From: Angelo Rossi (angelo.raymond.rossi_at_gmail.com)
Date: Mon Dec 26 2016 - 15:53:42 CST
Dear Jim:
Excellent! As they say in New York, I like it already.
By that, I mean the QM/MM new feature. The example that was included
worked "right out of the box" utilizing the NAMD/ORCA interface. This
really opens many new opportunities.
I do have one comment, however. ORCA, as it is distributed to the general
public, runs on a single SMP node. For example, I ran NAMD 2.12
(Linux-x86_64-muticore)/ORCA, and it worked fine.
But, what happens when one runs NAMD (Linux-x86_64-ibverbs)/ORCA? The QM
component would have to run on only one of the nodes in the selection.
Doesn't this eventually present a bottleneck situation where the MM
component waits for the QM to finish? I understand that is the case
always, but as more nodes are added, this becomes the dominant delay.
Other QM/MM codes may not have this problem (e.g. NWChem), since both QM
and MM components are parallelized (though the MM component not as well as
NAMD).
Help!
Warm regards,
Angelo
This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:22:46 CST