back
1 comments
Good to see you following up on this and tweaking it some more.

Having done this kind of thing a few times in the past I have to ask, which line in your test harness actually checks that the end result vector is actually sorted?

If thee is a small error in splitting and merging it might only throw an out of sequence element on a rare occasion - in such circumstances it might be wise to log the seed value used for the "random" sequence so it can be retraced and debugged.

If you're after further challenges have a shot at cranking out the largest number of uniformly distributed random points in a 3D unit sphere (or generalise to a unit N-sphere) and look at testing both the speed and whether they are in fact uniformly distributed at all scales.