Scalability and composability of flow accumulation algorithms based on asynchronous many-tasks

Models simulating the state of biological and physical environment can be built using frameworks that contain pre-developed data structures operations. To achieve good model performance it is important individual modelling operations perform scale well. Flow accumulation support use criteria for selecting how much material flows downstream are an part in several Earth surface simulation models. For these operations, no algorithms exist perform, scale, compose The objective this study to develop algorithms, evaluate their performance, scalability, composability. We base our on asynchronous many-task approach parallel concurrent computations, which avoids synchronization points supports composability relative strong weak scaling efficiencies when a flow operation over six CPU cores NUMA node 83% 84% respectively. case-study four cluster nodes 73% 84%. Our composable: latency executing two combined lower than sum latencies. • well additional hardware. combined. Asynchronous many-tasks useful implementation algorithms. Concurrent tasks hide temporal load imbalance inherent