Implementations of BLAST for parallel computers.

PubMed

Jülich, A

1995-02-01

The BLAST sequence comparison programs have been ported to a variety of parallel computers-the shared memory machine Cray Y-MP 8/864 and the distributed memory architectures Intel iPSC/860 and nCUBE. Additionally, the programs were ported to run on workstation clusters. We explain the parallelization techniques and consider the pros and cons of these methods. The BLAST programs are very well suited for parallelization for a moderate number of processors. We illustrate our results using the program blastp as an example. As input data for blastp, a 799 residue protein query sequence and the protein database PIR were used.

Portable programming on parallel/networked computers using the Application Portable Parallel Library (APPL)

NASA Technical Reports Server (NTRS)

Quealy, Angela; Cole, Gary L.; Blech, Richard A.

1993-01-01

The Application Portable Parallel Library (APPL) is a subroutine-based library of communication primitives that is callable from applications written in FORTRAN or C. APPL provides a consistent programmer interface to a variety of distributed and shared-memory multiprocessor MIMD machines. The objective of APPL is to minimize the effort required to move parallel applications from one machine to another, or to a network of homogeneous machines. APPL encompasses many of the message-passing primitives that are currently available on commercial multiprocessor systems. This paper describes APPL (version 2.3.1) and its usage, reports the status of the APPL project, and indicates possible directions for the future. Several applications using APPL are discussed, as well as performance and overhead results.

A performance study of sparse Cholesky factorization on INTEL iPSC/860

NASA Technical Reports Server (NTRS)

Zubair, M.; Ghose, M.

1992-01-01

The problem of Cholesky factorization of a sparse matrix has been very well investigated on sequential machines. A number of efficient codes exist for factorizing large unstructured sparse matrices. However, there is a lack of such efficient codes on parallel machines in general, and distributed machines in particular. Some of the issues that are critical to the implementation of sparse Cholesky factorization on a distributed memory parallel machine are ordering, partitioning and mapping, load balancing, and ordering of various tasks within a processor. Here, we focus on the effect of various partitioning schemes on the performance of sparse Cholesky factorization on the Intel iPSC/860. Also, a new partitioning heuristic for structured as well as unstructured sparse matrices is proposed, and its performance is compared with other schemes.

Distributed-Memory Computing With the Langley Aerothermodynamic Upwind Relaxation Algorithm (LAURA)

NASA Technical Reports Server (NTRS)

Riley, Christopher J.; Cheatwood, F. McNeil

1997-01-01

The Langley Aerothermodynamic Upwind Relaxation Algorithm (LAURA), a Navier-Stokes solver, has been modified for use in a parallel, distributed-memory environment using the Message-Passing Interface (MPI) standard. A standard domain decomposition strategy is used in which the computational domain is divided into subdomains with each subdomain assigned to a processor. Performance is examined on dedicated parallel machines and a network of desktop workstations. The effect of domain decomposition and frequency of boundary updates on performance and convergence is also examined for several realistic configurations and conditions typical of large-scale computational fluid dynamic analysis.

Paging memory from random access memory to backing storage in a parallel computer

DOEpatents

Archer, Charles J; Blocksome, Michael A; Inglett, Todd A; Ratterman, Joseph D; Smith, Brian E

2013-05-21

Paging memory from random access memory (`RAM`) to backing storage in a parallel computer that includes a plurality of compute nodes, including: executing a data processing application on a virtual machine operating system in a virtual machine on a first compute node; providing, by a second compute node, backing storage for the contents of RAM on the first compute node; and swapping, by the virtual machine operating system in the virtual machine on the first compute node, a page of memory from RAM on the first compute node to the backing storage on the second compute node.

Implementation and performance of parallel Prolog interpreter

DOE Office of Scientific and Technical Information (OSTI.GOV)

Wei, S.; Kale, L.V.; Balkrishna, R.

1988-01-01

In this paper, the authors discuss the implementation of a parallel Prolog interpreter on different parallel machines. The implementation is based on the REDUCE--OR process model which exploits both AND and OR parallelism in logic programs. It is machine independent as it runs on top of the chare-kernel--a machine-independent parallel programming system. The authors also give the performance of the interpreter running a diverse set of benchmark pargrams on parallel machines including shared memory systems: an Alliant FX/8, Sequent and a MultiMax, and a non-shared memory systems: Intel iPSC/32 hypercube, in addition to its performance on a multiprocessor simulation system.

Efficient diagonalization of the sparse matrices produced within the framework of the UK R-matrix molecular codes

NASA Astrophysics Data System (ADS)

Galiatsatos, P. G.; Tennyson, J.

2012-11-01

The most time consuming step within the framework of the UK R-matrix molecular codes is that of the diagonalization of the inner region Hamiltonian matrix (IRHM). Here we present the method that we follow to speed up this step. We use shared memory machines (SMM), distributed memory machines (DMM), the OpenMP directive based parallel language, the MPI function based parallel language, the sparse matrix diagonalizers ARPACK and PARPACK, a variation for real symmetric matrices of the official coordinate sparse matrix format and finally a parallel sparse matrix-vector product (PSMV). The efficient application of the previous techniques rely on two important facts: the sparsity of the matrix is large enough (more than 98%) and in order to get back converged results we need a small only part of the matrix spectrum.

Parallel performance investigations of an unstructured mesh Navier-Stokes solver

NASA Technical Reports Server (NTRS)

Mavriplis, Dimitri J.

2000-01-01

A Reynolds-averaged Navier-Stokes solver based on unstructured mesh techniques for analysis of high-lift configurations is described. The method makes use of an agglomeration multigrid solver for convergence acceleration. Implicit line-smoothing is employed to relieve the stiffness associated with highly stretched meshes. A GMRES technique is also implemented to speed convergence at the expense of additional memory usage. The solver is cache efficient and fully vectorizable, and is parallelized using a two-level hybrid MPI-OpenMP implementation suitable for shared and/or distributed memory architectures, as well as clusters of shared memory machines. Convergence and scalability results are illustrated for various high-lift cases.

Contention Modeling for Multithreaded Distributed Shared Memory Machines: The Cray XMT

DOE Office of Scientific and Technical Information (OSTI.GOV)

Secchi, Simone; Tumeo, Antonino; Villa, Oreste

Distributed Shared Memory (DSM) machines are a wide class of multi-processor computing systems where a large virtually-shared address space is mapped on a network of physically distributed memories. High memory latency and network contention are two of the main factors that limit performance scaling of such architectures. Modern high-performance computing DSM systems have evolved toward exploitation of massive hardware multi-threading and fine-grained memory hashing to tolerate irregular latencies, avoid network hot-spots and enable high scaling. In order to model the performance of such large-scale machines, parallel simulation has been proved to be a promising approach to achieve good accuracy inmore » reasonable times. One of the most critical factors in solving the simulation speed-accuracy trade-off is network modeling. The Cray XMT is a massively multi-threaded supercomputing architecture that belongs to the DSM class, since it implements a globally-shared address space abstraction on top of a physically distributed memory substrate. In this paper, we discuss the development of a contention-aware network model intended to be integrated in a full-system XMT simulator. We start by measuring the effects of network contention in a 128-processor XMT machine and then investigate the trade-off that exists between simulation accuracy and speed, by comparing three network models which operate at different levels of accuracy. The comparison and model validation is performed by executing a string-matching algorithm on the full-system simulator and on the XMT, using three datasets that generate noticeably different contention patterns.« less

Debugging Fortran on a shared memory machine

DOE Office of Scientific and Technical Information (OSTI.GOV)

Allen, T.R.; Padua, D.A.

1987-01-01

Debugging on a parallel processor is more difficult than debugging on a serial machine because errors in a parallel program may introduce nondeterminism. The approach to parallel debugging presented here attempts to reduce the problem of debugging on a parallel machine to that of debugging on a serial machine by automatically detecting nondeterminism. 20 refs., 6 figs.

Performance Comparison of a Set of Periodic and Non-Periodic Tridiagonal Solvers on SP2 and Paragon Parallel Computers

NASA Technical Reports Server (NTRS)

Sun, Xian-He; Moitra, Stuti

1996-01-01

Various tridiagonal solvers have been proposed in recent years for different parallel platforms. In this paper, the performance of three tridiagonal solvers, namely, the parallel partition LU algorithm, the parallel diagonal dominant algorithm, and the reduced diagonal dominant algorithm, is studied. These algorithms are designed for distributed-memory machines and are tested on an Intel Paragon and an IBM SP2 machines. Measured results are reported in terms of execution time and speedup. Analytical study are conducted for different communication topologies and for different tridiagonal systems. The measured results match the analytical results closely. In addition to address implementation issues, performance considerations such as problem sizes and models of speedup are also discussed.

Automatic Adaptation of Tunable Distributed Applications

DTIC Science & Technology

2001-01-01

size, weight, and battery life, with a single CPU, less memory, smaller hard disk, and lower bandwidth network connectivity. The power of PDAs is...wireless, and bluetooth [32] facilities; thus achieving different rates of data transmission. 1 With the trend of “write once, run everywhere...applications, a single component can execute on multiple processors (or machines) in parallel. These parallel applications, written in a specialized language

Three-Dimensional High-Lift Analysis Using a Parallel Unstructured Multigrid Solver

NASA Technical Reports Server (NTRS)

Mavriplis, Dimitri J.

1998-01-01

A directional implicit unstructured agglomeration multigrid solver is ported to shared and distributed memory massively parallel machines using the explicit domain-decomposition and message-passing approach. Because the algorithm operates on local implicit lines in the unstructured mesh, special care is required in partitioning the problem for parallel computing. A weighted partitioning strategy is described which avoids breaking the implicit lines across processor boundaries, while incurring minimal additional communication overhead. Good scalability is demonstrated on a 128 processor SGI Origin 2000 machine and on a 512 processor CRAY T3E machine for reasonably fine grids. The feasibility of performing large-scale unstructured grid calculations with the parallel multigrid algorithm is demonstrated by computing the flow over a partial-span flap wing high-lift geometry on a highly resolved grid of 13.5 million points in approximately 4 hours of wall clock time on the CRAY T3E.

Efficient Parallel Algorithm For Direct Numerical Simulation of Turbulent Flows

NASA Technical Reports Server (NTRS)

Moitra, Stuti; Gatski, Thomas B.

