Search Results (706)

In modern multi-core processors, memory request latency critically constrains overall performance. Prefetching, a promising technique, mitigates memory access latency by pre-loading data into faster cache structures. However, existing core-side prefetchers lack visibility to the DRAM state and may issue suboptimal requests, while conventional memory-side prefetchers often default to simple next-line policies that miss complex access patterns. We propose a comprehensive memory-side prefetch optimization scheme, which includes a prefetcher that utilizes advanced prefetching algorithms and an optimization module. Our prefetcher is capable of detecting more complex memory access patterns, thereby improving both prefetch accuracy and coverage. Additionally, considering the characteristics of DRAM memory access, the optimization module minimizes the negative impact of prefetch requests on DRAM by enhancing coordination with memory operations. Additionally, our prefetcher works in synergy with core-side prefetchers to deliver superior overall performance. Simulation results using Gem5 and SPEC CPU2017 workloads show that our approach delivers an average performance improvement of 10.5% and reduces memory access latency by 61%. Our prefetcher also operates in conjunction with core-side prefetchers to form a multi-level prefetching hierarchy, enabling further performance gains through coordinated and complementary prefetching strategies. Full article

Energy consumption has emerged as a critical design constraint in deploying high-performance neural networks, especially on edge devices with limited power resources. In this paper, a comparative study is conducted for two prevalent deep learning paradigms—convolutional neural networks (CNNs), exemplified by ResNet18, and transformer-based large language models (LLMs), represented by GPT3-small, Llama-7B, and GPT3-175B. By analyzing how the scaling of memory energy versus computing energy affects the energy consumption of neural networks with different batch sizes (1, 4, 8, 16), it is shown that ResNet18 transitions from a memory energy-limited regime at low batch sizes to a computing energy-limited regime at higher batch sizes due to its extensive convolution operations. On the other hand, GPT-like models remain predominantly memory-bound, with large parameter tensors and frequent key–value (KV) cache lookups accounting for most of the total energy usage. Our results reveal that reducing weight-memory energy is particularly effective in transformer architectures, while improving multiply–accumulate (MAC) efficiency significantly benefits CNNs at higher workloads. We further highlight near-memory and in-memory computing approaches as promising strategies to lower data-transfer costs and enhance power efficiency in large-scale deployments. These findings offer actionable insights for architects and system designers aiming to optimize artificial intelligence (AI) performance under stringent energy budgets on battery-powered edge devices. Full article

Intelligent Connected Vehicles (ICVs) impose stringent requirements on real-time computational services. However, limited onboard resources and the high latency of remote cloud servers restrict traditional solutions. Unmanned Aerial Vehicle (UAV)-assisted Mobile Edge Computing (MEC), which deploys computing and storage resources at the network edge, offers a promising solution. In UAV-assisted vehicular networks, jointly optimizing content and service caching, computation offloading, and UAV trajectories to maximize system performance is a critical challenge. This requires balancing system energy consumption and resource allocation fairness while maximizing cache hit rate and minimizing task latency. To this end, we introduce system efficiency as a unified metric, aiming to maximize overall system performance through joint optimization. This metric comprehensively considers cache hit rate, task computation latency, system energy consumption, and resource allocation fairness. The problem involves discrete decisions (caching, offloading) and continuous variables (UAV trajectories), exhibiting high dynamism and non-convexity, making it challenging for traditional optimization methods. Concurrently, existing multi-agent deep reinforcement learning (MADRL) methods often encounter training instability and convergence issues in such dynamic and non-stationary environments. To address these challenges, this paper proposes a MADRL-based joint optimization approach. We precisely model the problem as a Decentralized Partially Observable Markov Decision Process (Dec-POMDP) and adopt the Multi-Agent Proximal Policy Optimization (MAPPO) algorithm, which follows the Centralized Training Decentralized Execution (CTDE) paradigm. Our method aims to maximize system efficiency by achieving a judicious balance among multiple performance metrics, such as cache hit rate, task delay, energy consumption, and fairness. Simulation results demonstrate that, compared to various representative baseline methods, the proposed MAPPO algorithm exhibits significant superiority in achieving higher cumulative rewards and an approximately 82% cache hit rate. Full article

