I took some time to think about it and don't know that it was the case
for me. If the openmpi-bin package was missing, I wouldn't have been
able to even use openmpi and during my trouble, I would check the mpi
software each time I installed it to verify that it worked (using some
simple example commands). Furthermore, I checked the parent folder after
each installation and saw non-empty bin folders in them each time. Did
the error code somehow point to that? It could have been that the
open-mpi bin package somehow installed itself during my troubleshooting.
I also don't know whether it was the "enable-shared" option, the
"disable-dlopen" option, or both that helped me finally get Rmpi to
work. I could go back and uninstall everything to start over and
pinpoint it, though I don't think I can go through the frustration
again. Regardless, I hope that this information helps other users.
Thank you very much Dirk and Brian for your help. And thank you Hao for
such a great job with the Rmpi package. As always, guys like you are
what make R so successful.
On 08/30/2012 11:02 AM, Dirk Eddelbuettel wrote:
On 29 August 2012 at 20:07, Linh Tran wrote:
| Thank you for the advice. I tried using the command, and it still
| wouldn't load.
|
| I tried uninstalling all of the MPI interfaces, reinstalled openmpi
| using the "--enabled-shared --disable-dlopen" command, and Rmpi was able
| to install sucessfully inside R.
I just spent some time going over this, along with Prof Hao, the author of
Rmpi. You may simply have been missing the openmpi-bin package which
provides tools like orterun -- Open MPI now needs this at startup. This may
get added to configure tests in the future to alert users.
With this openmpi-bin package installed, I could easily install Rmpi on
Debian and Ubuntu using the distro's Open MPI packages. No need for local
Open MPI builds. Also note this post on the Open MPI list:
http://www.open-mpi.org/community/lists/devel/2012/04/10840.php
Static builds may lead to loss of resources at runtime as you may have
multiple copies running.
| Thank you to Prof Ripley for pointing that out! I think it was shared
| and dlopen options that kept preventing me from succeeding.
I believe this advise to be incorrect. We have been using dynamically linked
builds of Open MPI for several years on Debian with good success. I was
involved in an overhaul of Debian's Open MPI packages around 2006/2007; the
package has since been very well maintained by others.
Dirk
| On 08/29/2012 01:40 PM, Dirk Eddelbuettel wrote:
| > On 29 August 2012 at 11:37, Linh Tran wrote:
| > | I've spent a few days trying to install Rmpi with no luck. I originally
| > | tried using mpich, moved on to mpich2, and then to openmpi. I've gotten
| > | the furthest with openmpi, though am still running into this problem and
| > | can't figure it out. Can someone help!? Thanks so much in advanced.
| > |
| > | I'm using an HP Envy laptop with Ubuntu 12.04. Output is below, with the
| > | error at the bottom.
| >
| > I suggest you do
| >
| > sudo apt-get install r-cran-rmpi
| >
| > which installs the prebuild Rmpi as well as the Open MPI libary it depends
upon.
| >
| > Dirk
| >
|
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
