AMBER Archive (2006)

Subject: Re: AMBER: use_pme=0 problem

From: luckyang_at_gmail.com
Date: Thu May 25 2006 - 20:56:30 CDT


Dear Dr. Case,

Thank you very much for your help. I have a big system so that with/without
PME does affect the computer time alot. I speed it up alot by setting nffts
to 2, order to 3. I also tried to use eedmeth=5, but I find that using
eedmeth = 5 or 1 gives me slightly different Evdw. From the manual I thougt
this flag has nothing to do with VDW energy. Do you have any idea about
this?

Thanks,

Lu

On 5/25/06, David A. Case <case_at_scripps.edu> wrote:
>
> On Thu, May 25, 2006, luckyang_at_gmail.com wrote:
> >
> > I may have to keep the longrange VDW correction term. Does this mean I
> have
> > to turn the PME on? Is there a way to keep the longrange VDW correction
> > while turn off the PME? The difference of the computer time is huge with
> and
> > without PME for my systems.
>
> You would have to modify the code so that the vdwmeth correction is called
> even when use_pme is set to 0. Maybe someone on the list (e.g. Tom
> Darden)
> will volunteer, but you should not hold your breath.
>
> There should not be a big difference with/without PME. Since all your
> charges
> are zero, you can set up a very coarse PME grid (i.e. small values of
> nfft1,2,3), which should minimize the time spent in reciprocal
> space. Even
> with the default values, your timing comment is surprising (unless you are
> either using lot of parallel processes, or you have a very large unit
> cell...)
>
> ...dac
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu