Watertight ray triangle intersection


I’m looking for a watertight ray/triangle intersection test.

I already tried this solution, to no avail. Oddly enough I have more triangle miss with this watertight approach than with a simple optix::intersect_triangle() call.

Do you know a robust approach for this problem ?

Alternatively, is it possible to slightly modify optix intersection tests to allow an epsilon on edge tests (even if it gives me false positive, I’d prefer that to false negative) ?

Any insight would be useful, thank you

The problem was not watertightness, but something else in my code (error in backface culling).