On Apr 25, 2012, at 8:58 AM, seshendra seshu wrote:
> I have installed openMPI on the master node and separately on client nodes
> how can i check that all nodes can communicate for openmpi for running
> parallel programs.
Check out this prior post from this list:
http://www.open-mpi.org/
On Apr 24, 2012, at 3:33 PM, Tom Rosmond wrote:
> Yes, I would be interested in such a plugin. But be advised that I am
> strictly a fortran programmer, so if it requires any C/C++ talent, I
> would be in trouble. So maybe, before jumping into that, I would like
> to be able to look at what proc
That's very odd, indeed -- it's listed as being inside MPI_INIT, but we don't
get any further details from there. :-\
Any chance you could try upgrading to OMPI 1.4.5 and/or 1.5.5?
On Apr 24, 2012, at 1:57 PM, Jeffrey A Cummings wrote:
> I've been having an intermittent failure during MPI init
On Apr 24, 2012, at 3:01 PM, Tom Rosmond wrote:
> My question is this: If the cartesian mapping is done so the two
> spacial dimensions are the 'most rapidly varying' in equivalent 1-D
> processor mapping, will Open-mpi automatically assign those 2 dimensions
> 'on-node', and assign the 'ensemble
Could you repeat your tests with 1.4.5 and/or 1.5.5?
On Apr 23, 2012, at 1:32 PM, Martin Siegert wrote:
> Hi,
>
> I am debugging a program that hangs in MPI_Allreduce (openmpi-1.4.3).
> An strace of one of the processes shows:
>
> Process 10925 attached with 3 threads - interrupt to quit
> [pi
It looks like you are using LAM/MPI. This list is for supporting Open MPI, a
wholly different MPI software implementation. However, speaking as one of the
core LAM/MPI developers, I'll tell you that you should uninstall LAM and
install Open MPI install. We abandoned LAM/MPI several years ago.
On Apr 23, 2012, at 11:53 AM, Jeffrey Squyres wrote:
> Here's a description of MPI_MPROBE:
>
>http://blogs.cisco.com/performance/what-is-mpi_mprobe/
Oops -- this is the link I really meant to send:
http://blogs.cisco.com/performance/mpi-forum-roundup/
--
Jeff
Additionally, the MPI-3 MPI_MPROBE function will definitely help in this area.
We have it implemented in the Open MPI development trunk; it's slated for
release in the Open MPI 1.7 series. Here's a description of MPI_MPROBE:
http://blogs.cisco.com/performance/what-is-mpi_mprobe/
On Apr 2
On Apr 20, 2012, at 10:01 AM, seshendra seshu wrote:
> Am new to MPI i was trying to install OpenMPI on 1 master and 16 slaves i
> could not find the correct correct installation steps and configurations
> needed for before installing open MPI , so can any help me in installation
> procedure.
The new MPI-3 function MPI_Improbe might be helpful here, but it may still cost
you extra in temporary buffer space (allocated internally in MPI), depending on
the size of the incoming message.
Meaning: OMPI (and other MPI's) will still allocate *something* for each
matched incoming message, ev
What happens if you "dig quoVadis27"?
If you don't get a valid answer back, then it's not a resolvable name.
On Apr 19, 2012, at 6:42 AM, Bernhard Knapp wrote:
> Dear mail-list users,
>
> I have a problem when I try to run a parallel gromacs job on fedora core 15.
> The same job (same install
; > > Envoyé par : users-boun...@open-mpi.org
> > >
> > > Hi Jeffy,
> > >
> > > I checked the SEND RECV buffers and it looks ok to me. The code I
> > > have sent works only when I statically initialize the array.
> > >
> > > The code fails
gt; Hi Jeffy,
> >
> > I checked the SEND RECV buffers and it looks ok to me. The code I
> > have sent works only when I statically initialize the array.
> >
> > The code fails everytime I use malloc to initialize the array.
> >
> > Can you please lo
Ralph filed https://svn.open-mpi.org/trac/ompi/ticket/3075 about this.
I just reviewed Brice's patch and it looks good to me (it looks like I forgot
to update the logical/physical conversion in the module_set/get methods).
