[gmx-users] GROMACS with SGI native compliers
nmoore at physics.umn.edu
Thu Aug 4 23:28:53 CEST 2005
Did you test the LYS/PME benchmark
(http://www.gromacs.org/benchmarks/single.php)? Usage of the Partial Mesh
Ewald sum (PME) requires a bunch of Fourier transforms.
> Thank you, Nathan, for your tips!
>> Also, if you're doing a bunch of PME stuff you might want to either (1)
>> replace FFTW with the SGI-specific LAPACK routine (I think has name
>> similar to "complib.sgimath"
> The complib.sgimath library has been replaced by the scs library (you
> can type "man scsl" for more information). One of my colleagues wrote a
> wrapper that translated FFTW calls to the corresponding SCSL routine,
> but it was never released. I tested it in-house with GROMACS, but the
> only test cases I have are the GROMACS standard benchmark suite, none of
> which seems to make heavy use of FFTW...? Anyway, the result was that I
> saw no improvement using FFTW vs. SCSL.
> -- Haruna @ SGI
> gmx-users mailing list
> gmx-users at gromacs.org
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
More information about the gromacs.org_gmx-users