Do not use channel_layout in StreamReader

We only care about the number of channels, so no need to create channel_layout. One can directly pass the number of channels to filter. Also int64 channel_layout is a deprecated attributes.
pytorch · Apr 10, 2024 · fc80842 · fc80842
1 parent ea437b3
commit fc80842
Show file tree

Hide file tree

Showing 4 changed files with 14 additions and 14 deletions.
diff --git a/src/libtorio/ffmpeg/filter_graph.cpp b/src/libtorio/ffmpeg/filter_graph.cpp
@@ -22,17 +22,17 @@ std::string get_audio_src_args(
     AVSampleFormat format,
     AVRational time_base,
     int sample_rate,
-    uint64_t channel_layout) {
+    int nb_channels) {
   char args[512];
   std::snprintf(
       args,
       sizeof(args),
-      "time_base=%d/%d:sample_rate=%d:sample_fmt=%s:channel_layout=0x%" PRIx64,
+      "time_base=%d/%d:sample_rate=%d:sample_fmt=%s:channel_layout=%dc",
       time_base.num,
       time_base.den,
       sample_rate,
       av_get_sample_fmt_name(format),
-      channel_layout);
+      nb_channels);
   return std::string(args);
 }
 
@@ -66,10 +66,10 @@ void FilterGraph::add_audio_src(
     AVSampleFormat format,
     AVRational time_base,
     int sample_rate,
-    uint64_t channel_layout) {
+    int nb_channels) {
   add_src(
       avfilter_get_by_name("abuffer"),
-      get_audio_src_args(format, time_base, sample_rate, channel_layout));
+      get_audio_src_args(format, time_base, sample_rate, nb_channels));
 }
 
 void FilterGraph::add_video_src(

diff --git a/src/libtorio/ffmpeg/filter_graph.h b/src/libtorio/ffmpeg/filter_graph.h
@@ -47,7 +47,7 @@ class FilterGraph {
       AVSampleFormat format,
       AVRational time_base,
       int sample_rate,
-      uint64_t channel_layout);
+      int nb_channels);
 
   void add_video_src(
       AVPixelFormat format,

diff --git a/src/libtorio/ffmpeg/stream_reader/post_process.cpp b/src/libtorio/ffmpeg/stream_reader/post_process.cpp
@@ -19,10 +19,15 @@ FilterGraphFactory get_audio_factory(
   return [fmt = codec_ctx->sample_fmt,
           time_base,
           rate = codec_ctx->sample_rate,
-          channel_layout = codec_ctx->channel_layout](
-             const std::string& filter_desc) -> FilterGraph {
+          nb_channels =
+#if LIBAVUTIL_VERSION_INT >= AV_VERSION_INT(58, 2, 100)
+              codec_ctx->ch_layout.nb_channels
+#else
+              codec_ctx->channels
+#endif
+  ](const std::string& filter_desc) -> FilterGraph {
     FilterGraph f;
-    f.add_audio_src(fmt, time_base, rate, channel_layout);
+    f.add_audio_src(fmt, time_base, rate, nb_channels);
     f.add_audio_sink();
     f.add_process(filter_desc);
     f.create_filter();

diff --git a/src/libtorio/ffmpeg/stream_reader/stream_processor.cpp b/src/libtorio/ffmpeg/stream_reader/stream_processor.cpp
@@ -140,11 +140,6 @@ void open_codec(
     av_dict_set(&opts, "threads", "1", 0);
   }
 
-  if (!codec_ctx->channel_layout) {
-    codec_ctx->channel_layout =
-        av_get_default_channel_layout(codec_ctx->channels);
-  }
-
   int ret = avcodec_open2(codec_ctx, codec_ctx->codec, &opts);
   clean_up_dict(opts);
   TORCH_CHECK(