Search

A three-dimensional Network-on-Chip (3D NoC) equips modern multicore processors with good scalability, a small area, and high performance using vertical through-silicon vias (TSV). However, the failure rate of TSV, which is higher than that of horizontal links, causes unpredictable topology variations and requires adaptive routing algorithms to select the available paths dynamically. Most works have aimed at the congestion control for TSV partially 3D NoCs to bypass the TSV reliability issue, while others have focused on the fault tolerance in TSV fully connected 3D NoCs and ignored the performance degradation. In order to adequately improve reliability and performance in TSV fully connected 3D NoC architectures, we propose a TSV-aware Reinforcement Learning Assisted Routing Algorithm (RLARA) for fault-tolerant 3D NoCs. The proposed method can take advantage of both the high throughput of fully connected TSVs and the cost-effective fault tolerance of partially connected TSVs using periodically updated TSV-aware Q table of reinforcement learning. RLARA makes the distributed routing decision with the lowest TSV utilization to avoid the overheating of the TSVs and mitigate the reliability problem. Furthermore, the K-means clustering algorithm is further adopted to compress the routing table of RLARA by exploiting the routing information similarity. To alleviate the inherent deadlock issue of adaptive routing algorithms, the link Q-value from reinforcement learning is combined with the router status based in buffer utilization to predict the congestion and enable RLARA to perform best even under a high traffic load. The experimental results of the ablation study on simulator Garnet 2.0 verify the effectiveness of our proposed RLARA under different fault models, which can perform better than the latest 3D NoC routing algorithms, with up to a 9.04% lower average delay and 8.58% higher successful delivered rate. [ABSTRACT FROM AUTHOR]

In this study, we investigate the effects of intentional crosstalk noise on modern Multi-Processor System-on-Chips (MPSoCs). We demonstrate the potential for an adversary to inject false data into the communication network of the MPSoC, thereby creating intentional crosstalk scenarios that increase the likelihood of bit errors in other users’ data. This is achieved through a pre-attack reconnaissance phase in which the attacker determines the size of the MPSoC’s communication links. Based on this information, the attacker then issues false packets into the MPSoC’s network to intentionally corrupt users’ data. Our evaluations confirm that the attack can have serious consequences, altering the on-chip communication system of MPSoCs. Additionally, we propose a multi-layer defense mechanism that ensures the proper functioning of the MPSoC in the presence of both benign and malicious crosstalk scenarios. Our experimental results confirm the effectiveness of our proposed method in addressing at least 88% of the crosstalk situations with minimal power and performance overhead compared to alternative solutions.

Cache bypassing is widely employed to alleviate cache contention and pollution in GPUs. However, cache bypassing often puts more pressure on the network-on-chip (NoC) since the bypassed requests need to traverse the NoC to reach the lower-level memories, thus worsening the NoC congestion. In this paper, we propose an aggressive GPU cache bypassing technique (called SC-Table) to alleviate cache contention and pollution. The SC-Table relies on 2-bit saturating counters (SCs) to store the bypass history of warps. Memory requests issued by a warp are allowed to bypass the L1D when the corresponding SC's value reaches the bypass threshold. In addition, we adopt the monolithic 3D-based NoC (M3D NoC) to provide better NoC throughput and latency. The combination of the SC-Table and the M3D NoC improves GPU performance by 34.6%, on average, over the baseline where there is no cache bypassing and the traditional 2D NoC is adopted. [ABSTRACT FROM AUTHOR]

The three-dimensional Network-on-Chip (NoC) data acquisition system is designed to create a time-interleaved data acquisition system using NoC technology. In the design of NoC application systems, optimizing the mapping algorithm can effectively reduce network communication latency. Aiming at the mapping challenge of a large number of functional IP nodes in 3D NoC data acquisition system, the reinforcement learning and improved attention mechanism mapping algorithm (RA-Map) is proposed. The RA-Map mapping algorithm employs node function encoding and node position encoding to express the properties of an IP node in the task graph preprocessing. The local attention mechanism is used in the mapping network encoder, and the fusion of dynamic key node information is proposed in the decoder. The mapping result evaluation network achieves unsupervised training of the mapping network. These targeted improvements improve the quality of the mapping. Experimental results show that the RA-Map mapping algorithm can effectively model the IP core mapping. Compared with the DPSO algorithm and SA algorithm, the average communication cost of RA-Map mapping algorithm is reduced by 6.5 % and 8.5 %, respectively. [ABSTRACT FROM AUTHOR]

