Message Passing Interface Library Research Articles

CPU 클러스터 구축 및 3차원 공간분할 병렬 FDTD 알고리즘 구현

본 연구에서는 빠르게 전자파 해석을 수행할 수 있는 병렬 유한차분 시간영역(Finite-Difference Time-Domain: FDTD) 알고리즘을 구현하기 위하여 CPU 클러스터를 구축하였다. 병렬 FDTD 알고리즘은 단일 프로세서를 이용한 FDTD 알고리즘에 비해 해석 시간을 크게 줄일 수 있으며, 전기적으로 매우 큰 구조물에 대한 전자파 해석도 가능하다. 본 연구팀에서는 CPU 클러스터 기반의 병렬 FDTD 알고리즘에서 요구되는 프로세스 간의 통신을 위해 MPI(Message Passing Interface) 라이브러리를 이용하였으며, 3차원 공간분할을 적용하여 프로세스 간의 통신 시간을 최소화하였다. 단일 프로세서를 이용한 FDTD 알고리즘 대비 CPU 클러스터 기반의 병렬 FDTD 알고리즘의 계산속도 향상도를 기본 모드와 하이퍼 모드에서 분석하였으며, 전기적으로 매우 큰 콘크리트 구조물의 전자파 해석을 하였다. In this work, we construct a CPU cluster to implement a parallel finite-difference time domain(FDTD) algorithm for fast electromagnetic analyses. This parallel FDTD algorithm can reduce the computational time significantly and also analyze electrically larger structures, compared to a single FDTD counterpart. The parallel FDTD algorithm needs communication between neighboring processors, which is performed by the MPI(Message Passing Interface) library and a 3-D domain decomposition is employed to decrease the communication time between neighboring processors. Compared to a single-processor FDTD, the speed up factor of a-CPU-cluster-based parallel FDTD algorithm is investigated for the normal mode and the hypermode and finally analyze an electrically large concrete structure by the developed parallel algorithm.

Performance evaluation of ultra-large-scale first-principles electronic structure calculation code on the K computer

Silicon nanowires are potentially useful in next-generation field-effect transistors, and it is important to clarify the electron states of silicon nanowires to know the behavior of new devices. Computer simulations are promising tools for calculating electron states. Real-space density functional theory (RSDFT) code performs first-principles electronic structure calculations. To obtain higher performance, we applied various optimization techniques to the code: multi-level parallelization, load balance management, sub-mesh/torus allocation, and a message-passing interface library tuned for the K computer. We measured and evaluated the performance of the modified RSDFT code on the K computer. A 5.48 petaflops (PFLOPS) sustained performance was measured for an iteration of a self-consistent field calculation for a 107,292-atom Si nanowire simulation using 82,944 compute nodes, which is 51.67% of the K computer’s peak performance of 10.62 PFLOPS. This scale of simulation enables analysis of the behavior of a silicon nanowire with a diameter of 10–20 nm.

Development and analysis of a tool for speed up of EnergyPlus through parallelization

This paper describes how the EPparallel tool splits a single annual simulation into 12 simulations of one month each and runs them in parallel. The paper also describes the methodology to prepare input files, enable file sharing between nodes, collate results generated by the nodes, and ensure quality check on the simulations. The EPparallel tool uses Message Passing Interface library and runs on Linux. The tool has been tested on 16 commercial reference buildings over 16 US weather files. The results of these 256 runs which include the run times, computing time overheads, speed gains and accuracy of results are presented in this paper. The speed gain ranged from 2.9×to 7.8×and deviation (percentage of output values obtained in parallel simulation which were off by more than 1% as compared to output values obtained in annual simulation) ranged from 0% to 4%.

Simulation of the Navier–Stokes equations in three dimensions with a spectral collocation method

SUMMARYThis paper describes a nonlinear, three‐dimensional spectral collocation method for the simulation of the incompressible Navier–Stokes equations under the Boussinesq approximation, motivated by geophysical and environmental flows. These flows are driven by the interaction of stratified fluid with topography, which this model accurately accounts for by using a mapped coordinate system. The spectral collocation method is implemented with both a Fourier trigonometric expansion and the Chebyshev polynomials, as appropriate for the domain boundary conditions. The coordinate mapping prohibits the use of existing, fast solution methods that rely on the separation of variables, so a preconditioner based on the approximate solution of a corresponding finite‐difference problem with geometric multigrid is used. The model is parallelized with the Message Passing Interface library, and it runs effectively on shared and distributed‐memory systems. Copyright © 2013 John Wiley & Sons, Ltd.

