MPI detach — Towards automatic asynchronous local completion

When aiming for large-scale parallel computing, waiting time due to network latency, synchronization, and load imbalance are the primary opponents of high efficiency. A common approach hide latency with computation is use non-blocking communication. In presence a consistent imbalance, synchronization cost just visible symptom imbalance. Tasking approaches as in OpenMP, TBB, OmpSs, or C++20 coroutines promise expose higher degree concurrency, which can be distributed on available execution units significantly increase balance. Available MPI functionality does not integrate seamlessly into such tasking parallelization. this work, we present slim extension interface allow seamless integration communication concepts asynchronous OpenMP C++. Using our concept allows span task dependency graphs over full memory application. We furthermore investigate compile-time analysis necessary transform an application using blocking integrating tasks proposed extension.