Automated Scheduling Algorithm Selection and Chunk Parameter Calculation in OpenMP

Increasing node and cores-per-node counts in supercomputers render scheduling load balancing critical for exploiting parallelism. OpenMP applications can achieve high performance via careful selection of kind chunk parameters on a per-loop, per-application, per-system basis from portfolio advanced algorithms (Korndörfer et al. , 2022). This approach is time-consuming, challenging, may need to change during execution. We propose Auto4OMP novel automated applications. With Auto4OMP, we introduce three algorithm methods an expert-defined chunk parameter OpenMP's schedule clause's respectively. Auto4OMP extends the schedule(auto) chunk parameter implementation LLVM's runtime library automatically select algorithm calculate Loop characteristics are inferred loop execution over application's time-steps. The experiments performed this work show that improves by up $11\%$ compared outperforms manual selection. MPI+OpenMP explicitly minimizing thread- implicitly reducing process-load imbalance.