SIParCS 2020 - Shivani Singh

Evaluating Two Approaches to Automated Code Refactoring

Recorded Talk

SAMURAI code is used in CISL(Computational and Information Systems Lab) ASAP(Application Scalability and Performance) group, for data assimilation on observational data . Its first release, the CSU-Original version was CPU only code with compiler optimization. On implementing the optimization techniques and enabling the code for GPU, the code runtime has improved considerably. The timing analysis on the current optimized version of the SAMURAI code shows that there is still a lot of scope for further optimization. We evaluated two techniques for this project – Polyhedral Optimization & CUDA Programming. Polyhedral Optimization is a method for automatically optimizing and parallelizing sequential programs through static analysis. CUDA is a parallel computing platform and programming model developed by NVidia for general computing on its own GPUs.

Mentors: John Dennis, Brian Dobbins

Slides