arXiv subjects

Wenxin Zheng

Publications and source records attributed to Wenxin Zheng.

7 recordsLinked to original sources

arXiv2026

Determinism-Preserving GPU Spatial Sharing with Vitamin-E

GPU sharing faces a determinism--utilization tradeoff: fixed bindings can strand capacity as demand fluctuates, while resource-driven kernel reshaping improves utilization by altering a launch's parallel structure, potentially changing output bits. We rethink modern GPU scheduling and observe that it decouples logical structure from physical width: one unmodified launch spans a family of widths through changes in block placement and wave count. From this observation, we derive the parallel-structure invariant: for fixed-structure deterministic workloads, keeping each launch immutable makes its output bits independent of physical width. Guided by this invariant, Vitamin-E late-binds immutable launches to pooled physical contexts, preserving bitwise equality across allocations, whereas resource-driven reshaping can alter the selected token under temperature-zero greedy decoding. Across all workload--baseline comparisons, Vitamin-E achieves up to 3.50$\times$ the aggregate normalized LLM training throughput, 62.5\% lower inference p99 latency, and 1.43$\times$ the background-training throughput. With the same mechanism, \textsc{TPOT-First} reduces TPOT SLO violations by up to 46.1\% over \textsc{Throughput-Oriented} on three serving workloads, demonstrating mechanism effectiveness and policy flexibility.

cs.DC↗

arXiv2026

VUDA: Breaking CUDA-Vulkan Isolation for Spatial Sharing of Compute and Graphics on the Same GPU

GPU-based simulation environments for embodied AI interleave physics simulation (CUDA) and photorealistic rendering (Vulkan) on a single device. We observe that two foundational scenarios -- simulation data generation and RL training -- can be naturally adapted to execute their simulation and rendering phases concurrently, presenting a significant opportunity to improve GPU utilization through spatial multiplexing. However, a fundamental obstacle we term execution isolation prevents this: CUDA and Vulkan create separate GPU contexts whose channels are bound to different scheduling groups, confining compute and graphics to mutually exclusive time slices. Existing spatial-sharing techniques are limited to the CUDA ecosystem, while temporal-sharing approaches underutilize available resources. This paper presents VUDA, a system that breaks execution isolation to enable spatial parallelism between CUDA compute and Vulkan graphics workloads. VUDA is built on two key observations: although CUDA and Vulkan expose different programming abstractions, their execution paths converge to a common channel primitive at the driver and hardware level; meanwhile, their virtual-address spaces are inherently disjoint, making safe page-table merging feasible without remapping. VUDA exposes a thin API for developers to annotate co-schedulable CUDA streams, and realizes spatial sharing through channel redirection into Vulkan's scheduling domain and page-table grafting to unify address spaces, eliminating all data copying on the critical path. Experiments on representative embodied-AI workloads show that VUDA delivers up to 85% higher throughput than temporal-sharing baselines, while improving GPU utilization and reducing end-to-end latency.

cs.OS↗

arXiv2025

Retrofitting Control Flow Graphs in LLVM IR for Auto Vectorization

Modern processors increasingly rely on SIMD instruction sets, such as AVX and RVV, to significantly enhance parallelism and computational performance. However, production-ready compilers like LLVM and GCC often fail to fully exploit available vectorization opportunities due to disjoint vectorization passes and limited extensibility. Although recent attempts in heuristics and intermediate representation (IR) designs have attempted to address these problems, efficiently simplifying control flow analysis and accurately identifying vectorization opportunities remain challenging tasks. To address these issues, we introduce a novel vectorization pipeline featuring two specialized IR extensions: SIR, which encodes high-level structural information, and VIR, which explicitly represents instruction dependencies through data dependency analysis. Leveraging the detailed dependency information provided by VIR, we develop a flexible and extensible vectorization framework. This approach substantially improves interoperability across vectorization passes and expands the search space for identifying isomorphic instructions, ultimately enhancing both the scope and efficiency of automatic vectorization. Experimental evaluations demonstrate that our proposed vectorization pipeline achieves significant performance improvements, delivering speedups of up to 53% and 58% compared to LLVM and GCC, respectively.

cs.PL↗

arXiv2025

Basic Cycle Ratio: Cost-Effective Ranking of Influential Spreaders from Local and Global Perspectives

