Graphics Processing Unit Technology Research Articles

Numerical Simulation of the Large‐Scale Huangtian (China) Landslide‐Generated Impulse Waves by a GPU‐Accelerated Three‐Dimensional Soil‒Water Coupled SPH Model

AbstractThis work presents an improved soil‒water coupling model to simulate landslide‐generated impulse waves (LGIWs) in a unified smoothed particle hydrodynamics (SPH) framework, where both water flow and landslide motions are modeled by SPH using an interface coupling technique. Graphics processing unit technology based on an open‐source platform DualSPHysics is chosen to employ the landslide dynamics and soil‒water interface coupling to achieve the capability of large‐scale simulation and high‐resolution modeling for three‐dimensional LGIW problems. A subaerial landslide‐generated water waves, is simulated to demonstrate the accuracy and ability of this model. The Huangtian LGIW is then simulated to reproduce the entire disaster chain, including landslide dynamics, fluid‒solid interaction, and surge wave generation. Particle resolution dependence is examined, giving a particle distance of 5.0 m, which can provide a converged landslide deposit and surge wave. The simulation shows that in the Huangtian LGIW, the landslide deposit volume was approximately 41.6 million m3 (600 m width, 768 m length, and 400 m above the still water level), with an immersed landslide volume of 11.3 million m3; for the surge wave, the maximum wave and run‐up heights were 34.3 and 48 m, respectively. These results are within the estimated ranges of both the landslide and surge wave according to limited field survey data. The case study of the Huangtian LGIW provides a typical reference of how to reproduce a reliable whole process of large scale multi‐physical and multiscale LGIW, including full information of landslide dynamics, interface coupling behavior, and surge wave characteristics.

Efficient GPU Cloud architectures for outsourcing high-performance processing to the Cloud

AbstractThe world is becoming increasingly dependant in computing intensive applications. The appearance of new paradigms, such as Internet of Things (IoT), and advances in technologies such as Computer Vision (CV) and Artificial Intelligence (AI) are creating a demand for high-performance applications. In this regard, Graphics Processing Units (GPUs) have the ability to provide better performance by allowing a high degree of data parallelism. These devices are also beneficial in specialized fields of manufacturing industry such as CAD/CAM. For all these applications, there is a recent tendency to offload these computations to the Cloud, using a computing offloading Cloud architecture. However, the use of GPUs in the Cloud presents some inefficiencies, where GPU virtualization is still not fully resolved, as our research on what main Cloud providers currently offer in terms of GPU Cloud instances shows. To address these problems, this paper first makes a review of current GPU technologies and programming techniques that increase concurrency, to then propose a Cloud computing outsourcing architecture to make more efficient use of these devices in the Cloud.

Applying a Deep Learning Neural Network to Gait-Based Pedestrian Automatic Detection and Recognition

Gait recognition is a noncontact biometric procedure that determines the identity or health status of a person by analyzing his or her walking posture and habits, including skeletal and joint movements. The most remarkable feature of this method is the possibility of conducting recognition without demanding much cooperation from participants. Therefore, this recognition technique has attracted much attention from scholars. Additionally, because of the rapid development of graphics processing unit technology, related hardware and computation performance, the applications of deep-learning technology are considerably enhanced. The objective of this study was to apply a deep neural network (DNN), which employs deep-learning technology, to achieve gait-based automatic pedestrian detection and recognition. In contrast to using wearable devices to precisely capture skeletal and joint movements, pedestrian color-image sequences were used as input in this study. Subsequently, a pretraining convolutional neural network (CNN) was employed to capture pedestrian location and extract pedestrian dense optical flow to serve as concrete low-level feature inputs. Then, a finely-tuned DNN based on the wide residual network was employed to extract high-level abstract features. In addition, to overcome the difficulty of obtaining local temporal features by using a 2D CNN, part of the 3D convolutional structure was introduced into the CNN. This design enabled use of limited memory to acquire more effective features and enhance the DNN performance. The experimental results show that the proposed method has exceptional performance for pedestrian detection and recognition.

Inference of dynamic spatial GRN models with multi-GPU evolutionary computation.

Reverse engineering mechanistic gene regulatory network (GRN) models with a specific dynamic spatial behavior is an inverse problem without analytical solutions in general. Instead, heuristic machine learning algorithms have been proposed to infer the structure and parameters of a system of equations able to recapitulate a given gene expression pattern. However, these algorithms are computationally intensive as they need to simulate millions of candidate models, which limits their applicability and requires high computational resources. Graphics processing unit (GPU) computing is an affordable alternative for accelerating large-scale scientific computation, yet no method is currently available to exploit GPU technology for the reverse engineering of mechanistic GRNs from spatial phenotypes. Here we present an efficient methodology to parallelize evolutionary algorithms using GPU computing for the inference of mechanistic GRNs that can develop a given gene expression pattern in a multicellular tissue area or cell culture. The proposed approach is based on multi-CPU threads running the lightweight crossover, mutation and selection operators and launching GPU kernels asynchronously. Kernels can run in parallel in a single or multiple GPUs and each kernel simulates and scores the error of a model using the thread parallelism of the GPU. We tested this methodology for the inference of spatiotemporal mechanistic gene regulatory networks (GRNs)-including topology and parameters-that can develop a given 2D gene expression pattern. The results show a 700-fold speedup with respect to a single CPU implementation. This approach can streamline the extraction of knowledge from biological and medical datasets and accelerate the automatic design of GRNs for synthetic biology applications.

Granular layEr Simulator: Design and Multi-GPU Simulation of the Cerebellar Granular Layer.

In modern computational modeling, neuroscientists need to reproduce long-lasting activity of large-scale networks, where neurons are described by highly complex mathematical models. These aspects strongly increase the computational load of the simulations, which can be efficiently performed by exploiting parallel systems to reduce the processing times. Graphics Processing Unit (GPU) devices meet this need providing on desktop High Performance Computing. In this work, authors describe a novel Granular layEr Simulator development implemented on a multi-GPU system capable of reconstructing the cerebellar granular layer in a 3D space and reproducing its neuronal activity. The reconstruction is characterized by a high level of novelty and realism considering axonal/dendritic field geometries, oriented in the 3D space, and following convergence/divergence rates provided in literature. Neurons are modeled using Hodgkin and Huxley representations. The network is validated by reproducing typical behaviors which are well-documented in the literature, such as the center-surround organization. The reconstruction of a network, whose volume is 600 × 150 × 1,200 μm3 with 432,000 granules, 972 Golgi cells, 32,399 glomeruli, and 4,051 mossy fibers, takes 235 s on an Intel i9 processor. The 10 s activity reproduction takes only 4.34 and 3.37 h exploiting a single and multi-GPU desktop system (with one or two NVIDIA RTX 2080 GPU, respectively). Moreover, the code takes only 3.52 and 2.44 h if run on one or two NVIDIA V100 GPU, respectively. The relevant speedups reached (up to ~38× in the single-GPU version, and ~55× in the multi-GPU) clearly demonstrate that the GPU technology is highly suitable for realistic large network simulations.

Fast Implementation of SHA-3 in GPU Environment

Recently, Graphic Processing Units (GPUs) have been widely used for general purpose applications such as machine learning applications, acceleration of cryptographic applications (especially, blockchains), etc. The development of CUDA makes this General-Purpose computing on GPU possible. In particular, currently GPU technology has been widely used for server-side applications so as to provide fast and efficient service to a number of clients. In other words, servers need to process a large amount of user data and execute authentication process. Verifying the integrity of transmitted data is essential for ensuring that the data is not modified during transmission. Hash functions are the cryptographic algorithm which can verify the integrity of data and there are SHA-1, SHA-2, and SHA-3 standard hash functions. In 2015, Keccak algorithm was selected for SHA-3 competition by NIST. However, until now, software implementations of SHA-3 have not provided enough performance for various applications. In addition, SHA-3 and SHAKE using SHA-3 are being used in many Post-Quantum Cryptosystems (PQC) submitted to NIST PQC competition. Therefore, SHA-3 optimization research is required in the software environment. We propose an optimized SHA-3 software implementation on GPU environment. For performance efficiency, we propose several techniques including optimization of SHA-3 internal process, inline PTX optimization, optimized memory usage, and the application of asynchronous CUDA stream. As a result of applying the proposed optimization method, our SHA-3(512) (resp. SHA-3(256)) implementation without CUDA stream provides a maximum throughput of 88.51 Gb/s (resp. 171.62 Gb/s) on RTX2080Ti GPU. Furthermore, without the application of CUDA stream, our SHA-3(512) software on GTX1070 provides about 49.73% improved throughput compared with the previous best work on GTX1080, which shows the superiority of our proposed optimization methods. Our optimized SHA-3 software on GPU can be efficiently used for block-chain applications and several PQCs (especially, key generation process in Lattice-based cryptosystems).

