John
Guillaume sent me new 2.6 kernel 32-bit X86 mex files
which I have installed and this resolved the situation.
If I now enter
>> spm_atranspa(1)
ans =
1
>>
but this may not be the test you want since the new Mex files are now
in place
regards
Neil
On 04/07/2008, at 9:59 PM, John Ashburner wrote:
> This was a test to see if SPM had been properly installed. Many of
> the SPM
> questions relate to incorrect installation, so I wanted to save time
> by
> having some of these questions answered automatically (especially
> the one
> about the problems with the default settings of WinZip).
>
> This piece of checking is to test whether there is an executable
> spm_atranspa
> that does not crash out. If there is, then it assumes that the mex
> files
> have been compiled for that platform.
>
> What happens if you simply type the following?
>
> spm_atranspa(1);
>
> If it does not appear to do anything, then things should be fine and
> I don't
> know why the code reported the problem. If it gives some error
> message, then
> there is something wrong with the installation and the code did the
> right
> thing.
>
> best regards,
> -John
>
>
>
> On Thursday 26 June 2008 05:47, Neil Killeen wrote:
>> Hello
>>
>> I run matlab on an SGI Altix IA64 platform. There is no native
>> implementation
>> for matlab for IA64 so it runs in the Altix's IA32 emulation layer.
>> I.e. this means
>> the Altix can run X86 Matlab binaries. When we start Matlab, we
>> provide the
>> -glnx86 argument to force the correct architecture.
>>
>> Now, we have been running SPM5 for some time on this platform. The
>> last patch I applied was # 573. This version of SPM5 starts up and
>> runs fine on the Altix with no modification.
>>
>> I have now just upgraded to patch # 1782. I believe the spm.m
>> startup
>>
>> script has become cleverer :-) It now triggers an errror:
>>>> spm
>>
>> ??? Error using ==> spm>check_installation at 1190
>> SPM uses a number of "mex" files, which are compiled functions.
>> These need to be compiled for the various platforms on which SPM
>> is run. At the FIL, where SPM is developed, the number of
>> computer platforms is limited. It is therefore not possible to
>> release a version of SPM that will run on all computers. See
>> /home/nkilleen/apps/SPM5/spm5/src/Makefile
>> for information about how to compile mex files for GLNX86
>> in MATLAB 7.5.0.338 (R2007b).
>>
>> Error in ==> spm at 308
>> check_installation;
>>
>>
>> On looking inside spm.m in the previous version. the piece
>> of code for checking the mex files does not exist and so this
>> error did not occur.
>>
>> The same range of Mex file architectures exists in the version # 573
>> release which ran happily. If I copy in the spm.m from Version #
>> 573
>> into the # 1782 system SPM does start OK (but I have not explored
>> extensively). If I comment out the Mex test code in spm.m in
>> the new version SPM does not start.
>>
>> I'm not sure exactly what it's unhappy about. Given that the error
>> does refer to GLNX86 (the correct platform) does that mean I should
>> explicitly have mex files of the form *.mexglnx86 ? SPM comes with
>> mex files of the form *.mexglx only (perhaps these are a synonym)
>>
>> I also wonder whether in # 573 we never tried to run a function
>> which uses a mex file; so there may be an undiscovered problem.
>>
>> thanks for generic and specific advice on this
>> Neil
>>
>> p.s. the kernel on the Altix is V 2.6 . To get a correct Mex file,
>> one must
>> compile on an X86 2.6 kernel machine and then transfer. Can someone
>> tell me what kernel the *.mexglx files were compiled with ?
>
|