cp2k.psmp and cp2k.popt on the test systems H2O-128 and H2O-32

[Torstein Fjermestad]

Dear cp2k users, 

I have run some tests of cp2k.psmp and cp2k.popt on the test systems H2O-128 and H2O-32.

I did three tests (labelled TEST 1, TEST 2, and TEST 3), and the details are shown below.

The summary of the results is that cp2k.psmp fails for H2O-128, but not for H2O-32.

Is this an expected result? 

Has anyone tested cp2k.psmp for these systems before?

Thanks in advance for your help.

Regards,

Torstein 

 

TEST 1: 

Test system H2O-128 (https://github.com/cp2k/cp2k/blob/master/tests/QS/benchmark/H2O-128.inp)

Executable: cp2k.psmp

The calculation was run on 4 nodes, 16 tasks per node, and 2 threads per task.

This resulted in the following error message:

 *******************************************************************************

 *   ___                                                                       *

 *  /   \                                                                      *

 * [ABORT]                                                                     *

 *  \___/    Cholesky decompose failed: the matrix is not positive definite or *

 *    |                              ill-conditioned.                          *

 *  O/|                                                                        *

 * /| |                                                                        *

 * / \                                                  fm/cp_fm_cholesky.F:94 *

 *******************************************************************************

 ===== Routine Calling Stack =====

           10 cp_fm_cholesky_decompose

            9 make_full_inverse_cholesky

            8 make_preconditioner

            7 prepare_preconditioner

            6 init_scf_loop

            5 scf_env_do_scf

            4 qs_energies

            3 qs_forces

            2 qs_mol_dyn_low

            1 CP2K

TEST 2:

Test system H2O-128 (https://github.com/cp2k/cp2k/blob/master/tests/QS/benchmark/H2O-128.inp)

Executable: cp2k.popt

The calculation was run on 4 nodes and  32 tasks per node.

This resulted in a normal termination of the program

TEST 3:

Test system: H2O-32 (https://github.com/cp2k/cp2k/blob/master/tests/QS/benchmark/H2O-32.inp)

Executable: cp2k.psmp

The calculation was run on 4 nodes, 16 tasks per node, and 2 threads per task.

This resulted in a normal termination of the program

[hut...@chem.uzh.ch]

Hi

this is a very strange behavior. I'm not able to reproduce this
on my system. The error points to a severe problem with the
basis set or the geometry. However, that is not possible
as the same input runs with the popt executable. Leaves
a problem with the threaded libraries, but why is then the
other example working?
I can only suggest to further investigate the problem by running
more tests using different number of OMP threads 1,2,4 and different
number of MPI threads.

regards

Juerg
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Torstein Fjermestad"
Sent by: cp...@googlegroups.com
Date: 04/03/2019 05:19PM
Subject: [CP2K:11527] cp2k.psmp and cp2k.popt on the test systems H2O-128 and H2O-32

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.