Sorry for the delayed response.....
I got and built the tarball 1.8.3-272-g4e4f997 below.
A single node job runs ok, with correct cores etc
A multi-node job dies with the following error (no core dumps now):
>>>A specified physical processor does not exist in this topology:
>>> CPU number: 0
>>> Cpu set given: None
my mpirun line looks like:
/apps/share/openmpi/1.8.3-272-g4e4f997/bin/mpirun --prefix
/apps/share/openmpi/1.8.3-272-g4e4f997 --mca btl openib,tcp,sm,self --x
LD_LIBRARY_PATH --np 64 myexe -i br.i -l tommy1.o
My compile options for openmpi are:
version=1.8.3-272-g4e4f997
./configure \
--disable-vt \
--prefix=/apps/share/openmpi/$version \
--disable-shared \
--enable-static \
--with-openib \
--enable-mpirun-prefix-by-default \
--with-memory-manager=none \
--with-hwloc \
--with-lsf=/apps/share/LSF/9.1.3/9.1 \
--with-lsf-libdir=/apps/share/LSF/9.1.3/9.1/linux2.6-glibc2.3-x86_64/lib \
--with-wrapper-cflags="-shared-intel" \
--with-wrapper-cxxflags="-shared-intel" \
--with-wrapper-ldflags="-shared-intel" \
--with-wrapper-fcflags="-shared-intel" \
--enable-mpi-ext
Can you see anything that should or shouldn't be there?
Thanks
________________________________
From: devel <[email protected]> on behalf of Ralph Castain
<[email protected]>
Sent: Monday, December 15, 2014 10:07 PM
To: Open MPI Developers
Subject: Re: [OMPI devel] 1.8.4rc Status
My correction - the fix is in the nightly tarball from tonight. You can get it
here:
openmpi-v1.8.3-272-g4e4f997.tar.bz2<http://www.open-mpi.org/nightly/v1.8/openmpi-v1.8.3-272-g4e4f997.tar.bz2>
On Mon, Dec 15, 2014 at 2:40 PM, Ralph Castain
<[email protected]<mailto:[email protected]>> wrote:
Hey Tom
Note that rc2 had a bug in the out-of-band messaging system - might be what you
are hitting. I'd suggest working with rc4.
On Mon, Dec 15, 2014 at 12:57 PM, Tom Wurgler
<[email protected]<mailto:[email protected]>> wrote:
I have to take it back. While the first job was less than a node's worth of
cores and ran properly on the cores I wanted. more testing is revealing other
problems.
Anything that spans more than one node crashes and burns, with a core dump, and
nothing in the files to indicate why.
Note this is still rc2....
More testing on-going....
________________________________
From: devel <[email protected]<mailto:[email protected]>> on
behalf of Tom Wurgler <[email protected]<mailto:[email protected]>>
Sent: Monday, December 15, 2014 1:23 PM
To: Open MPI Developers
Subject: Re: [OMPI devel] 1.8.4rc Status
It seems to be working in rc2 after all.
I was still trying to use a rankfile, but it appears that is no longer needed.
Thanks!
________________________________
From: devel <[email protected]<mailto:[email protected]>> on
behalf of Ralph Castain <[email protected]<mailto:[email protected]>>
Sent: Monday, December 15, 2014 8:45 AM
To: Open MPI Developers
Subject: Re: [OMPI devel] 1.8.4rc Status
Should be there in rc4, and I thought it made it to rc2 for that matter. I'll
take a gander.
FWIW: I'm working off-list with IBM to tighten the LSF integration so we
correctly read and follow their binding directives. This will also be in 1.8.4
as we are in final test with it now.
Ralph
On Mon, Dec 15, 2014 at 5:40 AM, Tom Wurgler
<[email protected]<mailto:[email protected]>> wrote:
Forgive me if I've missed it, but I believe using physical OR logical core
numbering was going to be
reimplemented in the 1.8.4 series.
I've checked out rc2 and as far as I can tell, it isn't there as yet. Is this
correct?
thanks!
________________________________
From: devel <[email protected]<mailto:[email protected]>> on
behalf of Ralph Castain <[email protected]<mailto:[email protected]>>
Sent: Monday, December 15, 2014 8:35 AM
To: Open MPI Developers
Subject: [OMPI devel] 1.8.4rc Status
Hi folks
Trying to summarize the current situation on releasing 1.8.4. Remaining
identified issues:
1. TCP/BTL hang under mpi-thread-multiple. Asked George to look into it.
2. hwloc updates required. Brice committed them to the hwloc 1.7 repo. Gilles
volunteered to create the PR from there.
3. Fortran f08 binding disable for compilers not meeting certain conditions. PR
from Gilles awaiting review by Jeff
4. Topo signature issue reported by IBM. Ralph is waiting for more debug.
5. MPI/IO issue reported by Eric Chamberland. Gilles investigating.
6. make check issue on SPARC. Problem and fix reported by Paul Hargrove, Ralph
will commit
7. Linkage issue on Solaris-11 reported by Paul Hargrove. Missing the
multi-threaded C libraries, apparently need "-mt=yes" in both compile and link.
Need someone to investigate.
Please let me know if I've missed anything.
Ralph
_______________________________________________
devel mailing list
[email protected]<mailto:[email protected]>
Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel
Link to this post:
http://www.open-mpi.org/community/lists/devel/2014/12/16595.php
_______________________________________________
devel mailing list
[email protected]<mailto:[email protected]>
Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel
Link to this post:
http://www.open-mpi.org/community/lists/devel/2014/12/16604.php
