[gmx-users] Mac OS X system

Taeho Kim taeho.kim at utoronto.ca
Wed Dec 18 17:50:40 CET 2002


Hi David and all,
Thank you for your reply.
I am wondering how to explain the results(with np -1) below and how to improve my system, not cluster. Did I miss something? Is it matter of the way I compiled fftw/gromacs?  Please comment about <<Q>> below.

I got to have chance to compare my results with the benchmark page on Lys/Cut. I downloaded and run several jobs.  I do understand that my test is not exactly same, but it's just for comparison and making faster system. The only difference may be CPU and the way of compiling fftw ,gromacs, and so on.  

Our             Dual Mac     1.0G    1453 ps /day  (from the log file 60.55 ps/Node hour x 24hour/day)

Gromacs   BM  Dual Mac 1.25G   2784 ps
                     single Mac 1.25G  1712 ps 
http://www.gromacs.org/benchmarks/single.php

For your information,
FFTW:
% chmod u+x config.guess config.sub
% perl -pi -e 's/-06-03/' configure
% ./configure --enable-float --enable-mpi    
    <<Q>> do I need to add "--host-powerpc"?  (I have "PPC Altivec wrt CPU/Innerloops" in the log file
% make
% sudo make install

Gromacs-3.1.5_pre1
% ./configure --enable-mpi  <<Q>> how about --disable-shared? this is default option.
% make
% sudo make install
% sudo ln -s /usr/libexec/gcc/darwin/2.95.2/cpp /lib/cpp  
    <<Q>> do I need to link to ~gcc/darwin/3.1/cpp  instead of 2.95.2/cpp ? 

Thanks in advance!

Taeho
-----원본 메시지-----
보낸 사람: David <spoel at xray.bmc.uu.se>
받는 사람: gmx-users at gromacs.org <gmx-users at gromacs.org>
날짜: 2002년 12월 16일 월요일 오후 12:54
제목: Re: [gmx-users] Mac OS X cluster system


>On Mon, 2002-12-16 at 17:34, Taeho Kim wrote:
>> Hi all,
>> I am just wondering whether any one has experienced on Mac OS X  system, especially MPI parallel. 
>> http://www.gromacs.org/benchmarks/single.php
>> Where can I find( or make) "PPC version of LAM-mpi" as mentioned in the above link? 
>> 
>> I got the results on 2-node Mac cluster(2 machines), but it is not impressive at all. 
>> (My systems: Gromacs-3.1.5_pre1, fftw-2.1.3, MPICH, two-Mac 1.0 G, OS X 10.2, )
>>  It took the same amount of time on two-node Mac-cluster as that on a single machine. The above link mentioned the Mac cluster could be "significantly slower" via TCP/IP.   Is there any ways to improve the speed of simulation on multiple nodes of Mac system? 
>> 
>Checkout www.lam-mpi.org
>
>if you use the magic options to improve tcp (see mailing list) it should
>scale decently for large systems.
>> Thanks,
>> Taeho
>> 
>-- 
>Groeten, David.
>________________________________________________________________________
>Dr. David van der Spoel, Biomedical center, Dept. of Biochemistry
>Husargatan 3, Box 576,  75123 Uppsala, Sweden
>phone: 46 18 471 4205 fax: 46 18 511 755
>spoel at xray.bmc.uu.se spoel at gromacs.org   http://xray.bmc.uu.se/~spoel
>++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>_______________________________________________
>gmx-users mailing list
>gmx-users at gromacs.org
>http://www.gromacs.org/mailman/listinfo/gmx-users
>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 mailing list