OPTIMA: Design-Space Exploration of Discharge-Based In-SRAM Computing: Quantifying Energy-Accuracy Trade-Offs

TL;DR

OPTIMA is a modeling framework that enables fast and accurate design-space exploration of in-SRAM computing accelerators, optimizing energy and accuracy trade-offs while addressing circuit nonlinearities and variations.

Contribution

We introduce OPTIMA, a novel modeling framework that significantly accelerates and improves the design of energy-efficient in-SRAM accelerators with robust accuracy.

Findings

100x faster simulation speed for in-SRAM multipliers

Achieved 1.05pJ energy per 4-bit operation

Attained high classification accuracy on ImageNet and CIFAR-10 datasets

Abstract

In-SRAM computing promises energy efficiency, but circuit nonlinearities and PVT variations pose major challenges in designing robust accelerators. To address this, we introduce OPTIMA, a modeling framework that aids in analyzing bit-line discharge and power consumption in 6T-SRAM-based accelerators. It provides insights into limiting factors and enables fast design-space exploration of circuit configurations. Leveraging OPTIMA for in-SRAM multiplications exhibits ~100x simulation speed-up while maintaining an RMS modeling error of 0.88mV. Exploration yields an optimized multiplier with 1.05pJ energy consumption per 4-bit operation and classification accuracies of 71.8% (top-1) and 90.4% (top-5) for ImageNet and 92.5% for CIFAR-10 datasets respectively when applied in quantized DNNs. To further support research and development, we made our tool flow available open source at https://github.com/sevjaeg/optima.