cudaDevAttrConcurrentManagedAccess on Orin

Our multithreaded application uses managed memory extensively. It is quite tedious to correctly attach all buffers at all times. The Xavier family has cudaDevAttrConcurrentManagedAccess = 0, so there is no way around the occasional segfault during development.

Will the Orin family have cudaDevAttrConcurrentManagedAccess = 1?

Hi @tobias.fischer1 , the software does not support full Unified Memory model and hence this will continue to be 0 on Orin.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.