]> git.sesse.net Git - casparcg/blobdiff - modules/ffmpeg/producer/audio/audio_decoder.cpp
[ffmpeg] Copied flush logic when seeking from 2.0, as well as current frame in clip...
[casparcg] / modules / ffmpeg / producer / audio / audio_decoder.cpp
index 1ac7fdeaa796fe61c3db3c1602f62b11e747ca5a..e92a7c18134859e05370fc65b13e6fb18a5d720e 100644 (file)
-/*\r
-* Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>\r
-*\r
-* This file is part of CasparCG (www.casparcg.com).\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
-* Author: Robert Nagy, ronag89@gmail.com\r
-*/\r
-\r
-#include "../../stdafx.h"\r
-\r
-#include "audio_decoder.h"\r
-\r
-#include "audio_resampler.h"\r
-\r
-#include "../util/util.h"\r
-#include "../../ffmpeg_error.h"\r
-\r
-#include <core/video_format.h>\r
-\r
-#include <tbb/cache_aligned_allocator.h>\r
-\r
-#include <queue>\r
-\r
-#if defined(_MSC_VER)\r
-#pragma warning (push)\r
-#pragma warning (disable : 4244)\r
-#endif\r
-extern "C" \r
-{\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 { namespace ffmpeg {\r
-       \r
-struct audio_decoder::implementation : boost::noncopyable\r
-{      \r
-       int                                                                                                                     index_;\r
-       const safe_ptr<AVCodecContext>                                                          codec_context_;         \r
-       const core::video_format_desc                                                           format_desc_;\r
-\r
-       audio_resampler                                                                                         resampler_;\r
-\r
-       std::vector<int8_t,  tbb::cache_aligned_allocator<int8_t>>      buffer1_;\r
-\r
-       std::queue<safe_ptr<AVPacket>>                                                          packets_;\r
-\r
-       const int64_t                                                                                           nb_frames_;\r
-       tbb::atomic<size_t>                                                                                     file_frame_number_;\r
-public:\r
-       explicit implementation(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
-               : format_desc_(format_desc)     \r
-               , codec_context_(open_codec(*context, AVMEDIA_TYPE_AUDIO, index_))\r
-               , resampler_(format_desc.audio_channels,        codec_context_->channels,\r
-                                        format_desc.audio_sample_rate, codec_context_->sample_rate,\r
-                                        AV_SAMPLE_FMT_S32,                             codec_context_->sample_fmt)\r
-               , buffer1_(AVCODEC_MAX_AUDIO_FRAME_SIZE*2)\r
-               , nb_frames_(0)//context->streams[index_]->nb_frames)\r
-       {               \r
-               file_frame_number_ = 0;\r
-               CASPAR_LOG(debug) << "[audio_decoder] " << context->streams[index_]->codec->codec->long_name;      \r
-       }\r
-\r
-       void push(const std::shared_ptr<AVPacket>& packet)\r
-       {                       \r
-               if(!packet)\r
-                       return;\r
-\r
-               if(packet->stream_index == index_ || packet->data == nullptr)\r
-                       packets_.push(make_safe_ptr(packet));\r
-       }       \r
-       \r
-       std::shared_ptr<core::audio_buffer> poll()\r
-       {\r
-               if(packets_.empty())\r
-                       return nullptr;\r
-                               \r
-               auto packet = packets_.front();\r
-\r
-               if(packet->data == nullptr)\r
-               {\r
-                       packets_.pop();\r
-                       file_frame_number_ = static_cast<size_t>(packet->pos);\r
-                       avcodec_flush_buffers(codec_context_.get());\r
-                       return flush_audio();\r
-               }\r
-\r
-               auto audio = decode(*packet);\r
-\r
-               if(packet->size == 0)                                   \r
-                       packets_.pop();\r
-\r
-               return audio;\r
-       }\r
-\r
-       std::shared_ptr<core::audio_buffer> decode(AVPacket& pkt)\r
-       {               \r
-               buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
-               int written_bytes = buffer1_.size() - FF_INPUT_BUFFER_PADDING_SIZE;\r
-               \r
-               int ret = THROW_ON_ERROR2(avcodec_decode_audio3(codec_context_.get(), reinterpret_cast<int16_t*>(buffer1_.data()), &written_bytes, &pkt), "[audio_decoder]");\r
-\r
-               // There might be several frames in one packet.\r
-               pkt.size -= ret;\r
-               pkt.data += ret;\r
-                       \r
-               buffer1_.resize(written_bytes);\r
-\r
-               buffer1_ = resampler_.resample(std::move(buffer1_));\r
-               \r
-               const auto n_samples = buffer1_.size() / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);\r
-               const auto samples = reinterpret_cast<int32_t*>(buffer1_.data());\r
-\r
-               ++file_frame_number_;\r
-\r
-               return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
-       }\r
-\r
-       bool ready() const\r
-       {\r
-               return packets_.size() > 10;\r
-       }\r
-\r
-       uint32_t nb_frames() const\r
-       {\r
-               return 0;//std::max<int64_t>(nb_frames_, file_frame_number_);\r
-       }\r
-};\r
-\r
-audio_decoder::audio_decoder(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new implementation(context, format_desc)){}\r
-void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}\r
-bool audio_decoder::ready() const{return impl_->ready();}\r
-std::shared_ptr<core::audio_buffer> audio_decoder::poll(){return impl_->poll();}\r
-uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
-uint32_t audio_decoder::file_frame_number() const{return impl_->file_frame_number_;}\r
-\r
-}}
\ No newline at end of file
+/*
+* Copyright (c) 2011 Sveriges Television AB <info@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 "../input/input.h"
+#include "../../ffmpeg_error.h"
+
+#include <core/video_format.h>
+#include <core/frame/audio_channel_layout.h>
+
+#include <common/log.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 {
+       
+uint64_t get_ffmpeg_channel_layout(AVCodecContext* dec)
+{
+       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);
+       return layout;
+}
+
+struct audio_decoder::impl : boost::noncopyable
+{      
+       core::monitor::subject                                                                          monitor_subject_;
+       input&                                                                                                          input_;
+       int                                                                                                                     index_;
+       int                                                                                                                     actual_index_;
+       const core::video_format_desc                                                           format_desc_;
+       const spl::shared_ptr<AVCodecContext>                                           codec_context_          = open_codec(input_.context(), AVMEDIA_TYPE_AUDIO, actual_index_, false);
