Re: [AMBER-Developers] final (!?!) release candidate for Amber11

From: Andreas Goetz <agoetz.sdsc.edu>
Date: Fri, 16 Apr 2010 10:53:25 -0700

Hi Ben,

On Apr 16, 2010, at 8:28 AM, Ben Roberts wrote:

> On my Mac (Snow Leopard 10.6.2, Intel compilers 11.1.084), I get this now:
>
> Finished test suite for AmberTools at Fri 16 Apr 2010 11:10:02 EDT.
>
> 312 file comparisons passed
> 2 file comparisons failed
> 4 tests experienced errors
>
> The diffs are in the PM3 tests, which I can't recall seeing before:

These are new tests for sqm which I put in.

> possible FAILURE: check c2h2.pm3.go.out.dif
> /Applications/Amber11/2010-04-14/amber11/AmberTools/test/sqm/c2h2
> 23c23
> < Electronic energy = -12472.30940072 kcal/mol ( -540.83992024 eV)
>> Electronic energy = -12472.33189812 kcal/mol ( -540.84089580 eV)
> 24c24
> < Corcore repulsion = 6343.90221386 kcal/mol ( 275.09224292 eV)
>> Corcore repulsion = 6343.92470919 kcal/mol ( 275.09321839 eV)
> ### Maximum absolute error in matching lines = 2.25e-02 at line 23 field 4
> ### Maximum relative error in matching lines = 3.55e-06 at line 24 field 4
> ---------------------------------------
> possible FAILURE: check ch3f.pm3.go.out.dif
> /Applications/Amber11/2010-04-14/amber11/AmberTools/test/sqm/ch3f
> 24c24
> < Electronic energy = -25371.26965143 kcal/mol ( -1100.18080965 eV)
>> Electronic energy = -25371.37839156 kcal/mol ( -1100.18552498 eV)
> 25c25
> < Corcore repulsion = 11412.26897811 kcal/mol ( 494.87311817 eV)
>> Corcore repulsion = 11412.37771890 kcal/mol ( 494.87783352 eV)
> ### Maximum absolute error in matching lines = 1.09e-01 at line 25 field 4
> ### Maximum relative error in matching lines = 9.53e-06 at line 25 field 4
> ---------------------------------------
> Maybe these are some kind of machine rounding error, but the difference seems a bit big for that. Has anyone else seen it?

Thanks for reporting. These differences are not worrying. What matters is the total energy. We will have to tighten the geometry convergence criteria or exclude the individual energy components from testing. I will test intel compilers under linux to see what I get and get back in touch with you.

All the best,
Andy

>
> The failures were in the rism1d tests, which I persuaded to work by applying Jason's patch. But rism1d/spc-nacl then threw a complaint:
>
> [roberts.qtpclnt070 spc]$ less spc-nacl.xvv.dif
> 22884c22884
> < -0.22768338E-10 -0.17071213E-10 -0.11363355E-10 -0.56477011E-11 0.72786318E-13
>> -0.22768338E-10 -0.17071214E-10 -0.11363352E-10 -0.56476964E-11 0.72793957E-13
> 32715c32715
> < -0.62459056E-8 0.80495845E-10 0.64089812E-8 0.12736262E-7 0.19059050E-7
>> -0.62459004E-8 0.80504293E-10 0.64089803E-8 0.12736268E-7 0.19059046E-7
> ### Maximum absolute error in matching lines = 8.45e-15 at line 32715 field 2
> ### Maximum relative error in matching lines = 1.05e-04 at line 22884 field 5
>
> Is the tolerance here set too low?
>
> Ben
>
> On 14/4/2010, at 4:11 p.m., David A. Case wrote:
>
>> Hi everyone:
>>
>> I've put a new RC up in the usual place:
>>
>> http://ambermd.org/downloads/AmberTools.14apr10.tar.bz2
>> http://ambermd.org/downloads/Amber11.14apr10.tar.bz2
>>
>> I'm sure there will be some changes before release, but I hope this is last
>> one that needs to be tested.
>>
>> There *are* a number of new test cases (rism1d, pbsa, PIMD), many of which
>> were not being run in earlier RC's. So I will be disappointed but not too
>> surprised if new problems show up; I hope the old ones are gone.
>>
>> Scott: I think you should place bug 130 on low priority--there is a simple
>> workaround for users, even if some test cases fail.
>>
>> If you can, please test AmberTools first (before untarring the Amber11 file),
>> since that is all that many users will have.
>>
>> ...thanks....dac
>>
>>
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers

--
Dr. Andreas W. Goetz
San Diego Supercomputer Center
Tel  : +1-858-822-4771
Email: agoetz.sdsc.edu
Web  : www.awgoetz.de
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Apr 16 2010 - 11:00:03 PDT
Custom Search