PC Cluster System Research Articles

Real-Time Space Carving Using Graphics Hardware

Reconstruction of real-world scenes from a set of multiple images is a topic in computer vision and 3D computer graphics with many interesting applications. Attempts have been made to real-time reconstruction on PC cluster systems. While these provide enough performance, they are expensive and less flexible. Approaches that use a GPU hardware-acceleration on single workstations achieve real-time framerates for novel-view synthesis, but do not provide an explicit volumetric representation. This work shows our efforts in developing a GPU hardware-accelerated framework for providing a photo-consistent reconstruction of a dynamic 3D scene. High performance is achieved by employing a shape from silhouette technique in advance. Since the entire processing is done on a single PC, the framework can be applied in mobile environments, enabling a wide range of further applications. We explain our approach using programmable vertex and fragment processors and compare it to highly optimized CPU implementations. We show that the new approach can outperform the latter by more than one magnitude and give an outlook for interesting future enhancements.

Development of a parallelized 3D electrostatic PIC-FEM code and its applications

A parallelized three-dimensional self-consistent electrostatic particle-in-cell (PIC) code using unstructured tetrahedral mesh is proposed. Parallel implementation of the current unstructured PIC-FEM code is realized on distributed-memory PC-cluster system utilizing dynamic domain decomposition. Completed code is verified by simulating a quasi-1D RF argon gas discharge with results comparable to previous experimental observations and simulations. Parallel performance with dynamic domain decomposition of the PIC code is tested using a 3D RF argon gas discharge on a PC-cluster system. Results show that parallel efficiency can achieve 83% at 32 processors with dynamic domain decomposition. Some possible improvement of the code performance is demonstrated. Completed code is then applied to predict field emission without and with space-charge effect, and to simulate the RF magnetron argon plasma to demonstrate its capability in handling practical problems.

狭ポールチップ幅ＳＰＴヘッドのマイクロマグネティック記録磁界解析

Landau-Lifshitz-Gilbert (LLG) micromagnetic recording field analysis of a single-pole-type (SPT) head using a parallelized program with a PC cluster system is investigated. The use of a PC cluster system allows dynamic calculations of SPT write head models with cells 10 nm a side. An 8-CPU system realized a calculation speed 30 times faster than the original program, which used a 64-bit single processor system. The dc and dynamic recording fields are investigated for various pole tip structures and material characteristics. First, the dc (quasi-static) recording fields are compared for LLG and FEM calculations. Next, it is shown that an optimum non-zero throat height exists that maximizes recording field strength. The addition of a trailing shield is found to be effective for obtaining a larger recording field gradient. Dynamic recording fields are also derived for various head structures and materials.

Parallel implementation of a three-dimensional electrostatic PIC–MCC method using an unstructured tetrahedral mesh

A parallel three-dimensional electrostatic particle-in-cell Monte Carlo Collision (PIC-MCC) code using an unstructured tetrahedral mesh is developed and validated in this paper. Poisson's equation is discretized by a nodal finite element method using a linear interpolation function. Charmed-particle motion is traced cell by cell using a leap-frog method and the Boris scheme. Code is validated using quasi-one-dimensional radiofrequeney gas discharge by observing the bi-Maxwellian elecfrom energy probability distribution function at low pressure. Parallel performance of the PIC code is tested using a three-dimensional radiofrequency gas discharge on a PC-cluster system. Results show that parallel efficiency of 83% can be achieved at 32 processors with dynamic domain decomposition.

PC Cluster based Parallel Adaptive Evolutionary Algorithm for Service Restoration of Distribution Systems

This paper presents an application of the parallel Adaptive Evolutionary Algorithm (AEA) to search an optimal solution of the service restoration in electric power distribution systems, which is a discrete optimization problem. The main objective of service restoration is, when a fault or overload occurs, to restore as much load as possible by transferring the de-energized load in the out of service area via network reconfiguration to the appropriate adjacent feeders at minimum operational cost without violating operating constraints. This problem has many constraints and it is very difficult to find the optimal solution because of its numerous local minima. In this investigation, a parallel ABA was developed for the service restoration of the distribution systems. In parallel AEA, a genetic algorithm (GA) and an evolution strategy (ES) in an adaptive manner are used in order to combine the merits of two different evolutionary algorithms: the global search capability of the GA and the local search capability of the ES. In the reproduction procedure, proportions of the population by GA and ES are adaptively modulated according to the fitness. After ABA operations, the best solutions of AEA processors are transferred to the neighboring processors. For parallel computing, a PC cluster system consisting of 8 PCs was developed. Each PC employs the 2 ㎓ Pentium Ⅳ CPU and is connected with others through switch based fast Ethernet. To show the validity of the proposed method, the developed algorithm has been tested with a practical distribution system in Korea. From the simulation results, the proposed method found the optimal service restoration strategy. The obtained results were the same as that of the explicit exhaustive search method. Also, it is found that the proposed algorithm is efficient and robust for service restoration of distribution systems in terms of solution quality, speedup, efficiency, and computation time.

