Gus -
Open MPI 1.3.0 & 1.3.1 attempted to use some controls in the glibc malloc
implementation to handle memory registration caching for InfiniBand.
Unfortunately, it was not only bugging in that it didn't work, but it also
has the side effect that certain memory usage patterns can cause the
I think Serge is talking about compiling the application against one
version of Open MPI, linking dynamically, then running against another
version of Open MPI. Since it's dynamically linked, the ORTE/OMPI
interactions are covered (the version of mpirun, libopen-rte, and libmpi
all match). Th
Hi -
Unfortunately, as the bug at Sun's web site points out, their compiler is
borked. There's not a lot we can do about that fact, without causing a
whole host of other problems.
If you aren't using C++, I'd recommend compiling Open MPI with GCC and
then reseting the wrapper compilers to i
On Mon, 23 Mar 2009, Olaf Lenz wrote:
and the solution that is described there still looks as though it should
still work now, or shouldn't it? Just link all the OpenMPI plugins against
the base OpenMPI libraries, and it should work. Or am I wrong?
I believe your suggestion will not work, cer
Ugh! If you dont' get to this by Friday and I'm able to get the XGrid bug
knocked out quickly, I'll take a look. I remember being worried about
that case when I fixed up the OOB connection code, but thought I convinced
myself it was right. Apparently not - I wonder if I got a loop wrong and
Ricardo -
That's really interesting. THis is on a Leopard system, right? I'm the
author/maintainer of the xgrid code. Unfortunately, I've been hiding
trying to finish my dissertation the last couple of months. I can't offer
much advice without digging into it in more detail than I have tim
Joe -
There are two different files being discussed, which might be the cause of
the confusion. And this is really complicated, undocumented code I'm
shamefully responsible for, so the confusion is quite understandable :).
There's asm.c, which on all non-Sparc v8 platforms just pre-processes
On Thu, 29 Jan 2009, Gabriele Fatigati wrote:
Dear OpenMPI Developer,
i have a doubt regards mpi_leave_pinned parameter. Suppose i have a simple for:
for( int i=0; i< 100; i++)
MPI_Reduce(a, b, ...)
My question is: if i set mpi_leave_pinned= 1, buffer memories are
pinned in the entire pro
On Thu, 29 Jan 2009, Paul Wardman wrote:
I'm using R on a Ubuntu 8.10 machine, and, in particular, quite a lot of
papply calls to analyse data. I'm currently using the LAM implementation,
as it's the only one I've got to work properly. However, while it works
fine on one PC, it fails with the er
On Thu, 29 Jan 2009, Wayne Gilmore wrote:
I'm trying to use the compiler_args field in the wrappers script to deal
with 32 bit compiles on our cluster.
I'm using portland group compilers and use the following for 32 bit
builds: -tp p7 (I actually tried to use -tp x32 but it does not compile
co
Nick -
I think you have an incorrect deffinition of "correctly" :). According to
the MPI standard, an MPI implementation is free to either layer language
bindings (and only allow profiling at the lowest layer) or not layer the
language bindings (and require profiling libraries intercept each
hanks a lot,
Mi
Inactive hide details for "Brian W. Barrett"
"Brian W. Barrett"
"Brian W. Barrett"
Sent by: users-boun...@open-mpi.org
08/25/2008 01:44 PM
Please respond to
Open MPI Users
[IMAGE]
To
On Mon, 25 Aug 2008, Mi Yan wrote:
Does OpenMPI always use SEND/RECV protocol between heterogeneous
processors with different endianness?
I tried btl_openib_flags to be 2 , 4 and 6 respectively to allowe RDMA,
but the bandwidth between the two heterogeneous nodes is slow, same as
the bandwidth
On Wed, 16 Jul 2008, Adam Jundt wrote:
I have been working on getting a nightly tarball of Open MPI to build on
a Cray XT4 system running CNL. I found the following post on the forum:
http://www.open-mpi.org/community/lists/users/2007/09/4059.php. I had to
modify the configure options a little (
Thanks for adding the section, but the information contained within is
wrong (as was the information in this e-mail thread). OS X essentially
*always* adds a -rpath to both executables depending on libraries and
libraries depending on other libraries. There is no need to set
LD_LIBRARY_PATH,
Terry -
It tells us that I'm not as smart as I thought :). If you're willing to
help track this down, I'd like to try some other things that will require
a more involved patch (it'll take me a day or two to get the patch right).
