Unit Testing

[cdeterman]

As noted in gpuRcuda I currently don't have access to a NVIDIA GPU so I will depend on the R community to confirm that the unit tests created herein pass.

[cdeterman]

@gpuRcore/cuda_users at your convenience. Could you please try to install both gpuRcuda and gpuRcublas and run the single unit test I have implemented. This will be a good confirmation of package installs working properly and that you can run the unit tests.

You should be able to install gpuRcuda directly from git with

devtools::install_github('gpuRcore/gpuRcuda')

But you will want to clone the gpuRcublas repo so you will have the unit tests to check. The steps for this are simply.

git clone https://github.com/gpuRcore/gpuRcublas.git
cd gpuRcublas.git
Rscript -e 'devtools::test()'

That should be sufficient to build and test the package. Thanks for you assistance in assuring these packages integrity.

[cdeterman]

@gpuRcore/cuda_users I realize I neglected to note that the thrust dependency is not yet on CRAN. You will also need to install that with

devtools::install_github('wrathematics/thrust')

[cdeterman]

@gpuRcore/cuda_users I also forgot to mention that there is a fix in the dev version of Rcpp which is also not yet up on CRAN. Please install the git version of Rcpp

devtools::install_github('RcppCore/Rcpp')

Also, when installing gpuRcuda please note the configuration output to confirm it is installing with the CUDA backend (i.e. it will be using the nvcc compiler). If not, try setting Sys.setenv("BACKEND"="CUDA") prior to installing.