Performance measurements (was: python on the smalltalk VM)

[Douglas Alan]

claird at starbase.neosoft.com (Cameron Laird) writes:

> Yes, I get a factor of between 100 and 300 for empty loops, but, as
> soon as they do anything useful, the multiple drops down to a range
> more like ten--definitely not a hundred.

Are you telling me that you can implement an FFT in Python (without
using any C extensions) that runs within an order of magnitude of the
speed that an optimized C implementation would?  I wouldn't be too
surprised if you could do one that runs 50 times slower, but I *would*
be highly surprised if you could get it down to 10.

Please don't make too much out of my statements -- I'm quite aware
that for many things the performance limitations are irrelevant, and
that when they are not that they can often be eliminated by using C
extensions or clever use of Python's built-in higher level data types.
The dream of the Self people, however, is that someday soon everything
can be done in-language in a very high-level, orthogonal, and dynamic
language without ever paying much of a performance cost.

|>oug