Message Passing Interface Model Research Articles

The implementation of the three-dimensional unified gas-kinetic wave-particle method on multiple graphics processing units

To further improve the efficiency of the unified gas-kinetic wave-particle (UGKWP) method in hypersonic rarefied non-equilibrium flows, particularly the particle simulation process, we presented the first application of the three-dimensional UGKWP method to multiple graphics processing unit (GPU) devices in this study. The wave and particle evolution components of the method are addressed using cell and particle paralleling strategies, respectively, enabling the primary loop of the GPU-based UPKWP (GPU-UGKWP) to be executed entirely by the compute unified device architecture threads on GPU devices. Concurrently, communication issues between central processing unit (CPU) nodes are resolved by employing the message passing interface model. Additionally, we introduce a tailored memory management scheme for the GPU-UGKWP method, facilitating efficient access to the particle array. Performance comparisons reveal that, relative to a single Intel Xeon Gold 6148 CPU core, the Nvidia Tesla P100 achieves a total speedup of 34 using one GPU device, and 226 with eight GPU devices, and a single Nvidia Titan V GPU device attains a speedup of 62. The speedup outcomes on multiple CPU cores and GPU devices demonstrate that the GPU-based algorithm is better suited for computationally demanding tasks, particularly in particle-dominated simulations. As evidenced by the reduced calculation time for a hypersonic technology vehicle simulation performed on the P100 cluster, GPU devices significantly outperform their CPU counterparts.

Parallel large eddy simulations of transitional flow in a compressor cascade with endwalls

Laminar-turbulent transition and corner separation play a critical role in the aerodynamics of the compressor and are quite sensitive to the changes of flow conditions and external disturbances. However, a deep understanding of such fine flow phenomena poses a great challenge for turbulent methods and computer resources. In order to clarify the impacts of incoming flow states on the three-dimensional transitional flow in a compressor cascade, we construct a parallel Large-Eddy-Simulation (LES) methodology and apply it to a full-span compressor cascade. Both the turbulent and laminar incoming endwall boundary layers are considered at a free-stream turbulence level of 4%, which is typical in the multistage axial-flow compressor environment. The parallel performance of the MPI (Message Passing Interface) model and hybrid MPI-OpenMP (Open Multi-Processing) model is particularly analyzed at a parallel scale of 10 000 CPU (Central Processing Unit) cores. The parallel performance test shows that the efficiency of the MPI model is evidently higher than that of the hybrid MPI-OpenMP model. The LES results indicate that the incoming laminar endwall boundary layer results in a more remarkable reduction in the blade loading near the endwall and a larger total pressure loss than the turbulent one. The incoming endwall boundary layer state shows a significant impact on the evolution process of the endwall turbulence and a small impact on the corner separation and the suction-surface transition. This study demonstrates the ability of the parallel LES method to capture complex transitional-flow structures in compressor cascades and its potential application to the deeper understandings of compressor aerodynamics.

An improved coarse-grained parallel algorithm for computational acceleration of ordinary Kriging interpolation

Heavy computation limits the use of Kriging interpolation methods in many real-time applications, especially with the ever-increasing problem size. Many researchers have realized that parallel processing techniques are critical to fully exploit computational resources and feasibly solve computation-intensive problems like Kriging. Much research has addressed the parallelization of traditional approach to Kriging, but this computation-intensive procedure may not be suitable for high-resolution interpolation of spatial data. On the basis of a more effective serial approach, we propose an improved coarse-grained parallel algorithm to accelerate ordinary Kriging interpolation. In particular, the interpolation task of each unobserved point is considered as a basic parallel unit. To reduce time complexity and memory consumption, the large right hand side matrix in the Kriging linear system is transformed and fixed at only two columns and therefore no longer directly relevant to the number of unobserved points. The MPI (Message Passing Interface) model is employed to implement our parallel programs in a homogeneous distributed memory system. Experimentally, the improved parallel algorithm performs better than the traditional one in spatial interpolation of annual average precipitation in Victoria, Australia. For example, when the number of processors is 24, the improved algorithm keeps speed-up at 20.8 while the speed-up of the traditional algorithm only reaches 9.3. Likewise, the weak scaling efficiency of the improved algorithm is nearly 90% while that of the traditional algorithm almost drops to 40% with 16 processors. Experimental results also demonstrate that the performance of the improved algorithm is enhanced by increasing the problem size.

Performance analysis of parallel algorithms in physics simulation for molecular dynamics simulation liquid metals solidification processes

In this work, a parallel arithmetic program for the MD (molecular dynamics) simulation study of a large-sized system is proposed and reformed, based on the previous program with PVM (Parallel Virtual Machine) model applied in a small-sized system for liquid metals. Our methodology is combined with the MPI (message passing interface) and OpenMP (Open Multiple Processing) programming model using the spatial domain decomposition method, conquering the problems occurred in PVM model and enlarging the scale of simulated system. Comparing with the previous small-sized system consisting of 50,000–100,000 atoms, the large-sized system is consist of 5,000,000–10,000,000 atoms and the simulation results are more closely to the real situation of the simulated system. In this paper, the performance of parallel program using MPI+OpenMP model is analyzed, showing better speedup, parallel efficiency, and scalability. Finally, we adopt many physical evaluation methods to verify the validity of the simulation results, including pair distribution function, bond-type index analysis, atomic clusters analysis and visualizing analysis. From these physical results, it is clear that the simulation results are in good agreement with the experimental results, which confirm the correctness of the program in simulation.