The die-stacking structure of 3D network-on-chips (3D NoC) leads to high power density and unequal thermal conductance between different layers, which results in low reliability and performance degradation of 3D NoCs. Congestion-aware adaptive routing, which is capable of balancing the network's traffic load, can alleviate congestion and thermal problems so as to improve the performance of the network. In this study, we propose a traffic- and thermal-aware Q-routing algorithm (TTQR) based on Q-learning, a reinforcement learning method. The proposed algorithm saves the local traffic status and the global temperature information to the Q1-table and Q2-table, respectively. The values of two tables are updated by the packet header and saved in a small size, which saves the hardware overhead. Based on the ratio of the Q1-value to the Q2-value corresponding to each direction, the packet's output port is selected. As a result, packets are transferred to the chosen path to alleviate thermal problems and achieve more balanced inter-layer traffic. Through the Access Noxim simulation platform, we compare the proposed routing algorithm with the TAAR routing algorithm. According to experimental results using synthetic traffic patterns, our proposed methods outperform the TAAR routing algorithm by an average of 63.6% and 41.4% in average latency and throughput, respectively. [ABSTRACT FROM AUTHOR]

Three-dimensional network-on-chip (NoC) is the primary interconnection method for 3D-stacked multicore processors due to their excellent scalability and interconnect flexibility. With the support of 3D NoC, 3D non-uniform cache architecture (NUCA) is commonly used to organize the last-level cache (LLC) due to its high capacity and fast access latency. However, owing to the layered structure that leads to longer heat dissipation paths and variable inter-layer cooling efficiency, 3D NoC experiences a severe thermal problem that has a big impact on the reliability and performance of the chip. In traditional memory-to-LLC mapping in 3D NUCA, the traffic load in each node is inconsistent with its heat dissipation capability, causing thermal hotspots. To solve the above problem, we propose a temperature-balanced NUCA mapping mechanism named TB-NUCA. First, the Bayesian optimization algorithm is used to calculate the probability distribution of cache blocks in each node in order to equalize the node temperature. Secondly, the structure of TB-NUCA is designed. Finally, comparative experiments were conducted under random, transpose-2, and shuffle traffic patterns. The experimental results reveal that, compared with the classical NUCA mapping mechanism (S-NUCA), TB-NUCA can increase the mean-time-to-failure (MTTF) of routers by up to 28.13% while reducing the maximum temperature, average temperature, and standard deviation of temperature by a maximum of 4.92%, 4.48%, and 20.46%, respectively. [ABSTRACT FROM AUTHOR]

The die-stacking structure of 3D network-on-chips (3D NoC) leads to high power density and unequal thermal conductance between different layers, which results in low reliability and performance degradation of 3D NoCs. Congestion-aware adaptive routing, which is capable of balancing the network’s traffic load, can alleviate congestion and thermal problems so as to improve the performance of the network. In this study, we propose a traffic- and thermal-aware Q-routing algorithm (TTQR) based on Q-learning, a reinforcement learning method. The proposed algorithm saves the local traffic status and the global temperature information to the Q1-table and Q2-table, respectively. The values of two tables are updated by the packet header and saved in a small size, which saves the hardware overhead. Based on the ratio of the Q1-value to the Q2-value corresponding to each direction, the packet’s output port is selected. As a result, packets are transferred to the chosen path to alleviate thermal problems and achieve more balanced inter-layer traffic. Through the Access Noxim simulation platform, we compare the proposed routing algorithm with the TAAR routing algorithm. According to experimental results using synthetic traffic patterns, our proposed methods outperform the TAAR routing algorithm by an average of 63.6% and 41.4% in average latency and throughput, respectively.

