> in my experience numpy and pdl are pretty similar in speed and quite fast > when used properly. but in either case if you slip and use interpreted loops > instead of the vector ops, you'll get hammered. >
With PDL, I'm pretty aware (and I think the perl code looks strikingly different) when I use a loop instead of a vector op - how easy is it to 'slip' and use a slow loop in numpy? If the fast functions are not obvious, or numpy does some optimization behind the scenes, then you could argue that the test Jarle did is a fair one... Matt _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
