On Tuesday 29 May 2001 11:50, Michael Matz wrote:
> Hi,
>
> On Tue, 29 May 2001, Waldo Bastian wrote:
> > > I never got around at
> > > implementing this thing last year, cause I wasn't able to convinve
> > > myself, that the number of reduced relocation really would be worth it.
> >
> > I predict that it will not influence the number of relocations :-]
>
> Well, I wasn't _that_ pessimistic ;)  There _are_ private methods and
> classes which are for internal purposes, and these can be hidden
> alltogether.  Esp. in KLM's representing whole applications most
> classes/methods are "internal".  If these symbols are made DSO-local ld
> can bind references to them at link time, which reduces the number of
> symbol-based relocations quite a bit, so KLM loading could be even faster.

If they are non-virtual they currently result in a R_386_JUMP_SLOT relocation 
which is being relocated lazy anyway. If they are virtual they are unlikely 
to be private, and even if they are and not exported I doubt that gcc will 
create an entry other than R_386_32 for them. (That would be interesting to 
know though)

Cheers,
Waldo
-- 
bastian@kde.org | SuSE Labs KDE Developer | bastian@suse.com