1997-01-01

A distributed algorithm for a high-order-accurate finite-difference approach to the direct numerical simulation (DNS) of transition and turbulence in compressible flows is described. This work has two major objectives. The first objective is to demonstrate that parallel and distributed-memory machines can be successfully and efficiently used to solve computationally intensive and input/output intensive algorithms of the DNS class. The second objective is to show that the computational complexity involved in solving the tridiagonal systems inherent in the DNS algorithm can be reduced by algorithm innovations that obviate the need to use a parallelized tridiagonal solver.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chen, Chao; Pouransari, Hadi; Rajamanickam, Sivasankaran

We present a parallel hierarchical solver for general sparse linear systems on distributed-memory machines. For large-scale problems, this fully algebraic algorithm is faster and more memory-efficient than sparse direct solvers because it exploits the low-rank structure of fill-in blocks. Depending on the accuracy of low-rank approximations, the hierarchical solver can be used either as a direct solver or as a preconditioner. The parallel algorithm is based on data decomposition and requires only local communication for updating boundary data on every processor. Moreover, the computation-to-communication ratio of the parallel algorithm is approximately the volume-to-surface-area ratio of the subdomain owned by everymore » processor. We also provide various numerical results to demonstrate the versatility and scalability of the parallel algorithm.« less

Frequent Statement and Dereference Elimination for Imperative and Object-Oriented Distributed Programs

PubMed Central

El-Zawawy, Mohamed A.

2014-01-01

This paper introduces new approaches for the analysis of frequent statement and dereference elimination for imperative and object-oriented distributed programs running on parallel machines equipped with hierarchical memories. The paper uses languages whose address spaces are globally partitioned. Distributed programs allow defining data layout and threads writing to and reading from other thread memories. Three type systems (for imperative distributed programs) are the tools of the proposed techniques. The first type system defines for every program point a set of calculated (ready) statements and memory accesses. The second type system uses an enriched version of types of the first type system and determines which of the ready statements and memory accesses are used later in the program. The third type system uses the information gather so far to eliminate unnecessary statement computations and memory accesses (the analysis of frequent statement and dereference elimination). Extensions to these type systems are also presented to cover object-oriented distributed programs. Two advantages of our work over related work are the following. The hierarchical style of concurrent parallel computers is similar to the memory model used in this paper. In our approach, each analysis result is assigned a type derivation (serves as a correctness proof). PMID:24892098

Parallelization of KENO-Va Monte Carlo code

NASA Astrophysics Data System (ADS)

Ramón, Javier; Peña, Jorge

1995-07-01

KENO-Va is a code integrated within the SCALE system developed by Oak Ridge that solves the transport equation through the Monte Carlo Method. It is being used at the Consejo de Seguridad Nuclear (CSN) to perform criticality calculations for fuel storage pools and shipping casks. Two parallel versions of the code: one for shared memory machines and other for distributed memory systems using the message-passing interface PVM have been generated. In both versions the neutrons of each generation are tracked in parallel. In order to preserve the reproducibility of the results in both versions, advanced seeds for random numbers were used. The CONVEX C3440 with four processors and shared memory at CSN was used to implement the shared memory version. A FDDI network of 6 HP9000/735 was employed to implement the message-passing version using proprietary PVM. The speedup obtained was 3.6 in both cases.

Implementing Molecular Dynamics for Hybrid High Performance Computers - 1. Short Range Forces

DOE Office of Scientific and Technical Information (OSTI.GOV)

Brown, W Michael; Wang, Peng; Plimpton, Steven J

The use of accelerators such as general-purpose graphics processing units (GPGPUs) have become popular in scientific computing applications due to their low cost, impressive floating-point capabilities, high memory bandwidth, and low electrical power requirements. Hybrid high performance computers, machines with more than one type of floating-point processor, are now becoming more prevalent due to these advantages. In this work, we discuss several important issues in porting a large molecular dynamics code for use on parallel hybrid machines - 1) choosing a hybrid parallel decomposition that works on central processing units (CPUs) with distributed memory and accelerator cores with shared memory,more » 2) minimizing the amount of code that must be ported for efficient acceleration, 3) utilizing the available processing power from both many-core CPUs and accelerators, and 4) choosing a programming model for acceleration. We present our solution to each of these issues for short-range force calculation in the molecular dynamics package LAMMPS. We describe algorithms for efficient short range force calculation on hybrid high performance machines. We describe a new approach for dynamic load balancing of work between CPU and accelerator cores. We describe the Geryon library that allows a single code to compile with both CUDA and OpenCL for use on a variety of accelerators. Finally, we present results on a parallel test cluster containing 32 Fermi GPGPUs and 180 CPU cores.« less

Impact of Load Balancing on Unstructured Adaptive Grid Computations for Distributed-Memory Multiprocessors

NASA Technical Reports Server (NTRS)

Sohn, Andrew; Biswas, Rupak; Simon, Horst D.

1996-01-01

The computational requirements for an adaptive solution of unsteady problems change as the simulation progresses. This causes workload imbalance among processors on a parallel machine which, in turn, requires significant data movement at runtime. We present a new dynamic load-balancing framework, called JOVE, that balances the workload across all processors with a global view. Whenever the computational mesh is adapted, JOVE is activated to eliminate the load imbalance. JOVE has been implemented on an IBM SP2 distributed-memory machine in MPI for portability. Experimental results for two model meshes demonstrate that mesh adaption with load balancing gives more than a sixfold improvement over one without load balancing. We also show that JOVE gives a 24-fold speedup on 64 processors compared to sequential execution.

The FORCE - A highly portable parallel programming language

NASA Technical Reports Server (NTRS)

Jordan, Harry F.; Benten, Muhammad S.; Alaghband, Gita; Jakob, Ruediger

1989-01-01

This paper explains why the FORCE parallel programming language is easily portable among six different shared-memory multiprocessors, and how a two-level macro preprocessor makes it possible to hide low-level machine dependencies and to build machine-independent high-level constructs on top of them. These FORCE constructs make it possible to write portable parallel programs largely independent of the number of processes and the specific shared-memory multiprocessor executing them.

The FORCE: A highly portable parallel programming language

NASA Technical Reports Server (NTRS)

Jordan, Harry F.; Benten, Muhammad S.; Alaghband, Gita; Jakob, Ruediger

1989-01-01

Here, it is explained why the FORCE parallel programming language is easily portable among six different shared-memory microprocessors, and how a two-level macro preprocessor makes it possible to hide low level machine dependencies and to build machine-independent high level constructs on top of them. These FORCE constructs make it possible to write portable parallel programs largely independent of the number of processes and the specific shared memory multiprocessor executing them.

Automatic selection of dynamic data partitioning schemes for distributed memory multicomputers

NASA Technical Reports Server (NTRS)

Palermo, Daniel J.; Banerjee, Prithviraj

1995-01-01

For distributed memory multicomputers such as the Intel Paragon, the IBM SP-2, the NCUBE/2, and the Thinking Machines CM-5, the quality of the data partitioning for a given application is crucial to obtaining high performance. This task has traditionally been the user's responsibility, but in recent years much effort has been directed to automating the selection of data partitioning schemes. Several researchers have proposed systems that are able to produce data distributions that remain in effect for the entire execution of an application. For complex programs, however, such static data distributions may be insufficient to obtain acceptable performance. The selection of distributions that dynamically change over the course of a program's execution adds another dimension to the data partitioning problem. In this paper, we present a technique that can be used to automatically determine which partitionings are most beneficial over specific sections of a program while taking into account the added overhead of performing redistribution. This system is being built as part of the PARADIGM (PARAllelizing compiler for DIstributed memory General-purpose Multicomputers) project at the University of Illinois. The complete system will provide a fully automated means to parallelize programs written in a serial programming model obtaining high performance on a wide range of distributed-memory multicomputers.

Highly parallel sparse Cholesky factorization

NASA Technical Reports Server (NTRS)

Gilbert, John R.; Schreiber, Robert

1990-01-01

Several fine grained parallel algorithms were developed and compared to compute the Cholesky factorization of a sparse matrix. The experimental implementations are on the Connection Machine, a distributed memory SIMD machine whose programming model conceptually supplies one processor per data element. In contrast to special purpose algorithms in which the matrix structure conforms to the connection structure of the machine, the focus is on matrices with arbitrary sparsity structure. The most promising algorithm is one whose inner loop performs several dense factorizations simultaneously on a 2-D grid of processors. Virtually any massively parallel dense factorization algorithm can be used as the key subroutine. The sparse code attains execution rates comparable to those of the dense subroutine. Although at present architectural limitations prevent the dense factorization from realizing its potential efficiency, it is concluded that a regular data parallel architecture can be used efficiently to solve arbitrarily structured sparse problems. A performance model is also presented and it is used to analyze the algorithms.

Parallel ALLSPD-3D: Speeding Up Combustor Analysis Via Parallel Processing

NASA Technical Reports Server (NTRS)

Fricker, David M.

1997-01-01

The ALLSPD-3D Computational Fluid Dynamics code for reacting flow simulation was run on a set of benchmark test cases to determine its parallel efficiency. These test cases included non-reacting and reacting flow simulations with varying numbers of processors. Also, the tests explored the effects of scaling the simulation with the number of processors in addition to distributing a constant size problem over an increasing number of processors. The test cases were run on a cluster of IBM RS/6000 Model 590 workstations with ethernet and ATM networking plus a shared memory SGI Power Challenge L workstation. The results indicate that the network capabilities significantly influence the parallel efficiency, i.e., a shared memory machine is fastest and ATM networking provides acceptable performance. The limitations of ethernet greatly hamper the rapid calculation of flows using ALLSPD-3D.

Kanerva's sparse distributed memory: An associative memory algorithm well-suited to the Connection Machine

NASA Technical Reports Server (NTRS)

Rogers, David

1988-01-01

The advent of the Connection Machine profoundly changes the world of supercomputers. The highly nontraditional architecture makes possible the exploration of algorithms that were impractical for standard Von Neumann architectures. Sparse distributed memory (SDM) is an example of such an algorithm. Sparse distributed memory is a particularly simple and elegant formulation for an associative memory. The foundations for sparse distributed memory are described, and some simple examples of using the memory are presented. The relationship of sparse distributed memory to three important computational systems is shown: random-access memory, neural networks, and the cerebellum of the brain. Finally, the implementation of the algorithm for sparse distributed memory on the Connection Machine is discussed.

Automatic data partitioning on distributed memory multicomputers. Ph.D. Thesis

NASA Technical Reports Server (NTRS)

Gupta, Manish

1992-01-01

Distributed-memory parallel computers are increasingly being used to provide high levels of performance for scientific applications. Unfortunately, such machines are not very easy to program. A number of research efforts seek to alleviate this problem by developing compilers that take over the task of generating communication. The communication overheads and the extent of parallelism exploited in the resulting target program are determined largely by the manner in which data is partitioned across different processors of the machine. Most of the compilers provide no assistance to the programmer in the crucial task of determining a good data partitioning scheme. A novel approach is presented, the constraints-based approach, to the problem of automatic data partitioning for numeric programs. In this approach, the compiler identifies some desirable requirements on the distribution of various arrays being referenced in each statement, based on performance considerations. These desirable requirements are referred to as constraints. For each constraint, the compiler determines a quality measure that captures its importance with respect to the performance of the program. The quality measure is obtained through static performance estimation, without actually generating the target data-parallel program with explicit communication. Each data distribution decision is taken by combining all the relevant constraints. The compiler attempts to resolve any conflicts between constraints such that the overall execution time of the parallel program is minimized. This approach has been implemented as part of a compiler called Paradigm, that accepts Fortran 77 programs, and specifies the partitioning scheme to be used for each array in the program. We have obtained results on some programs taken from the Linpack and Eispack libraries, and the Perfect Benchmarks. These results are quite promising, and demonstrate the feasibility of automatic data partitioning for a significant class of scientific application programs with regular computations.

Communication Studies of DMP and SMP Machines

NASA Technical Reports Server (NTRS)

Sohn, Andrew; Biswas, Rupak; Chancellor, Marisa K. (Technical Monitor)

1997-01-01

Understanding the interplay between machines and problems is key to obtaining high performance on parallel machines. This paper investigates the interplay between programming paradigms and communication capabilities of parallel machines. In particular, we explicate the communication capabilities of the IBM SP-2 distributed-memory multiprocessor and the SGI PowerCHALLENGEarray symmetric multiprocessor. Two benchmark problems of bitonic sorting and Fast Fourier Transform are selected for experiments. Communication-efficient algorithms are developed to exploit the overlapping capabilities of the machines. Programs are written in Message-Passing Interface for portability and identical codes are used for both machines. Various data sizes and message sizes are used to test the machines' communication capabilities. Experimental results indicate that the communication performance of the multiprocessors are consistent with the size of messages. The SP-2 is sensitive to message size but yields a much higher communication overlapping because of the communication co-processor. The PowerCHALLENGEarray is not highly sensitive to message size and yields a low communication overlapping. Bitonic sorting yields lower performance compared to FFT due to a smaller computation-to-communication ratio.

Efficient Machine Learning Approach for Optimizing Scientific Computing Applications on Emerging HPC Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Arumugam, Kamesh

Efficient parallel implementations of scientific applications on multi-core CPUs with accelerators such as GPUs and Xeon Phis is challenging. This requires - exploiting the data parallel architecture of the accelerator along with the vector pipelines of modern x86 CPU architectures, load balancing, and efficient memory transfer between different devices. It is relatively easy to meet these requirements for highly structured scientific applications. In contrast, a number of scientific and engineering applications are unstructured. Getting performance on accelerators for these applications is extremely challenging because many of these applications employ irregular algorithms which exhibit data-dependent control-ow and irregular memory accesses. Furthermore,more » these applications are often iterative with dependency between steps, and thus making it hard to parallelize across steps. As a result, parallelism in these applications is often limited to a single step. Numerical simulation of charged particles beam dynamics is one such application where the distribution of work and memory access pattern at each time step is irregular. Applications with these properties tend to present significant branch and memory divergence, load imbalance between different processor cores, and poor compute and memory utilization. Prior research on parallelizing such irregular applications have been focused around optimizing the irregular, data-dependent memory accesses and control-ow during a single step of the application independent of the other steps, with the assumption that these patterns are completely unpredictable. We observed that the structure of computation leading to control-ow divergence and irregular memory accesses in one step is similar to that in the next step. It is possible to predict this structure in the current step by observing the computation structure of previous steps. In this dissertation, we present novel machine learning based optimization techniques to address the parallel implementation challenges of such irregular applications on different HPC architectures. In particular, we use supervised learning to predict the computation structure and use it to address the control-ow and memory access irregularities in the parallel implementation of such applications on GPUs, Xeon Phis, and heterogeneous architectures composed of multi-core CPUs with GPUs or Xeon Phis. We use numerical simulation of charged particles beam dynamics simulation as a motivating example throughout the dissertation to present our new approach, though they should be equally applicable to a wide range of irregular applications. The machine learning approach presented here use predictive analytics and forecasting techniques to adaptively model and track the irregular memory access pattern at each time step of the simulation to anticipate the future memory access pattern. Access pattern forecasts can then be used to formulate optimization decisions during application execution which improves the performance of the application at a future time step based on the observations from earlier time steps. In heterogeneous architectures, forecasts can also be used to improve the memory performance and resource utilization of all the processing units to deliver a good aggregate performance. We used these optimization techniques and anticipation strategy to design a cache-aware, memory efficient parallel algorithm to address the irregularities in the parallel implementation of charged particles beam dynamics simulation on different HPC architectures. Experimental result using a diverse mix of HPC architectures shows that our approach in using anticipation strategy is effective in maximizing data reuse, ensuring workload balance, minimizing branch and memory divergence, and in improving resource utilization.« less

RTNN: The New Parallel Machine in Zaragoza

NASA Astrophysics Data System (ADS)

Sijs, A. J. V. D.

I report on the development of RTNN, a parallel computer designed as a 4^4 hypercube of 256 T9000 transputer nodes, each with 8 MB memory. The peak performance of the machine is expected to be 2.5 Gflops.

The Tera Multithreaded Architecture and Unstructured Meshes

NASA Technical Reports Server (NTRS)

Bokhari, Shahid H.; Mavriplis, Dimitri J.

1998-01-01

The Tera Multithreaded Architecture (MTA) is a new parallel supercomputer currently being installed at San Diego Supercomputing Center (SDSC). This machine has an architecture quite different from contemporary parallel machines. The computational processor is a custom design and the machine uses hardware to support very fine grained multithreading. The main memory is shared, hardware randomized and flat. These features make the machine highly suited to the execution of unstructured mesh problems, which are difficult to parallelize on other architectures. We report the results of a study carried out during July-August 1998 to evaluate the execution of EUL3D, a code that solves the Euler equations on an unstructured mesh, on the 2 processor Tera MTA at SDSC. Our investigation shows that parallelization of an unstructured code is extremely easy on the Tera. We were able to get an existing parallel code (designed for a shared memory machine), running on the Tera by changing only the compiler directives. Furthermore, a serial version of this code was compiled to run in parallel on the Tera by judicious use of directives to invoke the "full/empty" tag bits of the machine to obtain synchronization. This version achieves 212 and 406 Mflop/s on one and two processors respectively, and requires no attention to partitioning or placement of data issues that would be of paramount importance in other parallel architectures.

Proceedings: Sisal `93

DOE Office of Scientific and Technical Information (OSTI.GOV)

Feo, J.T.

1993-10-01

This report contain papers on: Programmability and performance issues; The case of an iterative partial differential equation solver; Implementing the kernal of the Australian Region Weather Prediction Model in Sisal; Even and quarter-even prime length symmetric FFTs and their Sisal Implementations; Top-down thread generation for Sisal; Overlapping communications and computations on NUMA architechtures; Compiling technique based on dataflow analysis for funtional programming language Valid; Copy elimination for true multidimensional arrays in Sisal 2.0; Increasing parallelism for an optimization that reduces copying in IF2 graphs; Caching in on Sisal; Cache performance of Sisal Vs. FORTRAN; FFT algorithms on a shared-memory multiprocessor;more » A parallel implementation of nonnumeric search problems in Sisal; Computer vision algorithms in Sisal; Compilation of Sisal for a high-performance data driven vector processor; Sisal on distributed memory machines; A virtual shared addressing system for distributed memory Sisal; Developing a high-performance FFT algorithm in Sisal for a vector supercomputer; Implementation issues for IF2 on a static data-flow architechture; and Systematic control of parallelism in array-based data-flow computation. Selected papers have been indexed separately for inclusion in the Energy Science and Technology Database.« less

Parallel spatial direct numerical simulations on the Intel iPSC/860 hypercube

NASA Technical Reports Server (NTRS)

Joslin, Ronald D.; Zubair, Mohammad

1993-01-01

The implementation and performance of a parallel spatial direct numerical simulation (PSDNS) approach on the Intel iPSC/860 hypercube is documented. The direct numerical simulation approach is used to compute spatially evolving disturbances associated with the laminar-to-turbulent transition in boundary-layer flows. The feasibility of using the PSDNS on the hypercube to perform transition studies is examined. The results indicate that the direct numerical simulation approach can effectively be parallelized on a distributed-memory parallel machine. By increasing the number of processors nearly ideal linear speedups are achieved with nonoptimized routines; slower than linear speedups are achieved with optimized (machine dependent library) routines. This slower than linear speedup results because the Fast Fourier Transform (FFT) routine dominates the computational cost and because the routine indicates less than ideal speedups. However with the machine-dependent routines the total computational cost decreases by a factor of 4 to 5 compared with standard FORTRAN routines. The computational cost increases linearly with spanwise wall-normal and streamwise grid refinements. The hypercube with 32 processors was estimated to require approximately twice the amount of Cray supercomputer single processor time to complete a comparable simulation; however it is estimated that a subgrid-scale model which reduces the required number of grid points and becomes a large-eddy simulation (PSLES) would reduce the computational cost and memory requirements by a factor of 10 over the PSDNS. This PSLES implementation would enable transition simulations on the hypercube at a reasonable computational cost.

Unstructured grids on SIMD torus machines

NASA Technical Reports Server (NTRS)

Bjorstad, Petter E.; Schreiber, Robert

1994-01-01

Unstructured grids lead to unstructured communication on distributed memory parallel computers, a problem that has been considered difficult. Here, we consider adaptive, offline communication routing for a SIMD processor grid. Our approach is empirical. We use large data sets drawn from supercomputing applications instead of an analytic model of communication load. The chief contribution of this paper is an experimental demonstration of the effectiveness of certain routing heuristics. Our routing algorithm is adaptive, nonminimal, and is generally designed to exploit locality. We have a parallel implementation of the router, and we report on its performance.

Efficient Parallel Kernel Solvers for Computational Fluid Dynamics Applications

NASA Technical Reports Server (NTRS)

Sun, Xian-He

1997-01-01

Distributed-memory parallel computers dominate today's parallel computing arena. These machines, such as Intel Paragon, IBM SP2, and Cray Origin2OO, have successfully delivered high performance computing power for solving some of the so-called "grand-challenge" problems. Despite initial success, parallel machines have not been widely accepted in production engineering environments due to the complexity of parallel programming. On a parallel computing system, a task has to be partitioned and distributed appropriately among processors to reduce communication cost and to attain load balance. More importantly, even with careful partitioning and mapping, the performance of an algorithm may still be unsatisfactory, since conventional sequential algorithms may be serial in nature and may not be implemented efficiently on parallel machines. In many cases, new algorithms have to be introduced to increase parallel performance. In order to achieve optimal performance, in addition to partitioning and mapping, a careful performance study should be conducted for a given application to find a good algorithm-machine combination. This process, however, is usually painful and elusive. The goal of this project is to design and develop efficient parallel algorithms for highly accurate Computational Fluid Dynamics (CFD) simulations and other engineering applications. The work plan is 1) developing highly accurate parallel numerical algorithms, 2) conduct preliminary testing to verify the effectiveness and potential of these algorithms, 3) incorporate newly developed algorithms into actual simulation packages. The work plan has well achieved. Two highly accurate, efficient Poisson solvers have been developed and tested based on two different approaches: (1) Adopting a mathematical geometry which has a better capacity to describe the fluid, (2) Using compact scheme to gain high order accuracy in numerical discretization. The previously developed Parallel Diagonal Dominant (PDD) algorithm and Reduced Parallel Diagonal Dominant (RPDD) algorithm have been carefully studied on different parallel platforms for different applications, and a NASA simulation code developed by Man M. Rai and his colleagues has been parallelized and implemented based on data dependency analysis. These achievements are addressed in detail in the paper.

