Computing On Graphics Processing Units Research Articles

Optimization of compute unified device architecture for real-time ultrahigh-resolution optical coherence tomography

We propose an optimized signal processing scheme that utilizes the compute unified device architecture (CUDA) for real-time spectral domain optical coherence tomography (OCT). Because linear spline interpolation and the direct spectral reshaping method have low data and control dependencies, these algorithms maximally utilize graphic processing unit (GPU) resources for dispersion control. In addition, data transfer between main memory and GPU, regarded as one of the most wasteful and time-consuming processes in GPU computing, is executed in parallel with the signal processing by overlapping kernel execution and data transfers. Experimental results obtained from application of the proposed scheme to a laboratory constructed OCT system comprising five spectrally shifted SLDs indicate that the OCT system has an axial resolution of 4.8μm and transverse resolution of 13μm in air. Further, coherence artifacts are reduced by 3–14dB over the side-lobes in the point spread function. The optimization of CUDA enables OCT imaging rates up to 350kHz (A-lines/sec) with a single GTX680 GPU.

CUD@SAT: SAT solving on GPUs

The parallel computing power offered by graphic processing units (GPUs) has been recently exploited to support general purpose applications – by exploiting the availability of general API and the single-instruction multiple-thread-style parallelism present in several classes of problems (e.g. numerical simulations and matrix manipulations) – where relatively simple computations need to be applied to all items in large sets of data. This paper investigates the use of GPUs in parallelising a class of search problems, where the combinatorial nature leads to large parallel tasks and relatively less natural symmetries. Specifically, the investigation focuses on the well-known satisfiability testing (SAT) problem and on the use of the NVIDIA compute unified device architecture, one of the most popular platforms for GPU computing. The paper explores ways to identify strong sources of GPU-style parallelism from SAT solving. The paper describes experiments with different design choices and evaluates the results. The outcomes demonstrate the potential for this approach, leading to one order of magnitude of speedup using a simple NVIDIA platform.

Optimizing the computation of a parallel 3D finite difference algorithm for graphics processing units

SummaryThis paper explores the possibilities of using a graphics processing unit for complex 3D finite difference computation via MUSTA‐FORCE and WENO algorithms. We propose a novel algorithm based on the new properties of CUDA surface memory optimized for 2D spatial locality and compare it with 3D stencil computations carried out via shared memory, which is currently considered to be the best approach. A case study was performed for the extensive generation of a time series of 3D grids of arbitrary size used in the computation of collisions between heavy nuclei in terms of relativistic hydrodynamics. It proved that implementation based on surface memory is as much as 23% faster than an equivalent implementation using shared memory. Copyright © 2014 John Wiley & Sons, Ltd.

GPU Computing to Improve Game Engine Performance

Although the graphics processing unit (GPU) was originally designed to accelerate the image creation for output to display, today's general purpose GPU (GPGPU) computing offers unprecedented performance by offloading computing-intensive portions of the application to the GPGPU, while running the remainder of the code on the central processing unit (CPU). The highly parallel structure of a many core GPGPU can process large blocks of data faster using multithreaded concurrent processing. A game engine has many components and multithreading can be used to implement their parallelism. However, effective implementation of multithreading in a multicore processor has challenges, such as data and task parallelism. In this paper, we investigate the impact of using a GPGPU with a CPU to design high-performance game engines. First, we implement a separable convolution filter (heavily used in image processing) with the GPGPU. Then, we implement a multiobject interactive game console in an eight-core workstation using a multithreaded asynchronous model (MAM), a multithreaded synchronous model (MSM), and an MSM with data parallelism (MSMDP). According to the experimental results, speedup of about 61x and 5x is achieved due to GPGPU and MSMDP implementation, respectively. Therefore, GPGPU-assisted parallel computing has the potential to improve multithreaded game engine performance.

