NWChem 6.6-27746 error with aug-cc-pvdz but not with 6-31g

Clicked A Few Times

1:12:59 PM PST - Tue, Dec 12th 2017

Dear all,

I am using the Ubuntu package of NWChem 6.6-27746 on a workstation (hostname "attractor") with a Xeon E5-2620v4, 64 GB RAM and an EVGA GTX1080. I can also successfully compile the same version, and I receive errors very similar to the ones below which are produced with the apt-derived package.

uname -a
Linux attractor 4.13.0-19-generic #22-Ubuntu SMP Mon Dec 4 11:58:07 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

So what works? When I run the H2O-example file (basis set is 6-31g), it produces the correct energy without hickups. For the Neon file it runs without errors with 6-31g, but if this is changed to aug-cc-pvdz, as in the original exampe file, the run fails with these lines:

      Superposition of Atomic Density Guess
      -------------------------------------

 Sum of atomic energies:        -128.49634957
  from getmem: mem. needed=               189886  , mem. available=               189091
 Error no.                     1  in getmem  memory overflow : call no., amount requested :                   85                45960
 ------------------------------------------------------------------------
texas: nerror called                   0
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
  current input line : 
     9: task scf
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
 An error occured while computing integrals
 ------------------------------------------------------------------------
 For more information see the NWChem manual at http://nwchemgit.github.io/index.php/NWChem_Documentation


 For further details see manual section:                                                                                                                                                                                                                                                                
0:0:texas: nerror called:: -1
(rank:0 hostname:attractor pid:3911):ARMCI DASSERT fail. ../../ga-5-4/armci/src/common/armci.c:ARMCI_Error():208 cond:0
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 0 in communicator MPI COMMUNICATOR 4 DUP FROM 0
with errorcode -1.

What are these errors related to? I have read that the ARMCI DASSERT errors can be linked to memory issues, so here are my kernel.shmall and .shmmax settings:

cat /proc/sys/kernel/shmmax
68719476736

cat /proc/sys/kernel/shmall
18446744073692774399

But this "texas"-error I cannot find much about... There is a Texas integrals patch, but shouldn't the patches have been applied to the Ubuntu package already? (Last modification to package was Fri, 04 Aug 2017). See: [1]

Let me know if you want me to run any diagnostic commands!

Any help is appreciated - thanks in advance!
best regards
Åsmund

Forum Vet

11:19:21 AM PST - Wed, Dec 13th 2017
Åsmund Could you post the complete input file? Thanks

Clicked A Few Times

11:27:55 AM PST - Wed, Dec 13th 2017

Hi Edoapra!

Thanks very much for looking and sorry for forgetting to post the input files!

This works:

start ne
title "Neon"
geometry; ne 0 0 0; end
basis spherical 
  ne library 6-31g
end
scf; thresh 1e-10; end
task scf

This produces the error:

start ne
title "Neon"
geometry; ne 0 0 0; end
basis spherical 
  ne library aug-cc-pvdz
end
scf; thresh 1e-10; end
task scf

Please let me know if there is any other information I can provide,

Åsmund

Forum Vet

11:32:51 AM PST - Wed, Dec 13th 2017
By the way, what version of Ubuntu are you using? Xenial? Thanks

Clicked A Few Times

11:35:12 AM PST - Wed, Dec 13th 2017
Artful Aardvark (17.10) 64-bit Maybe I shouldn't be using the newest version for a workstation, but rather the latest LTS (16.04)? Thanks again!

Forum Vet

11:41:28 AM PST - Wed, Dec 13th 2017

The Xenial 6.6+r27746-2 package seems to work ...

program         = /usr/bin/nwchem
    date            = Wed Dec 13 10:40:11 2017

    compiled        = Mon_Feb_15_08:24:17_2016
    source          = /build/nwchem-MF0R1k/nwchem-6.6+r27746
    nwchem branch   = 6.6
    nwchem revision = 27746
    ga revision     = 10594

      Superposition of Atomic Density Guess
      -------------------------------------

 Sum of atomic energies:        -128.49634957

      Non-variational initial energy
      ------------------------------

 Total energy =    -128.496350
 1-e energy   =    -182.403715
 2-e energy   =      53.907365
 HOMO         =      -0.853040
 LUMO         =       0.287355


