audiorenderer: split audio rendering up into chunks for better parallelism

app/render/backend/audiorenderbackend.cpp

@@ -100,6 +100,18 @@ void AudioRenderBackend::ConnectWorkerToThis(RenderWorker *worker)
   connect(arw, &AudioRenderWorker::ConformUnavailable, this, &AudioRenderBackend::ConformUnavailable, Qt::QueuedConnection);
 }
 
+TimeRange AudioRenderBackend::PopNextFrameFromQueue()
+{
+  TimeRange range = cache_queue_.first();
+
+  // Limit range per worker to 2 seconds (FIXME: arbitrary, should be tweaked, maybe even in config?)
+  range.set_out(qMin(range.out(), range.in() + rational(2)));
+
+  cache_queue_.RemoveTimeRange(range);
+
+  return range;
+}
+
 void AudioRenderBackend::ConformUnavailable(StreamPtr stream, const TimeRange &range, const rational &stream_time, const AudioRenderingParams& params)
 {
   ConformWaitInfo info = {stream, params, range, stream_time};

app/render/backend/audiorenderbackend.h

@@ -47,6 +47,8 @@ protected:
 
   virtual void ConnectWorkerToThis(RenderWorker* worker) override;
 
+  virtual TimeRange PopNextFrameFromQueue() override;
+
   QHash<Node*, Node*> copy_map_;
 
 private: