4 Speeding up iteration procedures

At this stage, we assume that you’ve grown familiar with R’s most relevant loop constructs, including for loops and the *apply family of functions. Yet, there’s two other packages we’d like to introduce within the scope of this short course due to their convenient performance when it comes to accelerating certain operations. But first things first… here’s the topics (and the related packages) that we’re gonna cover in the upcoming section on speeding up iteration procedures in R.

Parallelization via doParallel (and foreach)

Just a bit of theory ahead.

“Speedup is a common measure of the performance gain from a parallel processor. It is defined as the ratio of the time required to complete the job with one processor to the time required to complete the job with N processors. […] In a machine where work can be dynamically assigned to available processors, it is attained as long as the number of pieces of work ready for processing is at least N.” (Denning and Tichy 1990)

Or, to cut a long story short, if you’re performing one and the same iteration again and again, you may as well distribute an equal amount of sub-iterations to each processor available on your local machine. In theory, distributing an operation to N nodes would result in a n-fold speed gain.

In this scope, we’ll have a brief look at R’s capabilities in terms of parallel processing using the doParallel package along with foreach (an other package to deal with loop constructs). Be aware, however, that there are plenty of opportunities to “go parallel” in R which you might want to have a look at.

“Cross-lingual” programming via Rcpp

The Rcpp package offers a seamless integration of C++ functionality in R. The underlying reason why such a thing exists in the first place is rather easy to explain: sometimes R code is just not fast enough. But don’t be afraid if you haven’t gotten in touch with C++ so far: we’re barely gonna scratch the surface of what is possibly when combining those two languages. In fact, our short excursion on the topic is merely meant to raise awareness of such things slumbering in the depth of the R universe. In case you’d like to learn more about the subject, we recommend to have a closer look at Hadley Wickham’s comprehensive introduction to High performance functions with Rcpp (Wickham 2014) or, if you’re a fan of hard copies, Dirk Eddelbuettel’s book about “Seamless R and C++ Integration with Rcpp” (Eddelbuettel 2013).

References

Denning, Peter J., and Walter F. Tichy. 1990. “Highly Parallel Computation.” Science 250 (4985). American Association for the Advancement of Science: 1217–22.

Wickham, Hadley. 2014. Advanced R. Chapman & Hall/CRC the R Series. CRC Press.

Eddelbuettel, Dirk. 2013. Seamless R and C++ Integration with Rcpp. Use R! Springer. doi:10.1007/978-1-4614-6868-4.