Parallel Performance Evaluation Research Articles

5T magnetic resonance imaging: radio frequency hardware and initial brain imaging.

We aimed to demonstrate the feasibility of generating high-resolution human brain magnetic resonance imaging (MRI) at 5 Tesla (T) using a quadrature birdcage transmit/48-channel receiver coil assembly. A quadrature birdcage transmit/48-channel receiver coil assembly was designed for human brain imaging at 5T. The radio frequency (RF) coil assembly was validated by electromagnetic (EM) simulations and phantom imaging experimental studies. The simulated B1+ field inside a human head phantom and inside a human head model generated by the birdcage coils driven in circularly polarized (CP) mode at 3T, 5T and 7T was compared. Signal-to-noise ratio (SNR) maps, the inverse g-factor maps for evaluation of parallel imaging performance, anatomic images, angiography images, vessel wall images and susceptibility weighted images (SWI) were acquired using the RF coil assembly at 5T and compared to those acquired using a 32-channel head coil on a 3T MRI scanner. For the EM simulations, 5T MRI provided less RF inhomogeneity compared to that of 7T. In the phantom imaging study, the distributions of the measured B1+ field were consistent with the distributions of the simulated B1+ field. In the human brain imaging study, the average SNR value of the brain in the transversal plane at 5T was 1.6 times of that at 3T. The 48-channel head coil at 5T had higher parallel acceleration capability than the 32-channel head coil at 3T. The anatomic images at 5T also showed higher SNR than those at 3T. Improved delineation of the hippocampus, lenticulostriate arteries, and basilar arteries was observed at 5T compared to 3T. SWI with a higher resolution of 0.3 mm ×0.3 mm ×1.2 mm could be acquired at 5T, which enabled better visualization of small blood vessels compared to that at 3T. 5T MRI can provide significant SNR improvement compared to that of 3T with less RF inhomogeneity than that of 7T. The ability to obtain high quality in vivo human brain images at 5T using the quadrature birdcage transmit/48-channel receiver coil assembly has significant in clinical and scientific research applications.

Digital Circuit Performance Evaluation of Parallel Gated Junctionless Field Effect Transistor

This paper reports the modeling and performance enhancement of complementary metal-oxide-semiconductor (CMOS) inverters using parallel-gated junctionless field-effect transistors (PGJLFET). A mathematical model for different parameters, namely, low and high input voltage, low and high output voltage, noise margin, rise time, fall time, propagation delay, and power dissipation of the inverter circuit was established. To establish the model for various parameters, potential models at the source-channel boundary and the potential at the drain-channel boundary are considered. The variations in the voltage transfer characteristics of the output current with respect to the input voltage and the power dissipation with respect to the input voltage for steady-state conditions and transient states were investigated for different gate dielectrics, gap lengths, and gate oxide thicknesses. The models were validated by comparing them with technology computer-aided design (TCAD) simulation results. The noise margin, rise time, fall time, and propagation delay of the PGJLFET-based inverter are estimated and compared with those of a conventional junctionless transistor. It has been observed that the PGJLFET exhibits an improvement in the noise margin and propagation delay compared with conventional junctionless transistors.

Simulation of high-viscosity-ratio multicomponent fluid flow using a pseudopotential model based on the nonorthogonal central-moments lattice Boltzmann method.

In this research, the development of a pseudopotential multicomponent model with the capability of simulating high-viscosity-ratio flows is discussed and examined. The proposed method is developed based on the non-orthogonal central moments model in the lattice Boltzmann method, and the exact difference model (EDM) is used to apply the intercomponent interaction force. In contrast to the original Shan-Chen model, in which the applying force has the viscosity-dependent error term, the error term of this model does not depend on the viscosity. A GPU parallel cuda code has been developed and is employed to study the proposed method. Different cases are considered to evaluate the ability of the model, including the Laplace test, a static droplet, and a two-component concurrent channel flow. Also, wetting and nonwetting relative permeabilities for flows with dynamic viscosity ratios between 0.0002 and 5000 are predicted. Numerical results are compared with those of available analytical solutions. Very good agreement between these results are observed. The model has the capability of simulating multicomponent flows with very low kinematic viscosities of the order of 10^{-5} and dynamic viscosity ratios of up to an order of 10^{4}, which is a much wider range compared with that of existing pseudopotential models. Furthermore, the results showed that the parallel processing on GPU significantly accelerated computations. The present parallel performance evaluation shows that the cuda parallel can achieve about 41 times improvement than the CPU serial implementation. The aforementioned enhancement increases the flexibility of the multicomponent lattice Boltzmann method and its applicability to a broader spectrum of engineering applications.

