Search

This paper presents the graphics processing unit (GPU) implementation of two-dimensional discrete cosine transform (2D DCT) and inverse discrete cosine transform (2D IDCT) for image compression applications. Based on the trigonometric properties, the transform matrices are simplified, resulting in reduced computation over the naive implementation. To assess its performance, the output image quality is measured in terms of several metrics and found to be better than all other existing transforms. To further improve the timings, a GPU implementation of the proposed transforms is obtained by exploiting the inter-level parallelism among threads and blocks in addition to efficiently accessing data from the shared memory resources. This has resulted in significant improvement in speedup (more than 5k) for both the transforms. The proposed GPU implementation of 2D DCT is compared in terms of processing time and is shown to outperform the existing work across all image dimensions. [ABSTRACT FROM AUTHOR]

In medical diagnosis, mammographic imaging is mainly concerned with the breast parenchymal patterns (counterbalance of glandular tissue and fatty tissue) by which an expert radiologist can easily determine the abnormalities in the breast of cancer patients and if the interpretation of mammogram and the quality of mammogram both are well provided. Accordingly, improved mammographic view via an efficient image processing algorithm plays a significant role in the medical diagnosis of mammograms. This study introduces a sharpening method based on the modified Laplacian filter (MLF) on compute unified device architecture (CUDA) to improve the visibility and detection of pernicious lesions in a mammogram. To process considerably large mammograms on CPU, the conventional Laplacian sharpening is more time-consuming due to the processing of all pixels with serial execution manner. Although this type of image sharpening is well established for improved image quality, its effect on a larger image for use in the GPU environment has not been extensively studied. The proposed framework is successfully devised and implemented in an efficient parallel execution manner on a computing platform of graphic processing units (GPU). To examine the impact of mammograms and filter size on performance along with the comparative processing time between serial execute on CPU and parallel computing on GPU (except data transfer time). To accelerate the performance of the proposed model, we adopt both global and shared memory in GPU to realize further improvements of the execution speed. The proposed framework applies a new nonlinear filter constraints module in the sharping stage while the Laplacian filter attenuate noise sensitivity and leads to achieving visually improved results in comparison with formal sharping. The proposed framework has been extensively compared with other recent baseline methods showing to improvement in the computational cost of the image sharping approach. Experimental results establish that the two proposed sharping methods outperform the state-of-the-art methods with respect to execution speed. [ABSTRACT FROM AUTHOR]

Sorting algorithms are fundamental tools in data processing. Sorting has been a deep area for algorithmic researchers, and many resources have been invested in more work on sorting algorithms. For this purpose, many existing sorting algorithms have been reviewed for algorithmic complexity performance. In this paper, implementation of strand sort algorithm using a Message Passing Interface (MPI) and Compute Unified Device Architecture (CUDA) approach is carried out. The proposed work is tested with a standard benchmark dataset. The main idea of the proposed algorithm is to divide the elements of the input dataset into several additional temporary sub-lists to process in parallel. The implemented algorithm using MPI and CUDA enhance the algorithm's performance. The average speedup obtained is 3.9187 using MPI and 19.9270 using CUDA respectively. [ABSTRACT FROM AUTHOR]

In arbitrary arithmetic computation and computational science, multiplying large integers is a widely used operation. Numerous cryptographic techniques involve operations on extremely large subsets of the integer numbers, including the Diffie-Hellman key exchange, RSA, ECC, and others. These techniques employ safe message encryption, decryption, and key exchange using security keys with a size of at least 1024 bits. Exponentiation and multiplication are necessary in order to carry out encryption, decryption, and key exchange. The Karatsuba algorithm is a fast and efficient method for multiplying large numbers, that reduces the number of multiplications from four to three at each recursive step. In this paper, we conduct a comprehensive assessment of the Karatsuba algorithm's performance when applied to both sequential and parallel contexts. We utilize the power of Nvidia Graphics Processing Unit (GPU) with Compute Unified Device Architecture (CUDA) programming to gauge the speedup of the parallel implementation and processor configurations. The speedup achieved by the Karatsuba algorithm running on Nvidia GPU CUDA platform over the sequential is 30.12. There exists an improvement in performance by making use of available GPU cores. The findings emphasize the potential advantages of parallelization in reducing the overall computation time. [ABSTRACT FROM AUTHOR]

The need for parallel sorting algorithms have been driven by the increasing need for large-scale datasets to be processed efficiently. Pigeonhole sorting is one of the sorting algorithms that carries sorting in linear time. This study focuses on enhancing the efficacy of the Pigeonhole Sorting method to improve the performance of the algorithm by employing parallel programming techniques specifically Message Passing Interface (MPI) and Compute Unified Device Architecture (CUDA). The primary objective is to develop and assess parallel solutions for Pigeonhole Sorting, with the aim of optimizing sorting efficiency in data-intensive applications. Commencing with a comprehensive analysis of the sequential design of the Pigeonhole Sorting algorithm, this work proceeds to create parallel implementations using CUDA for Graphics Processing Unit (GPU) acceleration and MPI for distributed memory parallelism. This work contributes valuable insights into adapting the Pigeonhole Sorting algorithm to parallel contexts. The findings emphasize the potential advantages of parallelization in reducing the overall computation time. [ABSTRACT FROM AUTHOR]

The analysis of electrothermal coupling problems finds extensive application in engineering. However, for large-scale electrothermal coupling problems, the time cost and storage requirements for solving them using the finite element method (FEM) are substantial. We optimize the finite element electrothermal coupling computation from two aspects: computational speed and storage usage. Based on the assembly-free FEM, we explore the symmetry of element matrices to reduce storage for second-order tetrahedral elements and propose a graphics processing unit (GPU) parallel algorithm to improve computational speed. At the same time, we allocate the parallel parts of an electrothermal coupling problem to two GPUs to improve the speed further. In addition, for the three types of boundary conditions in electrothermal coupling problems, we design parallel application methods suitable for assembly-free FEM. Finally, we compare our strategy with methods from other literature through the numerical experiment. Our method reduces the element matrices' storage by 45%. Compared with the solution process using the element level method and degree of freedom level method, our strategy achieves average acceleration ratios of 5.83 and 1.38, respectively. [ABSTRACT FROM AUTHOR]

The National Center for Atmospheric Research released a global atmosphere model named Community Atmosphere Model version 5.0 (CAM5), which aimed to provide a global climate simulation for meteorological research. Among them, the cloud microphysics scheme is extremely time-consuming, so developing efficient parallel algorithms faces large-scale and chronic simulation challenges. Due to the wide application of GPU in the fields of science and engineering and the NVIDIA's mature and stable CUDA platform, we ported the code to GPU to accelerate computing. In this paper, by analyzing the parallelism of CAM5 cloud microphysical schemes (CAM5 CMS) in different dimensions, corresponding GPU-based one-dimensional (1D) and two-dimensional (2D) parallel acceleration algorithms are proposed. Among them, the 2D parallel algorithm exploits finer-grained parallelism. In addition, we present a data transfer optimization method between the CPU and GPU to further improve the overall performance. Finally, GPU version of the CAM5 CMS (GPU-CMS) was implemented. The GPU-CMS can obtain a speedup of 141.69 × on a single NVIDIA A100 GPU with I/O transfer. In the case without I/O transfer, compared to the baseline performance on a single Intel Xeon E5-2680 CPU core, the 2D acceleration algorithm obtained a speedup of 48.75 × , 280.11 × , and 507.18 × on a single NVIDIA K20, P100, and A100 GPU, respectively. [ABSTRACT FROM AUTHOR]

We present hipBone, an open-source performance-portable proxy application for the Nek5000 (and NekRS) computational fluid dynamics applications. HipBone is a fully GPU-accelerated C++ implementation of the original NekBone CPU proxy application with several novel algorithmic and implementation improvements which optimize its performance on modern fine-grain parallel GPU accelerators. Our optimizations include a conversion to store the degrees of freedom of the problem in assembled form in order to reduce the amount of data moved during the main iteration and a portable implementation of the main Poisson operator kernel. We demonstrate near-roofline performance of the operator kernel on three different modern GPU accelerators from two different vendors. We present a novel algorithm for splitting the application of the Poisson operator on GPUs which aggressively hides MPI communication required for both halo exchange and assembly. Our implementation of nearest-neighbor MPI communication then leverages several different routing algorithms and GPU-Direct RDMA capabilities, when available, which improves scalability of the benchmark. We demonstrate the performance of hipBone on three different clusters housed at Oak Ridge National Laboratory, namely, the Summit supercomputer and the Frontier early-access clusters, Spock and Crusher. Our tests demonstrate both portability across different clusters and very good scaling efficiency, especially on large problems. [ABSTRACT FROM AUTHOR]

The ocean general circulation model (OGCM) is an essential tool for researching oceanography and atmospheric science. The LASG/IAP climate system ocean model version 3 (LICOM3) is a parallel version of the OGCM. Our goal is to implement and optimize a GPU version of LICOM3 based on compute unified device architecture (CUDA) called LICOM3-CUDA. Considering the characteristics of LICOM3 and CUDA, we design and implement some pivotal optimization methods, including redesigning the numerical algorithms of complicated functions, decoupling data dependency, avoiding memory write conflicts, and optimizing communication. In this paper, we selected two experiments, including 1 ∘ (small-scale) and 0.1 ∘ (large-scale) resolutions to evaluate the performance of LICOM3-CUDA. Under the experimental environment of two Intel Xeon Gold 6148 CPUs and four NVIDIA Quadro GV100s, the LICOM3-CUDA (1 ∘ ) achieves a simulation speed of 114.3 simulation-year-per-day (SYPD). Compare with the performance of LICOM3, the LICOM3-CUDA can run much faster with 6.5 times, and the compute-intensive module achieves over 70 × speedup. In addition, the energy consumption for the simulation year is reduced by 41.3%. As for high-resolution and large-scale simulation, the number of GPUs increased from 96 to 1536 as well as the LICOM3-CUDA (0.1 ∘ ) time consumption decreased from 3261 to 720 seconds with approximately 4.5 × of speedup. [ABSTRACT FROM AUTHOR]