Parallel computation of Entropic Lattice Boltzmann method on hybrid CPU–GPU accelerated system

Nowadays, an increasing number of researchers have demonstrated a tendency to choose hybrid CPU–GPU hybrid computing as a high performance computing alternative. Entropic Lattice Boltzmann method (ELBM) parallelization, like many parallel algorithms in the field of rapid scientific and engineering computing, has given rise to much attention for applications of computational fluid dynamics. This study aims to present an efficient implementation of ELBM flow simulation for the D3Q19 model in a hybrid CPU–GPU computing environment, which consists of AMD multi-core CPUs with NVIDIA Graphics Processing Units (GPUs). To overcome the GPU memory size limitation and communication overhead, we propose a set of techniques for the development of an efficient ELBM algorithm for hybrid CPU–GPU computation. Considering the contribution of computational capacity for both the CPU and GPU, an efficient load balancing model is built. The efficiency and accuracy of the proposed approach and established model are tested on a hybrid CPU–GPU accelerated system, where the intensive parts of the computation are dealt with the software framework OpenMP and CUDA. Finally, we show the comparison of resulting computational performance using a hybrid CPU–GPU approach against both a single CPU core and a single GPU device.

SU-D-BRD-03: A Gateway for GPU Computing in Cancer Radiotherapy Research

Purpose: Graphics Processing Unit (GPU) has become increasingly important in radiotherapy. However, it is still difficult for general clinical researchers to access GPU codes developed by other researchers, and for developers to objectively benchmark their codes. Moreover, it is quite often to see repeated efforts spent on developing low-quality GPU codes. The goal of this project is to establish an infrastructure for testing GPU codes, cross comparing them, and facilitating code distributions in radiotherapy community. Methods: We developed a system called Gateway for GPU Computing in Cancer Radiotherapy Research (GCR2). A number of GPU codes developed by our group and other developers can be accessed via a web interface. To use the services, researchers first upload their test data or use the standard data provided by our system. Then they can select the GPU device on which the code will be executed. Our system offers all mainstream GPU hardware for code benchmarking purpose. After the code running is complete, the system automatically summarizes and displays the computing results. We also released a SDK to allow the developers to build their own algorithm implementation and submit their binary codes to the system. The submitted code is then systematically benchmarked using a variety of GPU hardware and representative data provided by our system. The developers can also compare their codes with others and generate benchmarking reports. Results: It is found that the developed system is fully functioning. Through a user-friendly web interface, researchers are able to test various GPU codes. Developers also benefit from this platform by comprehensively benchmarking their codes on various GPU platforms and representative clinical data sets. Conclusion: We have developed an open platform allowing the clinical researchers and developers to access the GPUs and GPU codes. This development will facilitate the utilization of GPU in radiation therapy field.

Efficient GPU parallel computing of population balance—Monte Carlo method for coagulation

The population balance-Monte Carlo (PBMC) method has become increasingly popular because the discrete and stochastic nature of the MCmethod is especially suited for particle dynamics. However, for the two-particle events (typically, particle coagulation), the double looping over all simulation particles is required in normal PBMCmethods, and simulating particle coagulation is in general a challenging computational task due to its numerical complexity and the computing cost. The compute unified device architecture (CUDA) is a programming approach for performing scientific calculations on a graphics processing unit (GPU) as a data-parallel computing device. In this article we present an implementation of accelerating PBMCmethod based on the Inverse scheme and the Acceptance-rejection (AR) scheme for simulating particle coagulation on the GPU. The main idea is to implement the highly threaded data-parallel processing tasks by using GPU and serial computing of complex logic and transaction processing by CPU. Furthermore, the computation accuracy of the PBMC on GPU was validated with a benchmark, a CPU-based discrete-sectional method. To evaluate the accelerating performance, the computing time on the GPU against its sequential counterpart on the CPU was compared. The speedups show that the GPU can accelerate the PBMC by a factor from decades to more than one hundred, depending on the number of simulation particle.

