DescriptionThis project pursues extremely energy-efficient and high-throughput photonic interconnected computing systems enabled by scalable and modular integration of 2.5D or 3D silicon photonic (SiP) interposers with all-to-all, contention-less, and arbitration-free wavelength routing by cyclic Arrayed-Waveguide-Grating-Routers (AWGRs) [LIONS: Low-Latency Interconnect Optical Network Switches]. First, this paper introduces the working principle of the all-to-all photonic interconnect technology and show how it can be used to implement a multi-chip computing board. Second, the paper evaluates performance of the system under various workloads in terms of execution time and energy per transaction via simulations. Third, the latest results in terms of device fabrication and testing of SiP AWGR photonic routers are presented. Finally, the paper discusses modular and hierarchical scaling of the all-to-all photonic interconnects utilizing AWGRs with practical numbers of port counts.