\r
#include "audio_decoder.h"\r
\r
-#include <tbb/task_group.h>\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
#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
std::shared_ptr<AVCodecContext> codec_context_; \r
const core::video_format_desc format_desc_;\r
int index_;\r
- std::vector<int8_t, tbb::cache_aligned_allocator<int8_t>> buffer1_; // avcodec_decode_audio3 needs 4 byte alignment\r
- std::vector<int8_t, tbb::cache_aligned_allocator<int8_t>> buffer2_; // avcodec_decode_audio3 needs 4 byte alignment\r
- std::vector<int16_t, tbb::cache_aligned_allocator<int16_t>> audio_samples_; // avcodec_decode_audio3 needs 4 byte alignment\r
+ std::unique_ptr<audio_resampler> resampler_;\r
+\r
+ std::vector<int8_t, tbb::cache_aligned_allocator<int8_t>> buffer1_;\r
+\r
std::queue<std::shared_ptr<AVPacket>> packets_;\r
- std::shared_ptr<ReSampleContext> resampler_;\r
+\r
+ int64_t nb_frames_;\r
public:\r
- explicit implementation(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
+ explicit implementation(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
: format_desc_(format_desc) \r
- { \r
- AVCodec* dec;\r
- index_ = av_find_best_stream(context.get(), AVMEDIA_TYPE_AUDIO, -1, -1, &dec, 0);\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
- int errn = avcodec_open(context->streams[index_]->codec, dec);\r
- if(errn < 0)\r
- return;\r
- \r
- codec_context_.reset(context->streams[index_]->codec, avcodec_close);\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
- if(codec_context_ &&\r
- (codec_context_->sample_rate != static_cast<int>(format_desc_.audio_sample_rate) || \r
- codec_context_->channels != static_cast<int>(format_desc_.audio_channels)) ||\r
- codec_context_->sample_fmt != AV_SAMPLE_FMT_S16)\r
- { \r
- auto resampler = av_audio_resample_init(format_desc_.audio_channels, codec_context_->channels,
- format_desc_.audio_sample_rate, codec_context_->sample_rate,
- AV_SAMPLE_FMT_S16, codec_context_->sample_fmt,
- 16, 10, 0, 0.8);\r
+ buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
\r
- CASPAR_LOG(warning) << L" Invalid audio format.";\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
- if(resampler)\r
- resampler_.reset(resampler, audio_resample_close);\r
- else\r
- codec_context_ = nullptr;\r
- } \r
+ CASPAR_LOG_CURRENT_EXCEPTION();\r
+ CASPAR_LOG(warning) << "[audio_decoder] Failed to open audio-stream. Running without audio."; \r
+ }\r
}\r
\r
void push(const std::shared_ptr<AVPacket>& packet)\r
{ \r
- if(!codec_context_)\r
- return;\r
-\r
if(packet && packet->stream_index != index_)\r
return;\r
\r
packets_.push(packet);\r
} \r
\r
- std::vector<std::vector<int16_t>> poll()\r
+ std::vector<std::shared_ptr<core::audio_buffer>> poll()\r
{\r
- std::vector<std::vector<int16_t>> result;\r
+ std::vector<std::shared_ptr<core::audio_buffer>> result;\r
+\r
+ if(packets_.empty())\r
+ return result;\r
\r
if(!codec_context_)\r
- result.push_back(std::vector<int16_t>(format_desc_.audio_samples_per_frame, 0));\r
- else if(!packets_.empty())\r
+ return empty_poll();\r
+ \r
+ auto packet = packets_.front();\r
+\r
+ if(packet) \r
{\r
- decode(packets_.front());\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
- while(audio_samples_.size() > format_desc_.audio_samples_per_frame)\r
- {\r
- const auto begin = audio_samples_.begin();\r
- const auto end = audio_samples_.begin() + format_desc_.audio_samples_per_frame;\r
+ return result;\r
+ }\r
\r
- result.push_back(std::vector<int16_t>(begin, end));\r
- audio_samples_.erase(begin, end);\r
- }\r
- }\r
+ std::vector<std::shared_ptr<core::audio_buffer>> empty_poll()\r
+ {\r
+ auto packet = packets_.front();\r
+ packets_.pop();\r
\r
- return result;\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
- void decode(const std::shared_ptr<AVPacket>& packet)\r
- { \r
- if(!packet) // eof\r
- {\r
- auto truncate = audio_samples_.size() % format_desc_.audio_samples_per_frame;\r
- if(truncate > 0)\r
- {\r
- audio_samples_.resize(audio_samples_.size() - truncate); \r
- CASPAR_LOG(info) << L"Truncating " << truncate << L" audio-samples."; \r
- }\r
- avcodec_flush_buffers(codec_context_.get());\r
- }\r
- else\r
- {\r
- buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2, 0);\r
- int written_bytes = buffer1_.size() - FF_INPUT_BUFFER_PADDING_SIZE;\r
- // TODO: Packet might contain multiple frames\r
- const int errn = avcodec_decode_audio3(codec_context_.get(), reinterpret_cast<int16_t*>(buffer1_.data()), &written_bytes, 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
- buffer1_.resize(written_bytes);\r
-\r
- if(resampler_)\r
- {\r
- buffer2_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2, 0);\r
- auto ret = audio_resample(resampler_.get(),\r
- reinterpret_cast<short*>(buffer2_.data()), \r
- reinterpret_cast<short*>(buffer1_.data()), \r
- buffer1_.size() / av_get_bytes_per_sample(codec_context_->sample_fmt)); \r
- buffer2_.resize(ret);\r
- std::swap(buffer1_, buffer2_);\r
- }\r
-\r
- const auto n_samples = buffer1_.size() / av_get_bytes_per_sample(AV_SAMPLE_FMT_S16);\r
- const auto samples = reinterpret_cast<int16_t*>(buffer1_.data());\r
-\r
- audio_samples_.insert(audio_samples_.end(), samples, samples + n_samples); \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
+ return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
}\r
\r
bool ready() const\r
{\r
- return !codec_context_ || !packets_.empty();\r
+ return !packets_.empty();\r
}\r
};\r
\r
-audio_decoder::audio_decoder(const std::shared_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new implementation(context, format_desc)){}\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::vector<int16_t>> audio_decoder::poll(){return impl_->poll();}\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