[CPMD-list] CPMD on 64-bit Linux (openSUSE 10.3) Intel Xeon

[Vladimir Stegailov]

Hi, Axel!

> FFLAGS= -O2 -unroll -pc64 -march=pentium3 -mtune=core2
> LFLAGS= -L/opt/intel/mkl/10.0.2.018/lib/em64t -lmkl_intel_lp64 -lmkl_sequential 
>  -lmkl_core

A couple of may be stupid questions. Is it usually normal to mix options for 
different types of cpus ('pentium3' and 'core2' options)?
Is 'core2' relevant to Xeons? Not only to Core2 cpus?

> additional speedup can be achieved by using FFTW instead of the
> default internal default FFT. for that you'll have to install
> fftw-2 (not fftw3!), change -DFFT_DEFAULT to -DFFT_FFTW and
> add -ldfftw (or -lfftw depending on your Linux) to LFLAGS.
> the major source for the speedup is actually less the speed
> of the FFT different implementations, but the fact that FFTW
> supports more grids, so you may find a smaller grid that
> fits your problem and thus reduces the amount of data to be
> processed (and the memory needed).

Is it a general recommendation for any platform?
Can it improve the scalability of the code (since less amount of data should 
be communicated)?

best regards,
Vladimir


>
> this will link with the non-openmp version of the libraries
> and thus no -lpthread or -lguide should be needed.
>
> with intel MKL v10 this is required unless you set
> OMP_NUM_THREADS to 1 in your default environment.
>
> additional speedup can be achieved by using FFTW instead of the
> default internal default FFT. for that you'll have to install
> fftw-2 (not fftw3!), change -DFFT_DEFAULT to -DFFT_FFTW and
> add -ldfftw (or -lfftw depending on your Linux) to LFLAGS.
> the major source for the speedup is actually less the speed
> of the FFT different implementations, but the fact that FFTW
> supports more grids, so you may find a smaller grid that
> fits your problem and thus reduces the amount of data to be
> processed (and the memory needed).
>
> cheers,
>   axel.
>
> OI> Based on IFORT-AMD64
> OI>
> OI>
> OI> dear cpmd users,
> OI> any ideas or help will be appreciated
> OI>
> OI> Our system administrator asks  me pass on his thanks also.
> OI>
> OI> best regards,
> OI> oksana
> OI>
> OI>
> OI>
> OI> * cpmd-list-request at cpmd.org [Sat, 05 Apr 2008 05:00:45 +0200]:
> OI> > Send CPMD-list mailing list submissions to
> OI> > cpmd-list at cpmd.org
> OI> >
> OI> > To subscribe or unsubscribe via the World Wide Web, visit
> OI> > http://cpmd.org/mailman/listinfo/cpmd-list
> OI> > or, via email, send a message with subject or body 'help' to
> OI> > cpmd-list-request at cpmd.org
> OI> >
> OI> > You can reach the person managing the list at
> OI> > cpmd-list-owner at cpmd.org
> OI> >
> OI> > When replying, please edit your Subject line so it is more specific
> OI> > than "Re: Contents of CPMD-list digest..."
> OI>
> OI>
> OI> _______________________________________________
> OI> CPMD-list mailing list
> OI> CPMD-list at cpmd.org
> OI> http://cpmd.org/mailman/listinfo/cpmd-list
> OI>
>
> -- 
> =======================================================================
> Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
>   Center for Molecular Modeling   --   University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.