Summary: Three‐dimensional Network‐on‐Chip (3D NoC) is a promising solution for solving 2D NoC problems while optimizing the system's performance. Mapping applications in 3D NoC is a crucial step as it has a significant impact on overall system performance. Moreover, multi‐task supported processing elements (PEs) are needed to run multiple applications and provide more scalability. Most of the existing 3D mapping approaches consider only the single‐task platform. In this paper, we propose an efficient multi‐task mapping algorithm targeting regular 3D NoC that allows an incremental mapping and parallel execution of many applications onto different partitions on the 3D NoC. The proposed mapping algorithm is composed of three main steps aiming to reduce the communications overhead, exploiting the benefits of vertical links and improving the performances. The algorithm has been evaluated with various random and realistic benchmarks and compared with existing mapping algorithms for 3D NoC. The experimental results demonstrate that the proposed mapping strategy achieves significant performance in terms of communication cost, energy consumption and execution time. [ABSTRACT FROM AUTHOR]

Three-dimensional Network-on-Chips (3D NoCs) is a popular design choice due to its low packet latency, low network power consumption and high packing density. However, 3D NoCs suffer from high temperature issues. The 3D stacking of Si-layers elongates heat transfer path from different Si-layers to the heat sink resulting in increase in peak temperature of the chip. Since routers of NoCs have high power densities, a higher router activity may result in signification increase in temperature of the chip. Therefore, a judicious selection of the routing path is necessary to reduce the chip temperature. As the routers placed at the lower Si-layers have higher thermal conductance to the heat sink, a routing path consisting of more number of routers in the lower Si-layers may improve the temperature profile of the chip. In this paper, we have proposed two different thermal-aware routing approaches, which use downward detoured routing for some optimally selected communication paths to reduce the chip temperature. The first technique is a thermal-aware application-specific Mixed Integer Linear Programming based method (named TMD), while the second one is an application-agnostic heuristic approach (named TSD). To predict the effect of detour decisions on the temperature profile of the chip, TMD technique has applied two different thermal models with a constraint on the average packet delay (APD) of the network. Experimental results show that a significant temperature reduction (up to 22 ∘) can be achieved within minimal performance loss (10% increase in APD) using either of the proposed techniques, compared to the minimal path routing algorithms – XYZ, ZXY and EDGE and the greedy detour approaches – All Detour and Random Detour. [ABSTRACT FROM AUTHOR]

Three-Dimensional (3D) Network-on-Chip (NoC) architectures being the combination of NoC and 3D die stacking chip integration technology are more prone to faults that might occur with ageing, low supply voltage or physical damage of the chip. Being the heart of NoC, the router is known for consuming significantly high power. With an on-chip testability feature for faults, the amount of power consumed increases linearly with the addition of logical components. This paper, proposes a power optimized Built-In Self-Test (BIST) capable 7-port router architecture for 3D NoC by considering Marching Memory Through Type for buffer design and Cellular Automata Rule 45 for test sequence generation. The power report generated for the proposed design has been compared against that of the conventional router without BIST capability and widely used Linear Feedback Shift Register based BIST capable router architectures and further a constructive discussion in this paper, states the advantages of the proposed design in comparison to its analogues.

In the paper we describe F-medians searching algorithm for Three-dimensional (3D) Network-on-Chip (NoC) design. Modern 3D NoC development is complex and complicated task. Developer should solve different problems: Intellectual Property (IP) blocks placement on the die, organization of vertical links between dies in the 3D stack, energy consumption limitation, system performance improvement. We consider approaches for placement vertical links on the die and suggest a new methods that are based on the P-median problem.

