Chapter 5 Parallel Computing in R

Note The classification of explicit and implicit parallelization is given in this slides. The difference between these two types of parallelization is similar to that between multi-processing and multi-threading. They can be combined.

• Explicit paralellization

• Implicit paralellization

• Combined approach

• Limitations

  • The "load data at master -> send to workers -> collect at master -> send to worker->repeat" workflow is not suitable for interactive data analysis and iterative algorithms.

  

Note Figure source with adaptation.

^{R_hpc_packages}. The CRAN page of High-Performance and Parallel Computing with R contains such a list of packages. ↩

^{pnmath_archive}. Download the package from here. This is the demo. ↩

• pbdR

• GPU computing. Rth is a package of functions written in Thrust, callable from R. It thus provides to R programmers a set of parallel applications that run on both GPUs and multicore systems.