In trying to track down an issue with the permute_hfa() function.... In the perform_permutations() function defined within, I see the parallel apply is now at the top of the loop. This is fine if the number of populations is larger than the available cores, but usually this won't be the case. It'll be more efficient to move the parallel apply deeper into the loop, after generating the sets for each population. This is especially true for the normal use case where we only have one population.
Also, I think windows isn't working because of how threading, environments, and objects work. I commented a suggested TODO fix but haven't implemented it yet. Not sure it will work.
In trying to track down an issue with the permute_hfa() function.... In the perform_permutations() function defined within, I see the parallel apply is now at the top of the loop. This is fine if the number of populations is larger than the available cores, but usually this won't be the case. It'll be more efficient to move the parallel apply deeper into the loop, after generating the sets for each population. This is especially true for the normal use case where we only have one population.