English
Related papers

Related papers: Scale-out Systolic Arrays

200 papers

As AI-based applications become pervasive, CPU vendors are starting to incorporate matrix engines within the datapath to boost efficiency. Systolic arrays have been the premier architectural choice as matrix engines in offload accelerators.…

Hardware Architecture · Computer Science 2021-10-06 Geonhwa Jeong , Eric Qin , Ananda Samajdar , Christopher J. Hughes , Sreenivas Subramoney , Hyesoon Kim , Tushar Krishna

Transformer models rely heavily on the scaled dot-product attention (SDPA) operation, typically implemented as FlashAttention. Characterized by its frequent interleaving of matrix multiplications and softmax operations, FlashAttention fails…

Hardware Architecture · Computer Science 2025-12-09 Jiawei Lin , Yuanlong Li , Guokai Chen , Thomas Bourgeat

Benefiting from the self-attention mechanism, Transformer models have attained impressive contextual comprehension capabilities for lengthy texts. The requirements of high-throughput inference arise as the large language models (LLMs)…

Hardware Architecture · Computer Science 2024-07-16 Huizheng Wang , Jiahao Fang , Xinru Tang , Zhiheng Yue , Jinxi Li , Yubin Qin , Sihan Guan , Qize Yang , Yang Wang , Chao Li , Yang Hu , Shouyi Yin

Modern Deep Neural Network (DNN) accelerators are equipped with increasingly larger on-chip buffers to provide more opportunities to alleviate the increasingly severe DRAM bandwidth pressure. However, most existing research on buffer…

Hardware Architecture · Computer Science 2025-01-23 Jingwei Cai , Xuan Wang , Mingyu Gao , Sen Peng , Zijian Zhu , Yuchen Wei , Zuotong Wu , Kaisheng Ma

Both efficient neural networks and hardware accelerators are being explored to speed up DNN inference on edge devices. For example, MobileNet uses depthwise separable convolution to achieve much lower latency, while systolic arrays provide…

Hardware Architecture · Computer Science 2021-05-31 Surya Selvam , Vinod Ganesan , Pratyush Kumar

Modern hardware architectures for Convolutional Neural Networks (CNNs), other than targeting high performance, aim at dissipating limited energy. Reducing the data movement cost between the computing cores and the memory is a way to…

Hardware Architecture · Computer Science 2025-01-15 Cristian Sestito , Shady Agwa , Themis Prodromakis

Transformer-based models are becoming more and more intelligent and are revolutionizing a wide range of human tasks. To support their deployment, AI labs offer inference services that consume hundreds of GWh of energy annually and charge…

Systems and Control · Electrical Eng. & Systems 2025-08-29 Ching-Yi Lin , Sahil Shah

As deep neural networks develop significantly more diverse and complex, achieving high performance and efficiency on complicated DNN models faces pressing challenges. Modern DNN workloads are increasingly diverse in operation types, tensor…

Hardware Architecture · Computer Science 2026-05-25 Xingzhen Chen , Zhuoping Yang , Jinming Zhuang , Shixin Ji , Sarah Schultz , Zheng Dong , Weisong Shi , Peipei Zhou

The combination of Winograd's algorithm and systolic array architecture has demonstrated the capability of improving DSP efficiency in accelerating convolutional neural networks (CNNs) on FPGA platforms. However, handling arbitrary…

Hardware Architecture · Computer Science 2021-07-12 Xinheng Liu , Yao Chen , Cong Hao , Ashutosh Dhar , Deming Chen

Deep Neural Network (DNN) accelerators are extensively used to improve the computational efficiency of DNNs, but are prone to faults through Single-Event Upsets (SEUs). In this work, we present an in-depth analysis of the impact of SEUs on…

Hardware Architecture · Computer Science 2024-05-27 Naïn Jonckers , Toon Vinck , Gert Dekkers , Peter Karsmakers , Jeffrey Prinzie

Massively parallel systolic arrays and resource-efficient depthwise separable convolutions are two promising techniques to accelerate DNN inference on the edge. Interestingly, their combination is inefficient: Computational patterns of…

Hardware Architecture · Computer Science 2021-08-27 Vinod Ganesan , Pratyush Kumar

The number of processing elements (PEs) in a fixed-sized systolic accelerator is well matched for large and compute-bound DNNs; whereas, memory-bound DNNs suffer from PE underutilization and fail to achieve peak performance and energy…

Signal Processing · Electrical Eng. & Systems 2020-06-29 Nandan Kumar Jha , Shreyas Ravishankar , Sparsh Mittal , Arvind Kaushik , Dipan Mandal , Mahesh Chandra

Reduced-precision and variable-precision multiply-accumulate (MAC) operations provide opportunities to significantly improve energy efficiency and throughput of DNN accelerators with no/limited algorithmic performance loss, paving a way…

Distributed, Parallel, and Cluster Computing · Computer Science 2022-01-20 Ehab M. Ibrahim , Linyan Mei , Marian Verhelst

Edge inference for large language models (LLM) offers secure, low-latency, and cost-effective inference solutions. We emphasize that an edge accelerator should achieve high area efficiency and minimize external memory access (EMA) during…

Hardware Architecture · Computer Science 2025-07-15 Chun-Ting Chen , HanGyeol Mun , Jian Meng , Mohamed S. Abdelfattah , Jae-sun Seo

Deep neural networks (DNN) have become significant applications in both cloud-server and edge devices. Meanwhile, the growing number of DNNs on those platforms raises the need to execute multiple DNNs on the same device. This paper proposes…

Hardware Architecture · Computer Science 2023-02-22 Midia Reshadi , David Gregg

The evolution of IoT based smart applications demand porting of artificial intelligence algorithms to the edge computing devices. CNNs form a large part of these AI algorithms. Systolic array based CNN acceleration is being widely advocated…

Distributed, Parallel, and Cluster Computing · Computer Science 2019-02-11 Hazoor Ahmad , Muhammad Tanvir , Muhammad Abdullah Hanif , Muhammad Usama Javed , Rehan Hafiz , Muhammad Shafique

The NP-complete combinatorial optimization task of assigning offsets to a set of buffers with known sizes and lifetimes so as to minimize total memory usage is called dynamic storage allocation (DSA). Existing DSA implementations bypass the…

Due to their growing popularity and computational cost, deep neural networks (DNNs) are being targeted for hardware acceleration. A popular architecture for DNN acceleration, adopted by the Google Tensor Processing Unit (TPU), utilizes a…

Machine Learning · Computer Science 2018-02-20 Jeff Zhang , Tianyu Gu , Kanad Basu , Siddharth Garg

As artificial intelligence (AI) and machine learning (ML) technologies disrupt a wide range of industries, cloud datacenters face ever-increasing demand in inference workloads. However, conventional CPU-based servers cannot handle excessive…

Hardware Architecture · Computer Science 2022-06-08 Jung-Hoon Kim , Sungyeob Yoo , Seungjae Moon , Joo-Young Kim

Systolic arrays are a prominent choice for deep neural network (DNN) accelerators because they offer parallelism and efficient data reuse. Improving the reliability of DNN accelerators is crucial as hardware faults can degrade the accuracy…

Hardware Architecture · Computer Science 2024-02-13 Wei-Kai Liu