Hi - no, it sounds like things are running fine - I guess it is just a
function of the _very_ large number of inputs and regressors that you
have. At this point it's probably best to just let it finish, but I
can think of two possible ways to speed this up in future:
1. As you have a really beefy machine with lots of CPUs and RAM, it
would be worth installing SGE locally and then you'll get a speedup of
x8.
2. Your model would give very similar results if you did a fixed-
effects analysis for each subject separately at second-level (each
group of 4 inputs I guess) and then just did a simple ME group-mean at
the third level. This would be very similar but probably an awful lot
faster (though a lot more man-hour-work to setup!).
Cheers.
On 6 Dec 2007, at 23:00, James Porter wrote:
> Steve-
>
> I'm running FSL 4.0.2 on a KDE 3.3.1-9.el4 (Red Hat) platform. I
> have 8 2.66
> GHz Intel Xeon processors at my disposal. The flame processes that I
> have
> going are each utilizing 100% a given processor. I have 16GB of
> memory, each
> flame process is only utilizing around 0.2% of available memory. Any
> other
> particular info that might have diagnostic importance?
>
> -Jim
>
>
> On Thu, 6 Dec 2007 09:25:54 +0000, Steve Smith
> <[log in to unmask]> wrote:
>
>> Hi - because you have so many input images, FLAME is being run a
>> slice
>> at a time. However, you're right that it should not normally be
>> taking
>> this long to complete each slice, so yes it does sound like something
>> is up - possibly your computer doesn't have enough RAM etc.? If you
>> look at top again, is most of the time indeed being taken with
>> running
>> flame? How much %cpu is it getting, and how much RAM is it using?
>> What
>> hardware and OS are you using?
>>
>> One thing that might make a difference - in the latest patch
>> release -
>> i.e. FSL-4.0.2, there was a minor script change in FEAT which means
>> that the splitting of the data into individual slices is done just
>> once rather than many many times (thanks to Tim for suggesting this).
>> So if your processing is taking ages because of the file IO and
>> fslsplit rather than the actual FLAME program, you should see an
>> improvement if you download the very latest version of FSL.
>>
>> Cheers.
>>
>>
>>
>> On 5 Dec 2007, at 17:56, James N. Porter wrote:
>>
>>> Hello-
>>>
>>> I've set up a 3rd level analysis that has been running for 4 days
>>> now. I keep checking top, and the flame processes do restart
>>> approximately every 3-5 hours. So it seems that the iterations are
>>> moving along, but I've never had a 3rd level analysis last longer
>>> than half a day. Does 4+ days seem unreasonable for the following
>>> parameters?
>>>
>>> CPU: a pair of quad-core Intel Xeon processors at 2.66GHz each and a
>>> total of 16GB of RAM
>>>
>>> Inputs: 192 cope images, 4 each from 48 subjects
>>>
>>> Model: Flame 1, GLM set up as a paired samples design, a la the help
>>> page example, with EV 1 defining my contrasts of copes and EVs 2-25
>>> defining pairs of subjects to account for pair-by-pair
>>> intercorrelation (they were identical twins). My design summary can
>>> be seen here: http://www.tc.umn.edu/~norb0062/design.png
>>>
>>> Thanks for your input,
>>> --
>>> Jim Porter
>>> Graduate Student
>>> Clinical Science & Psychopathology Research
>>> University of Minnesota
>>>
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------------
>> =
>> =
>> =
>> =
>> =====================================================================
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|