1

I only found one post that has a similar problem to me which is here. While my error is similar, i.e. 

The downloaded source packages are in
        ‘/tmp/RtmpbUkeUo/downloaded_packages’
Warning message:
In install.packages("gputools") :
  installation of package ‘gputools’ had non-zero exit status

my situation is not the same.

I am sshing in to a school PC which has CUDA already installed in /usr/local/cuda/. I am using R v 3.0.2 with the platform x86_64-redhat-linux-gnu. When I install.packages("gputools") I choose the CRAN mirror in USA (CA 1) or USA (CA 2).

The beginning of the installation, I see 

package ‘gputools’ successfully unpacked and MD5 sums checked

which is a good sign, but it is later followed by many lines of 

hcluster.cu(XXX): warning: a __device__ variable "hcluster_YYY" cannot be directly read in a host function

where XXX is a number and YYY is a function name.

Would someone please suggest what may be the problem? Is there more information I can provide you that would help you help me? Thank you.

Community
  • 1
  • 1
user3131893
  • 63
  • 5
  • There are installation instruction inside the tarball. Have you read them? You may need particular gcc versions and what not... You may also need to set environment variables to point to the CUDA directory. – Dirk Eddelbuettel Dec 24 '13 at 17:50
  • @DirkEddelbuettel thank you for your response. I have read the INSTALL file and I'm running in to other problems like using an unfound command, `export`, and trying to set `LD_LIBRARY_PATH` which is an undefined variable. Once I figure out these problems, I'll try `install.packages("gputools")` again. – user3131893 Dec 24 '13 at 19:28

0 Answers0