LIMCA: LLM for Automating Analog In-Memory Computing Architecture Design Exploration

TL;DR

The paper addresses the bottleneck of manually designing analog IMC crossbars for DNN acceleration by introducing LIMCA, a fine-tune-free, No-Human-In-The-Loop LLM-driven framework that automatically generates, validates, and expands SPICE netlists for IMC architectures. By leveraging a structured IMC dataset built on IMAC-Sim and an NHIL verification loop, LIMCA enables rapid, hardware-aware design space exploration that meets user-defined Power, Area, and Accuracy constraints. Experimental results on MNIST show LIMCA achieving $\geq 96\%$ accuracy with $\leq 3\,\text{W}$, and design exploration time reductions of approximately $11.5\times$ to $49.7\times$ compared to manual pipelines. The work delivers an open-source framework and dataset that accelerate IMC design, reduce reliance on expert iteration, and support scalable edge-oriented AI hardware development.

Abstract

Resistive crossbars enabling analog In-Memory Computing (IMC) have emerged as a promising architecture for Deep Neural Network (DNN) acceleration, offering high memory bandwidth and in-situ computation. However, the manual, knowledge-intensive design process and the lack of high-quality circuit netlists have significantly constrained design space exploration and optimization to behavioral system-level tools. In this work, we introduce LIMCA, a novel fine-tune-free Large Language Model (LLM)-driven framework for automating the design and evaluation of IMC crossbar architectures. Unlike traditional approaches, LIMCA employs a No-Human-In-Loop (NHIL) automated pipeline to generate and validate circuit netlists for SPICE simulations, eliminating manual intervention. LIMCA systematically explores the IMC design space by leveraging a structured dataset and LLM-based performance evaluation. Our experimental results on MNIST classification demonstrate that LIMCA successfully generates crossbar designs achieving $\geq$96% accuracy while maintaining a power consumption $\leq$3W, making this the first work in LLM-assisted IMC design space exploration. Compared to existing frameworks, LIMCA provides an automated, scalable, and hardware-aware solution, reducing design exploration time while ensuring user-constrained performance trade-offs.

Figures (4)

• Figure 1: An overview of proposed LIMCA framework.
• Figure 2: An analog IMC crossbar array pair (positive and negative arrays).
• Figure 3: LIMCA framework.
• Figure 4: Distribution of Power, Area, Accuracy across hardware parameters.