Is extract_f2 maxmem per core? about admixtools HOT 2 OPEN

I clarified this in the documentation. The functions that have a maxmem parameter are not multi-threaded, so it's total memory, not per core. But the exact memory that will be used is not really capped by the value provided here, after I made some changes to the function without updating the calculation for how much memory will be used.

The maxmem parameter should still be useful in two ways:

1. If extract_f2 (or a similar function that used has the parameter) crashes because it runs out of memory, setting it to a smaller value than the default 8000 could solve the problem.
2. If the function splits the computation up into blocks and is slow, and you have lots of memory, increasing the value will speed the computation up.

from admixtools.

I remembered that I did indeed add an option to parallelize extract_f2() a couple of months ago following a suggestion from Benjamin Peter. In the latest version this is implemented so that the pairwise f2-calculations get split into several chunks, and the size of each chunk is proportional to maxmem/n_cores. So yes, maxmem is per core, but the exact memory used by each process is only proportional to maxmem/n_cores.

from admixtools.