Closed bwmeyers closed 4 months ago
We decided on using a "macro" approach, which makes the code GPU-falvour agnostic. Aliases/macros are created depending on which GPU language is being used (HIP or CUDA), but the code itself now just refers to these functions as gpu*
functions.
Once we have identified which parts of VCSBeam need to be changed, the next step is figuring out how to best make those changes. Is it a matter of simply creating some macros that, at compile time, make the decisions based on the host system?