[Python-Dev] Performance compares (original) (raw)

Tim Peters tim.one@home.com
Sun, 20 May 2001 07:05:53 -0400

[M.-A. Lemburg]

... Running the same test for 2.1 vs. 2.0 there's not much to notice, so the important changes seem to be originating in the move from 1.5.2 to 2.0.

IIRC, Guido, Skip Montanaro and I put major effort into finding speedups for 1.5.2, and Fredrik did more independently (like inlining high-frequency int operations in the eval loop). Also IIRC, that's the last time any concerted effort was put into speeding Python. 1.5.2 was an efficiency peak, then, and unstable equilibrium never endures without deliberate and persistent rebalancing work. If Python were "a real product", it would be at least one person's full-time job to keep it in peak shape. But it's not even a part-time job for anyone, and I don't see that changing. In compensation, machines have gotten faster much quicker than Python has slowed.