Tools
EvaCAM
EvaCAM is a circuit- and architecture-level modeling and evaluation tool designed for Content Addressable Memories (CAMs)—special-purpose in-memory computing units that enable parallel searches directly in memory. With growing interest in CAMs for data-intensive applications like machine learning and bioinformatics, the design space is expanding rapidly. Beyond traditional ternary CAMs (TCAMs), EvaCAM supports analog CAM (ACAM) and multi-bit CAM (MCAM) designs using various non-volatile memory (NVM) devices, offering benefits like higher density, better energy efficiency, and non-volatility. It also accommodates both exact and approximate match searches, broadening CAM utility for new applications. EvaCAM allows exploration of CAM array structures and sensing circuits, validated through HSPICE simulations and chip measurements. A detailed case study highlights its use in FeFET CAM design space exploration.
3D-NVSim
The design space of M3D for NMC/IMC fabrics is vast, covering novel materials, emerging logic and memory devices, 2D/3D interconnects, etc. To assess the value proposition of emerging memory/logic devices in M3D NMC/IMC fabrics, exploring the said design space is required but can be prohibitively expensive if detailed simulation and physical design are applied to all designs. Tools that can facilitate fast exploration and reductio of the design space with high fidelity are crucial. Our research project aims to develop solutions to address this need. Two proposed solutions include: (1) a compact model for back-end-of-line (BEOL) compatible oxide semiconductor (OS) Ferroelectric Field-Effect Transistors (FeFETs), and (2) 3D-NVSim, which can model two 3D SRAM folding strategies—3D bitline (3DBL) folding and 3D wordline (3DWL) folding.
CAMASIM
CAMASim is a comprehensive simulation framework for CAM-based accelerators, addressing the challenge of designing efficient hardware for memory-intensive search operations. Content Addressable Memory (CAM) excels at parallel computation in memory, but creating accelerators that balance accuracy, hardware cost, and flexibility for both exact and approximate searches remains complex amid CAM’s evolution across algorithms, architectures, circuits, and devices. CAMASim offers modularity, flexibility, and generality, providing a detailed design space for CAM-based accelerators. It includes automated functional simulation for accuracy and hardware performance prediction, leveraging a circuit-level CAM modeling tool. CAMASim streamlines design space exploration, empowering researchers to develop effective CAM-based solutions for diverse search-intensive applications.