Research Project Title:
SparseLoop: Tool for Searching Optimal Mapping of Sparse DNN Workload to Accelerator
abstract:The increase of demand for computing in the artificial intelligence field led to a search for deep neural network accelerators. However, evaluating an accelerator design is a time-consuming and expensive process. My project aims to make modelling of tensor accelerators easier by building upon SparseLoop, a tool for performance modeling of sparse tensor accelerators running various sparse workloads. My project will investigate how to analytically/statistically represent sparsity within a workload, how it affects the search for optimal data mapping, and look for improved search algorithms that can find the optimal data mapping faster.
Through this SuperUROP, I want to gain experience in research in computer architecture. I've taken 6.823 (computer systems architecture) and 6.825 (hardware architecture for deep learning) and I want to dive deeper into computer architecture research and its application in deep learning. I also want to learn how to work on a successful research project.