Graduate Seminar / CS Candidate Talk

February 16, 2018 - 11:30 AM to 1:30 PM
Woodward 106

Dr. Maryam Mehri Dehnavi, Rutgers University

Title: Transforming Computation and Communication Patterns for High-performance

Abstract: The emergence of new applications and stupendously large matrices in
applications such as data mining and large-scale scientific simulations has
rendered the classical software frameworks and numerical methods inadequate
in many situations. In this talk, I will demonstrate my work on building
domain-specific compilers and on reformulating classical mathematical
methods to significantly improve the performance and scalability of
large-scale applications on modern computing platforms.

In the first part of the talk, I will present Sympiler, a domain-specific
code generator that transforms computation patterns in sparse matrix
methods for high-performance. I will demonstrate how decoupling symbolic
analysis from numerical manipulation will enable the optimization of sparse
codes with static sparsity patterns. The performance of Sympiler-generated
code will be compared to hand-optimize library codes to demonstrate the
effectiveness of symbolic decoupling. In the second part of the talk, I
will show that through mathematical reformulation, communication patterns
in classical optimization methods can be transformed. As a result, the
performance of the algorithm is inherently improved when executed on
distributed platforms leading to a speedup of up to 5-fold compared to the
classical formulation

Bio: 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 and a visiting scholar
at UC Berkeley. She received her Ph.D. in Electrical and Computer
Engineering from McGill University in 2013. Maryam is the recipient of the
NSF CRII, NSERC CGS, NSERC PDF awards and her research has received the
grand final prize of the 2017 ACM SRC competitions.