Analytical Frequency for larger systems at Becke97-D/cc-pVTZ : ga reorder: duplicate failed 0


Just Got Here
Dear NWChem users,
I am trying to run a frequency calculation for a system with 105 atoms at Becke97-D/cc-pVTZ level.
Initially, when I tried with 4GB and 8 processors, it terminated with the following error

hess_init: could not allocate g_rhs_xc      555

after that, when I increased the number of processors to 12 keeping the same 4GB memory, it calculated till 1 electron contributions

HESSIAN: the one electron contributions are done in    1694.5s
and it terminated with the following error
ga_reorder: duplicate failed        0

then, I tried with 6GB and 12 processors with different combinations of heap, stack and global memories

(1) memory 6 GB
(terminated after
HESSIAN: the two electron contributions are done in 47901.5s
with an error
ga_create_atom_blocked: gdens1
------------------------------------------------------------------------
ga_create_atom_blocked: ga_create_irreg 2030

 
(2) memory stack 2.25 heap 0.75 global 3 GB
(terminated after
HESSIAN: the two electron contributions are done in 48967.8s
with the same error as in the above option

(3)memory stack 1.5 heap 0.75 global 3.75 GB
(terminated after
HESSIAN: the two electron contributions are done in 47658.1s
with the same error as in the above option

(4)memory stack 1 heap 0.75 global 4.25 GB
(terminated after
HESSIAN: the two electron contributions are done in 48079.1s
NWChem CPHF Module
------------------
scftype = RHF
nclosed = 175
nopen = 0
variables = 324625
# of vectors = 315
tolerance = 0.10D-03
level shift = 0.00D+00
max iterations = 50
max subspace = 3150
SCF residual: 9.58843312846938156E-005
Iterative solution of linear equations
No. of variables 324625
No. of equations 315
Maximum subspace 3150
Iterations 50
Convergence 1.0D-04
Start time 147534.7
iter nsub residual time
---- ------ -------- ---------
------------------------------------------------------------------------
ga_reorder: duplicate failed 0


After this I tried various combinations like 8GB and 12 processors, 10GB and 10 processors, 14GB and 8 processors.
Everytime it ended up with the same eror as in (4)

However, for a system with 60 atoms, the analytical frequency calculation at same level successfully finished with 4GB and 8 processors.

Please help me in solving this problem

Thank You,
Shilpi

Forum Vet
Shilpi
Could you post (or upload to an accessible website) your input file?

Just Got Here
Dear Edo,
It worked with a higher(6GB) global memory than what I specified in (4).
The job finished successfully.
Thank You.

Shilpi.


Forum >> NWChem's corner >> NWChem functionality