[R-sig-hpc] Problems Installing Rmpi

R. A. Bilonick rab at consolidated.net
Thu May 14 22:22:08 CEST 2009


On Thu, 2009-05-14 at 15:54 -0400, Rowe, Brian Lee Yung (Portfolio
Analytics) wrote:
> As much as I like Debian, I never found installing a new OS as the
> 'easier' solution for a software problem, though if you're on a VM (or
> Windows) the argument is different. Of course it also depends on your
> level of experience with Linux and its related compilation process.
> 
> Rather than speculating, you can see explicitly what mpicc is using with
> the -showme variants:
> 
> mpicc -showme
> mpicc -showme:compile
> mpicc -showme:link
> 
> man mpicc gives you the details. Noting the values of the above commands
> and analyzing your configure.ac script should give you enough clues to
> narrow down the root cause and fix it appropriately.
> 
> HTH,
> Brian
> 

I don't think it is related at all to mpicc. I have tried this too many
ways to enumerate but so long as mpi.h was located during the install, I
always ended up with "mpi.h" being appended to the path in the compile
statement. My biggest Linux weakness is that I'm not that familiar with
shell scripting.

I appreciate everyone's help on this. I enjoy the challenge of getting
it working and learning the details, but my Linux skills are limited
(even after 10 years or so). I only chose CentOS because I could not get
Fedora 10 to install with a RAID system (this bug was way beyond my
skills) and CentOS is similar to Fedora (although somewhat behind given
it is a clone of RHEL). Although in many ways I prefer Fedora, I've come
to depend on Ubuntu when it's just too difficult to get Fedora working.
I ran the Ubuntu 9.04 64-bit cd in live mode, and everything worked
nicely. I'm worried about it working with the RAID when I install it to
the hard drive. I did notice that packages for both openmpi and Rmpi are
in the repositories. So we will see how it goes. The main point of this
computer is to run R in parallel, so whatever makes this easy or just
possible I'm willing to try.

Rick B.



More information about the R-sig-hpc mailing list