Installing GPU Packages

After you have installed the CUDA Toolkit, you can start installing the R GPU packages. Since some of the packages have to be compiled from source, you must also install the R developer environment. For Fedora users, enter the following in a terminal:

As for Ubuntu users, use the following instead.

Ubuntu users should follow the R installation instruction for Ubuntu as found in CRAN for the latest release.

Installing RPUD Package

Define a environment variable named R_LIBS_USER in .bashrc that points to an existing folder in your home directory. In the example below, I have chosen a folder lib/R created ahead of time.

You would need to install a required dependency of SparseM in R.

Now retrieve the latest release of rpud from the download page, and install it in a terminal:

You may verify the installation by loading rpud in R. The welcome message will show whether it has been successful.

Installing RPUDPLUS Add-on

While rpud computes the distance matrix on GPU in high speed, its data processing capacity is confined by the video graphics memory. In order to make full use of the system RAM available to R (about 2GB) for certain operations, you can install the rpudplus add-on. It enables rpud to handle larger data sets in calculating distance metrics, and can perform more tasks efficiently. You can download and extract the package in a temporary folder, and run the installation script.

You may now verify your rpudplus installation in R.

Note

The rpudplus add-on requires double precision arithmetic hardware support. In order to fully exploit rpudplus, you should ensure the compute capability of your CUDA GPU exceeds 1.3 or above according to the CUDA hardware page.