Steve Dekorte posted a comparison of vector performance between Python and his language, Io (a Self derivative).
The differences are absolutely stunning, but I'm not a numbers geek, so all I see are results that blow Python out of the water. I can't help but wonder how Perl would compare. Or C, C++ (with the proper vector libs), Fortran for that matter.
Some of the other statistics^Wpretty graphs and charts Steve uses to describe Io look pretty impressive, too.
Update: David Ascher points out that these statistics^Wobservations are only comparing the quality of the C-based inner loops (vector engines) in Io and Python (numarray in Python's case), and do not reflect on the languages as a whole. (Ditto Io / Perl / Perl + PDL.)
David continues:
The art of optimizing array operations is fairly well established. The challenge comes in the portability, memory behaviors, exception handling rules etc. that you may or may not want to follow to get the extra FLOPS. And that way lie hundreds of email threads.Choices that people argue vociferously about include: C or C++? Templates or not? Control over IEEE exception handling? Casting & coercion behaviors? Optimizing for what size arays? Optimizing for what data types? What code bloat to tolerate? etc.
