Right - thank you.
You reminded me that the bottleneck is the non-bonded interactions.
Aggressive coalescence is what I’ll do.
> On Sep 24, 2016, at 3:26 PM, W. Andeas Svrcek-Seiler <svrci.tbi.univie.ac.at> wrote:
>
> Hi,
>> Aggressive coalescence will give smaller bond parameter tables but may
>> not run as fast because of poor cache behavior when the dynamics code
>> is running?
>> One set of parameters for every bond will give larger parameter tables
>> but might run faster?
> though I've been out this business for several years now, I'm positively
> sure that the speed impact of anyting having to do with bonds is
> negligible. Bonds time scales linearly with the number of atoms
> involved. Even if you manage to increase bonds-speed 10-fold the impact
> on total runtime will be negligible.
>
> best,
> Andreas
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Sep 24 2016 - 13:30:02 PDT
