Actually, I just did a compilation of the cuda component with cuda 13 headers and there were no compilations errors, so I guess that is good! So at least the signatures for the cuda calls made by the component have not changed. The only question is are there
certain C marcos (such as #define some-value) that have changed in 13 vs 12 or did nvidia make an effort to keep backward compatibility in this regard.
Then there is the issue of runtime, of course.
I can set aside the cuda-13.0.0 installation tree in the cloud if you want to get at it (too large to attach it to the email).
Thanks, Steve