Route packets, not wires: on-chip inteconnection networks
Proceedings of the 38th annual Design Automation Conference
Boosting the Performance of Myrinet Networks
IEEE Transactions on Parallel and Distributed Systems
Adaptive Bubble Router: A Design to Improve Performance in Torus Networks
ICPP '99 Proceedings of the 1999 International Conference on Parallel Processing
Principles and Practices of Interconnection Networks
Principles and Practices of Interconnection Networks
An architecture and compiler for scalable on-chip communication
IEEE Transactions on Very Large Scale Integration (VLSI) Systems
Hi-index | 0.00 |
In the case of simple tile-based architecture, such as small reconfigurable processor arrays, a virtual-channel mechanism, which requires additional logic and pipeline stages, will be one of the crucial factors for a low cost implementation of their on-chip routers. To guarantee deadlock-free packet transfer with no virtual channels on tori, we propose a non-minimal strategy consistent with the rule of dimension-order routing (DOR) algorithm. Since embedded streaming applications usually generate predictable data traffic, the path set can be customized to the traffic from alternative DOR paths. Although the proposed strategy does not use any virtual channels, it achieves almost the same performance as virtual-channel routers on tori in eleven of 18 application traces.