[gmx-users] Compile Gromacs for Blue Gene/L

[Erik Lindahl]

Hi Eric,

I think you'll be _very_ happy with Gromacs4, or the current CVS  
version if you want to test that :-)

One problem is that "relative scaling" is a rather pointless measure,  
since it will be better the slower the code is on a single CPU (less  
time spent in communication). Our parallelization goal is rather been  
to get the highest possible absolute performance on any given number  
of nodes.

 From what I've seen, I still think Blue Matter might show better  
relative scaling when you use several thousand of nodes for small  
systems, but Gromacs4 will beat that absolute performance using a  
(significantly) smaller number of nodes.

Cheers,

Erik

On Apr 8, 2007, at 11:19 PM, Eric Jakobsson wrote:

> As I understand it, Blue Matter is a very stripped down laboratory  
> code, so I think the scaling is the only large virtue that one  
> would want to acquire.
>
> At 02:06 AM 4/7/2007, you wrote:
>> Eric Jakobsson wrote:
>>> Not a direct response but a thought:  Wouldn't it be wonderful to  
>>> have a code that combined the best qualities of Gromacs and Blue  
>>> Matter?  Is it possible to assemble a group or community of  
>>> people to do that?
>> Erik L. has been in touch with IBM about their 3D FFT routines,  
>> but apparently they pointed out that there is already another FFTW  
>> based library that is even better, so we'll be trying that. Apart  
>> from that, I do not know too much about Blue matter. The IBM  
>> website boasts about scaling, which looks great indeed, but it  
>> doesn't say anything about absolute performance (ns/day), which  
>> may be in the papers of course.
>>
>> We are currently in the process of writing up a paper on the  
>> absolute performance of the gromacs development version, and even  
>> if GROMACS doesn't reach the same absolute performance as Blue  
>> matter (which it may!) it gives a lot more bang for the buck in  
>> terms of hardware costs.
>>
>> But to further the discussion: is there anything else in Blue  
>> matter that would be interesting to have implemented into GROMACS?
>> --
>> David.
>> _____________________________________________________________________ 
>> ___
>> David van der Spoel, PhD, Assoc. Prof., Molecular Biophysics group,
>> Dept. of Cell and Molecular Biology, Uppsala University.
>> Husargatan 3, Box 596,          75124 Uppsala, Sweden
>> phone:  46 18 471 4205          fax: 46 18 511 755
>> spoel at xray.bmc.uu.se    spoel at gromacs.org   http://folding.bmc.uu.se
>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 
>> +++++
>> _______________________________________________
>> gmx-users mailing list    gmx-users at gromacs.org
>> http://www.gromacs.org/mailman/listinfo/gmx-users
>> Please search the archive at http://www.gromacs.org/search before  
>> posting!
>> Please don't post (un)subscribe requests to the list. Use the www  
>> interface or send it to gmx-users-request at gromacs.org.
>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
> ---------------------------------
> Eric Jakobsson, Ph.D.
> Professor, Department of Molecular and Integrative Physiology, and  
> of Biochemistry, and of the Center for Biophysics and Computational  
> Biology
> Senior Research Scientist, National Center for Supercomputing  
> Applications
> Professor, Beckman Institute for Advanced Science and Technology
> 3261 Beckman Institute, mc251
> University of Illinois, Urbana, IL 61801
> ph. 217-244-2896  Fax 217 244 9757
>
>
>
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before  
> posting!
> Please don't post (un)subscribe requests to the list. Use the www  
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php