Hierarchical parallel processing of large scale data clustering on a PC cluster with GPU co-processing

This paper presents an effective scheme for clustering a huge data set using a PC cluster system, in which each PC is equipped with a commodity programmable graphics processing unit (GPU). The prop...

Parallel genetic algorithms for product configuration management on PC cluster systems

This work presents two parallel genetic algorithms (PGAs) for product configuration management: a parallel conventional genetic algorithm (PCGA) and a parallel multiple-searching genetic algorithm (PMGA). This parallel/distributed approach is based on a coarse-grained (or island) paradigm which is implemented on a cluster of PCs using message passing interface for the genetic information interchange. The product configuration problem assuming that customers would like to have minimum cost and a customized product can be obtained by finding the shortest path of the configuration network diagram. The performance of these algorithms is estimated by comparing the solutions of PGAs with those of sequential genetic algorithms (GAs) and mathematical programming. A weighting scale example from an empirical study is reported for illustrational purposes. Computational results show that the solutions obtained from the PMGA outperform other GAs in both accuracy and efficiency.

Development of a parallel Poisson's equation solver with adaptive mesh refinement and its application in field emission prediction

A parallel electrostatic Poisson's equation solver coupled with parallel adaptive mesh refinement (PAMR) is developed in this paper. The three-dimensional Poisson's equation is discretized using the Galerkin finite element method using a tetrahedral mesh. The resulting matrix equation is then solved through the parallel conjugate gradient method using the non-overlapping subdomain-by-subdomain scheme. A PAMR module is coupled with this parallel Poisson's equation solver to adaptively refine the mesh where the variation of potentials is large. The parallel performance of the parallel Poisson's equation is studied by simulating the potential distribution of a CNT-based triode-type field emitter. Results with ∼100 000 nodes show that a parallel efficiency of 84.2% is achieved in 32 processors of a PC-cluster system. The field emission properties of a single CNT triode- and tetrode-type field emitter in a periodic cell are computed to demonstrate their potential application in field emission prediction.

Optimized high-order finite difference wave equations modeling on reconfigurable computing platform

Finite difference (FD) methods are the most prevalent numerical modeling algorithms for simulating linear wave propagation phenomena in geophysics, electromagnetics, and aero- or marine-acoustics applications. Unfortunately, evaluating time evolution for waves in large-scale 2D or 3D domains is computationally demanding and data-intensive. As such, its programs are often exclusively executed on high performance supercomputers. In this paper, we propose a solution to accelerate the execution of realistic wave field modeling problems on a baseline reconfigurable computing (RC) platform. By adopting appropriate high-order temporal and spatial FD schemes along with efficient on-chip data buffering structure, we alleviate external memory bandwidth bottleneck at the cost of increased floating-point computations, which fortunately can be absorbed by the pipelined computing engine with negligible speed penalty. In this way, a balance point can always be reached where the utilization of onboard reconfigurable hardware resources and external memory bandwidth are all optimized. The desirable simplicity and scalability properties of our method make it compatible with most commercial RC platforms. Moreover, our implementation is consistent with the prevalent PC-Cluster systems and can achieve better price-performance ratio along with much lower power consumption.

Three-dimensional phononic band gap calculations using the FDTD method and a PC cluster system

This paper aims at studying the band gap phenomena of three-dimensional phononic crystals using the finite difference time domain (FDTD) method and a PC cluster system. In the paper, Bloch's theorem is applied to the wave equation and to the boundary conditions of the periodic structure. We calculate the variations of displacements and take discrete Fourier transform to acquire the resonances of the structures. Then, the dispersion relations of the bulk acoustic wave can be obtained and the band gaps are predicted accordingly. On the other hand, because of larger data calculation in three-dimensional phononic crystals, we introduce the PC cluster system and parallel FDTD programs written with respect to the architecture of a PC cluster system. Finally, we discuss the numerical calculation of two-dimensional and three-dimensional phononic crystals consisting of steel/epoxy and draw conclusions regarding the band gap phenomena between these phononic crystals.

