[mpich-discuss] Installing MPICH2-1.0.5p3 in a SMP cluster

[Suman Vajjala]

Hi,

 Thanks for the suggestion. I was wondering if there could be an alternative
explanation since the same code shows different cpu usages configured with
different compilers.

-Suman

Message: 3
> Date: Thu, 16 Jul 2009 09:23:45 -0500
> From: Pavan Balaji <balaji at mcs.anl.gov>
> Subject: Re: [mpich-discuss] Installing MPICH2-1.0.5p3 in a SMP
>        cluster
> To: mpich-discuss at mcs.anl.gov
> Message-ID: <4A5F37F1.5080409 at mcs.anl.gov>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>
> 1.0.5p3 is an ancient version of MPICH2. We are at mpich2-1.1 right now,
> and will release 1.1.1 in a day or two.
>
> With respect to the CPU usage, it's too difficult to diagnose this
> problem with just information on the CPU usage. Can you try profiling
> your application using MPE (and view the profile with Jumpshot). It'll
> show you if there are any unnecessary stalls in MPI (e.g., waiting for
> data to arrive).
>
>  -- Pavan
>
> On 07/16/2009 07:37 AM, Suman Vajjala wrote:
> > Hi,
> >
> >  I have a strange problem. In one case, I installed mpich2 using gcc and
> > gfortran compilers and I get a CPU usage of 100% when I run parallel
> > codes. In the second case, I installed mpich2 with Intel icc and ifort
> > compilers (64bit, ver-11.1) and I get a cpu usage of 85%. Could it be
> > because of the configuration options used for the mpich2 installation?
> > Do the codes always have to show a 100% cpu usage?
> >
> > Thanks and regards
> > Suman Vajjala
>
> --
> Pavan Balaji
> http://www.mcs.anl.gov/~balaji <http://www.mcs.anl.gov/%7Ebalaji>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20090716/77873e36/attachment.htm>