Comparison of 2D Mesh and Reconfigurabl Mesh Topologies for Network on Chip Design

• Published in: 2020 5th International Conference on Computer Science and Engineering (UBMK) pp. 136 - 140
• Main Authors: Kullu, Pinar, Tosun, Suleyman
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2020
• Subjects: 2D mesh; Bandwidth; energy; Global Positioning System; Hafnium compounds; Network topology; Network-on-chip; Network-on-Chip (NoC); reconfg- urable mesh; Topology; Two dimensional displays
• DOI: 10.1109/UBMK50275.2020.9219423

The most commonly used topology type in Network-on-Chip (NoC) design is two dimensional (2D) mesh topology since it easy to design and fabricate. Furthermore, it can be reused for different applications and is fault-tolerant as it provides more than one path between each pair of communicating nodes. On the other hand, a mesh topology generally is optimized for a single application during the design process, which only allows this specific application to run on the architecture. 2D reconfigurable mesh topology gives an opportunity to integrate multiple applications simultaneously. A reconfigurable mesh is generated by placing configuration switches between the routers of a classical mesh. When all application nodes are mapped to the structure, the switches can be configured based on the design needs such as meeting some constraints and optimizing some design parameters. Although reconfigurable meshes have some advantages over classical 2D meshes, they have disadvantages as well such as extra hardware cost. In this study, we compare and contrast classical 2D mesh topology and reconfigurable mesh topology in terms of energy consumption, bandwidth requirements, and speed (i.e. frequencies) of running applications. We testes both topology types on several randomly generated application graphs.