[molpro-user] RESTART: optimization task

Ol Ga o.gaeurisco at gmail.com
Wed Jul 21 20:10:07 BST 2010


Dear Grant,

Thank you for attention to my post in the mail list. I would provide
additional requested information.

The job was submitted from the same user directory on the same nodes.
But I should underline that the scheduler system deletes files in
tmp-directory, but doesn't delete anything in user directory.

A part of output file concerning to RESTART program:
==========
 PROGRAM * RESTART
 **********************************************************************************************************************************
 PROGRAMS   *        TOTAL   RESTART
 CPU TIMES  *         0.25      0.00
 REAL TIME  *         0.65 SEC
 DISK USED  *         0.00 MB
 **********************************************************************************************************************************
 SETTING GEOMTYP        =    XYZ

 Variable memory set to  500000000 words,  buffer space   230000 words

 SETTING BASIS          =    O=VDZ,C=VDZ,H=VDZ
========= etc.

What keyword or something else should I apply to make this type of job
(geom. opt. with numerical hessian and gradients) restartable?

I would hope I clarified the situation. But I can provide any
additional information missed by my unavailability to make a diagnosis
of the problem. Possible the scheduler system deleted something
important like file with last geom. or gradients or etc etc.

Sincerely,
Oleg


On 7/21/10, Grant Hill <jghill at wsu.edu> wrote:
> Dear Oleg,
>
> Are you making file 2 permanent anywhere? I'm not sure how the job could be
> restarted without it.
>
>
> Grant
>
>
> On 21 Jul 2010, at 08:08, Ol Ga wrote:
>
>> Dear Molpro Users and Developers,
>>
>> I observed a problem with a RESTART capability of a molpro job: MOLPRO
>> didn't make a restart of geometry optimization when I used mentioned
>> below input. It starts usual calculation - freq. calculation, then
>> opt.
>> Would I ask you please how to overcome the situation? The calculation
>> takes a lot of time but time-limit is a hard limit of a cluster (but
>> freq. calculation passed with several geom. opt. cycles).
>>
>> A restart file:
>>
>> ***for molpro
>> RESTART;
>> memory, 500,m
>> geomtyp=xyz
>> geometry={
>> 9
>>
>> C              1.2161039424  -0.2039410323   0.5982856002
>> C              2.0014322291  -0.2678446520  -0.5478486196
>> H              2.7259403611   0.4959259713  -0.7623311408
>> H              1.7916050225  -0.9833194730  -1.3218753720
>> H              0.6492539822  -1.0679929648   0.8949804060
>> H              1.5335932676   0.4278736078   1.4072982739
>> O             -0.2716324651   0.9144777229   0.1781837837
>> O             -1.8618804350  -0.5856308591   0.3288621744
>> O             -1.2522679048   0.1780536793  -0.4556371056
>> }
>> basis={O=VDZ,C=VDZ,H=VDZ}
>> {hf;wf,40,1,0,0}
>> {multi,MAXIT=512;occ,24;closed,15;wf,40,1,0,0}
>> {rs2с,maxit=512,maxit=512}
>> FREQUENCIES,TASKREC=5500.2;
>> {optg,root=2,gradient=1.d-3,MAXIT=100,method=QSD}
>> --
>>
>> A job aborted by time-scheduler:
>>
>> ***for molpro
>> memory, 500,m
>> geomtyp=xyz
>> geometry={
>> 9
>>
>> C              1.2161039424  -0.2039410323   0.5982856002
>> C              2.0014322291  -0.2678446520  -0.5478486196
>> H              2.7259403611   0.4959259713  -0.7623311408
>> H              1.7916050225  -0.9833194730  -1.3218753720
>> H              0.6492539822  -1.0679929648   0.8949804060
>> H              1.5335932676   0.4278736078   1.4072982739
>> O             -0.2716324651   0.9144777229   0.1781837837
>> O             -1.8618804350  -0.5856308591   0.3288621744
>> O             -1.2522679048   0.1780536793  -0.4556371056
>> }
>> basis={O=VDZ,C=VDZ,H=VDZ}
>> {hf;wf,40,1,0,0}
>> {multi,MAXIT=512;occ,24;closed,15;wf,40,1,0,0}
>> {rs2с,maxit=512,maxit=512}
>> FREQUENCIES,TASKREC=5500.2;
>> {optg,root=2,gradient=1.d-3,MAXIT=100,method=QSD}
>> ---
>>
>>
>> Sincerely,
>> PhD student Oleg Gadzhiev
>>
>> PS: I have read in the list about problems with restarts. i hope my
>> case can be successfully passed.
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
>
>



More information about the Molpro-user mailing list