GPU-RRTMG_SW: Accelerating a Shortwave Radiative Transfer Scheme on GPU

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× speedup when compared to a single Intel Xeon E5-2680 CPU core.

Molecular dynamics simulation of single-walled carbon nanotubes based on GPU

A graphics processing unit (GPU)-based molecular dynamics simulation of single-walled carbon nanotube (SWCNT) molecules is proposed. Under the background of GPU technology, the differential unit and stress of an SWCNT molecule are analyzed first. The stress of the SWCNT molecule at a distance from the end of the SWCNT molecule is obtained by integration, and the calculation of the critical length of stress transfer at the interface of the SWCNT molecule is completed. The SWCNT molecule is divided into two steps by temperature control and using a molecular dynamics simulation program. In this paper, GPU technology is used to derive the velocity Verlet algorithm on the basis of the Verlet algorithm, which completes the optimization design of the numerical algorithm of the differential equation of motion of SWCNTs. Finally, through the SWCNTs, the optimization design of the numerical algorithm of the differential equation of motion of SWCNTs is completed. The molecular dynamics simulation of SWCNTs based on GPU is realized by the equilibrium molecular dynamics simulation of nanotubes and the non-equilibrium molecular dynamics simulation of SWCNTs. The simulation results show that the molecular dynamics simulation method based on GPU can reduce the eccentricity of SWCNT molecules.

Model construction of photon propagation based on the geometrical symmetries and GPU technology for the quad-head PET system

Compact quad-head PET systems have been developed by many researchers. This compact configuration leads to depth of interaction (DOI) blurring, which greatly reduces the quality of the reconstructed image. How to accurately model the photon propagation is the key to relieving DOI blurring. In this study, photon propagation is modelled by using the multi-ray method. As we all know, the precision of the multi-ray model is proportional to its computational complexity. Focusing on this problem, we have developed the symmetries of the quad-head PET system according to the different structural features of the opposite and adjacent detector heads. Given all the symmetrical properties, the construction of the photon propagation model becomes a trade-off between the calculation precision and computational complexity of the multi-ray method. In addition, when photon propagation is stored in the system response matrix (SRM), the storage cost is also reduced largely. In the proposed quad-head PET system, the SRM is compressed about 559 times, and the matrix scale is approximately 1.7 GB. The calculation of the SRM can be finished in 60 s when the graphics processing unit (GPU) acceleration technology is adopted. In order to verify the imaging performances, several simulated experiments are carried out. The metrics FWHM, location accuracy, Contrast Recovery (CR), and Coefficient of Variation (COV) are adopted to quantify the imaging results. We find that the mean location difference can reach 1.29% based on a simulated multi-spheres experiment. Through this experiment, the FWHMs can reach to about 1.0 mm in the center or close to the edge of the field of view (FOV). In the image quality experiment, the CR can up to 0.95 in the hot region, and the COV reaches 5.82% in the background region.