Efficient magnetohydrodynamic simulations on distributed multi-GPU systems using a novel GPU Direct–MPI hybrid approach

Modern graphics processing units (GPUs) have been widely utilized in magnetohydrodynamic (MHD) simulations in recent years. Due to the limited memory of a single GPU, distributed multi-GPU systems are needed to be explored for large-scale MHD simulations. However, the data transfer between GPUs bottlenecks the efficiency of the simulations on such systems. In this paper we propose a novel GPU Direct–MPI hybrid approach to address this problem for overall performance enhancement. Our approach consists of two strategies: (1) We exploit GPU Direct 2.0 to speedup the data transfers between multiple GPUs in a single node and reduce the total number of message passing interface (MPI) communications; (2) We design Compute Unified Device Architecture (CUDA) kernels instead of using memory copy to speedup the fragmented data exchange in the three-dimensional (3D) decomposition. 3D decomposition is usually not preferable for distributed multi-GPU systems due to its low efficiency of the fragmented data exchange. Our approach has made a breakthrough to make 3D decomposition available on distributed multi-GPU systems. As a result, it can reduce the memory usage and computation time of each partition of the computational domain. Experiment results show twice the FLOPS comparing to common 2D decomposition MPI-only implementation method. The proposed approach has been developed in an efficient implementation for MHD simulations on distributed multi-GPU systems, called MGPU–MHD code. The code realizes the GPU parallelization of a total variation diminishing (TVD) algorithm for solving the multidimensional ideal MHD equations, extending our work from single GPU computation (Wong et al., 2011) to multiple GPUs. Numerical tests and performance measurements are conducted on the TSUBAME 2.0 supercomputer at the Tokyo Institute of Technology. Our code achieves 2 TFLOPS in double precision for the problem with 12003 grid points using 216 GPUs.

GPU computing of compressible flow problems by a meshless method with space-filling curves

A graphic processing unit (GPU) implementation of a meshless method for solving compressible flow problems is presented in this paper. Least-square fit is used to discretize the spatial derivatives of Euler equations and an upwind scheme is applied to estimate the flux terms. The compute unified device architecture (CUDA) C programming model is employed to efficiently and flexibly port the meshless solver from CPU to GPU. Considering the data locality of randomly distributed points, space-filling curves are adopted to re-number the points in order to improve the memory performance. Detailed evaluations are firstly carried out to assess the accuracy and conservation property of the underlying numerical method. Then the GPU accelerated flow solver is used to solve external steady flows over aerodynamic configurations. Representative results are validated through extensive comparisons with the experimental, finite volume or other available reference solutions. Performance analysis reveals that the running time cost of simulations is significantly reduced while impressive (more than an order of magnitude) speedups are achieved.

Discrete vs. Continuous Approach in Crowd Dynamics Modeling Using GPU Computing

The aim of the study is threefold: firstly to compare the possibilities of using general-purpose computing on graphics processing units (GPGPU) in continuous and discrete crowd dynamics simulation; secondly, to compare selected properties of analogical implementations of crowd dynamics using a central processing unit (CPU) and a graphics processing unit (GPU), and thirdly, to draw conclusions on the applicability of GPUs in engines of professional crowd simulations. For this purpose, the authors have implemented two models of pedestrian dynamics: continuous Social Forces model and discrete, cellular automata–based Social Distances model. The presented simulations refer to outdoor, large-area pedestrian movement.

Power Consumption Prediction Model of General-Purpose Computing GPU with Static Program Slicing

