SHF: Small: Ensuring Reliability and Portability of Scientific Software for Heterogeneous Architectures
The results of this research will be tools and techniques to help scientists find bugs more effectively in such programs. This research has important implications for the reliability of important scientific programs such as those used in biomedical imaging applications, climate modelling, and vehicle design. This project develops rigorous methods for analyzing parallel scientific code, specifically written using the now emerging OpenCL parallel programming standard. The goal is to detect potential sources of reliability and portability deficiencies in such code that are due to dependencies of the floating-point behavior on the underlying hardware, which may be unknown to the programmer. Traditional reliability methods such as program testing and debugging are ineffective for parallel OpenCL programs, because program behavior may vary across runs, making after-test behavior uncertain.
Northeastern University’s College of Engineering is home to numerous federally-funded research centers and an array of leading-edge projects and initiatives that advance discovery and new knowledge in health, sustainability, and security.