Hi all!
After some performance comparations between running SANDER 7 in a SGI
Multiprocessor system and a Linux Beowulf Cluster, we came to a
conclusion:
SGI provides a property math library for fast FFT calculations that
SANDER takes proof of. In the other way, SANDER under Linux is using the
Fortran FFT functions that come with the SANDER package. These functions
are slower than the SGI ones, most surely because SGI took a lot of
effort in programming their math libraries in assembler or any other
low-level language near to the machine-code.
We then started looking for an alternative to the SANDER FFT functions
and found the Intel Kernel Math Library
(
http://www.intel.com/software/products/mkl/mkl52/index.htm). It seems
they have, as SGI, a fast solution to FFT calculations.
As SANDER 7 is using the ewald method by default, I was wondering if
there would be any compiler option for using the Intel MKL instead of
the SANDER built-in FFT functions. In the case this is not possible, I
would like to know if the AMBER developers are thinking about the
possibility of using the Intel MKL in a next release.
Thank you in advance
-------------------------------------------------------------
Joffre Heredia Rodrigo Tel: (34)-93-5813812
Laboratory of Computational Medicine Fax: (34)-93-5812344
Biostatistic Dept.
UAB School of Medicine. Bellaterra Joffre.Heredia_at_uab.es
08193-Barcelona (SPAIN)
-------------------------------------------------------------
Received on Thu Jun 06 2002 - 12:46:32 PDT
