...@open-mpi.org] On Behalf
Of Ralph Castain
Sent: Monday, November 12, 2012 6:07 PM
To: Open MPI Users
Subject: Re: [OMPI users] EXTERNAL: Re: gathering problem
Most common problem is connector not fully seated. Causes capacitive
connection, passes many tests
Try reseating connectors
Sent
Most common problem is connector not fully seated. Causes capacitive
connection, passes many tests
Try reseating connectors
Sent from my iPhone
On Nov 12, 2012, at 3:15 PM, "Hodge, Gary C" wrote:
> Today I tried a test where all sender and receiver processes were located on
> nodes on the s
Today I tried a test where all sender and receiver processes were located on
nodes on the same blade chassis.
With this process deployment, I did not detect any delayed send/receive
processing, all tests ran as expected!!!
In the failing test below, the sender (proc39) was on the second blade ch
I set processor affinity by running a script after all processes are running.
My processes are bound to their own core. Each process has 3 threads, one is
my app and the other 2 appear after MPI_Init is called.
Each thread in a process is bound to the same core and the processes are
running with
