Memristive-Based Associative Memory for Approximate Computational Reuse

Abstract

Listen

Translate article

Multimedia applications running on thousands of deep and wide pipelines working concurrently in GPUs have been an important target for power minimization both at the architectural and algorithmic levels. At the hardware level, energy efficiency techniques that employ voltage overscaling face a barrier so-called “path walls”: reducing operating voltage beyond a certain point generates massive number of timing errors that are impractical to tolerate. We propose an architectural innovation, called A\(^2\)M\(^2\) module (approximate associative memristive memory) that exhibits few tolerable timing errors suitable for GPU applications under voltage overscaling. A\(^2\)M\(^2\) is integrated with every floating point unit (FPU), and performs partial functionality of the associated FPU by pre-storing high frequency patterns for computational reuse that avoids overhead due to re-execution. Voltage overscaled A\(^2\)M\(^2\) is designed to match an input search pattern with any of the stored patterns within a Hamming distance range of 0–2. This matching behavior under voltage overscaling leads to a controllable approximate computing for multimedia applications. Our experimental results for the AMD Southern Islands GPU show that four image processing kernels tolerate the mismatches during pattern matching resulting in a PSNR \(\ge \) 30 dB. The A\(^2\)M\(^2\) module with 8-row enables 28% voltage overscaling in 45 nm technology resulting in 32% average energy saving for the kernels, while delivering an acceptable quality of service. This chapter provides a method for accepting errors in GPUs .