Application performance is bound by the compiler's ability to effectively map a program onto a target architecture. Benchmarking and performance analysis allow computer engineers to make the most effective use of the underlying hardware and operating system software the application at hand. In this project, students will develop benchmarking and workload analysis skills. Applications will be targeted for both uniprocessor and multiprocessor systems. Students will also be introduced to using hardware-based system instrumentation.

Possible research projects students will be working on include the following:

Benchmarking Uniprocessor and Multiprocessor Systems
     In this project, students will help to run and evaluate the performance of CenSSIS applications and traditional benchmarks, as on uniprocessor and multiprocessor systems.

Profiling Using Hardware-based Counters
     This project will introduce students to the concept of capturing runtime profiles using hardware event counters. Students will be introduced to the PAPI toolset in this project.

Performance Middleware
     In this project, students will evaluate different counter-based profiling methods to better understand cluster-based workloads. A number of middleware profiling platforms will be used, including Paradyne and HPCview.