Download Programming on Parallel Machines - matloff
Transcript
140 CHAPTER 8. INTRODUCTION TO PARALLEL R 8.2 Some Parallel R Packages Here are a few parallel R packages: • Message-passing or quasi-message passing: Rmpi, snow, foreach • Shared-memory: Rdsm, bigmemory • GPU: gputools, rgpu A far more extensive list is at http://cran.r-project.org/web/views/HighPerformanceComputing. html. Some of these packages will be covered in the following sections. 8.3 Installing the Packages With the exception of Rgpu, all of the packages above are downloadable/installable from CRAN, the official R repository for contributed code. Here’s what to do, say for Rmpi: Suppose you want to install in the directory /a/b/c/. The easiest way to do so is use R’s install.packages() function, say: > i n s t a l l . p a c k a g e s ( ” Rmpi ” , ” / a / b / c / ” ) This will install Rmpi in the directory /a/b/c/Rmpi. You’ll need to arrange for the directory /a/b/c (not /a/b/c/Rmpi) to be added to your R library search path. I recommend placing a line . libPaths (”/ a /b/ c /”) in a file .Rprofile in your home directory (this is an R startup file). In some cases, due to issues such as locations of libraries, you may need to install a CRAN package “by hand.” See Section 8.7.1 and 8.7.3 below. 8.4 Rmpi The Rmpi package provides an interface from R to MPI. (MPI is covered in detail in Chapter 6. Its author is Hao Yu of the University of Western Ontario.