On Oct 2, 2007, at 9:53 PM, Mohamad Chaarawi wrote:
> Yea i think those problems where when i was running from the already
> installed mpich2.. but doesn't mpich pick up slurm from the path
Yes, but you need to link against SLURM's libpmi specifically. In
the template, I have stuff like this:
[MPI install: MPICH2]
mpi_get = mpich2
save_stdout_on_success = 1
merge_stdout_stderr = 0
# Adjust this for your site (this is what works at Cisco). Needed to
# launch in SLURM; adding this to LD_LIBRARY_PATH here propagates this
# all the way through the test run phases that use this MPI install,
# where the test executables will need to have this set.
prepend_path = LD_LIBRARY_PATH /opt/slurm/current/lib
module = MPICH2
mpich2_compiler_name = gnu
mpich2_compiler_version = &get_gcc_version()
mpich2_configure_arguments = --disable-f90 CFLAGS=-O3 --enable-fast --
# These are needed to launch through SLURM; adjust as appropriate.
mpich2_additional_wrapper_ldflags = -L/opt/slurm/current/lib
mpich2_additional_wrapper_libs = -lpmi
Note these last two fields ^^. You'll need to replace the -L value
with whatever is relevant for your cluster.
> On Tue, October 2, 2007 1:09 pm, Jeff Squyres (jsquyres) wrote:
>> I'm away from a computer right now so I don't have the specifics,
>> but we
>> saw some testbake results from UH today of mpich2 under slurm that
>> not run properly - it ran 16 copies of skampi instead of 1 16-node
>> job, so
>> the output was very skewed (and completely mis-parsed).
>> Can you check your mpich2 compile / link settings to ensure that
>> linking against the slurm pmi library properly?
>> mtt-devel mailing list
> Mohamad Chaarawi
> Instructional Assistant http://www.cs.uh.edu/~mschaara
> Department of Computer Science University of Houston
> 4800 Calhoun, PGH Room 526 Houston, TX 77204, USA
> mtt-devel mailing list