Install openmpi-2.1.0 on Centos 7

If you are here looking for new dreams, this is the place. It’s just that I like to write down here also what I do. So if you are not interested on IT stuff, just skip this post: I will come back to my fantasies in a few hours.

I was saying I need from time to time to install openmpi from scratch. It’s again time to do so. You see, this is the advantage of a blog: you know when was the last time you did something. So I download the newest openmpi version from here to  try to get rid of a CUDA misbehavior our SLURM cluster is having. Yes I know. Whatdidyojustsaiiiid? Sometimes I wonder what I’m speaking about also.

Let’s cut to the chase. The package, once downloaded and moved to its final location, needs to be configured. I try first time like this:

./configure --prefix=/usr/lib64/openmpi-2.1.0

And after a few minutes, I get this nice message, that is a new feature of this version:

Open MPI configuration:
-----------------------
Version: 2.1.0
Build MPI C bindings: yes
Build MPI C++ bindings (deprecated): no
Build MPI Fortran bindings: mpif.h, use mpi
MPI Build Java bindings (experimental): no
Build Open SHMEM support: yes
Debug build: no
Platform file: (none)

Miscellaneous
-----------------------
CUDA support: no
 
Transports
-----------------------
Cray uGNI (Gemini/Aries): no
Intel Omnipath (PSM2): no
Intel SCIF: no
Intel TrueScale (PSM): no
Mellanox MXM: no
Open UCX: no
OpenFabrics Libfabric: no
OpenFabrics Verbs: no
Portals4: no
Shared memory/copy in+copy out: yes
Shared memory/Linux CMA: yes
Shared memory/Linux KNEM: no
Shared memory/XPMEM: no
TCP: yes
 
Resource Managers
-----------------------
Cray Alps: no
Grid Engine: no
LSF: no
Slurm: yes
ssh/rsh: yes
Torque: no

I’m really happy to read this report. It confirms me that, previously, I didn’t compile it with the right options. So I reconfigure it with CUDA support like this:

 ./configure --prefix=/usr/lib64/openmpi-2.1.0 
--with-cuda=/usr/local/cuda-8.0/

And check that the corresponding output is now telling me I have CUDA:

Miscellaneous
-----------------------
CUDA support: yes

Then I make all install, and test it. All looks fine. Anyway, I still have one concern. Do I need to compile one openmpi per CUDA version in our systems? I hope not, but we’ll see!  I accept your thoughts about it 😉

Advertisements

About bitsanddragons

A traveller, an IT professional and a casual writer
This entry was posted in bits, centos, linux, slurm. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s