0

Aiming at cross-platform development in Python+numba for JetsonNano, recently installed cudatoolkit (cudatoolkit 10.1.243 h6bb024c_0) on host Ubuntu 18.04. For the project I used environment where other packages needed for the project are installed via conda, same channel, conda-forge.

When the packages in the environment are listed, there is no mention of cuda-cross-aarch64. Tried to find some helping tips, but none of the search results pointed to cross-platform development resources (compiler, linker) from conda packages. The search on Anaconda site for cuda-cross-aarch64 gave no results.

Is there a way to make the cross-platform development work (host Ubuntu 18.04, no cuda cores / GPU on the host + JetsonNano with cuda 10.0) with the cuda related packages from conda?

gudrun
  • 11
  • 2
  • 1
    None of the conda packages contain compilers, headers etc. They are just runtime libraries – talonmies Apr 12 '21 at 06:34
  • Thanks, talonmies. That means I need to have cuda-cross-arch64 installed separately. But, since it seems that anaconda does not have this package in its repo, would you have suggestion as to where to turn to? Should I follow the official documentation form Nvidia and do the installation form official SDK kit? – gudrun Apr 13 '21 at 20:17

0 Answers0