Although you can use numexpr for some easy (albeit modest) threaded speedups and streamlined memory usage. I've had more success coding custom array processing

Although you can use numexpr for some easy (albeit modest) threaded speedups and streamlined memory usage. I've had more success coding custom array processing routines in cython where I can easily exploit threads.