[Stackless] Video on channels

Guy Hulbert gwhulbert at eol.ca
Tue Sep 11 22:30:22 CEST 2007


On Tue, 2007-09-11 at 14:04 -0600, OvermindDL1 wrote:
> I do not expect it to 'compete' with C++, (if you want speed you use
> C/C++, if you want ease of programming, use python, as you can see the
[snip]
> generator one was over twice the speed of the normal, and even then it
> was still over 100 times slower then the C++ version).

This seems to contradict your first sentence ;-)

> 
> Just the difference of a small expected difference to something that
> is astronomical.  I have been using python for years, but I have never
> done number crunching in it; I use it by embedding it into my programs

Nasty surprise, eh ?

> for scripting capabilities and workflow, not speed.
[snip]
> > no chance in competing with C/C++ (based on experience with perl,

I got a 300 x improvement rewriting someone else's perl program in C ...
reduced run times from 24 hours to 5 minutes.  I was quite surprised.
I think the difference in my case was due almost entirely to the
overhead using arrays (which are linked lists of scalar objects in
perl).

I believe that NumPy (just looked at their site) is intended to solve
this kind of problem and I wonder if it works with stackless.  I'm
guessing that it may not since they are still sponsoring google
summer-of-code projects (for 2006, at least) but, then again, from what
I've read about stackless perhaps I'm wrong.

Hopefully someone will clarify this ...

>  which,
> > of course, is not quite the same thing as python :-).
-- 
--gh






More information about the Stackless mailing list