[ofa-general] uDAPL question

[Yong Qin]

For the 64-bit 20070404 build and rc1, it did work. That turned out to
be my own mistake, I didn't realize it needs more memlock than before. I
just ran my testing script and it failed. After increasing the memlock,
it has the same symptom as beta1 on 32-bit applications. I'll attach
more info for bug 521.

Thanks,

Yong
 

-----Original Message-----
From: Arlin Davis [mailto:arlin.r.davis at intel.com] 
Sent: Wednesday, April 04, 2007 7:26 PM
To: Woodruff, Robert J; Yong Qin; Hefty, Sean
Cc: general at lists.openfabrics.org
Subject: RE: [ofa-general] uDAPL question


>
>With the night build 20070404, both 32-bit and 64-bit hanged on
>RDMA_init. All the testing were done with Intel MPI 3.0.


The 20070404 build (64-bit) works fine on our Xeon clusters with RHEL 5,
MT25208 adapters (4.7.4)
and Intel MPI 3.0.

Can you send more log information so I can see if it is in the progress
engine waiting for
connections or data?

You could also try running with "-env I_MPI_USE_DYNAMIC_CONNECTIONS 0"
so that static connections
are used and "-env I_MPI_RDMA_USE_EVD_FALLBACK 1 " so that data
transfers are validated via
completions instead of polling memory. This may help us isolate your
problems.

Also, what adapters are you using?

Thanks,

-arlin