Distributed computing for membrane-based modeling of action potential propagation.

PubMed

Porras, D; Rogers, J M; Smith, W M; Pollard, A E

2000-08-01

Action potential propagation simulations with physiologic membrane currents and macroscopic tissue dimensions are computationally expensive. We, therefore, analyzed distributed computing schemes to reduce execution time in workstation clusters by parallelizing solutions with message passing. Four schemes were considered in two-dimensional monodomain simulations with the Beeler-Reuter membrane equations. Parallel speedups measured with each scheme were compared to theoretical speedups, recognizing the relationship between speedup and code portions that executed serially. A data decomposition scheme based on total ionic current provided the best performance. Analysis of communication latencies in that scheme led to a load-balancing algorithm in which measured speedups at 89 +/- 2% and 75 +/- 8% of theoretical speedups were achieved in homogeneous and heterogeneous clusters of workstations. Speedups in this scheme with the Luo-Rudy dynamic membrane equations exceeded 3.0 with eight distributed workstations. Cluster speedups were comparable to those measured during parallel execution on a shared memory machine.

Architectures for reasoning in parallel

NASA Technical Reports Server (NTRS)

Hall, Lawrence O.

1989-01-01

The research conducted has dealt with rule-based expert systems. The algorithms that may lead to effective parallelization of them were investigated. Both the forward and backward chained control paradigms were investigated in the course of this work. The best computer architecture for the developed and investigated algorithms has been researched. Two experimental vehicles were developed to facilitate this research. They are Backpac, a parallel backward chained rule-based reasoning system and Datapac, a parallel forward chained rule-based reasoning system. Both systems have been written in Multilisp, a version of Lisp which contains the parallel construct, future. Applying the future function to a function causes the function to become a task parallel to the spawning task. Additionally, Backpac and Datapac have been run on several disparate parallel processors. The machines are an Encore Multimax with 10 processors, the Concert Multiprocessor with 64 processors, and a 32 processor BBN GP1000. Both the Concert and the GP1000 are switch-based machines. The Multimax has all its processors hung off a common bus. All are shared memory machines, but have different schemes for sharing the memory and different locales for the shared memory. The main results of the investigations come from experiments on the 10 processor Encore and the Concert with partitions of 32 or less processors. Additionally, experiments have been run with a stripped down version of EMYCIN.

The OpenMP Implementation of NAS Parallel Benchmarks and its Performance

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Frumkin, Michael; Yan, Jerry

1999-01-01

As the new ccNUMA architecture became popular in recent years, parallel programming with compiler directives on these machines has evolved to accommodate new needs. In this study, we examine the effectiveness of OpenMP directives for parallelizing the NAS Parallel Benchmarks. Implementation details will be discussed and performance will be compared with the MPI implementation. We have demonstrated that OpenMP can achieve very good results for parallelization on a shared memory system, but effective use of memory and cache is very important.

Low Latency Messages on Distributed Memory Multiprocessors

DOE PAGES

Rosing, Matt; Saltz, Joel

1995-01-01

This article describes many of the issues in developing an efficient interface for communication on distributed memory machines. Although the hardware component of message latency is less than 1 ws on many distributed memory machines, the software latency associated with sending and receiving typed messages is on the order of 50 μs. The reason for this imbalance is that the software interface does not match the hardware. By changing the interface to match the hardware more closely, applications with fine grained communication can be put on these machines. This article describes several tests performed and many of the issues involvedmore » in supporting low latency messages on distributed memory machines.« less

Optimizing NEURON Simulation Environment Using Remote Memory Access with Recursive Doubling on Distributed Memory Systems.

PubMed

Shehzad, Danish; Bozkuş, Zeki

2016-01-01

Increase in complexity of neuronal network models escalated the efforts to make NEURON simulation environment efficient. The computational neuroscientists divided the equations into subnets amongst multiple processors for achieving better hardware performance. On parallel machines for neuronal networks, interprocessor spikes exchange consumes large section of overall simulation time. In NEURON for communication between processors Message Passing Interface (MPI) is used. MPI_Allgather collective is exercised for spikes exchange after each interval across distributed memory systems. The increase in number of processors though results in achieving concurrency and better performance but it inversely affects MPI_Allgather which increases communication time between processors. This necessitates improving communication methodology to decrease the spikes exchange time over distributed memory systems. This work has improved MPI_Allgather method using Remote Memory Access (RMA) by moving two-sided communication to one-sided communication, and use of recursive doubling mechanism facilitates achieving efficient communication between the processors in precise steps. This approach enhanced communication concurrency and has improved overall runtime making NEURON more efficient for simulation of large neuronal network models.

Optimizing NEURON Simulation Environment Using Remote Memory Access with Recursive Doubling on Distributed Memory Systems

PubMed Central

Bozkuş, Zeki

2016-01-01

Increase in complexity of neuronal network models escalated the efforts to make NEURON simulation environment efficient. The computational neuroscientists divided the equations into subnets amongst multiple processors for achieving better hardware performance. On parallel machines for neuronal networks, interprocessor spikes exchange consumes large section of overall simulation time. In NEURON for communication between processors Message Passing Interface (MPI) is used. MPI_Allgather collective is exercised for spikes exchange after each interval across distributed memory systems. The increase in number of processors though results in achieving concurrency and better performance but it inversely affects MPI_Allgather which increases communication time between processors. This necessitates improving communication methodology to decrease the spikes exchange time over distributed memory systems. This work has improved MPI_Allgather method using Remote Memory Access (RMA) by moving two-sided communication to one-sided communication, and use of recursive doubling mechanism facilitates achieving efficient communication between the processors in precise steps. This approach enhanced communication concurrency and has improved overall runtime making NEURON more efficient for simulation of large neuronal network models. PMID:27413363

An Adaptive Memory Interface Controller for Improving Bandwidth Utilization of Hybrid and Reconfigurable Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Castellana, Vito G.; Tumeo, Antonino; Ferrandi, Fabrizio

Emerging applications such as data mining, bioinformatics, knowledge discovery, social network analysis are irregular. They use data structures based on pointers or linked lists, such as graphs, unbalanced trees or unstructures grids, which generates unpredictable memory accesses. These data structures usually are large, but difficult to partition. These applications mostly are memory bandwidth bounded and have high synchronization intensity. However, they also have large amounts of inherent dynamic parallelism, because they potentially perform a task for each one of the element they are exploring. Several efforts are looking at accelerating these applications on hybrid architectures, which integrate general purpose processorsmore » with reconfigurable devices. Some solutions, which demonstrated significant speedups, include custom-hand tuned accelerators or even full processor architectures on the reconfigurable logic. In this paper we present an approach for the automatic synthesis of accelerators from C, targeted at irregular applications. In contrast to typical High Level Synthesis paradigms, which construct a centralized Finite State Machine, our approach generates dynamically scheduled hardware components. While parallelism exploitation in typical HLS-generated accelerators is usually bound within a single execution flow, our solution allows concurrently running multiple execution flow, thus also exploiting the coarser grain task parallelism of irregular applications. Our approach supports multiple, multi-ported and distributed memories, and atomic memory operations. Its main objective is parallelizing as many memory operations as possible, independently from their execution time, to maximize the memory bandwidth utilization. This significantly differs from current HLS flows, which usually consider a single memory port and require precise scheduling of memory operations. A key innovation of our approach is the generation of a memory interface controller, which dynamically maps concurrent memory accesses to multiple ports. We present a case study on a typical irregular kernel, Graph Breadth First search (BFS), exploring different tradeoffs in terms of parallelism and number of memories.« less

Parallel implementation of D-Phylo algorithm for maximum likelihood clusters.

PubMed

Malik, Shamita; Sharma, Dolly; Khatri, Sunil Kumar

2017-03-01

This study explains a newly developed parallel algorithm for phylogenetic analysis of DNA sequences. The newly designed D-Phylo is a more advanced algorithm for phylogenetic analysis using maximum likelihood approach. The D-Phylo while misusing the seeking capacity of k -means keeps away from its real constraint of getting stuck at privately conserved motifs. The authors have tested the behaviour of D-Phylo on Amazon Linux Amazon Machine Image(Hardware Virtual Machine)i2.4xlarge, six central processing unit, 122 GiB memory, 8  ×  800 Solid-state drive Elastic Block Store volume, high network performance up to 15 processors for several real-life datasets. Distributing the clusters evenly on all the processors provides us the capacity to accomplish a near direct speed if there should arise an occurrence of huge number of processors.

Run-time scheduling and execution of loops on message passing machines

NASA Technical Reports Server (NTRS)

Crowley, Kay; Saltz, Joel; Mirchandaney, Ravi; Berryman, Harry

1989-01-01

Sparse system solvers and general purpose codes for solving partial differential equations are examples of the many types of problems whose irregularity can result in poor performance on distributed memory machines. Often, the data structures used in these problems are very flexible. Crucial details concerning loop dependences are encoded in these structures rather than being explicitly represented in the program. Good methods for parallelizing and partitioning these types of problems require assignment of computations in rather arbitrary ways. Naive implementations of programs on distributed memory machines requiring general loop partitions can be extremely inefficient. Instead, the scheduling mechanism needs to capture the data reference patterns of the loops in order to partition the problem. First, the indices assigned to each processor must be locally numbered. Next, it is necessary to precompute what information is needed by each processor at various points in the computation. The precomputed information is then used to generate an execution template designed to carry out the computation, communication, and partitioning of data, in an optimized manner. The design is presented for a general preprocessor and schedule executer, the structures of which do not vary, even though the details of the computation and of the type of information are problem dependent.

Run-time scheduling and execution of loops on message passing machines

NASA Technical Reports Server (NTRS)

Saltz, Joel; Crowley, Kathleen; Mirchandaney, Ravi; Berryman, Harry

1990-01-01

Sparse system solvers and general purpose codes for solving partial differential equations are examples of the many types of problems whose irregularity can result in poor performance on distributed memory machines. Often, the data structures used in these problems are very flexible. Crucial details concerning loop dependences are encoded in these structures rather than being explicitly represented in the program. Good methods for parallelizing and partitioning these types of problems require assignment of computations in rather arbitrary ways. Naive implementations of programs on distributed memory machines requiring general loop partitions can be extremely inefficient. Instead, the scheduling mechanism needs to capture the data reference patterns of the loops in order to partition the problem. First, the indices assigned to each processor must be locally numbered. Next, it is necessary to precompute what information is needed by each processor at various points in the computation. The precomputed information is then used to generate an execution template designed to carry out the computation, communication, and partitioning of data, in an optimized manner. The design is presented for a general preprocessor and schedule executer, the structures of which do not vary, even though the details of the computation and of the type of information are problem dependent.