Thanks!
On Apr 12, 2012, at 1:46 AM, Brice Goglin wrote:
> Hello Tetsu
On Apr 18, 2012, at 3:15 AM, Peter Sels wrote:
> I suppose with C++ MPI it's possible to enforce more strict type
> checking using template or so.)
Not really, unfortunately. :-(
The idea is to be able to send *anything* through the (void*) choice buffer
designation -- that includes even sendi
> mpirun noticed that process rank 1 with PID 2936 on node machine4 exited on
> signal 11 (Segmentation fault).
>
> Can someone help please.
>
> Thanks
>
>
>
> On Tue, Apr 17, 2012 at 6:01 PM, Jeffrey Squyres wrote:
> Try malloc'ing
Moving the conversation to this bug:
https://svn.open-mpi.org/trac/ompi/ticket/3076
On Apr 16, 2012, at 4:57 AM, Seyyed Mohtadin Hashemi wrote:
> I recompiled everything from scratch with GCC 4.4.5 and 4.7 using OMPI 1.4.5
> tarball.
>
> I did some tests and it does not seem that i can mak
Sorry for the delay in replying; I was out last week.
MPI_SEND and MPI_RECV take pointers to the buffer to send and receive,
respectively.
When you send a scalar variable, like an int, you get the address of the buffer
via the & operator (e.g., MPI_Send(&i, ...)). When you send a new'ed/malloc
Try malloc'ing your array instead of creating it statically on the stack.
Something like:
int *data;
int main(..) {
{
data = malloc(ARRAYSIZE * sizeof(int));
if (NULL == data) {
perror("malloc");
exit(1);
}
// ...
}
On Apr 17, 2012, at 5:05 AM, Rohan Deshpande
On Apr 16, 2012, at 1:54 PM, Shiqing Fan wrote:
> They are definitions for enable dllexport/import declarations on Windows, and
> they existed since the initial version for Cygwin. Normally these definitions
> are hidden via mpicc command wrapper, but on Windows, when user tries to
> compile a
Shiqing --
What are these defines? Shouldn't they be invisible when compiling MPI
applications?
On Apr 9, 2012, at 4:13 PM, Shiqing Fan wrote:
> Hi Greg,
>
> Glad to hear that it works for you.
>
> And yes, these definitions are necessary for compiling any MPI application on
> Windows.
>
The output from that program looks fine to me on Linux:
[6:25] svbu-mpi:~/mpi % mpirun -np 4 hello
Slave:1
Slave:2
Slave says, Flag: 1 MyID: 2
Slave says, Flag2: 2 MyID: 2
Slave:3
Slave says, Flag: 1 MyID: 3
Slave says, Flag2: 2 MyID: 3
Master says, Flag: 1 M
On Apr 4, 2012, at 8:04 PM, Rohan Deshpande wrote:
> Yes they are on same subnet. ips for example - 192.168.1.1, 192.168.1.2,
> 192.168.1.3
This is generally considered a bad idea, not just for MPI, but for Linux in
general. Google around about this. One reason, for example, is that there i
ou please explain more.
>
> Thanks
>
> On Wed, Apr 4, 2012 at 10:47 PM, Jeffrey Squyres wrote:
> Do you have multiple IP interfaces on a single machine on the same IP subnet,
> perchance?
>
> If so, don't do that. :-) Put each of your IP addresses in a differen
Do you have multiple IP interfaces on a single machine on the same IP subnet,
perchance?
If so, don't do that. :-) Put each of your IP addresses in a different IP
subnet.
On Apr 3, 2012, at 10:30 PM, Rohan Deshpande wrote:
> Hi,
>
> I am running a MPI program using cluster and tcp communi
o. Sorry!
>
>
> --- On Wed, 4/4/12, Jeffrey Squyres wrote:
>
> From: Jeffrey Squyres
> Subject: Re: [OMPI users] Need Performance estimation of MPI_Bcast
> To: "Open MPI Users"
> Date: Wednesday, April 4, 2012, 4:35 PM
>
> It's a bit more complicat
It's a bit more complicated than that; there are many factors involved:
- how many peer MPI processes are involved in the broadcast
- how many of those peers are local (and will likely communicate via shared
memory, not a network device) and how many are remote
- what broadcast algorithm is used
I don't know exactly how Fedora packages Open MPI, but I've seen some
distributions separate Open MPI into a base package and a "devel" package. And
mpicc (and some friends) are split off into that "devel" package.