To improve the seismic connectivity reliability (SCR) analysis efficiency of water distribution systems (WDS) based on Monte Carlo (MC) simulation, the quasi-Monte Carlo (QMC) method sampled by a low-discrepancy sequence is applied. Furthermore, a parallel algorithm combined with the breadth-first search algorithm for SCR analysis of WDS based on the QMC method and Compute Unified Device Architecture (CUDA) platform was proposed. A city WDS was taken as a computational example, the accuracy and efficiency of the traditional MC algorithm and parallel algorithm were compared, and the influence of the Sobol sequence and pseudo-random number sequence was analysed. The analysis results show that when 1,000,000 simulations are performed, the maximum error of the calculation results of the two methods is 0.2%, and the parallel method can obtain a six-fold speedup ratio compared with the serial method, indicating that the proposed parallel method is correct, meets the accuracy requirements, and helps to improve the SCR analysis efficiency. When the number of simulations is the same, the simulation results based on the Sobol sequence are more accurate than those based on the pseudo-random number sequence. The proposed parallel method also achieves a good acceleration effect in the SCR analysis of large-scale WDS. [ABSTRACT FROM AUTHOR]

Astronomi, çok büyük boyutlu veriler ile çalışmak zorunda olan bir alandır. Bu büyük boyutlu verilerin işlenmesi ve bu verilerden bilgi çıkarımı işlemi çok uzun zaman aldığı için bu verileri modelleyen bilgisayar yazılımlarının performansı çok önemlidir. Bu çalışmada, Atacama Büyük Milimetre/Milimetre-altı Dizisi (Atacama Large Millimeter/Submillimeter Array- ALMA) isimli teleskop setinden gelen astronomik verileri modellemek için geliştirilen XCLASS (extended CASA Line Analysis Software Suite-genişletilmiş CASA Çizgi Analiz Yazılım Bölümü) adlı yazılımın çok yoğun hesaplamaların yapıldığı myXCLASS adlı bölümü, Grafik İşlemci Birimi (Graphics Processing Unit- GPU) programlama tekniği kullanılarak hızlandırılmıştır. GPU programlama ortamı olarak Birleşik Hesap Cihazı Mimarisi (Compute Unified Device Architecture- CUDA) kullanılmıştır. Uygulama, Tesla K20m isimli iki adet ekran kartı üzerinde test edilmiş, CPU- GPU çalışma süresi bakımından performans karşılaştırılması yapılmış ve ayrıntılı sonuçlar sunulmuştur. Elde edilen sonuçlar ile GPU programlama tekniği kullanılarak ALMA gibi gelişmiş gözlem araçlarından elde edilen çok büyük boyutlu astronomik verilerin modellenmesinde kullanılan yazılımlarda yüksek performans kazanımı sağlanacağı gösterilmiştir.

Molecular biology has revealed that complex life phenomena can be treated as the result of many gene interactions. Investigating these interactions and understanding the intrinsic mechanisms of biological systems using gene expression data have attracted a lot of attention. As a typical gene regulatory network (GRN) inference method, the S-system has been utilized to deal with small-scale network identification. However, it is extremely difficult to optimize it to infer medium-to-large networks. This paper proposes a novel parallel swarm intelligent algorithm, PGRNIG, to optimize the parameters of the S-system. We employed the clone selection strategy to improve the whale optimization algorithm (CWOA). To enhance the time efficiency of CWOA optimization, we utilized a parallel CWOA (PCWOA) based on the compute unified device architecture (CUDA) platform. Decomposition strategy and L 1 regularization were utilized to reduce the search space and complexity of GRN inference. We applied the PGRNIG algorithm on three synthetic datasets and two real time-series expression datasets of the species of Escherichia coli and Saccharomyces cerevisiae. Experimental results show that PGRNIG could infer the gene regulatory network more accurately than other state-of-the-art methods with a convincing computational speed-up. Our findings show that CWOA and PCWOA have faster convergence performances than WOA. [ABSTRACT FROM AUTHOR]

A general-purpose GPU (GPGPU) is employed in a variety of domains, including accelerating the spread of deep natural network models; however, further research into its effective implementation is needed. When using the compute unified device architecture (CUDA), which has recently gained popularity, the situation is analogous to use the GPUs and its memory space. This is due to the lack of a gold standard for selecting the most efficient approach for CUDA GPU parallel computation. Contrarily, as solving the least absolute shrinkage and selection operator (LASSO) regression fully consists of the basic linear algebra operations, the computation using GPGPU is more effective than other models. Additionally, its optimization problem often requires fast and efficient calculations. The purpose of this study is to provide brief introductions to the implementation approaches and numerically compare the computational efficiency of GPU parallel computation with that of the fast iterative shrinkage-thresholding algorithm for LASSO. This study contributes to providing gold standards for the CUDA GPU parallel computation, considering both computational efficiency and ease of implementation. Based on our comparison results, we recommend implementing the CUDA GPU parallel computation using Python, with either a dynamic-link library or PyTorch for the iterative algorithms.

