Re: [Mpqc-users] memory

[Francesco P. <fra...@al...>]

On Thursday 13 July 2006 18:16, Curtis Janssen wrote:
> Hi Francesco,
>
> You will use a lot less memory if you use threads instead of a
> ShmMessageGrp. So you should launch with:
>
> mpqc -threadgrp "<PthreadThreadGrp>:(num_threads=4)" filename.in | tee
> filename.out

Is that likely to accelerate the "geometry optimization" process? I want to 
try. In fact, with ShMeggageGrp, max_iterations = 40, and having set 5GB ram 
as indicated below (out of the total 8GB  available), the calculation has now 
ended with "Optimizaton NOT converged" after 40 iterations, all linealy 
converged (which adds to the ca 20 iterations carried out previouly, each 
time restarting with the lowest ebergy point). The LUMO - HOMO difference is 
steadily decreasing toward the longest wavelent UV absorption (that I know in 
ethanol only) but still far.

I understand I am asking too much to my hardware but the energy is decreasing 
and the error is at a good level. May be other 40 iterations ...

francesco
>
> The memory keyword should go in the mole section (just below the
> "mole<CLHF>" line).  It is a suggested value, MPQC will use more, so don't
> set it to all of the memory on your system.
>
> Curt
>
> On Thursday 13 July 2006 08:51, Matt Leininger wrote:
> > Put "memory = 1GB" (or whatever value you want, 100MB, 4GB) into the
> > mole<CLKS> or guess_wavefunction<CLHF> sections.  You probably only need
> > to increase the memory in the mole<CLKS> section.  If you don't have a
> > default memory setting I thought MPQC just used whatever memory it
> > needed.   Joe or Curt could clarify.
> >
> >   - Matt
> >
> > On Thu, 2006-07-13 at 15:40 +0200, Francesco Pietra wrote:
> > > Hardware: two 265 dual opteron, ram 8GB, for a b3lyp MCsearch OO
> > > calculation for a large molecule with command
> > >
> > > $mpqc -messagegrp "<ShmMessageGrp>:(n=4)" filename.in | tee
> > > filename.out
> > >
> > > (restarted using the lowest energy point after that the Linux Debian
> > > kernel had died after 29 hours of calculation and 13 iterations
> > > completed to convergence)
> > >
> > > $top
> > >
> > > shows little usage of memory:
> > >
> > > Mem: total 81711428k total, 4953228k used, 7675952k free, 75860k
> > > buffers,
> > >
> > > Swap 6024248k total, 0k used
> > >
> > > Since - for this hardware - it is a heavy calculation, could either the
> > > launch command or the input file (attached) be modified to get more
> > > memory for the calculation?
> > >
> > > Thanks a lot for helping
> > >
> > > francesco pietra
> > > -----------------------------------------------------------------------
> > >-- Using Tomcat but need to do more? Need to support web services,
> > > security? Get stuff done quickly with pre-integrated technology to make
> > > your job easier Download IBM WebSphere Application Server v.1.0.1 based
> > > on Apache Geronimo
> > > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=12164
> > >2 _______________________________________________
> > > Mpqc-users mailing list
> > > Mpq...@li...
> > > https://lists.sourceforge.net/lists/listinfo/mpqc-users
> >
> > -------------------------------------------------------------------------
> > Using Tomcat but need to do more? Need to support web services, security?
> > Get stuff done quickly with pre-integrated technology to make your job
> > easier Download IBM WebSphere Application Server v.1.0.1 based on Apache
> > Geronimo
> > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> > _______________________________________________
> > Mpqc-users mailing list
> > Mpq...@li...
> > https://lists.sourceforge.net/lists/listinfo/mpqc-users