Updates on the campus response to coronavirus (COVID-19)

Ph.D. Proposal Oral Exam - Bahar Asgari

Event Details

Monday, October 21, 2019

9:00am - 11:00am

Room 2100, Klaus

For More Information


Event Details

Title:  Efficiently Accelerating Sparse Problems by Enabling Stream Accesses to Memory using Hardware/Software Techniques


Dr. Kim, Advisor 

Dr. Krishna, Chair

Dr. Mukhopadhyay


The objective of the proposed research is improving the performance of sparse problems that have a wide range of applications but still, suffer from serious challenges when running on modern computers. In summary, the challenges include the underutilization of concurrent compute engines because of the distribution of non-zero values in sparse data and the underutilization of available memory bandwidth because of dependencies in computation or slow mechanisms for decompressing the sparse data. Our key insight to address the aforementioned challenges is that based on the type of the problem, we can either modify the distribution of non-zero elements, transform the computations mathematically to extract more parallelism, or change their representations. By applying such techniques, while sustaining the nature of the problem, the execution adapts more effectively to given hardware resources. To this end, this proposal introduces hardware/software techniques to enable stream accesses to memory for accelerating the inference of deep neural networks (DNNs), iterative solvers of partial differential equations (PDEs), and graph algorithms. Based on our preliminary results, we propose to extend our approaches to real-time applications, namely the navigation algorithms in autonomous technologies.

Last revised October 15, 2019