Message-ID: <D77D49DA-7B4B-4CCD-B6A3-A17F736A6FE0@gmail.com>
Date: 2013-01-31T16:59:49Z
From: michele caseposta
Subject: rgenoud and snow
Hello everyone.
I am trying to use rgenoud in a parallel environment.
While making some tests, I noticed that rgenoud seems not to be using the cluster created by makeCluster, doing all the computation in the main node.
Following is the example:
ncores <- 5
cl <- makeCluster(rep('localhost', ncores), type='SOCK')
evalFunc3 <- function(x){
r <- rnorm(1000000)
print(Sys.getpid())
return(x[1]/2-x[2]^3+0.5)
}
genResSeq <- genoud(evalFunc3, nvars=2, max=T, pop.size=ncores, cluster = cl)
stopCluster(cl)
Is this the expected behavior? I would have guessed each node to be busy with the evaluation of the function for each individual, but it does not seem to be the case.