RStudio project template to run parallel code in R, using foreach, doParallel and/or doMPI.

To install this package in RStudio, you need devtools in R. You can install it with

install.packages("devtools")

Then, run the following.

devtools::install_github("jlaria/p4R")

To start a new project go to File -> New Project … and then select p4R Project Template

Name your project and select its folder, then Create Project.

By default, a p4R project template contains the following files and folders.

This is the main script of your project if you are going to use the doParallel backend (if you are running the simulations using one physical host only, e.g. a personal computer).

By default, the parallel backend is set to use half of the available cores.

library(doParallel)
nodes = detectCores()
cl = makeCluster(nodes/2) # Specify number of threads here...
registerDoParallel(cl)

You can change it to cl = makeCluster(nodes) to use the full parallel capacity. However, take into account that many pc’s have two virtual cores for each phisical core, and using all the available virtual cores may cause the system to overheat.

You don’t have to edit this file, unless you want to change the number of threads.

This is the main script of your project if you are going to use the doMPI backend (for instance, if you are using several physical hosts that communicate using MPI).

This file is not supposed to run directly in R. Use either some job queue (such as PBS), or the script file mpi_run.sh. You don’t have to edit this file.

This file contains the declarations of all user functions that are supposed to be accessed by the child processes. If you want to track the progress of your simulations, define a wlog function (default).

wlog = function(text,...){
  cat(paste0(date(),"   ", text,...,"
"), file="log.txt", append = T)
}

This is the source code of your simulation design, divided into the following sections.

• Log file (initialization)
• Global Libraries (add your required libraries here)
• Parameters (declare the global parameters of the simulations here)
• Generate data (If your simulations require data generation, you can do it before the main loop, and save time later. In addition, this is useful if several processes later will share the same dataset)
• Parallel loop (edit to match your simulation requirements)

Folder to place the data files.

At the end of the simulations, results/results.RData contains a copy of the global environment, with the simulation results.

This log file is generated runtime by default, and contains the progress of the simulations. You can log specific details of your simulations using the wlogfunction.