A Test Suite to Support the Optimization of Scientific Applications

This work creates a set of concise test cases that accurately mimic the structure and workflow of scientific applications. Scientific applications often use layers of nested function pointers. Function pointers are a programming construct that can be very useful, but prevent code optimization. Code optimization is important for scientific applications because they are resource intensive and power hungry. Code optimization reduces execution time resulting in less energy consumption and more scientific progress. The set of test cases will be used to exercise optimization techniques. The codebases of scientific applications are large and do not lend themselves to easy testing of optimization strategies. To allow easy testing of optimization strategies, we create test cases. These test cases are small self-contained versions of scientific applications. The manageable size of these test cases allows effective testing of optimization strategies concerning function pointers. Effective testing allows creating new optimizations. These new optimizations reduce execution time and energy consumption of the application