An Enhanced Parallel Loop Self-Scheduling Scheme for Cluster Environments

Approaches for dealing with scheduling and load-balancing in PC-based cluster systems are famous and well known. In such environments, Self-Scheduling Schemes are suitable for parallel loops with independent iterations. However, while schemes such as FSS, GSS, and TSS fit most computer systems, they cannot provide good load-balancing. Chao-Tung Yang and Shun-Chi Chang proposed a parallel loop scheduling scheme for heterogeneous PC cluster systems in Yang and Chang [13]. Though the proposed scheme allows users to choose parameters before execution initialization, weaknesses in it motivated us to develop further improvements. For instance, using fixed and monotonous parameters can easily lead to invalid scheduling due to use of previously input information. Thus, in this paper we propose a new scheme that fits most widely available computer systems and allows the scheduling parameter to be adjusted dynamically in order to provide higher overall performance.

Parallel implementation of molecular dynamics simulation for short-ranged interaction

A parallel molecular dynamics simulation method, designed for large-scale problems, employing dynamic spatial domain decomposition for short-ranged molecular interactions is proposed. In this parallel cellular molecular dynamics (PCMD) simulation method, the link-cell data structure is used to reduce the searching time required for forming the cut-off neighbor list as well as for domain decomposition, which utilizes the multi-level graph-partitioning technique. A simple threshold scheme (STS), in which workload imbalance is monitored and compared with some threshold value during the runtime, is proposed to decide the proper time for repartitioning the domain. The simulation code is implemented and tested on the memory-distributed parallel machine, e.g., PC-cluster system. Parallel performance is studied using approximately one million L-J atoms in the condensed, vaporized and supercritical states. Results show that fairly good parallel efficiency at 49 processors can be obtained for the condensed and supercritical states (∼60%), while it is comparably lower for the vaporized state (∼40%).

1104 マスタースレーブモデルにおける並列遺伝的アルゴリズムの検討(OS22 最適化計算法)

This paper discuses the performance of parallel genetic algorithms. Master-slave model is one of implementation models of GAs. In the most cases, the significant scale up is derived by using a master-slave model. However, the optimum number of process is existed and this number is decided by the relation of computational time and communicational time. In this paper, this optimum number of process is analyzed. At the same time, this analysis is confirmed through the experiments using PC cluster system.

SERVICE RESTORATION IN DISTRIBUTION AUTOMATION SYSTEM USING PARALLEL HYBRID GENETIC ALGORITHM-TABU SEARCH

This paper presents an application of parallel hybrid Genetic Algorithm-Tabu Search (GA-TS) in searching the optimal service restoration solution in an electric power distribution system, a discrete optimization problem. In this investigation, parallel hybrid GA-TS was developed for the service restoration of a distribution system. For parallel computing, a PC-cluster system consisting of 8 PCs was developed. Each PC employed a 2 GHz Pentium IV CPU, and was connected with others through Ethernet-switch-based fast Ethernet.The newly developed algorithm was tested with a practical distribution system in Korea. From the simulation results, the proposed method found the optimal fault restoration strategy. The obtained results were the same as those of the explicit exhaustive search method (Wu, Tomsovic, and Chen, 1991). Also, it was found that the proposed algorithm is efficient and robust for distribution system service restoration in the solution quality, speed-up, efficiency, and computation time.

Real-time 3D shape reconstruction, dynamic 3D mesh deformation, and high fidelity visualization for 3D video

3D video [IEEE Multimedia (1997) 18] is the ultimate image media recording dynamic visual events in the real world as is; it records time varying 3D object shape with high fidelity surface properties (i.e., color and texture). Its applications cover wide varieties of personal and social human activities: entertainment (e.g., 3D game and 3D TV), education (e.g., 3D animal picture books), sports (e.g., sport performance analysis), medicine (e.g., 3D surgery monitoring), culture (e.g., 3D archive of traditional dances), and so on. In this paper, we propose: (1) a PC cluster system for real-time reconstruction of dynamic 3D object action from multi-view video images, (2) a deformable 3D mesh model for reconstructing the accurate dynamic 3D object shape, and (3) an algorithm of rendering natural-looking texture on the 3D object surface from the multi-view video images. Experimental results with quantitative performance evaluations demonstrate the effectiveness of these methods in generating high fidelity 3D video from multi-view video images.

