Hi,
> we run spm2/spm99 just fine on Fedora Core2 64 bit (2 opterons) with
> matlab 7.0.1/SP1, a fairly modern combination. We had these problems
> with the renderer (GUI crashes), which were solved with Johns tops on
> replacing the 'painters' handle. See link below. Instability issues
> also can arise with incompatible binaries, that are written in C and
> have nothing to do with matlab. We recompiled our binaries with the
> appropriate flags for 64 bits opterons, and all went fine (I posted
> the makefile some time ago, others used it succesfully), they seem to
> be stable so far. Before that, with inappropriate compiler flags,
> there were stability issues. I do not have experience with Matlab/spm
> on Suse Linux, so perhaps someone else can comment on that, it might
> be harder than on RH/Fedora systems from what I read on the list
> recently.
I believe there's absolutely no reason why Red Hat / Fedora systems
would have more problems in general - not that I think Red Hat / Fedora
distributions are technically superior. There may be problems with a
specific release, but I see no reason why SUSE or Mandrakelinux would
behave better (or worse) in general, they use the same glibc libraries
after all. Of course glibc, kernel or X11 patches may be specific to
some distributions, but this can't account for problems with *all* Red
Hat / Fedora releases.
The problem is that we have something like 10 users running SPM
concurrently every day. That could be the problem here, maybe SPM is
used very extensively at our place. Otherwise I can assure you we have
Matlab and SPM installed properly now, so we don't have local
installation issues. Anyway, I'm happy to learn SPM99 runs just fine
under Matlab 7. I had reports of users complaining of very different
results when running SPM99 under different versions of Matlab. These
reports are still under investigation; since many other users seem to
have never noticed such issues, I hope we'll converge on a solution.
Again, these may be corner cases of degenerate data.
> I am not sure wether the community of users does a better job than
> the FIL in supporting matlab versions. With some tweaks (mostly to
> the binaries) it is usually doable. But perhaps an online archive
> with more packages for specific platforms/matlab combinations might
> be a good idea, with the appropriate binaries/gui tweaks. Not
> everyone is familiar enough with the command line to tweak spm. I'd
> be happy to post our package.
I don't understand. The FIL doesn't maintain SPM99 anymore, does it?
--
Dimitri Papadopoulos
|