Search for: All records

Large language models (LLMs) have achieved high accuracy in diverse NLP and computer vision tasks due to self- attention mechanisms relying on GEMM and GEMV operations. However, scaling LLMs poses significant computational and energy challenges, particularly for traditional Von-Neumann architectures (CPUs/GPUs), which incur high latency and energy consumption from frequent data movement. These issues are even more pronounced in energy-constrained edge environments. While DRAM-based near-memory architectures offer improved energy efficiency and throughput, their processing elements are limited by strict area, power, and timing constraints. This work introduces CIDAN-3D, a novel Processing-in-Memory (PIM) architecture tailored for LLMs. It features an ultra-low-power Neuron Processing Element (NPE) with high compute density (#Operations/Area), enabling ecient in-situ execution of LLM operations by leveraging high parallelism within DRAM. CIDAN- 3D reduces data movement, improves locality, and achieves substantial gains in performance and energy efficiency—showing up to 1.3X higher throughput and 21.9X better energy efficiency for smaller models, and 3X throughput and 7X energy improvement for large decoder-only models compared to prior near-memory designs. As a result, CIDAN-3D offers a scalable, energy-efficient platform for LLM-driven Gen-AI applications. 

Transformers-based language models have achieved remarkable accuracy in various NLP tasks, employing self-attention mecha- nisms primarily based on matrix multiplication. However, their significant size leads to data movement issues, causing latency and energy efficiency challenges in conventional Von-Neumann systems. To mitigate these issues, several in-memory and near- memory architectures have been proposed. This paper introduces PACT-3D, a near-memory architecture featuring novel computing units integrated with DRAM banks. PACT-3D significantly reduces latency by 1.7× and improves energy efficiency by 18.7× compared to state-of-the-art near-memory architectures. 

This article presents TULIP, a new architecture for a variable precision quantized neural network (QNN) inference. It is designed with the goal of maximizing energy efficiency per classification. TULIP is constructed by arranging a collection of unique processing elements (TULIP-PEs) in a single-instruction–multiple-data (SIMD) fashion. Each TULIP-PE contains binary neurons that are interconnected using multiplexers. Each neuron also has a small dedicated local register connected to it. The binary neurons are implemented as standard cells and used for implementing threshold functions, i.e., an inner-product and thresholding operation on its binary inputs. The neurons can be reconfigured with a single change in the control signals to implement all the standard operations used in a QNN. This article presents novel algorithms for implementing the operations of a QNN on the TULIP-PEs in the form of a schedule of threshold functions. TULIP was implemented as an ASIC in TSMC 40nm-LP technology. A QNN accelerator that employs a conventional multiply and accumulate-based arithmetic processor was also implemented in the same technology to provide a fair comparison. The results show that TULIP is 30X−50X more energy-efficient than an equivalent design, without any penalty in performance, area, or accuracy. Furthermore, TULIP achieves these improvements without using traditional techniques such as voltage scaling or approximate computing. Finally, this article also demonstrates how the run-time tradeoff between accuracy and energy efficiency is done on the TULIP architecture. 

Graph Convolutional Networks (GCNs) have successfully incorporated deep learning to graph structures for social network analysis, bio-informatics, etc. The execution pattern of GCNs is a hybrid of graph processing and neural networks which poses unique and significant challenges for hardware implementation. Graph processing involves a large amount of irregular memory access with little computation whereas processing of neural networks involves a large number of operations with regular memory access. Existing graph processing and neural network accelerators are therefore inefficient for computing GCNs. This paper presents Parag, processing in memory (PIM) architecture for GCN computation. It consists of customized logic with minuscule computing units called Neural Processing Elements (NPEs) interfaced to each bank of the DRAM to support parallel graph processing and neural network computation. It utilizes the massive internal parallelism of DRAM to accelerate the GCN execution with high energy efficiency. Simulation results for inference of GCN over standard datasets show a latency and energy reduction by three orders of magnitude over a CPU implementation. When compared to a state-of-the-art PIM architecture, PARAG achieves on an average 4x reduction in latency and 4.23x reduction in the energy-delay-product (EDP). 

In this paper, we describe a design of a mixed-signal circuit for an binary neuron (a.k.a perceptron, threshold logic gate) and a methodology for automatically embedding such cells in ASICs. The binary neuron, referred to as an FTL (flash threshold logic) uses floating gate or flash transistors whose threshold voltages serve as a proxy for the weights of the neuron. Algorithms for mapping the weights to the flash transistor threshold voltages are presented. The threshold voltages are determined to maximize both the robustness of the cell and its speed. The performance, power, and area of a single FTL cell are shown to be significantly smaller (79.4%), consume less power (61.6%), and operate faster (40.3%) compared to conventional CMOS logic equivalents. Also included are the architecture and the algorithms to program the flash devices of an FTL. The FTL cells are implemented as standard cells, and are designed to allow commercial synthesis and P&R tools to automatically use them in synthesis of ASICs. Substantial reductions in area and power without sacrificing performance are demonstrated on several ASIC benchmarks by the automatic embedding of FTL cells. The paper also demonstrates how FTL cells can be used for fixing timing errors after fabrication.