[R-sig-hpc] multi-core processing on single Win7-64 computer

Timothy_Handley at nps.gov Timothy_Handley at nps.gov
Tue Mar 15 17:17:02 CET 2011

Hey folks, I would like to speed up some R processes by splitting tasks
among the 2 cores on my desktop computer. There seem to be a variety of
ways one can do this, however the obvious choices aren't working for me.
I'm hoping that some of you may have advice.

System: Windows 7-64, Core2 Duo CPU, R 2.12.2.

MPI Option: I've successfully installed MPICH2 (1.3.2.p1). The test program
(calculating pi) works via command prompt window, although the
configuration GUI (wmpiconfig) doesn't work (known bug on 64-bit systems).
I then installed Rmpi (0.5-9), but whenever I try to load this package
(call "library(Rmpi)") I receive an error: Error: package 'Rmpi' is not
installed for 'arch=x64'.  This same error was reported last November
(http://comments.gmane.org/gmane.comp.lang.r.hpc/499), but I don't see any

Sockets: The documentation for both snow and snowfall say that one can run
parallel processes with no additional software using sockets. I've
installed the most recent versions of both packages, and tried to set up a
simple socket cluster.  The documentation for snow suggests the commands
  makeCluster(c("localhost","localhost"), type="SOCK")
  makeCluster(2, type="SOCK")
should create a simple cluster on my workstation. However, both commands
hang. After hitting enter, R simply sits until I lose patience (~2-3min).
Note that ESC still functions to stop the command. Similarly, the very nice
website here:
explicitly says that this snowfall command
  sfInit( parallel=TRUE, cpus=2, type="SOCK" )
should work anywhere. Unfortunately, this also causes R to hang. I disabled
my security software (Symantec), then retried the socket commands, but R
still hangs.

Any ideas would be greatly appreciated. Thanks,

Tim Handley
Research Assistant
Channel Islands National Park
(Will be working from both CHIS and SAMO)
SAMO Phone: 805-370-2300 x2412(Mon, Tue)
CHIS Phone: 805-658-5759 (Wed, Thu, Fri )

More information about the R-sig-hpc mailing list