Expand description
Per-rung segment chunk queue connecting the decode pump to N encoder workers.
v2 multi-GPU model (2026-05-11): the pump groups decoded source
frames into fixed-size chunks (one chunk = one CMAF segment’s
worth of frames = keyframe_interval frames) and pushes them
into this queue with a monotonic segment index. Encoder workers
pop chunks and emit segments. The segment index travels with the
frames so each worker knows which output file to write.
Single-producer, multi-consumer. Bounded capacity for memory safety: pump blocks when full, workers block when empty.