]> git.sesse.net Git - casparcg/blobdiff - modules/ffmpeg/producer/audio/audio_decoder.cpp
[ffmpeg] Reimplemented support for playing all audio streams in a clip and treating...
[casparcg] / modules / ffmpeg / producer / audio / audio_decoder.cpp
index 20fb263fc0d2702741798d404b27569398c47ad9..0a1008876583afed71cbc24252315e4d5a3e6a48 100644 (file)
-/*\r
-* copyright (c) 2010 Sveriges Television AB <info@casparcg.com>\r
-*\r
-*  This file is part of CasparCG.\r
-*\r
-*    CasparCG is free software: you can redistribute it and/or modify\r
-*    it under the terms of the GNU General Public License as published by\r
-*    the Free Software Foundation, either version 3 of the License, or\r
-*    (at your option) any later version.\r
-*\r
-*    CasparCG is distributed in the hope that it will be useful,\r
-*    but WITHOUT ANY WARRANTY; without even the implied warranty of\r
-*    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the\r
-*    GNU General Public License for more details.\r
-\r
-*    You should have received a copy of the GNU General Public License\r
-*    along with CasparCG.  If not, see <http://www.gnu.org/licenses/>.\r
-*\r
-*/\r
-#include "../../stdafx.h"\r
-\r
-#include "audio_decoder.h"\r
-\r
-#if defined(_MSC_VER)\r
-#pragma warning (push)\r
-#pragma warning (disable : 4244)\r
-#endif\r
-extern "C" \r
-{\r
-       #define __STDC_CONSTANT_MACROS\r
-       #define __STDC_LIMIT_MACROS\r
-       #include <libavformat/avformat.h>\r
-       #include <libavcodec/avcodec.h>\r
-}\r
-#if defined(_MSC_VER)\r
-#pragma warning (pop)\r
-#endif\r
-\r
-namespace caspar {\r
-\r
-struct audio_decoder::implementation : boost::noncopyable\r
-{      \r
-       AVCodecContext& codec_context_;\r
-               \r
-       const core::video_format_desc format_desc_;\r
-\r
-       std::vector<short> current_chunk_;\r
-\r
-       std::vector<std::vector<short>> chunks_;\r
-\r
-public:\r
-       explicit implementation(AVCodecContext& codec_context, const core::video_format_desc& format_desc) \r
-               : codec_context_(codec_context)\r
-               , format_desc_(format_desc)             \r
-       {\r
-               if(codec_context_.sample_rate != static_cast<int>(format_desc_.audio_sample_rate) || \r
-                  codec_context_.channels != static_cast<int>(format_desc_.audio_channels))\r
-               {       \r
-                       BOOST_THROW_EXCEPTION(\r
-                               file_read_error()  <<\r
-                               msg_info("Invalid sample-rate or number of channels.") <<\r
-                               arg_value_info(boost::lexical_cast<std::string>(codec_context_.sample_rate)) << \r
-                               arg_name_info("codec_context"));\r
-               }\r
-       }\r
-               \r
-       void push(std::shared_ptr<AVPacket>&& audio_packet)\r
-       {       \r
-               if(!audio_packet)\r
-                       return;\r
-               \r
-               if(audio_packet->size == 0)\r
-               {\r
-                       avcodec_flush_buffers(&codec_context_);\r
-                       return;\r
-               }\r
-\r
-               auto s = current_chunk_.size();\r
-               current_chunk_.resize(s + 4*format_desc_.audio_sample_rate*2+FF_INPUT_BUFFER_PADDING_SIZE/2);\r
-               \r
-               int written_bytes = (current_chunk_.size() - s)*2 - FF_INPUT_BUFFER_PADDING_SIZE;\r
-               const int errn = avcodec_decode_audio3(&codec_context_, &current_chunk_[s], &written_bytes, audio_packet.get());\r
-               if(errn < 0)\r
-               {       \r
-                       BOOST_THROW_EXCEPTION(\r
-                               invalid_operation() <<\r
-                               boost::errinfo_api_function("avcodec_decode_audio2") <<\r
-                               boost::errinfo_errno(AVUNERROR(errn)));\r
-               }\r
-\r
-               current_chunk_.resize(s + written_bytes/2);\r
-\r
-               const auto last = current_chunk_.end() - current_chunk_.size() % format_desc_.audio_samples_per_frame;\r
-               \r
-               for(auto it = current_chunk_.begin(); it != last; it += format_desc_.audio_samples_per_frame)           \r
-                       chunks_.push_back(std::vector<short>(it, it + format_desc_.audio_samples_per_frame));           \r
-\r
-               current_chunk_.erase(current_chunk_.begin(), last);\r
-       }\r
-\r
-       bool empty() const\r
-       {\r
-               return chunks_.empty();\r
-       }\r
-\r
-       std::vector<short> front()\r
-       {\r
-               return chunks_.front();\r
-       }\r
-\r
-       void pop()\r
-       {\r
-               chunks_.pop_back();\r
-       }\r
-};\r
-\r