Parallel Simulation of Unsteady Turbulent Flames

NASA Technical Reports Server (NTRS)

Menon, Suresh

1996-01-01

Time-accurate simulation of turbulent flames in high Reynolds number flows is a challenging task since both fluid dynamics and combustion must be modeled accurately. To numerically simulate this phenomenon, very large computer resources (both time and memory) are required. Although current vector supercomputers are capable of providing adequate resources for simulations of this nature, the high cost and their limited availability, makes practical use of such machines less than satisfactory. At the same time, the explicit time integration algorithms used in unsteady flow simulations often possess a very high degree of parallelism, making them very amenable to efficient implementation on large-scale parallel computers. Under these circumstances, distributed memory parallel computers offer an excellent near-term solution for greatly increased computational speed and memory, at a cost that may render the unsteady simulations of the type discussed above more feasible and affordable.This paper discusses the study of unsteady turbulent flames using a simulation algorithm that is capable of retaining high parallel efficiency on distributed memory parallel architectures. Numerical studies are carried out using large-eddy simulation (LES). In LES, the scales larger than the grid are computed using a time- and space-accurate scheme, while the unresolved small scales are modeled using eddy viscosity based subgrid models. This is acceptable for the moment/energy closure since the small scales primarily provide a dissipative mechanism for the energy transferred from the large scales. However, for combustion to occur, the species must first undergo mixing at the small scales and then come into molecular contact. Therefore, global models cannot be used. Recently, a new model for turbulent combustion was developed, in which the combustion is modeled, within the subgrid (small-scales) using a methodology that simulates the mixing and the molecular transport and the chemical kinetics within each LES grid cell. Finite-rate kinetics can be included without any closure and this approach actually provides a means to predict the turbulent rates and the turbulent flame speed. The subgrid combustion model requires resolution of the local time scales associated with small-scale mixing, molecular diffusion and chemical kinetics and, therefore, within each grid cell, a significant amount of computations must be carried out before the large-scale (LES resolved) effects are incorporated. Therefore, this approach is uniquely suited for parallel processing and has been implemented on various systems such as: Intel Paragon, IBM SP-2, Cray T3D and SGI Power Challenge (PC) using the system independent Message Passing Interface (MPI) compiler. In this paper, timing data on these machines is reported along with some characteristic results.

Parallel grid generation algorithm for distributed memory computers

NASA Technical Reports Server (NTRS)

Moitra, Stuti; Moitra, Anutosh

1994-01-01

A parallel grid-generation algorithm and its implementation on the Intel iPSC/860 computer are described. The grid-generation scheme is based on an algebraic formulation of homotopic relations. Methods for utilizing the inherent parallelism of the grid-generation scheme are described, and implementation of multiple levELs of parallelism on multiple instruction multiple data machines are indicated. The algorithm is capable of providing near orthogonality and spacing control at solid boundaries while requiring minimal interprocessor communications. Results obtained on the Intel hypercube for a blended wing-body configuration are used to demonstrate the effectiveness of the algorithm. Fortran implementations bAsed on the native programming model of the iPSC/860 computer and the Express system of software tools are reported. Computational gains in execution time speed-up ratios are given.

Mobile and replicated alignment of arrays in data-parallel programs

NASA Technical Reports Server (NTRS)

Chatterjee, Siddhartha; Gilbert, John R.; Schreiber, Robert

1993-01-01

When a data-parallel language like FORTRAN 90 is compiled for a distributed-memory machine, aggregate data objects (such as arrays) are distributed across the processor memories. The mapping determines the amount of residual communication needed to bring operands of parallel operations into alignment with each other. A common approach is to break the mapping into two stages: first, an alignment that maps all the objects to an abstract template, and then a distribution that maps the template to the processors. We solve two facets of the problem of finding alignments that reduce residual communication: we determine alignments that vary in loops, and objects that should have replicated alignments. We show that loop-dependent mobile alignment is sometimes necessary for optimum performance, and we provide algorithms with which a compiler can determine good mobile alignments for objects within do loops. We also identify situations in which replicated alignment is either required by the program itself (via spread operations) or can be used to improve performance. We propose an algorithm based on network flow that determines which objects to replicate so as to minimize the total amount of broadcast communication in replication. This work on mobile and replicated alignment extends our earlier work on determining static alignment.

Scalable parallel distance field construction for large-scale applications

DOE PAGES

Yu, Hongfeng; Xie, Jinrong; Ma, Kwan -Liu; ...

2015-10-01

Computing distance fields is fundamental to many scientific and engineering applications. Distance fields can be used to direct analysis and reduce data. In this paper, we present a highly scalable method for computing 3D distance fields on massively parallel distributed-memory machines. Anew distributed spatial data structure, named parallel distance tree, is introduced to manage the level sets of data and facilitate surface tracking overtime, resulting in significantly reduced computation and communication costs for calculating the distance to the surface of interest from any spatial locations. Our method supports several data types and distance metrics from real-world applications. We demonstrate itsmore » efficiency and scalability on state-of-the-art supercomputers using both large-scale volume datasets and surface models. We also demonstrate in-situ distance field computation on dynamic turbulent flame surfaces for a petascale combustion simulation. In conclusion, our work greatly extends the usability of distance fields for demanding applications.« less

Scalable Parallel Distance Field Construction for Large-Scale Applications.

PubMed

Yu, Hongfeng; Xie, Jinrong; Ma, Kwan-Liu; Kolla, Hemanth; Chen, Jacqueline H

2015-10-01

Computing distance fields is fundamental to many scientific and engineering applications. Distance fields can be used to direct analysis and reduce data. In this paper, we present a highly scalable method for computing 3D distance fields on massively parallel distributed-memory machines. A new distributed spatial data structure, named parallel distance tree, is introduced to manage the level sets of data and facilitate surface tracking over time, resulting in significantly reduced computation and communication costs for calculating the distance to the surface of interest from any spatial locations. Our method supports several data types and distance metrics from real-world applications. We demonstrate its efficiency and scalability on state-of-the-art supercomputers using both large-scale volume datasets and surface models. We also demonstrate in-situ distance field computation on dynamic turbulent flame surfaces for a petascale combustion simulation. Our work greatly extends the usability of distance fields for demanding applications.

Review and analysis of dense linear system solver package for distributed memory machines

NASA Technical Reports Server (NTRS)

Narang, H. N.

1993-01-01

A dense linear system solver package recently developed at the University of Texas at Austin for distributed memory machine (e.g. Intel Paragon) has been reviewed and analyzed. The package contains about 45 software routines, some written in FORTRAN, and some in C-language, and forms the basis for parallel/distributed solutions of systems of linear equations encountered in many problems of scientific and engineering nature. The package, being studied by the Computer Applications Branch of the Analysis and Computation Division, may provide a significant computational resource for NASA scientists and engineers in parallel/distributed computing. Since the package is new and not well tested or documented, many of its underlying concepts and implementations were unclear; our task was to review, analyze, and critique the package as a step in the process that will enable scientists and engineers to apply it to the solution of their problems. All routines in the package were reviewed and analyzed. Underlying theory or concepts which exist in the form of published papers or technical reports, or memos, were either obtained from the author, or from the scientific literature; and general algorithms, explanations, examples, and critiques have been provided to explain the workings of these programs. Wherever the things were still unclear, communications were made with the developer (author), either by telephone or by electronic mail, to understand the workings of the routines. Whenever possible, tests were made to verify the concepts and logic employed in their implementations. A detailed report is being separately documented to explain the workings of these routines.

Portable parallel stochastic optimization for the design of aeropropulsion components

NASA Technical Reports Server (NTRS)

Sues, Robert H.; Rhodes, G. S.

1994-01-01

This report presents the results of Phase 1 research to develop a methodology for performing large-scale Multi-disciplinary Stochastic Optimization (MSO) for the design of aerospace systems ranging from aeropropulsion components to complete aircraft configurations. The current research recognizes that such design optimization problems are computationally expensive, and require the use of either massively parallel or multiple-processor computers. The methodology also recognizes that many operational and performance parameters are uncertain, and that uncertainty must be considered explicitly to achieve optimum performance and cost. The objective of this Phase 1 research was to initialize the development of an MSO methodology that is portable to a wide variety of hardware platforms, while achieving efficient, large-scale parallelism when multiple processors are available. The first effort in the project was a literature review of available computer hardware, as well as review of portable, parallel programming environments. The first effort was to implement the MSO methodology for a problem using the portable parallel programming language, Parallel Virtual Machine (PVM). The third and final effort was to demonstrate the example on a variety of computers, including a distributed-memory multiprocessor, a distributed-memory network of workstations, and a single-processor workstation. Results indicate the MSO methodology can be well-applied towards large-scale aerospace design problems. Nearly perfect linear speedup was demonstrated for computation of optimization sensitivity coefficients on both a 128-node distributed-memory multiprocessor (the Intel iPSC/860) and a network of workstations (speedups of almost 19 times achieved for 20 workstations). Very high parallel efficiencies (75 percent for 31 processors and 60 percent for 50 processors) were also achieved for computation of aerodynamic influence coefficients on the Intel. Finally, the multi-level parallelization strategy that will be needed for large-scale MSO problems was demonstrated to be highly efficient. The same parallel code instructions were used on both platforms, demonstrating portability. There are many applications for which MSO can be applied, including NASA's High-Speed-Civil Transport, and advanced propulsion systems. The use of MSO will reduce design and development time and testing costs dramatically.

On the relationship between parallel computation and graph embedding

DOE Office of Scientific and Technical Information (OSTI.GOV)

Gupta, A.K.

1989-01-01

The problem of efficiently simulating an algorithm designed for an n-processor parallel machine G on an m-processor parallel machine H with n > m arises when parallel algorithms designed for an ideal size machine are simulated on existing machines which are of a fixed size. The author studies this problem when every processor of H takes over the function of a number of processors in G, and he phrases the simulation problem as a graph embedding problem. New embeddings presented address relevant issues arising from the parallel computation environment. The main focus centers around embedding complete binary trees into smaller-sizedmore » binary trees, butterflies, and hypercubes. He also considers simultaneous embeddings of r source machines into a single hypercube. Constant factors play a crucial role in his embeddings since they are not only important in practice but also lead to interesting theoretical problems. All of his embeddings minimize dilation and load, which are the conventional cost measures in graph embeddings and determine the maximum amount of time required to simulate one step of G on H. His embeddings also optimize a new cost measure called ({alpha},{beta})-utilization which characterizes how evenly the processors of H are used by the processors of G. Ideally, the utilization should be balanced (i.e., every processor of H simulates at most (n/m) processors of G) and the ({alpha},{beta})-utilization measures how far off from a balanced utilization the embedding is. He presents embeddings for the situation when some processors of G have different capabilities (e.g. memory or I/O) than others and the processors with different capabilities are to be distributed uniformly among the processors of H. Placing such conditions on an embedding results in an increase in some of the cost measures.« less

A real-time MPEG software decoder using a portable message-passing library

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kwong, Man Kam; Tang, P.T. Peter; Lin, Biquan

1995-12-31

We present a real-time MPEG software decoder that uses message-passing libraries such as MPL, p4 and MPI. The parallel MPEG decoder currently runs on the IBM SP system but can be easil ported to other parallel machines. This paper discusses our parallel MPEG decoding algorithm as well as the parallel programming environment under which it uses. Several technical issues are discussed, including balancing of decoding speed, memory limitation, 1/0 capacities, and optimization of MPEG decoding components. This project shows that a real-time portable software MPEG decoder is feasible in a general-purpose parallel machine.

