Standalone Win32 wxPython apps?

Grant Edwards grante at visi.com
Sun Mar 10 20:19:43 EST 2002


In article <mailman.1015801697.3437.python-list at python.org>, Geoff Gerrietts wrote:
> Quoting Grant Edwards (grante at visi.com):
>> In article <1a9n8us1629mrvei40hdkuuerc1j6p318m at 4ax.com>, Courageous wrote:
>
>> I would think that for an isolated app, .so's would slow down
>> load times.  With a statically linked program all you do is a
>> single memmap() call (well, ignoring a few details).  With
>> .so's you've actually got to resolve all of the symbols and do
>> the linking step (not required for statically linked images).
>> 
>> Since you've got the same number of bytes to swap in off disk,
>> that should be a wash.
> 
> I can't speak for the windows world, but in the Linux world, .so's end
> up saving some time because the dynamic linker doesn't actually move
> things into the program's address space until needed -- it does this
> on a page-by-page basis.

That is true for all text pages.  It doesn't matter how they
are linked.

> Consequently, less start-time latency at a slight cost when
> you use the feature.

I don't see why.

> My data comes from: http://www.suse.de/~bastian/Export/linking.txt

I just read it, and it does say that .so's are demand-loaded.
But, static images are also demanded-loaded.  It does show that
lazy binding alleviates the start-up penalty I had described.

-- 
Grant Edwards                   grante             Yow!  This is my WILLIAM
                                  at               BENDIX memorial CORNER
                               visi.com            where I worship William
                                                   Bendix like a GOD!!



More information about the Python-list mailing list