we’ve run into an issue with the recent cuda 13.0 release. during linking, calls to nppiConvert_8u32f_C1R result in undefined reference errors. however, the context-based variant nppiConvert_8u32f_C1R_Ctx links without any problem.
it looks like non-CTX functions such as nppiConvert_8u32f_C1R are no longer resolvable in this version, even though they were in previous releases.
why does non-CTX functions still present in the header file if they are not available?
has anyone else experienced this with cuda 13.0, and is there a confirmed workaround or change in npp linking behavior that explains it?