[gmx-users] FFT plan mismatch with mdrun_mpi

[Mark Abraham]

On 4/06/2011 3:42 AM, Christian Blouin wrote:
> Hello, I'll first apologize in case that this is a common issue with
> an available solution, but I have a problem running mdrun_mpi on a
> rock cluster and there
> is no amount of searching that I can do to find information on this
> error (manual, online docs, general google search). Could someone
> point me in the right direction for documentations?
>
> The error is the following:
>
> Program mdrun_mpi, VERSION 4.5.4
> Source code file: gmx_fft_fftw2.c, line: 494
>
> Fatal error:
> FFT plan mismatch - bad plan or direction.
> For more information and tips for troubleshooting, please check the GROMACS
> website at http://www.gromacs.org/Documentation/Errors

I've never seen this one before.

> I'm trying to run a simple mdrun as per one of the tutorial online on
> 50 CPUs. I can provide more details, but I'm hoping that there is
> already an answer to this question somewhere.

I suspect most tutorials are based on systems that are too small to want 
to use this many processors. Parallelism is not a magic bullet that lets 
one run any calculation on any combination of processors. I'd suggest 
using fewer and hoping that solves the problem. Otherwise, trying 
linking with an up-to-date install of FFTW3.

Mark