Author Topic: SCF Iterations in TURBOMOLE 7.2  (Read 5230 times)

saleheen_noman

  • Jr. Member
  • **
  • Posts: 12
  • Karma: +0/-0
SCF Iterations in TURBOMOLE 7.2
« on: May 04, 2018, 04:33:14 AM »
Hi!

I was trying perform some ridft calculations using TURBOMOLE 7.2. I used to run these calculations using 6.X versions and had nothing to complain about. However, in version 7.2, in cases where the convergence is slow, my calculations get aborted after 1861 SCF iterations even when the limit is set to higher number (e.g. 3000), and it shows the following error message-

Code: [Select]
pri_inittask  sem_open  n failed 23
pri_inittask  sem_open  n failed 23
pri_inittask  sem_open  n failed 23
pri_inittask  sem_open  n failed 23
forrtl: error (78): process killed (SIGTERM)
Image              PC                Routine            Line        Source             
Unknown            0000000002951AEA  Unknown               Unknown  Unknown
Unknown            00002B9B7CBE47E0  Unknown               Unknown  Unknown
libpthread-2.12.s  00002B9B7CBE46AB  raise                 Unknown  Unknown
libmpi.so.1        00002B9B7C4DB577  HPMPI_Abort           Unknown  Unknown
Unknown            000000000088E910  Unknown               Unknown  Unknown
Unknown            000000000088AD38  Unknown               Unknown  Unknown
Unknown            000000000087A75F  Unknown               Unknown  Unknown
Unknown            0000000000E5BA72  Unknown               Unknown  Unknown
Unknown            00000000004A2865  Unknown               Unknown  Unknown
Unknown            0000000000440D75  Unknown               Unknown  Unknown
Unknown            0000000000430F1E  Unknown               Unknown  Unknown
libc-2.12.so       00002B9B7CE11D1D  __libc_start_main     Unknown  Unknown
Unknown            0000000000430E29  Unknown               Unknown  Unknown


The SIGTERM part of the error message repeats for a few more times. And it always crashes after 1861 steps, no matter what number I set for the scf iterations. I tried to play with the scforditalshift and scfdamp values, didn't help. Is it a bug in the code? Did anyone face any similar problems?

I'm adding my control file here in case anyone wants to have a look at it.

Code: [Select]
$title
$operating system unix
$symmetry c1
$user-defined bonds    file=coord
$coord    file=coord
$optimize
 internal   off
 redundant  off
 cartesian  on
 global     off
 basis      off
$atoms
pt 1-51                                                                        \
   basis =pt ecp-60-mwb-SVP                                                    \
   ecp   =pt ecp-60-mwb                                                        \
   jbas  =pt ecp-60-mwb-SVP
c  52-53                                                                       \
   basis =c def2-SVP                                                           \
   jbas  =c def2-SVP
o  54-55                                                                       \
   basis =o def2-SVP                                                           \
   jbas  =o def2-SVP
h  56-61                                                                       \
   basis =h def2-SVP                                                           \
   jbas  =h def2-SVP
$basis    file=basis
$ecp    file=basis
$rundimensions
   dim(fock,dens)=1078869
   natoms=61
   nshell=603
   nbf(CAO)=1467
   dim(trafo[SAO<-->AO/CAO])=1679
   rhfshells=2
   nbf(AO)=1361
$uhfmo_alpha   file=alpha
$uhfmo_beta   file=beta
$uhf
$alpha shells
 a       1-484                                  ( 1 )
$beta shells
 a       1-468                                  ( 1 )
$scfiterlimit       3000
$thize     0.10000000E-04
$thime        5
$
$scfintunit
 unit=30       size=0        file=twoint
$scfdiis
$maxcor    500 MiB  per_core
$drvopt
   cartesian  on
   basis      off
   global     off
   hessian    on
   dipole     on
   nuclear polarizability
$interconversion  off
   qconv=1.d-7
   maxiter=25
$coordinateupdate
   dqmax=0.3
   interpolate  on
   statistics    5
$forceupdate
   ahlrichs numgeo=0  mingeo=3 maxgeo=4 modus=<g|dq> dynamic fail=0.3
   threig=0.005  reseig=0.005  thrbig=3.0  scale=1.00  damping=0.0
$forceinit on
   diag=default
$energy    file=energy
$grad    file=gradient
$forceapprox    file=forceapprox
$dft
   functional pbe
   gridsize   m4
$scfconv   7
$scfdamp   start=1.500  step=0.050  min=0.100
$scforbitalshift  closedshell=0.5  automatic! 1.0
$ricore      500
$rij
$jbas    file=auxbasis
$last step     define
$end

Thank you so much!

Saleheen

uwe

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 558
  • Karma: +0/-0
Re: SCF Iterations in TURBOMOLE 7.2
« Reply #1 on: May 04, 2018, 11:02:52 AM »
Hi,

the error message

pri_inittask  sem_open  n failed 23

indicates that you might have problems with I/O. Not hardware problems, but probably a too small limit for the number of open files. Or a  lot of (old) files in /dev/shm ?

Are there other jobs (also from other users) running on the same node?

Regards,

Uwe

saleheen_noman

  • Jr. Member
  • **
  • Posts: 12
  • Karma: +0/-0
Re: SCF Iterations in TURBOMOLE 7.2
« Reply #2 on: May 04, 2018, 06:56:01 PM »
Dr. Uwe

I looked into /proc/sys/fs/file-max which says the limit is 562454 and file-nr says 6400 files remain opened. Also /dev/shm is not full of old files. I am using the whole node myself. I don't know why it still says the same thing. Just to clarify, the problem seem to vanish with one combination of scforbitalshift and scfdamp, but I don't understand with a different combination why it doesn't perform iterations more than 1861 times. Do the number of open files exceed 562454 when a calculation reaches 1861 iteration steps?

Thanks!

Saleheen