[molpro-user] Paralell jobs
Grant Hill
jghill at wsu.edu
Fri May 8 19:36:10 BST 2009
Cristina,
A simple suggestion to get things started, but are you sure you're
running the mpp version and not mppx?
Grant
On 8 May 2009, at 07:45, Cristina Sanz Sanz wrote:
> Dear all,
>
> Some months ago we got the new 2008 version of MOLPRO and we installed
> it in the cluster we have
> in our group here in Birmingham. I have been running jobs in what I
> thought it was paralell, but I have
> realised the last few days that it might be that paralell running is
> not
> working well.
> I have run the same calculation using only 1 processor and then
> using 4
> processors. At the very
> beginning in the MOLPRO file I see the number of processors is
> different, so that made me think that
> things were OK. When I look to the CPU times at the end of the files I
> see that there are not big differences,
> what then makes me think that it isn't running in paralell.
> I include the CPU times in both output files:
>
> -->> 1 Processor
>
> CPU TIMES * 7091.24 199.30 0.84 0.66 0.47
> 0.29 0.66 0.49 0.38 0.27 0.91
> REAL TIME * 8398.50 SEC
> DISK USED * 121.78 MB
>
> -->> 4 Processors
>
> CPU TIMES * 7352.23 199.35 0.40 0.34 0.26
> 0.18 0.34 0.28 0.23 0.18 0.45
> REAL TIME * 8990.41 SEC
> DISK USED * 150.89 MB
> SF USED * 68.66 MB
>
> When I see the walltimes, it turns that the job running in 4
> processors
> is 10 minutes longer that the one running only
> 1 processor.
>
> Does anybody have any clue why this could be? Is it related to a bad
> installation? Any suggestion?
>
> Thank you very much in advance.
>
> Cristina
>
> --
> Dr. Cristina Sanz Sanz
> School of Chemistry Phone: +44 (0)121 414 7475
> University of Birmingham Fax: +44 (0)121 414 4403
> Birmingham B15 2TT email: C.SanzSanz at bham.ac.uk
>
> Web page: http://che-gwlinux4.bham.ac.uk/~csanz/index.html
>
> _______________________________________________
> 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