Create a parallel Fork cluster, if useful — makeOptimalCluster • pemisc

Given the size of a problem, it may not be useful to create a cluster. This will make a Fork cluster (so Linux only)

makeOptimalCluster(
  useParallel = getOption("pemisc.useParallel", FALSE),
  MBper = 500,
  maxNumClusters = parallel::detectCores(),
  assumeHyperThreads = FALSE,
  ...
)

Arguments

useParallel: Logical or numeric. If FALSE, returns NULL. If numeric, then will return a cluster object with this many cores, up to maxNumClusters
MBper: Numeric. Passed to memRequiredMB in optimalClusterNum()
maxNumClusters: Numeric or Integer. The theoretical upper limit for number of clusters to create (e.g., because there are only 3 problems to solve, not parallel::detectCores)
assumeHyperThreads: Logical. If TRUE, then it will more efficiently divide the maxNumClusters by useParallel, so that there is a lower number of cores used. This calculation may not be the ideal balance. A message will indicate the change from maxNumClusters, if there is one.
...: Passed to makeForkClusterRandom. Only relevant for iseed.