[CP2K-user] [CP2K:21148] Graphene nano stripe (GNS) optimisation
Giuseppe Zollo
1saqquara at gmail.com
Thu Feb 13 11:17:26 UTC 2025
Il giorno giovedì 13 febbraio 2025 alle 11:52:21 UTC+1 Frederick Stein ha
scritto:
That sounds pretty odd. It is also interesting that you can run a few MD
steps. For debugging:-
- If you have compiled CP2K yourself,
yes
can you compile a pdbg?
yes
Or do you have pdbg available?
yes
Can you run your input file with this executable?
yes
- If that is not available or the result not promissing enough: Can you add
the TRACE and TRACE_MASTER keywords to your GLOBAL section and provide the
last ca. 100 lines of the output (the actual output file will be pretty
large). This would help to narrow down the error.
I've run the pdbg executable also adding TRACE and TRACE_MASTER and I got
this output where it crushes
000000:000128<< 11 3
cp_fm_finish_copy_general 0.001 Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 11 3
cp_cfm_scale_and_add_fm start Hostmem: 871 MB GPUmem: 0 MB
000000:000128<< 11 3
cp_cfm_scale_and_add_fm 0.000 Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 11 4
cp_fm_finish_copy_general start Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 12 290 mp_waitall_1
start Hostmem: 871 MB GPUmem: 0 MB
000000:000128<< 12 290 mp_waitall_1
0.000 Hostmem: 871 MB GPUmem: 0 MB
000000:000128<< 11 4
cp_fm_finish_copy_general 0.001 Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 11 4
cp_cfm_scale_and_add_fm start Hostmem: 871 MB GPUmem: 0 MB
000000:000128<< 11 4
cp_cfm_scale_and_add_fm 0.000 Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 11 1 cp_cfm_geeig
start Hostmem: 871 MB GPUmem: 0 MB
000000:000128>> 12 1
cp_cfm_cholesky_decompose start Hostmem: 871 MB GPUmem: 0 MB
Giuseppe Zollo schrieb am Donnerstag, 13. Februar 2025 um 11:47:43 UTC+1:
Hello Frederick.
thanks for you reply.
I've just done what you suggest but I've got exactly the same issue.
And what is really strange is that no error or output from the scheduler si
produced.
This typically happens when you have memory problems but here I'm using 460
GB, that I think is more than enough
Il giorno giovedì 13 febbraio 2025 alle 11:28:01 UTC+1 Frederick Stein ha
scritto:
Dear Giuseppe,
Have you already tried to run a series of test calculations with a smaller
k-point grid, especially 1x1x1 with complex wave functions? CP2K prints
some memory info at each MD step (Estimated peak process memory) and at the
very end of a successful run. Do you have more output files such as from
your batch system pointing to OOM-events?
Best,
Frederick
Giuseppe Zollo schrieb am Donnerstag, 13. Februar 2025 um 11:16:41 UTC+1:
Hello,
thank you so much for your time.
Unfortunately setting the wavefunctions to REAL was the only way to make
the simulation proceed. Indeed if a do not include this flag (so as to
assume complex wavefunctions) the simulation hangs, as reported here, at
the initials of the first SCF loop and the run crashes.
I'm using quite a lot memory so I assume that should not be a memory
problem, should it be?
*******************************************************************************
Iteration Convergence Energy
[au]
*******************************************************************************
1 0.240530
-6.658276883875
2 0.153299
-6.681113186074
3 0.339530E-02
-6.698173661406
4 0.134821E-03
-6.698181647940
5 0.672005E-06
-6.698181660696
Energy components [Hartree] Total Energy ::
-6.698181660696
Band Energy ::
-1.311151959578
Kinetic Energy ::
3.482441966707
Potential Energy ::
-10.180623627403
Virial (-V/T) ::
2.923415156586
Core Energy ::
-8.233425335600
XC Energy ::
-2.899318720305
Coulomb Energy ::
4.434562395209
Total Pseudopotential Energy ::
-11.749206534330
Local Pseudopotential Energy ::
-12.352996294377
Nonlocal Pseudopotential Energy ::
0.603789760047
Confinement ::
0.333392320232
Orbital energies State L Occupation Energy[a.u.]
Energy[eV]
1 0 2.000 -0.483136
-13.146798
1 1 2.000 -0.172440
-4.692332
Total Electron Density at R=0:
4.671964
Re-scaling the density matrix to get the right number of electrons
# Electrons Trace(P) Scaling
factor
252 252.000
1.000
SCF WAVEFUNCTION OPTIMIZATION
Step Update method Time Convergence Total energy
Change
------------------------------------------------------------------------------
Il giorno giovedì 13 febbraio 2025 alle 10:44:01 UTC+1 Jürg Hutter ha
scritto:
Hi
I don't have time to look closer into your problem but a quick inspection
of the input file shows a major problem in your definition for the
k-points.
The Wavefunction REAL option should not be used in you case, as the correct
WFN for a MP Grid of 1 1 8 will have complex wavefunctions.
regards
JH
________________________________________
From: cp... at googlegroups.com <cp... at googlegroups.com> on behalf of Giuseppe
Zollo <1saq... at gmail.com>
Sent: Tuesday, February 11, 2025 12:15 PM
To: cp2k
Subject: [CP2K:21135] Graphene nano stripe (GNS) optimisation
Dear All,
I've been working for long time with QE.
Now I'm trying to get some results with CP2K (I'm new of this nice code).
A problem seems to arise for an edge patterned GNS I'm interested in. I
hope to get some results with hybrid functionals that is very hard to
obtain with plane waves as the unit cell is large.
I get easy convergence only by using kpoints (no other attempt has been
successful) and full periodic system (partial periodicity does not converge
and system without kpoints and two unit cells along the stripe does not
converge as well).
But the point Is that the structure, that I know is stable from plane waves
QE calculations, is completely destroyed after few optimisation steps.
Maybe I'm missing something.
Here I attach input, output and optimisation coordinates
Any help will be appreciated,
Regards
--
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... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
To view this discussion visit
https://groups.google.com/d/msgid/cp2k/b438d7f7-fb8c-4ad3-a726-d58efc7b27f7n%40googlegroups.com
<
https://groups.google.com/d/msgid/cp2k/b438d7f7-fb8c-4ad3-a726-d58efc7b27f7n%40googlegroups.com?utm_medium=email&utm_source=footer>.
--
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/8dd9591e-db6f-4537-8868-a0b2eabb98c1n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250213/10039c73/attachment-0001.htm>
More information about the CP2K-user
mailing list