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 std::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::queue<spl::shared_ptr<AVPacket>>                                           packets_;\r
  73         \r
  74 public:\r
  75         impl()\r
  76                 : input_(nullptr)\r
  77         {\r
  78         }\r
  79 \r
  80         explicit impl(input& in, const core::video_format_desc& format_desc) \r
  81                 : input_(&in)\r
  82                 , format_desc_(format_desc)     \r
  83                 , codec_context_(open_codec(input_->context(), AVMEDIA_TYPE_AUDIO, index_))\r
  84                 , swr_(swr_alloc_set_opts(nullptr,\r
  85                                                                                 av_get_default_channel_layout(format_desc_.audio_channels), AV_SAMPLE_FMT_S32, format_desc_.audio_sample_rate,\r
  86                                                                                 get_channel_layout(codec_context_.get()), codec_context_->sample_fmt, codec_context_->sample_rate,\r
  87                                                                                 0, nullptr), [](SwrContext* p){swr_free(&p);})\r
  88                 , buffer_(AVCODEC_MAX_AUDIO_FRAME_SIZE*4)\r
  89         {               \r
  90                 if(!swr_)\r
  91                         CASPAR_THROW_EXCEPTION(bad_alloc());\r
  92 \r
  93                 THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");\r
  94         }\r
  95                 \r
  96         std::shared_ptr<core::audio_buffer> poll()\r
  97         {               \r
  98                 auto result = std::make_shared<core::audio_buffer>();\r
  99 \r
 100                 if(!codec_context_)\r
 101                 {\r
 102                         result = empty_audio();\r
 103                         return result;\r
 104                 }\r
 105 \r
 106                 std::shared_ptr<AVPacket> packet;\r
 107                 if(!input_->try_pop_audio(packet))\r
 108                         return result;\r
 109 \r
 110                 if(packet == flush_packet())\r
 111                 {\r
 112                         avcodec_flush_buffers(codec_context_.get());\r
 113                         return result;\r
 114                 }\r
 115                 \r
 116                 if(packet == null_packet())\r
 117                 {\r
 118                         if(codec_context_->codec->capabilities & CODEC_CAP_DELAY)\r
 119                         {\r
 120                                 AVPacket pkt;                \r
 121                                 av_init_packet(&pkt);\r
 122                                 pkt.data = nullptr;\r
 123                                 pkt.size = 0;\r
 124                                 \r
 125                                 core::audio_buffer audio;\r
 126                                 while(decode(pkt, audio))\r
 127                                         boost::range::push_back(*result, audio);\r
 128                         }\r
 129                         return result;\r
 130                 }\r
 131 \r
 132                 while(packet->size > 0)\r
 133                 {\r
 134                         core::audio_buffer audio;\r
 135                         if(decode(*packet, audio))\r
 136                                 boost::range::push_back(*result, audio);                                \r
 137                 }\r
 138                 \r
 139                 event_subject_  << monitor::event("file/audio/sample-rate")     % codec_context_->sample_rate\r
 140                                                 << monitor::event("file/audio/channels")        % codec_context_->channels\r
 141                                                 << monitor::event("file/audio/format")          % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))\r
 142                                                 << monitor::event("file/audio/codec")           % u8(codec_context_->codec->long_name);         \r
 143                 \r
 144                 return result;\r
 145         }\r
 146 \r
 147         bool decode(AVPacket& pkt, core::audio_buffer& result)\r
 148         {               \r
 149                 std::shared_ptr<AVFrame> decoded_frame(avcodec_alloc_frame(), av_free);\r
 150 \r
 151                 int got_frame = 0;\r
 152                 auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), decoded_frame.get(), &got_frame, &pkt), "[audio_decoder]");\r
 153                                         \r
 154                 if(len == 0)\r
 155                 {\r
 156                         pkt.size = 0;\r
 157                         return false;\r
 158                 }\r
 159 \r
 160         pkt.data += len;\r
 161         pkt.size -= len;\r
 162 \r
 163                 if(!got_frame)\r
 164                         return false;\r
 165                                                         \r
 166                 const uint8_t *in[] = {decoded_frame->data[0]};\r
 167                 uint8_t* out[]          = {buffer_.data()};\r
 168 \r
 169                 auto channel_samples = swr_convert(swr_.get(), \r
 170                                                                                         out, static_cast<int>(buffer_.size()) / format_desc_.audio_channels / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32), \r
 171                                                                                         in, decoded_frame->nb_samples);\r
 172                         \r
 173                 auto ptr = reinterpret_cast<int32_t*>(buffer_.data());\r
 174                 result = core::audio_buffer(ptr, ptr + channel_samples * format_desc_.audio_channels);\r
 175                 \r
 176                 return true;\r
 177         }\r
 178         \r
 179         uint32_t nb_frames() const\r
 180         {\r
 181                 return 0;\r
 182         }\r
 183 \r
 184         std::wstring print() const\r
 185         {               \r
 186                 return L"[audio-decoder] " + u16(codec_context_->codec->long_name);\r
 187         }\r
 188 };\r
 189 \r
 190 audio_decoder::audio_decoder() : impl_(new impl()){}\r
 191 audio_decoder::audio_decoder(input& input, const core::video_format_desc& format_desc) : impl_(new impl(input, format_desc)){}\r
 192 audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}\r
 193 audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}\r
 194 std::shared_ptr<core::audio_buffer> audio_decoder::operator()(){return impl_->poll();}\r
 195 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
 196 std::wstring audio_decoder::print() const{return impl_->print();}\r
 197 void audio_decoder::subscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.subscribe(o);}\r
 198 void audio_decoder::unsubscribe(const monitor::observable::observer_ptr& o){impl_->event_subject_.unsubscribe(o);}\r
 199 \r
 200 }}