The rationale is that you don't need mpicc (and friends) to *run* Open MPI
appl
Dylan --
Sorry for the delay in replying.
On an offhand guess, does the problem go away if you run with:
--mca mpi_leave_pinned 0
?
On Mar 20, 2012, at 3:35 PM, Dylan Nelson wrote:
> Hello,
>
> I've been having trouble for awhile now running some OpenMPI+IB jobs on
> multiple tasks. The p
You might want to take an MPI tutorial or two; there's a few good ones
available on the net.
My favorites are the basic and intermediate level MPI tutorials at NCSA.
On Mar 25, 2012, at 1:13 PM, Rohan Deshpande wrote:
> Hi,
>
> I want to distribute the data on different machines using open mp
On Mar 21, 2012, at 11:34 AM, Brock Palen wrote:
> tcp with this code?
Does it matter enough for debugging runs?
> Can we disable the psm mtl and use the verbs emulation on qlogic? While the
> qlogic verbs isn't that great it is still much faster in my tests than tcp.
>
> Is there a particula
We unfortunately don't have much visibility into the PSM device (meaning: Open
MPI is a thin shim on top of the underlying libpsm, which handles all the MPI
point-to-point semantics itself). So we can't even ask you to run padb to look
at the message queues, because we don't have access to them
Can you build with g++ instead of icpc?
All the C++ MPI bindings are inlined anyway, so the performance difference
between the two might be negligible.
On Mar 20, 2012, at 11:35 AM, Gunter, David O wrote:
> I wish it were that easy. When I go that route, I get error messages like
> the follo
Did you run autogen.pl?
(if you're working with the Open MPI trunk for development reasons, you might
want to post to the de...@open-mpi.org list, not the general users list)
On Mar 20, 2012, at 8:31 AM, Ilias Miroslav wrote:
> Dear all,
>
> I updated ompi-trunk to the most recent trunk:
>
>
+1
See the description of cm vs. ob1 in the OMPI README. Here's the latest
description (I think we recently added a little more description here):
https://svn.open-mpi.org/trac/ompi/browser/trunk/README#L421
The PSM MTL does not have CUDA support; the smcuda BTL is for ob1 only.
On Mar 1
Many thanks for doing this Eugene.
On Mar 15, 2012, at 11:58 AM, Eugene Loh wrote:
> On 03/13/12 13:25, Jeffrey Squyres wrote:
>> On Mar 9, 2012, at 5:17 PM, Jeremiah Willcock wrote:
>>> On Open MPI 1.5.1, when I call MPI_Testsome with incount=0 and the two
>>> o
On Mar 14, 2012, at 1:06 PM, Patrik Jonsson wrote:
> I think I tracked it down, though. The problem was in the boost.mpi
> [snip]
Yuck! Glad you tracked it down.
> I do have a more general question, though: Is there a good way to back
> out the location of the request object if I stop deep in t
On Mar 14, 2012, at 9:38 AM, Patrik Jonsson wrote:
> I'm trying to track down a spurious segmentation fault that I'm
> getting with my MPI application. I tried using valgrind, and after
> suppressing the 25,000 errors in PMPI_Init_thread and associated
> Init/Finalize functions,
I haven't looked
sion
>> mpiexec (OpenRTE) 1.4.3
>>
>> Report bugs to http://www.open-mpi.org/community/help/
>> [tstitt@memtfe] /pscratch/tstitt > mpif90 -o mpitag mpitag.f90
>> [tstitt@memtfe] /pscratch/tstitt > mpiexec -n 2 ./mpitag
>>0
>>0
Tim --
I am unable to replicate this problem with a 1.4 build with icc.
Can you share your test code?
