ndPython: I NEED TO TALK WITH ONE OF THE PYTHON CORE DEVELOPERS

I need help with an implementation of your interpreter under Nanodesktop PSPE/PSP. I'm working to a version of Stackless Python interpreter (2.5.2) called ndPython. It is able to work under PSPE and PSP and it shall be the most powerful interpreter ever realized before on these platforms. I need to know something about the C functions that are recalled by the interpreter when it executes a .pyc file. Our version of ndPython is very slow in execution respect to Carlos's StackLess Python (another product for PSP). We believe that the trouble is in any routine of our Nanodesktop libc that could be a bottleneck. But we don't know which can be the interested routine (string ? memory allocation ?) Can you tell me which modules provide to decode and execute a .pyc file ? The number of modules that are statically linked in the interpreter can affect its performance ? (for example, for a larger number of strcmp between the keywords). The execution of a .pyc file is a memory allocation intensive task ? I've already enable the PYMEM option and I've see that the speed is improved (but Carlos's StackLess Python remains faster). We've done these tests: http://www.psp-ita.com/forum/viewtopic.php?t=28323&start=20 from ndpsp import * x = time.clock () for i in range(1000): pass print >> clock, "1.", time.clock()-t We've the results: stackless: 0.0037279 ndPython: 0.1015625 We don't understand why ndPython is so slower. Please, help us. The product is complete, but we cannot release cause this problem. Thank you in advance. Filippo Battaglia -- Caselle da 1GB, trasmetti allegati fino a 3GB e in piu' IMAP, POP3 e SMTP autenticato? GRATIS solo con Email.it http://www.email.it/f Sponsor: Vieni in vacanza nei Riccione Family Hotels! Ti aspettano servizi per bambini, mini club e tanto divertimento! Prenota on-line la tua vacanza! Clicca qui: http://adv.email.it/cgi-bin/foclick.cgi?mid=9206&d=24-6

SUBJECT LINES IN ALL CAPS CREATE NEGATIVE IMPRESSION Filippo Battaglia wrote: [snip questions I cannot answer] Question about existing, older Python versions should be directed to python-list, at least for a first try. Stackless Python is not produced by PSF core developers. It has third-party alterations to the core that might affect the answers to your question.
You should mention that the above is Italian, not English. There once was, and maybe still is, a python discussion list in Italian.
from ndpsp import * x = time.clock () for i in range(1000): pass print >> clock, "1.", time.clock()-t
When you post code, cut and paste the actual code that you ran. The above code will not run. Line-endings are missing. t is not defined. Perhaps you meant 't = time.clock()'. Good luck with your project. Terry Jan Reedy

The OP was redirected here from the stackless list since his questions were not stackless specific. Stackless python is in sync with the latest 2.x, 3.x branches as well as the trunk so performance problems he may with performance in general are perhaps best resolved by this lot. Kristján
-----Original Message----- From: python-dev-bounces+kristjan=ccpgames.com@python.org [mailto:python-dev-bounces+kristjan=ccpgames.com@python.org] On Behalf Of Terry Reedy Sent: 24. júní 2009 16:10 To: python-dev@python.org Subject: Re: [Python-Dev] ndPython: I NEED TO TALK WITH ONE OF THE PYTHON CORE DEVELOPERS Question about existing, older Python versions should be directed to python-list, at least for a first try. Stackless Python is not produced by PSF core developers. It has third-party alterations to the core that might affect the answers to your question.

Kristján Valur Jónsson wrote:
The OP was redirected here from the stackless list since his questions were not stackless specific.
Ok. He should have said so. Third party directions are not always right. As it turns out, he did post to python-list but ignored the rather strong hint about not shouting in the subject line.
Stackless python is in sync with the latest 2.x, 3.x branches as well as the trunk so performance problems he may with performance in general are perhaps best resolved by this lot.
He said he is using Stackless with the obsolete 2.5.2 release, not even the final and best 2.5.4 release with whatever bug fixes and speedups it has. Switching to the latter, at least, would be my first suggestion. Terry

Filippo Battaglia <pegasus2000 <at> email.it> writes:
http://www.psp-ita.com/forum/viewtopic.php?t=28323&start=20
from ndpsp import * x = time.clock () for i in range(1000): pass print >> clock, "1.", time.clock()-t
First, it has been noted that your message would better be sent to python-list. You must provide a working code example that doesn't need any "ndpsp" import to work. Also, since it is about porting Python to a particular platform, please enlighten us as to what type of CPU it uses, what compiler, what optimization options were used etc. As a sidenote, a common way of timing small snippets of Python code is by using the "timeit" module. From the command line, you could write: $ python -m timeit "for i in range(1000): pass" 10000 loops, best of 3: 87.2 usec per loop If you don't spend at least a bit of time writing your message clearly and giving enough details, nobody will spend their unpaid voluntary time trying to understand what happens. Regards Antoine.
participants (5)
-
"Martin v. Löwis"
-
Antoine Pitrou
-
Filippo Battaglia
-
Kristján Valur Jónsson
-
Terry Reedy