In-Memory Mirroring: Cloning Without Reading

TL;DR

This paper introduces in-memory mirroring (IMM), a novel technique that eliminates read and write-back steps in in-memory computing, significantly improving energy efficiency and doubling performance by enabling parallel data cloning within memory.

Contribution

IMM provides a new method for data movement in IMC architectures that avoids energy and performance penalties associated with traditional read/write operations.

Findings

IMM achieves 2× performance improvement over conventional methods.

IMM reduces energy consumption by avoiding read/write penalties.

Parallel cloning of entire rows is performed with O(1) complexity.

Abstract

In-memory computing (IMC) has gained significant attention recently as it attempts to reduce the impact of memory bottlenecks. Numerous schemes for digital IMC are presented in the literature, focusing on logic operations. Often, an application's description has data dependencies that must be resolved. Contemporary IMC architectures perform read followed by write operations for this purpose, which results in performance and energy penalties. To solve this fundamental problem, this paper presents in-memory mirroring (IMM). IMM eliminates the need for read and write-back steps, thus avoiding energy and performance penalties. Instead, we perform data movement within memory, involving row-wise and column-wise data transfers. Additionally, the IMM scheme enables parallel cloning of entire row (word) with a complexity of $\mathcal{O}(1)$. Moreover, our analysis of the energy consumption of the proposed technique using resistive random-access memory crossbar and experimentally validated JART VCM v1b model. The IMM increases energy efficiency and shows 2$\times$ performance improvement compared to conventional data movement methods.