[gmx-users] invalid number of nodes
gmx3 at hotmail.com
Wed Aug 13 17:39:03 CEST 2008
> Date: Wed, 13 Aug 2008 17:06:34 +0200
> From: ckutzne at gwdg.de
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] invalid number of nodes
> Berk Hess wrote:
> > Hi,
> > In the CVS version the fourier grid is no longer required to be
> > divisible by the number of processors.
> > But in the other mail you had fourier_nz set to zero, which is complete
> > nonsens.
> grompp (at least 3.3.x) will use fourierspacing for each direction that
> has a fourier_n* value of 0, so why not set fourier_nz = 0?
Oops, I forgot that.
It got a bit late last night, I get I am still not fully recovered.
But the 1024 grid dimension is not required.
Separate PME nodes should be used, which means that one can probably
use the same fourier grid as for smaller numbers of processors.
Although the pme code still runs slightly faster when the grid dimensions
are a multiple of the number of pme nodes. But increases the grid dimensions
will often be more costly than just using the normal optimal grid dimensions.
> > From: regafan at hotmail.com
> > To: gmx-users at gromacs.org
> > Subject: RE: [gmx-users] invalid number of nodes
> > Date: Wed, 13 Aug 2008 14:12:04 +0000
> > Anyway, in the new version of Gromacs, the -np option of grompp does not
> > work, does it?
> The -np flag does not work any more in CVS since it is no longer needed.
> The atoms are now divided over the nodes in mdrun, in 3.3. this was done
> by grompp.
> gmx-users mailing list gmx-users at gromacs.org
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
Express yourself instantly with MSN Messenger! Download today it's FREE!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the gromacs.org_gmx-users