FPGA Acceleration and Optimization of the CLUE Reconstruction Algorithm using SYCL and Alpaka

This paper investigates the FPGA acceleration of the CLUE clustering algorithm for the upgraded CMS High Granularity Caloriemeter (HGCAL). This work builds off a previous SYCL FPGA implementation by identifying the performance bottlenecks and introducing modifications to optimize the implementation. The main focus for optimization was lowering the initiation interval and latency. Additionally, this project adapts the CLUE algorithm to target FPGA execution through Alpaka. The results of the project reflect the potential to improve CLUE throughput through using FPGA hardware.