I did help some friends generate such sets - using the tool linked a patch power of around 0.65 seemed pretty useful for that purpose, and avoided having to manually cull patches with the inherent annoyance / chance of mis-step. Maybe you might consider a simple patch power adjustment option for the cube patch list.But as with many of our approaches, there are no set rules.
For example, when performing MPVM probe matching, a patch set with an inverse gamma applied (so there are less dark patches) can be advantageous.
Fair enough, I don't know exactly how your code works, though given the end result is the same I'm not sure it much matters (I worked out your method by analysing a few patch sets and coming up with an algorithm that came up with the same result.). I think I did the "rounding" by casting a float to an int.And we define the patch colours in 8 bit, we do not round down.
Cool, look forward to see it. Lumagen it seems only do 8 bit patterns, but I recall my Accupel does 10 bit.In future iterations of ColourSpace we will add 10 bit, and likely 12 bit native values for patch generation.