The reason `fib` is a useful benchmark here is not because `fib` is an inherently parallel problem, but because the functionality is obvious and it's easy to pa

The reason `fib` is a useful benchmark here is not because `fib` is an inherently parallel problem, but because the functionality is obvious and it's easy to parallelise. Computing fib with memoisation from 1 will certainly be faster than (parallel) recursion without memoisation and data sharing.