Simulated turbulence phase screens have been widely used in a series of optical experiments, but their speed and accuracy still need to be improved. Because of that, We use sparse spectral models for atmospheric turbulence simulation, and we use the Computational Unified Device Architecture (CUDA) to accelerate the calculation process. Finally, we load the calculated turbulence model onto the Liquid Crystal Atmospheric Turbulence Simulator (LC-ATS). The atmospheric turbulence model was used as a wavefront phase screen, and that model was deployed to the Liquid Crystal Spatial Light Modulator (LC-SLM). In 512 × 512 resolution, it takes 1.36 ms per frame, and it reaches 423 Hz on the liquid crystal on silicon (LC-OS). • The sparse spectral method can be used to simulate the atmosphere turbulence data in real-time. • The Liquid Crystal Atmospheric Turbulence Simulator (LC-ATS) can simulate the atmosphere in real-time. • The Computational Unified Device Architecture (CUDA) are used on GPU to improving the speed of atmospheric data calculation. [ABSTRACT FROM AUTHOR]

We present new algorithms for the parallelization of Eulerian–Lagrangian interaction operations in the immersed boundary method. Our algorithms rely on two well-studied parallel primitives: key-value sort and segmented reduce. The use of these parallel primitives allows us to implement our algorithms on both graphics processing units (GPUs) and on other shared-memory architectures. We present strong and weak scaling tests on problems involving scattered points and elastic structures. Our tests show that our algorithms exhibit near-ideal scaling on both multicore CPUs and GPUs. [ABSTRACT FROM AUTHOR]

Biological interaction databases accommodate information about interacted proteins or genes. Clustering on the networks formed by the interaction information for finding regions highly connected could reveal the functional affinities or structural similarities between protein or gene entities. With the ever-increasing amounts of information in these databases, the runtime of a clustering task is more and more unaffordable. In this paper, we propose a heterogeneous parallel algorithm focusing on accelerating clustering tasks using distributed CPU–GPU clusters. Our parallel implementation is based on the original serial algorithm of the Markov clustering (MCL). In our parallel implementation, we utilize both the CPUs and GPUs to exploit the power of heterogeneous platforms. With the BioGRID biological interaction database, we have tested the proposed algorithm on a computer cluster equipped with NVIDIA Tesla P100 GPU accelerators. The result shows that, the algorithm is efficient in GPU memory usage and inter-node data transmission, and it can complete the clustering task in 3.2 minutes with the best speedup of 70.02 times compared to the serial counterpart.We believe our work can provide key insights for realizing fast MCL analyses on large-scale biological data, with distributed CPU–GPU computer clusters. [ABSTRACT FROM AUTHOR]

Simulation of the nonlinear mechanical response of materials with explicit representation of microstructural features is extremely challenging. These models typically involve a very large number of degrees of freedom, and are prone to convergence difficulties when searching for roots to nonlinear equilibrium equations. We focus on an idealized material model that is motivated by the microstructure of synthetic nacre: individual ‘bricks’ (representing ceramic platelets) interact through nonlinear cohesive springs (representing a small volume fraction of polymer that bonds the platelets). The model simulates composite fracture through rupture of the cohesive springs. The problem is cast in terms of energy minimization and is essentially described by ‘nearest neighbor’ interactions. The principal focus of this paper is to illustrate the computational gains achievable by the strategic marriage of robust, global Monte Carlo minimization algorithms to the graphics processing unit architecture, and to describe how they were realized on the Nvidia GPU. Results comparing the computation times for graphics processing unit and central processing unit implementations demonstrate that a new adaptive version of the simulated annealing algorithm yields a speedup of approximately 5 times, whereas the graphics processing unit implementation yields a speed-up of about 16 times over conventional four-core central processing unit implementations. The resulting speed enhancement for adaptive graphics processing unit minimization of a factor of 80 enables a far broader range of simulations than has previously been possible. Simulations involving as many as 300,000 bricks can be performed in hours, as compared to weeks required by central processing unit implementation. Many aspects of this approach are translatable to other physical problems involving energy minimization in systems with large numbers of degrees of freedom.

As high performance computing technology continues to develop rapidly, graphics processing units (GPUs) are more widely used for daily computing tasks. However, some applications, such as weather forecasting, require large-scale computation. Thus, due to the computationally intensive characteristics of the rapid radiative transfer model for general circulation models (RRTMG) in the Earth system model, this study uses GPU-related technology to accelerate computation of the model. First, two kinds of algorithms using GPU technology to accelerate the RRTMG shortwave radiation scheme (RRTMG_SW) are proposed. Then, an optimization method for data transmission between host and device is proposed. Finally, after using CUDA Fortran and CUDA C to implement these algorithms, two GPU versions of RRTMG_SW, namely CF-RRTMG_SW (CUDA Fortran version) and CC-RRTMG_SW (CUDA C version), were developed. The experimental results demonstrate that the proposed acceleration algorithms are effective. Without I/O transfer, running CC-RRTMG_SW on a NVIDIA GeForce Titan V GPU achieved a $38.88\times $ speedup when compared to a single Intel Xeon E5-2680 CPU core.

The fire control system operator’s monitor unit displays the video feed from online video cameras and fire control system-generated on-screen data overlays. The design specified in this paper and developed by the authors is a hardware implementation of GUI on-screen data overlays from various data sources to minimise the fire control system’s CPU load. The monitor unit can acquire video feeds from two cameras, with one daylight (full colour) VIS camera and one IR (monochromatic) camera and the video output generated by the fire control system computer, and overlay or mix the two video sources on-screen. Each of the video camera feeds can be processed in real time with gamma and dynamic brightness range correction. The daylight VIS camera feed processing supports video pixel change detection to substitute for a motion detection system. The fire control system monitor can also analyse the video feeds and overlay the hazy areas of the VIS camera feed output with the corresponding areas of IR camera feed output. The fire control system monitor supports IR camera feed PIP (picture-in-picture) display overlaid on the VIS camera display, mixing of both camera feeds, and panoramic view output from both video feeds. On-screen text and a moving targeting crosshair can also be displayed. The fire control system monitor communicates with its master control system via a serial bus. The monitor can also work autonomously (not connected to a master control system); commands and status output functions are implemented with a bezel control keyboard with individually backlit keys. The functionalities were developed with relatively simple hardware and software solutions.

