]> git.sesse.net Git - casparcg/blob - modules/ffmpeg/producer/audio/audio_decoder.cpp
2366c10d8032260be15a9ca11197455926c528ce
[casparcg] / 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         int                                                                                                                     actual_index_;
66         const core::video_format_desc                                                           format_desc_;
67         const spl::shared_ptr<AVCodecContext>                                           codec_context_          = open_codec(input_.context(), AVMEDIA_TYPE_AUDIO, actual_index_, false);
68
69         std::shared_ptr<SwrContext>                                                                     swr_                            {
70                                                                                                                                                                                 swr_alloc_set_opts(
71                                                                                                                                                                                                 nullptr,
72                                                                                                                                                                                                 create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
73                                                                                                                                                                                                 AV_SAMPLE_FMT_S32,
74                                                                                                                                                                                                 format_desc_.audio_sample_rate,
75                                                                                                                                                                                                 create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
76                                                                                                                                                                                                 codec_context_->sample_fmt,
77                                                                                                                                                                                                 codec_context_->sample_rate,
78                                                                                                                                                                                                 0,
79                                                                                                                                                                                                 nullptr),
80                                                                                                                                                                                 [](SwrContext* p){swr_free(&p); }
81                                                                                                                                                                         };
82
83         cache_aligned_vector<uint8_t>                                                           buffer_;
84
85         std::shared_ptr<AVPacket>                                                                       current_packet_;
86         
87 public:
88         explicit impl(
89                         input& in,
90                         const core::video_format_desc& format_desc,
91                         int audio_stream_index)
92                 : input_(in)
93                 , index_(audio_stream_index)
94                 , actual_index_(input_.get_actual_audio_stream_index(index_))
95                 , format_desc_(format_desc)
96                 , buffer_(480000 * 4)
97         {
98                 if(!swr_)
99                         CASPAR_THROW_EXCEPTION(bad_alloc());
100
101                 THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");
102         }
103                 
104         std::shared_ptr<AVFrame> poll()
105         {               
106                 if(!current_packet_ && !input_.try_pop_audio(current_packet_, index_))
107                         return nullptr;
108                 
109                 std::shared_ptr<AVFrame> audio;
110
111                 if(!current_packet_)    
112                 {
113                         avcodec_flush_buffers(codec_context_.get());    
114                 }
115                 else if(!current_packet_->data)
116                 {
117                         if(codec_context_->codec->capabilities & CODEC_CAP_DELAY)                       
118                                 audio = decode(*current_packet_);
119                         
120                         if(!audio)
121                                 current_packet_.reset();
122                 }
123                 else
124                 {
125                         audio = decode(*current_packet_);
126                         
127                         if(current_packet_->size == 0)
128                                 current_packet_.reset();
129                 }
130         
131                 return audio;
132         }
133
134         std::shared_ptr<AVFrame> decode(AVPacket& pkt)
135         {               
136                 auto frame = create_frame();
137                 
138                 int got_frame = 0;
139                 auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), frame.get(), &got_frame, &pkt), "[audio_decoder]");
140                                         
141                 if(len == 0)
142                 {
143                         pkt.size = 0;
144                         return nullptr;
145                 }
146
147         pkt.data += len;
148         pkt.size -= len;
149
150                 if(!got_frame)
151                         return nullptr;
152                                                         
153                 const uint8_t **in      = const_cast<const uint8_t**>(frame->extended_data);
154                 uint8_t* out[]          = {buffer_.data()};
155
156                 auto channel_samples = swr_convert(swr_.get(), 
157                                                                                         out, static_cast<int>(buffer_.size()) / codec_context_->channels / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32),
158                                                                                         in, frame->nb_samples);
159
160                 frame->data[0]          = buffer_.data();
161                 frame->linesize[0]      = channel_samples * codec_context_->channels * av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);
162                 frame->nb_samples       = channel_samples;
163                 frame->format           = AV_SAMPLE_FMT_S32;
164
165                 monitor_subject_  << core::monitor::message("/file/audio/sample-rate")  % codec_context_->sample_rate
166                                                 << core::monitor::message("/file/audio/channels")       % codec_context_->channels
167                                                 << core::monitor::message("/file/audio/format")         % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))
168                                                 << core::monitor::message("/file/audio/codec")          % u8(codec_context_->codec->long_name);                 
169
170                 return frame;
171         }
172         
173         uint32_t nb_frames() const
174         {
175                 return 0;
176         }
177
178         std::wstring print() const
179         {               
180                 return L"[audio-decoder] " + u16(codec_context_->codec->long_name);
181         }
182 };
183
184 audio_decoder::audio_decoder(input& input, const core::video_format_desc& format_desc, int audio_stream_index) : impl_(new impl(input, format_desc, audio_stream_index)){}
185 audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}
186 audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}
187 std::shared_ptr<AVFrame> audio_decoder::operator()(){return impl_->poll();}
188 uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}
189 std::wstring audio_decoder::print() const{return impl_->print();}
190 core::monitor::subject& audio_decoder::monitor_output() { return impl_->monitor_subject_;}
191
192 }}