"Creating cuDPP" allocates all available RAM (32GB) never stops, exits with error

I saw this issue discussed and presumably “solved”, but no explanation of any kind of why it happens and what to do when it happens.

With GVDB 1.0 I did not have this problem. Yes, cuDPP creation took a minute or so, but never failed. With GVDB 1.1 it fails.

With GVDB 1.1 the first run of any compiled application starts as usual, but freezes with “creating cuDPP” message, meanwhile RAM allocation increases to all available, 32 GB in my case, and after 10 to 15 minutes the application exist with some error, which explains nothing.

Please, I need a solution ASAP.

Sorry, I overlooked the solution: in cuDPP library ALL __shfl functions should be replaced to corresponding _sync versions, and it works like a charm!