Re: [AMBER-Developers] Updated PIMD tests

From: Scott Brozell <sbrozell.rci.rutgers.edu>
Date: Tue, 18 Aug 2009 03:32:32 +0100

Hi,

Ahhh, I knew MR RUN THE TEST SUITE would have a good reason ;-)

All the test diffs looked like minor fp round off.
In no cases were there divergent results, but in many but not all
cases the initial energies differed in the 4th decimal place;
in a couple the initial kinetic energy differed in the 3rd dec place;
this may be the worst one:

./PIMD/part_rpmd_water/spcfw_rpmd.out.dif
52c52
< nstlim = 10, nscm = 999999999, nrespa = 1 
---
>      nstlim  =        10, nscm    =         0, nrespa  =         1
89c89
<  Etot   =     -1632.9189  EKtot   =       579.4651  EPtot      =     -2212.3840
---
>  Etot   =     -1632.9118  EKtot   =       579.4721  EPtot      =     -2212.3840
...
163,164c163,164
<  Etot   =        92.0704  EKtot   =        47.1693  EPtot      =        44.9051
<  BOND   =        42.5414  ANGLE   =         4.5016  DIHED      =         0.
---
>  Etot   =        92.0717  EKtot   =        47.1699  EPtot      =        44.9058
>  BOND   =        42.5421  ANGLE   =         4.5017  DIHED      =         0.
Naturally, I encourage Francesco and everyone else to scrutinize
the test cases.  In addition, to identify platform dependencies
more testing is needed; for the last few months nightly testing
has been absent for various reasons.
Scott
On Mon, Aug 17, 2009 at 03:17:18PM -0700, Ross Walker wrote:
> 
> Modified Files:
> 	spcfw_pimd.out.save 
> Log Message:
> Updated for this change; output from glenn with ifort:
> date: 2009/07/08 23:37:26;
> RCW: Set nscm=0 instead of just arbitrarily setting it to 9999999.
> Linux opt-login02.osc.edu 2.6.18-128.1.1.el5 #1 SMP Mon Jan 26 13:58:24 EST
> 2009 x86_64 x86_64 x86_64 GNU/Linux
> Intel(R) Fortran Compiler for applications running on Intel(R) 64, Version
> 10.0    Build 20070426 Package ID: l_fc_p_10.0.023
> 
> You updated the PIMD tests for the changes I made to nscm. The reason I
> never updated these is because all the test cases were failing for me with
> errors that were beyond what I would consider acceptable. Generally in the
> last few digits of the initial kinetic energy leading to a divergence in
> results. Francesco was going to look into this.
> 
> Did the test cases all pass for you with here with the exception of the
> printing of nscm?
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Aug 19 2009 - 23:22:18 PDT
Custom Search