PDF HTML阅读 XML下载 导出引用 引用提醒 静态程序切片的GPU通用计算功耗预测模型 DOI: 10.3724/SP.J.1001.2013.04361 作者: 作者单位: 作者简介: 通讯作者: 中图分类号: 基金项目: 国家自然科学基金(60970012); 教育部博士点专项基金(20113120110008); 上海重点科技攻关项目(09511501000,09220502800); 上海市一流学科建设项目(XTKX2012) Power Consumption Prediction Model of General-Purpose Computing GPU with Static Program Slicing Author: Affiliation: Fund Project: 摘要 | 图/表 | 访问统计 | 参考文献 | 相似文献 | 引证文献 | 资源附件 | 文章评论 摘要:随着图形处理器通用计算的发展,GPU(graphics processing unit)通用计算程序功耗的度量与优化成为绿色计算领域中的一个基础问题.当前,GPU 计算能耗评测主要通过硬件来实现,而开发人员无法在编译之前了解应用程序能耗,难以实现能耗约束下的代码优化与重构.为了解决开发人员评估应用程序能耗的问题,提出了针对应用程序源代码的静态功耗预测模型,根据分支结构的疏密程度以及静态程序切片技术,分别建立分支稀疏和稠密两类应用程序的功耗预测模型.程序切片是介于指令与函数之间的度量粒度,在分析GPU应用程序时具有较强的理论支持和可行性.用非线性回归和小波神经网络建立两种切片功耗模型.针对特定GPU 非线性回归模型的准确性较好.小波神经网络预测模型适合各种体系的GPU,具有较好的通用性.对应用程序分支结构进行分析后,为分支稀疏程序提供加权功率统计模型,以保证功耗评估算法的效率.分支稠密程序则采用基于执行路径概率的功耗预测法,以提高预测模型的准确性.实验结果表明,两种预测模型及算法能够有效评估GPU 通用计算程序的功耗,模型预测值与实际测量值的相对误差低于6%. Abstract:With the development of general-purpose computing of GPUs (graphics processing units), power consumption measurements and optimization have become an essential issue in the green computing field. The current power consumption of GPUs is mainly measured by the hardware. However the programmers have had difficulty understanding the power consumption profile of the applications used to optimize and refactor before the compile phase. To solve this issue, power consumption models were proposed for GPU applications with regard to sparseness- branch and denseness-branch programs based on program slicing, respectively. The program slicing is fine-grained level that lies between the function and the instruction levels and has good feasibility and accuracy in the power consumption estimation. The power consumption prediction models for program slicing were proposed through no-linear regression and wavelet neural networks. To specific GPUs, the power prediction model based on no-linear regression is more precise than the prediction model based on wavelet neural networks. However the wavelet neural networks model has better generality to various kinds of GPUs. After analyzing the structure of the applications, the weighted power model for sparseness-branch programs was provided to achieve better effectiveness. The probability slicing power model for denseness-branch programs was also proposed to improve the accuracy that is based on the probability of the execution paths. The results indicate that the two different models can effectively predict the power consumption. And the average relative error between the predicted value and the measured value is less than 6%. 参考文献 相似文献 引证文献

Eclat Algorithm for FIM on CPU-GPU co-operative & parallel environment

Extracting the frequent itemsets from a transactional database is a fundamental task in data mining field because of its broad applications in mining association rules, time series, correlations etc. The Apriori or Eclat approaches are the commonly used generate-and-check approach to obtain frequent itemsets from a database with a given threshold value. Implementations take advantage of the GPU's massively multi-threaded SIMD (Single Instruction, Multiple Data) architecture which will employ a bitmap data structure to represent vertical transaction list ,to exploit the GPU's SIMD parallelism , and to perform the support counting operation. The implementation runs entirely on the GPU and eliminates intermediate data transfer between the GPU memory and the CPU memory, which can reduce computation time and improve overall performance .OpenCL is a platform independent Open Computing Language for GPU computation. Thus, the aim of our approach is to develop efficient parallel new advanced Eclat strategy of Frequent Itemset Mining that utilize new-generation graphics processing units (GPUs) to speed-up the process.

Numerical modeling of dam-break flood through intricate city layouts including underground spaces using GPU-based SPH method

This paper applies the meshfree Smoothed Particle Hydrodynamics (SPH) method with Graphical Processing Unit (GPU) parallel computing technique to investigate the highly complex 3-D dam-break flow in urban areas including underground spaces. Taking the advantage of GPUs parallel computing techniques, simulations involving more than 107 particles can be achieved. We use a virtual geometric plane boundary to handle the outermost solid wall in order to save considerable video card memory for the GPU computing. To evaluate the accuracy of the new GPU-based SPH model, qualitative and quantitative comparison to a real flooding experiment is performed and the results of a numerical model based on Shallow Water Equations (SWEs) is given with good accuracy. With the new GPU-based SPH model, the effects of the building layouts and underground spaces on the propagation of dambreak flood through an intricate city layout are examined.

Probing the Statistical Validity of the Ductile-to-Brittle Transition in Metallic Nanowires Using GPU Computing.