Parallel Implementation of Triangular Cellular Automata for Computing Two-Dimensional Elastodynamic Response on Arbitrary Domains

NASA Astrophysics Data System (ADS)

Leamy, Michael J.; Springer, Adam C.

In this research we report parallel implementation of a Cellular Automata-based simulation tool for computing elastodynamic response on complex, two-dimensional domains. Elastodynamic simulation using Cellular Automata (CA) has recently been presented as an alternative, inherently object-oriented technique for accurately and efficiently computing linear and nonlinear wave propagation in arbitrarily-shaped geometries. The local, autonomous nature of the method should lead to straight-forward and efficient parallelization. We address this notion on symmetric multiprocessor (SMP) hardware using a Java-based object-oriented CA code implementing triangular state machines (i.e., automata) and the MPI bindings written in Java (MPJ Express). We use MPJ Express to reconfigure our existing CA code to distribute a domain's automata to cores present on a dual quad-core shared-memory system (eight total processors). We note that this message passing parallelization strategy is directly applicable to computer clustered computing, which will be the focus of follow-on research. Results on the shared memory platform indicate nearly-ideal, linear speed-up. We conclude that the CA-based elastodynamic simulator is easily configured to run in parallel, and yields excellent speed-up on SMP hardware.

A Pervasive Parallel Processing Framework for Data Visualization and Analysis at Extreme Scale

DOE Office of Scientific and Technical Information (OSTI.GOV)

Moreland, Kenneth; Geveci, Berk

2014-11-01

The evolution of the computing world from teraflop to petaflop has been relatively effortless, with several of the existing programming models scaling effectively to the petascale. The migration to exascale, however, poses considerable challenges. All industry trends infer that the exascale machine will be built using processors containing hundreds to thousands of cores per chip. It can be inferred that efficient concurrency on exascale machines requires a massive amount of concurrent threads, each performing many operations on a localized piece of data. Currently, visualization libraries and applications are based off what is known as the visualization pipeline. In the pipelinemore » model, algorithms are encapsulated as filters with inputs and outputs. These filters are connected by setting the output of one component to the input of another. Parallelism in the visualization pipeline is achieved by replicating the pipeline for each processing thread. This works well for today’s distributed memory parallel computers but cannot be sustained when operating on processors with thousands of cores. Our project investigates a new visualization framework designed to exhibit the pervasive parallelism necessary for extreme scale machines. Our framework achieves this by defining algorithms in terms of worklets, which are localized stateless operations. Worklets are atomic operations that execute when invoked unlike filters, which execute when a pipeline request occurs. The worklet design allows execution on a massive amount of lightweight threads with minimal overhead. Only with such fine-grained parallelism can we hope to fill the billions of threads we expect will be necessary for efficient computation on an exascale machine.« less

Runtime support for parallelizing data mining algorithms

NASA Astrophysics Data System (ADS)

Jin, Ruoming; Agrawal, Gagan

2002-03-01

With recent technological advances, shared memory parallel machines have become more scalable, and offer large main memories and high bus bandwidths. They are emerging as good platforms for data warehousing and data mining. In this paper, we focus on shared memory parallelization of data mining algorithms. We have developed a series of techniques for parallelization of data mining algorithms, including full replication, full locking, fixed locking, optimized full locking, and cache-sensitive locking. Unlike previous work on shared memory parallelization of specific data mining algorithms, all of our techniques apply to a large number of common data mining algorithms. In addition, we propose a reduction-object based interface for specifying a data mining algorithm. We show how our runtime system can apply any of the technique we have developed starting from a common specification of the algorithm.

Implementing Shared Memory Parallelism in MCBEND

NASA Astrophysics Data System (ADS)

Bird, Adam; Long, David; Dobson, Geoff

2017-09-01

MCBEND is a general purpose radiation transport Monte Carlo code from AMEC Foster Wheelers's ANSWERS® Software Service. MCBEND is well established in the UK shielding community for radiation shielding and dosimetry assessments. The existing MCBEND parallel capability effectively involves running the same calculation on many processors. This works very well except when the memory requirements of a model restrict the number of instances of a calculation that will fit on a machine. To more effectively utilise parallel hardware OpenMP has been used to implement shared memory parallelism in MCBEND. This paper describes the reasoning behind the choice of OpenMP, notes some of the challenges of multi-threading an established code such as MCBEND and assesses the performance of the parallel method implemented in MCBEND.

Implementation and Characterization of Three-Dimensional Particle-in-Cell Codes on Multiple-Instruction-Multiple-Data Massively Parallel Supercomputers

NASA Technical Reports Server (NTRS)

Lyster, P. M.; Liewer, P. C.; Decyk, V. K.; Ferraro, R. D.

1995-01-01

A three-dimensional electrostatic particle-in-cell (PIC) plasma simulation code has been developed on coarse-grain distributed-memory massively parallel computers with message passing communications. Our implementation is the generalization to three-dimensions of the general concurrent particle-in-cell (GCPIC) algorithm. In the GCPIC algorithm, the particle computation is divided among the processors using a domain decomposition of the simulation domain. In a three-dimensional simulation, the domain can be partitioned into one-, two-, or three-dimensional subdomains ("slabs," "rods," or "cubes") and we investigate the efficiency of the parallel implementation of the push for all three choices. The present implementation runs on the Intel Touchstone Delta machine at Caltech; a multiple-instruction-multiple-data (MIMD) parallel computer with 512 nodes. We find that the parallel efficiency of the push is very high, with the ratio of communication to computation time in the range 0.3%-10.0%. The highest efficiency (> 99%) occurs for a large, scaled problem with 64(sup 3) particles per processing node (approximately 134 million particles of 512 nodes) which has a push time of about 250 ns per particle per time step. We have also developed expressions for the timing of the code which are a function of both code parameters (number of grid points, particles, etc.) and machine-dependent parameters (effective FLOP rate, and the effective interprocessor bandwidths for the communication of particles and grid points). These expressions can be used to estimate the performance of scaled problems--including those with inhomogeneous plasmas--to other parallel machines once the machine-dependent parameters are known.

Parallel computation and the basis system

DOE Office of Scientific and Technical Information (OSTI.GOV)

Smith, G.R.

1993-05-01

A software package has been written that can facilitate efforts to develop powerful, flexible, and easy-to use programs that can run in single-processor, massively parallel, and distributed computing environments. Particular attention has been given to the difficulties posed by a program consisting of many science packages that represent subsystems of a complicated, coupled system. Methods have been found to maintain independence of the packages by hiding data structures without increasing the communications costs in a parallel computing environment. Concepts developed in this work are demonstrated by a prototype program that uses library routines from two existing software systems, Basis andmore » Parallel Virtual Machine (PVM). Most of the details of these libraries have been encapsulated in routines and macros that could be rewritten for alternative libraries that possess certain minimum capabilities. The prototype software uses a flexible master-and-slaves paradigm for parallel computation and supports domain decomposition with message passing for partitioning work among slaves. Facilities are provided for accessing variables that are distributed among the memories of slaves assigned to subdomains. The software is named PROTOPAR.« less

Second International Workshop on Software Engineering and Code Design in Parallel Meteorological and Oceanographic Applications

NASA Technical Reports Server (NTRS)

OKeefe, Matthew (Editor); Kerr, Christopher L. (Editor)

1998-01-01

This report contains the abstracts and technical papers from the Second International Workshop on Software Engineering and Code Design in Parallel Meteorological and Oceanographic Applications, held June 15-18, 1998, in Scottsdale, Arizona. The purpose of the workshop is to bring together software developers in meteorology and oceanography to discuss software engineering and code design issues for parallel architectures, including Massively Parallel Processors (MPP's), Parallel Vector Processors (PVP's), Symmetric Multi-Processors (SMP's), Distributed Shared Memory (DSM) multi-processors, and clusters. Issues to be discussed include: (1) code architectures for current parallel models, including basic data structures, storage allocation, variable naming conventions, coding rules and styles, i/o and pre/post-processing of data; (2) designing modular code; (3) load balancing and domain decomposition; (4) techniques that exploit parallelism efficiently yet hide the machine-related details from the programmer; (5) tools for making the programmer more productive; and (6) the proliferation of programming models (F--, OpenMP, MPI, and HPF).

Parallel computing for probabilistic fatigue analysis

NASA Technical Reports Server (NTRS)

Sues, Robert H.; Lua, Yuan J.; Smith, Mark D.

1993-01-01

This paper presents the results of Phase I research to investigate the most effective parallel processing software strategies and hardware configurations for probabilistic structural analysis. We investigate the efficiency of both shared and distributed-memory architectures via a probabilistic fatigue life analysis problem. We also present a parallel programming approach, the virtual shared-memory paradigm, that is applicable across both types of hardware. Using this approach, problems can be solved on a variety of parallel configurations, including networks of single or multiprocessor workstations. We conclude that it is possible to effectively parallelize probabilistic fatigue analysis codes; however, special strategies will be needed to achieve large-scale parallelism to keep large number of processors busy and to treat problems with the large memory requirements encountered in practice. We also conclude that distributed-memory architecture is preferable to shared-memory for achieving large scale parallelism; however, in the future, the currently emerging hybrid-memory architectures will likely be optimal.

Load Balancing Unstructured Adaptive Grids for CFD Problems

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Oliker, Leonid

1996-01-01

Mesh adaption is a powerful tool for efficient unstructured-grid computations but causes load imbalance among processors on a parallel machine. A dynamic load balancing method is presented that balances the workload across all processors with a global view. After each parallel tetrahedral mesh adaption, the method first determines if the new mesh is sufficiently unbalanced to warrant a repartitioning. If so, the adapted mesh is repartitioned, with new partitions assigned to processors so that the redistribution cost is minimized. The new partitions are accepted only if the remapping cost is compensated by the improved load balance. Results indicate that this strategy is effective for large-scale scientific computations on distributed-memory multiprocessors.

Shared Memory Parallelization of an Implicit ADI-type CFD Code

NASA Technical Reports Server (NTRS)

Hauser, Th.; Huang, P. G.