The paper presents a discussion on the issue of possible acceleration of radiolocation signal processing algorithms in seekers using graphics processing units. A concept and implementation examples of algorithms performing digital data filtering on general purpose central and graphics processing units are introduced. The results of performance comparison of central and graphics processing units during computing discrete convolution are presented at the end of the paper.

In optical remote sensing image reconstruction, image registration is an important issue to address in order to ensure satisfactory reconstruction performance. In this study, a multi-frame image registration algorithm for high-resolution images and its parallel design method are proposed. The algorithm realizes an improved feature point detection method based on an adaptive gradient bilateral tensor filter and carries out weighted Gaussian surface sub-pixel interpolation to obtain more accurate corner positions, which better guarantees the registration accuracy. On this basis, multi-scale expansion is carried out to generate descriptors for image registration. In addition, the operation-level parallel analysis and design are carried out on a GPU platform based on compute unified device architecture (CUDA), and the memory model of the GPU is utilized reasonably. The task-level parallel analysis and design are carried out based on the GPU stream model. Moreover, based on the open multi-processing (OpenMP) platform, a multi-core CPU carries out parallel design at the operation level and task level, which realizes post-processing operations such as optical remote sensing images loading, accurate matching, and coordinate mapping, thereby effectively improving registration speed. Compared with feature point algorithms and deep learning algorithm, our algorithm and its parallel design significantly improve the registration accuracy and speed of high-resolution optical remote sensing images. [ABSTRACT FROM AUTHOR]

Atmospheric radiation physical process plays an important role in climate simulations. As a radiative transfer scheme, the rapid radiative transfer model for general circulation models (RRTMG) is widely used in weather forecasting and climate simulation systems. However, its expensive computational overhead poses a severe challenge to system performance. Therefore, improving the radiative transfer model's computational performance has significant scientific research and practical value. Numerous radiative transfer models have benefited from a widely used and powerful GPU. Nevertheless, few of them have exploited CPU/GPU cluster resources within heterogeneous high-performance computing platforms. In this paper, we endeavor to demonstrate an approach that runs a large-scale, computationally intensive, longwave radiative transfer model on a GPU cluster. First, a CUDA-based acceleration algorithm of the RRTMG longwave radiation scheme (RRTMG_LW) on multiple GPUs is proposed. Then, a heterogeneous, hybrid programming paradigm (MPI+CUDA) is presented and utilized with the RRTMG_LW on a GPU cluster. After implementing the algorithm in CUDA Fortran, a multi-GPU version of the RRTMG_LW, namely GPUs-RRTMG_LW, was developed. The experimental results demonstrate that the multi-GPU acceleration algorithm is valid, scalable, and highly efficient when compared to a single GPU or CPU. Running the GPUs-RRTMG_LW on a K20 cluster achieved a 77.78 × speedup when compared to a single Intel Xeon E5-2680 CPU core. [ABSTRACT FROM AUTHOR]

The various clustering methods are widely applied in analyzing biological interaction networks, such as the protein–protein interaction and the genetic interaction networks. With the rapid growth of these biological datasets in scale, much longer runtime is required to make cluster analyses on them. In this paper, we propose a novel parallel Markov clustering (MCL) method based on the ELLPACK-R sparse matrix format that can run on multiple graphic processing units (GPUs) equipped standalone computers. The method is implemented using the Compute Unified Device Architecture (CUDA) programming framework, and fine-grained warp-level optimization is introduced for improving the performance. The BioGRID, a large-scale and freely accessible database of protein and genetic interactions, is adopted as the dataset in the experiment. The method has been assessed on a desktop computer equipped with two NVIDIA GTX 1070 GPUs. The results show that the proposed multi-GPU method can conduct MCL clustering on the full-size BioGRID database with about 6.5 min, that is much faster than the CPU serial MCL implementation which needs almost an hour and a half execution time. [ABSTRACT FROM AUTHOR]

Sliding spotlight synthetic aperture radar (SAR) is a hot imaging mode in space-borne SAR. It can acquire high-resolution and large azimuth scene size simultaneously. The main challenges for sliding spotlight SAR processing in real time are the huge volume of raw data and complex imaging algorithm. The rapid development of the graphics processing units (GPUs) provides a new computing platform for high-performance SAR processing. This work presents a parallel processing of sliding spotlight SAR imaging algorithm on GPU via compute unified device architecture and tests the speedup of each module of the algorithm on the K20c GPU platform from the aspect of a small granularity and a large granularity of point target images. Compared with traditional CPU-based SAR imaging technology, the proposed GPU-based sliding spotlight SAR imaging obtains a speedup of 186.80 when processing data with a granularity of 16,384 × 8192 and gets a speedup of 49.72 when processing data with a granularity of 65,536 × 32,768, thereby greatly improving the real-time imaging processing.