A multi-GPU biclustering algorithm for binary datasets

Graphics Processing Units technology (GPU) and CUDA architecture are one of the most used options to adapt machine learning techniques to the huge amounts of complex data that are currently generated. Biclustering techniques are useful for discovering local patterns in datasets. Those of them that have been implemented to use GPU resources in parallel have improved their computational performance. However, this fact does not guarantee that they can successfully process large datasets. There are some important issues that must be taken into account, like the data transfers between CPU and GPU memory or the balanced distribution of workload between the GPU resources. In this paper, a GPU version of one of the fastest biclustering solutions, BiBit, is presented. This implementation, named gBiBit, has been designed to take full advantage of the computational resources offered by GPU devices. Either using a single GPU device or in its multi-GPU mode, gBiBit is able to process large binary datasets. The experimental results have shown that gBiBit improves the computational performance of BiBit, a CPU parallel version and an early GPU version, called ParBiBit and CUBiBit, respectively. gBiBit source code is available at https://github.com/aureliolfdez/gbibit.

Implementation and optimization of GPU‐based parallel one‐step leapfrog ADI‐FDTD for far‐field scattering problems

The one-step leapfrog alternative-direction-implicit finite-difference time-domain (ADI-FDTD), free from the Courant-Friedrichs-Lewy (CFL) stability condition and sub-step computations, is efficient when dealing with fine grid problems. However, solution of the numerous tridiagonal systems still imposes a great computational burden and makes the method hard to execute in parallel. In this paper, we proposed an efficient graphic processing unit (GPU)-based parallel implementation of the one-step leapfrog ADI-FDTD for the far-field EM scattering simulation of objects, in which we present and analyze the manners of calculation area division and thread allocation and a data layout transformation of z components is proposed to achieve better memory access mode, which is a key factor affecting GPU execution efficiency. The simulation experiment is carried out to verify the accuracy and efficiency of the GPU-based implementation. The simulation results show that there is a good agreement between the proposed one-step leapfrog ADI-FDTD method and Yee's FDTD in solving the far-field scattering problem and huge benefits in performance were encountered when the method was accelerated using GPU technology.

Laser Scanned Real Environment for Intelligent Virtualization of Crane Lifting

BackgroundThis paper presents a Tower Crane Intelligent Lifting Path Planner (TCILP) for the lifting task in the environments which are highly complex. The environment can be old industrial plants existing over many decades, or sites with tentative storage spaces. Often, these environments do not have workable digital models or are impractical to have 3D representations are impractical. MethodsThe current investigation introduces the cutting-edge technology of laser scanning to convert the real environments into point cloud format regarded as the virtualized version of the construction sites or plants. The challenge is to deal with the huge data set of point clouds of multiple scans in order to produce a complete virtualized model. The tower crane is also virtualized for the path planning purpose. A parallelized Genetic Algorithm is employed to address the intelligent path planning issue for the lifting task of tower cranes in complicated environments taking the advantage of Graphics Processing Unit technology, which has high computing performance yet low cost. ResultsOptimal lifting paths are generated in several seconds.

Use of deep learning algorithms for real-time detection of vessels in confined spaces using the Tensorflow framework

Over 4515 small boat accidents were registered in the United State of America in 2012, resulting in 651 causalities and 22% of the accidents took place between two boats. It is, therefore, one of the most interesting applications for image analysis and recognition using deep learning, collision avoidance in passenger boats. Advances in parallel computing, graphic processing unit technology and deep learning have facilitated real-time image processing. The main objective of this study was to compare the performance metrics for different deep learning algorithms using pre-trained data sets. The algorithms used were: faster region-based convolutional neural networks, region-based fully convolutional network, and single shot multibox detector using the feature extractors: residual neural network, inception and convolutional neural networks for mobile vision applications to detect generic boats in confined waterways. These models were coded in Python programming language, using the framework Tensorflow and OpenCV library for image processing. The algorithms were pre-trained using the free images database posted on the web, Microsoft COCO. The use of these pre-trained models allowed making use of computers without graphic processing unit. As a result, it was found that the faster region-based convolutional neural networks and region-based fully convolutional network method compared to the single shot multibox detector method offer a small advantage precision if speed detection is not required, but the single shot multibox detector method is useful for case detectors in real time, however it did not perform as accurate when detecting small objects.

