[SOLVED] Can I rely on helper_cuda.h for future compilation?

When I need cuFFT, I normally include helper_cuda.h so I can use the return values for easier debugging.
Is it ok, considering that one day I will compile with CUDA 10, 11, or this is not future-proof?

It is not future proof. The CUDA sample codes are not maintained for production use. They are subject to change.

This isn’t just a hypothetical statement. There is a historical example for reference. They already went through a major breaking change in the CUDA 4.2 - CUDA 5.0 era, and the web is littered with people who were wondering what happened and what to do, because they built their codes on components provided by the CUDA samples (called SDK up to that juncture).



Thanks for the clarification, txbob!