Graphics processing unit (GPU)-based computing for climate system models is a longstanding research area of interest. The rapid radiative transfer model for general circulation models (RRTMG), a popular atmospheric radiative transfer model, can calculate atmospheric radiative fluxes and heating rates. However, the RRTMG has a high calculation time, so it is urgent to study its GPU-based efficient acceleration algorithm to enable large-scale and long-term climatic simulations. To improve the calculative efficiency of radiation transfer, this paper proposes a GPU-based acceleration algorithm for the RRTMG longwave radiation scheme (RRTMG_LW). The algorithm concept is accelerating the RRTMG_LW in the g-point dimension. After implementing the algorithm in CUDA Fortran, the G-RRTMG_LW was developed. The experimental results indicated that the algorithm was effective. In the case without I/O transfer, the G-RRTMG_LWon one K40 GPU obtained a speedup of 30.98x over the baseline performance on one single Intel Xeon E5-2680 CPU core. When compared to its counterpart running on 10 CPU cores of an Intel Xeon E5-2680 v2, the G-RRTMG_LW on one K20 GPU in the case without I/O transfer achieved a speedup of 2.35x. [ABSTRACT FROM AUTHOR]

Traditional split look-up tables (S-LUT) technology is modified and redesigned to generate computer generated hologram (CGH) with higher computation efficiency without increasing storage space. Two main aspects are involved in the proposedmethod (1) The content of original vertical modulation factor look-up tables is changed to record the contribution of space points at same Y coordinate to every row pixels of the holographic plane. With the new look-up tables the number of online CGH computation is greatly reduced and the implementation of CUDA(compute unified device architecture) parallel computation is simplified as well. (2) Phase iteration technology is introduced into the generation of look-up tables offline by substituting addition and subtraction for time-consuming square, square root operation. As a result, the generation efficiency of offline look-up tables is significantly improved. Experiment on generating CGH with two kinds of 3D point clouds shows that, the time for the generation of look-up tables offline in the proposed algorithm is saved about 1-1.5 times, compared to original algorithm, without increasing storage. In addition, the time for the generation of CGH online with new look-up tables is save at least about 15% and parallelization can be easily implemented as well. The proposed algorithm turn out to be valuable for 3D display based on digital holography technology. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) images play an increasingly important role in ocean environmental monitoring and research. However, SAR images are inherently corrupted by speckle noise. SAR ocean images have some unique characteristics. The signatures of ocean phenomena in SAR images mainly exhibit as stripe or plaque shaped features. These features typically have a high degree of self-similarity or redundancy. The nonlocal means (NLM) method can measure the structural similarity between different image patches and take advantage of redundant information in images. Considering that the NLM algorithm is computationally intensive and time-consuming, an accelerated NLM algorithm for SAR ocean image despeckling is proposed in this paper. A method is used to discriminate between texture and flat pixels in SAR images. Large similarity and search windows are used on texture pixels, whereas small similarity and search windows are used on flat pixels. Furthermore, the improved NLM algorithm is accelerated by a graphic processing unit (GPU) based on the compute unified device architecture (CUDA) parallel computation framework. The computational efficiency is improved by approximately 200 times. [ABSTRACT FROM AUTHOR]

Sliding spotlight synthetic aperture radar (SAR) is a hot imaging mode in space-borne SAR. It can acquire high-resolution and large azimuth scene size simultaneously. The main challenges for sliding spotlight SAR processing in real time are the huge volume of raw data and complex imaging algorithm. The rapid development of the graphics processing units (GPUs) provides a new computing platform for high-performance SAR processing. This work presents a parallel processing of sliding spotlight SAR imaging algorithm on GPU via compute unified device architecture and tests the speedup of each module of the algorithm on the K20c GPU platform from the aspect of a small granularity and a large granularity of point target images. Compared with traditional CPU-based SAR imaging technology, the proposed GPU-based sliding spotlight SAR imaging obtains a speedup of 186.80 when processing data with a granularity of 16,384 × 8192 and gets a speedup of 49.72 when processing data with a granularity of 65,536 × 32,768, thereby greatly improving the real-time imaging processing. [ABSTRACT FROM AUTHOR]

Climatic simulations rely heavily on high-performance computing. As one of the atmospheric radiative transfer models, the rapid radiative transfer model for general circulation models (RRTMG) is used to calculate the radiative transfer of electromagnetic radiation through a planetary atmosphere. Radiation physics is one of the most time-consuming physical processes, so the RRTMG presents large-scale and long-term simulation challenges to the development of efficient parallel algorithms that fit well into multicore clusters. This paper presents a method for improving the calculative efficiency of radiation physics, an RRTMG long-wave radiation scheme (RRTMG_LW) that is accelerated on a graphics processing unit (GPU). First, a GPU-based acceleration algorithm with one-dimensional domain decomposition is proposed. Then, a second acceleration algorithm with two-dimensional domain decomposition is presented. After the two algorithms were implemented in Compute Unified Device Architecture (CUDA) Fortran, a GPU version of the RRTMG_LW, namely G-RRTMG_LW, was developed. Results demonstrated that the proposed acceleration algorithms were effective and that the G-RRTMG_LW achieved a significant speedup. In the case without I/O transfer, the 2-D G-RRTMG_LW on one K40 GPU obtained a speed increase of 18.52× over the baseline performance on a single Intel Xeon E5-2680 CPU core. [ABSTRACT FROM AUTHOR]

