Olm-e said :
> ok, thanks for the rather clear answer ...
> 
> from experience, I only have tested and witnessed (huge) difference in
> video compression time (i.e. with ffmpeg2theora) on 64bit compared to
> 32bit installations ... on sound parts, I must admit I have not realy
> made tests to compare...

Ok, could you make some benchmarks if you have time?
like... for example how ffmpeg2theora, ffmpeg, oggenc, compare on the
same 64bits machine with:
- amd64 kernel on a 386 Debian OS
- amd64 kernel on a amd64 Debian OS


> So we will wait for the things from Karsten to happen, of course it is
> not so urgent I guess, and we'll install 32bit versions on thoses system
> for now...

I noticed that there is a big difference already when you use a kernel
correctly tuned to your CPU.
We're waiting now for 2.6.29 to be final, and we'll provide a new RT
kernel based on it. We will provide at least a 686 version and an 
amd64 version.


> We'll stay the eyes open for possible code sprint funding,
> (did you tried http://www.nlnet.nl/ ? )

We did not try, that could be a good idea if we find enough partner
organisations to work on this.

a.



> 
> thanks for all,
> 
> Olivier.
> 
> 
> Aymeric Mansoux wrote:
> > Hello Olivier,
> >
> > Olm-e said :
> >   
> >> we are a bunch of people here in brussels wanting to get pure-dyne on
> >> 64bit systems. (we are "organising" in a "brussels audio linux user
> >> group" involving the 2 independant radio "radio panik" and "radio
> >> campus" as well as ConstantVZW and Okno.be f.ex. )
> >> could you tell how this would be done ?
> >> is there some script around that we could modify to have it build ?
> >> would one 64bit machine put on a fix IP here be of some help for you to
> >> make it happens (if of course you have the time/motivation) ?
> >>
> >> in short : how can we help on that part ?
> >>     
> >
> > Ok, 1st of all, pure:dyne as it is now, in its 32bit version, runs
> > perfectly fine on intel 64bit CPU with multiple cores. We have also
> > started working on new kernels that will be available soon and that take
> > advantages of features from more recent CPU, while running in a 32bit
> > system.
> >
> > When it comes to having a full 64bit system, karsten is trying one at
> > the moment, and to be honest without any proper benchmark, it is hard to
> > tell the difference. Few applications take advantage of it and the gain
> > is more about precision rather than noticable performance boost. 
> > On the other hand, beyond the 64bit aspect, packaging for such a
> > platform would mean that we can directly compile with better GCC
> > optimisations, for example, and this "could" (I said could :) in some
> > cases lead to noticable performance gain. That why we are interested in
> > this.
> >
> > In terms of infrastructure, thanks for your offer. As it is now, we have
> > already everything we need to make some tests, but what is really
> > missing is time and people to look into it. Basically what we would
> > really need is funding so we can organise a code sprint and work on this
> > to get the ball rolling. We have the experience and we've been talking 
> > with 64Studio for a while now about merging some efforts, so that would
> > be a good occasion to finally get together and see what can be done.
> >
> > Otherwise, as it is now, the roadmap for the next milestone is to
> > consolidate what we have done so far, make the live instalattion easier,
> > USB boot on mactel, package different kernels optimised for different
> > intel CPU and work on all kind of things related to workshops.
> >
> >
> > a.
> >
> >
> >  
> >   
> >> thanks,
> >>
> >> Olivier M.
> >> ogeem.be
> >> okno.be
> >>
> >> ---
> >> [email protected]
> >> irc.goto10.org #pure:dyne
> >>
> >>     
> >
> >   
> 
> 
> ---
> [email protected]
> irc.goto10.org #pure:dyne
> 

---
[email protected]
irc.goto10.org #pure:dyne

Reply via email to