Resilience through Self-Configuration in Future Massively Defective Nanochips


Piotr Zajac, Jacques Henri Collet, Jean Arlat, Yves Crouzet

 

Abstract


This paper addresses the resilience challenges in the future nanochips made up of massively defective nanoelements and organized in a replicative multicore architecture. The main idea is to suggest that the chip should work with almost no external testing or control mechanisms, using a self-configuration methodology to ensure the resilience of operation even in the presence of a significant fraction of defective cores. By self–configuration, we mean self-diagnosis with mutual tests, self-shutdown of inactive cores and self-configuration of communications. We study the efficiency of the proposed methodology and show that the method is applicable up to a fraction of 30% of defective cores in the on-chip network.

Keywords: Nanoscale chips, muticore processor architectures, resilience, self-reconfiguration