Efficient SIMD computations on FPGA: Architectures, design techniques, and applications
Year: 2025
Type: article
Abstract: Field-Programmable Gate Arrays (FPGAs) provide a flexible and efficient platform for implementing Single Instruction, Multiple Data (SIMD) computations, offering advantages over traditional CPUs and GPUs through customizable architectures. This article explores the design considerations, optimization techniques, and practical applications of SIMD operations on FPGAs. We examine how vector processing units, specialized memory organizations, and interconnect architectures can be tailored to application requirements, while investigating methods for datapath optimization, memory access enhancement, and pipeline efficiency. The discussion extends to real-world FPGA-based SIMD applications in digital signal processing, machine learning acceleration, and image/video processing, highlighting how the reconfigurable nature of FPGAs enables performance and energy efficiency improvements for data-parallel workloads across various domains.
Author: Yadav, Gaurav
Institution: University of Southern California
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7083333391
FWCI:
Cites:
Cited by:
Related to: