Document best practices for linking apps against Trilinos with CUDA
Created by: mhoemmen
@trilinos/framework @trilinos/tpetra
It looks like applications are not linking against Trilinos with CUDA in the way that Trilinos itself builds and links with CUDA. See discussion here: https://github.com/gahansen/Albany/issues/195
This suggests that Trilinos needs to document best practices for linking applications against Trilinos with CUDA. It could be that applications have multiple good options, in which case Trilinos should document the strengths and weaknesses of each.