On Mar 10, 2012, at 7:30 PM, Timothy Stitt wrote:
> Hi all,
>
> I was experimenting with MPI_TAG_UB in my code recently and found that its
> value is set to 0 in my v1.4.3 and v1.4.5 builds
On Mar 9, 2012, at 5:17 PM, Jeremiah Willcock wrote:
> On Open MPI 1.5.1, when I call MPI_Testsome with incount=0 and the two output
> arrays NULL, I get an argument error (MPI_ERR_ARG). Is this the intended
> behavior? If incount=0, no requests can complete, so the output arrays can
> never
e time. But I hope we're getting darn close to done with 1.5.5.
> Rob
>
> On 3/9/12 1:16 PM, Jeffrey Squyres wrote:
>> On Mar 9, 2012, at 12:05 PM, Robert Collyer wrote:
>>
>>> Downloaded 1.5.5rc3, installed libtool 2.4.2, autoconf 2.68, and automake
&
On Mar 9, 2012, at 12:05 PM, Robert Collyer wrote:
> Downloaded 1.5.5rc3, installed libtool 2.4.2, autoconf 2.68, and automake
> 1.11.3, and added them to my PATH.
>
> Running autogen.sh and I get the following error
There's no need for you to run autogen.sh -- just run configure and make.
au
On Mar 6, 2012, at 10:17 AM, nadia.der...@bull.net wrote:
> Since MPI_Comm_create sets the created communicator to MPI_COMM_NULL for the
> processes that are not in the group , may be preceding your collectives by a:
> if (MPI_COMM_NULL != new_comm) {
>
> }
> could be enough.
>
> But ma
On Mar 6, 2012, at 4:43 PM, Timothy Stitt wrote:
> Can anyone tell me whether it is legal to pass zero values for some of the
> send count elements in an MPI_AlltoallV() call? I want to perform an
> all-to-all operation but for performance reasons do not want to send data to
> various ranks who
Sorry for the delay. Answers inline.
On Feb 29, 2012, at 1:09 PM, Eric Chamberland wrote:
> I would like to know which of "waitone" vs "waitany" is optimal and of
> course, will never produce deadlocks.
In your sample application, either will work just fine -- they will not
deadlock.
> Afte
Is the error reproducible with Open MPI 1.5.5rc3?
On Mar 9, 2012, at 10:05 AM, Robert Collyer wrote:
> My apologies for not including that, it was on linux, specifically Fedora
> Core 9.
>
> On 3/9/12 8:32 AM, Jeffrey Squyres wrote:
>> I am unable to replicate this proble
I am unable to replicate this problem on both OS X and Linux using the commands
you provided.
What OS are you using?
On Mar 8, 2012, at 10:27 PM, Robert Collyer wrote:
> The libtool that is constructed for the configuration I'm using (below) has
> bug when it is constructing compiler_lib_sear
'
>
> GNU version still builds fine.
>
>
> On Tue, Mar 6, 2012 at 5:38 AM, Jeffrey Squyres wrote:
>> I disabled C++ inline assembly for PGI (we already had C inline assembly for
>> PGI).
>>
>> So I don't think this should have caused a new e
c_swap_64'
>
> GNU version still builds fine.
>
>
> On Tue, Mar 6, 2012 at 5:38 AM, Jeffrey Squyres wrote:
>> I disabled C++ inline assembly for PGI (we already had C inline assembly for
>> PGI).
>>
>> So I don't think this should have caused a
t;>> unusual part, you're compiling 32 bits only (thus no atomic 64 bits are
>>>> available), and we are forcing atomic operations on a 64 bits value. That
>>>> would be quite strange …
>>>>
>>>> Are you trying to compile 32 or 64 bit
Can you use gfortran instead? I don't think we've tested with g95 in years (if
ever).
Yes, you can manually edit mpif-config.h, if you need to. I'm guessing that
OMPI's configure script got the wrong answer from g95, and therefore put in a 0
for those values. I don't know if we want to suppo
On Mar 2, 2012, at 2:50 PM, Ralph Castain wrote:
>> Ralph and I just had a phone conversation about this. We consider it a bug
>> -- you shouldn't need to put --prefix in the app file. Meaning: --prefix is
>> currently being ignored if you use an app file (and therefore you have to
>> put --p
On Mar 2, 2012, at 9:48 AM, Yiguang Yan wrote:
> (All with the same test script test.bash I post in my previous emails, so run
> with app file fed to mpirun command.)
