On Wed, Dec 12, 2018, Alfredo Quevedo wrote:
>
>I would like to compile Amber18 under KNL Xeon Phi architecture, with
>the aim of using sander (and its QM/MM capabilities). Reading the amber
>homepage (Intel support section) I understand that only pmemd can be
>used to take profit of Xeon Phi architecture? In that case, is there any
>workaround possible to run QM/MM calculations under Xeon Phi?
No. Note that the Intel-specific changes to pmemd are all devoted to
the MM part of the simulation. In QM/MM, it is quite common for the QM
part to be the bottleneck. So even if the XeonPhi capabilities in pmemd
were to be back-ported to sander, it would probably not help much.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Dec 13 2018 - 10:00:01 PST