On Thu, Oct 13, 2011 at 6:35 PM, Josmar R. da Rocha <bije_br.yahoo.com.br>wrote:
> Dear amber users,
> I got an error when I tried to compile the AmberTools 1.5 in a server with
> 24 GB ECC RAM and two Tesla C2070, after having patched it with the
> bugfix.all.
>
> make[1]: Entering directory `/usr/local/amber11/AmberTools/src/rism'
> cpp -traditional -P -DBINTRAJ -DRISM -DRISM_CRDINTERP -DRISM_DX
> -DRISM_LINPROJ rism_report_c.f > _rism_report_c.f
> gfortran -c -O3 -mtune=generic -ffree-form -o rism_report_c.o
> _rism_report_c.f
> cpp -traditional -P -DBINTRAJ -DRISM -DRISM_CRDINTERP -DRISM_DX
> -DRISM_LINPROJ safemem.f > _safemem.f
> gfortran -c -O3 -mtune=generic -ffree-form -o safemem.o _safemem.f
> _safemem.f:105.40:
>
> type(memTracker),private :: totalMem
> 1
> Error: Object 'totalmem' at (1) must have the SAVE attribute for
> default initialization of a component
> make[1]: *** [safemem.o] Error 1
> make[1]: Leaving directory `/usr/local/amber11/AmberTools/src/rism'
> make: *** [serial] Error 2
>
> The gcc version is 4.4.4 20100503 (Red Hat 4.4.4-2) running in Fedora 13
> x86_64.
> In the second server, that has 8 GB ECC RAM, the previous version of
> AmberTools (AT-1.4) was installed. I renamed the folder and so tried
> to install the new version and so the same error was appears. I think
> the problem is due to ECC memory, because we have workstations with
> Non-ECC memory and the AmberTools-1.5 + Amber11 were compilled without
> any problem.
> Do you have any guess on how can I solve it?
>
Try giving it the save attribute. That is, put ",save" before :: totalMem.
I'm not sure why your compiler is complaining about that -- I've used the
same version of the compiler (as well as PGI compilers and intel compilers)
and I've never seen that issue before.
Moreover, this appears to be an error in the compilation step, which, as far
as I know, is completely unaffected by ECC settings (ECC shouldn't impact
any typical operation).
I don't know the Fortran 90/95/2003 standard thoroughly enough to know what
is allowed syntax and what isn't, so I have no idea why your fortran
compiler requires the save attribute on that variable and no other fortran
compiler I've used has.
HTH,
Jason
>
> Thanks in advance
>
> Josmar R. da Rocha
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Oct 13 2011 - 16:30:04 PDT
