memory problem for running CCSD(T) jobs


Click here for full thread
Clicked A Few Times
Thanks a lot.

Do you think how much global memory should I increase to ?

By the way, do I need to increase stack memory too?

Zongtang



Quote:Bert Aug 8th 2:20 pm
It looks like we're somehow underestimating the amount of global memory needed for the triples part of the calculation. I would recommend you increase the global memory block.

Thanks,

Bert


Quote:Zfang Aug 7th 7:39 pm
I am running a CCSD(T) with aug-cc-pVTZ-PP basis set single point job with nwchem. The total basis function is 954.
There is no problem before it did triple.
However, the triples calculation couldn't proceed,
the error message is "ga_create g_objv failed 0".

I used 255 nodes with 4 cores per node at EMSL.

The input file is attached below.

Hope some one can give me suggestions to figure out the problem and fix that. Thanks a lot.


echo
start mo3o9.ethanol.complex.ccsdt.at
memory stack 1500 mb heap 100 mb global 2000 mb noverify
title "mo3o9.ethanol.complex CCSD(T)/aug-cc-pVTZ(-PP)//B3LYP/cc-pVDZ(-PP)"
charge 0

geometry
 MO     0.714207    2.031548   -0.007232
MO 2.307475 -1.049091 0.030346
MO -1.321761 -1.003274 -0.065257
O 0.717072 2.965045 -1.422640
O 0.700697 3.087421 1.319778
O 2.205515 0.849172 0.044627
O -0.775577 0.899103 0.042082
O 2.986805 -1.635643 1.466597
O 3.242802 -1.616334 -1.262012
O -2.010045 -1.740565 -1.426571
O 0.507937 -1.570080 -0.143301
O -1.921222 -1.765465 1.325778
O -3.219413 0.220409 -0.065770
H -3.096979 1.046882 0.423276
C -4.522423 -0.375798 0.225670
H -4.600897 -1.206058 -0.483010
H -4.494516 -0.781630 1.245441
C -5.630007 0.641713 0.036365
H -6.597368 0.161151 0.236592
H -5.527963 1.483792 0.736819
H -5.640933 1.030164 -0.989866
end

.......

scf
  vectors output mo3o9.ethanol.complex.ccsdt.at.movecs
rhf
singlet
maxiter 100
  1. thresh 1.0e-10
  2. tol2e 1.0e-10
end

task scf

ccsd
  maxiter 60
thresh 1e-6
freeze 24
end
task ccsd(t)