Rmpi needs help (I'd say a bugfix) if your MPI is not lam (and you
have not told us). See the Rmpi/README file, but the effect is more
serious in 2.12.0 (and the maintainer was informed weeks ago).
I don't understand your notation, but am assuming lines starting with
# are both your comments *and* error messages from R.
On Thu, 21 Oct 2010, Cutler, Gene wrote:
I've been trying to get some sort of parallel computing setup
running across my two Macs, one with 8 cores the other with 4. So
far, I've mainly been trying to get MPI to work, but without a lot
of success. For example:
Attempt 1:
/usr/local/openmpi/bin/mpirun --prefix /usr/local/openmpi -hostfile ~/hostfile
/Library/Frameworks/R.framework/Versions/2.12/Resources/library/snow/RMPISNOW
# this drops me into R and I can see messages from multiple R instances
starting up, so far so good...
library('Rmpi')
library('snow')
cl <- getMPIcluster()
cl
# NULL
cl <- makeMPIcluster()
# cl <- makeMPIcluster()
# Error in makeMPIcluster() : no nodes available.
# Execution halted
# --------------------------------------------------------------------------
# mpirun noticed that the job aborted, but has no info as to the process
# that caused that situation.
# --------------------------------------------------------------------------
Note, that in fact any error such as just typing a stray character will cause R
to quite with that same 'aborted' notice.
Attempt 2:
/usr/local/openmpi/bin/mpirun --prefix /usr/local/openmpi -hostfile ~/hostfile
-n 1 R --slave
library(doMPI)
cl <- startMPIcluster()
# Error in system("lamnodes", TRUE, ignore.stderr = TRUE) :
# error in running command
# Calls: startMPIcluster -> mpi.universe.size -> grep -> system
# Execution halted
# --------------------------------------------------------------------------
# mpirun noticed that the job aborted, but has no info as to the process
# that caused that situation.
# --------------------------------------------------------------------------
I've also looked into PVM, but that doesn't seem to support multiple cpus per
machine. Using sockets with snow was also giving me problems, but I'll try
that again only after completely giving up on MPI.
Does anyone have some suggestions for getting this to work?
Thanks in advance,
Gene
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
--
Brian D. Ripley, rip...@stats.ox.ac.uk
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
