This is the mail archive of the
guile@sourceware.cygnus.com
mailing list for the Guile project.
Re: Making Guile slower
Greg Harvey <Greg.Harvey@thezone.net> writes:
> Marius Vollmer <mvo@zagadka.ping.de> writes:
>
> > Is anyone worried about this change re performance?
> >
> > 1999-12-10 Greg Harvey <Greg.Harvey@thezone.net> (applied --12/10/99 gjb)
> >
> > * smob.c (scm_smob_prehistory): initialize allocated smob
> >
> > * tags.h: new tag: scm_tc16_allocated
> >
> > * gc.c (scm_gc_for_newcell): set the car of the new cell
> > to scm_tc16_allocated
> > * pairs.h (SCM_NEWCELL): set the car to scm_tc16_allocated
> > (scm_gc_mark): mark allocated cells.
> >
> > I am somewhat, because SCM_NEWCELL now seems to perform one more
> > operation.
>
> Did you even think about what this is accomplishing before you started
> worrying about the performance or are you now just on a hunt to show
> that we're trying to make guile perform poorly?
Errm, I wrote this hastily, and I'm sorry that I gave the impression
of wanting to cause trouble again. What really motivated me to bring
this up was that Mikael showed a slow down of Guile, and this was the
change that might have been a performance impact in my opinion.
Honest.
I really don't think that you guys do not care about the performance
of Guile. The reason why I continued this discussion so stubbornly is
that it wasn't clear to me that Greg B understood the point I was
trying to make (together with and inspired by Mikael). In my view you
took a chance with Guile's performance and stability but there wasn't
any real reason for that. Guile might not be `finished' the same way
that bash is finished, say, but it still is worth it to not
gratitiously risk the existing state, in my view.
[ It might have helped to quickly note in the ChangeLog that this was
a bugfix. You should also change scm_debug_newcell, I think. ]