Deep learning parallel computing and evaluation for embedded system clustering architecture processor

In the era of intelligence, the processing of a large amount of information and various intelligent applications need to rely on embedded devices. This trend has made machine learning algorithms play an increasingly important role. High-performance embedded computing is an effective means to solve the lack of computing power of embedded devices. Aiming at the problem that the calculation amount of new intelligent embedded applications based on machine learning technology is higher, the computing power of traditional embedded systems is difficult to meet their needs, this paper studies the parallel optimization and implementation techniques of convolutional neural networks in Parallella platform. The parallel optimization strategy of convolutional neural network on the clustering architecture processor of heterogeneous multi-core system is given. Then the high-performance implementation of convolutional neural network on Parallella platform is studied, and the function of convolutional neural network system is implemented. A set of performance evaluation methods for embedded parallel processors is proposed. From the application point of S698P, the eCos operating system is selected as the platform. The single-core mode and multi-core mode are compared on the simulator GRSIM, and the parallel performance evaluation is given. Experiments have shown that the efficiency of deep learning tasks is significantly improved compared to traditional parallel methods.

Parallel performance evaluation for curing process of thermoset nanocomposites materials

The investigation towards rate of growth for the curing process of thermoset nanocomposites materials is of great concern to most engineering and chemistry researchers since it is closely related to our natural resources, and has valuable properties. It is important to tailor and control the waste composite materials temperature profile during curing process to reproduce and improve its efficiency. Due to this phenomenon, temperature profile during curing process between two layers of composite materials is considered in this study. The mathematical model of 1D convection–diffusion of the heat equation of thick thermoset composite during its curing process has been employed. We predicted the temperature behavior for separating two layers of nanocomposites materials. Mathematical modelling to represent the problem were discretized numerically by employing central Finite Different Method (FDM) with weight parameter θ. The Linear System Equation (LSE) resulted from the discretization are solved by three numerical methods, Alternating Group Explicit (AGE), Red Black Gauss Seidel (RBGS) and Jacobi (JB). Results acquired from RBGS and JB methods have become the benchmarks for evaluating the results from AGE methods. Then, we developed parallel algorithms on the mathematical model based on AGE, RBGS and JB methods. Finally, we evaluated and analyzed the performance indicators based on the parallel performance evaluation (PPE) in solving the PDE. Parallel algorithm is computed in Parallel Virtual Machine (PVM) programming on Linux operating system (Fedora 21) as a platform in order to accelerate the sequential execution as well as the convergence rate. Parallel AGE has significantly outperformed their counterparts compared with the benchmark classical methods (RBGS and JB) for two dimensional problems. Despite having huge computational complexity, results obtained from Parallel AGE methods have proven valuable since it enables to provide solution with higher accuracy and stability.

NvPD: novel parallel edit distance algorithm, correctness, and performance evaluation

Edit distance has applications in many domains such as bioinformatics, spell checking, plagiarism checking, query optimization, speech recognition, and data mining. Traditionally, edit distance is computed by dynamic programming based sequential solution which becomes infeasible for large problems. In this paper, we introduce NvPD, a novel algorithm for parallel edit distance computation by resolving dependencies in the conventional dynamic programming based solution. We also establish the correctness of modified dependencies. NvPD exhibits certain characteristics such as balanced workload among processors, less synchronization overhead, maximum utilization of resources and it can exploit spatial locality. It requires $$\min (m,n)$$ steps to complete as compared to diagonal based approach that completes in $$\max (m,n)$$ . Experimental evaluation using variety of random and real life data sets over shared memory multi-core systems and graphic processing units (GPUs) show that NvPD outperforms state-of-the-art parallel edit distance algorithms.

