Re: [AMBER] getting MC barostat to print less

From: Smith, Louis <>
Date: Fri, 13 Oct 2017 05:14:25 +0000

Hi Jason,

I see from testing the local cluster's version that amber 16 doesn't, but amber 14 does. Happy to know it doesn't matter regardless. Thanks.

From: Jason Swails <>
Sent: Thursday, October 12, 2017 10:36 PM
To: AMBER Mailing List
Subject: Re: [AMBER] getting MC barostat to print less

On Thu, Oct 12, 2017 at 12:36 AM, Smith, Louis <> wrote:

> Hello people of Amber,
> I was wondering if anyone knows whether it's possible to get the MC
> barostat to print less. I've found that it records every attempt in the
> outfile, rather than summarizing the attempts every ntpr steps, which is
> less than ideal* when running relatively long trajectories using pmemd.cuda
> where less frequent updating was specified.
> Is there any way to reduce the volume of these messages in line with what
> ntpr specifies?
> Does writing them from a GPU run as frequently as once every 100 steps
> (the default value for mcbarint) negatively effect performance?

‚ÄčThe GPU doesn't do this calculation, the CPU does (it's a simple set of
multiplications). The effect on performance will be (less than)
negligible. In fact, the GPU code already downloads all forces (and
uploads them for failed volume change attempts), which will be MUCH slower
than writing a line of text in the output file.

FWIW, newer versions of pmemd.cuda don't print this information anymore
(what version are you using?).


Jason M. Swails
AMBER mailing list
AMBER mailing list
Received on Thu Oct 12 2017 - 22:30:02 PDT
Custom Search