Skip to content

Improve rendering latency and stability

David Edmundson requested to merge work/d_ed/faster_rdp into master

We cannot drop random frames after the encoder. It leads to content glitches or a frozen video stream. It's best to always send everything and rely on the pre-encoder frame queue.

If we have multiple frames ready to send, send them immediately. Otherwise when we change the framerate we don't burn through our queue and our latency will never decrease.

This patch also takes on a more aggressive frame rate policy. We currently calculate the average throughput and send the target based on that. This means we never get through any existing backlog or provide a safety net.

Merge request reports