Automatic Performance Profiling (APP)

The rise in spatial and temporal resolution of Earth System Models (ESMs) enhances physical process representation but increases computational demands. This strains scalability and efficiency on modern HPC architectures.

Identifying and addressing parallel performance bottlenecks in large climate models is complex and time-consuming, requiring specialized expertise. The Automatic Performance Profiling (APP) tool addresses this by providing an automated workflow for performance analysis. APP runs end-to-end profiling experiments, generating comprehensive reports that combine high-level metrics with detailed insights into MPI communication, cache behavior, and function profiling.