Parallel computation

Parallel version of replicate with parallel

Loading the package:

> library(parallel)

of which we’ll use the 6 following functions:

• detectCores which detects the number of CPU cores;
• makeCluster which creates a cluster of a given number of cores;
• clusterEvalQ which evaluates a literal expression on each cluster node. It is a parallel version of evalq.
• clusterExport which assigns the values on the master R process of the variables named in its named list argument to variables of the same names in the global environment (aka workspace) of each node;
• parSapply parallel version of sapply.
• stopCluster which shuts done the cluster.

The first step is to create a cluster. It’s good practice to let one core available, in case:

> cl <- makeCluster(detectCores() - 1)

Then, we need to populate each core of the cluster with any package we may need:

> clusterEvalQ(cl, library(package1, package2))

as well as any object (data and functions) we may need from the current workspace, for example:

> clusterExport(cl, c("data1", "data2", "fct1", "fct2"))

Then, you ready to launch your calculations with nb replicates:

> out <- parSapply(cl, 1:nb, function(x) {
+ # here you include what should be done in each replication.
+ # this calculation will require the packages package1 and package2,
+ # as well as data data1 and data2 and function fct1 and fct2 of the
+ # workspace.
+ })

Once your calculations are done, you have to terminate the cluster:

> stopCluster(cl)