Michael  Gilbert

Michael Gilbert

Scholar Title

MIT EECS | Angle Undergraduate Research and Innovation Scholar

Research Title

LoopTree: Flexible Framework for Evaluation of Accelerators with Pipelined Dataflow

Cohort

2021–2022

Department

Electrical Engineering and Computer Science

Research Areas
  • Computer Architecture
Supervisor

Joel S. Emer

Abstract

In recent years, many frameworks have been proposed to enable fast design space exploration of deep neural network (DNN) accelerators. However, these frameworks are limited to evaluating accelerator designs supporting single-layer dataflows that serially process each DNN layer. Meanwhile, fused-layer dataflows efficiently reuse output activations stored on-chip by simultaneously processing multiple DNN layers. Thus, fused-layer dataflows can lead to significant reductions in off-chip data transfers, improving bandwidth usage efficiency and thus inference execution time.

To enable modeling of DNN accelerators that support fused-layer dataflows, this paper presents LoopTree. To concisely represent a fused-layer dataflow, LoopTree proposes a representation using a tree of loop nests. In addition, an evaluation framework is also presented. Given an architecture, a workload, and a mapping, LoopTree can generate prediction of important metrics such as latency, energy usage, compute count, memory capacity usage, and memory bandwidth usage.

Quote

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.

Back to Scholars