[QE-developers] Bug for U_projection_type = 'ortho-atomic'

Timrov Iurii iurii.timrov at epfl.ch
Mon Aug 17 13:17:48 CEST 2020 

Dear Max,


Actually there was a more serious bug than I thought. It is related to the internal poor-man bands parallelization, and this bug appears when  nproc_pool>nbnd (i.e. for a specific combination of the number of cores, number of k point pools, and number of bands).


I have made a merge request with the fix on Gitlab, and here is the link for the two modified routines that you can put in PW/src and recompile the code. Please let us know if there are still problems.

https://drive.google.com/drive/folders/1ORTA5kILqvqBuWpl0CQYdeX8vv0CTF4L?usp=sharing


Many thanks to Paolo for explaining to me how the poor-man bands parallelization works!


Greetings,

Iurii


--
Dr. Iurii TIMROV
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: developers <developers-bounces at lists.quantum-espresso.org> on behalf of Max Amsler <amsler.max at gmail.com>
Sent: Friday, August 14, 2020 2:31:00 PM
To: General discussion list for Quantum ESPRESSO developers
Subject: Re: [QE-developers] Bug for U_projection_type = 'ortho-atomic'

Thanks Iurii,
I recompiled with your recommended changes, but it the run no just hangs after reaching self consistency without reporting any error messages. Any idea what the reason could be?

Best
Max

On 14 Aug 2020, at 11:00, Timrov Iurii <iurii.timrov at epfl.ch<mailto:iurii.timrov at epfl.ch>> wrote:

Dear Max,

Thank you for reporting the bug! It is easy to fix it, in PW/src/force_hub.f90, on line 438 replace
IF (okvan) DEALLOCATE (dproj_us)
by
IF (ALLOCATED(dproj_us)) DEALLOCATE (dproj_us)

Another solution is to move
IF (okvan) DEALLOCATE (dproj_us)
before line 436

> starting_magnetization(1) = 5.

BTW, starting_magnetization should be in the range from -1 to +1.

Greetings,
Iurii

--
Dr. Iurii Timrov
Postdoctoral Researcher
STI - IMX - THEOS and NCCR - MARVEL
Swiss Federal Institute of Technology Lausanne (EPFL)
CH-1015 Lausanne, Switzerland
+41 21 69 34 881
http://people.epfl.ch/265334
________________________________
From: developers <developers-bounces at lists.quantum-espresso.org<mailto:developers-bounces at lists.quantum-espresso.org>> on behalf of Max Amsler <amsler.max at gmail.com<mailto:amsler.max at gmail.com>>
Sent: Friday, August 14, 2020 9:04:27 AM
To: General discussion list for Quantum ESPRESSO developers
Subject: [QE-developers] Bug for U_projection_type = 'ortho-atomic'

Dear all,
I have been trying to run a vc-relaxation with version 6.6 of QE, using the newly implemented U_projection_type = 'ortho-atomic’. After the SCF has converged and the forces on the atoms are about to be computed, pw.x crashes with an error complaining that an array is not being allocated:

forrtl: severe (153): allocatable array or pointer is not allocated
Image              PC                Routine            Line        Source
pw.x               0000000000EA50B4  Unknown               Unknown  Unknown
pw.x               0000000000757C94  dndtau_k_                 438  force_hub.f90
pw.x               000000000074B521  force_hub_                160  force_hub.f90
pw.x               00000000004C92ED  forces_                   116  forces.f90
pw.x               00000000005CCE5C  run_pwscf_                172  run_pwscf.f90
pw.x               000000000041005D  MAIN__                    106  pwscf.f90
pw.x               000000000040FE92  Unknown               Unknown  Unknown
libc-2.26.so       00002AAAB2F60F8A  __libc_start_main     Unknown  Unknown
pw.x               000000000040FDAA  Unknown               Unknown  Unknown

pw.x was compiled on daint at CSCS, with the intel compiler version 19.0.1.144.

Running on 144 cores with -nk 4.

The input file is as follows:

————————————————————————
 &control
    calculation='vc-relax'
    restart_mode='from_scratch',
    pseudo_dir = './',
    outdir='./',
    prefix='CrO2',
    nstep=100,
    disk_io='low',
    max_seconds=79200
    forc_conv_thr = 1.d-5
    etot_conv_thr = 1.d-5
 /
 &system
U_projection_type = 'ortho-atomic'
    nspin = 2
    starting_magnetization(1) = 5.
    starting_magnetization(2) = 0.
    ibrav= 0,
    nat=  6,
    ntyp = 2
    nosym=.false.
    ecutwfc  =  100.0 ,
    ecutrho  = 1200.0 ,
    occupations  =  'smearing' ,
    smearing  =  'gaussian' ,
    degauss  =  0.01 ,
    lda_plus_u=.true.
    lda_plus_u_kind     = 0,
    Hubbard_U(1) = 6.5022
 /
 &electrons
    conv_thr =  1.0d-12,
    mixing_beta = 0.4,
 /
 &ions
 /
 &cell
 press_conv_thr = 0.1
 /
ATOMIC_SPECIES
Cr0 51.9961 Cr.pbe-spn-rrkjus_psl.1.0.0.UPF
O  15.99   O.pbe-n-rrkjus_psl.1.0.0.UPF
K_POINTS automatic
   8 8 12         0    0    0
CELL_PARAMETERS (angstrom)
   4.515875669   0.000000000   0.000000000
   0.000000000   4.515868266   0.000000000
   0.000000000   0.000000000   3.010520517

ATOMIC_POSITIONS (crystal)
Cr0 0.0000000000 0.0000000000 0.0000000000
Cr0 0.5000000000 0.5000000000 0.5000000000
O 0.3038809317 0.3038796501 0.0000000000
O 0.8038809317 0.1961203369 0.5000000000
O 0.6961190683 0.6961203499 0.0000000000
O 0.1961190683 0.8038796501 0.5000000000
————————————————————————

_______________________________________________
developers mailing list
developers at lists.quantum-espresso.org<mailto:developers at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/developers
_______________________________________________
developers mailing list
developers at lists.quantum-espresso.org<mailto:developers at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/developers

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20200817/b94dd04c/attachment-0001.html>

More information about the developers mailing list