.PVM.spawn {rpvm} | R Documentation |
Starts up ntask
copies of an executable file task
or
slave R processes on the virtual machine.
.PVM.spawn (task, ntask = 1, flag = "Default", where = "", arglist = NULL, verbose = FALSE) .PVM.spawnR (slave, ntask = 1, flag = "Default", where = "", slavedir = "demo", outdir = "/tmp", verbose = FALSE) .PVM.spawnflags
task |
name of an executable file |
ntask |
number of tasks to spawn |
flag |
options flag |
where |
name of a host or architecture |
arglist |
a character vectpr to be passed to the spawned task. |
slave |
name of slave R script |
slavedir |
full path to the slave script |
outdir |
full path to the directory where the output from slave R process will be |
verbose |
if true, print out more information which might be helpful in debugging |
flag is a list of strings used to specify options. Allowed values are
"Default" | PVM chooses where to spawn processes |
"Host" | spawn on a particular host (where ) |
"Arch" | spawn on a particular PVM_ARCH (where ) |
"Debug" | starts tasks under a debugger |
"Trace" | trace data is generated |
"MppFront" | starts tasks on MPP front-end |
"HostCompl" | complements host set in where |
task
is a character string containing the name of the
executable file name of the PVM process to be started. It must
already reside on the host on which it is to be started and can be
find by local pvmd (possibly by specifying ep=
option when
the host is added, see .PVM.addhosts
). The default
location PVM looks at is $PVM_ROOT/bin/$PVM_ARCH
.
To spawn a R process, one can use the script ‘slaveR.sh’ which
has to be included in the search path of pvmd, for example by passing
ep="$R_LIBS/rpvm"
(the default place this script is installed)
to .PVM.addhosts
. Note that here "R_LIBS"
referes to
the R library on the slave node.
.PVM.spawnR
can be used to conveniently spawn slave R
processes. The paths for the slave script and output file should be
given as arguments toslavedir
and outdir
.
The name of slave script must be given as argument slave
. It
is crucial that this script is present and can be found on each slave
node. The input file path has to be relative to system.file
(package = "rpvm")
on each node, and it has to be the same on all
nodes. The default input path is the "demo"
subdirectory
of rpvm
installed directory.
The output path can be either relative to the home directory or an
absolute pathname (the default is "/tmp"
). The output file
from slave R
process will have the same name as the slave
script with process id and .Rout
appended to it.
.PVM.spawnflags
is a mapping of strings to integer constants
defined by PVM.
A vector containing tids
of successfully spawned processes (the
length could be smaller than the number requested) or -1 if there is
an system error.
Na (Michael) Li nali@umn.edu and A.J. Rossini rossini@u.washington.edu
PVM documentation
## Not run: children <- .PVM.spawnR (slave = "helloR.R") ## End(Not run)