AMBER Archive (2002)

Subject: Re: Intel Kernel Math Library

From: case (case_at_scripps.edu)
Date: Thu Jun 06 2002 - 10:02:25 CDT


On Thu, Jun 06, 2002, Joffre Heredia wrote:
>
>
> 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.
>

I don't know of any efforts along these lines by the "core" Amber developers,
but we are always happy to take suggestions.

Note, however, that slowness in FFT's is often not a serious problem. For
example, on the "jac" benchmark (150 residue protein in water), about 5%
of the time is spent in the FFT routines (on a 1.2 GHz Athlon processor).
And that benchmark doesn't use the RESPA multiple-time-step algorithm, which
could cut the FFT percentage about in half. So it's not clear that there is
much to be gained by speeding up the FFT libraries.

..good luck...dac

-- 

================================================================== David A. Case | e-mail: case_at_scripps.edu Dept. of Molecular Biology, TPC15 | fax: +1-858-784-8896 The Scripps Research Institute | phone: +1-858-784-9768 10550 N. Torrey Pines Rd. | home page: La Jolla CA 92037 USA | http://www.scripps.edu/case ==================================================================