Advanced Routing Algorithms for General Purpose Photonic Processors

TL;DR

This paper introduces new bidirectional and depth-first search algorithms for efficient routing and cycle detection in programmable photonic processors, improving dynamic path allocation and fault tolerance.

Contribution

It presents novel path and cycle finding algorithms tailored for waveguide mesh architectures, optimizing routing based on various performance metrics and handling malfunctioning units.

Findings

Algorithms demonstrate faster routing in photonic networks.

Effective cycle detection for dynamic path allocation.

Comparison shows improved efficiency over existing methods.

Abstract

Cost-effective and programmable photonic-driven solutions like electronic counterparts (FPGAs) can be implemented using waveguide mesh architectures along with tunable couplers for routing to implement general-purpose photonic processors. These processors/ networks are represented using undirected weighted graphs, where weights are included to implement constraints in the routing. Faster automated routing and cycle finding algorithms are crucial for dynamic path allocations in live networks to implement various functionalities using these processors. We propose path and cycle finding algorithms based on bidirectional and depth-first search techniques, considering various performance metrics for each device to optimize the path according to the required metric. Multiple cases of path distribution and implementation of cycles of various sizes have been demonstrated. Various methods to eliminate the non-functioning or malfunctioning units are proposed. The broad applicability of the proposed path-finding algorithm has been demonstrated using the same algorithm to create a list of all the possible input-output combinations in a 4*4 photonic switching network. A comparison of available search algorithms in terms of execution time and complexity has been described.