Improving the Reliability of MPI Libraries via Message Flow Checking

Despite the success of the Message Passing Interface (MPI), many MPI libraries have suffered from software bugs. These bugs severely impact the productivity of a large number of users, causing program failures or other errors. As a result, MPI application developers often have to spend days or weeks in vain debugging their own code. To address this daunting problem, this paper presents a new method called FlowChecker, which detects communication related bugs in MPI libraries. First, FlowChecker extracts program intentions of message passing (MP-intentions), which specify messages to be delivered from the sources to the destinations. Then FlowChecker tracks the message flows that actually occur in the underlying MPI libraries. Finally, FlowChecker checks whether the messages are correctly delivered from the sources to the destinations by comparing the message flows against the MP-intentions. If a mismatch is found, FlowChecker reports a bug and provides diagnostic information to help MPI library developers to understand and fix it. We have built a FlowChecker prototype on Linux and evaluated it with five real-world and two injected bug cases in three widely used MPI libraries, including Open MPI, MPICH2, and MVAPICH2. Our experimental results show that FlowChecker effectively detects all seven evaluated bug cases. Additionally, it provides useful diagnostic information for narrowing down or even pinpointing root causes of the bugs. Moreover, our experiments with High Performance Linpack and NAS Parallel Benchmarks show that FlowChecker induces low runtime overhead (0.9-5.6 percent on Open MPI, 0.9-8.1 percent on MPICH2, and 1.6-9.7 percent on MVAPICH2).

A PARALLEL MONTE CARLO CODE FOR SIMULATING COLLISIONAL N -BODY SYSTEMS

We present a new parallel code for computing the dynamical evolution of collisional N-body systems with up to N~10^7 particles. Our code is based on the the Henon Monte Carlo method for solving the Fokker-Planck equation, and makes assumptions of spherical symmetry and dynamical equilibrium. The principal algorithmic developments involve optimizing data structures, and the introduction of a parallel random number generation scheme, as well as a parallel sorting algorithm, required to find nearest neighbors for interactions and to compute the gravitational potential. The new algorithms we introduce along with our choice of decomposition scheme minimize communication costs and ensure optimal distribution of data and workload among the processing units. The implementation uses the Message Passing Interface (MPI) library for communication, which makes it portable to many different supercomputing architectures. We validate the code by calculating the evolution of clusters with initial Plummer distribution functions up to core collapse with the number of stars, N, spanning three orders of magnitude, from 10^5 to 10^7. We find that our results are in good agreement with self-similar core-collapse solutions, and the core collapse times generally agree with expectations from the literature. Also, we observe good total energy conservation, within less than 0.04% throughout all simulations. We analyze the performance of the code, and demonstrate near-linear scaling of the runtime with the number of processors up to 64 processors for N=10^5, 128 for N=10^6 and 256 for N=10^7. The runtime reaches a saturation with the addition of more processors beyond these limits which is a characteristic of the parallel sorting algorithm. The resulting maximum speedups we achieve are approximately 60x, 100x, and 220x, respectively.

Parallel Particle-Grid Hybrid Method for Nuclear Power System Accident Analysis

In this paper, the numerical analysis code of particle-grid hybrid method is developed by parallel computational technology. Particle-grid hybrid method is a new method for two-phase flow problem analysis. When this method was used to study the two-phase problem in nuclear power system the calculation time was very long. Hence, the MPI (Message Passing Interface) library is chosen as the parallel environment to parallel the original serial code. The grid and particle calculation parts are paralleled, separately. Jacobi point iteration method and ADI (alternating direction implicit) combined with divide and flow line programming techniques for grid calculation are discussed. Then several cases are set for testing the parallel efficiency of different pressure correction equation solution methods. This method has been proved to be accurate and reliable compared with the results of original serial code; it can shorten the calculating time to a certain degree. And this method provides valuable references and guidelines for further improvement of hybrid method and expansion of its application fields.