In this paper, we investigate a low Earth orbit (LEO) satellite-enabled mobile edge computing (MEC) network, where multiple cache-enabled LEO satellites are deployed to address heterogeneous content requests from ground users. To evaluate the network’s capability in meeting user demands, we adopt the average quality of experience (QoE) of the users as the performance metric, defined based on the effective transmission rate under communication interference. Our analysis reveals that the average QoE is determined by the content caching decisions at the satellites, thereby allowing us to formulate an average QoE maximization problem, subject to practical constraints on the satellite caching capacity. To tackle this NP-hard problem, we design a two-stage content caching algorithm that combines divide-and-conquer and greedy policies for efficient solution. The numerical results validate the effectiveness of the proposed approach. Compared with several benchmark schemes, our algorithm achieves notable improvements in terms of the average QoE while significantly reducing caching costs, particularly under resource-constrained satellite settings. Full article

With the continuous development of the Internet of Things (IoT) and communication technologies, the demand for low latency in practical applications is becoming increasingly significant. Mobile edge computing, as a promising computational model, is receiving growing attention. However, most existing studies fail to consider two critical factors: task dependency and service caching. Additionally, the majority of proposed solutions are not related to the optimal solution. We investigate the task offloading problem in mobile edge computing. Considering the requirements of applications for service caching and task dependency, we define an optimization problem to minimize the delay under the constraint of maximum completion cost and present a $(1+ϵ)$-approximation algorithm and an exact algorithm. Specifically, the offloading scheme is determined based on the relationships between tasks as well as the cost and delay incurred by data transmission and task execution. Simulation results demonstrate that in all cases, the offloading schemes obtained by our algorithm consistently outperform other algorithms. Moreover, the approximation ratio to the optimal solution from the approximation algorithm is validated to be less than $(1+ϵ)$, and the exact algorithm consistently produces the optimal solution. Full article

Fruit color serves as a crucial visual indicator in chili peppers and is closely linked to the bioactive components that determine their economic and nutritional value. However, the specific components and potential molecular mechanisms that impact fruits’ development and color changes are less thoroughly understood. Here, we utilized three chili pepper varieties (CS03, CS29, and L816) at different developmental stages (young fruit stage, turning color stage, and mature stage) as research materials and integrated transcriptome and non-targeted metabolome analyses to explore the variation in bioactive components and color to explain the molecular regulatory mechanisms underlying different colors of chili peppers during the young fruit stage. Our results showed that flavonoids were the most enriched differential metabolites; aromadendrin 4′-glucoside, diospyrin, precarthamin, kaempferol-3-O-rutinoside, and kaempferol-3-O-Glucoside were significantly enriched in the young fruit stage of pepper CS03; and cyanidin, delphinidin, and cyanidin 3-glucoside were major contributors to the color formation. The upregulation of anthocyanin was related to the structural genes CaC4H, Ca4CL, CaCHS, CaF3H, CaANS, and CaUFGT, and key transcription factors such as CaMYBs and CabHLHs may have contributed to the differential accumulation of anthocyanins in CS03; in addition, RT-qPCR validation was correlated with anthocyanins, but also with flavonoids. This article mainly focuses on the changes in chili pigments, particularly anthocyanins, and explores the molecular mechanisms involved. This provides a reference for research on color in solanaceae vegetables and lays a theoretical foundation for further research on the bioactive components of chili peppers, as well as for optimizing harvesting practices and dietary recommendations. Full article