>
> (1) If I put the --prefix in the app file, on each line of it, it works fine
> as Jeff said.
>
> (2) Since in the manual, i
onfigure, make clean and make all
>>
>>
>> On Thu, Mar 1, 2012 at 10:03 PM, Jeffrey Squyres wrote:
>>> Did you do a full autogen / configure / make clean / make all ?
>>>
>>>
>>> On Mar 1, 2012, at 8:53 AM, Abhinav Sarje wrote:
>>>
&g
n Mar 1, 2012, at 4:51 PM, Jeffrey Squyres wrote:
> I see the problem.
>
> It looks like the use of the app context file is triggering different
> behavior, and that behavior is erasing the use of --prefix. If I replace the
> app context file with a complete command line, it wo
I see the problem.
It looks like the use of the app context file is triggering different behavior,
and that behavior is erasing the use of --prefix. If I replace the app context
file with a complete command line, it works and the --prefix behavior is
observed.
Specifically:
$mpirunfile $mcap
Did you do a full autogen / configure / make clean / make all ?
On Mar 1, 2012, at 8:53 AM, Abhinav Sarje wrote:
> Thanks Ralph. That did help, but only till the next hurdle. Now the
> build fails at the following point with an 'undefined reference':
> ---
> Making all in tools/ompi_info
On Mar 1, 2012, at 1:17 AM, Jingcha Joba wrote:
> Aah...
> So when openMPI is compile with OFED, and run on a Infiniband/RoCE devices, I
> would use the mpi would simply direct to ofed to do point to point calls in
> the ofed way?
I'm not quite sure how to parse that. :-)
The openib BTL uses
gt; customized OFED ( in RDMA scenario's or anything (if u can specify)).
>
>
> On Thu, Mar 1, 2012 at 1:45 AM, Jeffrey Squyres wrote:
> On Feb 29, 2012, at 2:57 PM, Jingcha Joba wrote:
>
> > So if I understand correctly, if a message size is smaller than it will use
>
On Mar 1, 2012, at 3:33 AM, Pinero, Pedro_jose wrote:
> I am launching 200 light processes in two computers with 8 cores each one
> (Intel i7 processor). They are dedicated and are interconnected through a
> point-to-point Gigabit Ethernet link.
>
> I read about oversubscribing nodes in the op
On Feb 29, 2012, at 2:57 PM, Jingcha Joba wrote:
> So if I understand correctly, if a message size is smaller than it will use
> the MPI way (non-RDMA, 2 way communication), if its larger, then it would use
> the Open Fabrics, by using the ibverbs (and ofed stack) instead of using the
> MPI's s
est" method (whether it's a send/recv style exchange, an
RDMA-stle exchange, or a mixture of the two).
Make sense?
> On Wed, Feb 29, 2012 at 10:49 AM, Jeffrey Squyres wrote:
> FWIW, if Brian says that our one-sided stuff is a bit buggy, I believe him
> (because he wrote it).
Gah. I didn't realize that my 1.4.x build was a *developer* build.
*Developer* builds give a *lot* more detail with plm_base_verbose=100
(including the specific rsh command being used). You obviously didn't get that
output because you don't have a developer build. :-\
Just for reference, he
er than I'd like (indeed, I'm in the process of rewriting most of it
> as part of Open MPI's support for MPI-3's revised RDMA), so it's likely
> that the bugs are in Open MPI's onesided support. Can you try a more
> recent release (something from the 1.5 tre
On Feb 29, 2012, at 9:39 AM, Stefano Dal Pont wrote:
> I'm a newbie with openMPI so the problem it's probably me :)
> Im using a Fortran 90 code developed under Ubuntu 10.04. I've recently
> installed the same code on my Archlinux machine but I have some issues
> concerning openMPI.
> A simple
On Feb 29, 2012, at 11:01 AM, Pinero, Pedro_jose wrote:
> I am using OMPI v.1.5.5 to communicate 200 Processes in a 2-Computers cluster
> connected though Ethernet, obtaining a very poor performance.
