Dr. Maryam Mehri Dehnavi,
CoRE Building Lecture Hall
Abstract: Computational scientists often encounter problems requiring the solution to large sparse systems. To enable fast and accurate simulations, the execution time of these solvers needs to be accelerated on modern high-performance architectures. In this talk, I will show that through mathematical reformulation and algorithm re-engineering, memory access patterns in iterative sparse solvers can be restructured to significantly improve their performance on parallel architectures. I will also discuss our recent advances in building domain-specific compilers for sparse matrix computations to automatically generate fast code for sparse numerical methods
Bio: Dr. Maryam Mehri Dehnavi is an Assistant Professor in the Electrical and Computer Engineering Department at Rutgers University. Her research focuses on high-performance computing and domain-specific compiler design. Previously she was a postdoctoral researcher at MIT with the Supertech research group. She received her Ph.D. in Electrical and Computer Engineering from McGill University in 2012.