Re: Some performance notes

[Tim Janik <timj gtk org>]

On Wed, 8 Aug 2001, Johannes Stezenbach wrote:

> Tim Janik wrote:
> > 
> > On Mon, 6 Aug 2001, Johannes Stezenbach wrote:
> > 
> > >   int i;
> > >   GObject *obj[10000];
> > >   for (i = 0; i < 10000; i++)
> > >     obj[i] = g_object_new (G_TYPE_OBJECT, NULL);
> > >
> > > takes about 20 msecs. Not too bad ;-)
> > >
> > > The same using
> > >
> > >     obj[i] = g_cim_list_new ();
> > >
> > > takes 700 msecs
> 
> > don't let the code's complexity fool you, it's written to handle unlimted
> > numbers of construct and non-construct parameters, however the
> > normal case is none or 1 construct property per object, so that
> > code like:
> >       if (pspec->flags & (G_PARAM_CONSTRUCT | G_PARAM_CONSTRUCT_ONLY))
> >         {
> ...
> > 
> > usually boils down to a NOP.
> 
> 
> I created a stand-alone test program which shows that either using two
> G_PARAM_CONSTRUCT_ONLY or passing two property arguments to g_object_new()
> slows down object creating by a factor of 16.
> (Creating 100.000 objects takes 1 sec. vs. 16 sec. on my PIII 733.)
> 
> I profiled it using gprof and attach relevant excerpts from the
> profile results.

ok, we should really rerun that test after:
1) pspec-pool-lookup indexes by quarks
2) g_object_new_valist eliminated an extra per-property call to
   pspec-pool-lookup
3) signal performance was improved
and maybe
4) TypeNode.supers[] became TypeNode* pointers.

currently this profile just confirms that we're slow in areas
where we already know that we're slow and can do better ;)

> 
> Regards,
> Johannes

---
ciaoTJ