| Title |
Analysis of Performance and Resource Utilization for FPGA-Based QR Decomposition Architectures |
| Authors |
윤석재(Suk-Jae Yoon) ; 유호영(Hoyoung Yoo) |
| DOI |
https://doi.org/10.5573/ieie.2026.63.8.32 |
| Keywords |
FPGA; QR decomposition; Gram-Schmidt; Householder transformation; Givens rotation |
| Abstract |
This paper proposes FPGA-based architectures for real-time processing of QR decomposition, a critical component in high-speed computing systems such as MIMO communications and radar signal processing, and presents a comparative analysis of performance and resource efficiency across major algorithms. In this research, four representative QR decomposition algorithms?CGS(Classical Gram-Schmidt), MGS(Modified Gram-Schmidt), HT(Householder Transformation), and CORDIC-based GR(Givens Rotation)?were implemented in hardware. Each architecture was designed based on the standard computational structure of the respective algorithm, and implementation and verification were performed targeting a Kintex UltraScale FPGA device. Through experimental evaluations, hardware resource utilization and computational latency were measured for each algorithm. Furthermore, numerical stability during hardware implementation was assessed by analyzing orthogonality errors in the computation results. The experimental results show that the Householder-based architecture offers superior numerical precision and high processing performance, while the CORDIC-based architecture enables resource-efficient implementation with relatively lower resource utilization. The comparative analysis provided in this paper can serve as a valuable reference for selecting and designing optimal QR decomposition algorithms tailored to specific system requirements and constraints. |