PDNOC: An Efficient Partially Diagonal Network-on-Chip Design




Thomas Canhao Xu, Ville Leppänen, Pasi Liljeberg, Juha Plosila, Hannu Tenhunen

Roman Wyrzykowski, Jack Dongarra, Konrad Karczewski, Jerzy Waśniewski

Parallel Processing and Applied Mathematics

PublisherSpringer

2014

Lecture Notes in Computer Science

Parallel Processing and Applied Mathematics

LNCS

8384

513

522

10

978-3-642-55223-6

0302-9743

DOIhttps://doi.org/10.1007/978-3-642-55224-3_48



With the constantly increasing of number of cores in multicore processors, more emphasis should be paid to the on-chip interconnect. Performance and power consumption of an on-chip interconnect are directly affected by the network topology. The efficiency can also be optimized by proper mapping of applications. Therefore in this paper, we propose a novel Partially Diagonal Network-on-Chip (PDNOC) design that takes advantage of both heterogeneous network topology and congestionaware application mapping. We analyse the partially diagonal network in terms of area usage, power consumption, routing algorithm and implementation complexity. The key insight that enables the PDNOC is that most communication patterns in real-world applications are hot-spot and bursty. We implement a full system simulation environment using SPLASH-2 benchmarks. Evaluation results shown that, the proposed PDNOC provides up to 25% improvement in execution time over concentrated mesh, and 3.6x better energy delay product over fully connected diagonal network.



Last updated on 2024-26-11 at 20:45