Let me know if you'd be wiling to look further (hopefully only a
The Objective C thing is not important. If you notice, it actually found
gcc as the Objective C compiler (these days, most installs of GCC
including the Objective C backend). The only code in Open MPI that uses
Objective C is for OS X, but the way Autoconf works, it's easier to check
for the
Well, as a quick hack, you can try adding --disable-dlopen to the
configure line. It will disable the building of individual components
(instead linking them into the main shared libraries). It means that you
have to be slightly more careful about which components you build, but in
practice u
On Wed, 9 Apr 2008, Bailey, Eric wrote:
/home/MPI/openmpi-1.2.6/configure
--prefix=/home/MPI/openmpi-1.2.6/openmpi-1.2.6-install-7448
--host=ppc74xx-linux --build=i686-pc-linux-gnu
--host should be formated in the output of config.guess, which would be
something along the lines of processor f
On Mon, 3 Mar 2008, Brock Palen wrote:
I was writing a simple code to demonstrate deadlock. I create an
array of some number of doubles, when sizeof(double)*SIZE is >
btl_sm_eager_limit (on a smp machine) I expect the code to deadlock
but not until the buffer size is greater than the eager limi
On Wed, 27 Feb 2008, David Gunter wrote:
We are trying to build OMPI-1.2.4 for a BProc/Ethernet-based cluster.
Here are the configure options:
./configure --prefix=${PREFIX} \
--libdir=${LIBDIR} \
--enable-shared \
--with-bproc \
--with-tm=/opt/PBS \
--with-io_romio_flags=--with-file-
I played with this to fix some things in ORTE at one point, and it's a
very dangerous slope -- you're essentially guaranteeing you have a
deadlock case. Now instead of running off the stack, you'll deadlock.
The issue is that we call opal_progress to wait for something to happen
deep in the bo
A second attempt, this time with less suck :).
Brian
On Fri, 22 Feb 2008, Brian W. Barrett wrote:
Oops.. forgot to test the C++ bindings, which of course broke. Let me
try again and I'll send a better patch.
brian
On Fri, 22 Feb 2008, Brian W. Barrett wrote:
On Thu, 21 Feb 2008,
Oops.. forgot to test the C++ bindings, which of course broke. Let me
try again and I'll send a better patch.
brian
On Fri, 22 Feb 2008, Brian W. Barrett wrote:
On Thu, 21 Feb 2008, Jeff Squyres wrote:
While I agree that having AC actually define them to a value is a Good
Thing (b
On Thu, 21 Feb 2008, Jeff Squyres wrote:
While I agree that having AC actually define them to a value is a Good
Thing (better than just defining it to be empty), I do see the pickle
that it has put us in. :-\ I don't see an obvious solution.
I do :). Try the attached patch. It sneaks in at
David -
It looks like the code to do this was never pushed into the v1.2 release
(although it is in the trunk). I have no idea what time frame you're
looking at, but if you need an updated ROMIO before 1.3 is available,
someone would need to bring over the changes and do a 1.2.6 release.
In
Now that this discussion has gone way off into the MPI standard woods :).
Was your test using Open MPI 1.2.4 or 1.2.5 (the one with the segfault)?
There was definitely a bug in 1.2.4 that could cause exactly the behavior
you are describing when using the shared memory BTL, due to a silly
delay
On Mon, 7 Jan 2008, Danny G Smith wrote:
We would like to standardize to the Intel compiler across the platforms
that we are using. Is it possible to compile?
Yes, this should work without problem. Note that if you're using Leopard,
you need to upgrade to the latest version of the Intel com
On Wed, 12 Dec 2007, Alex Pletzer wrote:
I'm on a AMD64 box (Linux quartic.txcorp.com 2.6.19-1.2288.fc5 #1 SMP
Sat Feb 10 14:59:35 EST 2007 x86_64 x86_64 x86_64 GNU/Linux) and
compiled openmpi-1.2.4 using the Lahey-Fujitsu compiler (lfc). The
compilation of openmpi went fine.
