Page 1 of 1

[FIXED 6.13] nlo calculations fails when using > 1proc

Posted: Mon Feb 13, 2012 11:33 am
by roginovicci
Hello. I'm trying to calculate raman spectra in GaN/AlN superlattice but my calculations always fails when I perfom calculation of Raman tensor using more than one processor. My input file is as follows:

Code: Select all

   acell    5.9403442780E+00  5.9403442780E+00  3.8115227793E+01
   angdeg 90.000  90.000 120.000
#   spgroup 156
rprim sqrt(0.75)  0.5  0.0   #hexagonal primitive vectors must be
     -sqrt(0.75)  0.5  0.0   #specified with high accuracy to be
      0.0         0.0  1.0   #sure that the symmetry is recognized and preserved in the optimization process

# Definition of the atom
   ntypat 3                 # Two types of atoms
   znucl 31 7 13               # Ga and N atomic numbers of the
   natom 16                  # There are 16 atoms in a cell
   typat 1 2 1 2 1 2 1 2 3 2 3 2 3 2 3 2      # Type 1 is Ga, type 2 is N 3 is Al

xred    3.3333333330E-01  6.6666666670E-01  0.0000000000E+00
3.3333333330E-01  6.6666666670E-01  9.5005079702E-02
6.6666666663E-01  3.3333333337E-01  1.2500000000E-01
6.6666666663E-01  3.3333333337E-01  2.2006995754E-01
3.3333333330E-01  6.6666666670E-01  2.5000000000E-01
3.3333333330E-01  6.6666666670E-01  3.4509756564E-01
6.6666666663E-01  3.3333333337E-01  3.7500000000E-01
6.6666666663E-01  3.3333333337E-01  4.7280325129E-01
3.3333333330E-01  6.6666666670E-01  5.0000000000E-01
3.3333333330E-01  6.6666666670E-01  5.9506147890E-01
6.6666666663E-01  3.3333333337E-01  6.2500000000E-01
6.6666666663E-01  3.3333333337E-01  7.2003534357E-01
3.3333333330E-01  6.6666666670E-01  7.5000000000E-01
3.3333333330E-01  6.6666666670E-01  8.4502086796E-01
6.6666666663E-01  3.3333333337E-01  8.7500000000E-01
6.6666666663E-01  3.3333333337E-01  9.6721208507E-01



## Definition of XC functional and planewave basis set
   ixc       7              # Teter XC functional
   ecut      30              # Kinetic energy cut-off (in Ha)
   ecutsm    0.5
   dilatmx   1.05
# Definition of the k-point grid
   kptopt    1
   ngkpt     5 5 4
   nshiftk   1
   shiftk    0.0 0.0 0.5

# Parameters of the SCF cycles
   iscf      7      # Pulay mixing (default for norm-conserving PSP)
   nstep     100    # Modify default
# --------------------------
# Structural optimization (atomic positions in cell)
#   ionmov 2
   ntime 100
   tolvrs 1.0d-15
   tolmxf 1.0d-6
   diemac 9.0
#
# Optimization of cell parameters
# change acell only
#   optcell 2
# opcell 2 - to change acell and xred
# --------------------------

################################################################################
# LINEAR AND NON-LINEAR RESPONSE CALCULATION
################################################################################

ndtset 5     jdtset 1 2 3 4 5

# 1: SCF calculation: GS WF in the BZ
#-------------------------------------
   prtden1    1              # Print electronic density
   kptopt1    1
   toldfe1    1.0d-12

# 2: non SCF calculation: GS WF in the whole BZ
#----------------------------------------------
   getden2    1              # Get density from set 1
   kptopt2    2
   iscf2     -2
   getwfk2    1              # Get WF from set 1
   tolwfr2    1.0d-22        # Use WF criterion since iscf < 0
   nband2     52             # Explicit number of valence bands

# 3: ddk on the MP grid of k-points
#----------------------------------
   getwfk3    2
   kptopt3    2
   rfdir3     1 1 1
   rfelfd3    2              # Derivative of GS WF with respect to k
   tolwfr3    1.0d-22
   nband3     52

# 4: ddE on the MP grid of k-points
#-----------------------------------
   getwfk4    2
   getddk4    3              # Get derivatives from set 3
   kptopt4    2
   prtden4    1
   rfdir4     1 1 1
   rfphon4    1
   rfatpol4   1 16
   rfelfd4    3              # Generate of the 1st-order response using ddk of 3
   tolvrs4    1.0d-12
   nband4     52
   prepanl4   1

