[mvapich-discuss] How to setup mpd with make.mvapich.gen2

[zhang yigang]

Mr. Perkins:

I certainly hope that when I use more CPUs (like 192), the wall time would reduce so that I save my time for the simulation. From 32, to 64, and to 96 CPUs, the wall time do reduce although not much from 64 to 96 CPUs. From 96 to 192 CPUs, the wall time increase! This is something I do not like because I want to use all the CPUs effectively. 

It will be great if you and your students can find a solution for that. 

yigang Zhang

----- Original Message ----- 
From: "Jonathan Perkins" <perkinjo at cse.ohio-state.edu>
To: "zhang yigang" <zhangyg at mail.iggcas.ac.cn>; <mvapich-discuss at cse.ohio-state.edu>
Sent: Friday, March 14, 2008 7:59 PM
Subject: Re: [mvapich-discuss] How to setup mpd with make.mvapich.gen2


> On Fri, Mar 14, 2008 at 11:03:37AM +0800, zhang yigang wrote:
> > Dear Dr. Perkins:
> > 
> > Yes, indeed, I can run VASP now without mpd. Thanks a lot for your concern.
> > 
> > Now I am trying to solve another problem. VASP uses the same wall clock time with 96 and 192 CPUs. The scaling with the number of CPUs is not satisfactory for the big system I am simulating. 
> 
> Yigang Zhang:
> I've queried the students in the group to give some pointers and to help with
> your question related to scaling.  Just for clarification, is your concern that
> you're not seeing any growth in the wall time when you move from 92 to 192 cpus?
> 
> Please refer to me as Jonathan or Mr. Perkins.  I have not earned any Doctoral
> Degrees at this point.
> 
> > 
> > With best regards!
> > yigang zhang
> > 
> > ----- Original Message ----- 
> > From: "Jonathan Perkins" <perkinjo at cse.ohio-state.edu>
> > To: "zhangyigang" <zhangyg at mail.iggcas.ac.cn>
> > Cc: <mvapich-discuss at cse.ohio-state.edu>
> > Sent: Thursday, March 13, 2008 10:12 PM
> > Subject: Re: [mvapich-discuss] How to setup mpd with make.mvapich.gen2
> > 
> > 
> > > On Wed, Mar 12, 2008 at 10:19:42AM +0000, zhangyigang wrote:
> > > > Dear All: 
> > > > I download mvapich-1.0 and try to set it up in my machines. Following the
> > > > user guide, I choose make.mvapich.gen2 for my machine and run the setup by
> > > > typing: 
> > > > ./make.mvapich.gen2 
> > > > Everything goes smoothly. I got mpif90, mpicc, etc. But I cannot get mpd
> > > > (which I am so customed to because I always used mpich2 and mvapich2). So I
> > > > modified make.mvapich.gen2 as follows: 
> > > > --with-device=ch4-gen2 --with-device=ch-4mpd 
> > > > The compiling gives an error message. 
> > > > Could you kindly inform me how to set up openfabrics/gen2 and mpd together?
> > > > BTW, how to run a program with mvapich? I used to use mvapich2 with mpd.
> > > > The user guide says just setup ssh and then mpirun_rsh -np 1 -hostfile
> > > > cluster ./a.out. I tried, but the program did not run. Do we need first to
> > > > run a daemon or something else before mpirun_rsh (like in mvapich2, we have
> > > > to run mpdboot first before mpiexec)? 
> > > > Thanks greatly for your answer. 
> > > > yigang Zhang
> > > > _______________________________________________
> > > > mvapich-discuss mailing list
> > > > mvapich-discuss at cse.ohio-state.edu
> > > > http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
> > > 
> > > Yingang Zhang:
> > > According to a post earlier today it looks like you have overcome your issues
> > > using MVAPICH 1.0.  As you may now already know, mpd is no longer supported
> > > starting with this release.
> > > 
> > > For further information on using mpirun_rsh you can take a look our userguide
> > > located at http://mvapich.cse.ohio-state.edu/support/mvapich_user_guide.html.
> > > 
> > > Below follows a direct link to the most relevant section:
> > > http://mvapich.cse.ohio-state.edu/support/mvapich_user_guide.html#x1-210005.2
> > > 
> > > Please let us know if you experience any further difficulties.
> > > 
> > > -- 
> > > Jonathan Perkins
> > > http://www.cse.ohio-state.edu/~perkinjo
> 
> -- 
> Jonathan Perkins
> http://www.cse.ohio-state.edu/~perkinjo