AMBER: rbornstat average and fluctuations incorrect when nrespa != nrespai

From: Ingvar Lagerstedt <ingvar.clearspeed.com>
Date: Fri, 20 Oct 2006 17:17:08 +0100

I think I have found a bug for the rbornstat option when nrespa !=
nrespai. I noticed that I got NaN's for the fluctuations and that the
averages was sligthly off in some cases, including the gb_mb test case.
Rerunning with fewer steps and the average can become large than the max
value. I traced the problem to that the value of tspan matched the
nrespa setting rather than the nrespai setting when writing the summary
statistics in runmd.f

The following is a suggested fix:

diff runmd.f.Amber9 runmd.f
157c157
< integer nvalid 
---
 >    integer nvalid, nvalidi
459a460
 >    nvalidi = 0
1967a1969,1971
 >
 >    ! added for rbornstat
 >    if (mod(irespa,nrespai) == 0 .or. irespa < 2) nvalidi = nvalidi + 1
2349a2354,2356
 >
 >          ! Born radii stats collected every nrespai step not nrespa step
 >          tspan = nvalidi
-- 
Ingvar Lagerstedt, Scientist
ClearSpeed Technology plc, www.clearspeed.com
Tel: +44 117 317 2055, Fax: +44 117 317 2002
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Oct 22 2006 - 06:07:26 PDT
Custom Search