This vignettes illustrates how to launch parallel workers on the current, local machine. This works the same on all operating systems where R is supported, e.g. Linux, macOS, and MS Windows.
The below illustrates how to launch a cluster of two parallel workers on the current machine, run some basic calculations in paralllel, and then shut down the cluster.
library(parallelly)
library(parallel)
cl <- makeClusterPSOCK(2)
print(cl)
#> Socket cluster with 2 nodes where 2 nodes are on host 'localhost'
#> (R version 4.4.2 (2024-10-31), platform x86_64-pc-linux-gnu)
y <- parLapply(cl, X = 1:100, fun = sqrt)
y <- unlist(y)
z <- sum(y)
print(z)
#> [1] 671.4629
parallel::stopCluster(cl)
Comment: In the parallel package, a parallel worker is referred to a parallel node, or short node, which is why we use the same term in the parallelly package.
An alternative to specifying the number of parallel workers is to
specify a character vector with that number of "localhost" entries,
e.g.
cl <- makeClusterPSOCK(c("localhost", "localhost"))
The availableCores() function will return the number of workers that
the system allows. It respects many common settings that controls the
number of CPU cores that the current R process is alloted, e.g. R
options, environment variables, and CGroups settings. For details, see
help("availableCores"). For example,
library(parallelly)
cl <- makeClusterPSOCK(availableCores())
print(cl)
#> Socket cluster with 8 nodes where 8 nodes are on host 'localhost'
#> (R version 4.4.2 (2024-10-31), platform x86_64-pc-linux-gnu)