Parallel.Rmd

---
title: "Parallel"
author: "Eva Kleingeld"
date: "August 12, 2016"
output: pdf_document
---

This script contains the code to build/run a model in parallel AND top build/run a model in parallel on multiple computers. 



```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r }
#library(doParallel)
library(parallel)
library(doParallel)
library(foreach) #loop parallel
# foreach () %doPar%

cl<-makeCluster(4)
registerDoParallel(cl)
registerDoSEQ() ######### What does this function do?


#example
system.time(
  t <-foreach(i=1:100000,combine=cbind) %dopar% { sqrt(i) }
)[3]

stopCluster(cl)
CloseC

## Normal run is faster?
system.time( 
  for (i in c(1:100000)) { 
    t[i] <- sqrt(i) })


library(doParallel)
library(foreach)

cl<-makeCluster(6)
registerDoParallel(cl)

#Meerdere computers
user    <- 'Eva'
primary <- 'pc150169.knmi.nl' #number of your computer .knmi.nl
machineAddresses <- list(
  list(host=primary,user=user,
       ncore=4),
  list(host= 'pc150395.knmi.nl',user=user,
       ncore=1) #list all the computers you want to use with the number of cores
)
spec <- lapply(machineAddresses,
               function(machine) {
                 rep(list(list(host=machine$host,
                               user=machine$user)),
                     machine$ncore)
               })
spec <- unlist(spec,recursive=FALSE)
parallelCluster <- parallel::makeCluster(type='PSOCK',
                                         master=primary,
                                         spec=spec)
print(parallelCluster)
clusterEvalQ(parallelCluster, library(doParallel)) #To check wether packages are available
registerDoParallel(parallelCluster)
#Here room for some calculations with a foreach loop

stopCluster(parallelCluster)

```