\r
#include "audio_decoder.h"\r
\r
+#include "audio_resampler.h"\r
+\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
- #define __STDC_CONSTANT_MACROS\r
- #define __STDC_LIMIT_MACROS\r
#include <libavformat/avformat.h>\r
#include <libavcodec/avcodec.h>\r
}\r
#endif\r
\r
namespace caspar {\r
-\r
-struct audio_decoder::implementation : boost::noncopyable\r
-{\r
- typedef std::vector<short, tbb::cache_aligned_allocator<short>> buffer;\r
\r
- AVCodecContext* codec_context_;\r
-\r
- buffer audio_buffer_; \r
- buffer current_chunk_;\r
+struct audio_decoder::implementation : boost::noncopyable\r
+{ \r
+ std::shared_ptr<AVCodecContext> codec_context_; \r
+ const core::video_format_desc format_desc_;\r
+ int index_;\r
+ std::unique_ptr<audio_resampler> resampler_;\r
\r
- const size_t audio_frame_size_;\r
+ std::vector<int8_t, tbb::cache_aligned_allocator<int8_t>> buffer1_;\r
\r
- static const size_t SAMPLE_RATE = 48000;\r
- static const size_t N_CHANNELS = 2;\r
+ std::queue<std::shared_ptr<AVPacket>> packets_;\r
\r
+ int64_t nb_frames_;\r
public:\r
- explicit implementation(AVCodecContext* codec_context, double fps) \r
- : codec_context_(codec_context)\r
- , audio_buffer_(4*SAMPLE_RATE*2+FF_INPUT_BUFFER_PADDING_SIZE/2)\r
- , audio_frame_size_(static_cast<size_t>(static_cast<double>(SAMPLE_RATE) / fps) * N_CHANNELS)\r
- {\r
- if(!codec_context)\r
- BOOST_THROW_EXCEPTION(null_argument() << arg_name_info("codec_context")); \r
+ explicit implementation(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
+ : format_desc_(format_desc) \r
+ , nb_frames_(0)\r
+ { \r
+ try\r
+ {\r
+ AVCodec* dec;\r
+ index_ = THROW_ON_ERROR2(av_find_best_stream(context.get(), AVMEDIA_TYPE_AUDIO, -1, -1, &dec, 0), "[audio_decoder]");\r
+\r
+ THROW_ON_ERROR2(avcodec_open(context->streams[index_]->codec, dec), "[audio_decoder]");\r
+ \r
+ codec_context_.reset(context->streams[index_]->codec, avcodec_close);\r
+\r
+ buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
+\r
+ resampler_.reset(new audio_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
+ }\r
+ catch(...)\r
+ {\r
+ index_ = THROW_ON_ERROR2(av_find_best_stream(context.get(), AVMEDIA_TYPE_VIDEO, -1, -1, nullptr, 0), "[audio_decoder]");\r
+\r
+ CASPAR_LOG_CURRENT_EXCEPTION();\r
+ CASPAR_LOG(warning) << "[audio_decoder] Failed to open audio-stream. Running without audio."; \r
+ }\r
}\r
- \r
- std::vector<std::vector<short>> execute(const aligned_buffer& audio_packet)\r
+\r
+ void push(const std::shared_ptr<AVPacket>& packet)\r
{ \r
- int written_bytes = audio_buffer_.size()*2 - FF_INPUT_BUFFER_PADDING_SIZE;\r
- const int errn = avcodec_decode_audio2(codec_context_, audio_buffer_.data(), &written_bytes, audio_packet.data(), audio_packet.size());\r
+ if(packet && packet->stream_index != index_)\r
+ return;\r
\r
- if(errn < 0 || codec_context_->sample_rate != SAMPLE_RATE || codec_context_->channels != 2)\r
- { \r
- BOOST_THROW_EXCEPTION(\r
- invalid_operation() <<\r
- boost::errinfo_api_function("avcodec_decode_audio2") <<\r
- boost::errinfo_errno(AVUNERROR(errn)));\r
+ packets_.push(packet);\r
+ } \r
+ \r
+ std::vector<std::shared_ptr<core::audio_buffer>> poll()\r
+ {\r
+ std::vector<std::shared_ptr<core::audio_buffer>> result;\r
+\r
+ if(packets_.empty())\r
+ return result;\r
+\r
+ if(!codec_context_)\r
+ return empty_poll();\r
+ \r
+ auto packet = packets_.front();\r
+\r
+ if(packet) \r
+ {\r
+ result.push_back(decode(*packet));\r
+ if(packet->size == 0) \r
+ packets_.pop();\r
}\r
+ else \r
+ { \r
+ avcodec_flush_buffers(codec_context_.get());\r
+ result.push_back(nullptr);\r
+ packets_.pop();\r
+ } \r
\r
- current_chunk_.insert(current_chunk_.end(), audio_buffer_.data(), audio_buffer_.data() + written_bytes/2);\r
+ return result;\r
+ }\r
\r
- std::vector<std::vector<short>> chunks;\r
- \r
- const auto last = current_chunk_.end() - current_chunk_.size() % audio_frame_size_;\r
+ std::vector<std::shared_ptr<core::audio_buffer>> empty_poll()\r
+ {\r
+ auto packet = packets_.front();\r
+ packets_.pop();\r
\r
- for(auto it = current_chunk_.begin(); it != last; it += audio_frame_size_) \r
- chunks.push_back(std::vector<short>(it, it + audio_frame_size_)); \r
+ if(!packet) \r
+ return boost::assign::list_of(nullptr);\r
+ \r
+ return boost::assign::list_of(std::make_shared<core::audio_buffer>(format_desc_.audio_samples_per_frame, 0)); \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
- current_chunk_.erase(current_chunk_.begin(), last);\r
+ buffer1_ = resampler_->resample(std::move(buffer1_));\r
\r
- return chunks;\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
+ return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
+ }\r
+\r
+ bool ready() const\r
+ {\r
+ return !packets_.empty();\r
}\r
};\r
\r
-audio_decoder::audio_decoder(AVCodecContext* codec_context, double fps) : impl_(new implementation(codec_context, fps)){}\r
-std::vector<std::vector<short>> audio_decoder::execute(const aligned_buffer& audio_packet){return impl_->execute(audio_packet);}\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::vector<std::shared_ptr<core::audio_buffer>> audio_decoder::poll(){return impl_->poll();}\r
+int64_t audio_decoder::nb_frames() const{return impl_->nb_frames_;}\r
}
\ No newline at end of file
projects / casparcg / blobdiff