PUMA: A Programmable Ultra-efficient Memristor-based Accelerator for Machine Learning Inference

TL;DR

PUMA is a novel memristor-based accelerator that combines in-memory analog computing with general-purpose execution units, enabling efficient and programmable machine learning inference across diverse applications.

Contribution

The paper introduces PUMA, a programmable memristor-based accelerator with a specialized ISA and compiler, expanding memristor crossbar capabilities to a wide range of ML workloads.

Findings

Achieves up to 2446x energy efficiency improvement over GPUs

Attains 66x latency reduction compared to GPUs

Maintains similar energy and latency as application-specific memristor accelerators

Abstract

Memristor crossbars are circuits capable of performing analog matrix-vector multiplications, overcoming the fundamental energy efficiency limitations of digital logic. They have been shown to be effective in special-purpose accelerators for a limited set of neural network applications. We present the Programmable Ultra-efficient Memristor-based Accelerator (PUMA) which enhances memristor crossbars with general purpose execution units to enable the acceleration of a wide variety of Machine Learning (ML) inference workloads. PUMA's microarchitecture techniques exposed through a specialized Instruction Set Architecture (ISA) retain the efficiency of in-memory computing and analog circuitry, without compromising programmability. We also present the PUMA compiler which translates high-level code to PUMA ISA. The compiler partitions the computational graph and optimizes instruction scheduling and register allocation to generate code for large and complex workloads to run on thousands of spatial cores. We have developed a detailed architecture simulator that incorporates the functionality, timing, and power models of PUMA's components to evaluate performance and energy consumption. A PUMA accelerator running at 1 GHz can reach area and power efficiency of $577~GOPS/s/mm^2$ and $837~GOPS/s/W$, respectively. Our evaluation of diverse ML applications from image recognition, machine translation, and language modelling (5M-800M synapses) shows that PUMA achieves up to $2,446\times$ energy and $66\times$ latency improvement for inference compared to state-of-the-art GPUs. Compared to an application-specific memristor-based accelerator, PUMA incurs small energy overheads at similar inference latency and added programmability.