Re: FPU problem, colinux+oM 2.4.26-1

The version of colinux i'm using is colinux 0.6.1 released on 2004-05-29 as
on
       http://sourceforge.net/project/showfiles.php?group_id=98788
and I was compling it from the source. For the configuration of kernel, if I
turn
of the oM support, with everything else the same, the NaN/gettimeofday
problem
will be gone.

On a change log it said there was a bug fix about floating point status. So
I want to find some later
versions of snapshot of 0.6.1, to find out what the fix is. But they were
removed from colinux.org.

Yes, using kernel 2.6 should be better, I'd like to wait for the
auto-migration function to be ready.
For now I have to be back on my simulation itself.


Thanks
Yasheng

On 2/10/06, Andreas Schäfer <[email protected]> wrote:
>
> On 18:19 Fri 10 Feb     , Yasheng wrote:
> > This problem was reported by ckauhaus,
> http://cluster.inf-ra.uni-jena.de/harpy/ticket/6.
> > But it is not a colinux bug, it is a colinux+oM bug. If one run colinux
> without oM, there
> > will be no problem.
>
> Really? As far as I can recall we had these errors due to an elderly
> coLinux version. Did you use the same coLinux versions (e.g. different
> snapshots) for both tests?
>
> > The reason could be, colinux by itself  protects the FPU status
> properly. But somehow
> > oM will break this protection. I guess by oM itself there should be no
> such problem,
> > otherwise it should be found already.
>
> That again sounds possible...
>
> > So, the conclusion is, colinux+oM 2.4.26 simply doesn't work, or could
> someone
> > figure out how to fix the FPU status?
>
> ...maybe Tab knows. Otherwise I'd suggest you the 2.6 in the long
> run. Do you plan using more than the couple of nodes you're currently
> using? If not, have you considered replacing coLinux by VMWare Player?
>
> Have a nice weekend!
> -Andreas
>