Expand description
Receiver pipeline.
N stream tasks ──frames──> rayon pool (open → decrypt → decompress → pwrite) ──> disk
read-ahead of (cfg.workers threads)
queue_depth eachFrames are self-describing, so a chunk is written the moment it decodes —
there is no reassembly buffer, and a 100 GB file costs the same memory as a
small one. A file is renamed from its .part sidecar into place only after
every chunk has landed and, when enabled, its hash has been checked.
Functions§
- receive
- Receive into
dest_root.