[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [ESPResSo] Non-cubic simulation box

From: Limbach, Hans Joerg, LAUSANNE, NRC-FS
Subject: RE: [ESPResSo] Non-cubic simulation box
Date: Wed, 5 Mar 2008 08:18:18 +0100

Dear Olaf,

In principle you are right and the P3M is already prepared for handling
non-cubic, rectangular systems (e.g. all (or at least many) quantities
are vectors with at the moment three identical values). On the other
hand one person would need to dig him/herself again in all the theory
and at the end make sure everything is correct. For the FFT I am not so
sure, because also the home made communication would have to be adapted
for three different directions (e.g. which 2 of the 3 you choose for the
2D domain decomposition?). And don't forget the tuning and the error
estimates. My guess is that the minimum requirement to change it would
be one person working fulltime for 3 month on it. 

Best regards,

> -----Original Message-----
> From: address@hidden 
> [mailto:address@hidden On Behalf Of Olaf Lenz
> Sent: mardi, 4. mars 2008 17:03
> To: Jacob Kirkensgaard
> Cc: address@hidden
> Subject: Re: [ESPResSo] Non-cubic simulation box
> Hash: RIPEMD160
> Post scriptum:
> As a colleague of mine has correctly commented, not all 
> algorithms implemented in ESPResSo do actually support 
> non-cubic, rectangular systems. In particular, the current 
> implementation of the P3M algorithm for electrostatics does 
> not allow for non-cubic systems.
> However, extending the P3M implementation to be able to 
> handle generic rectangular systems should be possible without 
> too much effort.
> Best regards
>       Olaf Lenz
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: Using GnuPG with Mozilla -
> iD8DBQFHzXK1tQ3riQ3oo/oRA9IcAJ0bW37+su0zw86eA7k1lWeFtTNs8QCfQaIe
> Z9sshDL0HdxNMknRpTAPy4w=
> =GQmA
> _______________________________________________
> ESPResSo mailing list
> address@hidden

reply via email to

[Prev in Thread] Current Thread [Next in Thread]