Title: Scalable Energy-Efficient Microarchitectures with Computational Error Tolerance

Bobin Deng

Ph.D. Student

Schools of Computer Science

Georgia Institute of Technology

Date: Thursday, April 9th, 2020

Time: 9:00 am to 11:00 am (EDT)

Location: *No Physical Location*

BlueJeans: https://gatech.bluejeans.com/9026676698

Committee:

Dr. Thomas Conte (advisor), School of Computer Science, Georgia Institute of Technology

Dr. Hyesoon Kim, School of Computer Science, Georgia Institute of Technology

Dr. Alexandros Daglis, School of Computer Science, Georgia Institute of Technology

Dr. Jeanine Cook, Sandia National Laboratories

Abstract:

Due to the problems of high leakage current and threshold voltage, Dennard scaling has reached its limit on conventional semiconductor technology. Energy reduction at the transistor level by simply lowering supply voltage has proven to be infeasible for these devices (e.g., MOSFETs). Some recently proposed millivolt switch techniques are aimed at mitigating these issues, by maintaining high on/off ratio of drain currents with a much lower supply voltage. However, Vdd reduction is limited by high intermittent error probabilities in millivolt switches. Energy-efficient microarchitectures that are computationally error-tolerant are therefore urgently needed. 

This thesis systematically leverages the error detection and correction properties of Redundant Residue Number System (RRNS) by varying the number of non-redundant (n) and redundant (r) components (residues) within such a two-dimensional (n, r)-RRNS design plane. Being able to efficiently handle resilience in this (n, r)-RRNS plane significantly improves reliability, allowing further Vdd reduction to save energy. 

To this end, first, I will discuss the necessary implementation details of a single error correction RRNS core. Second, propose a scalable RRNS microarchitecture that simultaneously supports both, error-correction, as well as checkpointing with restart capabilities upon detecting uncorrectable errors. Third, design novel RRNS-based adaptive checkpointing&restart mechanisms that automatically guarantee reliability while minimizing the energy-delay product (EDP). Finally, I will explore the RRNS design space systematically to find the optimal (n, r) configuration point.