Kokkos C++ Performance Portability Programming EcoSystem

When I was at CERN during 2003-2004, a common discussion theme was about merging the threading libraries on ATLAS HLT projects (there were three if I recall correctly).

which never happened, because the reasoning was each of them had different tradeoffs and no one wanted to loose their specific library.

I wonder how many of them are now in use.

This is mostly anecdotal, but I help develop an astrophysical code and we’re very interested in adopting Kokkos, and are close to pulling the trigger. (We’re not affiliated with any of these labs).

Being able to maintain a single software stack compatible with CPUs, NVIDIA GPU, and AMD GPUs is incredibly appealing (Especially given, how much computing resources vary between clusters).

I know other groups are interested too.