[mvapich-discuss] OpenFabrics

[Kevin Ball]

On Wed, 2006-08-30 at 18:30, Sayantan Sur wrote:
> Hi Kevin,
> 
> >
> >Itanium2 processors, MT25204 Mellanox cards with fw_ver 1.1.0.  And as
> >mentioned before, OFED 1.1-rc2 with MVAPICH 0.9.7, on a 2.6.12 kernel.
> >  
> >
> Could you let us know the following:
> 
> 1) Do you see the same behavior with MVAPICH-0.9.8?

With MVAPICH-0.9.8, I no longer see the behavior with the vanilla osu
benchmarks (osu_bw and osu_bibw).  However, I still see the problem with
a somewhat modified version I have (We submitted a version of this to
Dr. Panda, or you can find it at
http://www.pathscale.com/performance/InfiniPath/mpi_multibw/mpi_multibw.html


> 2) Can you run MVAPICH-0.9.8 over TCP/IP over IPoIB successfully on 
> those machines?

  To this point I have not succeeded in doing this.  I set up a hosts
file with the ib0 (IPoIB) net addresses in it, but what appears to
happen is mpirun ssh's over via IPoIB to start the jobs, but the jobs
then communicated via the ethernet link between them.

  Have you played with this previously and have any advice on how to get
it to work?  I'll keep tinkering, but if you have any thoughts they
would be appreciated.

> 
> If the problem persists, then we will want to take a closer look at it. 
> Since we don't have access to IA64 machines with MT25204 cards, will it 
> be possible for you to provide access to the machines?

  Unfortunately at this time the machines are in a location where we
cannot give anyone else access.  I will ask about this, but
unfortunately I think it is unlikely to be an option any time soon.

Thanks!

-Kevin


> 
> Thanks,
> Sayantan.