Dear All,
I just updated the "netfrc" option in pmemd.cuda (w/ -gti). The GTI implementation now supports netfrc=1 to remove the net force. Nevertheless, all current cuda "golden" baselines were done without removing the net force, i.e., netfrc=0 (since it was not implemented) BUT lots of "golden" baselines contain "netfrc=1".
As a result, when configured with -gti you will see lots of "failed" cases where the results differ from the baseline by "netfrc =1 vs 0".
Proposed solution: the "golden" baselines should be corrected w/ "netfrc=0" and the non-gti pmemd.cuda should be modified to always output "netfrc=0", to reflect the true test conditions.
If no one objects, I can do the proposed modification by end of this week.
Taisung
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Jan 24 2018 - 10:30:02 PST
