Rnvvm

Rnvvm_0.1-0.tar.gz

This package is an R interface to libNVVM - a library that converts LLVM code (in intermediate representation (IR) form) to PTX code that can be loaded and run on a GPU. One aim of this package is to be able to compile R code to LLVM using the Rllvm and RLLVMCompile packages, and then to convert it to PTX code. We then run this code on the GPU with the RCUDA package.

As of release candidated 5.5, libNVVM is released as part of the CUDA SDK distributed by NVIDIA.

Installation

When installing this package, the configuration script will look for the collection of nvvm files, i.e. nvvm.h header and the compiled DLL/DSO. It looks in the typical locations, i.e. /usr/local/cuda/nvvm or /usr/local/cuda-5.5/nvvm. If you have installed nvvm in a different location, specify this by setting the environment variable LIBNVVM_HOME to the nvvm/ directory which contains the include/ and lib64/ directories.

Duncan Temple Lang <dtemplelang@ucdavis.edu>

Last modified: Sun Jun 30 06:34:11 PDT 2013