R/parallelisiere.R

In holgerman/toolboxH: Helperfunctions for R

#' @title FUNCTION_TITLE
#' @description FUNCTION_DESCRIPTION
#' @param proc PARAM_DESCRIPTION, Default: 3
#' @param on_server PARAM_DESCRIPTION, Default: Sys.info()["sysname"] == "Linux"
#' @param mkl_threads PARAM_DESCRIPTION, Default: 1
#' @return OUTPUT_DESCRIPTION
#' @details DETAILS
#' @examples
#' \dontrun{
#' if(interactive()){
#'  #EXAMPLE1
#'  }
#' }
#' @seealso
#'
#' @rdname parallelisiere
#' @export
parallelisiere = function(proc=3, on_server = Sys.info()['sysname']=="Linux", mkl_threads = 1) {
  # mkl_threads: Revolution R aka MRO uses internal multithreading via Intel Math Kernel Libraries. This sometimes seames to conflict with parallizing by foreach
  if(on_server ==T) {
    for(i in c( "doMC", "foreach")) {
      suppressPackageStartupMessages(library(i, character.only = TRUE))
    }

    doMC::registerDoMC(cores=proc)
    if("RevoUtilsMath" %in% installed.packages()) {
      RevoUtilsMath::setMKLthreads(1)
      message("MKL threads set to ", RevoUtilsMath::getMKLthreads())
    }

  }

  if(on_server ==F) {
    for(i in c( "doSNOW", "foreach")) {
      suppressPackageStartupMessages(library(i, character.only = TRUE))
    }

    registerDoSNOW(snow::makeCluster(proc, type = "SOCK"))
    if("RevoUtilsMath" %in% installed.packages()) {
      RevoUtilsMath::setMKLthreads(1)
      message("MKL threads set to ", RevoUtilsMath::getMKLthreads())
    }

  }
  message("Parallel backbone initiated for ",proc, " parallel processes....")

}