-audio_decoder::audio_decoder(AVCodecContext& codec_context, const core::video_format_desc& format_desc) : impl_(new implementation(codec_context, format_desc)){}\r
-void audio_decoder::push(std::shared_ptr<AVPacket>&& audio_packet){impl_->push(std::move(audio_packet));}\r
-bool audio_decoder::empty() const {return impl_->empty();}\r
-std::vector<short> audio_decoder::front() {return impl_->front();}\r
-void audio_decoder::pop(){impl_->pop();}\r
-}
\ No newline at end of file
+/*
+* Copyright 2013 Sveriges Television AB http://casparcg.com/
+*
+* This file is part of CasparCG (www.casparcg.com).
+*
+* CasparCG is free software: you can redistribute it and/or modify
+* it under the terms of the GNU General Public License as published by
+* the Free Software Foundation, either version 3 of the License, or
+* (at your option) any later version.
+*
+* CasparCG is distributed in the hope that it will be useful,
+* but WITHOUT ANY WARRANTY; without even the implied warranty of
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+* GNU General Public License for more details.
+*
+* You should have received a copy of the GNU General Public License
+* along with CasparCG. If not, see <http://www.gnu.org/licenses/>.
+*
+* Author: Robert Nagy, ronag89@gmail.com
+*/
+
+#include "../../StdAfx.h"
+
+#include "audio_decoder.h"
+
+#include "../util/util.h"
+#include "../../ffmpeg_error.h"
+
+#include <core/video_format.h>
+#include <core/mixer/audio/audio_util.h>
+
+#include <common/cache_aligned_vector.h>
+
+#include <queue>
+
+#if defined(_MSC_VER)
+#pragma warning (push)
+#pragma warning (disable : 4244)
+#endif
+extern "C"
+{
+       #include <libavformat/avformat.h>
+       #include <libavcodec/avcodec.h>
+       #include <libswresample/swresample.h>
+}
+#if defined(_MSC_VER)
+#pragma warning (pop)
+#endif
+
+namespace caspar { namespace ffmpeg {
+
+struct audio_decoder::implementation : boost::noncopyable
+{
+       int                                                                             index_;
+       const spl::shared_ptr<AVCodecContext>   codec_context_;
+       const int                                                               out_samplerate_;
+
+       cache_aligned_vector<int32_t>                   buffer_;
+
+       std::queue<spl::shared_ptr<AVPacket>>   packets_;
+
+       std::shared_ptr<SwrContext>                             swr_                            {
+                                                                                                                                       swr_alloc_set_opts(
+                                                                                                                                                       nullptr,
+                                                                                                                                                       codec_context_->channel_layout
+                                                                                                                                                                       ? codec_context_->channel_layout
+                                                                                                                                                                       : av_get_default_channel_layout(codec_context_->channels),
+                                                                                                                                                       AV_SAMPLE_FMT_S32,
+                                                                                                                                                       out_samplerate_,
+                                                                                                                                                       codec_context_->channel_layout
+                                                                                                                                                                       ? codec_context_->channel_layout
+                                                                                                                                                                       : av_get_default_channel_layout(codec_context_->channels),
+                                                                                                                                                       codec_context_->sample_fmt,
+                                                                                                                                                       codec_context_->sample_rate,
+                                                                                                                                                       0,
+                                                                                                                                                       nullptr),
