Refactor StreamReader - let StreamProcessor own codec context (#3157)

Summary:
Pull Request resolved: #3157

AVCodecContext plays central role in decoding and encoding.
Currently in StreamReader, the object is owned inside of Decoder class
and it's not accessible from other objects.

This commit move the ownership of AVCodecContext out of Decoder to
StreamProcessor class so that other components can check access its field.

Also, the Decoder class, which is super thin wrapper around AVCodecContext
object, is now absorbed to StreamProcessor class.

Reviewed By: xiaohui-zhang

Differential Revision: D43924664

fbshipit-source-id: e53254955d9ce16871e393bcd8bb2794ce6a51ff

Author: mthrok

torchaudio/csrc/ffmpeg/CMakeLists.txt

@@ -12,7 +12,6 @@ set(
   stream_reader/buffer/common.cpp
   stream_reader/buffer/chunked_buffer.cpp
   stream_reader/buffer/unchunked_buffer.cpp
-  stream_reader/decoder.cpp
   stream_reader/sink.cpp
   stream_reader/stream_processor.cpp
   stream_reader/stream_reader.cpp

torchaudio/csrc/ffmpeg/stream_reader/decoder.cpp

@@ -49,27 +49,27 @@ std::unique_ptr<Buffer> get_buffer(
 
 std::unique_ptr<FilterGraph> get_filter_graph(
     AVRational input_time_base,
-    AVCodecParameters* codecpar,
+    AVCodecContext* codec_ctx,
     AVRational frame_rate,
     const std::string& filter_description) {
-  auto p = std::make_unique<FilterGraph>(codecpar->codec_type);
+  auto p = std::make_unique<FilterGraph>(codec_ctx->codec_type);
 
-  switch (codecpar->codec_type) {
+  switch (codec_ctx->codec_type) {
     case AVMEDIA_TYPE_AUDIO:
       p->add_audio_src(
-          static_cast<AVSampleFormat>(codecpar->format),
+          codec_ctx->sample_fmt,
           input_time_base,
-          codecpar->sample_rate,
-          codecpar->channel_layout);
+          codec_ctx->sample_rate,
+          codec_ctx->channel_layout);
       break;
     case AVMEDIA_TYPE_VIDEO:
       p->add_video_src(
-          static_cast<AVPixelFormat>(codecpar->format),
+          codec_ctx->pix_fmt,
           input_time_base,
           frame_rate,
-          codecpar->width,
-          codecpar->height,
-          codecpar->sample_aspect_ratio);
+          codec_ctx->width,
+          codec_ctx->height,
+          codec_ctx->sample_aspect_ratio);
       break;
     default:
       TORCH_CHECK(false, "Only audio/video are supported.");
@@ -84,25 +84,25 @@ std::unique_ptr<FilterGraph> get_filter_graph(
 
 Sink::Sink(
     AVRational input_time_base_,
-    AVCodecParameters* codecpar_,
+    AVCodecContext* codec_ctx_,
     int frames_per_chunk,
     int num_chunks,
     AVRational frame_rate_,
     const c10::optional<std::string>& filter_description_,
     const torch::Device& device)
     : input_time_base(input_time_base_),
-      codecpar(codecpar_),
+      codec_ctx(codec_ctx_),
       frame_rate(frame_rate_),
       filter_description(filter_description_.value_or(
-          codecpar->codec_type == AVMEDIA_TYPE_AUDIO ? "anull" : "null")),
+          codec_ctx->codec_type == AVMEDIA_TYPE_AUDIO ? "anull" : "null")),
       filter(get_filter_graph(
           input_time_base_,
-          codecpar_,
+          codec_ctx,
           frame_rate,
           filter_description)),
       output_time_base(filter->get_output_timebase()),
       buffer(get_buffer(
-          codecpar_->codec_type,
+          codec_ctx->codec_type,
           frames_per_chunk,
           num_chunks,
           double(output_time_base.num) / output_time_base.den,
@@ -139,7 +139,7 @@ FilterGraphOutputInfo Sink::get_filter_output_info() const {
 
 void Sink::flush() {
   filter = get_filter_graph(
-      input_time_base, codecpar, frame_rate, filter_description);
+      input_time_base, codec_ctx, frame_rate, filter_description);
   buffer->flush();
 }
 

torchaudio/csrc/ffmpeg/stream_reader/decoder.h

@@ -12,7 +12,7 @@ class Sink {
 
   // Parameters for recreating FilterGraph
   AVRational input_time_base;
-  AVCodecParameters* codecpar;
+  AVCodecContext* codec_ctx;
   AVRational frame_rate;
   std::string filter_description;
   std::unique_ptr<FilterGraph> filter;
@@ -23,7 +23,7 @@ class Sink {
   std::unique_ptr<Buffer> buffer;
   Sink(
       AVRational input_time_base,
-      AVCodecParameters* codecpar,
+      AVCodecContext* codec_ctx,
       int frames_per_chunk,
       int num_chunks,
       AVRational frame_rate,