Let me making sure I'm parsing this statement properly: are you launching 200
MPI processes on
FWIW, I'm immediately suspicious of *any* MPI application that uses the MPI
one-sided operations (i.e., MPI_PUT and MPI_GET). It looks like these two OSU
benchmarks are using those operations.
Is it known that these two benchmarks are correct?
On Feb 29, 2012, at 11:33 AM, Venkateswara Rao D
Just to put this up front: using the trunk is subject to have these kinds of
problems. It is the head of development, after all -- things sometimes break.
:-)
Ralph: FWIW, I can replicate this problem on my Mac (OS X Lion) with the SVN
trunk HEAD (svnversion tells me I have 26070M):
-
[6:
I haven't followed OFED development for a long time, so I don't know if there
is a buggy OFED in RHEL 5.4.
If you're doing development with the internals Open MPI (or if it'll be
necessary to dive into the internals for debugging a custom device/driver), you
might want to move this discussion t
On Feb 29, 2012, at 5:39 AM, adrian sabou wrote:
> I am experiencing a rather unpleasant issue with a simple OpenMPI app. I have
> 4 nodes communicating with a central node. Performance is good and the
> application behaves as it should. (i.e. performance steadily decreases as I
> increase the
On Feb 29, 2012, at 2:17 AM, Syed Ahsan Ali wrote:
> [pmdtest@pmd02 d00_dayfiles]$ echo ${MPIRUN} -np ${NPROC} -hostfile
> $i{ABSDIR}/hostlist -mca btl sm,openib,self --mca btl_openib_use_srq 1 ./hrm
> >> ${OUTFILE}_hrm 2>&1
> [pmdtest@pmd02 d00_dayfiles]$
Because you used >> and 2>&1, the out
FWIW: Ralph committed a change to mpirun the other day that will now check if
you're missing integer command line arguments. This will appear in Open MPI
v1.7. It'll look something like this:
% mpirun -np hostname
---
Open MPI has detected
The intent of the --prefix option (or using the full path name to mpirun) was
exactly for the purpose of not requiring changes to the .bashrc.
Can you run with "--mca plm_base_verbose 100" on your command line? This will
show us the exact rsh/ssh command line that is being executed -- it might
Yes, this is known behavior for our CLI parser. We could probably improve that
a bit...
On Feb 28, 2012, at 4:55 AM, Ralph Castain wrote:
>
> On Feb 28, 2012, at 2:52 AM, Reuti wrote:
>
>> Am 28.02.2012 um 10:21 schrieb Ralph Castain:
>>
>>> Afraid I have to agree with the prior reply - soun
Alex --
I'm sorry this mail thread fell through the cracks.
You might be using GNU Autotools that are actually too new for the 1.4 series.
The Autotools tightened up on some of their syntax and usage in more recent
versions that we didn't adapt to until well after 1.4 branched from the main
S
On Feb 20, 2012, at 5:06 PM, Rayson Ho wrote:
> Thanks, I just downloaded it again and it is not a corrupted file anymore!
>
> (But what's happened to the "what source control system do you guys
> use" question usually asked by Jeff? :-D )
I need to get back to asking that one. :-)
It's just
> http://gridscheduler.sourceforge.net/
>
> Scalable Grid Engine Support Program
> http://www.scalablelogic.com/
>
>
>
> On Mon, Feb 20, 2012 at 3:45 PM, Jeffrey Squyres wrote:
>> Little known secret: we edit before these things go to air. :-)
>>
>>
Little known secret: we edit before these things go to air. :-)
The recordings almost always take about an hour, but we snip some things out.
IIRC, we had some tech problems which wasted some time in this recording, and
some off-recording kibitzing. :-)
Also, it looks like Brock had a proble
Better to handle these things on the devel list -- I'll reply over there.
Thanks!
On Feb 18, 2012, at 3:11 PM, Aleksej Saushev wrote:
> Hello!
>
> WEXITSTATUS is defined in , see the patch attached.
>
> (Sorry, I couldn't find simple mail interface for bug reports.)
>
>
> --
> HE CE3OH...
80 matches
Mail list logo
