CODEWITHSUNDEEP
rparallel-processingdomc
statistician_in_training
statistician_in_training

Reputation: 813

Random number generation with doSMP

I'm trying to use doSMP package for parallel programming in R. Here's a snippet of the code I wrote to test whether each MC replicate starts from the same seed, and whether RNG will give the same number The random numbers come out the same. I want each MC replicate to be random(I generate random numbers in run.MC.replicate). How can I fix this?

par.mc.result<- foreach(mc =1:nmc,.packages=c("MASS") ) %dopar% {

    source("./src/simulation_math_util_fn.R")

    source("./src/oosMDS.R")
    source("./src/smacofM.R")
    source("./src/oosIM.R")
    sink(file=paste("debug",mc,collapse=""))
    print(runif(1))
            run.MC.replicate(myParams)

    }

I'm guessing this behaviour happens because of the way doSMP generates child R processes. Please help

Upvotes: 2

Views: 189

Answers (1)

Joshua Ulrich
Joshua Ulrich

Reputation: 176698

You need a RNG suited for parallel computing. Go to the High Performance Computing Task View and see the Parallel computing: Random numbers section. It lists the rsprng and rlecuyer packages.

Upvotes: 3

Related Questions