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

   1 /*
   2 * Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>
   3 *
   4 * This file is part of CasparCG (www.casparcg.com).
   5 *
   6 * CasparCG is free software: you can redistribute it and/or modify
   7 * it under the terms of the GNU General Public License as published by
   8 * the Free Software Foundation, either version 3 of the License, or
   9 * (at your option) any later version.
  10 *
  11 * CasparCG is distributed in the hope that it will be useful,
  12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 * GNU General Public License for more details.
  15 *
  16 * You should have received a copy of the GNU General Public License
  17 * along with CasparCG. If not, see <http://www.gnu.org/licenses/>.
  18 *
  19 * Author: Robert Nagy, ronag89@gmail.com
  20 */
  21
  22 #include "../../StdAfx.h"
  23
  24 #include "audio_decoder.h"
  25
  26 #include "../util/util.h"
  27 #include "../input/input.h"
  28 #include "../../ffmpeg_error.h"
  29
  30 #include <core/video_format.h>
  31 #include <core/frame/audio_channel_layout.h>
  32
  33 #include <common/log.h>
  34 #include <common/cache_aligned_vector.h>
  35
  36 #include <queue>
  37
  38 #if defined(_MSC_VER)
  39 #pragma warning (push)
  40 #pragma warning (disable : 4244)
  41 #endif
  42 extern "C"
  43 {
  44         #include <libavformat/avformat.h>
  45         #include <libavcodec/avcodec.h>
  46         #include <libswresample/swresample.h>
  47 }
  48 #if defined(_MSC_VER)
  49 #pragma warning (pop)
  50 #endif
  51
  52 namespace caspar { namespace ffmpeg {
  53
  54 uint64_t get_ffmpeg_channel_layout(AVCodecContext* dec)
  55 {
  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);
  57         return layout;
  58 }
  59
  60 struct audio_decoder::impl : boost::noncopyable
  61 {
  62         core::monitor::subject                                                                          monitor_subject_;
  63         input&                                                                                                          input_;
  64         int                                                                                                                     index_;
  65         const core::video_format_desc                                                           format_desc_;
  66         const spl::shared_ptr<AVCodecContext>                                           codec_context_          = open_codec(input_.context(), AVMEDIA_TYPE_AUDIO, index_, false);
  67
  68         std::shared_ptr<SwrContext>                                                                     swr_                            {
  69                                                                                                                                                                                 swr_alloc_set_opts(
  70                                                                                                                                                                                                 nullptr,
  71                                                                                                                                                                                                 create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
  72                                                                                                                                                                                                 AV_SAMPLE_FMT_S32,
  73                                                                                                                                                                                                 format_desc_.audio_sample_rate,
  74                                                                                                                                                                                                 create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
  75                                                                                                                                                                                                 codec_context_->sample_fmt,
  76                                                                                                                                                                                                 codec_context_->sample_rate,
  77                                                                                                                                                                                                 0,
  78                                                                                                                                                                                                 nullptr),
  79                                                                                                                                                                                 [](SwrContext* p){swr_free(&p); }
  80                                                                                                                                                                         };
  81
  82         cache_aligned_vector<uint8_t>                                                           buffer_;
  83         core::audio_channel_layout                                                                      channel_layout_;
  84
  85         std::shared_ptr<AVPacket>                                                                       current_packet_;
  86
  87 public:
  88         explicit impl(input& in, const core::video_format_desc& format_desc, const std::wstring& channel_layout_spec)
  89                 : input_(in)
  90                 , format_desc_(format_desc)
  91                 , buffer_(480000 * 4)
  92                 , channel_layout_(get_audio_channel_layout(
  93                                 codec_context_->channels,
  94                                 codec_context_->channel_layout,
  95                                 channel_layout_spec))
  96         {
  97                 if(!swr_)
  98                         CASPAR_THROW_EXCEPTION(bad_alloc());
  99
 100                 THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");
 101         }
 102
 103         std::shared_ptr<AVFrame> poll()
 104         {
 105                 if(!current_packet_ && !input_.try_pop_audio(current_packet_))
 106                         return nullptr;
 107
 108                 std::shared_ptr<AVFrame> audio;
 109
 110                 if(!current_packet_)
 111                 {
 112                         avcodec_flush_buffers(codec_context_.get());
 113                 }
 114                 else if(!current_packet_->data)
 115                 {
 116                         if(codec_context_->codec->capabilities & CODEC_CAP_DELAY)
 117                                 audio = decode(*current_packet_);
 118
 119                         if(!audio)
 120                                 current_packet_.reset();
 121                 }
 122                 else
 123                 {
 124                         audio = decode(*current_packet_);
 125
 126                         if(current_packet_->size == 0)
 127                                 current_packet_.reset();
 128                 }
 129
 130                 return audio;
 131         }
 132
 133         std::shared_ptr<AVFrame> decode(AVPacket& pkt)
 134         {
 135                 auto frame = create_frame();
 136
 137                 int got_frame = 0;
 138                 auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), frame.get(), &got_frame, &pkt), "[audio_decoder]");
 139
 140                 if(len == 0)
 141                 {
 142                         pkt.size = 0;
 143                         return nullptr;
 144                 }
 145
 146         pkt.data += len;
 147         pkt.size -= len;
 148
 149                 if(!got_frame)
 150                         return nullptr;
 151
 152                 const uint8_t **in      = const_cast<const uint8_t**>(frame->extended_data);
 153                 uint8_t* out[]          = {buffer_.data()};
 154
 155                 auto channel_samples = swr_convert(swr_.get(),
 156                                                                                         out, static_cast<int>(buffer_.size()) / codec_context_->channels / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32),
 157                                                                                         in, frame->nb_samples);
 158
 159                 frame->data[0]          = buffer_.data();
 160                 frame->linesize[0]      = channel_samples * codec_context_->channels * av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);
 161                 frame->nb_samples       = channel_samples;
 162                 frame->format           = AV_SAMPLE_FMT_S32;
 163
 164                 monitor_subject_  << core::monitor::message("/file/audio/sample-rate")  % codec_context_->sample_rate
 165                                                 << core::monitor::message("/file/audio/channels")       % codec_context_->channels
 166                                                 << core::monitor::message("/file/audio/format")         % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))
 167                                                 << core::monitor::message("/file/audio/codec")          % u8(codec_context_->codec->long_name);
 168
 169                 return frame;
 170         }
 171
 172         uint32_t nb_frames() const
 173         {
 174                 return 0;
 175         }
 176
 177         std::wstring print() const
 178         {
 179                 return L"[audio-decoder] " + u16(codec_context_->codec->long_name);
 180         }
 181 };
 182
 183 audio_decoder::audio_decoder(input& input, const core::video_format_desc& format_desc, const std::wstring& channel_layout_spec) : impl_(new impl(input, format_desc, channel_layout_spec)){}
 184 audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}
 185 audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}
 186 std::shared_ptr<AVFrame> audio_decoder::operator()(){return impl_->poll();}
 187 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}
 188 const core::audio_channel_layout& audio_decoder::channel_layout() const { return impl_->channel_layout_; }
 189 std::wstring audio_decoder::print() const{return impl_->print();}
 190 core::monitor::subject& audio_decoder::monitor_output() { return impl_->monitor_subject_;}
 191
 192 }}