code wiki / bench
topic: bench
55 modules sharing the bench name family (derived from the tree's prefix discipline).
The 'bench' topic family in the Nishi sovereign ecosystem is designed to evaluate and compare the performance and accuracy of various computational primitives and algorithms within the NishiLang framework. Modules like `bench_emit` generate and stream workloads through the system's primitives, while `bench_hll_vs_cpc` and `bench_hll_vs_cpcd` conduct head-to-head comparisons of different data structures under controlled conditions. Additionally, `nx_bench_census_gate` and `nx_bench_census_lib` track benchmark coverage across the ecosystem, ensuring comprehensive evaluation without external I/O.
auto-narrated by the local model from this topic's module headers; links verified against the wiki index.
narrated overview -- maintained by the narration lane, module links verified against this wiki.
| module | description | lines | funcs |
|---|---|---|---|
| bench_emit.nx | stream workload through OUR primitives, emit numbers to stdout. | 106 | 5 |
| bench_hll_vs_cpc.nx | head-to-head workload runner. | 77 | 1 |
| bench_hll_vs_cpcd.nx | HLL_8 vs CPC DENSE at matched accuracy. | 73 | 1 |
| bench_multiseed.nx | 20-workload accuracy averaging for OUR HLL + CPC. | 149 | 6 |
| nx_bench_census_gate.nx | gates the benchmark-coverage census on in-memory fixtures (no file I/O): | 46 | 3 |
| nx_bench_census_lib.nx | ecosystem BENCHMARK-COVERAGE census (the anti-navel-gazing engine scaled to | 126 | 8 |
| nx_bench_collatz.nx | NishiLang twin of bench/xlang/collatz.c (branch-heavy). | 57 | 2 |
| nx_bench_comb.nx | isolated bench: comb fixed-base k*G (bits-up path). | 47 | 3 |
| nx_bench_companion.nx | substrate-honest measurement primitive for | 212 | 7 |
| nx_bench_companion_test.nx | smoke for the bench-capture primitive. | 150 | 1 |
| nx_bench_companion_timed.nx | wall-clock-wrapped substrate measurement. | 107 | 5 |
| nx_bench_companion_timed_test.nx | smoke for wall-clock bench wrapper. | 84 | 1 |
| nx_bench_companion_tsv.nx | emit NxBenchReport as TSV row to a | 304 | 7 |
| nx_bench_companion_tsv_parse.nx | read bench TSV rows back from bytes. | 224 | 6 |
| nx_bench_companion_tsv_parse_test.nx | roundtrip smoke for TSV parser. | 180 | 1 |
| nx_bench_companion_tsv_test.nx | smoke for TSV emitter. | 135 | 1 |
| nx_bench_core.nx | B1 second stone of NISHI_PERFORMANCE_ | 212 | 3 |
| nx_bench_core_test.nx | KAT for B1 bench harness. | 172 | 1 |
| nx_bench_dist_emit_test.nx | 10-run distribution measurement. | 204 | 6 |
| nx_bench_dist_persist_test.nx | run distribution + persist to disk. | 197 | 3 |
| nx_bench_divpow2.nx | NishiLang twin of bench/xlang/divpow2.c. Signed | 57 | 2 |
| nx_bench_emit.nx | stream workload through OUR primitives, emit numbers to stdout. | 112 | 5 |
| nx_bench_fannkuch.nx | NishiLang twin of bench/xlang/fannkuch.c | 87 | 2 |
| nx_bench_fetch.nx | THE GENERIC SOVEREIGN URL->FILE FETCHER (rule 15 DRY; eats the hardcoded-URL | 130 | 9 |
| nx_bench_fnv.nx | NishiLang twin of bench/xlang/fnv.c. FNV-1a over an LCG byte | 52 | 2 |
| nx_bench_harness.nx | the bits-up search leaderboard engine. | 266 | 6 |
| nx_bench_harness_test.nx | KAT for the leaderboard engine. | 86 | 1 |
| nx_bench_hll_throughput.nx | head-to-head HLL throughput bench. | 115 | 3 |
| nx_bench_hll_vs_cpc.nx | head-to-head workload runner. | 85 | 1 |
| nx_bench_hll_vs_cpcd.nx | HLL_8 vs CPC DENSE at matched accuracy. | 80 | 1 |
| nx_bench_hll_vs_datasketches.nx | honest head-to-head. | 190 | 3 |
| nx_bench_ingest.nx | THE GENERIC BENCHMARK INGEST (rule 15 DRY; the second half of nx_bench_fetch). | 310 | 19 |
| nx_bench_intent.nx | intent/facet-aware retrieval metrics, bits-up. | 254 | 7 |
| nx_bench_intent_test.nx | KAT for intent/facet-aware metrics. | 76 | 5 |
| nx_bench_kg.nx | isolated bench: generic vs comb fixed-base k*G. | 55 | 3 |
| nx_bench_matmul.nx | NishiLang twin of bench/xlang/matmul.c. Memory-bound | 69 | 2 |
| nx_bench_metrics.nx | sovereign retrieval-quality metric kernels. | 207 | 11 |
| nx_bench_metrics_test.nx | KAT for the retrieval-quality kernels. | 106 | 2 |
| nx_bench_multiseed.nx | 20-workload accuracy averaging for OUR HLL + CPC. | 157 | 6 |
| nx_bench_popcount.nx | microbenchmark for nx_popcount64. | 96 | 1 |
| nx_bench_scale_dist_emit_test.nx | distribution measurement at 3 scales. | 197 | 8 |
| nx_bench_scale_test.nx | substrate scaling under increasing workload. | 168 | 4 |
| nx_bench_sign.nx | isolated ECDSA P-256 sign bench (generic kG path; | 51 | 3 |
| nx_bench_spectral.nx | NishiLang twin of bench/xlang/spectral.c (spectral-norm, | 124 | 7 |
| nx_bench_stats.nx | min/max/mean/p50/p95 stats over an i64 sample array. | 192 | 11 |
| nx_bench_stats_test.nx | smoke for the distribution-stats primitive. | 157 | 1 |
| nx_bench_stress_autoreg_test.nx | 10-iteration REAL transformer | 145 | 1 |
| nx_bench_stress_emit_test.nx | runs the stress workload AND emits | 141 | 1 |
| nx_bench_tsv_persist.nx | write bench TSV rows to a real file on disk. | 108 | 6 |
| nx_bench_tsv_persist_test.nx | smoke: write rows + read back + verify. | 141 | 1 |
| nx_bench_v1_vs_v2_test.nx | first paired substrate measurement. | 191 | 2 |
| nx_bench_verify.nx | isolated ECDSA P-256 verify bench (u1*G + u2*Q, | 58 | 3 |
| nx_bench_vs_c.nx | HORIZONTAL benchmark organ (Referee/Engineer-owned): races the sovereign | 78 | 5 |
| nx_bench_warc.nx | BENCHMARK CORPUS CAPTURE RUNNER: a URL-list file -> render-fetched -> ONE durable WARC | 269 | 9 |
| nx_bench_x25519.nx | isolated X25519 ECDH (variable-base) bench. | 43 | 3 |