[gmx-users] problems with gmxtest.pl

[Mark Abraham]

Bernhard Bandow wrote:
> Dear users and developers of gromacs,
> 
> following the installation of gromacs 4.0.3 the tests show some
> inconsistencies.
> The software was compiled with Intel 11.0.069 compiler using mvapich2
> Further options in configure were:
> --enable-fortran
> --enable-mpi
> --with-fft=mkl
> --enable-shared
> The system gromacs is intended to run on uses mpiexec to initiate
> message passing. So gmxtest.pl was modified at line 31 to:
> 
>   $mdprefix = "mpiexec -np $parallel"
> 
> The output in mdrun.out starts with
> 
>   -- NNODES=4, MYRANK=2, HOSTNAME=h01001
>   NNODES=4, MYRANK=0, HOSTNAME=h01001
>   NODEID=0 argc=1
>   NNODES=4, MYRANK=1, HOSTNAME=h01001
>   NODEID=1 argc=1
>   NNODES=4, MYRANK=3, HOSTNAME=h01001
>   NODEID=3 argc=1
>   NODEID=2 argc=1
>                            :-)  G  R  O  M  A  C  S  (-:
> 
> ...it seems to be a parallel run indeed.
> Nontheless grompp.out of the corresponding run contains the lines:
> 
> -------------------------------------------------------
> Program grompp, VERSION 4.0.3
> Source code file: statutil.c, line: 727
> 
> Invalid command line argument:
> -np
> -------------------------------------------------------
> 
> which is related to line 159 of gmxtest.pl.

Yep. Even the 4.0.4 version of gmxtest.pl as linked from here 
http://wiki.gromacs.org/index.php/test_set is not actually functional 
with GROMACS-4.0.4.

> Are there any comments to the usage and position of the -np argument
> under these circumstances in order to have more than an educated guess
> for that?

See http://wiki.gromacs.org/index.php/grompp#Parallel_calculations. 
There is no correct usage of grompp -np.

Mark