nx_bench_dist_persist_test.nx
buildroot/runtime/nx_bench_dist_persist_test.nx
about
nx_bench_dist_persist_test.nx -- run distribution + persist to disk.
Composes the distribution measurement primitive with the disk-write
persistence primitive. Produces a REAL TSV file at
/tmp/nx_bench_dist.tsv
containing one header + N data rows + summary stats.
This is the substrate-honest end-to-end workflow the s-class
comparison campaign needs: measurement happens, bytes hit disk,
external tools (sort / awk / future llama.cpp comparison rigs) read
the file and diff against their own.
After the smoke runs, the operator can:
cat /tmp/nx_bench_dist.tsv
sort -k7n /tmp/nx_bench_dist.tsv
awk -F'\t' '{print $1, $8}' /tmp/nx_bench_dist.tsv
And see real persisted substrate measurements.
dependencies 13 imports · 0 importers
diagram shows first 10 each side; +3 more imports, +0 more importers in the complete lists below.
imports: nx_syscalls.nxnx_tier.nxnx_clock.nxnx_actor.nxnx_message.nxnx_session.nxnx_bench_companion.nxnx_bench_companion_timed.nxnx_bench_companion_tsv.nxnx_bench_stats.nxnx_bench_tsv_persist.nxnx_gguf_fixture_tiny.nxnx_actor_role_llm_v2.nx
imported by: nobody (leaf or entry point)
call flow from main pre-order; caps 40 nodes / depth 6 declared; ↻ = already shown
structs
| none |
consts
| none |
functions
| 36 | func _emit_dec_to_buf(buf: *u8, off: nx_int, cap: nx_int, n: i64) -> nx_int |
| 66 | func _drive_one_run(seed: i64, n_iters: nx_int, now: nx_size) -> i64 |
| 96 | func main() -> i64 |