1999-01-01

A parallelization study designed for ADI-type algorithms is presented using the OpenMP specification for shared-memory multiprocessor programming. Details of optimizations specifically addressed to cache-based computer architectures are described and performance measurements for the single and multiprocessor implementation are summarized. The paper demonstrates that optimization of memory access on a cache-based computer architecture controls the performance of the computational algorithm. A hybrid MPI/OpenMP approach is proposed for clusters of shared memory machines to further enhance the parallel performance. The method is applied to develop a new LES/DNS code, named LESTool. A preliminary DNS calculation of a fully developed channel flow at a Reynolds number of 180, Re(sub tau) = 180, has shown good agreement with existing data.

Global Load Balancing with Parallel Mesh Adaption on Distributed-Memory Systems

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Oliker, Leonid; Sohn, Andrew

1996-01-01

Dynamic mesh adaption on unstructured grids is a powerful tool for efficiently computing unsteady problems to resolve solution features of interest. Unfortunately, this causes load imbalance among processors on a parallel machine. This paper describes the parallel implementation of a tetrahedral mesh adaption scheme and a new global load balancing method. A heuristic remapping algorithm is presented that assigns partitions to processors such that the redistribution cost is minimized. Results indicate that the parallel performance of the mesh adaption code depends on the nature of the adaption region and show a 35.5X speedup on 64 processors of an SP2 when 35% of the mesh is randomly adapted. For large-scale scientific computations, our load balancing strategy gives almost a sixfold reduction in solver execution times over non-balanced loads. Furthermore, our heuristic remapper yields processor assignments that are less than 3% off the optimal solutions but requires only 1% of the computational time.

Summer Proceedings 2016: The Center for Computing Research at Sandia National Laboratories

DOE Office of Scientific and Technical Information (OSTI.GOV)

Carleton, James Brian; Parks, Michael L.

Solving sparse linear systems from the discretization of elliptic partial differential equations (PDEs) is an important building block in many engineering applications. Sparse direct solvers can solve general linear systems, but are usually slower and use much more memory than effective iterative solvers. To overcome these two disadvantages, a hierarchical solver (LoRaSp) based on H2-matrices was introduced in [22]. Here, we have developed a parallel version of the algorithm in LoRaSp to solve large sparse matrices on distributed memory machines. On a single processor, the factorization time of our parallel solver scales almost linearly with the problem size for three-dimensionalmore » problems, as opposed to the quadratic scalability of many existing sparse direct solvers. Moreover, our solver leads to almost constant numbers of iterations, when used as a preconditioner for Poisson problems. On more than one processor, our algorithm has significant speedups compared to sequential runs. With this parallel algorithm, we are able to solve large problems much faster than many existing packages as demonstrated by the numerical experiments.« less

Hybrid MPI+OpenMP Programming of an Overset CFD Solver and Performance Investigations

NASA Technical Reports Server (NTRS)

Djomehri, M. Jahed; Jin, Haoqiang H.; Biegel, Bryan (Technical Monitor)

2002-01-01

This report describes a two level parallelization of a Computational Fluid Dynamic (CFD) solver with multi-zone overset structured grids. The approach is based on a hybrid MPI+OpenMP programming model suitable for shared memory and clusters of shared memory machines. The performance investigations of the hybrid application on an SGI Origin2000 (O2K) machine is reported using medium and large scale test problems.

Parallel computation and the Basis system

DOE Office of Scientific and Technical Information (OSTI.GOV)

Smith, G.R.

1992-12-16

A software package has been written that can facilitate efforts to develop powerful, flexible, and easy-to-use programs that can run in single-processor, massively parallel, and distributed computing environments. Particular attention has been given to the difficulties posed by a program consisting of many science packages that represent subsystems of a complicated, coupled system. Methods have been found to maintain independence of the packages by hiding data structures without increasing the communication costs in a parallel computing environment. Concepts developed in this work are demonstrated by a prototype program that uses library routines from two existing software systems, Basis and Parallelmore » Virtual Machine (PVM). Most of the details of these libraries have been encapsulated in routines and macros that could be rewritten for alternative libraries that possess certain minimum capabilities. The prototype software uses a flexible master-and-slaves paradigm for parallel computation and supports domain decomposition with message passing for partitioning work among slaves. Facilities are provided for accessing variables that are distributed among the memories of slaves assigned to subdomains. The software is named PROTOPAR.« less

Scalable parallel communications

NASA Technical Reports Server (NTRS)

Maly, K.; Khanna, S.; Overstreet, C. M.; Mukkamala, R.; Zubair, M.; Sekhar, Y. S.; Foudriat, E. C.

1992-01-01

Coarse-grain parallelism in networking (that is, the use of multiple protocol processors running replicated software sending over several physical channels) can be used to provide gigabit communications for a single application. Since parallel network performance is highly dependent on real issues such as hardware properties (e.g., memory speeds and cache hit rates), operating system overhead (e.g., interrupt handling), and protocol performance (e.g., effect of timeouts), we have performed detailed simulations studies of both a bus-based multiprocessor workstation node (based on the Sun Galaxy MP multiprocessor) and a distributed-memory parallel computer node (based on the Touchstone DELTA) to evaluate the behavior of coarse-grain parallelism. Our results indicate: (1) coarse-grain parallelism can deliver multiple 100 Mbps with currently available hardware platforms and existing networking protocols (such as Transmission Control Protocol/Internet Protocol (TCP/IP) and parallel Fiber Distributed Data Interface (FDDI) rings); (2) scale-up is near linear in n, the number of protocol processors, and channels (for small n and up to a few hundred Mbps); and (3) since these results are based on existing hardware without specialized devices (except perhaps for some simple modifications of the FDDI boards), this is a low cost solution to providing multiple 100 Mbps on current machines. In addition, from both the performance analysis and the properties of these architectures, we conclude: (1) multiple processors providing identical services and the use of space division multiplexing for the physical channels can provide better reliability than monolithic approaches (it also provides graceful degradation and low-cost load balancing); (2) coarse-grain parallelism supports running several transport protocols in parallel to provide different types of service (for example, one TCP handles small messages for many users, other TCP's running in parallel provide high bandwidth service to a single application); and (3) coarse grain parallelism will be able to incorporate many future improvements from related work (e.g., reduced data movement, fast TCP, fine-grain parallelism) also with near linear speed-ups.

Implementing Molecular Dynamics on Hybrid High Performance Computers - Particle-Particle Particle-Mesh

DOE Office of Scientific and Technical Information (OSTI.GOV)

Brown, W Michael; Kohlmeyer, Axel; Plimpton, Steven J

The use of accelerators such as graphics processing units (GPUs) has become popular in scientific computing applications due to their low cost, impressive floating-point capabilities, high memory bandwidth, and low electrical power requirements. Hybrid high-performance computers, machines with nodes containing more than one type of floating-point processor (e.g. CPU and GPU), are now becoming more prevalent due to these advantages. In this paper, we present a continuation of previous work implementing algorithms for using accelerators into the LAMMPS molecular dynamics software for distributed memory parallel hybrid machines. In our previous work, we focused on acceleration for short-range models with anmore » approach intended to harness the processing power of both the accelerator and (multi-core) CPUs. To augment the existing implementations, we present an efficient implementation of long-range electrostatic force calculation for molecular dynamics. Specifically, we present an implementation of the particle-particle particle-mesh method based on the work by Harvey and De Fabritiis. We present benchmark results on the Keeneland InfiniBand GPU cluster. We provide a performance comparison of the same kernels compiled with both CUDA and OpenCL. We discuss limitations to parallel efficiency and future directions for improving performance on hybrid or heterogeneous computers.« less

Parallelization Issues and Particle-In Codes.

NASA Astrophysics Data System (ADS)

Elster, Anne Cathrine

1994-01-01

"Everything should be made as simple as possible, but not simpler." Albert Einstein. The field of parallel scientific computing has concentrated on parallelization of individual modules such as matrix solvers and factorizers. However, many applications involve several interacting modules. Our analyses of a particle-in-cell code modeling charged particles in an electric field, show that these accompanying dependencies affect data partitioning and lead to new parallelization strategies concerning processor, memory and cache utilization. Our test-bed, a KSR1, is a distributed memory machine with a globally shared addressing space. However, most of the new methods presented hold generally for hierarchical and/or distributed memory systems. We introduce a novel approach that uses dual pointers on the local particle arrays to keep the particle locations automatically partially sorted. Complexity and performance analyses with accompanying KSR benchmarks, have been included for both this scheme and for the traditional replicated grids approach. The latter approach maintains load-balance with respect to particles. However, our results demonstrate it fails to scale properly for problems with large grids (say, greater than 128-by-128) running on as few as 15 KSR nodes, since the extra storage and computation time associated with adding the grid copies, becomes significant. Our grid partitioning scheme, although harder to implement, does not need to replicate the whole grid. Consequently, it scales well for large problems on highly parallel systems. It may, however, require load balancing schemes for non-uniform particle distributions. Our dual pointer approach may facilitate this through dynamically partitioned grids. We also introduce hierarchical data structures that store neighboring grid-points within the same cache -line by reordering the grid indexing. This alignment produces a 25% savings in cache-hits for a 4-by-4 cache. A consideration of the input data's effect on the simulation may lead to further improvements. For example, in the case of mean particle drift, it is often advantageous to partition the grid primarily along the direction of the drift. The particle-in-cell codes for this study were tested using physical parameters, which lead to predictable phenomena including plasma oscillations and two-stream instabilities. An overview of the most central references related to parallel particle codes is also given.

Job Management Requirements for NAS Parallel Systems and Clusters

NASA Technical Reports Server (NTRS)

Saphir, William; Tanner, Leigh Ann; Traversat, Bernard

1995-01-01

A job management system is a critical component of a production supercomputing environment, permitting oversubscribed resources to be shared fairly and efficiently. Job management systems that were originally designed for traditional vector supercomputers are not appropriate for the distributed-memory parallel supercomputers that are becoming increasingly important in the high performance computing industry. Newer job management systems offer new functionality but do not solve fundamental problems. We address some of the main issues in resource allocation and job scheduling we have encountered on two parallel computers - a 160-node IBM SP2 and a cluster of 20 high performance workstations located at the Numerical Aerodynamic Simulation facility. We describe the requirements for resource allocation and job management that are necessary to provide a production supercomputing environment on these machines, prioritizing according to difficulty and importance, and advocating a return to fundamental issues.

Toward an automated parallel computing environment for geosciences

NASA Astrophysics Data System (ADS)

Zhang, Huai; Liu, Mian; Shi, Yaolin; Yuen, David A.; Yan, Zhenzhen; Liang, Guoping

2007-08-01

Software for geodynamic modeling has not kept up with the fast growing computing hardware and network resources. In the past decade supercomputing power has become available to most researchers in the form of affordable Beowulf clusters and other parallel computer platforms. However, to take full advantage of such computing power requires developing parallel algorithms and associated software, a task that is often too daunting for geoscience modelers whose main expertise is in geosciences. We introduce here an automated parallel computing environment built on open-source algorithms and libraries. Users interact with this computing environment by specifying the partial differential equations, solvers, and model-specific properties using an English-like modeling language in the input files. The system then automatically generates the finite element codes that can be run on distributed or shared memory parallel machines. This system is dynamic and flexible, allowing users to address different problems in geosciences. It is capable of providing web-based services, enabling users to generate source codes online. This unique feature will facilitate high-performance computing to be integrated with distributed data grids in the emerging cyber-infrastructures for geosciences. In this paper we discuss the principles of this automated modeling environment and provide examples to demonstrate its versatility.

