Methodology for adapting on-chip interconnect architectures

Authors

    Authors

    S. Suboh; V. Narayana; M. Bakhouya; J. Gaber;T. El-Ghazawi

    Comments

    Authors: contact us about adding a copy of your work at STARS@ucf.edu

    Abbreviated Journal Title

    IET Comput. Digit. Tech.

    Keywords

    computer architecture; network-on-chip; quality of service; adapting; onchip interconnect architecture methodology; Network-on-Chip; NoC; scalability problem; bus based system-on-chip; quality of service; network infrastructure; predictable electrical parameters; programmable; infrastructure; traffic pattern application; NETWORK; SYSTEMS; Computer Science, Hardware & Architecture; Computer Science, Theory &; Methods

    Abstract

    Network-on-chip (NoC) has been proposed to solve the scalability problem experienced in bus-based system-on-chip. The main challenge is the ability to predict the quality of service that the network infrastructure provides while meeting other system constraints, namely power and area. Although these architectures are regular with predictable electrical parameters, they may suffer from higher latency and lower throughput. To tackle this issue, the network structure needs to be adaptable in response to the needs of the application. This paper presents a methodology for augmenting an NoC with a programmable infrastructure that allows application-specific adaptation. Based on the developed infrastructure, an algorithm is also presented for static adaptation based on application traffic patterns. To evaluate the proposed methodology of the adaptable NoC, the WK-recursive on-chip interconnect is used as a case study. Simulations are conducted and reported results demonstrate the usefulness of the proposed approach.

    Journal Title

    Iet Computers and Digital Techniques

    Volume

    8

    Issue/Number

    3

    Publication Date

    1-1-2014

    Document Type

    Article

    Language

    English

    First Page

    109

    Last Page

    117

    WOS Identifier

    WOS:000337941800001

    ISSN

    1751-8601

    Share

    COinS