GPU Accelerated Intuitionistic Fuzzy and Otsu Algorithms for Foreign Leaf Detection in Cotton

The foreign substances, arising during the production and shaping of wool and cotton raw materials that are used in textile and cotton gin factories or coming from the outside, decrease considerably the quality of the obtained fabric or yarn. Nowadays, a different methods are used to separate foreign substances in the textile sector, most of these methods are not efficient in terms of speed and quality. Computerized vision systems play a vital role in the field of textiles as in other fields. In this study, Intuitionistic Fuzzy Algorithm is used to define the foreign substances in the images that obtained from a camera. CPU (Central Processing Unit) based applications have speed problems due to the structure of the algorithm. For this reason, GPU (Graphics Processing Unit) technology was used to overcome the speed problem. The otsu algorithm generates a dynamic threshold from the numerical values of the image obtained using the Intuitionistic fuzzy algorithm. By this means, the threshold value of each frame obtained from the camera was calculated on real time and implemented on the image timely. These algorithms were accelerated maximum 262 times using NVIDIA GTX 480 GPU supported display card.

GPU technology in the high frame rate ultrasound imaging era

As a rapidly maturing class of many-core parallel computing hardware, graphical processing unit (GPU) has enjoyed a surge of interest in the medical ultrasound community in the past decade. In particular, the software-level programmability of GPUs has significantly lowered the entry barrier for ultrasound imaging researchers (who might not be parallel computing specialists) to pursue fast realization of novel imaging algorithms that have known theoretical potential but have yet to demonstrate their real-time feasibility. This presentation shall highlight how GPUs have emerged as a new computing workhorse in realizing various ultrasound imaging innovations. Of particular note is the enabling role that GPUs have played in fostering practical pursuit of high-frame-rate ultrasound (HiFRUS) imaging innovations that are based on direct processing of pre-beamformed radiofrequency (RF) data acquired from individual array elements. Using the state-of-art GPU technology, it is readily possible to achieve >1,000 fps HiFRUS beamforming throughput. GPU can also play a pivotal role in facilitating fast realization of computationally intensive HiFRUS algorithms, such as adaptive beamforming, color-encoded flow speckle imaging, and eigen-processing. The real-time performance of these GPU computing kernels will be discussed, and their practical implementation on software-oriented open-platform research systems will be presented.

Parallelized Monte-Carlo dosimetry using graphics processing units to model cylindrical diffusers used in photodynamic therapy: From implementation to validation.

The Monte-Carlo method is the standard method for computing the dosimetry of both ionizing and non-ionizing radiation. Because this technique is highly time-consuming in conventional implementations, several improvements have recently been developed to speed-up simulations. Among the improvements, the use of graphics processing units (GPU) to parallelize algorithms provides a cost-efficient solution to accelerate the Monte-Carlo method. Parallel implementation of Monte-Carlo using GPU technology is described in the context of photodynamic therapy (PDT) dosimetry. This algorithm has been optimized to compute light emitted from optical fibers with cylindrical diffusers that are used in interstitial PDT applications. A comparison of the experimental measurements used to assess the results of the Monte-Carlo method is detailed. Illumination profiles of several commercially available diffusers are measured using an optical phantom that mimics the optical properties of the brain. Additionally, this Monte-Carlo method is compared to ex-vivo measurements made by a device dedicated to intraoperative PDT treatment of brain tumors. The results of the GPU Monte-Carlo validation are in accordance with the recommendations of the American Association of Physicists in Medicine. The acceleration obtained with the GPU implementation is in accordance with the literature and is sufficiently fast to be integrated in a treatment planning system dedicated to planning routine clinical interstitial PDT treatments.