As more generators and loads are integrated by power electronic converters with complicated controls, electromagnetic transients (EMTs) simulation becomes an important tool for studying dynamic characteristics of large‐scale power systems. To accelerate system‐level EMT simulations, a fine‐grained parallel algorithm on graphics processing units (GPU) is proposed. By decomposing the computational models of the EMT simulation into heterogeneous, homogeneous and network solution computations, the simulations are mapped into three unified GPU kernels. To incorporate control signals and non‐linear features of electrical components, heterogeneous computations are formulated as layered direct acyclic graphs (LDAG) of primitive operations. An LDAG kernel is designed to carry out theses primitive operations efficiently by grouped threads. Then, homogeneous computations for state updates of electrical components are modelled as sets of fused multiply‐add (FMA) operations, which are concurrently processed by an FMA kernel. Moreover, a hybrid network solution kernel is designed to solve the network equations, which can adaptively select dense or sparse solvers. Large‐scale test systems are created and simulated on an NVIDIA K20x GPU. The results show that the proposed GPU‐based EMT simulations are accurate and achieve 10x speedups over the CPU‐based ones. [ABSTRACT FROM AUTHOR]

Copyright of Journal of National University of Defense Technology / Guofang Keji Daxue Xuebao is the property of NUDT Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper sets up a parameterized model of engine-bladed disk system. Then a method of mistuned parameter identification is proposed, which consists of static frequency test of blades, dichotomy, and finite element analysis. Dynamic genetic ant colony optimization combined algorithm is presented which is used as a bladed arrangement optimization method. It combines the advantage of ant colony algorithm and genetic algorithm. By this way, the local and global search ability is introduced and the disadvantage of a single algorithm is overcome. The improved parallel algorithm comes up with graphics processing unit based on compute unified device architecture and its performance is analyzed. The results show that using optimization results can reduce the amplitude and localization of forced vibration response of bladed-disk system. At the same time, optimization based on compute unified device architecture framework can improve computing speed. Providing support for engineering application in terms of effectiveness and efficiency. [ABSTRACT FROM AUTHOR]

This paper presents the Distributed Heterogeneous Inspection System (DHIS), which comprises two CUDA workstations and is equipped with CPU distributed computing, CPU concurrent computing, and GPU concurrent computing functions. Thirty-two grayscale images, each with 5,000×12,288 pixels and simulated defect patterns, were created to evaluate the performances of three system configurations: (1) DHIS; (2) two CUDA workstations with CPU distributed computing and GPU concurrent computing; (3) one CUDA workstation with GPU concurrent computing. Experimental results indicated that: (1) only DHIS can satisfy the time limit, and the average turnaround time of DHIS is 37.65% of the time limit; (2) a good linear relationship exists between the processing speed ratio and the instruction sequence quantity ratio. [ABSTRACT FROM AUTHOR]

An accelerating method based on CUDA（compute unified device architecture）with Kepler GPU（graphics processing unit）was proposed to speed up the DFT（discrete Fourier transform）processing in the communication simulation platform.Based on this method，the whole DFT processing was split into subtasks named molecular-subtasks corresponding to communication links and a molecular-subtask was further split into smaller parallel subtasks named atomic-subtasks which correspond to the DFT processing in a link.Then，the atomic-subtasks were processed in parallel by the threads in a GPU kernel function，as well as the molecular-subtasks were processed in parallel via several GPU kernel functions to shorter the simulation time.Simulation results show this method can speed up the DFT processing more than 300 times compared with single thread CPU program and 3 times compared with traditional GPU program.

Parallel discovery of inherent clusters using massively threaded architectures is the solution for handling computational challenges raised by fat datasets in cluster analysis. The Graphics Processing Unit and Compute Unified Device Architecture form a convincing platform to parallelize clustering algorithms. The parallel K-means algorithm aims at increasing the speedup, but often faces the hitch of falling into local minima. The heuristic search procedure to discover the global optima in the solution space is known as Tabu Search. The K-means clustering solution is fine-tuned by applying parallel implementation of Tabu Search K-means clustering in order to increase efficacy. The aim is to combine optimization characteristic of Tabu Search for calculation of centroids with clustering attitude of K-means and to enhance the solution using processing power of GPU. The parallelization strategy used exhibits increase in speedup. The parallel Tabu-KM algorithm is tested on standard datasets, and performance is compared with sequential K-means, parallel K-means and sequential Tabu K-means algorithms. The experimental results confirm yet another parallelization technique to unravel data clustering problems. [ABSTRACT FROM AUTHOR]

In recent years, while the collected data are increased, we need effective computation methods to process these data. Due to the fact that most of the real world problems are difficult to solve, swarm intelligence and evolutionary computation algorithms are interested because they guarantee the near optimal solution for the problem in a reasonable time but not guarantee the optimal solution. In another perspective, if the data or process can be parallelized, the parallel computation is a good choice instead of serial programming approaches in terms of time effectiveness. In this study, the tree-seed algorithm, which is a recently proposed population-based iterative search algorithm, is implemented within CUDA platform in parallel. The performance of the parallel version of the algorithm has been investigated on the benchmark functions and compared with the performance of the serial version of the algorithm. The dimensionality of the problems is taken as 10 and the performance analysis and comparisons have been conducted under the condition of different sizes of the population. Experimental studies show that the parallel version of the algorithm is accelerated to 184.65 times in accordance with the serial version of the algorithm on some problems. [ABSTRACT FROM AUTHOR]