$ ../configure --
On Thu, 13 Dec 2007, Jeff Squyres wrote:
Specifically: it would probably require some significant hackery in
the OMPI build process to put in a #define that indicates whether OMPI
is being built statically or not. But the AM/LT process shields this
information from the build process by design (
On Thu, 15 Nov 2007, Tim Prins wrote:
I have seen situations where after installing Open MPI, the wrapper
compilers did not create any executables, and seemed to do nothing.
I was never able to figure out why the wrappers were broken, and
reinstalling Open MPI always seemed to make it work.
If
On Sun, 21 Oct 2007, Tony Sheh wrote:
I'm working in xcode and i'm trying to build an application that
links against the OMPI libraries. So far i've included the following
files in the build:
libmpi.dylib
libopen-pal.dylib
libopen-rte.dylib
and the errors i get are
Undefined symbols:
all
SINGLE is returned in provided, such a design works.
> Now, since the thread synchronization seems to be fine, I think multiple
> threads (with only one thread executing MPI calls) works even if
> "provided" returns MPI_THREAD_SINGLE.
>
> -Thanks and Regards,
> Sarang.
>
&g
t; this kind of a design or there would be issues with threads in OpenMPI.
>
> -Thanks and Regards,
> Sarang.
>
> Quoting "Brian W. Barrett" :
>
>> There are two ways you can find out if an Open MPI build has
>> MPI_THREAD_MULTIPLE support:
>>
>>
There are two ways you can find out if an Open MPI build has
MPI_THREAD_MULTIPLE support:
* On the command line: You can run "ompi_info | grep Thread". If you see
mpi: no, the build doesn't support MPI_THREAD_MULTIPLE. If you see mpi:
yes, the build does support MPI_THREAD_MULTIPLE.
* At run-ti
> Is there a command line compile or runtime option to disable tight
> integration with Sun Grid Engine? I reviewed the FAQ and saw there was
> a way not to build certain components at compile time. However, I'm
> not quite sure what it would be in the case of gridengine. Any help
> would be apprec
Hi -
This is pretty odd. I haven't gotten a chance to make the "official"
1.2 OS X package yet, so I havent run into this problem. I'll try to
replicate in the next day or two. In the mean time, just to make
sure, you are using the latest version of XCode, right?
Brian
On Apr 6, 2007
On Apr 6, 2007, at 10:42 AM, Bas van der Vlies wrote:
On Apr 6, 2007, at 6:18 PM, Jeff Squyres wrote:
On Apr 6, 2007, at 12:14 PM, Bas van der Vlies wrote:
Have you run into a situation where OMPI gets the wrong flags
because
it's not using pbs-config?
Yes, We install the torque header fil
On Mar 16, 2007, at 6:53 AM, Grismer, Matthew J Civ AFRL/VAAC wrote:
I attempted to build 1.2rc4 on an Intel Mac Pro running OS X 10.4.9
and using gcc and the Intel Fortran compiler. The configure
succeeded fine, but make failed with a “What architecture?” error.
I’ve attached the configu
On Nov 9, 2006, at 9:06 AM, Åke Sandgren wrote:
I sent this in October and got very little feedback.
There is a problem in ompi_config_pthreads.m4 and
ompi_check_pthread_pids.m4 at least on Ubuntu Dapper
with both gcc and pgi.
The attached patch against v1.1.2 fixes the problem.
What happens
On Feb 14, 2007, at 10:50 AM, Jeff Squyres wrote:
On Feb 14, 2007, at 12:43 PM, Adrian Knoth wrote:
Blah. We definitely need to work on our error messages.