We perform a large-scale statistical analysis (>2000 independent simulations) of the elongation and rupture of gold nanowires, probing the validity and scope of the recently proposed ductile-to-brittle transition that occurs with increasing nanowire length [Wu et al. Nano Lett. 2012, 12, 910-914]. To facilitate a high-throughput simulation approach, we implement the second-moment approximation to the tight-binding (TB-SMA) potential within HOOMD-Blue, a molecular dynamics package which runs on massively parallel graphics processing units (GPUs). In a statistical sense, we find that the nanowires obey the ductile-to-brittle model quite well; however, we observe several unexpected features from the simulations that build on our understanding of the ductile-to-brittle transition. First, occasional failure behavior is observed that qualitatively differs from that predicted by the model prediction; this is attributed to stochastic thermal motion of the Au atoms and occurs at temperatures as low as 10 K. In addition, we also find that the ductile-to-brittle model, which was developed using classical dislocation theory, holds for nanowires as small as 3 nm in diameter. Finally, we demonstrate that the nanowire critical length is higher at 298 K relative to 10 K, a result that is not predicted by the ductile-to-brittle model. These results offer practical design strategies for adjusting nanowire failure and structure and also demonstrate that GPU computing is an excellent tool for studies requiring a large number of independent trajectories in order to fully characterize a system's behavior.

Hough transform‐based approach for estimating 3D rotation angles of vertebrae from biplanar radiographs using GPU‐acceleration

ABSTRACTFor scoliosis assessment, a system was developed to estimate the three‐dimensional (3D) rotation angles of vertebrae from biplanar radiographs. The proposed approach was based on the generalized Hough transform (GHT) technique to match the projected contours of the standard 3D primitive with vertebral edges detected in biplanar radiographs. A deformation tolerant matching strategy was proposed to extend the GHT for an inexact contour matching. The system was implemented on a graphics processing unit (GPU). Seventeen dry vertebral specimens were used to evaluate accuracy. Precision was tested by three operators on 181 vertebrae from radiographs of 15 patients. With GPU computing, the average estimation time was 10 s for a vertebra. Results showed that the 3D estimation error was less than 2.5° with the mean value of 1.6°. The intraoperator variability was within 1.6° and the interoperator variability was within 2.2°. These promising results indicate that the proposed system can help orthopedic surgeons measure the 3D vertebral rotation angles from two‐dimensional radiographs for better assessment of scoliosis. © 2013 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 23, 272–279, 2013

A Survey of General Purpose Computation of GPU for Computational Fluid Dynamics

The graphics processing unit (GPU) has evolved from configurable graphics processor to a powerful engine for high performance computer. In this paper, we describe the graphics pipeline of GPU, and introduce the history and evolution of GPU architecture. We also provide a summary of software environments used on GPU, from graphics APIs to non-graphics APIs. At last, we present the GPU computing in computational fluid dynamics applications, including the GPGPU computing for Navier-Stokes equations methods and the GPGPU computing for Lattice Boltzmann method.

A hybrid construction of a decision tree for multimedia contents

The growing availability of large amounts of multimedia contents in science and industry have made data mining applications such as data classification highly demanding. The contribution of this paper is two-fold. First, we propose an approach for constructing a decision tree based classification model for multimedia contents. Second, in order to speed up the performance of the proposed model, we propose a hybrid CPU-GPU approach for construction of decision tree on Graphic Processing Unit (GPU). Our approach not only accelerates the construction of decision tree via GPU computing, but also does so by considering the power and energy consumption of the GPU. Through the experiments, we demonstrate that the proposed hybrid CPU-GPU approach outperforms CPU-based sequential implementation by several times.