2D-NoCs have been the mainstream approach used to interconnect multi-core systems. 3D-NoCs have emerged to compensate for deficiencies of 2D-NoCs such as long latency and power overhead. A low-latency routing algorithm for 3D-NoC is designed to accommodate high-speed communication between cores. Both simulation and analytical models are applied to estimate the communication latency of NoCs. Generally, simulations are time-consuming and slow down the design process. Analytical models provide, within a fraction of the time, nearly accurate results which can be used by simulation to fine-tune the design. In this paper, a high performance and adaptive routing algorithm has been proposed for partially connected 3D-NoCs. Latency of the routing algorithm under different traffic patterns, different number of elevators and different elevator assignment mechanisms are reported. An analytical model, tailored to the adaptivity of the algorithm and under low traffic scenarios, has been developed and the results have been verified by simulation. According to the results, simulation and analytical results are consistent within a 10 percent margin. [ABSTRACT FROM AUTHOR]

3D ICs can take advantage of a scalable communication platform, commonly referred to as the Networks-on-Chip (NoC). In the basic form of 3D-NoC, all routers are vertically connected. Partially connected 3D-NoC has emerged because of physical limitations of using vertical links. Routing is of great importance in such partially connected architectures. A high-performance, fault-tolerant and adaptive routing strategy with respect to the communication flow among the cores is crucial while freedom from livelock and deadlock has to be guaranteed. In this paper we introduce a new routing algorithm for partially connected 3D-NoCs. The routing algorithm is adaptive and tolerates the faults on vertical links as compared to the predesigned routing algorithms. Our results show a $40-50\%$ improvement in the fraction of intact inter-level communications when the fault tolerant algorithm is used. This routing algorithm is light-weight and has only one virtual channel along the $Y$ dimension. [ABSTRACT FROM AUTHOR]

The use of short Through-Silicon Vias (TSVs) in 3D integration Technology introduces a significant reduction in routing area, power consumption, and delay. Although, there are still several challenges in 3D integration technology; mainly low yield, which is a direct result of extra fabrication steps of TSVs. Therefore, reducing TSV count has a considerable effect on improving yield and hence reducing cost. A TSV multiplexing technique called TSVBOX was introduced in Said et al. (2013) to reduce the TSV count without affecting the direct benefits of TSVs. Although, the TSVBOX introduces some delay to the signals to be multiplexed, this delay effect of TSV multiplexing is not addressed yet. In this paper, we analyze the TSVBOX timing requirements and propose a design methodology for TSVBOX-based 3D Network-on-Chip (NoC). Then performance and power comparisons are conducted to investigate the direct effects of TSV multiplexing on these two metrics. After that the basic fabrication metrics are compared to investigate the effect of the proposed design methodology on yield and cost. We show that the TSVBOX extremely enhances the fabrication metrics at minimal degradation in performance and power consumption, especially for Hotspot-like traffic patterns. [ABSTRACT FROM AUTHOR]

Die-stacking technology is expanding the space diversity of on-chip communications by leveraging through-silicon-via (TSV) integration and wafer bonding. The 3D network-on-chip (NoC), a combination of die-stacking technology and systematic on-chip communication infrastructure, suffers from increased thermal density and unbalanced heat dissipation across multi-stacked layers, significantly affecting chip performance and reliability. Recent studies have focused on runtime thermal management (RTM) techniques for improving the heat distribution balance, but performance degradations, owing to RTM mechanisms and unbalanced inter-layer traffic distributions, remain unresolved. In this study, we present a Q-function-based traffic- and thermal-aware adaptive routing algorithm, utilizing a reinforcement machine learning technique that gradually incorporates updated information into an RTM-based 3D NoC routing path. The proposed algorithm initially collects deadlock-free directions, based on the RTM and topology information. Subsequently, Q-learning-based decision making (through the learning of regional traffic information) is deployed for performance improvement with more balanced inter-layer traffic. The simulation results show that the proposed routing algorithm can improve throughput by 14.0%–28.2%, with a 24.9% more balanced inter-layer traffic load and a 30.6% more distributed inter-layer thermal dissipation on average, compared with those obtained in previous studies of a 3D NoC with an 8 × 8 × 4 mesh topology. [ABSTRACT FROM AUTHOR]