Run-Time Deadlock Detection

Abstract

Relentless technology downscaling provides a promising prospect to realize heterogeneous system-on-chip (SoC) and homogeneous chip-multiprocessor (CMP) based on the networks-on-chip (NoCs) paradigm with augmented scalability, modularity and performance. In many cases in such systems, scheduling and managing communication resources are the major design and implementation challenges instead of the computing resources. Moreover, adaptive packets routing in such systems are susceptible to routing deadlock, which could lead to performance degradation or system failure. Past research efforts were mainly focused on complex design-time approaches to avoid deadlock or simple heuristic run-time approaches to detect and recover from deadlock. The later, however, consider only local or partial information about the network which may produce substantial false detections, especially with the network close to saturation where blocked packets could be fagged as deadlocked.This chapter studies deadlock detection and recovery strategy in NoCs as opposed to deadlock avoidance. It presents a deadlock detection method that utilizes run-time transitive-closure (TC) computation to discover the existence of deadlock-equivalence sets, which imply loops of requests in NoCs. This detection scheme guarantees the discovery of all true-deadlocks without false alarms in contrast with state-of-the-art approximation and heuristic methods. A distributed TC-network architecture, which couples with the network-on-chip (NoC) infrastructure, is also presented to realize the detection mechanism efficiently. Detailed hardware realization architectures and schematics are also discussed. The captured results based on a cycle-accurate simulator demonstrate the effectiveness of the proposed method. It drastically outperforms timing-based deadlock detection mechanisms by eliminating false detections and, thus, reducing energy wastage in retransmission for various traffic scenarios including real-world application.