Dr. Catherine Olschanowsky
Scientific applications are computationally intensive and require expensive HPC resources. Optimizing scientific applications requires that we balance three competing goals: Performance, Productivity, and Portability. Performance is important because it reduces time to solution and power consumption. However, optimization has the potential to negatively impact scientific productivity due to obfuscating the code. Portable code, code that can be moved to different computers, tends to be slow and difficult to maintain. We propose to automate optimization by using the Sparse Polyhedral Framework as a compiler intermediate representation. In this work, we present SPF-IE, a tool for translating scientific applications from legacy C/C++ code to our internal representation, and present a high-level overview of our internal representation.
Rift, Anna, "Optimizing Scientific Computations with the Sparse Polyhedral Framework" (2021). 2021 Undergraduate Research Showcase. 91.