Article
Computer Science, Hardware & Architecture
Fangchuan Li, Shuangjia Liu, Ning Ma, Yanli Liu, Guanyu Xing, Yanci Zhang
Summary: In this paper, a novel hybrid occlusion culling method for large scale scenes is presented. The method utilizes an iterative hierarchical Z-buffer occlusion culling algorithm and rasterization for coarse-grained and fine-grained culling, respectively. A forward warping method is also proposed to generate a low resolution approximated depth map for accelerating the culling process. Experimental results demonstrate that the algorithm outperforms existing solutions in both performance and culling rate.
Article
Engineering, Multidisciplinary
Haohai Fu, Huamin Yang, Chunyi Chen
Summary: This paper presents a real-time rendering algorithm for large-scale terrain based on GPU tessellation, with a new rule in the tessellation stage to define terrain roughness.
ALEXANDRIA ENGINEERING JOURNAL
(2021)
Article
Computer Science, Information Systems
Eun-Seok Lee, Byeong-Seok Shin
Summary: This paper introduces an object-culling technique using vertex chunk to render a massive number of objects in real-time. It compresses the bounding boxes of objects into data units called vertex chunks to reduce input data for rendering passes and utilizes GPU parallel processing to quickly restore the data and select culled objects. The method improves performance by about 15% and shows a higher effect when multiple objects are placed.
Article
Computer Science, Hardware & Architecture
Jingcheng Shen, Linbo Long, Xin Deng, Masao Okita, Fumihiko Ino
Summary: In this work, a compression-based, memory-efficient method is proposed to accelerate out-of-core stencil codes that exceed the memory capacity of a GPU. The method integrates an on-the-fly compression technique to reduce CPU-GPU data transfers and employs a single-working buffer strategy to reduce GPU memory usage, resulting in increased temporal blocking steps. Experimental results show that the proposed method significantly reduces GPU memory usage by 21% and allows for double the number of temporal blocking steps compared to the codes without compression. The method achieves speedups up to 2.09x for single-precision floating-point format and up to 1.92x for double-precision floating-point format on an NVIDIA Tesla V100 GPU in comparison with the codes without compression.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Computer Science, Interdisciplinary Applications
Shuqin Wang, Jinhai Zhang
Summary: This study proposes a method to compress vector graphics of earthquake catalogs by disregarding invisible symbols, resulting in reduced processing time and file size. The experimental results demonstrate the effectiveness of this method in reducing the size of graphics files and significantly decreasing the time required for plotting.
EARTH SCIENCE INFORMATICS
(2022)
Article
Computer Science, Information Systems
Xintao Ding, Boquan Li, Wen Zhou, Cheng Zhao
Summary: Exploring reliable correspondences is crucial in two-view geometry estimation. The widely used RANSAC method may be inefficient when actual inliers are heavily contaminated. This study proposes a CSAC method based on gradient difference to identify true inliers from heavily contaminated correspondences, and it outperforms other methods in terms of inlier precision and model accuracy.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Interdisciplinary Applications
Gregorio Quintana-Orti, Monica Chillaron, Vicente Vidal, Gumersindo Verdu
Summary: In this study, an optimized GPU software for CT image reconstruction using QR factorization with out-of-core (OOC) techniques is presented. Experimental results show that this implementation is up to 6.5 times faster compared to the CPU version and provides improved image quality. This approach allows for reduced exposure time and radiation dose for the patient.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Mechanics
Shanyouming Sun, Dan Liu, Yinglong Sheng, Shangsheng Feng, Hongbin Zhu, Tian Jian Lu
Summary: The introduction of the new N-core structure has addressed the contradiction between the thermal and mechanical properties of sandwich panel cores, resulting in excellent performance. Experimental results show that, under the same mass, the load-bearing capacity of the hybrid core and corrugated core is superior to the web core, while both the hybrid core and web core outperform the corrugated core in terms of active cooling capability.
COMPOSITE STRUCTURES
(2021)
Article
Engineering, Mechanical
Mihaela Iftimiciuc, Arne Derluyn, Jochen Pflug, Dirk Vandepitte
Summary: Honeycomb cores are widely used in various industries to build advanced lightweight structures that take advantage of high stiffness-to-weight and strength-to-weight ratios. This study focuses on the compressive behavior of a novel hierarchical sandwich honeycomb core, both in virtual and experimental testing. The finite element model is validated and can be used for further structural optimization. The comparison between the proposed hierarchical structure and conventional expanded honeycombs highlights the advantages of structural hierarchy, showing a high potential for use in the construction of sandwich panels and parts.
JOURNAL OF SANDWICH STRUCTURES & MATERIALS
(2023)
Article
Engineering, Electrical & Electronic
Mohamed Ibrahim, Salah Elhoushy, Walaa Hamouda
Summary: This paper investigates the uplink (UL) performance of cell-free massive multiple-input multiple-output (CF mMIMO) systems supported by mmWave-fronthaul networks. Analytical expressions for the distribution of the provided fronthaul capacity and the average UL data rates are derived using stochastic geometry. The results show that increasing the density of blockages degrades the average UL data rates, but increasing the density of central processing units (CPUs) can limit this effect. The study also reveals that the network deployment should be adjusted based on the available fronthaul bandwidth and the density of blockages. Increasing the density of access points (APs) beyond a certain limit does not achieve further improvement in the UL data rates. Furthermore, increasing the number of antennas per AP may even cause a degradation in the system performance.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Computer Science, Software Engineering
David Corbalan-Navarro, Juan L. Aragon, Marti Anglada, Enrique de Lucas, Joan-Manuel Parcerisa, Antonio Gonzalez
Summary: This article introduces a microarchitectural technique called Omega-Test to reduce overdraw in tile-based rendering. By leveraging frame-to-frame coherence, the proposed approach utilizes discarded tile information from current frames to predict the visibility of tiles in the next frame. Experimental evaluation shows average energy savings of 26.4% and average speedup of 16.3% for the evaluated benchmarks.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Ciprian Paduraru, Miruna Paduraru
Summary: This paper aims to improve the performance, variety, and usability of crowd animation systems by conducting blending operations on the GPU side and utilizing the time dilation offset feature.
Article
Engineering, Electrical & Electronic
Lixiang Lin, Jianke Zhu, Yisu Zhang
Summary: In this paper, an effective coarse-to-fine approach is proposed to recover the textured mesh from multi-views. A differentiable Poisson Solver is employed to represent the object's shape, and the shape geometry is optimized by minimizing the differences between the rendered mesh and the predicted depth. A physically-based inverse rendering scheme is introduced to jointly estimate the environment lighting and object's reflectance, and the texture of the reconstructed mesh is interpolated from a learnable dense texture grid, enabling real-time rendering of high resolution images.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Jing Jin, Junhui Hou, Jie Chen, Huanqiang Zeng, Sam Kwong, Jingyi Yu
Summary: In this paper, a novel learning-based method is proposed to reconstruct densely-sampled light fields with arbitrary angular resolution from sparsely-sampled light fields with irregular structures. The proposed method, which is an end-to-end trainable network, achieves high reconstruction quality and computational efficiency by synthesizing coarse sub-aperture images and refining the angular relationship. Experimental results demonstrate the superiority of the proposed method in various light field applications, including image-based rendering and depth estimation enhancement.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Software Engineering
Max Plauth, Joan Bruguera Mico, Andreas Polze
Summary: This article demonstrates that transparent on-the-fly I/O link compression in the CloudCL framework can improve performance of scale-out GPU workloads by 1.11x to 2.07x.
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
(2023)