+                                                                                                                                       [](SwrContext* p)
+                                                                                                                                       {
+                                                                                                                                               swr_free(&p);
+                                                                                                                                       }
+                                                                                                                               };
+
+public:
+       explicit implementation(int stream_index, const spl::shared_ptr<AVFormatContext>& context, int out_samplerate)
+               : index_(stream_index)
+               , codec_context_(open_codec(*context, AVMEDIA_TYPE_AUDIO, index_, false))
+               , out_samplerate_(out_samplerate)
+               , buffer_(10 * out_samplerate_ * codec_context_->channels) // 10 seconds of audio
+       {
+               if(!swr_)
+                       CASPAR_THROW_EXCEPTION(bad_alloc());
+
+               THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");
+
+               codec_context_->refcounted_frames = 1;
+       }
+
+       void push(const std::shared_ptr<AVPacket>& packet)
+       {
+               if(!packet)
+                       return;
+
+               if(packet->stream_index == index_ || packet->data == nullptr)
+                       packets_.push(spl::make_shared_ptr(packet));
+       }
+
+       std::shared_ptr<core::mutable_audio_buffer> poll()
+       {
+               if(packets_.empty())
+                       return nullptr;
+
+               auto packet = packets_.front();
+
+               if(packet->data == nullptr)
+               {
+                       packets_.pop();
+                       avcodec_flush_buffers(codec_context_.get());
+                       return flush_audio();
+               }
+
+               auto audio = decode(*packet);
+
+               if(packet->size == 0)
+                       packets_.pop();
+
+               return audio;
+       }
+
+       std::shared_ptr<core::mutable_audio_buffer> decode(AVPacket& pkt)
+       {
+               auto decoded_frame = create_frame();
+
+               int got_frame = 0;
+               auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), decoded_frame.get(), &got_frame, &pkt), "[audio_decoder]");
+
+               if (len == 0)
+               {
+                       pkt.size = 0;
+                       return nullptr;
+               }
+
+               pkt.data += len;
+               pkt.size -= len;
+
+               if (!got_frame)
+                       return nullptr;
+
+               const uint8_t **in = const_cast<const uint8_t**>(decoded_frame->extended_data);
+               uint8_t* out[] = { reinterpret_cast<uint8_t*>(buffer_.data()) };
+
+               const auto channel_samples = swr_convert(
+                               swr_.get(),
+                               out,
+                               static_cast<int>(buffer_.size()) / codec_context_->channels,
+                               in,
+                               decoded_frame->nb_samples);
+
+               return std::make_shared<core::mutable_audio_buffer>(
+                               buffer_.begin(),
+                               buffer_.begin() + channel_samples * decoded_frame->channels);
+       }
+
+       bool ready() const
+       {
+               return packets_.size() > 10;
+       }
+
+       std::wstring print() const
+       {
+               return L"[audio-decoder] " + u16(codec_context_->codec->long_name);
+       }
+
+       uint64_t ffmpeg_channel_layout() const
+       {
+               if (codec_context_->channel_layout == 0)
+                       return av_get_default_channel_layout(codec_context_->channels);
+               else
+                       return codec_context_->channel_layout;
+       }
+};
+
+audio_decoder::audio_decoder(int stream_index, const spl::shared_ptr<AVFormatContext>& context, int out_samplerate) : impl_(new implementation(stream_index, context, out_samplerate)){}
+void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}
+bool audio_decoder::ready() const{return impl_->ready();}
+std::shared_ptr<core::mutable_audio_buffer> audio_decoder::poll() { return impl_->poll(); }
+int    audio_decoder::num_channels() const { return impl_->codec_context_->channels; }
+uint64_t audio_decoder::ffmpeg_channel_layout() const { return impl_->ffmpeg_channel_layout(); }
+std::wstring audio_decoder::print() const{return impl_->print();}
+
+}}