[tc4v]

so a 5x speedup from using 8x threads? I am not convinced this couldn't be achieved with numpy+multiprocessing ( or even maybe threading sinc numpy releases the GIL I think)