ABINIT Discussion Forums

I encounter a bug as below:

--- !BUG
message: |
The variables ikg, mkmem, and mpw must satisfy ikg<=(mkmem-1)*mpw,
while the arguments of the routine are
ikg =14739, mkmem =5, mpw =3560
Probable cause: Known error in invars1 for parallel spin-polarized case.
Temporary solution: Change the number of parallel processes.
src_file: m_fftcore.F90
src_line: 3412
...

After some googling I found out that this is most probably caused by the use of getcell -1. Strangely the bug only appeared when I active parallelism over kgb. Following the suggested temporary solution I had used different distribution of processors but to no avail. Any advice on how to overcome the bug?

My input script is as follow:


#Definition of the different datasets
#************************************

ndtset 3
jdtset 1 2 3

acell 3*3.98 angstrom

xred 0.000000 0.000000 0.000000
0.500000 0.500000 0.500000
0.000000 0.500000 0.500000
0.500000 0.000000 0.500000
0.500000 0.500000 0.000000

getcell -1
getwfk3 -1
getden3 -1

#Parallelization
#****************

# paral_kgb 1 autoparal 2 max_ncpus 8 # used to determine distribution of processors
paral_kgb1 1 npkpt1 2 npband1 1 npfft1 2
paral_kgb2 1 npkpt2 2 npband2 2 npfft2 1

#Definition of the elementary cell
#*********************************

rprim 1.0 0.0 0.0
0.0 1.0 0.0
0.0 0.0 1.0

#Definition of the atoms
#***********************
nband 32 # total valence electron/2 + 10
natom 5
ntypat 3
znucl 56 22 8
typat 1 2 3 3 3
ixc 11

#Definition of the SCF procedure
#*******************************
nstep 100
fftalg 401
diemac 10.0
iscf3 -2

#Definition of the plane wave basis set
#**************************************
ecut 45
ecutsm 0.5
dilatmx 1.05
ngkpt1 4 4 4
ngkpt2 6 6 6
ngkpt3 6 6 6
nshiftk 1
shiftk 0.5 0.5 0.5
toldfe1 1.0d-12
toldfe2 1.0d-12
tolwfr3 1.0d-22

# Relaxation of atomic positions
#*******************************
ionmov1 2
optcell1 1
tolmxf1 5.0d-5
ntime1 100

# Berry phase calculation of the polarization
#********************************************
berryopt3 -1
rfdir3 1 1 1


# add to conserve old < 6.7.2 behavior for calculating forces at each SCF step
optforces 1

A fast fix would be to do your relaxation (dataset 1) alone and then copy paste the xred coordinates from the output file to your input file.
Then run the 2 other datasets since they use the same positions without getcell...

This might not be the answer you're looking for but can save you time meanwhile and have your work going on.

Cheers
Jordan