[Pw_forum] acml - mkl

[Axel Kohlmeyer]

On Tue, 13 May 2008, ambavale sagar wrote:

AS> Dear Pwscf users,

AS> I have x86_64 (i.e. AMD64) architacture machine with Xeon 
AS> (x5450-quad core) processors having dual cpu. I have installed intel 
AS> fortran compiler, c compiler and mkl. But as i compile espresso4cvs2 
AS> version it does not recognise BLAS,LAPACK libraries. Must i install 

if configure does not find MKL, it either means you have not 
installed it correctly or to a location that configure does not
know about. check the config.log file and simply adapt the make.sys
file manually according to the MKL user's guide (RTFM!!).

AS> ACML as i have AMD architacture? During compilation of acml i 

both the 64-bit ACML and MKL should in principle work interchangably.
one would expect ACML to work better on AMD cpus and MKL on intel,
but in real life that varies.

AS> couldn't find driver.f( They say that driver.f is application 
AS> program). What should be the procedure?

you didn't read the documentation correctly. ACML does not need to
be compiled. this obviously explains how to compile an example
application to use ACML. since QE uses the standard BLAS/LAPACK
calls you only have to adapt the respective entries in make.sys
it even has comments to explain. if you have installed MKL 
(version 10.x) correctly a simple -lmkl would do for BLAS (and
automatically LAPACK) in case of a serial compile. if you want
to compile a parallel version, things get more complicated.
please see the multitude of previous discussions on this subject
in the mailing list archive.

SA> One more question: Is it possible to have Xeon processors with ia-64 
SA> architacture? 

this is impossible. IA64 is itanium. 

cheers,
   axel.


regards
Sagar Ambavale
Research Student
The M.S.University of Baroda
India



      Best Jokes, Best Friends, Best Food and more. Go to http://in.promos.yahoo.com/groups/bestofyahoo/

-- 
=======================================================================
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.