I've gone back and done more testing, here are the relevant bits:
1) Tried adding 4Gb swap file, still got the crash
2) I'm running on EC2 (ubuntu intrepid)
3) I'm using FSL 4.1.4
4) Generating a fsf file the manual way with the Feat gui also doesn't
work on the EC2 server
5) The exact same files work on my mac, requires a total of < 3Gb for
swap + real mem
So, I think there's some weirdness with amazon EC2, or at least with
ubuntu intrepid there. It remains to try the more recent karmic
images. But I'm guessing this isn't solved with that. If anyone else
is interested getting FSL working on EC2 please let me know. In the
end, it'd be a pretty good value proposition.
While I'm at it, FSL tcl guis won't work with nx connections either
because of bizarre rgb map issues. If anyone's interested in fast gui
over nx, that's another thing to sort out (but e.g., fslview works
fine).
Cheers,
Dav
On Oct 18, 2009, at 12:11 PM, Dav Clark wrote:
> Hi there,
>
> I'm rather confused as to why this is happening. The design matrix
> looks fine, and I'm pretty sure I'm not actually running out of memory
> here - as it seems to be prior to film_gls doing anything substantial.
> Any ideas? I have auto-generated the feat file, but it loads fine and
> looks right in the Feat gui, and I've used the version that I saved
> after viewing in Feat, same problem.
>
> I attached the fslerrorreport output and my fsf file.
>
> Thanks!
> Dav
> <fsl_kDPfwx.gz><scan0-from-gui.fsf>
|