Re: AMBER: benchmark walltime to ps/day calculation

From: Andrew Ring <aring.berkeley.edu>
Date: Thu, 31 Jan 2008 13:16:22 -0800

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

In terms of nstlim and dt, would that be:

ps per day = (nstlim * dt)/(total time of job in seconds/86400)

I ask because this contradicts the posting from 2005 and my usage of
nstlim and dt could be wrong.

Thank you.

Steven Winfield wrote:
> Hello
>
> How about:
>
> (Number of picoseconds your simulation ran for) / (Number of days
> it took to run)
>
> Steve :-)
>
> Andrew Ring wrote: Hello All,
>
> Could someone provide a simple equation for calculating ps/day from
> benchmark results. I am failing to understand the two
> explanations I have located.
>
> One post from 2005
> (http://amber.ch.ic.ac.uk/archive/200508/0205.html) on the Amber
> mailing list contains the equation: ps per day = total_time *
> nstlim * dt / 86400 This equation gives me numbers which are
> unreasonably low and vary directly with total process time (higher
> wall time results in higher ps/day).
>
> Unfortunately, I am missing something important to understanding
> the explanation at (http://amber.ch.ic.ac.uk/amber8.bench1.html):
> The value of ps per day = 86.4/(seconds per step), for a time step
> of 1 fs. Double this value if you plan to use 2 fs time steps.
>
> Thank you, Andrew Ring
>>
- -----------------------------------------------------------------------
>>
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
> -----------------------------------------------------------------------
> The AMBER Mail Reflector To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHojqmlHj4gJAUE8oRAsuiAJ9VPL2ABUIKVxljsAYVrx3ZBYMhcQCgrSNf
3QFojV1XBACWOt4o5XjW/Ps=
=wSHj
-----END PGP SIGNATURE-----

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Feb 03 2008 - 06:07:20 PST
Custom Search