As chip fabrication has advanced to the nano level, the increased link density has heightened the risk of failures. The potential performance drawbacks resulting from these link failures have become a critical challenge in the design of reliable network-on-chip (NoC) systems. Fault-tolerant routing algorithms have proven to be effective strategies for handling this issue by diverting packets away from failed links to prevent congestion. However, these algorithms often result in excessive packet diversion, especially in the presence of a higher failure rate, which can significantly constrain the network’s behavior. This article introduces a novel NoC design with reconfigurable multifunctional channels (RMC_NoC). This design dynamically adapts the channel functions in response to network conditions to ensure that packets from failed links follow their original paths. In addition, it presents a channel buffer bubble flow control mechanism that can resolve congestion by redistributing congested traffic within the channel buffer. The evaluation results demonstrate that our approach ensures superior network communication even in the presence of permanent link failures, with minimal area overhead and power consumption. Moreover, our system exhibits lower latency and higher throughput compared to state-of-the-art fault-tolerant methods across various link failure rates. Notably, even at a severe failure rate of 30%, RMC_NoC exhibits only a 16.3% increase in latency compared to an ideal failure-free environment (Baseline) while still maintaining system communication capabilities to a considerable extent.

As technology feature sizes diminish to the nanometer regime, the leakage power crisis has become a major challenge in network-on-chip (NoC) design. Power gating (PG) is used to mitigate growing leakage power as an effective static power-saving technique. Applying PG in a multiple NoC (Multi-NoC) rather than a traditional NoC is a promising solution. However, limited by the channel width of the subnets, the increase in packet length will bring a severe serialization issue and performance loss. Previous Multi-NoC schemes have to wake up more subnets to minimize the performance loss, which also sacrifices their energy efficiency. In this article, we introduce an architecture, namely, BandExp , which allows subnets to expand their bandwidth by utilizing the idle physical links of other subnets. More bandwidth helps subnets mitigate the serialization issue and reduce the performance loss. Meanwhile, other subnets gain longer sleep cycles and thus save more energy. Evaluation results indicate that compared to the state-of-the-art Catnap , the proposed architecture reduces the average packet latency and execution time of different benchmarks by 19.3% and 3.2%, respectively. Also, the net static energy of the network is reduced by 23.2% on average, while the incurred area overhead is only 1.3%.

Chiplet-based system-on-chip (SoC) architectures, leveraging 2.5-D/3-D integration technologies, provide scalable solutions for a wide range of applications. Achieving high performance and cost-effectiveness in these systems relies heavily on optimizing die-to-die interconnect topologies and designs, which are essential for seamless interchiplet communication. This article introduces a reconfigurable hybrid topology (RHT) architecture designed for chiplet-based multicore systems. RHT achieves high performance and energy efficiency by dynamically reconfiguring the network topology to traffic variations, adaptively selecting transport subnets, and optimizing link bandwidth allocation, thereby minimizing congestion and maximizing packet throughput. Furthermore, RHT leverages global traffic information to dynamically combine Torus loops, maximizing opportunities for rapid packet transmission delivery while guaranteeing minimal hop counts. Moreover, RHT accelerates packet transmission via bufferless combined loops, extending the continuous sleeping periods of routers, improves power gating efficiency, and significantly reduces static power consumption. Simulation results indicate that the Mesh-DyRing achieves over a 40% reduction in network latency and more than a 20% decrease in power consumption overhead compared to the baseline design. When compared to WiNoC, an advanced hybrid wired-wireless topology design, the Mesh-DyRing-PG configuration reduces power consumption by 56.2% while maintaining equivalent average network latency.

The integrated architecture that features both in-memory logic and host processors, or so-called “processing-in-memory” (PIM) architecture, is an emerging and promising solution to bridge the performance gap between the memory and host processors. In spite of the considerable potential of PIM, the workload offloading policy, which partitions the program and determines where code snippets are executed, is still a main challenge in PIM. In order to determine the best PIM offloading partitions, existing methods require in-depth program profiling to create the control flow graph (CFG) and then transform it into a graph-cut problem. These CFG-based solutions depend on detailed profiling of a crucial element, the execution time of basic blocks, to accurately assess the benefits of PIM offloading. The issue is that these execution times can change significantly in PIM, leading to inaccurate offloading decisions. To tackle this challenge, we present a novel PIM workload offloading framework called “RDPIM” for CPU-PIM graph processing accelerators, which systematically considers the variations in the execution time of basic blocks. By analyzing the relationship between data dependencies among workloads and the connectivity of input graphs, we identified three key features that can lead to variations in execution time. We developed a novel reuse distance (RD)-based model to predict the exact performance of basic blocks for optimal offloading decisions. We evaluate RDPIM using real-world graphs and compare it with some state-of-the-art PIM offloading approaches. Experiments have demonstrated that our method achieves an average speedup of $2\times $ compared to CPU-only executions and up to $1.6\times $ compared to state-of-the-art PIM offloading schemes.