Vision–language models learn visual concepts from the supervision of natural language. It can significantly enhance the generalizability of real-time intelligent sensing, such as analyzing camera-captured real-time images for visually impaired users. However, adapting vision–language models to distribution shifts at test time, caused by several factors such as lighting or weather changes, remains challenging. In particular, most existing test-time adaptation methods rely on gradient-based fine-tuning and backpropagation, making them computationally expensive and unsuitable for real-time applications. To address this challenge, the Training-Free Dynamic Adapter (TDA) has recently been introduced as a lightweight alternative that uses a dynamic key–value cache and pseudo-label refinement for test-time adaptation without backpropagation. Building on this, we propose TDA-L, a new framework that integrates Low-Rank Adaptation (LoRA) to reduce the size of feature representations and related computational overhead at test time using pre-learned low-rank matrices. TDA-L applies LoRA transformations to both query and cached features during inference, cost-efficiently improving robustness to distribution shifts while maintaining the training-free nature of TDA. Experimental results on seven benchmarks show that TDA-L maintains accuracy but achieves lower latency, less memory consumption, and higher throughput, making it well-suited for AI-based real-time sensing. Full article

This paper introduces a new memory-efficient algorithm for connected-components labeling in binary images, which is based on run-length encoding. Unlike conventional pixel-based methods that scan and label individual pixels using global buffers or disjoint-set structures, our approach encodes rows as linked segments and merges them using a union-by-size strategy. We accelerate run detection by using a precomputed 16-bit cache of binary patterns, allowing for fast decoding without relying on bitwise CPU instructions. When compared against other run-length encoded algorithms, such as the Scan-Based Labeling Algorithm or Run-Based Two-Scan, our method achieves up to 35% faster on most real-world datasets. While other binary-optimized algorithms, such as Bit-Run Two-Scan and Bit-Merge Run Scan, are up to 45% faster than our algorithm, they require much higher memory usage. Compared to them, our method tends to reduce memory consumption on some large document datasets by up to 80%. Full article

Almost 70 years after the surprise discovery of a cache of textile-wrapped relics inside an early 13th-century reliquary bust, the St Eustace head reliquary (accession number 1850,1127.1), four of the textile relic wrappings were analysed by combining multiband imaging and fibre-optic reflectance spectroscopy (FORS), as well as dye analysis by high-pressure liquid chromatography coupled to mass spectrometry (HPLC-MS) and fibre analysis by scanning electron microscopy—energy dispersive X-ray spectroscopy (SEM-EDX). In all cases, the use of silk was confirmed, in line with the idea that these precious textiles were purposefully chosen for reuse in a sacred setting. Additionally, dye analysis was able to point to the possible geographic origins of the textile fragments. For 1850,1127.1.a, a mixture of sappanwood (Biancaea sappan) and flavonoid yellow dyes was commensurate with a Chinese or Central Asian origin. Mediterranean origins were thought likely for 1850,1127.1.c and 1850,1127.1.f, from the mixture of kermes (Kermes vermilio) and cochineal (likely Porphyrophora sp.), found in the mauve band of the former, and the combination of weld (Reseda luteola), madder (Rubia tinctorum) and an indigoid dye found in the latter. Finally, the unusual combination of sappanwood, orchil and a yellow dye containing flavonoid glucuronides suggested a less straightforward origin for textile 1850,1127.1.g. The other textile fragments from the reliquary were only investigated using FORS without removing them from their Perspex glass mounts. Nonetheless, indications for the presence of insect-red anthraquinone dyes, safflower (Carthamus tinctorius) and an indigoid dye were obtained from some of these fragments. The study provides a window into the landscape of availability, use and re-use in sacred contexts of precious textiles in the 13th century and evidences the geographic reach of these silks, allowing a new perspective on the St Eustace head reliquary. Full article

