apparently the overhead of memoization (when it is threadsafe) is greater than the benefit of reduced work