Benchmarking an old chess engine on new hardware — LessWrong