[Octopus-devel] Re: [Octopus-users] gs in parallel

[Micael Oliveira]

Hi all!

  I don't think it is a good idea to have new flavors of the octopus 
executable. We already have enough problems with the existing ones!
  What I suggest is to print a warning message to the output if one 
tries to run a ground-state calculation using the mpi executable 
compiled without METIS.

Cheers,

Micael

Heiko Appel wrote:
> Hi,
> 
> On Thu, Dec 22, 2005 at 01:57:28PM +0100, Miguel Marques wrote:
> 
>>  Hi,
>>
>>
>>>Wouldn't it make sense if we make metis mandatory for the compilation 
>>>of the MPI version of the code? The people otherwise think they have 
>>>all they need to run in parallel and there's actually no error message 
>>>or warning that informs you that metis is required to run parallel in 
>>>domains - at least not while you are compiling the code ...
>>
>>I would like to be able to compile octopus without METIS. And let us not 
>>forget that METIS is *not* GPL... Although the code does not stop, it does 
>>state clearly in the section "Parallelization" how the code is running. 
>>What I did was to write a small paragraph in the FAQ explaining the problem. 
>>Please check it out...
> 
> 
> ok, I agree, especially with the GPL issue of metis. However, then
> I'd change the configure lines to avoid confusion. Having
> 
>  --enable-mpi
> 
> implies for me (and possibly for other people) a full-fledged MPI 
> executable which clearly has a dependence on metis. We could have, 
> e.g.
> 
>  --enable-states-only-mpi
>  --enable-full-mpi
> 
> or something similar to indicate from the start what one can expect
> from the resulting executable. This could prevent confusion.
> 
> best wishes,
> 
> 	Heiko
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Octopus-devel mailing list
> Octopus-devel at tddft.org
> http://www.tddft.org/mailman/listinfo/octopus-devel