Divide‐and‐conquer approach for solving singular value decomposition based on MapReduce

SummarySingular value decomposition (SVD) shows strong vitality in the area of information analysis and has significant application value in most of the scientific big data fields. However, with the rapid development of Internet, the information online reveals fast growing trend. For a large‐scale matrix, applying SVD computation directly is both time consuming and memory demanding. There are many works available to speed up the computation of SVD based on the message passing interface model. However, to deal with large‐scale data processing, a MapReduce model has many advantages over a message passing interface model, such as fault tolerance, load balancing and simplicity. For a MapReduce environment, existing approaches only focus on low rank SVD approximation and tall‐and‐skinny matrix SVD computation, and there are no implementations of full rank SVD computation. In this paper, we propose a MapReduce‐based implementation for solving divide‐and‐conquer SVD algorithm. To achieve high performance, we design a two‐stage task scheduling strategy based on the mathematical characteristics of divide‐and‐conquer SVD algorithm. To further strengthen the performance, we propose a row‐index‐based divide algorithm, a pipelined task scheduling method, and revised block matrix multiplication in MapReduce framework. Experimental result shows the efficiency of our algorithm. Our implementation can accommodate full rank SVD computation of large‐scale matrix very efficiently. Copyright © 2014 John Wiley & Sons, Ltd.

MPI와 OpenMP기반 하이브리드 모델을 이용한 항공 레이저 스캐닝 자료의 병렬 처리

In the present study, a parallel processing method running on a multi-core PC-Cluster is introduced to produce digital surface model (DSM) and digital terrain model (DTM) from huge airborne laser scanning data. A hybrid model using both message passing interface (MPI) and OpenMP was devised by revising a conventional MPI model which utilizes only MPI, and tested on a multi-core PC-Cluster for performance validation. In the results, the hybrid model has not shown better performances in the interpolation process to produce DSM, but the overall performance has turned out to be better by the help of reduced MPI calls. Additionally, scheduling function of OpenMP has revealed its ability to enhance the performance by controlling inequal overloads charged on cores induced by irregular distribution of airborne laser scanning data.

Large-scale high performance computation on 3D explosion and shock problems

Explosion and shock often involve large deformation, interface treatment between multi-material, and strong discontinuity. The Eulerian method has advantages for solving these problems. In parallel computation of the Eulerian method, the physical quantities of the computaional cells do not change before the disturbance reaches to these cells. Computational efficiency is low when using fixed partition because of load imbalance. To solve this problem, a dynamic parallel method in which the computation domain expands with disturbance is used. The dynamic parallel program is designed based on the generally used message passing interface model. The numerical test of dynamic parallel program agrees well with that of the original parallel program, also agrees with the actual situation.

A grid computing based approach for the power system dynamic security assessment

This paper addresses the problem of parallel dynamic security assessment applications from static homogeneous cluster environment to dynamic heterogeneous grid environment. Functional parallelism and data parallelism are supported by each of the message passing interface model and TCP/IP model. To consider the differences in heterogeneous computing resources and complexity of large-scale power system communities, a kernel-based multilevel algorithm is proposed for network partitioning. Since the bottleneck in distributed computation is low speed network communication, a bi-level latency exploitation technique is introduced for numerically solving system differential equations. The proposed grid-based implementation includes the core simulation engine, grid computing middleware, a Python interface and Python front-end utilities. Tests for a 39-bus network, a 4000-bus network and a 10,000-bus network are reported, and the results of these experiments demonstrate that the proposed scheme is able to execute the distributed simulations on computational grid infrastructure and provide efficient parallelism.

MPI microtask for programming the Cell Broadband Engine™ processor

The Cell Broadband Engine™ processor employs multiple accelerators, called synergistic processing elements (SPEs), for high performance. Each SPE has a high-speed local store attached to the main memory through direct memory access (DMA), but a drawback of this design is that the local store is not large enough for the entire application code or data. It must be decomposed into pieces small enough to fit into local memory, and they must be replaced through the DMA without losing the performance gain of multiple SPEs. We propose a new programming model, MPI microtask, based on the standard Message Passing Interface (MPI) programming model for distributed-memory parallel machines. In our new model, programmers do not need to manage the local store as long as they partition their application into a collection of small microtasks that fit into the local store. Furthermore, the preprocessor and runtime in our microtask system optimize the execution of microtasks by exploiting explicit communications in the MPI model. We have created a prototype that includes a novel static scheduler for such optimizations. Our initial experiments have shown some encouraging results.

Java on networks of workstations (JavaNOW): a parallel computing framework inspired by Linda and the Message Passing Interface (MPI)

Networks of workstations are a dominant force in the distributed computing arena, due primarily to the excellent price/performance ratio of such systems when compared to traditionally massively parallel architectures. It is therefore critical to develop programming languages and environments that can potentially harness the raw computational power availab le on these systems. In this article, we present JavaNOW (Java on Networks of Workstations), a Java based framework for parallel programming on networks of workstations. It creates a virtual parallel machine similar to the MPI (Message Passing Interface) model, and provides distributed associative shared memory similar to Linda memory model but with a flexible set of primitive operations. JavaNOW provides a simple yet powerful framework for performing computation on networks of workstations. In addition to the Linda memory model, it provides for shared objects, implicit multithreading, implicit synchronization, object dataflow, and collective communications similar to those defined in MPI. JavaNOW is also a component of the Computational Neighborhood [63], a Java-enabled suite of services for desktop computational sharing. The intent of JavaNOW is to present an environment for parallel computing that is both expressive and reliable and ultimately can deliver good to excellent performance. As JavaNOW is a work in progress, this article emphasizes the expressive potential of the JavaNOW environment and presents preliminary performance results only.