[Columbus] cidrtfl and parallel CI

[Andy Young]

Hello Professor Lishka,

The MEX with polyhes is extremely stable.  This is
using the COLUMBUS distribution as you have provided
without any modification.  Also, I am able to run
parallel single point calculations with the
distribution largely as you have provided, but some
times it won't run in parallel but I think that is
becaue of memory limitations on my various set ups.  
It is only when I am combining the two types of runs
that I have problems.  

-Andy

--- Hans Lischka <hans.lischka at univie.ac.at> wrote:
> Hi Andy,
> I did not say that the two cidrt programs produce
> the same cidrtfl, but 
> that they are equivalent (except for an
> inconsistency in the interacting 
> space part). You can construct the same DRT with
> different ways of 
> input. Ron and Thomas have devised several
> alternatives, which will 
> (can) lead to the same result.
> You are on a completely wrong track when you
> concentrate on the drt 
> issue! I still don't know what you have been doing
> exactly, but the 
> problems you are encountering could arise from
> naming conventions of 
> states in runc, for example.
> According to my information, the incorporation of
> the parallel CI into 
> an MXS determination  has never been tested. You
> should proceed in steps.
> Did you already manage to run a MXS/Polyhes search
> in a sequential test 
> job? I mean using standard colinp/runc scripts?
> You need to separate the question of MXS and
> parallel CI. Did you manage 
> to run a single-point parallel CI calculation
> already (single-state and 
> multistate).




> 
> Which Columbus version do you use. Which computer
> system do you use?
> 
> Best regards, Hans
> 
> 
> Andy Young wrote:
> 
> >Hello Professor Lischka and Professor Mueller and
> >everybody!
> >
> >  Thank you for helping me out.  I have attached
> >tarred runs that were successful.  These show that
> the
> >cidrtfl files are different when I run cidrt.x and
> >cidrtms.x.  
> >The input parameters were the same when I put them
> in
> >with COLINP.  
> > My goal is to be able to run minimizations on an
> >arbitrary state after running the CI in parallel
> and
> >also to be able to minimize a crossing seam of two
> >roots *of the same symmetry* after running the CI
> in
> >parallel.  I am wanting to use polyhes for all of
> >these minimizations.  
> >
> >  
> >
> >>The combination of a multistate CI calculation
> with
> >>a gradient 
> >>calculation is not supported currently! The reason
> >>is that the gradient 
> >>is computed only for one state. 
> >>    
> >>
> >
> >Okay, but COLUMBUS is picking up the cidrtfl file
> that
> >is made from CIDRTMS before it runs CIGRD.  My
> guess
> >is that CIGRD has been tested with a variety of
> >cidrtfl files made from CIDRTMS and not from CIDRT.
> 
> >The runc script is making the cidrtfl file each
> >iteration and it is doing it with the CIDRTMS
> program
> >and not the CIDRT program.
> >  
> >Many thanks,
> >Andy   
> >
> >
> >
> >__________________________________________________
> >Do You Yahoo!?
> >Tired of spam?  Yahoo! Mail has the best spam
> protection around 
> >http://mail.yahoo.com 
> >
> 
> 
> -- 
> Hans Lischka
> Institute for Theoretical Chemistry
> University of Vienna
> Waehringerstrasse 17
> A-1090 Vienna, Austria
> Tel.: +43-1-4277-52757
> Fax: +43-1-4277-9527
> E-mail: hans.lischka at univie.ac.at
> Web: http://homepage.univie.ac.at/hans.lischka
> 
> 



		
__________________________________ 
Yahoo! Mail 
Stay connected, organized, and protected. Take the tour: 
http://tour.mail.yahoo.com/mailtour.html