Greetings Ralph,
Thank you so much. I look forward to seeing the new OpenMPI 1.5.x with Xgrid
support. There was a point say 10.6.2 and 10.5.8 where OpenMPI 1.4.1 and
v1.2.9 both worked, with only a minor glitch.
Since I have a 10.6.3 version, I can retry on that and see what results I get.
I'm afraid we were unable to support xgrid after the 1.2 series as no
developer had access to an xgrid server. I recently received a complimentary
copy of OSX-server from Apple, and I expect to restore xgrid support at some
point in the 1.5 series.
It looks like you are hitting some issue with 1.2
Greetings all,
I am running into some trouble using OpenMPI with OSX 10.6.4 in a Kerberized
XGrid environment. Note, I did not have this trouble before in the OSX 10.5.8
Kerberized XGrid environment.
The pattern of this trouble is as follows:
1. User submits a mpi job entering "mpirun -np 4
Afraid I can only reiterate: we don't support binding of individual threads
to cores at this time. You can use bind-to-socket to constrain all threads
from a process to a socket, so they can at least use those cores - but the
threads will move around between the cores in that socket, and more threa
I use the taskset command, or just use 'top' and watch the performance.
On Thu, Jul 29, 2010 at 12:02 PM, Ralph Castain wrote:
> I don't see anything in your code that would bind, but I also don't see
> anything that actually tells you whether or not your are bound. It appears
> that MPI_Get_pro
I don't see anything in your code that would bind, but I also don't see
anything that actually tells you whether or not your are bound. It appears that
MPI_Get_processor_name is simply returning the name of the node as opposed to
the name/id of any specific core. How do you know what core the th
No problem, anyways I think you are headed in the right direction now.
--td
David Akin wrote:
Sorry for the confusion. What I need is for all OpenMP threads to
*not* stay on one core. I *would* rather each OpenMP thread to run on
a separate core. Is it my example code? My gut reaction is no bec
Sorry for the confusion. What I need is for all OpenMP threads to *not* stay
on one core. I *would* rather each OpenMP thread to run on a separate core.
Is it my example code? My gut reaction is no because I can manipulate
(somewhat) the cores the threads are assigned by adding -bysocket
-bind-to-s
Ralph Castain wrote:
On Jul 29, 2010, at 5:09 AM, Terry Dontje wrote:
Ralph Castain wrote:
How are you running it when the threads are all on one core?
If you are specifying --bind-to-core, then of course all the threads will be on
one core since we bind the process (not the thread). If you
If you check, I expect you will find that your threads and processes are not
bound to a core, but are now constrained to stay within a socket.
This means that if you run more threads than cores in a socket, you will see
threads idled due to contention.
On Jul 29, 2010, at 8:29 AM, David Akin w
Adding -bysocket -bind-to-socket worked! Now to figure out why that
is? I also assumed it was my code. You can try my simple example code
below.
On Thu, Jul 29, 2010 at 8:49 AM, Ralph Castain wrote:
>
> On Jul 29, 2010, at 5:09 AM, Terry Dontje wrote:
>
> Ralph Castain wrote:
>
> How are you runn
On Jul 29, 2010, at 5:09 AM, Terry Dontje wrote:
> Ralph Castain wrote:
>>
>> How are you running it when the threads are all on one core?
>>
>> If you are specifying --bind-to-core, then of course all the threads will be
>> on one core since we bind the process (not the thread). If you are
>
Ralph Castain wrote:
How are you running it when the threads are all on one core?
If you are specifying --bind-to-core, then of course all the threads will be on
one core since we bind the process (not the thread). If you are specifying -mca
mpi_paffinity_alone 1, then the same behavior result
Below are all places that could contain mca related settings.
grep -i mca /usr/mpi/gcc/openmpi-1.4-qlc/etc/openmpi-mca-params.conf
# This is the default system-wide MCA parameters defaults file.
# Specifically, the MCA parameter "mca_param_files" defaults to a
# "$HOME/.openmpi/mca-params.conf:$sy
14 matches
Mail list logo