I think we could use gai_strerror(3) for this. If we could agree to
get
rid of SUSv2 and rely on RFC 3493 ;)
It would not be too diff
On Feb 7, 2007, at 3:26 PM, Michael wrote:
Building openmpi-1.3a1r13525 on OS X 10.4.8 (PowerPC), using my
standard compile line
./configure F77=g95 FC=g95 LDFLAGS=-lSystemStubs --with-mpi-f90-
size=large --with-f90-max-array-dim=3 ; make all
and after installing I found that I couldn't compil
On Feb 8, 2007, at 6:33 PM, Troy Telford wrote:
The error is (upon job start), something to the extent of (transcribed
from phone):
mca_mpool_openib_register: cannot allocate memory
.
.
.
Error creating low priority CQ for MTHCA0: Cannot allocate memory.
What has to happen for t
For things like these, I usually use the "dot file" mca parameter
file in my home directory:
http://www.open-mpi.org/faq/?category=tuning#setting-mca-params
That way, I don't accidently forget to set the parameters on a given
run ;).
Brian
On Feb 8, 2007, at 6:15 PM, Mark Kosmowski wr
On Jan 25, 2007, at 3:51 PM, sdamjad wrote:
Recently Lam Mpi released a beta version 7.1.3b that fixes bug on
MAC OS X bug
on intel macs for 64
builds.
Does open mpi already has this bug fixed ??
The same fix was incorporated into Open MPI v1.1.3b5.
Brian
--
Brian Barrett
Open MPI Te
I think this is a reasonable thing to commit. However, keep in mind
that %p isn't totally portable. I think it should be good on all the
platforms GM/MX support, but probably not a great idea to use it in
the general codebase.
But still reasonable to make the code at this level understand
On Jan 17, 2007, at 10:56 AM, Tim Campbell wrote:
In the 1.2b3 build I notice that the opal* page links are not longer
included. Is this by design? Also, and more importantly, the actual
opalcc.1 man page which the links point to is not copied into the
man1 directory. I trace this to the addi
errno 13 on Linux is EACCESS. According to the man page, ftruncate()
only retrns errno 13 if the file is owned by another user. I can't
see exactly how this could occur, but you might want to look at /tmp/
and make sure everything in openmpi-sessions-eddie* is owned by user
eddie.
Bri
t* of vmem for buffering/mem mapping. We
schedule based on requested vmem, so this can be a problem. Do you
know
how vmem usage for buffering compares with OpenMPI?
Cheers,
Aaron
-Original Message-
From: users-boun...@open-mpi.org [mailto:users-bounces@open-
mpi.org] On
Behalf Of Brian
On Jan 17, 2007, at 2:39 AM, Gleb Natapov wrote:
Hi Robin,
On Wed, Jan 17, 2007 at 04:12:10AM -0500, Robin Humble wrote:
so this isn't really an OpenMPI questions (I don't think), but you
guys
will have hit the problem if anyone has...
basically I'm seeing wildly different bandwidths over
On Jan 16, 2007, at 4:29 PM, Brock Palen wrote:
What is the state of openMPI on a sgi altix? How does it compare to
mpt. I assume for all operations OMPI will use the sm btl thus all
others (other than self) could be disabled. Is there any other
tweaks users use? Or is OMPI even recommend o
On Jan 15, 2007, at 10:13 AM, Marcelo Maia Garcia wrote:
I am trying to setup SGE to run DLPOLY compiled with mpif90
(OpenMPI 1.2b2, pathscale Fortran compilers and gcc c/c++). In
general I am much more luckier running DLPOLY interactively then
using SGE. The error that I got is: Signal:7
Open MPI will not run on PA-RISC processors. There are some
functions in Open MPI that must be implemented in assembly.
