Hello,
i got an algorithm that generates trees, of different sizes, recursively. at
the moment i have the algorithm in its secuential version.
here we have 4 identical computers with Xeon 8-core in each node + 4gb ram.
they have HyperThreading so they count as 16-processors per node.
so i can lau
I'm investigating some very large performance variation and have reduced
the issue to a very simple MPI_Allreduce benchmark. The variability
does not occur for serial jobs, but it does occur within single nodes.
I'm not at all convinced that this is an Open MPI-specific issue (in
fact the same var
On Jun 2, 2010, at 1:31 PM, guillaume ranquet wrote:
> granquet@bordeplage-9 ~/openmpi-1.4.2 $ ~/openmpi-1.4.2-bin/bin/mpirun
> - --mca btl openib,sm,self --mca pml ^cm ~/bwlat/mpi_helloworld
> Hello world from process 0 of 1
> granquet@bordeplage-9 ~/openmpi-1.4.2 $
>
> I can tell it works :)
O
On Jun 2, 2010, at 1:51 PM, Jeff Squyres wrote:
>>> Ok, there is no segfault when it can't find IB.
>
> I'm not sure I follow this comment.
MX initialization is interfering on IB nodes (that do not have MX). I wanted to
make sure the opposite was not true (and it is not). :-)
Scott
On Jun 2, 2010, at 1:31 PM, guillaume ranquet wrote:
> > ~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,self --mca pml ^cm
> > ~/bwlat/mpi_helloworld
>
> the first command seems to be wrong, I had an error message:
> MCA framework parameters can only take a single negation operator
Correct.
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 06/02/2010 06:00 PM, Scott Atchley wrote:
> On Jun 2, 2010, at 11:52 AM, Scott Atchley wrote:
>
>> What if you explicitly disable MX?
>>
>> ~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,self --mca btl ^mx
>> ~/bwlat/mpi_helloworld
>
> And ca
Le Wednesday 02 June 2010 15:55:37, vous avez écrit :
> On Jun 2, 2010, at 9:50 AM, Yves Caniou wrote:
> > I copy the output of my last mail at the end of this one, to avoid
> > searching. Here is the line that I used to configure OMPI:
> >
> > $>./configure --prefix=/home/p10015/openmpi --with-thr
On Jun 2, 2010, at 11:52 AM, Scott Atchley wrote:
> What if you explicitly disable MX?
>
> ~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,self --mca btl ^mx
> ~/bwlat/mpi_helloworld
And can you try this as well?
~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,self --mca pml ^cm
~/bwlat
On Jun 2, 2010, at 11:14 AM, guillaume ranquet wrote:
>> What happens if you run:
>>
>> ~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,self
>> ~/bwlat/mpi_helloworld
>>
>> (i.e., MX support is still compiled in, but remove MX from the run-time)
>
> sadly, exactly the same thing :(
> it doe
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
I snipped some parts of the exchange and responding to 2 mails in this
one. (this may not be proper netiquette on this ML?)
On 06/02/2010 03:54 PM, Jeff Squyres wrote:
> What happens if you run:
>
> ~/openmpi-1.4.2-bin/bin/mpirun --mca btl openib,sm,
On Jun 2, 2010, at 9:54 AM, Jeff Squyres wrote:
>> this is the output I get on a node with ethernet and infiniband hardware.
>> note the Error regarding mx.
>>
>> $ ~/openmpi-1.4.2-bin/bin/mpirun ~/bwlat/mpi_helloworld
>> [bordeplage-9.bordeaux.grid5000.fr:32365] Error in mx_init (error No MX
>>
On 6/2/10 7:33 AM, Jeff Squyres wrote:
On Jun 1, 2010, at 5:27 PM, John R. Cary wrote:
After patching, I get:
...meaning that you have applied the patch that you sent yesterday?
Yes.
I applied the "fix the spaces after \ characters" patch to the trunk and
release branches.
On Jun 2, 2010, at 9:58 AM, Yves Caniou wrote:
> I copy the output of my last mail at the end of this one, to avoid searching.
> Here is the line that I used to configure OMPI:
>
> $>./configure --prefix=/home/p10015/openmpi --with-threads=posix
> --enable-mpi-threads --enable-progress-threads
I forgot the list...
-
Le Wednesday 02 June 2010 14:59:46, vous avez écrit :
> On Jun 2, 2010, at 8:03 AM, Ralph Castain wrote:
> > I built it with gcc 4.2.1, though - I know we have a problem with shared
> > memory hanging when built with gcc 4.4.x, so I wonder if the issue here
> > is y
On Jun 2, 2010, at 8:42 AM, guillaume ranquet wrote:
> yes, I have multiple clusters, some with infiniband, some with mx, some
> nodes with both Myrinet et Infiniband hardware and others with ethernet
> only.
>
> I reproduced it on a vanilla 1.4.1 and 1.4.2 with and without the
> - --with-mx swit
On Jun 1, 2010, at 5:27 PM, John R. Cary wrote:
> After patching, I get:
...meaning that you have applied the patch that you sent yesterday?
I applied the "fix the spaces after \ characters" patch to the trunk and
release branches.
But as for why you're getting errors, Ralph asked yesterday ho
On Jun 2, 2010, at 8:03 AM, Ralph Castain wrote:
> I built it with gcc 4.2.1, though - I know we have a problem with shared
> memory hanging when built with gcc 4.4.x, so I wonder if the issue here is
> your use of gcc 4.5?
It's not yet clear that we have a shared memory problem with gcc 4.4.x
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Hi,
yes, I have multiple clusters, some with infiniband, some with mx, some
nodes with both Myrinet et Infiniband hardware and others with ethernet
only.
I reproduced it on a vanilla 1.4.1 and 1.4.2 with and without the
- --with-mx switch.
this is t
Weird - it works fine for me:
sjc-vpn5-109:mpi rhc$ mpirun -n 3 ./abort
Hello, World, I am 1 of 3
--
MPI_ABORT was invoked on rank 1 in communicator MPI_COMM_WORLD
with errorcode 2.
NOTE: invoking MPI_ABORT causes Open MPI to
Dear All,
As already said on this mailing list, I found that a simple Hello_world
program does not necessarily end (the program just hangs after the
MPI_Finalize(), and I can printf the MPI_FINALIZED which confirm that the MPI
part of the code has finished, but the exit() or return() never end
20 matches
Mail list logo
