Getting gputools and gmatrix in R to run smoothly on Linux - r

Question

I am having trouble getting R to recognize gmatrix for R, running through Rstudio. My details are below.
Things that are accomplished and DONE:
install of openSuse Tumbleweed x86_64
install of NVIDIA graphics driver 361.42
install of CUDA 7.5 & symbolic link for /usr/local/cuda
set LD_LIBRARY_PATH & PATH
R 3.2.5
RStudio 0.98.501-1.62
patched /usr/local/cuda/include/host_config.h, line: 115 by increasing the "4"'s to "5"'s (this pertains to the gcc compiler)
install package gputools
~ partial install of package gmatrix
*I had to edit the host_config.h file because cuda is not supported for the gcc compiler that Tumbleweed ships with. I basically broke the evaluation that checks for version and set it to 5.9 instead of 4.9. Was this a supremely naive thing to do?
Things not Done, help required:
-I wasn't able to install gmatrix properly. I got R to install by manually adding the LD & PATH directories using the Sys.setenv()
> Sys.setenv("LD_LIBRARY_PATH"="/usr/lib64/R/lib:/lib:/usr/local/lib64:/usr/local/cuda/lib64")
> Sys.setenv("PATH"="/home/eamergent/bin:/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/usr/games:/usr/local/cuda/bin")
However, I am unable to load the gmatrix package, yet gputools works fine:
> library(gputools) "no issues"
> library(gmatrix)
> Error in dyn.load(file, DLLpath = DLLpath, ...) : unable to load shared object '/home/eamergent/R/x86_64-suse-linux-gnu-library/3.2/gmatrix/libs/gmatrix.so':libcudart.so.7.5: cannot open shared object file: No such file or directory Error: package or namespace load failed for ‘gmatrix’
*The Error makes it seem that libcudart.so.7.5 is to blame
However,
/usr/local/cuda/lib64/libcudart.so.7.5 does in fact exist:
> eamergent#linux-pnx7: ls /usr/local/cuda/lib64/libcuda
libcudadevrt.a libcudart.so libcudart.so.7.5 libcudart.so.7.5.18 libcudart_static.a
*I used tab to get a shorter list. there are many more in lib64
Any ideas on how to get R to recognize the already installed gmatrix package? Or how to properly reinstall without using the R environment path fix that I used above?

Answer 1

I had the same issue on Ubuntu 14.04 where the system and R (Sys.setenv()) shared library paths were set properly, but R failed to load the CUDA shared library. What worked was to set the CUDA shared library path using ldconfig.
as root or prefix commands with sudo:
# echo "/usr/local/cuda/lib64/libcuda" > /etc/ld.so.conf.d/cuda.conf
# ldconfig
then
# R
> install.packages("gmatrix")
or if gmatrix is already installed
# R
> library(gmatrix)
> gtest()