Efficient fingerprint matching using GPU

Graphical processing unit (GPU) has proven a beneficial tool in handling computationally intensive algorithms, by introducing massive parallelism in the calculations. In this study, an effective and low-cost fingerprint identification (FI) solution is proposed that can exploit the parallel computational power of GPU proficiently. It is achieved by mapping a generalised minutia neighbour-based novel encoding and matching algorithm on low-cost GPU technology. The proposed solution achieves high accuracy in comparison with two open source matchers and it is shown to be scalable by comparing matching performance on different GPUs. The proposed GPU implementation employs multithreading and loop unrolling, which minimises the use of nested loops and avoids sequential matching of encoded minutia features. After a thorough and careful designing of data structures, memory transfers and computations, a GPU-based fingerprint matching system is developed. It achieves on average 50,196 fingerprint matches per second on a single GPU. As compared to the sequential central processing unit implementation, the proposed system achieves a speed up of around 92 times, while maintaining the accuracy. The proposed system with matcher integrated on GPU can be considered as a good, low-cost, robust and efficient solution for large-scale applications of automated FI systems.

A Virtualization Desktop Application Platform of the Virtualized Graphics Processing Unit Technology on FlexPod Architecture

Intelligent equipment and network application based on Artificial Intelligence, Big Data, Internet of Things and other emerging technologies are becoming one of the most important efforts in the construction of experimental and training information infrastructure in Colleges and Universities. Aiming at the problem that the application environment of teaching experiment and training in Colleges and Universities of Art and Design requires higher performance ability of virtual desktop and graphics processing, the specific characteristics on training room of art and design, and the design ideas and principles of virtualization desktop technology, are analysed on the unique perspective of specific projects of Shanghai Institute of Visual Arts. A virtualization desktop application platform of the Virtualized Graphics Processing Unit Technology on FlexPod Architecture is put forward. With the application of this platform, the performance of the desktop is improving, and the complexity of IT management is reducing, and the model for the application of the virtual desktop technology for colleges and universities of art and design is provided.

GPU-Based High-performance Imaging for Mingantu Spectral RadioHeliograph**Supported by the National Key Research and Development Program of China (2016YFE0100300), the Joint Research Fund in Astronomy (No. U1531132, U1631129, U1231205) under cooperative agreement between the National Natural Science Foundation of China (NSFC) and the Chinese Academy of Sciences (CAS), the National Natural Science Foundation of China (Nos. 11403009 and 11463003).

As a dedicated solar radio interferometer, the MingantU SpEctral RadioHeliograph (MUSER) generates massive observational data in the frequency range of 400 MHz–15 GHz. High-performance imaging forms a significantly important aspect of MUSER’s massive data processing requirements. In this study, we implement a practical high-performance imaging pipeline for MUSER data processing. At first, the specifications of the MUSER are introduced and its imaging requirements are analyzed. Referring to the most commonly used radio astronomy software such as CASA and MIRIAD, we then implement a high-performance imaging pipeline based on the Graphics Processing Unit technology with respect to the current operational status of the MUSER. A series of critical algorithms and their pseudo codes, i.e., detection of the solar disk and sky brightness, automatic centering of the solar disk and estimation of the number of iterations for clean algorithms, are proposed in detail. The preliminary experimental results indicate that the proposed imaging approach significantly increases the processing performance of MUSER and generates images with high-quality, which can meet the requirements of the MUSER data processing.

Vision‐Based Natural Frequency Identification Using Laser Speckle Imaging and Parallel Computing

AbstractThis study focuses on the identification of the natural frequencies of structures through the analysis of the speckle pattern that a laser creates and a camera records. The laser pointer spreads its light over a target area on the structure and creates the speckle pattern. The ambient vibrations affect the pattern and a camera records the changes. The stream of images is fed into a graphics processing unit (GPU). The developed parallel code includes different algorithms: the speckle contrast image (SCI), the speckle flow imaging (SFI), and an innovative application of k‐means clustering that uses the geometrical centroid of each cluster as virtual sensors. The tracking of the centroid in time domain through the images creates a vibration signal. The signals from different clusters are processed together to extract the natural frequencies of the structure. This study applies the proposed method to different sample structures both in laboratory and in the field to demonstrate how the obtained signals are reliable and easy to handle. The GPU technology enhances the performance of the entire method and allows the achievement of real‐time processing. All these features create an inexpensive, portable, and efficient tool to inspect any structure or its components.