Crash during Spatial Smoothing

[JPoirier]

Hi,

NL crashes every time I try to spatially smooth a functional run (for a single subject). I've tried to SSmooth through the command line and via the NL interface as well. I can perform the action on the same dataset on another computer (running Tiger), but not on our main analysis computer (Snow Leopard).

Any ideas what the problem might be?

Thank you!
Josee

[rhoge]

Hi Josee,

Sorry about this crash, and thank you very much for posting the crash log as that makes it immediately clear what is going wrong.

Is your dataset a fairly large one? It seems that the program is not able to allocate the required memory and is thus crashing (it should handle this more gracefully).

As for why it should work on Tiger and not Snow Leopard, that is a bit mysterious but it may just have something to do with how memory is managed on the two operating system versions. You could try turning off 'force floating point output' and/or using a smaller kernel, but I kind of doubt this will help and you should of course be able to use any kernel you want.

You could also try quitting and re-starting NeuroLens, and then making sure that only the one input file is open when you run the spatial smoothing.

The odd thing is that your Snow Leopard analysis system appears to have a lot of RAM. The problem is that your version of NeuroLens is 32 bit, meaning that no matter how much RAM the machine has, you'll only ever be able to access 4 GB. A couple of years ago this was never a problem, but more and more we have very large datasets to work with.

I feel bad to keep saying this, but the unreleased NeuroLens 2 is a 64-bit program, which means that it should be almost impossible to run out of memory address space. We've tested it on insanely huge datasets and it works fine. Unfortunately I've had a very hard time getting that to a releasable state due to other commitments.

Hope this helps, and thanks again for reporting the problem.

Rick

[JPoirier]

Hi Rick,

Thanks for the reply.

I had guessed it was due to insufficient memory allocation, but was hoping there was a way around the problem (until NL2.0 arrives!).

I had tried turning off 'force floating point output', but that did not help. I had tried keeping quitting then only opening one window at once, and getting NL to cleanup data upon quitting, but that was also also overall unsuccessful.

As for your question, my functional datasets are about 285MB.

Thanks for the help and for a great piece of software!

Josee