Hi,
Our main 64-bit development machines are:
CPU OS
Dual Core AMD Operon 2214 Centos 5
Dual Core AMD Opteron 280 Fedora core 5
Intel Xeon 2.8Ghz Fedora core 4
> From: "Wen, Wei" <[log in to unmask]>
> Date: 10 September 2007 11:35:48 BDT
> To: [log in to unmask]
> Subject: Re: [FSL] tbss_3
> Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
>
> Thanks Steve, it would be nice if I can have a list of your working
> combinations.
>
> WW
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf
> Of Steve Smith
> Sent: Monday, 10 September 2007 6:23 PM
> To: [log in to unmask]
> Subject: Re: [FSL] tbss_3
>
> Hi - we use several different hardware flavours of 64-bit linux
> desktop / cluster nodes and have had pretty much NO problems
> installing linux and getting FSL to work. If you'd like a list of the
> exact linux flavours we used we can send this info.
>
> Cheers.
>
>
>
> On 10 Sep 2007, at 09:10, Wen, Wei wrote:
>
>
>> Thanks Steve.
>>
>> You are right on getting a 64-bit machine. But to sort out hardware
>> compatibility issue is a struggle! Anyway, will take your advice to
>> get a
>> new machine.
>>
>> Thanks a lot!
>> WW
>>
>> -----Original Message-----
>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>> Behalf
>> Of Steve Smith
>> Sent: Monday, 10 September 2007 5:57 PM
>> To: [log in to unmask]
>> Subject: Re: [FSL] tbss_3
>>
>> Hi - yes and no:
>>
>> You should be able to make this work by only merging some of the
>> images at a time (also you might change the merge command to use
>> avwmerge++). But then you would need to run certain steps in parallel
>> and some with the data fully combined - you'll have to play quite a
>> bit with the tbss scripts to do this. At the end you will have to
>> combine all the data to do the stats in randomise, and you are very
>> likely to run into the same problem then.
>>
>> I would also suggest that you might cut down the field-of view once
>> everything is sampled into standard space - but given the huge number
>> of subjects that you have this probably won't cut things down enough.
>>
>> In short - if you're prepared to get stuck into amending all the TBSS
>> scripts you might be able to make this work - but might get stuck at
>> the end with the randomise call. (Note that the new version of
>> randomise, which would be compatible with all your preprocessing
>> using FSL 3.3, can be run to use a lot less RAM than the old version
>> - but I bet it still will require too much for this number of
>> subjects).
>>
>> In short - it will be much easier to just find a 64-bit machine -
>> they're not expensive and increasingly neuroimaging analysis needs
>> it!
>>
>> Cheers.
>>
>>
>>
>>
>> On 10 Sep 2007, at 08:29, Wen, Wei wrote:
>>
>>
>>> Hi Steve,
>>>
>>> Is there a workaround please?
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>>> Behalf
>>> Of Steve Smith
>>> Sent: Monday, 10 September 2007 5:16 PM
>>> To: [log in to unmask]
>>> Subject: Re: [FSL] tbss_3
>>>
>>> Hi,
>>>
>>>
>>> On 10 Sep 2007, at 04:24, Wen, Wei wrote:
>>>
>>>
>>>> Hi Steve and FSLers,
>>>>
>>>>
>>>>
>>>> When I was doing tbss_3, I had the following error messages (after
>>>> individual normalization). Is there a solution or we have to do it
>>>> with a 64-bit system due to the large file size it was trying to
>>>> generate?
>>>>
>>>
>>> That's right - that's the solution.
>>>
>>> Cheers.
>>>
>>>
>>>
>>>
>>>> (over 280 scans and each is about 14MB)? (I am using tbss v1.0 )
>>>>
>>>>
>>>>
>>>> Many thanks and regards,
>>>>
>>>>
>>>> Wei Wen, Sydney Australia
>>>>
>>>>
|