DPC++ Performance on Representative Structured/Unstructured Mesh
This session presents an overview of performance achieved with Data Parallel C++ (DPC++) on Intel® CPUs with MG-CFD, an unstructured-mesh CFD mini-app, and OpenSBLI, a structured mesh academic computational fluid dynamics (CFD) code. Compare results to OpenMP* implementations and explore key differences and bottlenecks based on Intel® VTune™ Profiler and Intel® Advisor feedback.
István Reguly is an associate professor at Pazmany Peter Catholic University (PPCU), focusing on domain-specific languages for high-performance computing. He primarily works on the performance, portability, and productivity of the discretized solution of partial differential equations.