Re: [AMBER] Restart interrupted sander.MPI calculation

From: David A Case <>
Date: Sun, 22 Aug 2021 10:59:20 -0400

On Sat, Aug 21, 2021, Bruno Cramer wrote:

>How can I split an amber sander.MPI calculation in two steps (or more), the
>first step being the input of the second step.

Have the restart file from the first step become the input coordinate file
for the second step, and set ntx=5, irest=1 in the second run.

>If a submitted MD calculation is interrupted after e.g. 12 hrs, what can be
>done without starting all again?

If you set ntwr, you will save restart files at the frequency you request,
and these can be used for subsequent runs. But the trajectory files you get
might not be correct. My general approach is to break long simulations into
pieces of 12-24 hours. If one piece is unexpectedly interrupted, I just
start again from the beginning of that run.

But the approach to use depends on how often you expect to see unexpected
interruptions, and how you value the human time spent correctly piecing
together shorter runs, versus computer time of re-running an interrupted


AMBER mailing list
Received on Sun Aug 22 2021 - 08:00:02 PDT
Custom Search