Hi Galaxia
Try:
mpirun -mca btl ^openib -np 2 hello_world
This should turn off openib at runtime.
A better alternative is to download the OpenMPI tarball and
build it from source.
It will only build support to the hardware you have.
If you do so, use a non-standard installation directory.
This
I am working on a computer running CentOS 5 with 2 Quad-core CPUs (only
one computer, it is not connected with others). Previously the OpenMPI
version was 1.27 and my programs worked fine. After the automatic
upgrade to 1.32 (through yum), I can compile programs but it shows error
in running:
Chance your arm and include a CFLAGS with your configure including -D__GNUC__
CFLAGS=-D__GNUC__
A small test case just using those headers works, thus wise.
BTW, PGI 9.0-1 also fails on those headers.
DM
On Tue, 29 Dec 2009, Richard Walsh wrote:
All,
Not overwhelmed with responses here ...
Hi Richard,
I've built our OpenMPI with PGI 10.0 and included OpenIB support. I've verified
it works. You'll notice we build with UDAPL and OpenIB, but generally only
OpenIB is used.
Our complete configure flags, and OpenMPI is included with Scyld ClusterWare is
shown here:
./conf
Hi Richard
I haven't upgraded from PGI 8.0-4, so no comments on PGI 10.
As for IB, I configured OpenMPI only with the --with-openib flag,
but not --enable-openib-ibcm.
Have you tried that?
Gus Correa
-
Gustavo Correa
Lamont-Doh
All,
Not overwhelmed with responses here ... ;-) ... No one using PGI 10.0 yet?
We need it to make use of the GPU compiler directives they are supporting.
Can some perhaps comment on whether this is the correct way to configure
for an IB system? Everything works with Intel and/or if I compile w
Hi Ilya
OK, with 28 nodes and 4GB/node,
you have much more memory than I thought.
The maximum N is calculated based on the total memory
you have (assuming the nodes are homogeneous, have the same RAM),
not based on the memory per node.
I haven't tried OpenMPI 1.3.3.
Last I ran HPL was with OpenM
The executables must be available on all nodes - normally, this is done by
putting them in an NFS-mounted directory.
On Dec 29, 2009, at 6:35 AM, vishal shorghar wrote:
> HI All,
>
> Today I reran the trivial test on two nodes with (via the --scratch option)
> to a NFS share that is accessible
Hello, Gus!
Sorry for the lack of debug info.
I have 28 nodes. Each node have 2 processors Xeon 2.4 GHz with 4 Gb RAM.
OpenMPI 1.3.3 was compiled as:
CC=icc CFLAGS=" -O3" CXX=icpc CXXFLAGS=" -O3" F77=ifort FFLAGS=" -O3"
FC=ifort FCFLAGS=" -O3" ./configure --prefix=/opt/openmpi/intel/
--enable-debu
HI All,
Today I reran the trivial test on two nodes with (via the --scratch
option) to a NFS share that is accessible to all nodes in hostlist (as
suggested by Ethan).But still no luck.
I have shared
"/root/mtt-svn/samples/installs/nRpF/tests/trivial/test_get__trivial" on
my head node which
10 matches
Mail list logo
