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 "../util/util.h"\r
  27 #include "../input/input.h"\r
  28 #include "../../ffmpeg_error.h"\r
  29 \r
  30 #include <core/video_format.h>\r
  31 \r
  32 #include <common/log.h>\r
  33 \r
  34 #include <tbb/cache_aligned_allocator.h>\r
  35 \r
  36 #include <queue>\r
  37 \r
  38 #if defined(_MSC_VER)\r
  39 #pragma warning (push)\r
  40 #pragma warning (disable : 4244)\r
  41 #endif\r
  42 extern "C" \r
  43 {\r
  44         #include <libavformat/avformat.h>\r
  45         #include <libavcodec/avcodec.h>\r
  46         #include <libswresample/swresample.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 uint64_t get_channel_layout(AVCodecContext* dec)\r
  55 {\r
  56         auto layout = (dec->channel_layout && dec->channels == av_get_channel_layout_nb_channels(dec->channel_layout)) ? dec->channel_layout : av_get_default_channel_layout(dec->channels);\r
  57         return layout;\r
  58 }\r
  59 \r
  60 struct audio_decoder::impl : boost::noncopyable\r
  61 {       \r
  62         monitor::basic_subject                                                                          event_subject_;\r
  63         input*                                                                                                          input_;\r
  64         int                                                                                                                     index_;\r
  65         const spl::shared_ptr<AVCodecContext>                                           codec_context_;         \r
  66         const core::video_format_desc                                                           format_desc_;\r
  67 \r
  68         std::shared_ptr<SwrContext>                                                                     swr_;\r
  69 \r
  70         std::vector<uint8_t, tbb::cache_aligned_allocator<int8_t>>      buffer_;\r
  71 \r
  72         std::shared_ptr<AVPacket>                                                                       current_packet_;\r
  73         \r
  74 public:\r
  75         explicit impl(input& in, const core::video_format_desc& format_desc) \r
  76                 : input_(&in)\r
  77                 , format_desc_(format_desc)     \r
  78                 , codec_context_(open_codec(input_->context(), AVMEDIA_TYPE_AUDIO, index_))\r
  79                 , swr_(swr_alloc_set_opts(nullptr,\r
  80                                                                                 av_get_default_channel_layout(format_desc_.audio_channels), AV_SAMPLE_FMT_S32, format_desc_.audio_sample_rate,\r
  81                                                                                 get_channel_layout(codec_context_.get()), codec_context_->sample_fmt, codec_context_->sample_rate,\r
  82                                                                                 0, nullptr), [](SwrContext* p){swr_free(&p);})\r
  83                 , buffer_(AVCODEC_MAX_AUDIO_FRAME_SIZE*4)\r
  84         {               \r
  85                 if(!swr_)\r
  86                         CASPAR_THROW_EXCEPTION(bad_alloc());\r
  87 \r
  88                 THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");\r
  89         }\r
  90                 \r
  91         std::shared_ptr<AVFrame> poll()\r
  92         {               \r
  93                 if(!current_packet_ && !input_->try_pop_audio(current_packet_))\r
  94                         return nullptr;\r
  95                 \r
  96                 std::shared_ptr<AVFrame> audio;\r
  97 \r
  98                 if(!current_packet_)    \r
  99                 {\r
 100                         avcodec_flush_buffers(codec_context_.get());    \r
 101                 }\r
 102                 else if(!current_packet_->data)\r
 103                 {\r
 104                         if(codec_context_->codec->capabilities & CODEC_CAP_DELAY)                       \r
 105                                 audio = decode(*current_packet_);\r
 106                         \r
 107                         if(!audio)\r
 108                                 current_packet_.reset();\r
 109                 }\r
 110                 else\r
 111                 {\r
 112                         audio = decode(*current_packet_);\r
 113                         \r
 114                         if(current_packet_->size == 0)\r
 115                                 current_packet_.reset();\r
 116                 }\r
 117         \r
 118                 return audio ? audio : poll();\r
 119         }\r
 120 \r
 121         std::shared_ptr<AVFrame> decode(AVPacket& pkt)\r
 122         {               \r
 123                 auto frame = create_frame();\r
 124                 \r
 125                 int got_frame = 0;\r
 126                 auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), frame.get(), &got_frame, &pkt), "[audio_decoder]");\r
 127                                         \r
 128                 if(len == 0)\r
 129                 {\r
 130                         pkt.size = 0;\r
 131                         return nullptr;\r
 132                 }\r
 133 \r
 134         pkt.data += len;\r
 135         pkt.size -= len;\r
 136 \r
 137                 if(!got_frame)\r
 138                         return nullptr;\r
 139                                                         \r
 140                 const uint8_t *in[] = {frame->data[0]};\r
 141                 uint8_t* out[]          = {buffer_.data()};\r
 142 \r
 143                 auto channel_samples = swr_convert(swr_.get(), \r
 144                                                                                         out, static_cast<int>(buffer_.size()) / format_desc_.audio_channels / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32), \r
 145                                                                                         in, frame->nb_samples); \r
 146 \r
 147                 frame->data[0]          = buffer_.data();\r
 148                 frame->linesize[0]      = channel_samples * format_desc_.audio_channels * av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);\r
 149                 frame->nb_samples       = channel_samples;\r
 150                 frame->format           = AV_SAMPLE_FMT_S32;\r
 151                                                         \r
 152                 event_subject_  << monitor::event("file/audio/sample-rate")     % codec_context_->sample_rate\r
 153                                                 << monitor::event("file/audio/channels")        % codec_context_->channels\r
 154                                                 << monitor::event("file/audio/format")          % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))\r
 155                                                 << monitor::event("file/audio/codec")           % u8(codec_context_->codec->long_name);                 \r
 156 \r
 157                 return frame;\r
 158         }\r
 159         \r
 160         uint32_t nb_frames() const\r
 161         {\r
 162                 return 0;\r
 163         }\r
 164 \r
 165         std::wstring print() const\r
 166         {               \r
 167                 return L"[audio-decoder] " + u16(codec_context_->codec->long_name);\r
 168         }\r
 169 };\r
 170 \r
 171 audio_decoder::audio_decoder(input& input, const core::video_format_desc& format_desc) : impl_(new impl(input, format_desc)){}\r
 172 audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}\r
 173 audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}\r
 174 std::shared_ptr<AVFrame> audio_decoder::operator()(){return impl_->poll();}\r
 175 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
 176 std::wstring audio_decoder::print() const{return impl_->print();}\r
 177 void audio_decoder::subscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.subscribe(o);}\r
 178 void audio_decoder::unsubscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.unsubscribe(o);}\r
 179 \r
 180 }}