A Distributed Memory Parallel Fourth-Order IADEMF Algorithm

The fourth-order finite difference Iterative Alternating Decomposition Explicit Method of Mitchell and Fairweather (IADEMF4) sequential algorithm has demonstrated its ability to perform with high accuracy and efficiency for the solution of a one-dimensional heat equation with Dirichlet boundary conditions. This paper develops the parallelization of the IADEMF4, by applying the Red-Black (RB) ordering technique. The proposed IADEMF4-RB is implemented on multiprocessor distributed memory architecture based on Parallel Virtual Machine (PVM) environment with Linux operating system. Numerical results show that the IADEMF4-RB accelerates the convergence rate and largely improves the serial time of the IADEMF4. In terms of parallel performance evaluations, the IADEMF4-RB significantly outperforms its counterpart of the second-order (IADEMF2-RB), as well as the benchmarked fourth-order classical iterative RB methods, namely, the Gauss-Seidel (GS4-RB) and the Successive Over-relaxation (SOR4-RB) methods.

A Novel Parameter Optimization Method for the Driving System of High-Speed Parallel Robots

Driving system parameters optimization, especially the optimal selection of specifications of motor and gearbox, is very important for improving high-speed parallel robots' performance. A very challenging issue is parallel robots' performance evaluation that should be able to illustrate robots' performance accurately and guide driving system parameters optimization effectively. However, this issue is complicated by parallel robots' anisotropic translational and rotational dynamic performance, and the multiparameters of motors and gearboxes. In this paper, by separating the influence of translational and rotational degrees-of-freedom (DOFs) on robots' performance, a new dynamic performance index is proposed to reflect the driving torque in instantaneous acceleration. Then, the influence of driving system's multiparameters on robots' driving torque in instantaneous acceleration and cycle time in continuous motion is investigated. Based on the investigation, an inertia matching index is further derived which is more suitable for minimizing the driving torque of parallel robots with translational and rotational DOFs. A comprehensive parameterized performance atlas is finally established. Based on this atlas, the performance of a high-speed parallel robot developed in this paper can be clearly evaluated, and the optimal combination of motors and gearboxes can be quickly selected to ensure low driving torque and high pick-and-place frequency.

Simulations of turbulent duct flow with lattice Boltzmann method on GPU cluster

Lattice Boltzmann models are used to simulate three-dimensional pressure driven laminar and turbulent duct flows using graphics processing units. The performances of the D3Q15, D3Q19 and D3Q27 models are first validated by comparisons with laminar analytical and benchmark solutions; the D3Q27 model is shown to produce the most accurate results. Turbulent duct flows at Reτ=360 and 600 are further simulated by the D3Q27 model, and the predicted results compare favorably with the existing Navier–Stokes equation based benchmark solutions. For multi-GPU computations, the performance of 2-D domain decomposition is superior to that of 1-D domain decomposition where the performance saturates at a lower GPU count. For the present parallel performance evaluation, 1730 and 2464 MLUPS are achieved with 12 GTX Titan and 48 Tesla M2070 GPUs from nVIDIA, respectively.

The 5th International Conference on Science & Engineering in Mathematics, Chemistry and Physics 2017 (ScieTech 2017), The Ramada Bintang Hotel, Kuta, Bali, Indonesia 21 - 22 January 2017

The 5th International Conference on Science & Engineering in Mathematics, Chemistry and Physics 2017 (ScieTech 2017), The Ramada Bintang Hotel, Kuta, Bali, Indonesia 21 - 22 January 2017

High-performance simulation of nation-sized smart grids

This paper develops a large scale simulation approach for smart grids with a fine temporal resolution. This permits a detailed simulation of every single smart grid participant facilitating real-time estimates on a realistic scale about load distribution, demand and supply. This is especially interesting when considering the sustainability and efficiency of renewable energy generation. The smart grid model is based on a multi-agent system. Each agent is modelled by an optimisation problem which minimises its costs to meet a prescribed demand. In order to bestow some flexibility to each agent, it can generate energy, by e.g. solar panels, or store energy. All agents participate in a cooperative bargaining game to compute their respective optimal solution. By doing so, the energy prices are dynamically adapted to the agents’ solutions so that an energy market is formed. This bargaining game is designed that a unique Nash equilibrium exists. The computations to obtain the equilibrium point are performed in parallel. To this end, a synchronisation scheme is derived to ensure convergence of the proposed parallel bargaining algorithm. The smart grid structure and capacity limitations are also taken into account. The solutions of the bargaining game are compliant with the smart grid’s capacities by solving a modified network flow problem and forcing the bargaining solution within the capacity limitations in an iterative procedure. The correctness of the presented parallel algorithm is validated in tests along with a parallel performance evaluation. As a proof of concept of the practical applicability of this approach, a smart grid with over 30 million private households and multiple energy types is simulated.