Presently, these are only supported for:
* Alpha (32 or 64 bit, beta support)
* IA-32
* IA-64
* MIPS (n32 or 64 bit APIs only, beta support)
* Opteron / EM64T
Sorry to jump into the discussion late. The mx btl does not support
communication between processes on the same node by itself, so you
have to include the shared memory transport when using MX. This will
eventually be fixed, but likely not for the 1.2 release. So if you do:
mpirun --pr
> g++ -O3 -DNDEBUG -fno-inline -pthread -o .libs/ompi_info components.o
> ompi_info.o output.o param.o version.o -Wl,--export-dynamic
> ../../../ompi/.libs/libmpi.so
> //programs/openmpi-1.1.2/orte/.libs/liborte.so
> //programs/openmpi-1.1.2/opal/.libs/libopal.so -ldl -lnsl
> -lutil -lm -Wl,--rpath
On Dec 13, 2006, at 4:15 PM, Dennis McRitchie wrote:
When creating a static build of an MPI program, I get a number of
fatal
error messages, as listed below. They are all regarding conflicts
between duplicate definitions and different sizes of malloc, free,
realloc, etc. for the RHEL4 and open
On Dec 14, 2006, at 1:48 PM, Michael Galloway wrote:
On Thu, Dec 14, 2006 at 02:01:54PM -0500, Michael Galloway wrote:
good day all, i've been trying to build ompi with the 6.2-X
version of the pgi
compiler set (pgcc 6.2-4 64-bit target on x86-64 Linux). i've
tried both 1.1.2 and
the current
It would be difficult to do well without some MPI help, in my
opinion. You certainly could use the Linux processor affinity API
directly in the MPI application. But how would the process know
which core to bind to? It could wait until after MPI_INIT and call
MPI_COMM_RANK, but MPI implem
Tony -
It looks like you ran into a bug in Libtool. Unfortunately, in order
to better support Fortran 90 when building shared libraries, we use a
beta version of Libtool 2, which means we're living a bit more on the
edge than we'd like. I have a patch for this issue that I'll be
submitt
On Nov 21, 2006, at 5:49 PM, Adam Moody wrote:
Hello,
We have some clusters which consist of a large pool of 8-way nodes
connected via ethernet. On these particular machines, we'd like our
users to be able to run 8-way MPI jobs on node, but we *don't* want
them
to run MPI jobs across nodes v
Locking a page with mlock() is not all that is required for RDMA
using InfiniBand (or Myrinet, for that matter). You have to call
that device's registration function first. In Open MPI, that can be
done implicitly with the mpi_leave_pinned option, which will pin
memory as needed and then
The issue is that openpty() is failing on the machine. I think /dev/ptmx
and /dev/pts have to exist for pty support to work properly. We're having
the same issue on Coyote, but not on Pink and Flash. As far as I can
tell, those are the files / directories that are on the working systems
but not
Hi -
Can you please send the config.log file generated when you ran configure?
Generally, we ask that everyone send the information requested here:
http://www.open-mpi.org/community/help/
As it provides enough information to solve most of the problems seen
during the build process.
Thank
Is there a platform on which this breaks? It seems to have worked well
for years... I'll take a closer look early next week.
Brian
On Fri, 6 Oct 2006, ?ke Sandgren wrote:
On Fri, 2006-10-06 at 11:35 +0200, ?ke Sandgren wrote:
Hi!
Attached is a patch that fixes some errors in the configure
On Thu, 8 Jun 2006, Michael Kluskens wrote:
call MPI_WAITALL(3,sp_request,MPI_STATUSES_IGNORE,ier)
1
Error: Generic subroutine 'mpi_waitall' at (1) is not consistent with
a specific subroutine interface
Issue, 3rd arg
On Wed, 31 May 2006, Justin Bronder wrote:
On 5/30/06, Brian Barrett wrote:
On May 28, 2006, at 8:48 AM, Justin Bronder wrote:
> Brian Barrett wrote:
>> On May 27, 2006, at 10:01 AM, Justin Bronder wrote:
>>
>>
>>> I've attached the required logs. Essentially the problem seems to
>>> be tha
On Wed, 24 May 2006, Terry Reeves wrote:
Here is the out put for both g95 and gfortran
???
From the output you sent, you ran "./configure FC=g95". Configure did
not find a valid F77 compiler, and therefore skipped both the F77 and
F90 bindings.
Can you try:
./configure FC=g95 F77=g95
an
On Thu, 25 May 2006, Brock Palen wrote:
We are trying to track down a problem with our network, the question
has come up, if its posable that the mpi lib (OMPI) could be using
ether multicast or broadcast (or both)
In what cases could multicast traffic be seen if OMPI is our mpi lib.
At this
68 matches
Mail list logo
