Lower overhead of parallelism if possible #2
Description
Currently, the cutoff for getting performance gains is at about 20 microseconds (see #1) - although that's quite good, there are possibly ways to further lower the overhead by doing some performance optimisations.