Use AI to analyze performance bottlenecks #16
Description
The focus has always been to handle operations on large arrays as quickly as possible. I have learned a few things since I first wrote Polymath. For example, there are places where I check for a zero in a denominator before I divide, where it would be faster to divide and catch the DivideByZero exception. This gets tricky though because sometimes this shows up as a warning, not an error, and other times the array silently fills with infs or NaNs instead. (I don't want to see infs or NaNs in the data arrays; that's what the masks are for.) I bet there are also many other places where performance of the low-level code could be improved.