Differential interferometric synthetic aperture radar (DInSAR) is an effective technique to measure the surface displacement caused by strong earthquakes. However, in the area with respect to the most significant deformation along the fault zone, the coherence between pre- and post-earthquake SAR images is completely lost because of the earthquake-induced violent and chaotic destruction on the land surface and consequently, no surface displacement data can be measured. In our previous work, anadaptive local kriging(ALK) was proposed to solve this problem. However, according to the spatial distribution of the reference points for the interpolation procedure of ALK, using moving window with different sizes through all the pixels in the image is time consuming. Thus, ahigh-performance computing(HPC) is needed. As a result, a parallel image interpolation approach, referred as thegraphics processing unit(GPU) based ALK method, to retrieving slant-range surface motion maps derived by DInSAR is proposed in this study. The proposed GPU-ALK makes use of the parallelism of the original ALK architectures to retrieve the loss of data through the fault zone. In this paper, an HPC GPU-ALK is proposed to speed up the interpolation. It makes use of the performance profiling to analyze the serial version of ALK and perform a parallel GPU computation for reducing the computation time efficiently. By employing one NVIDIA TITAN GPU, the proposed GPU-ALK achieves a speedup of 104.32× compared to its CPU counterpart. [ABSTRACT FROM AUTHOR]

Abstract Optical Burst Switching (OBS) is a promising technology for next generation of Transparent Optical Networks (TON). However, many scientific challenges remain to be overcome such as the problem of Burst Routing and Wavelength Assignment (BRWA) with several conflicting objectives and constraints. In this paper, we first formulate the BRWA as a Multi Objective Integer Linear Programming (MO-ILP) optimization problem. In the formulated problem, the proposed BRWA policy will satisfy several constraints in order to guarantee a high-speed management of processes, required by the transparent optical traffic. Then, since the obtained ILP problem contains a large number of optical constraints and conflicting objectives, we propose to use an exact parallel Neural Hierarchical (epNH) MO-ILP solution with Graphics Processing Unit (GPU) parallel implementation using Compute Unified Device Architecture (CUDA). This also allows doing a concurrent search for multiple solutions, reducing processing cost, making hybrid interfaces to other search techniques, and achieving better overall effectiveness. In addition, our architecture based on Artificial Neural Networks (ANN) allows flexibility and scalability. The processing time remains fixed regardless of the input size. Our BRWA GPU-based epNH MO-ILP solver is based on the joint use of advanced MO-ILP optimization methods, ANN large-scale inherent parallelism and CUDA-GPU High-Performance Computing (HPC) architecture. [ABSTRACT FROM AUTHOR]

In this paper, we employ the lattice Boltzmann method implemented on compute unified device architecture-enabled graphical processing unit to investigate the multiphase fluid pipe flow. The basics of lattice Boltzmann method as well as the Shan–Chen multiphase model and the fundamentals of graphical processing unit with compute unified device architecture are thoroughly introduced. The procedure of implementation of lattice Boltzmann method on graphical processing unit and the comparison of the computing performance between graphical processing unit and CPU are presented. It is demonstrated that the graphical processing unit-based lattice Boltzmann method has remarkable advantages over CPU especially with selected appropriate parameters. The results of validation cases agree well with previous numerical results or analytical solutions. The vertical and horizontal multiphase pipe flow are simulated and discussed. [ABSTRACT FROM AUTHOR]

The nearest neighbor partitioning (NNP) method is a high performance approach which is used for improving traditional neural network classifiers. However, the construction process of NNP model is very time-consuming, particularly for large data sets, thus limiting its range of application. In this study, a parallel NNP method is proposed to accelerate NNP based on Compute Unified Device Architecture(CUDA). In this method, blocks and threads are used to evaluate potential neural networks and to perform parallel subtasks, respectively. Experimental results manifest that the proposed parallel method improves performance of NNP neural network classifier. Furthermore, the application of parallel NNP in performance evaluation of cement microstructure indicates that the proposed approach has favorable performance. [ABSTRACT FROM AUTHOR]

Transparent Optical Networks has been known as a potential solution for high speed flexible optical backbone networks since a long time ago. However, many scientific challenges remain to be overcome such as the problem of Optical Burst Grooming (OBG) with several conflicting objectives and constraints. OBG is a fundamental problem in the engineering, control, and management of optical traffic networks, and arises in most network design applications, including optical burst routing, survivability design, and traffic scheduling. The traffic grooming problem is to coalesce several high and low speed sub bursts close in time together to form a larger burst that will be switched as one unit. In this paper, we first formulate OBG as a Multi Objective Integer Linear Programming (MO-ILP) optimization problem. Then we propose to use a parallel and hierarchical solver based on Artificial Neural Networks (ANN) with Graphics Processing Unit (GPU) parallel implementation using Compute Unified Device Architecture (CUDA). The processing time remains fixed regardless of the input size and the number of optical constraints and conflicting objectives. Our OBG solver benefits of the joint use of advanced ILP-MO optimization methods, ANN large-scale inherent parallelism and CUDA-GPU High-Performance Computing (HPC) architecture. Through a comprehensive simulation study, we show that, our proposed grooming approach can significantly improve the optical traffic performance, resulting in less contention and low burst lost probability. [ABSTRACT FROM AUTHOR]