[molpro-user] Segmentation fault in Frequencies

[Werner Győrffy]

Computing frequencies by using the fully analytical Hartree-Fock Hessian 
is limited. Using numerical Hessian is however a less restricted option: 
this works, for example, in integral-direct way (direct,gdirect), or by 
using the density fitting approximation (df-hf). In general, if 
analytical gradients are available for a given method, the gradients are 
differentiated once to build the Hessian, otherwise the energy is 
differentiated twice. For the HF method the default choice

hf
freq

is using numerical Hessian computed from analytical gradients.

Regards,

Werner Gyorffy.

On 11/20/2013 09:04 AM, Shachar Klaiman wrote:
> Dear Werner,
>
> Thank you for the reply. I am sure the tests ran ok ( I run them on 2
> nodes with 16 processes on each node):
>
> SHA1      : f07512e756848f49f1178b36f2423c88963342df
>   ARCHNAME  : Linux/x86_64
>   FC        :
> /sw/laki-SL6x/hlrs/compiler/intel/parallel_studio_xe_2013-4-192/composer_xe_2013.5.192/bin/intel64/ifort
>   FCVERSION : 13.1.3
>   BLASLIB   :
> -L/sw/laki-SL6x/hlrs/compiler/intel/parallel_studio_xe_2013-4-192/composer_xe_2013.5.192/mkl/lib/intel64
> -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core
>   id        : heid
>
>   Nodes     nprocs
>   n082201     15
>   n052902     15
>   Number of processes for MPI-2 version of Molpro:   nprocs(total)=   32
>    nprocs(compute)=   30   nprocs(helper)=    2
>   ga_uses_ma=false, calling ma_init with nominal heap.
>   GA-space will be limited to   8.0 MW (determined by -G option)
>
>   Using customized tuning parameters: mindgm=16; mindgv=24; mindgc=8;
> mindgr=1; noblas=0; minvec=7
>   default implementation of scratch files=sf
>
>   ***,benzene
> ------
> No errors detected. Max error:  energy=3.58226089D-08
>
>   PROG                  E
>   OPTG (HF)        -227.89136038
>   ZPE ANALYTICAL      0.11641748
>   ZPE NUMERICAL       0.11641991
>
>
>          OPTG(HF)        HF-SCF
>     -227.89136038   -227.88664699
>   **********************************************************************************************************************************
>   Variable memory released
>
>
>
>
> I traced the error back to gdirect so there seems to be a problem with
> running frequencies in direct mode. Without gdirect there was no seg fault.
>
> Best,
> Shachar
>