Query Regarding Memory Allocation for Hessian Calculation

[Balakrishnan Viswanathan]

Dear Forum members,

     I have recently started a new project and am trying to run simple test jobs to ensure that I have the correct parameters set up. I am using version 6.8.1 on Compute Canada resources. I am now running up against a perplexing issue with memory allocation during calculation of the (analytic) Hessian matrix. The job seems to die after asking me to allocate space for computation of the one-electron integrals. However, I have allocated the same amount of space in my input file.

My input file is:

 scratch_dir /home/bviswana/scratch/
 memory 2048 mb
 basis spherical
   * library cc-pv6z
 end
 scf
    singlet
    rhf
    thresh 1d-7
    maxiter 200
 end
 geometry units angstroms
 symmetry c2v
   zmat
     h
     f 1 r
     variables
     r 0.95
   end
 end
 set int:txs:limxmem              317576602
 title "Harmonic Frequency"
 task scf energy
 task scf freq

The output file is attached. The .out file prepared by the scheduler reads

The following have been reloaded with a version change:
  1) gcccore/.5.4.0 => gcccore/.7.3.0
  2) icc/.2016.4.258 => icc/.2018.3.222
  3) ifort/.2016.4.258 => ifort/.2018.3.222
  4) imkl/11.3.4.258 => imkl/2018.3.222
  5) intel/2016.4 => intel/2018.3
  6) openmpi/2.1.1 => openmpi/3.1.4

Last System Error Message from Task 5:: Function not implemented
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 5 in communicator MPI COMMUNICATOR 4 DUP FROM 0
with errorcode -1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------

I am unable to determine why the job fails.

If I am not mistaken, the set command (in the input deck) asks for about 317 MB. I have provided 2 GB per core (and 48 cores) of total memory (I am running whole node calculations as I am using the localscratch). Where am I going wrong? I look forward to any advice that can help me proceed.

Regards,

Bala

[Edoardo Aprà]

NWChem can perform Hessian calculation using basis set up to h orbitals.

Since your basis set requires i orbitals, you will not able to get this calculation done.

An option is to use a cc-pv5z basis set instead of cc-pv6z.

By the way, I what is the purpose of running a 6-zeta Hartree-Fock frequency calculations when a triple zeta calculation will show converged results?

[Balakrishnan Viswanathan]

Thank you, Prof. Apra, for your clarification. Does this restriction also apply to numerical computation of the Hessian?

I am looking at clusters involving the monomer above and wanted to determine the operating parameters needed to run my study. I ran Hartree-Fock tests in advance of running correlated (TCE module) tests, as I intend to perform my study at correlated levels of theory.

[Edoardo Aprà]

No, this restriction does not apply to the numerical calculation of frequencies.

[Balakrishnan Viswanathan]

Thank you for your feedback.

I modified the above input to read "task scf freq numerical". When I tried running the job, the output file ended with the following output repeated four times.

 intdd_1eov: could not do hnd integrals                   0
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
  current input line :
     0:
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
 An error occured while computing integrals
 ------------------------------------------------------------------------
 For more information see the NWChem manual at
 http://www.nwchem-sw.org/index.php/NWChem_Documentation


 For further details see manual section:
                                                                                
                                                                                
                                                                                
                   
2:2:intdd_1eov: could not do hnd integrals:: -1
(rank:2 hostname:cdr2340.int.cedar.computecanada.ca pid:69666):ARMCI DASSERT fail. ../../ga-5.6.5/armci/src/common/armci.c:ARMCI_Error():208 cond:0

How can I rectify this error?

Regards,

Bala

[Edoardo Aprà]

Could you upload input and output files to a publicly accessible website (e.g. google drive) and post a link to the uploaded files?

[Balakrishnan Viswanathan]

My input and output files are attached herewith.

The error occurred when I was computing anharmonic frequencies using the VSCF module. Would it be better to to insert freq; read; end before the final step?

--
You received this message because you are subscribed to a topic in the Google Groups "NWChem Forum" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/nwchem-forum/Ub-8LAIdxlo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to nwchem-forum...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/nwchem-forum/73f3b522-9929-43a3-9355-a1568fe7ec03n%40googlegroups.com.

[Edoardo Aprà]

The jobs is not failing in task scf numerical, but in task property, instead

Same issue discussed on February 18th https://groups.google.com/g/nwchem-forum/c/Ub-8LAIdxlo/m/bl5b9iWbAgAJ

Properties require analytical second derivatives that are not available for i-orbitals

[Edoardo Aprà]

Same issue for task vscf
The only solution is to use a basis set that does not contain i-orbitals

[Balakrishnan Viswanathan]

Thank you, Prof. Apra.

To view this discussion on the web visit https://groups.google.com/d/msgid/nwchem-forum/65e8f75c-ea22-4b24-9c68-89aa1885aa85n%40googlegroups.com.