[AMBER] DISAVE does not work on pmemd.cuda / GPU

From: Mykola Dimura <mykola.dimura.gmail.com>
Date: Wed, 16 Sep 2015 18:57:33 +0200

Dear Amber community,
I am using time-averaged distance restraints with Amber an have
recently found out that time averaging does not work with pmemd.cuda.
Is it intended?
First I ran a simulation with NMR restraints and a line "&wt
type='DISAVE', value1=10.0,value2=3" in the input file by pmemd.cuda.
Then I removed the DISAVE line and ran the simulation again with
pmemd.cuda. Results were completely identical.
To verify, I ran the very same simulations with pmemd.MPI and got two
different trajectories. Trajectory produced by pmemd.MPI without
DISAVE was similar to the one from pmemd.cuda.
Trajectory produce by pmemd.MPI with DISAVE was different, as expected.
I did not find any information on DISAVE support by pmemd.cuda on the
webpage nor in the manual, so I decided to ask here.

I also noticed that results produced by pmemd.MPI for different runs
with exactly the same inputs are similar, but not identical. For
example restraint energy after 2 ps was 20.53, 20.49 and 21.01
(kcal/mol) for three consecutive runs on exactly the same hardware
(desktop with a single i7-4930K CPU).
Best regards, Mykola Dimura.

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 16 2015 - 10:00:03 PDT
Custom Search