Accelerating The Virtual Brain with code generation and GPU computing

The genesis of both theory of neural field modeling and forward solutions for neuroimaging modalities and empirical methods to obtain connectivity datasets necessary to simulate the human brain dynamics has lead to the development of a whole-brain simulator, built within a freely available neuroinformatics platform called the Virtual Brain (TVB) [1-3]. TVB and the research that led to it, has opened up several possibilities as well as further research questions. On the one hand, such simulations reproduce aspects of human resting state activity, and on the other, it has been observed that certain neural pathologies manifest changes in the resting state, therefore suggesting that the Virtual Brain presents an opportunity to model the dynamics of a brain, on a case by case basis, enabled by non-invasive tractography methods such as diffusion tensor or spectrum imaging. Other use cases are also possible, e.g. in the case of stroke patients to model the effects of resulting lesions on resting state dynamics. However, obstacles remain: First, the relative effects of the parameters on the network model stability and its associated parameter dispersion in TVB are not well understood, and second, because of realism of finite conduction velocities in the brain connectivity in TVB, the models remain computationally expensive. In the work presented here, we address both obstacles in computational terms by introducing and implementing strategies to improve the performance of the simulator in TVB, based on generating customized code per simulation [4]. These improvements are drastic when access to recent graphics processing units (GPUs) is available and represent orders of magnitude increase in the ability of users to explore both parameter spaces or qualitative dynamical regimes, and therefore significantly amplify the impact of whole-brain modeling approaches.

Using graphics processing units to investigate molecular coevolution

SUMMARYWe present a massively parallel implementation of the computation of (co)evolutionary signals from biomolecular sequence alignments based on mutual information (MI) and a normalization procedure to neutral evolution. The MI is computed for two‐point and three‐point correlations within any multiple sequence alignment. We meet the high computational demand in the normalization procedure efficiently with an implementation on Graphics Processing Units (GPUs) using NVIDIA's CUDA framework. In particular, the normalization of the MI for three‐point ‘cliques’ of amino acids or nucleotides requires large sampling numbers in the normalization, which we achieve by using GPUs. GPU computation serves as an enabling technology here insofar as MI normalization is also possible using traditional computational methods [1] or cluster computation, but only GPU computation makes MI normalization for sequence analysis feasible in a statistically sufficient sample and in acceptable time given affordable commodity hardware. We illustrate (i) the computational efficiency and (ii) the biological usefulness of two‐point and three‐point MI by applications to the well‐known protein calmodulin and the variable surface glycoprotein (VSG) of Trypanosoma brucei, which are subject to involved evolutionary pressure. Here, we find striking coevolutionary patterns and distinct information on the molecular evolution of these molecules that question previous work that relied on non‐efficient MI computations. Copyright © 2013 John Wiley & Sons, Ltd.

TH-A-108-01: Radiation Dose Calculations On Graphics Processing Units (GPUs): Advances and Challenges

Dose calculation is a critical component for radiation therapy. To ensure its accuracy, especially in complicated cases with large tissue heterogeneity, modern dose-calculation algorithms contain a lot of details in their models, which demand a high computation power to achieve an acceptable level of efficiency. Especially for Monte Carlo simulation, the most accurate dose-calculation method, its high computational burden hinders the clinical applications in many contexts. Recently, a number of research have been devoted to accelerating dose calculations for both photon and proton radiotherapy on computer graphics processing units (GPU), specialized processors with a highly parallel structure originally designed for manipulating computer graphics. The achieved speed-up factors and accuracy, as well as the low cost in hardware setup and maintenance, indicate a great potential for the clinical applications of GPU-based dose calculations. Yet, there are still significant challenges. For instance, there exist conflicts between the GPU's data-parallel processing structure and the randomness of Monte Carlo simulations and hence, it is challenging to design GPU-friendly parallelization schemes for this method. It is also desirable to develop suitable physics models that are simple for GPU implementations but attain sufficient accuracy. The proposed symposium aims to bring together experts to exchange the most recent advances in this topic, including pencil-beam algorithms, superposition/convolution algorithms, and both photon and proton Monte Carlo simulations. Challenges in these methods and potential solutions will also be discussed. Learning Objectives: 1. Understand how GPU is used to accelerate dose calculations in radiotherapy, including pencil-beam algorithm, superposition/ convolution, and Monte Carlo methods. 2. Understand the current status of these algorithms. 3. Understand the challenges for GPU-based dose calculations and the potential solutions. Dr. Sami Hissoiny is employed by Elekta. The GPU code from Dr. Todd McNutt was licensed to Elekta and Gulmay.