Error message in "dft 3cincor"
Just Got Here
9:51:22 AM PST - Mon, Nov 28th 2016
Dear Developers,

I would appreciate if somebody could help me to handle the following error message:

I have a system with 30 atoms, using M06-HF/Def2-TZVP calculation settings.
The error is the follows:
_________________________________________________________________-

     Non-variational initial energy

     ------------------------------



Total energy =    -400.526562

1-e energy   =   -1259.350525

2-e energy   =     501.794083

HOMO         =       0.244075

LUMO         =       0.246641



  Time after variat. SCF:      9.9



     3 Center 2 Electron Integral Information

     ----------------------------------------

         Maximum number of 3-center 2e- integrals is:        32227428.

           This is reduced with Schwarz screening to:        21063504.

           Incore requires a per proc buffer size of:         2923513.

                 The minimum integral buffer size is:          400737

         Minimum dble words available (all nodes) is:      1566304964

                  This is reduced (for later use) to:      1565901823

                            Suggested buffer size is:         2923513

                             no. integral batches is:            1000



          2.924 MW buffer allocated for incore 3-center

         2e- integral storage on stack.

 nlast n3cdbl                2930676               2923513

------------------------------------------------------------------------

dft_3cincor: blew out of dbls                 999

------------------------------------------------------------------------

------------------------------------------------------------------------

 current input line :

  102: task dft energy

------------------------------------------------------------------------

------------------------------------------------------------------------

This error has not yet been assigned to a category

------------------------------------------------------------------------

For more information see the NWChem manual at

http://nwchemgit.github.io/index.php/NWChem_Documentation




For further details see manual section:

No section for this category

Forum Vet
11:48:24 AM PST - Mon, Nov 28th 2016
This is -- in general -- a symptom that you are using too many processes for this case, therefore one way to fix it is to re-submit asking for few processes/cores.

Another way to fix it is to add the following line just before the task dft line in your input file

set dft:n3c_dbl 4000000

Just Got Here
1:09:29 AM PST - Tue, Nov 29th 2016
Quote:Edoapra Nov 28th 10:48 am
This is -- in general -- a symptom that you are using too many processes for this case, therefore one way to fix it is to re-submit asking for few processes/cores.

Another way to fix it is to add the following line just before the task dft line in your input file

set dft:n3c_dbl 4000000


Dear Edoapra,

Thank you for your answer

I'm afraid that the problem is not so simple.
I try to modelate the extra charge fluctuation between the DNA stacked bases, and I use the input presented for TCNE dimer using the Real-time TDDFT module (end of the documentation page).

First, I have 3 different Linux systems, where this part of the NWChem behave in different ways. In one of them compiled with Intel Fortran - OpenMPI the program does not run at all, whatever dft:n3c_dbl or number of processors I set. In the second system (Intel Fortran - Mpich), I was able to run for adenine-adenine case, but for the slightly larger guanine-guanine case the program fails. Here the problem comes not at the first "task dft" call but at the third one, when the dimer is built using the monomer calculations. In the third case (same Intel Fortran - Mpich) I have not seen any problem (the true is that I had a single run till now).

Best regards
Attila

Forum Vet
12:41:05 PM PST - Tue, Nov 29th 2016
Thanks for providing more details about your problem.
My initial diagnosis proved to be off the mark, since you are clearly not over-subscribing your run, but the problem lies in the incorrect estimate of the memory requirement to store the fitted Coulomb integrals.
Therefore, the only solution you have is to add the line
set dft:n3c_dbl 4000000
I have picked the value of 4000000 since the it is larger than the value of 2930676 printed when NWChem stopped. If you still have a failure with a value larger than 4000000, please increase the value further.


Forum >> NWChem's corner >> General Topics