[Smeagol-discuss] =?x-gbk?Q?efficiency=A3=A1?=

[Ivan Rungger]

Hello Guangping,

since you have an intel machine I would say that using ifort and mkl is 
the best choice. We have compiled smeagol with different compilers 
(ifort, pathscale,pgi,xfl), and they all give pretty much the same speed 
(and results). The important thing is to have efficient lapack/blas 
libraries, because that is where most of the compute time is spent. The 
mkl libraries are surely good for intel machines.

Regarding the version of mpi it should not really matter as well, we 
have compiled the code with mpich, openmpi, mvapich and others, and 
usually all of them work. The best maybe is to start with openmpi 
copmiled with intel+mkl, and then if that for some reason should not 
work you might try other versions of mpi.

Cheers,

Ivan

Guangping Zhang wrote:
> Dear users:
> Has anyone studied the different efficiencies of the smeagol compiled 
> with different compilers,mpi and math libraries.
> If my machine is Intel 64 bit,is the ifort compiler and mkl is the 
> best choice? If so,how can choose the mpi,openmpi or mpich or any else?
> Looking forward to any advice!
> ------------------
> *BEST REGARDS!*
> **
> *Guangping Zhang*
> *-----------------------------------*
> *Atom and Melecular Physics*
> *Physics and Electronics College*
> *Shandong Normal University*
> *Shandong,Jinan,China*
> *------------------------------------*
> ------------------------------------------------------------------------
>
> _______________________________________________
> Smeagol-discuss mailing list
> Smeagol-discuss at lists.tchpc.tcd.ie
> http://lists.tchpc.tcd.ie/listinfo/smeagol-discuss