Welcome to the doMPI project

The 'doMPI' package provides a parallel backend for the 'foreach' package. It is similar to the 'doSNOW' package, but uses 'Rmpi' directly. This allows it to do more, and (hopefully) execute more efficiently.

Features of the doMPI backend

• Data that is common to all tasks can be broadcast to the cluster workers, or piggy-backed with the first task to the workers.
• It is easy to write scripts that can be executed interactively or in batch mode using mpirun.
• A chunkSize option is supported to increase the task granularity for better performance.
• If chunkSize is used, and a cluster worker has multiple cores, the mclapply function from the 'parallel' package can be used to execute the tasks in parallel, thus enabling hybrid parallelism.
• An arbitrarily large number of tasks can be executed, since task arguments are not fetched until they are ready to be submitted, and results are processed as they are returned.
• Backend-specific options for initializing and finalizing the worker execution environments.
• Support for parallel random number generation.

The project summary page you can find here.