Straightforward Heterogeneous Computing with the oneAPI Coexecutor Runtime

Heterogeneous systems are the core architecture of most computing systems, from high-performance nodes to embedded devices, due their excellent performance and energy efficiency. Efficiently programming these has become a major challenge complexity architectures efforts required provide them with co-execution capabilities that can fully exploit applications. There many proposals simplify management acceleration devices multi-core CPUs. However, in cases, portability ease use compromise efficiency different devices—even more so when co-executing. Intel oneAPI, new powerful standards-based unified model, built on top SYCL, addresses issues. In this paper, oneAPI is provided strategies run same kernel between enabling exploitation static dynamic policies. This work evaluates for well-known set regular irregular HPC benchmarks, using two heterogeneous composed an integrated GPU CPU. Static load balancers evaluated, highlighting single significant key points promising technology. Experimental results show worthwhile algorithms improves even further shared memory.