code wiki / _hdl_build / nx_pardl.nx
nx_pardl.nx
buildroot/runtime/_hdl_build/nx_pardl.nx
about
nx_pardl.nx -- PARALLEL fragment download (the performance capability yt-dlp/aria2 have via -N /
--max-connection-per-server). Fork-pool: fetch `concurrency` fragments AT ONCE per batch (fork-per-fragment
= the proven X-DLP idiom, and the per-fetch TLS mmap leak dies with each child), each child writes its
fragment into a SHARED slot; the parent waits the batch, INTEGRITY-verifies each (nx_robust_dl), and marks
the resume-state -- so parallelism composes with retry/integrity/resume, not instead of them.
Measured, not checkboxed: pdl_download is timed serial-vs-parallel in the gate to REPORT the speedup axis.
dependencies 2 imports · 1 importers
imports: nx_syscalls.nxnx_robust_dl.nx
imported by: nx_pardl_gate.nx
structs
| none |
consts
| none |
functions
| 14 | func pdl_download(nfrags: i64, fragcap: i64, concurrency: i64, latency_ms: i64, sharedbuf: *u8, state: *u8) -> i64 |