Re: ABF (NAMD 2.7x) speed ?

From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Tue May 25 2010 - 10:47:32 CDT

Hi again,

In principle, the new version is at least as fast as the old one. Did
you also benchmark the old ABF code on the same data with the new NAMD
binary? (the old scripts still work) What about serial performance
(i.e. is it a prefactor or a scaling problem)? If the problem is
confirmed, can you please send me some input data off-list, so that I
can investigate?

Thanks,
Jerome

On 25 May 2010 16:51, Ziemys, Arturas <Arturas.Ziemys_at_uth.tmc.edu> wrote:
>
> Hi,
>
> I'm testing NAMD 2.7x ABF job on Ranger (TACC) and was surprised by very slow performance of ABF calculation. The production run of the same system without ABF was 0.08 days/ns, while with ABF it is 0.47 days/ns (x6 time difference!). The older version of ABF was giving no noticeable difference on the speed. Is it known issue ?
>
> Arturas

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:10 CST