+
+       std::shared_ptr<SwrContext>                                                                     swr_                            {
+                                                                                                                                                                               swr_alloc_set_opts(
+                                                                                                                                                                                               nullptr,
+                                                                                                                                                                                               create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
+                                                                                                                                                                                               AV_SAMPLE_FMT_S32,
+                                                                                                                                                                                               format_desc_.audio_sample_rate,
+                                                                                                                                                                                               create_channel_layout_bitmask(codec_context_->channels),//get_ffmpeg_channel_layout(codec_context_.get()),
+                                                                                                                                                                                               codec_context_->sample_fmt,
+                                                                                                                                                                                               codec_context_->sample_rate,
+                                                                                                                                                                                               0,
+                                                                                                                                                                                               nullptr),
+                                                                                                                                                                               [](SwrContext* p){swr_free(&p); }
+                                                                                                                                                                       };
+
+       cache_aligned_vector<uint8_t>                                                           buffer_;
+
+       std::shared_ptr<AVPacket>                                                                       current_packet_;
+       
+public:
+       explicit impl(
+                       input& in,
+                       const core::video_format_desc& format_desc,
+                       int audio_stream_index)
+               : input_(in)
+               , index_(audio_stream_index)
+               , actual_index_(input_.get_actual_audio_stream_index(index_))
+               , format_desc_(format_desc)
+               , buffer_(480000 * 4)
+       {
+               if(!swr_)
+                       CASPAR_THROW_EXCEPTION(bad_alloc());
+
+               THROW_ON_ERROR2(swr_init(swr_.get()), "[audio_decoder]");
+       }
+               
+       std::shared_ptr<AVFrame> poll()
+       {               
+               if(!current_packet_ && !input_.try_pop_audio(current_packet_, index_))
+                       return nullptr;
+               
+               std::shared_ptr<AVFrame> audio;
+
+               if (!current_packet_->data)
+               {
+                       if(codec_context_->codec->capabilities & CODEC_CAP_DELAY)                       
+                               audio = decode(*current_packet_);
+                       
+                       if (!audio)
+                       {
+                               avcodec_flush_buffers(codec_context_.get());
+                               current_packet_.reset();
+                               audio = flush();
+                       }
+               }
+               else
+               {
+                       audio = decode(*current_packet_);
+                       
+                       if(current_packet_->size == 0)
+                               current_packet_.reset();
+               }
+       
+               return audio;
+       }
+
+       std::shared_ptr<AVFrame> decode(AVPacket& pkt)
+       {               
+               auto frame = create_frame();
+               
+               int got_frame = 0;
+               auto len = THROW_ON_ERROR2(avcodec_decode_audio4(codec_context_.get(), 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**>(frame->extended_data);
+               uint8_t* out[]          = {buffer_.data()};
+
+               auto channel_samples = swr_convert(swr_.get(), 
+                                                                                       out, static_cast<int>(buffer_.size()) / codec_context_->channels / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32),
+                                                                                       in, frame->nb_samples);
+
+               frame->data[0]          = buffer_.data();
+               frame->linesize[0]      = channel_samples * codec_context_->channels * av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);
+               frame->nb_samples       = channel_samples;
+               frame->format           = AV_SAMPLE_FMT_S32;
+
+               monitor_subject_  << core::monitor::message("/file/audio/sample-rate")  % codec_context_->sample_rate
+                                               << core::monitor::message("/file/audio/channels")       % codec_context_->channels
+                                               << core::monitor::message("/file/audio/format")         % u8(av_get_sample_fmt_name(codec_context_->sample_fmt))
+                                               << core::monitor::message("/file/audio/codec")          % u8(codec_context_->codec->long_name);                 
+
+               return frame;
+       }
+       
+       uint32_t nb_frames() const
+       {
+               return 0;
+       }
+
+       std::wstring print() const
+       {               
+               return L"[audio-decoder] " + u16(codec_context_->codec->long_name);
+       }
+};
+
+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)){}
+audio_decoder::audio_decoder(audio_decoder&& other) : impl_(std::move(other.impl_)){}
+audio_decoder& audio_decoder::operator=(audio_decoder&& other){impl_ = std::move(other.impl_); return *this;}
+std::shared_ptr<AVFrame> audio_decoder::operator()(){return impl_->poll();}
+uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}
+std::wstring audio_decoder::print() const{return impl_->print();}
+core::monitor::subject& audio_decoder::monitor_output() { return impl_->monitor_subject_;}
+
+}}