 Starting SCF solution at       0.2s



 ----------------------------------------------
         Quadratically convergent ROHF

 Convergence threshold     :          1.000E-10
 Maximum no. of iterations :           30
 Final Fock-matrix accuracy:          1.000E-12
 ----------------------------------------------


 #quartets = 1.035D+03 #integrals = 6.007D+03 #direct =  0.0% #cached =100.0%


 Integral file          = ./ne.aoints.0
 Record size in doubles =  65536        No. of integs per rec  =  43688
 Max. records in memory =      3        Max. records in file   =  89550
 No. of bits per label  =      8        No. of bits per value  =     64


              iter       energy          gnorm     gmax       time
             ----- ------------------- --------- --------- --------
                 1     -128.4963497305  2.94D-07  1.99D-07      0.1
                 2     -128.4963497305  9.47D-12  8.33D-12      0.1


       Final RHF  results 
       ------------------ 

         Total SCF energy =   -128.496349730514
      One-electron energy =   -182.403715086633
      Two-electron energy =     53.907365356119
 Nuclear repulsion energy =      0.000000000000

        Time for solution =      0.0s

Quote:Twotoneblue Dec 13th 10:35 am

Artful Aardvark (17.10) 64-bit

Maybe I shouldn't be using the newest version for a workstation, but rather the latest LTS (16.04)?

Thanks again!

Clicked A Few Times

11:47:34 AM PST - Wed, Dec 13th 2017
Brilliant, thank you for trying it out! I think I will roll back to Xenial then. I will post back here as soon as I try it out. I am at your service if any other tests related to the above error are of interest to development.

Forum Vet

11:55:02 AM PST - Wed, Dec 13th 2017

Yes, I have just tried your input file on an Artful VM and I can confirm you the breakage you reported.
You might want to tell the Ubuntu maintainer to look into NWChem release 6.8 that we will release very soon.
https://github.com/nwchemgit/nwchem

    date            = Wed Dec 13 10:51:19 2017

    compiled        = Sat_Aug_05_04:04:35_2017
    source          = /build/nwchem-fU7ylv/nwchem-6.6+r27746
    nwchem branch   = 6.6
    nwchem revision = 27746
    ga revision     = 10594

     Superposition of Atomic Density Guess
      -------------------------------------

 Sum of atomic energies:        -128.49634957
  from getmem: mem. needed=               189886  , mem. available=               189091
 Error no.                     1  in getmem  memory overflow : call no., amount requested :                   85                45960
 ------------------------------------------------------------------------
 texas: nerror called                   0
 ------------------------------------------------------------------------
 ------------------------------------------------------------------------
  current input line : 
     8: task scf
 -------------------------------------------------------

Clicked A Few Times

12:06:53 PM PST - Wed, Dec 13th 2017
OK, I'm glad you could reproduce the error. I have filed a question at Launchpad: https://answers.launchpad.net/ubuntu/+source/nwchem/+question/661648 Thanks very much for changing the description and replying to it, and have a very good day, Dr. Aprà =)

Clicked A Few Times

2:15:58 PM PST - Wed, Dec 13th 2017
I can confirm that the Xenial package does not produce the errors mentioned above for the Artful package. The example energies are on the dot, with the exception of the two last examples: SCF energy of H2CO using ECPs for C and O: Example energies: Final RHF results Total SCF energy = -22.507927218024 One electron energy = -71.508730162974 Two electron energy = 31.201960019808 Nuclear repulsion energy = 17.798842925142 I obtain: Final RHF results Total SCF energy = -22.507927225240 One-electron energy = -71.508730186080 Two-electron energy = 31.201960035698 Nuclear repulsion energy = 17.798842925142 So the Nuclear repulsion energy is the same, while the others differ beyond the 7th decimal. MP2 optimization and CCSD(T) on nitrogen: Example states: The final MP2 energy is -109.383276, and the CCSD(T) energy is -109.399662. I obtain: Total MP2 energy -109.383276325499 Total CCSD(T) energy: -109.399661152291571 So the MP2 energy may be the same, while the CCSD(T) energy differs in the 6th decimal. Hope this is useful, Thanks again! best regards Åsmund

Forum >> NWChem's corner >> General Topics