# 5: 3DTE calculation
#--------------------
   getden5     1
   get1den5    4
   getwfk5     2
   get1wf5     4
   kptopt5     2
   optdriver5  5
   rf1elfd5    1
   rf1phon5    1
   rf1atpol5   1 16
   rf1dir5     1 1 1
   rf2elfd5    1
   rf2dir5     1 1 1
   rf3elfd5    1
   rf3dir5     1 1 1
   nband5      52
nbdbuf 0


Everything goes clean till 5th task 3DTE calculation.
The error is as follows:

ioarr: data read from disk file ganaln_nloo_DS4_DEN52

mv_3dte : finite difference expression of the k-point derivative
is performed using the PEAD formulation of the third-order energy
(see Nunes and Gonze PRB 63, 155107 (2001) Eq. 102)

rank 0 in job 9 v371_46979 caused collective abort of all ranks
exit status of rank 0: return code 1

Fatal error in MPI_Recv: Message truncated, error stack:
MPI_Recv(186): MPI_Recv(buf=0xaadc1a0, count=3319680, MPI_DOUBLE_PRECISION, src=1, tag=26, MPI_COMM_WORLD, status=0x2768630) fa
iled
do_cts(509)..: Message truncated; 26589888 bytes received but buffer size is 26557440

Is it a bug?
Is it possible to split input file into two ones to calculate tasks 1-4 and 5 separately?

Re: nlo calculations fails when using more than one processo

Posted: Fri Feb 17, 2012 9:25 pm
by roginovicci
I've got tested nlo calculations using NaNO2 structure with openmpi as mpi call library. And it still fails. Its defenetly a bug. How could I report it?
The openmpi message is:
/opt/openmpi/bin/mpirun --hostfile /opt/openmpi/hostfile -np 4 /work/builds/abinit/opt/abinit/bin/abinit < nano2_nlo.files > log
[main:29894] *** An error occurred in MPI_Recv
[main:29894] *** on communicator MPI_COMM_WORLD
[main:29894] *** MPI_ERR_TRUNCATE: message truncated
[main:29894] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
--------------------------------------------------------------------------
mpirun has exited due to process rank 1 with PID 29894 on
node main exiting without calling "finalize". This may
have caused other processes in the application to be
terminated by signals sent by mpirun (as reported here).
--------------------------------------------------------------------------

log file fails with:
ioarr: data read from disk file nano2_nloo_DS4_DEN16

mv_3dte : finite difference expression of the k-point derivative
is performed using the PEAD formulation of the third-order energy
(see Nunes and Gonze PRB 63, 155107 (2001) Eq. 102)

Re: nlo calculations fails when using more than one processo

Posted: Mon Mar 19, 2012 12:25 pm
by hdjani
Hi,
Yes, there is a bug for calculating 3DTE in parallel in the last version ABINIT 6.12.1. It will be fixed for sure in the next version. However, old version like abinit-6.2.3 works well for 3dte in parallel.
Cheers
Hania Djani

Universite de Liège
Institut de Physique, Bat. B5, 3/7
Allée du 6 aout, 17
B- 4000 Sart Tilman, Liège
Belgium

hdjani@ulg.ac.be

Re: nlo calculations fails when using more than one processo

Posted: Mon Mar 19, 2012 2:13 pm
by roginovicci
God bless you, Hania Djani!
I was awaiting for your answer 1 month. This is your firs post and it's really important for me! Thank you so much!

May I ask you a question? It's not that easy task for me to install abinit 6.2 due to technical problems I'll be trying to solve this week. Whilst I would like to know Is it possible to split nlo caltulations. I.e. calculate ground state and ddk and dde simultaneously witin one task (using tnlo1.in file for exaple) and later calculate 3dte (using tnlo2.in file for exaple) using output of previous task (output of task tnlo1.in).

Re: [FIXED 6.13] nlo calculations fails when using > 1proc

Posted: Thu Apr 05, 2012 9:50 am
by hdjani
Hi,
Yes of course you can split the calculations. you should have the ground state with its WFK and DEN, the ddk and ddE with its 1WF and 1DEN and phonons with 1WF and 1DEN. when you run 3dte, you should read : WFK and DEN of the ground state + 1WF and 1DEN of ddE + 1WF and 1DEN of phonons.
cheers
Hania Djani