Freq calculation fails
Just Got Here
1:00:37 PM PDT - Tue, Jun 3rd 2014
Dear NWchem users,

I am trying to compute the frequencies at CCSD(T) level with basis aug-cc-pvtz, the system is composed of several H and two C, 


CCSD(T) correlation energy / hartree =        -0.527427946764739

CCSD(T) total energy / hartree       =       -81.167505120376788

Cpu & wall time / sec          394.6          423.0

step    -5 energy         -81.16750512

           gradient         0.000039
Last System Error Message from Task 60:: Function not implemented
Last System Error Message from Task 61:: Function not implemented
Last System Error Message from Task 171:: Function not implemented
Last System Error Message from Task 70:: Function not implemented
Last System Error Message from Task 62:: Function not implemented
......
[170:node35] unexpected disconnect completion event from [69:node24]
Assertion failed in file ../../dapl_conn_rc.c at line 1174: 0
internal ABORT - process
Im using a Mellanox infiniband switch (40GB/s) and nwchem(6.3) was compiled with intel fortran v.14
im running using 20 nodes. each with 10 cores (total 200 cores) , each node has 64GB of RAM, my input is as follow :
START C2C8
TITLE "FREQ CCSD(T)/aug-cc-pvtz"
CHARGE 1
permanent_dir /home/soporte/25MAYO/NWCHEM/29may/.tmp
scratch_dir /home/soporte/25MAYO/NWCHEM/29may/.tmp
memory stack 1000 mb heap 1000 mb global 4000 mb verify hardfail

GEOMETRY
C -0.114647490 -1.817948140 0.000000000
H -0.466136430 -1.134637200 0.911164050
H 1.082999070 -1.979001030 0.000000000
H -0.416136430 -1.174637200 -0.911164050
C -0.019682100 1.196162620 0.000000000
H 0.122555850 3.111154510 0.132078720
H 0.030855000 0.317414370 0.000000000
H 0.182555850 2.111154510 -0.132078720
H -2.071412810 1.264114560 0.000000000
END

BASIS
  • library aug-cc-pvtz
END


SCF
SINGLET
rhf
thresh 1.0e-10
TOL2E 1.0e-10
maxiter 99
direct
END

driver
maxiter 99
end


TCE
scf
lshift 0.2
diis 4
maxiter 99
thresh 1e-6
CCSD(T)
cuda 1
END

set fock:densityscreen f
TASK TCE FREQUENCIES

Im thinking that it was a memory issue, but im not sure ...
Thank you very much for you help
Jose Luis

Forum Regular
8:38:07 AM PDT - Wed, Jun 4th 2014
Hi Jose,
Assuming the failure was caused by some intermittent problem you should be able to restart this calculation. Before you try that make sure you keep a copy of the essential files (e.g. cd /home/soporte/25MAYO/NWCHEM/29may/; tar -zcf backup.tgz .tmp). Then change the input file by replacing START with RESTART and remove the GEOMETRY block.
After that just start the calculation again. It should continue where it left off.
Let me know if you encounter any problems.
Huub

Just Got Here
2:56:03 PM PDT - Fri, Jun 27th 2014
Quote:Huub Jun 4th 8:38 am
Hi Jose,
Assuming the failure was caused by some intermittent problem you should be able to restart this calculation. Before you try that make sure you keep a copy of the essential files (e.g. cd /home/soporte/25MAYO/NWCHEM/29may/; tar -zcf backup.tgz .tmp). Then change the input file by replacing START with RESTART and remove the GEOMETRY block.
After that just start the calculation again. It should continue where it left off.
Let me know if you encounter any problems.
Huub

Dear Huub,
I'm thinking that is the switch that is not working correctly (Mellanox, infiniband )
thank you very much for your answer.
Jose Luis


Forum >> NWChem's corner >> Running NWChem