A Novel Non-minimal/Minimal Turn Model for Highly Adaptive Routing for 2D and 3D NoCs

One of the current issues in NoC routing is the use of an acyclic Channel Dependency Graph (CDG) for ensuring freedom from deadlock. This requirement forces certain routing turns to be prohibited, thus reducing the degree of adaptiveness. In this paper, we propose a novel turn model for 2D and 3D mesh topologies which allows cycles in the CDG, provided the Extended Channel Dependency Graph (ECDG) remains acyclic. The end result is that the proposed turn model reduces the number of restrictions on routing turns and hence is able to provide path diversity through additional minimal and non-minimal routes between the source and destination. The average performance gain of the proposed method is up to 28% across all selected benchmarks compared with XY and 11% compared with other adaptive algorithms, for a 7× 7 mesh. Keywords—Networks on Chip, routing, turn models, deadlock freedom, non-minimal paths.