Parallelized finite difference time domain room acoustic simulation

A parallelized room acoustics simulator based on the finite difference time domain (FDTD) method is developed, utilizing a Blue Gene/L supercomputer. Wave-based methods such as FDTD are desirable for use in room acoustics simulations since they account for effects such as diffraction and interference. However, such methods require large amounts of computational power and memory, especially when simulating large volumes or high frequencies. To utilize the power of modern computing systems and move toward large-scale simulations of realistic concert halls, a parallel FDTD implementation is written in the C++ programming language with the Message Passing Interface (MPI) library. The volume to be simulated is partitioned into blocks, which efficiently update shared interfaces between nearest neighbors using the Blue Gene architecture's point-to-point communication network. Several compact explicit FDTD schemes are compared using simulations of various spatial volumes, executed on varying numbers of processors. The use of a Blue Gene/L supercomputer demonstrates substantial speedup over an equivalent serial implementation.

Numerical Simulation of 3D Hydraulic Fracturing Based on an Improved Flow-Stress-Damage Model and a Parallel FEM Technique

The failure mechanism of hydraulic fractures in heterogeneous geological materials is an important topic in mining and petroleum engineering. A three-dimensional (3D) finite element model that considers the coupled effects of seepage, damage, and the stress field is introduced. This model is based on a previously developed two-dimensional (2D) version of the model (RFPA2D-Rock Failure Process Analysis). The RFPA3D-Parallel model is developed using a parallel finite element method with a message-passing interface library. The constitutive law of this model considers strength and stiffness degradation, stress-dependent permeability for the pre-peak stage, and deformation-dependent permeability for the post-peak stage. Using this model, 3D modelling of progressive failure and associated fluid flow in rock are conducted and used to investigate the hydro-mechanical response of rock samples at laboratory scale. The responses investigated are the axial stress–axial strain together with permeability evolution and fracture patterns at various stages of loading. Then, the hydraulic fracturing process inside a rock specimen is numerically simulated. Three coupled processes are considered: (1) mechanical deformation of the solid medium induced by the fluid pressure acting on the fracture surfaces and the rock skeleton, (2) fluid flow within the fracture, and (3) propagation of the fracture. The numerically simulated results show that the fractures from a vertical wellbore propagate in the maximum principal stress direction without branching, turning, and twisting in the case of a large difference in the magnitude of the far-field stresses. Otherwise, the fracture initiates in a non-preferred direction and plane then turns and twists during propagation to become aligned with the preferred direction and plane. This pattern of fracturing is common when the rock formation contains multiple layers with different material properties. In addition, local heterogeneity of the rock matrix and macro-scale stress fluctuations due to the variability of material properties can cause the branching, turning, and twisting of fractures.

An overview of CMPI

Cloud computing enables users to perform distributed computing tasks on many virtual machines, without owning a physical cluster. Recently, various distributed computing tasks such as scientific applications are being moved from supercomputers and private clusters to public clouds. Message passing interface (MPI) is a key and common component in distributed computing tasks. The virtualized computing environment of the public cloud hides the network topology information from the users, and existing topology-aware optimizations for MPI are no longer feasible in the cloud environment. We propose a network performance aware MPI library named CMPI. CMPI embraces a new model for capturing the network performance among different virtual machines in the cloud. Based on the network performance model, we develop novel network performance aware algorithms for communication operations. This poster gives an overview of CMPI design, and presents some preliminary results on collective operations such as broadcast.We demonstrate the effectiveness of our network performance aware optimizations on Amazon EC2.

Study on the Numerical Simulation of Explosion and Impact Processes Using PC Cluster System

Parallel computing has been applied in many fields, and the parallel computing platform system, PC cluster based on MPI (Message Passing Interface) library under Linux operating system is a cost-effectiveness approach to parallel compute. In this paper, the key algorithm of parallel program of explosion and impact is presented. The techniques of solving data dependence and realizing communication between subdomain are proposed. From the test of program, the portability of MMIC-3D parallel program is satisfied, and compared with the single computer, PC cluster can improve the calculation speed and enlarge the scale greatly.

Accelerating Relevance‐Vector‐Machine‐Based Classification of Hyperspectral Image with Parallel Computing

Benefiting from the kernel skill and the sparse property, the relevance vector machine (RVM) could acquire a sparse solution, with an equivalent generalization ability compared with the support vector machine. The sparse property requires much less time in the prediction, making RVM potential in classifying the large‐scale hyperspectral image. However, RVM is not widespread influenced by its slow training procedure. To solve the problem, the classification of the hyperspectral image using RVM is accelerated by the parallel computing technique in this paper. The parallelization is revealed from the aspects of the multiclass strategy, the ensemble of multiple weak classifiers, and the matrix operations. The parallel RVMs are implemented using the C language plus the parallel functions of the linear algebra packages and the message passing interface library. The proposed methods are evaluated by the AVIRIS Indian Pines data set on the Beowulf cluster and the multicore platforms. It shows that the parallel RVMs accelerate the training procedure obviously.