This study proposes a memory-based forensic procedure for real-time recovery of deleted data in Microsoft SQL Server environments. This approach is particularly relevant for sensor-driven and embedded systems—such as those used in IoT gateways and edge computing platforms—where lightweight SQL engines store critical operational and measurement data locally and are vulnerable to insider manipulation. Traditional approaches to deleted data recovery have primarily relied on transaction log analysis or static methods involving the examination of physical files such as .mdf and .ldf after taking the database offline. However, these methods face critical limitations in real-time applicability and may miss volatile data that temporarily resides in memory. To address these challenges, this study introduces a methodology that captures key deletion event information through transaction log analysis immediately after data deletion and directly inspects memory-resident pages loaded in the server’s Buffer Pool. By analyzing page structures in the Buffer Pool and cross-referencing them with log data, we establish a memory-driven forensic framework that enables both the recovery and verification of deleted records. In the experimental validation, records were deleted in a live SQL Server environment, and a combination of transaction log analysis and in-memory page inspection allowed for partial or full recovery of the deleted data. This demonstrates the feasibility of real-time forensic analysis without interrupting the operational database. The findings of this research provide a foundational methodology for enhancing the speed and accuracy of digital forensics in time-sensitive scenarios, such as insider threats or cyber intrusion incidents, by enabling prompt and precise recovery of deleted data directly from memory. These capabilities are especially critical in IoT environments, where real-time deletion recovery supports sensor data integrity, forensic traceability, and uninterrupted system resilience. Full article

This paper seeks to address the limitations of conventional remote sensing data dissemination algorithms, particularly their inability to model fine-grained multi-modal heterogeneous feature correlations and adapt to dynamic network topologies under resource constraints. This paper proposes multi-modal-MAPPO, a novel multi-modal deep reinforcement learning (MDRL) framework designed for a proactive data push in large-scale integrated satellite–terrestrial networks (ISTNs). By integrating satellite cache states, user cache states, and multi-modal data attributes (including imagery, metadata, and temporal request patterns) into a unified Markov decision process (MDP), our approach pioneers the application of the multi-actor-attention-critic with parameter sharing (MAPPO) algorithm to ISTNs push tasks. Central to this framework is a dual-branch actor network architecture that dynamically fuses heterogeneous modalities: a lightweight MobileNet-v3-small backbone extracts semantic features from remote sensing imagery, while parallel branches—a multi-layer perceptron (MLP) for static attributes (e.g., payload specifications, geolocation tags) and a long short-term memory (LSTM) network for temporal user cache patterns—jointly model contextual and historical dependencies. A dynamically weighted attention mechanism further adapts modality-specific contributions to enhance cross-modal correlation modeling in complex, time-varying scenarios. To mitigate the curse of dimensionality in high-dimensional action spaces, we introduce a multi-dimensional discretization strategy that decomposes decisions into hierarchical sub-policies, balancing computational efficiency and decision granularity. Comprehensive experiments against state-of-the-art baselines (MAPPO, MAAC) demonstrate that multi-modal-MAPPO reduces the average content delivery latency by 53.55% and 29.55%, respectively, while improving push hit rates by 0.1718 and 0.4248. These results establish the framework as a scalable and adaptive solution for real-time intelligent data services in next-generation ISTNs, addressing critical challenges in resource-constrained, dynamic satellite–terrestrial environments. Full article

This article examines the material data preserved in king’s coffins used to bury and/or rebury five different kings, which represent the surviving material evidence we have of the art produced to manufacture divinized kingship during the New Kingdom: Seqenenre Taa, Kamose, Thutmose I/Panedjem I, Thutmose III, and Ramses II. All of them were removed from their original 17th and 18th Dynasty sepulchers, stripped of valuable materials, modified, and reused in later cache burials of the 20th, 21st, and 22nd Dynasties by 20th and 21st Dynasty High Priests of Amen, who used these recrafted coffins as a means of claiming their political and ideological legitimacy. Supported with detailed evidence of the five surviving king’s coffins as objects of social and political value and sometimes relying on the coffins recovered from Tutankhamun’s tomb for comparison, this article attempts to reconstruct some of the original material state of this art as a tool of power. Full article