HIGH SPEED COMPUTING OF ICE THICKNESS EQUATION FOR ICE SHEET MODEL

Two-dimensional (2-D) ice flow thermodynamics coupled model acts as a vital role for visualizing the ice sheet behaviours of the Antarctica region and the climate system. One of the parameters used in this model is ice thickness. Explicit method of finite difference method (FDM) is used to discretize the ice thickness equation. After that, the equation will be performed on Compute Unified Device Architecture (CUDA) programming by using Graphics Processing Unit (GPU) platform. Nowadays, the demand of GPU for solving the computational problem has been increasing due to the low price and high performance computation properties. This paper investigates the performance of GPU hardware supported by the CUDA parallel programming and capable to compute a large sparse complex system of the ice thickness equation of 2D ice flow thermodynamics model using multiple cores simultaneously and efficiently. The parallel performance evaluation (PPE) is evaluated in terms of execution time, speedup, efficiency, effectiveness and temporal performance.

PARALLEL COMPUTING OF NUMERICAL SCHEMES AND BIG DATA ANALYTIC FOR SOLVING REAL LIFE APPLICATIONS

This paper proposed the several real life applications for big data analytic using parallel computing software. Some parallel computing software under consideration are Parallel Virtual Machine, MATLAB Distributed Computing Server and Compute Unified Device Architecture to simulate the big data problems. The parallel computing is able to overcome the poor performance at the runtime, speedup and efficiency of programming in sequential computing. The mathematical models for the big data analytic are based on partial differential equations and obtained the large sparse matrices from discretization and development of the linear equation system. Iterative numerical schemes are used to solve the problems. Thus, the process of computational problems are summarized in parallel algorithm. Therefore, the parallel algorithm development is based on domain decomposition of problems and the architecture of difference parallel computing software. The parallel performance evaluations for distributed and shared memory architecture are investigated in terms of speedup, efficiency, effectiveness and temporal performance.

High Speed Computation of Muscle Stress Problem Using Crack Propagation Modelling on DPCS Platform

Muscle stress is a disease resulting from performing too much effort, sport fitness or high blood pressure problem. This paper focuses on the mathematical modeling of crack propagation that applies on the human muscle stress. The modeling describes the relationship between the temperature and crack propagation among targeted region, node and element. The contribution of this paper is to predict the stress level based on the mathematical modeling for nodal displacements of crack propagation with respect to time and space. The initial stage of crack propagation that leads to an initiation of crack propagation is being modeled satisfactorily the discretization of finite element method (FEM) of the specific region. The sequential algorithm of the crack propagation simulation for muscle stress will be solve using Gauss Seidel method (GS). The parallel algorithm will be solve using Red Black Gauss Seidel method (RBGS). Some comparisons based on numerical analysis and parallel performance evaluation will be analyzed in terms of convergent rate of FE solution results to the exact solution of clinical dataset, run time prediction, time execution, speedup, efficiency, effectiveness and temporal performance. C and Parallel Virtual Machine (PVM) software with open source operating system on distributed parallel computer system (DPCS) will support the huge computation cost of the parallel algorithms.

Parallel Performance Comparison of Alternating Group Explicit Method Between Parallel Virtual Machine and Matlab Distributed Computing for Solving Large Sparse Partial Differential Equations

This paper presents the parallel implementation and performance analysis of a large sparse matrix of one dimensional parabolic Partial Differential Equation (PDE) between Parallel Virtual Machine (PVM) and Matlab Distributed Computing (MDC) platform. PVM is a software to perform parallel programming using heterogeneous or homogeneous clusters of parallel computing system connected by a logical network system to users as virtual memory parallel computer system. MDC is a scientific library programming language. Alternating Group Explicit (AGE) method is used to solve the parabolic equation with large sparse matrix. The distributed parallel algorithm of AGE method is implemented on the PVM with RedHat Linux operation and MDC Version 2011a. The parallel performance evaluation (PPE) is compared between the two platforms in terms of time execution, speedup, efficiency, effectiveness, and temporal performance. From the performance analysis, we can conclude that PVM is better compared to MDC.

