0

I'm having a bit of a trouble running a job on a nVidia cluster. My code is divided into a shared library containing the cuda code and a .cpp file which calls functions from that lib.

The thing is that when I submit the job (llsubmit) I get an error message telling me that ".. cannot open shared object file: No such file or directory"

I've tried adding the path where my .so file is located using the LD_LIBRARY_PATH variable but it still doesn't work.

Could I specify this by adding an entry in my .cmd file?

talonmies
  • 70,661
  • 34
  • 192
  • 269
emmerich
  • 512
  • 1
  • 5
  • 14
  • 1
    This sounds like something you should ask your system administrator about. Loadlever doesn't know anything about CUDA and how it has been set up will be completely site specific. – talonmies Jun 21 '14 at 09:34
  • 1
    This question appears to be off-topic because it is effectively a support request for a specific cluster. – talonmies Jun 21 '14 at 09:35
  • I hoped that someone with more experience with the loadlever might know how to add a specific library (.so) to a .cmd job file or to tell me that this is not possible. As far as I've searched I haven't found such option so this is why I've asked. – emmerich Jun 21 '14 at 13:45

0 Answers0