Application of Parallelized SOR Method to Electromagnetic Field Analysis of Superconductors

PC cluster systems are becoming popular in the field of high-performance computing. The authors have been studying the electromagnetic field analysis of high T/sub c/ superconductors (HTS) by the finite element method (FEM) for AC loss estimations. Superconductors are highly nonlinear electromagnetic media, and considerable computation time is required to calculate the temporal evolution of the electromagnetic field distribution in superconductors. In this study, parallel computing techniques were applied to the electromagnetic field analysis of HTSs by the finite element method. In one of the FEM codes used by the authors, the successive over-relaxation (SOR) method, was used to solve a system of equations. This part was parallelized using the multicolor SOR method, and it helped in reducing computation time. Data are explicitly passed between the processor elements (PEs) through an MPI. First, the HTS model was analyzed under one operating condition, using the code implemented multicolor SOR method, using four PEs, and the parallelization efficiency was confirmed. Next, the generality of the efficiency of the multicolor SOR method in our numerical simulation was examined under several operating conditions, using different models.

그리드 컴퓨팅을 이용한 열유동 해석 기법에 관한 기초 연구

Simulation of three-dimensional turbulent flow with LES and DNS lakes much time and expense with currently available computing resources and requires big computing resources especially for high Reynolds number. The emerging alternative to provide the required computing power and working environment is the Grid computing technology. We developed the CFD code which carries out the parallel computing under the Grid environment. We constructed the Grid environment by connecting different PC-cluster systems located at two different institutes of Pusan National University in Busan and KISTI in Daejeon. The specification of PC-cluster located at two different institutes is not uniform. We run our parallelized computer code under the Grid environment and compared its performance with that obtained using the homogeneous computing environment. When we run our code under the Grid environment, the communication time between different computer nodes takes much larger time than the real computation time. Thus the Grid computing requires the highly fast network speed.

A PC cluster system employing IEEE 1394

AbstractIn this paper, we describe the design and evaluation of a PC cluster system in which IEEE 1394 is applied. Networks for parallel cluster computing require low latency and high bandwidth. It is also important that the networks be commercially available at low cost. Few network devices satisfy all of the above requirements. However, the IEEE 1394 standard provides a good compromise for fulfilling these requirements. We have used IEEE 1394 devices, which support a 400 Mbps data transfer rate, to connect the nodes of a PC cluster system which we have designed and implemented. We have implemented two communication libraries. One is a fast communication library called CF for IEEE 1394. The other is a MPI layer library on the CF library. Experimental results show that CF achieves a 17.2 microsecond round‐trip time. On application benchmarks, the system was considerably faster than TCP/IP over Fast Ethernet. Even though the system was constructed at very low cost, it provides good performance. Using the IEEE 1394 standard is thus a good solution for low‐cost cluster systems. Copyright © 2004 John Wiley & Sons, Ltd.

Improvement of parallelization performance of GAMESS: global sum and (semi-)direct integral calculation in multireference perturbation calculation.

Performance benchmark tests of the multireference perturbation theory (MRPT2) calculation in the GAMESS suite of programs have been performed on a 128-node PC cluster system. The benchmark results point to two kinds of problems with GAMESS: network delay and slow access to disk files. To overcome these bottlenecks, we employ (semi-)direct atomic orbital integral calculations and a new algorithm for the "global sum" operation. These modifications produce remarkable performance improvements, especially in the 128-node PC cluster system, which was almost twice as fast in our version as in the original version. The modification of "global sum" also improved the performance of other types of molecular orbital calculations in GAMESS, such as density functional theory.

Two-scale finite element analysis of heterogeneous solids with periodic microstructures

Toward the application of a two-scale analysis method for nonlinear heterogeneous solids with periodic microstructures, we make a feasibility study and introduce a parallel algorithm to achieve the computational efficiency. For the feasibility, we focus our attention to the inhomogeneous deformation of the overall structure, which may imply the loss of periodicity assumed in the initial state. For parallel computations, we present an efficient algorithm for the deconcentration of computational loads by using a PC-cluster system. A simple numerical example for a three-dimensional heterogeneous body will suggest the applicability of the two-scale analysis method to practical problems.