Li Benchmarks

matmul_naive greenvalidity pass

Tier 1 · wall_time · N=256 · OS linux

passpass — perf claims allowed when ratio is green

Source: latest.csv:perf_present

Wall-clock row exists only — no explicit correctness signal yet.

Ratio vs best competitor: 0.757 (75.7% of rust speed) (table SOTA: rust — best competitor; charts use rust = 1.0)
Best competitor value: 0.0106 s

No analytical verify rows in ingest yet. Run tier-1 with --verify so latest.csv exports verify_ulps / verify_within_1ulp.

Problem size: N=256 (family matmul_naive)
Category: micro
Pillar: numerics
Package: lic
Li / catalog oracle (mean ± σ): 0.014 ± 0.001079 s (n=73) / 0.012 ± 0.001304 s (n=61) (cpp)
Ratio vs catalog oracle: 1.1667×
Best competitor: rust (0.0106 s)
Li relative speed vs SOTA: 0.757 (1.0 = rust speed)
Validity: validity pass
Threshold: 1.2×
Source: lic/benchmarks/workloads/tier1_micro/matmul_naive

PH ids: PH-5b, PH-7e

Compare oracle: cpp

Relative speed vs best competitor (rust) — SOTA = 1.0, higher is better. Absolute s values are in the table below.

Language comparison
Lang	Mean ± σ	Runs	Unit	Variant	OS
li	0.014 ± 0.001079	73	s	release	linux
cpp	0.012 ± 0.001304	61	s	release	linux
rust	0.0106 ± 0.000767	102	s	release	linux
julia	0.0111 ± 0.001013	93	s	release	linux
Metric: wall_time (value = mean of timed runs)