[openfabrics-ewg] OFED-1.0-rcX uDAPL testing with Intel MPI

[Aviram Gutman]

On which OSes  did you test it?


Regards,
   Aviram

-----Original Message-----
From: openfabrics-ewg-bounces at openib.org
[mailto:openfabrics-ewg-bounces at openib.org] On Behalf Of Bob Woodruff
Sent: Tuesday, May 30, 2006 7:19 PM
To: 'Matt Leininger'
Cc: openfabrics-ewg at openib.org
Subject: RE: [openfabrics-ewg] OFED-1.0-rcX uDAPL testing with Intel MPI

Matt wrote,
>On Thu, 2006-05-25 at 09:51 -0700, Woodruff, Robert J wrote:
>> >Has Intel been testing your MPI with the uDAPL available in the 
>> >OFED-1.0-rcX releases?
>> 
>> I tested the code base that was in RC4 and there were problems.
>> I submitted a patch to fix the issue and I think it was suppose to be

>> included in RC5. Is RC5 released yet ?

>  Yes, it was released this morning.

>> if so I can retry it with Intel MPI.

>  that would be great.  

>  - Matt

I tried OFED RC5 with Intel MPI on x86_64 and ia64 and everything seems
to be working fine on the Mellanox card. I also put together some ofed
rc5 backport patches and RPMs for 2.6.9-34EL and am hoping to get some
folks here to load it up on a larger cluster, my initial testing was
only on small clusters, but it looks stable. 

The only problem I saw is that there is still an issue with running
Intel MPI/uDAPL over the Pathscale (Qlogic) card. It appears to be a
uDAPL or MPI problem rather than a problem with the Pathscale driver. I
am now able to run ibv_rc_pingpong over the pathscale PCI-E card
reliably on a 2 node cluster. We plan to investigate the MPI/uDAPL issue
today. 

woody
_______________________________________________
openfabrics-ewg mailing list
openfabrics-ewg at openib.org
http://openib.org/mailman/listinfo/openfabrics-ewg