Existing databases supporting Online Transaction Processing (OLTP) workloads based on non-volatile memory (NVM) almost all use Multi-Version Concurrency Control (MVCC) protocol to ensure data consistency. MVCC allows multiple transactions to execute concurrently without lock conflicts, reducing the wait time between read and write operations, and thereby significantly increasing the throughput of NVM OLTP engines. However, it requires garbage collection (GC) to clean up the obsolete tuple versions to prevent storage overflow, which consumes additional system resources. Furthermore, existing GC approaches in NVM OLTP engines are inefficient because they are based on methods designed for dynamic random access memory (DRAM) OLTP engines, without considering the significant differences in read/write bandwidth and cache line size between NVM and DRAM. These approaches either involve excessive random NVM access (traversing tuple versions) or lead to too many additional NVM write operations, both of which degrade the performance and durability of NVM. In this paper, we propose TB-Collect, a high-performance GC approach specifically designed for NVM OLTP engines. On the one hand, TB-Collect separates tuple headers and contents, storing data in an append-only manner, which greatly reduces NVM writes. On the other hand, TB-Collect performs GC at the block level, eliminating the need to traverse tuple versions and improving the utilization of reclaimed space. We have implemented TB-Collect on DBx1000 and MySQL. Experimental results show that TB-Collect achieves 1.15 to 1.58 times the throughput of existing methods when running TPCC and YCSB workloads. Full article

To address the current limitations of spaceborne solid-state storage systems that cannot effectively support the parallel storage of multiple high-speed data streams, the throughput bottleneck of NAND FLASH-based solid-state storage systems was analyzed in relation to the high-speed data input requirements of payloads. A four-stage pipeline operation and bus parallel expansion scheme was proposed to enhance the throughput. Additionally, to support the parallel storage of multichannel data and continuity of pipeline loading, the shortcomings of existing caching schemes were analyzed, leading to the design of a storage system based on Synchronous Dynamic Random Access Memory (SDRAM). Model simulations indicate that, under extreme conditions, the proposed scheme could continuously receive and cache multiple high-speed file data streams into the SDRAM. File data were dynamically written into FLASH based on the priority and status of each partition cache autonomously, without overflow during caching. The system eventually entered a regular dynamic balance scheduling state to achieve parallel reception, caching, and autonomous scheduling of storage for multiple high-speed payload data streams. The data throughput rate of the storage system can reach 4 Gbps, thus satisfying future requirements for multichannel high-speed payload data storage in spaceborne solid-state storage systems. Full article

Smart cities are urban areas that use contemporary technology to improve citizens’ overall quality of life. These modern digital civil hubs aim to manage environmental conditions, traffic flow, and infrastructure through interconnected and data-driven decision-making systems. Today, many applications employ intelligent sensors for real-time data acquisition, leveraging visualization to derive actionable insights. However, despite the proliferation of such platforms, challenges like high data volume, noise, and incompleteness continue to hinder practical visual analysis. As missing data is a frequent issue in visualizing those urban sensing systems, our approach prioritizes their correction as a fundamental step. We deploy a hybrid imputation strategy combining SARIMAX, k-nearest neighbors, and random forest regression to address this. Building on this foundation, we propose an interactive web-based pipeline that processes, analyzes, and presents the sensor data provided by Basel’s “Smarte Strasse”. Our platform receives and projects environmental measurements, i.e., NO₂, O₃, PM_2.5, and traffic noise, as well as mobility indicators such as vehicle speed and type, parking occupancy, and electric vehicle charging behavior. By resolving gaps in the data, we provide a solid foundation for high-fidelity and quality visual analytics. Built on the Flask web framework, the platform incorporates performance optimizations through Flask-Caching. Concerning the user’s dashboard, it supports interactive exploration via dynamic charts and spatial maps. This way, we demonstrate how future internet technologies permit the accessibility of complex urban sensor data for research, planning, and public engagement. Lastly, our open-source web-based application keeps reproducible, privacy-aware urban analytics. Full article