AMBER Archive (2002)

Subject: Re: replacing fft routines

From: David A. Case (case_at_scripps.edu)
Date: Tue Oct 22 2002 - 17:58:34 CDT


On Tue, Oct 22, 2002, Florian Barth wrote:
>
> does anybody know how to implement a different fft routine into the
> sander code? As far as I know amber uses a one dimensional fft routine
> from sgi for the PME algorithm (at least on my linux box). I would
> expect a multi dimensional fft program like FFTW to perform much better,
> but don`t know if it is possible to use it sander.
> Any comment would be appreciated.
>

You are welcome to try things, but be sure to study carefully the timings
given at the end of sander runs first. On my Linux box, for the "jac"
benchmark (150 residue protein in water) the FFT time is only about 3% of
the total run time. So no amount of tuning of the FFT (by itself)
is going to help very much. Of course more drastic code changes might give
improvements, so I certainly encourage you to look carefully at how things
are done -- maybe you will see something that will help.

...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 ==================================================================