Spreading processes are fundamental to complex networks. Identifying influential spreaders with dual local and global roles presents a crucial yet challenging task. To address this, our study proposes a novel method, the Basic Cycle Ratio (BCR), for assessing node importance. BCR leverages basic cycles and the cycle ratio to uniquely capture a node's local significance within its immediate neighborhood and its global role in maintaining network cohesion. We evaluated BCR on six diverse real-world social networks. Our method outperformed traditional centrality measures and other cycle-based approaches, proving more effective at selecting powerful spreaders and enhancing information diffusion. Besides, BCR offers a cost-effective and practical solution for social network applications.

cs.SI↗

arXiv2025

Leveraging OS-Level Primitives for Robotic Action Management

End-to-end imitation learning frameworks (e.g., VLA) are increasingly prominent in robotics, as they enable rapid task transfer by learning directly from perception to control, eliminating the need for complex hand-crafted features. However, even when employing SOTA VLA-based models, they still exhibit limited generalization capabilities and suboptimal action efficiency, due to the constraints imposed by insufficient robotic training datasets. In addition to addressing this problem using model-based approaches, we observe that robotic action slices, which consist of contiguous action steps, exhibit strong analogies to the time slices of threads in traditional operating systems. This insight presents a novel opportunity to tackle the problem at the system level. In this paper, we propose AMS, a robot action management system enhanced with OS-level primitives like exception, context switch and record-and-replay, that improves both execution efficiency and success rates of robotic tasks. AMS first introduces action exception, which facilitates the immediate interruption of robotic actions to prevent error propagation. Secondly, AMS proposes action context, which eliminates redundant computations for VLA-based models, thereby accelerating execution efficiency in robotic actions. Finally, AMS leverages action replay to facilitate repetitive or similar robotic tasks without the need for re-training efforts. We implement AMS in both an emulated environment and on a real robot platform. The evaluation results demonstrate that AMS significantly enhances the model's generalization ability and action efficiency, achieving task success rate improvements ranging from 7x to 24x and saving end-to-end execution time ranging from 29% to 74% compared to existing robotic system without AMS support.

cs.OS↗

arXiv2025

Weighted cycle-based identification of influential node groups in complex networks

Identifying influential node groups in complex networks is crucial for optimizing information dissemination, epidemic control, and viral marketing. However, traditional centrality-based methods often focus on individual nodes, resulting in overlapping influence zones and diminished collective effectiveness. To overcome these limitations, we propose Weighted Cycle (WCycle), a novel indicator that incorporates basic cycle structures and node behavior traits (edge weights) to comprehensively assess node importance. WCycle effectively identifies spatially dispersed and structurally diverse key node group, thereby reducing influence redundancy and enhancing network-wide propagation. Extensive experiments on six real-world networks demonstrate WCycle's superior performance compared to five benchmark methods across multiple evaluation dimensions, including influence propagation efficiency, structural differentiation, and cost-effectiveness. The findings highlight WCycle's robustness and scalability, establishing it as a promising tool for complex network analysis and practical applications requiring effective influence maximization.

cs.SI↗

arXiv2024

G3R: Generating Rich and Fine-grained mmWave Radar Data from 2D Videos for Generalized Gesture Recognition

Millimeter wave radar is gaining traction recently as a promising modality for enabling pervasive and privacy-preserving gesture recognition. However, the lack of rich and fine-grained radar datasets hinders progress in developing generalized deep learning models for gesture recognition across various user postures (e.g., standing, sitting), positions, and scenes. To remedy this, we resort to designing a software pipeline that exploits wealthy 2D videos to generate realistic radar data, but it needs to address the challenge of simulating diversified and fine-grained reflection properties of user gestures. To this end, we design G3R with three key components: (i) a gesture reflection point generator expands the arm's skeleton points to form human reflection points; (ii) a signal simulation model simulates the multipath reflection and attenuation of radar signals to output the human intensity map; (iii) an encoder-decoder model combines a sampling module and a fitting module to address the differences in number and distribution of points between generated and real-world radar data for generating realistic radar data. We implement and evaluate G3R using 2D videos from public data sources and self-collected real-world radar data, demonstrating its superiority over other state-of-the-art approaches for gesture recognition.

cs.MM↗