[phenixbb] refinement using a multiprocessor machine or on a cluster

Ralf W. Grosse-Kunstleve rwgk at cci.lbl.gov
Fri Mar 6 22:20:09 PST 2009


Sorry, I didn't see this message before.
If you compile from sources with gcc 4.3, the raw FFT actually runs in
parallel. But the overall speedup is nothing to write home about.
Unfortunately, in refinement Amdahl's law really bites you:
http://en.wikipedia.org/wiki/Amdahl%27s_law
Ralf

> phenix.refine does not use any advantage of multiple available CPUs. Sorry.
>
> Pavel.
>
>
> On 3/4/09 11:26 AM, Gino Cingolani wrote:
> > Dear Phenix community,
> >
> > I was wondering if it is possible to run a refinement job on multiple
> > processor. The refinement of my 13,000 asymmetric unit takes about 2
> > days. It could be greatly speeded up by taking advantage of the 4 CPUs
> > in my PC. The phenix manual suggest to use 
> > nproc=3
> > to run using a multiprocessor machine or on a cluster.
> > But this applies only to LigandFit and AutoBuild Wizards.
> > What about simple refinement?
> >
> > Thanks in advance,
> >
> > Gino
> >
> >   
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://www.phenix-online.org/mailman/listinfo/phenixbb
>



More information about the phenixbb mailing list