A three-dimensional discontinuous Galerkin model applied to the baroclinic simulation of Corpus Christi Bay

In this work, we present results of a numerical study of Corpus Christi Bay, Texas and surrounding regions and compare simulated model results to recorded data. The validation data for the year 2000 include the water elevation, velocity, and salinity at selected locations. The baroclinic computations were performed using the University of Texas Bays and Estuaries 3D (UTBEST3D) simulator based on a discontinuous Galerkin finite element method for unstructured prismatic meshes. We also detail some recent advances in the modeling capabilities of UTBEST3D, such as a novel turbulence scheme and the support for local vertical discretization on parts of the computational domain. All runs were conducted on parallel clusters; an evaluation of parallel performance of UTBEST3D is included.

Experimental Evaluation of parallelism in real time execution

Abstract Parallel programming is a great challenging of modern computing. The main goal of that parallel programming is to improve performance of computer applications. A well-structured parallel application can achieve better performance over sequential execution on existing and upcoming parallel computer architecture. This paper describes the experimental evaluation of parallel application performance with thread-safe data structure in real time execution. Described different performance issues of this paper, help to make efficient parallel application for better performance. Before describing the experimental evaluation, this paper describes some methodologies relevant of parallel programming. The evaluation of parallel applications has been done by experimental result evaluation and performance measurements. The knowledge of proper partitioning, oversubscription, proper work load balancing, way of memory sharing will help to make more efficient parallel application.

五自由度并联机械腿静力学性能评价与优化设计

To evaluate the static performance and design optimally a parallel mechanical leg for the six-legged robot,a static analysis method based on the mapping of both constraint and actuation forces was proposed,and the mechanical leg was designed based on this method.Firstly,the constraint and actuation Jacobian matrixes were established based on the mapping of both constraint and actuation.Then,according to the virtual work principle,the actuation statics transmission equation was established,the performance evaluation indexes of actuation statics were designed,and the relationship curve between the performance evaluation index and structure parameters was calculated.By the same way,the transmission equation and performance evaluation index of constraint statics were obtained,and the relationship curve between the performance evaluation index and structure parameters was also calculated.Finally,based on the performance evaluation indexes both of constraint and actuation statics,the structure parameters were optimally designed by Monte Carlo method.Calculations show that when the structure parameters of a fixed platform and a movement platform are 200 mm,and 80 mm and the UPU branched chain lengths l1min,l1max are 500 mm and 900 mm,the statics comprehensive performance of the parallel mechanical leg is the best.These studies lays the theoretical foundation for further study of the six-legged robot.

Method of Parallel System Performance Evaluation Based on Multi-Level Fuzzy Synthetic Evaluation Model

Nowadays, the parallel system performance evaluation is the hot spot topic. It’s helpful to improve the parallel system performance to establish a scientific evaluation model. In this paper, using of analytic hierarchy process and the fuzzy mathematics theory, the multi-level fuzzy synthetic evaluation model is established, the performance index system is defined and the parallel system performance evaluation process is realized.

Parallel application benchmarks and performance evaluation of the Intel Xeon 7500 family processors

Abstract With the recent advent of novel multi- and many-core hardware architectures, application programmers have to deal with many hardware-specific implementation details and have to be familiar with software optimization techniques to benefit from new high-performance computing machines. Highly effcient parallel application design is in fact an interdisciplinary process involving domain specific and IT experts. Therefore, this paper aims to present early experiences with computationally demanding applications, development efforts and evaluation of their performance on the new family of Intel Xeon 7500 processors. We selected two application benchmarks applicable to real quantum chemistry and Computational Fluid Dynamics (CFD) problems as they can potentially take advantage of parallel processing on novel hardware architectures and built-in new features. Additionally, we discuss various parallel software improvements to mentioned applications, including appropriate changes to data structures as well as to communication and synchronization routines to deal with multi-level parallelism and hybrid hardware architectures. The obtained results confirmed that new hardware solutions can improve the overall application performance. However, in order to obtain a high level of parallel scalability various application modifications and tuning procedures are required as hardware configurations, including processors characteristics, interconnects and topologies, and they have a great influence on large-scale simulations.