StructureMBPT crashes when MPI uses more than 1 core

[mira42780534]

Dear all,

I am running StructureMBPT in parallel, but the calculation crashes. If I set "InitializeMPI(1)" or "Do not use MPI", the calculation completes successfully. Below is the output of the crashed calculation (StructureMBPT with 8 cores):

mbpt configs: 6 6 0
Rank=00, Construct Radial Basis, 0/8... 1.1087E+09  3.5228E+06
omp=0
ompf=5
extra=0
rand=0
msort=0
asort=10
prepyk=0
nbreit=-1
warn=0.05
nwarn=0.01
xwarn=0.0001
mwarn=0.5
wwarn=0.01
ewarn=3.67493e-05
ewarn11=0
ewarn12=0
warntr=1
ignore=50
ignoretr=10
ignorep=0
angzc=0.75
angzm=0
adjaz=0
savesum=0
maxn=4
maxm=0
minn=0
mini=0
minn2=0
mini2=0
reinit_ncps=0
reinit_meme=0
nlev=0
mcut=1 1 1 0
diag=0
n3=0
nsplit=0
freetr=0
ccn=
tid=
wmixmode=1
wmix=0.0001
nwmix=10000
Rank=03, RadialBasis Time= 2.9398E-03 nb=1
Rank=05, RadialBasis Time= 2.9447E-03 nb=1
Rank=04, RadialBasis Time= 2.9468E-03 nb=1
Rank=01, RadialBasis Time= 5.0340E-03 nb=2
Rank=06, RadialBasis Time= 5.0658E-03 nb=2
Rank=02, RadialBasis Time= 5.3672E-03 nb=2
Rank=00, RadialBasis Time= 5.3919E-03 nb=18
Rank=07, RadialBasis Time= 5.4124E-03 nb=2
Rank=00, Time =  5.64098E-03 1 1 1 1 0 4 0 0 0 0
Rank=00, CI Structure...5.6455E-03 1.1090E+09 3.5263E+06
Rank=05, Eff Ham:   0   1   1   1   0  1.764E-02  1.109E+09 -6.10168E+03 -6.10168E+03 -6.10168E+03 -6.10168E+03
Rank=03, Eff Ham:   8   5   5  15   0  1.765E-02  1.109E+09 -6.09702E+03 -6.09554E+03 -6.09702E+03 -6.09554E+03
Rank=07, Eff Ham:   7   1   1   1   0  1.764E-02  1.109E+09 -6.08774E+03 -6.08774E+03 -6.08774E+03 -6.08774E+03
Rank=07, Eff Ham:  13   5   5  15   0  1.768E-02  1.109E+09 -6.09914E+03 -6.08774E+03 -6.09914E+03 -6.08774E+03
Rank=06, Eff Ham:   3   1   1   1   0  1.764E-02  1.109E+09 -6.08738E+03 -6.08738E+03 -6.08738E+03 -6.08738E+03
Rank=01, Eff Ham:   5   3   3   6   0  1.764E-02  1.109E+09 -6.09923E+03 -6.09813E+03 -6.09923E+03 -6.09813E+03
Rank=00, Eff Ham:   9   5   5  15   0  1.765E-02  1.109E+09 -6.09914E+03 -6.09384E+03 -6.09914E+03 -6.09384E+03
Rank=04, Eff Ham:   4   2   2   3   0  1.764E-02  1.109E+09 -6.09705E+03 -6.09604E+03 -6.09705E+03 -6.09604E+03
Rank=03, Eff Ham:  12   5   5  15   0  1.767E-02  1.109E+09 -6.09702E+03 -6.09554E+03 -6.09702E+03 -6.09554E+03
Rank=05, Eff Ham:  16   3   3   6   0  1.768E-02  1.109E+09 -6.10168E+03 -6.09566E+03 -6.10168E+03 -6.09566E+03
Rank=02, Eff Ham:   1   1   1   1   0  1.764E-02  1.109E+09 -6.09927E+03 -6.09927E+03 -6.09927E+03 -6.09927E+03
Rank=06, Eff Ham:  20   1   1   1   0  1.771E-02  1.109E+09 -6.09609E+03 -6.08738E+03 -6.09609E+03 -6.08738E+03
Rank=07, Eff Ham:  17   2   2   3   0  1.769E-02  1.109E+09 -6.09914E+03 -6.08774E+03 -6.09914E+03 -6.08774E+03
Rank=00, Construct Effective Hamiltonian 15  1.80508E-02  1.10941E+09
Rank=00, MBPT structure cfgpair: 0 0 0 0 0 43 43 83 83  1.81509E-02  1.10941E+09
Rank=00, MBPT structure beg:  1.81548E-02  1.10941E+09
 FAIL IN RECOUPLING PROGRAMME
 FAIL IN RECOUPLING PROGRAMME
 FAIL IN RECOUPLING PROGRAMME

Input of FAC

Below is the Python script I used for the calculations. I tried setting InitializeMPI to 2, 4, 8, 16, and 32 cores, but all calculations crashed. The Structure function can run in parallel. Therefore, I suppose the problem is not caused by the MPI library on my machine.

from pfac.fac import *

InitializeMPI(8)

# Setting atomic parameters
z = 50 # Sn
k = 36 # Sn14+
a = ATOMICSYMBOL[z]
SetAtom(a)
SetUTA(0)

# Setting Configurations
Closed("1s 2s 2p 3s 3p 3d")
Config("4s2 4p6", group='ng')
Config("4s1 4p6 4d1", group='e11')
Config("4s2 4p5 4f1", group='e12')
Config("4s2 4p5 4d1", group='e21')
Config("4s1 4p6 4f1", group='e22')
Config("4s2 4p5", group='nii')

ConfigEnergy(0)
OptimizeRadial(['ng'])
ConfigEnergy(1)

g = ['ng', 'e11', 'e12', 'e21', 'e22', 'nii']
# Structure('b.en', g) # CI method
StructureMBPT('b.en',  ['mbpt_corr_b.h','mbpt_tot.h'], g, 6, [4], [4], 6) # First form StructureMBPT
# StructureMBPT('b.en', 'mbpt_tot.h', ['mbpt_corr_b.h'], g, 6) # Third form StructureMBPT

# Outputs
MemENTable('b.en')
PrintTable('b.en', 'a.en', 1)

FinalizeMPI()