Python Performance
Markus Kohler
markus_kohler at hp.com
Tue Jul 27 06:28:12 EDT 1999
Skip Montanaro wrote:
>
> I got a bit confused attributing this message. I think we have too many
> Markus's running around loose... ;-) I think I have the quoted text
> attributed correctly. Apologies if not.
>
> Stenberg> Yes that is much too slow. Good OO style requires a lot of
> Stenberg> function calls. I'm sure Visualworks a commercial Smalltalk
> Stenberg> implementation can do millions of calls on this machine. Ok
> Stenberg> it's a JIT and it's commercial but's still Smalltalk and
> Stenberg> therefore at least as dynamic as python. I guess even Squeak
> Stenberg> (www.squeak.org) would be faster. If you would sent us code, I
> Stenberg> could prove it.
>
That's me (Markus Kohler).
> Better yet, since the Python source code is available, you could prove it by
> speeding up the interpreter. Many people have optimized different parts of
> Python over the lifetime of the software. It's significantly faster than it
> was a couple years ago.
>
> I think one of the impediments to improving the speed of the language has
> been the dramatic performance improvement in hardware. There's not a lot of
> incentive to pour lots of time into speeding up method calls when for a few
> hundred bucks you can simply drop in a faster processor, disk or networking
> card.
Also it's true that hardware still improves, it's also true that if I would try
sell
an application to a customer with thousands of users and I would tell him to buy
a
new machine because I use a slow interpreter, they would probably kill me.
There are also applications where every second you save is worth a lot of many.
Production testing of IC's comes to my mind ...
>
> Kohler> P.S. This might surface as one example for 'why scripting
> Kohler> languages don't always cut it' in my master's thesis; if I have
> Kohler> made some grievous error(s?), corrections are welcome.
>
Even it was not me saying this I would still like to add a comment.
> I don't recall anyone ever saying scripting languages would always cut it.
Agreed, but they should cut as much as possible shouldn't they ?
>
> Kohler> P.P.S. Too bad Python doesn't have anything that creates _nice_
> Kohler> code, like Stalin for Lisp ;-)
>
> There is a substantially shorter history to Python than to Lisp. Give it
> time. In addition, Lisp has been knocking around the academic world for
> 40-odd years. With that many graduate students looking for a thesis topic,
> it was likely that some of those theses would focus on performance
> enhancement.
Yes there's hope that Python get's faster. I'm just want to avoid that people
repeat history by just saying "by new hardware if it's to slow".
Markus
--
Markus Kohler mailto:markus_kohler at hp.com
More information about the Python-list
mailing list