On Sep 22, 2010, at 1:53 PM, Gabriele Fatigati wrote:
> Thanks Jeff,
>
> and.. what about RDMA? It works only with point-to-point or also with
> collectives?
When collectives use the point-to-point sends (e.g., they effectively invoke
MPI_SEND as part of MPI_BCAST, for example), that will do a
Thanks Jeff,
and.. what about RDMA? It works only with point-to-point or also with
collectives?
2010/9/22 Jeff Squyres
> On Sep 22, 2010, at 3:46 AM, Gabriele Fatigati wrote:
>
> > i'm tuning collectives of OpenMPI 1.4.2 with OTPO. I have a little
> question about BTL. This layer is involves ju
Yes, the built-in GCC atomics might work. I don't know if anyone has tried
them; they would be most useful because they would allow us to use multiple
different platforms.
Patches would definitely be appreciated here.
On Sep 22, 2010, at 12:25 PM, Dave Love wrote:
> Jeff Squyres writes:
>
Jeff Squyres writes:
> I believe that the first step would be to get some assembly for the
> ARM platform for some of OMPI's key routines (locks, atomics, etc.).
> Beyond that, it *might* "just work"...?
Is http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=579505
relevant/useful?
Printouts of less than 100 bytes would be unusual...but possible
On Wed, Sep 22, 2010 at 8:15 AM, Jeff Squyres wrote:
> Are you running on machines with OpenFabrics devices (that Open MPI is
> using)?
>
> Is ompi-ps printing 100 bytes or more?
>
> What does ps show when your program is hung?
>
>
Are you running on machines with OpenFabrics devices (that Open MPI is using)?
Is ompi-ps printing 100 bytes or more?
What does ps show when your program is hung?
On Sep 17, 2010, at 3:13 PM, Matheus Bersot Siqueira Barros wrote:
> Open MPI Version = 1.4.2
> OS = Ubuntu 10.04 LTS and CentOS 5
On Sep 22, 2010, at 9:50 AM, Richard Walsh wrote:
> The implication of your reply
> is that if the symbols/libraries are all there then things should work.
*IF* SLES provides binary compatibility guarantees between 10.2 and 11.
If not, then it *may* work, and/or it may fail in mysterious ways
Jeff Squyres wrote:
>Probably your best bet would be:
>
>- investigate if there's a missing symbol or library in the current
>mca_btl_openib.so (e.g., run nm on mca_btl_openib.so and ensure that all those
>libraries are >present in SLES 11)
>- if it's a missing library, see if you can suppl
On Sep 20, 2010, at 1:20 PM, Richard Walsh wrote:
> I was not expecting things to work, and find that codes compiled using
> OpenMPI 1.4.1 commands under SLES 10.2 produce the following message
> when run under SLES11:
>
> mca: base: component_find: unable to open
> /share/apps/openmpi-intel/1.4
On Sep 9, 2010, at 4:31 PM, Ashley Pittman wrote:
>> What is the exact semantics of an asynchronous barrier,
>
> I'm not sure of the exact semantics but once you've got your head around the
> concept it's fairly simple to understand how to use it, you call
> MPI_IBarrier() and it gives you a ha
On Wed, 2010-09-22 at 14:16 +0200, Ake Sandgren wrote:
> On Wed, 2010-09-22 at 07:42 -0400, Jeff Squyres wrote:
> > This is a problem with the Pathscale compiler and old versions of GCC. See:
> >
> >
> > http://www.open-mpi.org/faq/?category=building#pathscale-broken-with-mpi-c%2B%2B-api
> >
On Sep 22, 2010, at 8:04 AM, Jens Domke wrote:
> I already tried this parameter, but I don't see any improvements in the
> benchmarks. Additionally while doing further investigations into the opensm I
> didn't see the QP requests for other LIDs than the base LIDs.
I'm afraid that I'm not an Inf
On Wed, 2010-09-22 at 07:42 -0400, Jeff Squyres wrote:
> This is a problem with the Pathscale compiler and old versions of GCC. See:
>
>
> http://www.open-mpi.org/faq/?category=building#pathscale-broken-with-mpi-c%2B%2B-api
>
> I note that you said you're already using GCC 4.x, but it's not
I already tried this parameter, but I don't see any improvements in the
benchmarks. Additionally while doing further investigations into the
opensm I didn't see the QP requests for other LIDs than the base LIDs.
Regards
Jens
Jeff Squyres wrote:
Yes, check out the btl_openib_max_lmc MCA param
On Sep 20, 2010, at 2:14 PM, Ken Mighell wrote:
> Has there been any consideration of porting OpenMPI to the ARM processor?
I don't believe that anyone is actively working on this, but I could be wrong.
> Plans are afoot to launch 7 ARM processors on a "Stage Coach" card in a 3U
> CubeSat. NASA
This is a problem with the Pathscale compiler and old versions of GCC. See:
http://www.open-mpi.org/faq/?category=building#pathscale-broken-with-mpi-c%2B%2B-api
I note that you said you're already using GCC 4.x, but it's not clear from your
text whether pathscale is using that compiler or
Yes, check out the btl_openib_max_lmc MCA parameter:
shell$ ompi_info --param btl openib --parsable | grep lmc
On Sep 21, 2010, at 11:45 AM, Jens Domke wrote:
> Hello,
>
> the InfiniBand architecture has a LMC feature to assign mutiple virtual LIDs
> to one port and so provides multiple paths
On Sep 22, 2010, at 3:46 AM, Gabriele Fatigati wrote:
> i'm tuning collectives of OpenMPI 1.4.2 with OTPO. I have a little question
> about BTL. This layer is involves just in point-to-point communication or
> also in collectives routines?
>
> Because i've noted that changing some blt parameter
Hi Nysal,
Thanks for your suggestions.
I'm now able to get the checksum computed and redirected to stdout, thanks (I
forgot the "-mca pml_base_verbose 5" option, you were right).
I haven't been able to observe the segmentation fault (with hdr->tag=0) so far
(when using pml csum) but I 'll let
Dear all,
i'm tuning collectives of OpenMPI 1.4.2 with OTPO. I have a little question
about BTL. This layer is involves just in point-to-point communication or
also in collectives routines?
Because i've noted that changing some blt parameters like btl_sm_eager_limit
and doing one collective routi
20 matches
Mail list logo
