n0is Research
Working papers on the n0is C++ engine, the experiments that shaped it, and the methods used to measure it.
01 / Systems methods · Working paper 0.6
Tyler Garrett
A CSV load appeared 2.10× slower in C++ than in a Swift port. Under a shared harness and matched dictionary work, the recorded ratio was 1.08×. The paper follows what changed between those measurements.
02 / Systems methods · Working paper 0.2
Tyler Garrett
Across 120 paired blocks, Zig led the 1M first load and n0is C++ led warmed loads at both sizes. The 10M first-load interval spanned equal times. The paper plots every pair.
03 / Evaluation methods · Working paper 0.2
Tyler Garrett
Four workers finished batches sooner, while individual evaluations took longer. The paper plots all 48 timed rounds and shows how a second study changed the estimate for a fifth worker.
The papers sit alongside the engine's benchmark data and experiment reports.