Basis set effects on the Hartree–Fock description of confined many-electron atoms

In this work, the basis sets designed by Clementi, Bunge and Thakkar, for atomic systems, have been used to obtain the electronic structure of confined many-electron atoms by using Roothaan's approach in the Hartree–Fock context with a new code written in C, which uses the message-passing interface library. The confinement was imposed as Ludeña suggested to simulate walls with infinity potential. For closed-shell atoms, the Thakkar basis set functions give the best total energies (TE) as a function of the confinement radius, obtaining the following ordering: TE(Thakkar) < TE(Bunge) < TE(Clementi). However, for few open-shell atoms this ordering is not preserved and a trend, for the basis sets, is not observed. Although there are differences between the TE predicted by these basis set functions, the corresponding pressures are similar to each other; it means that changes in the total energy are described almost in the same way by using any of these basis sets. By analysing the total energy as a function of the inverse of the volume we propose an equation of state; for regions of small volumes, this equation predicts that the pressure is inversely proportional to the square of the volume.

Alternative, uniformly expressive and more scalable interfaces for collective communication in MPI

► Significantly extends the expressivity of MPI collective. ► Contributes toward solving known scalability problems of irregular MPI collectives. ► Introduces alternative set of datatype constructors, some new to MPI. In both the regular and the irregular MPI (Message-Passing Interface) collective communication and reduction interfaces there is a correspondence between the argument lists and certain MPI derived datatypes. As a means to address and alleviate well-known memory and performance scalability problems in the irregular (or vector ) collective interface definitions of MPI we propose to push this correspondence to its natural limit, and replace the interfaces of the MPI collectives with a different set of interfaces that specify all data sizes and displacements solely by means of derived datatypes. This reduces the number of collective (communication and reduction) interfaces from 16 to 10, significantly generalizes the operations, unifies regular and irregular collective interfaces, makes it possible to decouple certain algorithmic decisions from the collective operation, and moves the interface scalability issue from the collective interfaces to the MPI derived datatypes. To complete the proposal we discuss the memory scalability of the derived datatypes and suggest a number of alternative datatypes for MPI, some of which should be of independent interest. A running example illustrates the benefits of this alternative set of collective interfaces. Implementation issues are discussed showing that an implementation can be undertaken within any reasonable MPI library implementation.

Solution of the Skyrme–Hartree–Fock–Bogolyubov equations in the Cartesian deformed harmonic-oscillator basis.: (VII) hfodd (v2.49t): A new version of the program

We describe the new version (v2.49t) of the code HFODD which solves the nuclear Skyrme Hartree-Fock (HF) or Skyrme Hartree-Fock-Bogolyubov (HFB) problem by using the Cartesian deformed harmonic-oscillator basis. In the new version, we have implemented the following physics features: (i) the isospin mixing and projection, (ii) the finite temperature formalism for the HFB and HF+BCS methods, (iii) the Lipkin translational energy correction method, (iv) the calculation of the shell correction. A number of specific numerical methods have also been implemented in order to deal with large-scale multi-constraint calculations and hardware limitations: (i) the two-basis method for the HFB method, (ii) the Augmented Lagrangian Method (ALM) for multi-constraint calculations, (iii) the linear constraint method based on the approximation of the RPA matrix for multi-constraint calculations, (iv) an interface with the axial and parity-conserving Skyrme-HFB code HFBTHO, (v) the mixing of the HF or HFB matrix elements instead of the HF fields. Special care has been paid to using the code on massively parallel leadership class computers. For this purpose, the following features are now available with this version: (i) the Message Passing Interface (MPI) framework, (ii) scalable input data routines, (iii) multi-threading via OpenMP pragmas, (iv) parallel diagonalization of the HFB matrix in the simplex breaking case using the ScaLAPACK library. Finally, several little significant errors of the previous published version were corrected.

