Re: [R] npRmpi trouble - mpi.comm.spawn causes segfault

2012-07-07 Thread claurin
Hi Peter, 

I had this same problem, but only when using Rmpi interactively. My scripts
work when run in batch mode, following the examples at
http://math.acadiau.ca/ACMMaC/Rmpi/ .  Since this message is old, you may
have discovered this already. Nevertheless, if others have the same problem,
knowing this may help them.

Best,

Charles

--
View this message in context: 
http://r.789695.n4.nabble.com/npRmpi-trouble-mpi-comm-spawn-causes-segfault-tp4532040p4635733.html
Sent from the R help mailing list archive at Nabble.com.

__
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.


[R] npRmpi trouble - mpi.comm.spawn causes segfault

2012-04-04 Thread Peter Man

Dear all,

I have a large dataset of randomly generated weighed sample for which I 
wish to compute a kernel density estimate.
I have used the "np" package successfully for smaller datasets, however 
for the larger ones, they take too long when
using the cross validation options for bandwidth selection ("cv.ls" or 
"cv.ml"). Of course, they are much quicker with "normal-reference".


To speed things up, I tried to use the "npRmpi" package. I had a lot of 
trouble installing it - but after appealing to google and finding the 
following page:


http://webappl.blogspot.co.uk/2012/01/install-rmpi-with-mpich2-environment.html

npRmpi was finally installed (basically I just added the flags "-lmpl 
-lopa" to the definition of PKG_LIBS in the configure script, and in the 
R CMD INSTALL call, I passed the configure.args as the correct paths to 
my MPICH2 installation).


However, when I call the "mpi.spawn.Rslaves(nslaves=1)" command, I get 
the following errors:


##
> mpi.spawn.Rslaves(nslaves=1)

 *** caught segfault ***
address 0x2df0c9f, cause 'memory not mapped'

Traceback:
 1: .Call("mpi_comm_spawn", as.character(slave), 
as.character(slavearg), as.integer(nslaves), as.integer(info), 
as.integer(root), as.integer(intercomm), PACKAGE = "npRmpi")
 2: mpi.comm.spawn(slave = system.file("Rslaves.sh", package = 
"npRmpi"), slavearg = arg, nslaves = nslaves, info = 0, root = 
root, intercomm = intercomm)

 3: mpi.spawn.Rslaves(nslaves = 1)

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
##

I also installed Rmpi (I installed Rmpi before npRmpi, and in a similar 
fashion), and calling the same function "mpi.spawn.Rslaves" gives the 
following error:


[mpiexec@manet.somewhere.group] match_arg 
(/home/userspace/plwm2/CPP/mpich2-1.4.1p1_gcc4.6.2/src/pm/hydra/utils/args/args.c:122): 
unrecognized argument pmi_args
[mpiexec@manet.somewhere.group] HYDU_parse_array 
(/home/userspace/plwm2/CPP/mpich2-1.4.1p1_gcc4.6.2/src/pm/hydra/utils/args/args.c:140): 
argument matching returned error
[mpiexec@manet.somewhere.group] parse_args 
(/home/userspace/plwm2/CPP/mpich2-1.4.1p1_gcc4.6.2/src/pm/hydra/ui/mpich/utils.c:1387): 
error parsing input array
[mpiexec@manet.somewhere.group] HYD_uii_mpx_get_parameters 
(/home/userspace/plwm2/CPP/mpich2-1.4.1p1_gcc4.6.2/src/pm/hydra/ui/mpich/utils.c:1438): 
unable to parse user arguments



Please help!!!

Thanks.

Peter

__
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.