Re: [AMBER-Developers] Current parallel test failures in the tree

From: Lin Fu <linfu3200.gmail.com>
Date: Wed, 24 Mar 2010 22:19:34 -0400

I also did this test, Mark is right, This neb test could also be passed with
64bit version of ifort 11.1 and mpich2-1.2.1 without any problem.

2010/3/24 Mark Williamson <mjw.sdsc.edu>

> Daniel Roe wrote:
>
>>
>> Out of curiosity does anyone know if this neb/vlimit stuff happens with
>> the
>> intel 11 compilers?
>>
>
> These specific tests all pass using a sander.MPI binary that has been
> compiled with ifort 11.1.064 . I've not tested any other v11 ifort
> compilers, hence I still stand by my conclusion that it is most likely a
> compiler bug, given that these tests also pass with compiles using gfortran
> 4.1.2, 4.4.3 and ifort 11.1.064 .
>
>
> But if one wants to look closer at this, one thing that does show up when
> compiling with ifort's runtime checking option ( see "FC=mpif90 -check all"
> in config.h) is the following:
>
> [mjw:test]$ make test.neb
> cd neb-testcases/neb_gb_partial && ./Run.neb_gb_partial
>
> Running multisander version of sander Amber11
> Total processors = 4
> Number of groups = 4
>
> forrtl: severe (408): fort: (3): Subscript #1 of the array NEB_NRG_ALL has
> value 0 which is less than the lower bound of 1
>
> forrtl: severe (408): fort: (3): Subscript #1 of the array NEB_NRG_ALL has
> value 0 which is less than the lower bound of 1
>
> forrtl: severe (408): fort: (3): Subscript #1 of the array NEB_NRG_ALL has
> value 0 which is less than the lower bound of 1
>
> forrtl: severe (408): fort: (3): Subscript #1 of the array NEB_NRG_ALL has
> value 0 which is less than the lower bound of 1
>
> rank 3 in job 28964 bunny.sdsc.edu_45545 caused collective abort of all
> ranks
> exit status of rank 3: return code 152
> rank 0 in job 28964 bunny.sdsc.edu_45545 caused collective abort of all
> ranks
> exit status of rank 0: return code 152
> program error.
> make: *** [test.neb] Error 1
>
>
> I've no idea if this has any relation to the "problem" being discussed, but
> it might be nice to resolve it before the release.
>
> regards,
>
> Mark
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Mar 24 2010 - 19:30:03 PDT
Custom Search