Influence of the compiler on multi-CPU performance of WRFv3

Abstract. The Weather Research and Forecasting system version 3 (WRFv3) is an open source and state of the art numerical Regional Climate Model used in climate related sciences. These years the model has been successfully optimized on a wide variety of clustered compute nodes connected with high speed interconnects. This is currently the most used hardware architecture for high-performance computing (Shainer et al., 2009). As such, understanding the influence of hardware like the CPU, its interconnects, or the software on WRFs performance is crucial for saving computing time. This is important because computing time in general is rare, resource intensive, and hence very expensive. This paper evaluates the influence of different compilers on WRFs performance, which was found to differ up to 26 %. The paper also evaluates the performance of different Message Passing Interface library versions, a software which is needed for multi CPU runs, and of different WRF versions. Both showed no significant influence on the performance for this test case on the used High Performance Cluster (HPC) hardware. Emphasis is also laid on the applied non-standard method of performance measuring, which was required because of performance fluctuations between identical runs on the used HPC. Those are caused by contention for network resources, a phenomenon examined for many HPCs (Wright et al., 2009).

Parallel multithreaded IDA* heuristic search: algorithm design and performance evaluation

Due to the witnessed prevalence of the commercial multi-core microprocessors, parallel programming becomes a dire need for efficiently using all available hardware resources for one application. One of the parallel programming approaches is multithreading, which has been proved to play a great role in providing sequential computers with virtual parallelisation, yielding faster execution and easy communication. Such advantageous features are provided through creating a dynamic number of concurrent threads at run-time of an application. Based on these facts, this paper presents a parallel multithreaded approach for the well-known iterative deepening A* (IDA*) heuristic search algorithm using POSIX threads (Pthreads) and message-passing interface libraries running on 16 dual-core processors. The feasibility of the parallel multithreaded approach is investigated as an alternative for hosting applications requiring intensive graph search. The analytical evaluation and experimental results revealed the improved performance achieved by the proposed parallel multithreaded IDA* algorithm.

A high efficiency parallel unstructured solver dedicated to internal combustion engine simulation

IFP-C3D, a hexahedral unstructured parallel solver dedicated to multiphysics calculation is being developed at IFP to compute the compressible combustion in internal engines. IFP-C3D uses an unstructured formalism, the finite volume method on staggered grid, time splitting, SIMPLE loop, subcycled advection, turbulent and Lagrangian spray and a liquid film model. Original algorithms and models such as the conditional temporal interpolation methodology for moving grids, the remapping algorithm for transferring quantities on different meshes during the computation enable IFP-C3D to deal with complex moving geometries with large volume deformation induced by all moving geometrical parts (intake/exhaust valve, piston). Large super-scalar machines up to 1000 processors are being widely used and IFP-C3D has been optimized for running on these Cluster machines. IFP-C3D is parallelized using the Message Passing Interface (MPI) library to distribute a calculation over a large number of processors. Moreover, IFP-C3D uses an optimized linear algebraic library to solve linear matrix systems and the METIS partitionner library to distribute the computational load equally for all meshes used during the calculation and in particular during the remap stage when new meshes are loaded. Numerical results and performance timing are presented to demonstrate the computational efficiency of the code.

RCS COMPUTATION USING A PARALLEL IN-CORE AND OUT-OF-CORE DIRECT SOLVER

Application to RCS computation of a higher order solver based on the surface integral approach is presented. The solver uses a direct method to solve the corresponding algebraic system of equations. Two versions of the solver are available: in-core and out-of-core. Both are e-ciently implemented as parallel codes using Message Passing Interface libraries. Several benchmark structures are analyzed showing the reliability, performance, and versatility to run in a wide variety of computer platforms, of the solver. The results shown are illustrative of what is the maximum frequency of analysis of the structures for a given type of simulation platform.

Dynamic workload balancing deques for branch and bound algorithms in the message passing interface

The message passing interface (MPI) is the standard in message passing parallel computation. MPI does not provide a canonical way to dynamically distribute run-time generated workload evenly across all the participating computer nodes. This paper proposes a MPI library to provide near-optimal dynamical workload balancing over branch and bound (B&B) algorithms; B&B potentially produces huge workload unbalance during a parallel execution. The library, named RaWSDM, provides a double ended queue (deque) data structure on which the programmer may pop, process, and later, pull back some parallel tasks; an underlying efficient system scheduler is responsible for keeping the workload balanced by exchanging elements among all deques. Theoretical bounds are traced and practical experiments are performed with the unlimited knapsack problem. Results show a performance gain up to 80% (best-case scenario) against a pure MPI implementation using round-robin scheduling, with near linear speedup and memory consumption.