Error while using flos for "flos_h2o_neb"

Asked by rezabma

Hi
I have compiled siesta according below specification:
=====================================================
Siesta Version : v4.1-b4
Architecture : x86_64_MPI_INTEL
Compiler version: ifort (IFORT) 15.0.0 20140723
Compiler flags : mpiifort -O2 -fPIC -xHost -ip -no-prec-div -fp-model source
PP flags : -DFC_HAVE_ABORT -DMPI -DSIESTA__MRRR -DSIESTA__FLOOK
Libraries : libfdict.a /opt/flook-master/libflook.a /opt/flook-master/aotus/obj/libaottus.a /opt/lua53/lib/liblua.a -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lmkl_blacs_intelmpi_lp64 -lpthread -lm -ldl
==========================================

It works fine for different tests and also for "lua_si111". However, I cannot run "flos_h2o_neb" example, I see this error:
======================================================================================

New grid distribution: 3
           1 1: 72 1: 15 1: 16
           2 1: 72 1: 15 17: 72
           3 1: 72 16: 72 1: 16
           4 1: 72 16: 72 17: 72
Setting up quadratic distribution...
ExtMesh (bp) on 0 = 124 x 65 x 71 = 572260
PhiOnMesh: Number of (b)points on node 0 = 17784
PhiOnMesh: nlist on node 0 = 59841
rdiag: Error in Cholesky factorisation
Stopping Program from Node: 0
rdiag: Error in Cholesky factorisation
Stopping Program from Node: 1
========================================================================

Question information

Language:
English Edit question
Status:
Solved
For:
Siesta Edit question
Assignee:
No assignee Edit question
Solved by:
rezabma
Solved:
Last query:
Last reply:
Revision history for this message
rezabma (rezabma) said :
#1

I forgotted to use /opt instead of $HOME in LUA_PATH. Now, I have new problem in running "flos_h2o_neb":

=====================================================
New grid distribution: 3
           1 1: 72 1: 15 1: 16
           2 1: 72 1: 15 17: 72
           3 1: 72 16: 72 1: 16
           4 1: 72 16: 72 17: 72
Setting up quadratic distribution...
ExtMesh (bp) on 0 = 124 x 65 x 71 = 572260
PhiOnMesh: Number of (b)points on node 0 = 17784
PhiOnMesh: nlist on node 0 = 59841
rdiag: Error in Cholesky factorisation
Stopping Program from Node: 0
rdiag: Error in Cholesky factorisation
Stopping Program from Node: 1
==========================================================

Any help?

Revision history for this message
rezabma (rezabma) said :
#2

Hi
It seems that it is only for H2O examples!!!

Revision history for this message
rezabma (rezabma) said :
#3

Hi
Problem solved. I think this is a bug in this version of siesta. In all cluster calculations, I changed the k-points from 111 to other numbers such as 211, and could run the jobs successfully.

Thanks
REZA

Revision history for this message
Nick Papior (nickpapior) said :
#4

The flos_h2o_neb example is *only* for an H2O example. Note that the images needs to correspond to the correct atomic structure and there is no check in the Lua code whether the correct number of atoms is present in the file. This is probably why it didn't work.

I have fixed this in r1072 such that Lua will stop Siesta if the atomic structures does not have the correct number of atoms.

Note, this is NOT a bug in Siesta.

Revision history for this message
rezabma (rezabma) said :
#5

Hi Nick,
You can test other cluster examples without input k-point or k-point 111. You will get similar error. However, if you change the k-point, without changing other settings, you car run the job successfully.
Thanks
Reza

Revision history for this message
Nick Papior (nickpapior) said :
#6

I have just tried with a different structure.
I got no error.
So I can't reproduce.

It isn't clear exactly what you did? How did you generate the image_*.xyz files?

Revision history for this message
rezabma (rezabma) said :
#7

Dear Nick,
this is an example for fe_cluster calculation with two fdf. one without kgrid that ends with error, and the other with kgrid of 211 with runs good.

Revision history for this message
rezabma (rezabma) said :
#8

It seems that my attached files doesn't go along with the mail message. Please test "fe_cluster_noncollinear", for example. Could you run the job with 4.1-b4 version with 111 kgrid (gamma point only)?

Revision history for this message
SuzyHo (suzyho) said :
#9

I just encountered the same error message for Tests/flos_h2o_meshcutoff and Tests/lua_h2o. And the modification of input kpoints works for me!

Revision history for this message
rezabma (rezabma) said :
#11

Hiit seems that in the new version of siesta above bug has been solved. It accepts kpoints '1 1 1' for gamma calculations. Please test your calculations with the master version of siesta in gitlab.
With the Best Regards

 Reza