With the growing demand for energy-efficient edge computing on resource-constrained FPGAs, LUT-based multipliers have been widely explored as alternatives to DSP-based designs. However, the FPGA implementation flow for evaluating large numbers of designs remains highly time-consuming and limits rapid deployment and exploration. Hence, an initial automated framework is proposed for rapid integration and evaluation of LUT-based multipliers in scalable systolic arrays, combining a two-stage supervised learning framework with RapidWright-based stitching. Preliminary results on a series of multiplier designs show up to 2.1× faster implementation and 5.9–23.8\% higher clock frequency.
Chen, R, Wang, X, da Silva, B & Stroobandt, D 2026, From LUT-Based Multipliers to Systolic Arrays: An Automated FPGA Implementation Flow. in 2026 36th International Conference on Field-Programmable Logic and Applications (FPL). Ghent University, pp. 372-373. https://doi.org/10.23919/FPLCPS00026.2026.00071
Chen, R., Wang, X., da Silva, B., & Stroobandt, D. (2026). From LUT-Based Multipliers to Systolic Arrays: An Automated FPGA Implementation Flow. In 2026 36th International Conference on Field-Programmable Logic and Applications (FPL) (pp. 372-373). Ghent University. https://doi.org/10.23919/FPLCPS00026.2026.00071
@inproceedings{d89b2004dbc64da8a3bb07f2a232d9c3,
title = "From LUT-Based Multipliers to Systolic Arrays: An Automated FPGA Implementation Flow",
abstract = "With the growing demand for energy-efficient edge computing on resource-constrained FPGAs, LUT-based multipliers have been widely explored as alternatives to DSP-based designs. However, the FPGA implementation flow for evaluating large numbers of designs remains highly time-consuming and limits rapid deployment and exploration. Hence, an initial automated framework is proposed for rapid integration and evaluation of LUT-based multipliers in scalable systolic arrays, combining a two-stage supervised learning framework with RapidWright-based stitching. Preliminary results on a series of multiplier designs show up to 2.1× faster implementation and 5.9–23.8\% higher clock frequency. ",
author = "Ruiqi Chen and Xiaoke Wang and \{da Silva\}, Bruno and Dirk Stroobandt",
year = "2026",
month = oct,
day = "27",
doi = "10.23919/FPLCPS00026.2026.00071",
language = "English",
isbn = "978-1-971299-32-7",
pages = "372--373",
booktitle = "2026 36th International Conference on Field-Programmable Logic and Applications (FPL)",
publisher = "Ghent University",
}