[Pw_forum] about OMP_NUM_THREADS
Jason Larkin
jmlarkin at andrew.cmu.edu
Fri Sep 25 16:06:16 CEST 2009
Hello Forum,
First, thanks very much for the great software package. Here is my
issue, one which is probably due to my own ignorance about QE/quantum
physics in general:
I'm trying to run a varaible cell relaxation on the compound GeTe.
Below is my input file:
&control
calculation = 'vc-relax'
prefix='gt',
pseudo_dir = '$HOME/pseudo/',
outdir='$HOME/QE/GST/GeTe/out/'
forc_conv_thr = 1.0d-5,
etot_conv_thr = 1.0d-5,
nstep=300,
/
&system
ibrav = 4, celldm(1)=7.993540933, celldm(3)=2.581560284
nat = 18, ntyp= 2,
ecutwfc = 100, ecutrho = 800,
/
&electrons
mixing_beta = 0.7
conv_thr = 1.0d-6,
/
&ions
upscale = 100.d0,
/
&cell
cell_dynamics='damp-w'
/
ATOMIC_SPECIES
Te 1.0 Te.pz-bhs.UPF
Ge 1.0 Ge.pz-bhs.UPF
ATOMIC_POSITIONS {crystal}
Te 0 0 0
Ge 0.666666 0.333333 0.055555556
Te 0.33333 0.666666 0.111111111
Ge 0 0 0.166666667
Te 0.6666666 0.3333333 0.222222222
Ge 0.3333333 0.6666666 0.277777778
Te 0 0 0.333333333
Ge 0.6666666 0.3333333 0.388888889
Te 0.3333333 0.6666666 0.444444444
Ge 0 0 0.5
Te 0.6666666 0.3333333 0.555555556
Ge 0.3333333 0.6666666 0.611111111
Te 0 0 0.666666667
Ge 0.6666666 0.3333333 0.722222222
Te 0.3333333 0.6666666 0.777777778
Ge 0 0 0.833333333
Te 0.6666666 0.3333333 0.888888889
Ge 0.3333333 0.6666666 0.944444444
K_POINTS {automatic}
12 12 2 0 0 0
My problem (issue) is this; the simulation continues to time out before
completing the relaxation. As such, I've been coninuously increasing
the amount of time and processors that I am requesting to perform this
calculation. However, it has reached a point (32 processors and 3:30
hours) where I'm almost certain that I am doing something wrong. I was
able to find one forum posting where a user was running a relaxation and
the calculation was not stopping even though the Energy and Force
calculation error thresh holds where being satisfied. It appears that my
energy does indeed converge. Let me attach part of my output:
*********************************************************************
! total energy = -203.60367909 Ry
Harris-Foulkes estimate = -203.60367961 Ry
estimated scf accuracy < 0.00000054 Ry
convergence has been achieved in 9 iterations
Forces acting on atoms (Ry/au):
atom 1 type 1 force = 0.00000000 0.00000000 0.00000000
atom 2 type 2 force = 0.00000000 0.00000000 0.00052582
atom 3 type 1 force = 0.00000000 0.00000000 -0.00108418
atom 4 type 2 force = 0.00000000 0.00000000 -0.00111021
atom 5 type 1 force = 0.00000000 0.00000000 -0.00066898
atom 6 type 2 force = 0.00000000 0.00000000 0.00030722
atom 7 type 1 force = 0.00000000 0.00000000 -0.00150570
atom 8 type 2 force = 0.00000000 0.00000000 -0.00163780
atom 9 type 1 force = 0.00000000 0.00000000 0.00177984
atom 10 type 2 force = 0.00000000 0.00000000 0.00000000
atom 11 type 1 force = 0.00000000 0.00000000 -0.00177984
atom 12 type 2 force = 0.00000000 0.00000000 0.00163780
atom 13 type 1 force = 0.00000000 0.00000000 0.00150570
atom 14 type 2 force = 0.00000000 0.00000000 -0.00030722
atom 15 type 1 force = 0.00000000 0.00000000 0.00066898
atom 16 type 2 force = 0.00000000 0.00000000 0.00111021
atom 17 type 1 force = 0.00000000 0.00000000 0.00108418
atom 18 type 2 force = 0.00000000 0.00000000 -0.00052582
Total force = 0.004763 Total SCF correction = 0.002516
entering subroutine stress ...
total stress (Ry/bohr**3) (kbar) P=
5650.32
0.02322957 0.00000000 0.00000000 3417.19 0.00 0.00
0.00000000 0.02322957 0.00000000 0.00 3417.19 0.00
0.00000000 0.00000000 0.06877108 0.00 0.00 10116.57
Wentzcovitch Damped Cell-Dynamics Minimization
convergence thresholds: EPSE = 0.10E-04 EPSF = 0.10E-04 EPSP =
0.50E+00
Entering Dynamics; it = 1 time = 0.00000 pico-seconds
*********************************************************************
The simulation goes through 23 complete Self-consistent Calculations.
Here is the last self-consistent run (before the 24th run timed out):
*********************************************************************
! total energy = -216.83428802 Ry
Harris-Foulkes estimate = -216.83428838 Ry
estimated scf accuracy < 0.00000036 Ry
convergence has been achieved in 10 iterations
Forces acting on atoms (Ry/au):
atom 1 type 1 force = 0.00000000 0.00000000 0.00000000
atom 2 type 2 force = 0.00000000 0.00000000 -0.00120296
atom 3 type 1 force = 0.00000000 0.00000000 -0.00028879
atom 4 type 2 force = 0.00000000 0.00000000 0.00057200
atom 5 type 1 force = 0.00000000 0.00000000 0.00115400
atom 6 type 2 force = 0.00000000 0.00000000 -0.00030438
atom 7 type 1 force = 0.00000000 0.00000000 -0.00064802
atom 8 type 2 force = 0.00000000 0.00000000 -0.00012442
atom 9 type 1 force = 0.00000000 0.00000000 0.00095840
atom 10 type 2 force = 0.00000000 0.00000000 0.00000000
atom 11 type 1 force = 0.00000000 0.00000000 -0.00095840
atom 12 type 2 force = 0.00000000 0.00000000 0.00012442
atom 13 type 1 force = 0.00000000 0.00000000 0.00064802
atom 14 type 2 force = 0.00000000 0.00000000 0.00030438
atom 15 type 1 force = 0.00000000 0.00000000 -0.00115400
atom 16 type 2 force = 0.00000000 0.00000000 -0.00057200
atom 17 type 1 force = 0.00000000 0.00000000 0.00028879
atom 18 type 2 force = 0.00000000 0.00000000 0.00120296
Total force = 0.003045 Total SCF correction = 0.001829
entering subroutine stress ...
total stress (Ry/bohr**3) (kbar) P=
1021.47
0.00437252 0.00000000 0.00000000 643.22 0.00 0.00
0.00000000 0.00437252 0.00000000 0.00 643.22 0.00
0.00000000 0.00000000 0.01208629 0.00 0.00 1777.95
Entering Dynamics; it = 24 time = 0.02226 pico-seconds
*********************************************************************
It appears that my force(s) are not consistent with the force threshold
that I set. I'm guessing this is because my initial guess at the
structure is quite far from the relaxed structure. So, would it be
reasonable to take the unit cell coordinates from the last completed
self-consistent calcualtion and use those as starting coordinates for a
brand new vc-relax run? This would be an effort to continue the
calculation, but not lose the 3:30 of run time I've already used. Also,
how would I handle a restart in this situation? Forgive me if this
question is too basic. Thanks for the help
Jason Larkin, PhD Student
Carnegie Mellon University
More information about the Pw_forum
mailing list