How to speed up calculations

[Superharz]

I have two large nested awkward arrays a and b.
I want to calculate a - b.
This operation takes about 30Gb of my RAM. However, the CPU is only at 10% and it seems like only 4 of my 16 logical cores are used.
My question: Is it possible to speed up the calculation and if yes: How?

[jpivarski]

Awkward Array does not automatically parallelize, and since most operations do not release the Python GIL (yet), a single process is not going to be able to take advantage of multiple cores.

To answer your question, I'd have to know a lot more about what a, b, and - are (if this is not using 100% of one CPU, they're probably not numbers, and so - is some overloaded behavior). However, I would guess that a and b are being lazily loaded from a file to use so much memory and not be CPU-bound.

[Superharz]

Thanks for your explanation.
I think you are right. The array I used was too big for my RAM. I found this out by flattening the the array with axis = None.
I tried it again with smaller arrays a and b and flattened them to load them completly into my RAM. The later substraction a - b happened almost instant.