if you are doing benchmarks using interpreted languages but care about smaller timescales perhaps your problem is using interpreted languages.