Re: kaffe in sarge

[Guilhem Lavaux <guilhem@kaffe.org>]

Dalibor Topic wrote:

> Am Dienstag, den 22.03.2005, 22:15 +0100 schrieb Thiemo Seufer:
>  
>
> > Dalibor Topic wrote:
> >    
> >
> > > Am Dienstag, den 22.03.2005, 13:46 +0100 schrieb Arnaud Vandyck:
> > >      
> > >
> > > > FYI.
> > > >
> > > > A new kaffe upstream snapshot with debian support for mips and
> > > > mipsel has been uploaded. It seems it couldn't be built on mipsel :'(
> > > >
> > > > http://buildd.debian.org/fetch.php?&pkg=kaffe&ver=2%3A1.1.4.PRECVS9-1&arch=mipsel&stamp=1111460324&file=log&as=raw
> > > >
> > > > I contact upstream.
> > > >        
> > > Thanks for the quick notice. I'm cross-ing to all debian arches atm to
> > > fix build problems I failed to notice ;( Fixes for ia64 and mipsel are
> > > in CVs already, a fix for arm is in the works.
> > >      
> > I heard the mips/mipsel fix means to switch to a slow generic cmpxchg,
> > have I understood this correctly?
> >    
>
> Yes. We'd need someone to look over the mips(el) cmpxchg to make an
> adaptation for the new glibc-based locking primitivs. I've CC:ed
> Guilhem, who can supply the details. I think he needs a specific variant
> of cmpxchg which returns the old value.
>
> I'm looking forward to patches, in particular to get the jit flying
> again on mips ;(
>
> The new package should be uploaded today, thanks to Arnaud.
>
> cheers,
> dalibor topic
>
>  
Hi,

I need an inlined function which does this, according to katomic.h:

/* Atomically store NEWVAL in *MEM if *MEM is equal to OLDVAL.
  Return the old *MEM value anyway.  */

where MEM, NEWVAL and OLDVAL are parameters of that function.

If someone feels to do it (but it is not necessary as there are default 
routines in katomic.h), he/she may also implement atomic_increment, 
atomic_decrement using inline asm.

Cheers,
Guilhem Lavaux.