Testing NWChem installation for correctness

From NWChem

Viewed 514 times, With a total of 4 Posts
Jump to: navigation, search

Just Got Here
Threads 1
Posts 2
We have two new installations (6.6 and 6.8) on a Cray system and we need to test them for correctness. They have been installed using the Intel 17 compiler tools. What is the procedure to test and make sure results are correct?

I have noticed the QA tests, should we obtain identical results? Is there a parameter to check?

Forum Vet
Threads 9
Posts 1522
$NWCHEM_TOP/QA contains a few scripts and a HOW-TO-RUN-TESTS file
You might want to run first doafewqmtests.mpi and the more complete doqmtests.mpi

Just Got Here
Threads 1
Posts 2
Thanks for the help.

I have been running some of the test from the QA directory, then parsing the output and diff them. So, for instance 6.8 compiled with Intel 17 would produce some fails, but also 6.6 patched and compiled with Intel 17. All those results are also different from the output produced with 6.6 and Intel 16...

So, the question is, should I assume that the last version (6.8) compiled with an up to date compiler (Intel 17) would produce the right (or close to) outputs? Here I show part of the output of
diff -w autosym.6.8.out.nwparse autosym.out.ref.nwparse


87,88c87,88
< P.Frequency      1692       1692       1692       1909       1909       3436
< P.Frequency      3718       3718       3719
---
> P.Frequency      1692       1692       1692       1909       1909       3463
> P.Frequency      3719       3719       3719
91,94c91,94
< Frequency         0          0          0        287        306        462
< Frequency      1692       1692       1692       1909       1909       3436
< Frequency      3718       3718       3719
< Zero-Point correction to Energy = 33.557
---
> Frequency         0          0          0        292        292        292
> Frequency      1692       1692       1692       1909       1909       3463
> Frequency      3719       3719       3719
> Zero-Point correction to Energy = 33.597
96,97c96,97
< P.Frequency      1692       1692       1692       1909       1909       3436
< P.Frequency      3718       3718       3719
---
> P.Frequency      1692       1692       1692       1909       1909       3463
> P.Frequency      3719       3719       3719


or
diff -w dft_s12gh_6.8.out.nwparse dft_s12gh.out.ref.nwparse


50,52c50,52
< Frequency       -36        -36        -17        169        196        200
< Frequency      1057       1680       1681       3530       3681       3685
< Zero-Point correction to Energy = 21.889
---
> Frequency       -37        -37        -24        -17         37         37
> Frequency      1053       1679       1679       3529       3661       3661
> Zero-Point correction to Energy = 21.805
54c54
< P.Frequency      1056       1682       1682       3530       3683       3686
---
> P.Frequency      1053       1679       1679       3529       3661       3661


How serious is this? If so, what can be done?
Edited On 2:32:54 AM PST - Thu, Jan 18th 2018 by Luiceur

Forum Vet
Threads 9
Posts 1522
Those numbers look OK to me.
Our parsing script needs to be improved.

Clicked A Few Times
Threads 2
Posts 7
Luiceur, you could try reading the post reply by Mernst in the forum post below:

http://www.nwchem-sw.org/index.php/Special:AWCforum/st/id1377/QA_tests_failing.html

Mernst's QACheck.py script is very useful - we've just started using it.

Hope this helps.


Forum >> NWChem's corner >> Compiling NWChem



Who's here now Members 0 Guests 1 Bots/Crawler 0


AWC's: 2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC