100                         return nullptr;\r
 101                                 \r
 102                 auto packet = packets_.front();\r
 103                 \r
 104                 if(!codec_context_)             \r
 105                 {\r
 106                         packets_.pop();\r
 107                         return packet->data == nullptr ? flush_audio() : empty_audio();\r
 108                 }\r
 109                 else\r
 110                 {\r
 111                         if(packet->data == nullptr)\r
 112                         {\r
 113                                 packets_.pop();\r
 114                                 file_frame_number_ = static_cast<uint32_t>(packet->pos);\r
 115                                 avcodec_flush_buffers(codec_context_.get());\r
 116                                 return flush_audio();\r
 117                         }\r
 118 \r
 119                         auto audio = decode(*packet);\r
 120 \r
 121                         if(packet->size == 0)                                   \r
 122                                 packets_.pop();\r
 123 \r
 124                         return audio;\r
 125                 }\r
 126         }\r
 127 \r
 128         std::shared_ptr<core::audio_buffer> decode(AVPacket& pkt)\r
 129         {               \r
 130                 buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
 131                 int written_bytes = static_cast<int>(buffer1_.size()) - FF_INPUT_BUFFER_PADDING_SIZE;\r
 132                 \r
 133                 int ret = THROW_ON_ERROR2(avcodec_decode_audio3(codec_context_.get(), reinterpret_cast<int16_t*>(buffer1_.data()), &written_bytes, &pkt), "[audio_decoder]");\r
 134 \r
 135                 // There might be several frames in one packet.\r
 136                 pkt.size -= ret;\r
 137                 pkt.data += ret;\r
 138                         \r
 139                 buffer1_.resize(written_bytes);\r
 140 \r
 141                 buffer1_ = resampler_->resample(std::move(buffer1_));\r
 142                 \r
 143                 const auto n_samples = buffer1_.size() / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);\r
 144                 const auto samples = reinterpret_cast<int32_t*>(buffer1_.data());\r
 145                 \r
 146                 event_subject_  << monitor::event("file/audio/sample-rate")     % codec_context_->sample_rate\r
 147                                                 << monitor::event("file/audio/channels")        % codec_context_->channels\r
 148                                                 << monitor::event("file/audio/format")          % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))\r
 149                                                 << monitor::event("file/audio/codec")           % u8(codec_context_->codec->long_name);\r
 150 \r
 151                 ++file_frame_number_;\r
 152 \r
 153                 return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
 154         }\r
 155 \r
 156         bool ready() const\r
 157         {\r
 158                 return !codec_context_ || !packets_.empty();\r
 159         }\r
 160         \r
 161         void clear()\r
 162         {\r
 163                 while(!packets_.empty())\r
 164                         packets_.pop();\r
 165         }\r
 166 \r
 167         uint32_t nb_frames() const\r
 168         {\r
 169                 return 0;//std::max<int64_t>(nb_frames_, file_frame_number_);\r
 170         }\r
 171 \r
 172         std::wstring print() const\r
 173         {               \r
 174                 return L"[audio-decoder] " + u16(codec_context_->codec->long_name);\r
 175         }\r
 176 };\r
 177 \r
 178 audio_decoder::audio_decoder() : impl_(new impl()){}\r
 179 audio_decoder::audio_decoder(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new impl(context, format_desc)){}\r
 180 audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}\r
 181 audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}\r
 182 void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}\r
 183 bool audio_decoder::ready() const{return impl_->ready();}\r
 184 std::shared_ptr<core::audio_buffer> audio_decoder::poll(){return impl_->poll();}\r
 185 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
 186 uint32_t audio_decoder::file_frame_number() const{return impl_->file_frame_number_;}\r
 187 std::wstring audio_decoder::print() const{return impl_->print();}\r
 188 void audio_decoder::clear(){impl_->clear();}\r
 189 void audio_decoder::subscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.subscribe(o);}\r
 190 void audio_decoder::unsubscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.unsubscribe(o);}\r
 191 \r
 192 }}