Streaming data analytics via message passing with application to graph algorithms

DOE PAGES

Plimpton, Steven J.; Shead, Tim

2014-05-06

The need to process streaming data, which arrives continuously at high-volume in real-time, arises in a variety of contexts including data produced by experiments, collections of environmental or network sensors, and running simulations. Streaming data can also be formulated as queries or transactions which operate on a large dynamic data store, e.g. a distributed database. We describe a lightweight, portable framework named PHISH which enables a set of independent processes to compute on a stream of data in a distributed-memory parallel manner. Datums are routed between processes in patterns defined by the application. PHISH can run on top of eithermore » message-passing via MPI or sockets via ZMQ. The former means streaming computations can be run on any parallel machine which supports MPI; the latter allows them to run on a heterogeneous, geographically dispersed network of machines. We illustrate how PHISH can support streaming MapReduce operations, and describe streaming versions of three algorithms for large, sparse graph analytics: triangle enumeration, subgraph isomorphism matching, and connected component finding. Lastly, we also provide benchmark timings for MPI versus socket performance of several kernel operations useful in streaming algorithms.« less

Toward Millions of File System IOPS on Low-Cost, Commodity Hardware

PubMed Central

Zheng, Da; Burns, Randal; Szalay, Alexander S.

2013-01-01

We describe a storage system that removes I/O bottlenecks to achieve more than one million IOPS based on a user-space file abstraction for arrays of commodity SSDs. The file abstraction refactors I/O scheduling and placement for extreme parallelism and non-uniform memory and I/O. The system includes a set-associative, parallel page cache in the user space. We redesign page caching to eliminate CPU overhead and lock-contention in non-uniform memory architecture machines. We evaluate our design on a 32 core NUMA machine with four, eight-core processors. Experiments show that our design delivers 1.23 million 512-byte read IOPS. The page cache realizes the scalable IOPS of Linux asynchronous I/O (AIO) and increases user-perceived I/O performance linearly with cache hit rates. The parallel, set-associative cache matches the cache hit rates of the global Linux page cache under real workloads. PMID:24402052

Toward Millions of File System IOPS on Low-Cost, Commodity Hardware.

PubMed

Zheng, Da; Burns, Randal; Szalay, Alexander S

2013-01-01

We describe a storage system that removes I/O bottlenecks to achieve more than one million IOPS based on a user-space file abstraction for arrays of commodity SSDs. The file abstraction refactors I/O scheduling and placement for extreme parallelism and non-uniform memory and I/O. The system includes a set-associative, parallel page cache in the user space. We redesign page caching to eliminate CPU overhead and lock-contention in non-uniform memory architecture machines. We evaluate our design on a 32 core NUMA machine with four, eight-core processors. Experiments show that our design delivers 1.23 million 512-byte read IOPS. The page cache realizes the scalable IOPS of Linux asynchronous I/O (AIO) and increases user-perceived I/O performance linearly with cache hit rates. The parallel, set-associative cache matches the cache hit rates of the global Linux page cache under real workloads.

Dynamic Load Balancing for Adaptive Computations on Distributed-Memory Machines

NASA Technical Reports Server (NTRS)

1999-01-01

Dynamic load balancing is central to adaptive mesh-based computations on large-scale parallel computers. The principal investigator has investigated various issues on the dynamic load balancing problem under NASA JOVE and JAG rants. The major accomplishments of the project are two graph partitioning algorithms and a load balancing framework. The S-HARP dynamic graph partitioner is known to be the fastest among the known dynamic graph partitioners to date. It can partition a graph of over 100,000 vertices in 0.25 seconds on a 64- processor Cray T3E distributed-memory multiprocessor while maintaining the scalability of over 16-fold speedup. Other known and widely used dynamic graph partitioners take over a second or two while giving low scalability of a few fold speedup on 64 processors. These results have been published in journals and peer-reviewed flagship conferences.

A Parallel Vector Machine for the PM Programming Language

NASA Astrophysics Data System (ADS)

Bellerby, Tim

2016-04-01

PM is a new programming language which aims to make the writing of computational geoscience models on parallel hardware accessible to scientists who are not themselves expert parallel programmers. It is based around the concept of communicating operators: language constructs that enable variables local to a single invocation of a parallelised loop to be viewed as if they were arrays spanning the entire loop domain. This mechanism enables different loop invocations (which may or may not be executing on different processors) to exchange information in a manner that extends the successful Communicating Sequential Processes idiom from single messages to collective communication. Communicating operators avoid the additional synchronisation mechanisms, such as atomic variables, required when programming using the Partitioned Global Address Space (PGAS) paradigm. Using a single loop invocation as the fundamental unit of concurrency enables PM to uniformly represent different levels of parallelism from vector operations through shared memory systems to distributed grids. This paper describes an implementation of PM based on a vectorised virtual machine. On a single processor node, concurrent operations are implemented using masked vector operations. Virtual machine instructions operate on vectors of values and may be unmasked, masked using a Boolean field, or masked using an array of active vector cell locations. Conditional structures (such as if-then-else or while statement implementations) calculate and apply masks to the operations they control. A shift in mask representation from Boolean to location-list occurs when active locations become sufficiently sparse. Parallel loops unfold data structures (or vectors of data structures for nested loops) into vectors of values that may additionally be distributed over multiple computational nodes and then split into micro-threads compatible with the size of the local cache. Inter-node communication is accomplished using standard OpenMP and MPI. Performance analyses of the PM vector machine, demonstrating its scaling properties with respect to domain size and the number of processor nodes will be presented for a range of hardware configurations. The PM software and language definition are being made available under unrestrictive MIT and Creative Commons Attribution licenses respectively: www.pm-lang.org.

Support for Debugging Automatically Parallelized Programs

NASA Technical Reports Server (NTRS)

Hood, Robert; Jost, Gabriele; Biegel, Bryan (Technical Monitor)

2001-01-01

This viewgraph presentation provides information on the technical aspects of debugging computer code that has been automatically converted for use in a parallel computing system. Shared memory parallelization and distributed memory parallelization entail separate and distinct challenges for a debugging program. A prototype system has been developed which integrates various tools for the debugging of automatically parallelized programs including the CAPTools Database which provides variable definition information across subroutines as well as array distribution information.

Concurrent computation of attribute filters on shared memory parallel machines.

PubMed

Wilkinson, Michael H F; Gao, Hui; Hesselink, Wim H; Jonker, Jan-Eppo; Meijster, Arnold

2008-10-01

Morphological attribute filters have not previously been parallelized, mainly because they are both global and non-separable. We propose a parallel algorithm that achieves efficient parallelism for a large class of attribute filters, including attribute openings, closings, thinnings and thickenings, based on Salembier's Max-Trees and Min-trees. The image or volume is first partitioned in multiple slices. We then compute the Max-trees of each slice using any sequential Max-Tree algorithm. Subsequently, the Max-trees of the slices can be merged to obtain the Max-tree of the image. A C-implementation yielded good speed-ups on both a 16-processor MIPS 14000 parallel machine, and a dual-core Opteron-based machine. It is shown that the speed-up of the parallel algorithm is a direct measure of the gain with respect to the sequential algorithm used. Furthermore, the concurrent algorithm shows a speed gain of up to 72 percent on a single-core processor, due to reduced cache thrashing.

Performance Modeling and Measurement of Parallelized Code for Distributed Shared Memory Multiprocessors

NASA Technical Reports Server (NTRS)

Waheed, Abdul; Yan, Jerry

1998-01-01

This paper presents a model to evaluate the performance and overhead of parallelizing sequential code using compiler directives for multiprocessing on distributed shared memory (DSM) systems. With increasing popularity of shared address space architectures, it is essential to understand their performance impact on programs that benefit from shared memory multiprocessing. We present a simple model to characterize the performance of programs that are parallelized using compiler directives for shared memory multiprocessing. We parallelized the sequential implementation of NAS benchmarks using native Fortran77 compiler directives for an Origin2000, which is a DSM system based on a cache-coherent Non Uniform Memory Access (ccNUMA) architecture. We report measurement based performance of these parallelized benchmarks from four perspectives: efficacy of parallelization process; scalability; parallelization overhead; and comparison with hand-parallelized and -optimized version of the same benchmarks. Our results indicate that sequential programs can conveniently be parallelized for DSM systems using compiler directives but realizing performance gains as predicted by the performance model depends primarily on minimizing architecture-specific data locality overhead.

ORCA Project: Research on high-performance parallel computer programming environments. Final report, 1 Apr-31 Mar 90

DOE Office of Scientific and Technical Information (OSTI.GOV)

Snyder, L.; Notkin, D.; Adams, L.

1990-03-31

This task relates to research on programming massively parallel computers. Previous work on the Ensamble concept of programming was extended and investigation into nonshared memory models of parallel computation was undertaken. Previous work on the Ensamble concept defined a set of programming abstractions and was used to organize the programming task into three distinct levels; Composition of machine instruction, composition of processes, and composition of phases. It was applied to shared memory models of computations. During the present research period, these concepts were extended to nonshared memory models. During the present research period, one Ph D. thesis was completed, onemore » book chapter, and six conference proceedings were published.« less

Memory access in shared virtual memory

DOE Office of Scientific and Technical Information (OSTI.GOV)

Berrendorf, R.

1992-01-01

Shared virtual memory (SVM) is a virtual memory layer with a single address space on top of a distributed real memory on parallel computers. We examine the behavior and performance of SVM running a parallel program with medium-grained, loop-level parallelism on top of it. A simulator for the underlying parallel architecture can be used to examine the behavior of SVM more deeply. The influence of several parameters, such as the number of processors, page size, cold or warm start, and restricted page replication, is studied.

Memory access in shared virtual memory

DOE Office of Scientific and Technical Information (OSTI.GOV)

Berrendorf, R.

1992-09-01

Shared virtual memory (SVM) is a virtual memory layer with a single address space on top of a distributed real memory on parallel computers. We examine the behavior and performance of SVM running a parallel program with medium-grained, loop-level parallelism on top of it. A simulator for the underlying parallel architecture can be used to examine the behavior of SVM more deeply. The influence of several parameters, such as the number of processors, page size, cold or warm start, and restricted page replication, is studied.

Implementation of a parallel unstructured Euler solver on shared and distributed memory architectures

NASA Technical Reports Server (NTRS)

Mavriplis, D. J.; Das, Raja; Saltz, Joel; Vermeland, R. E.

1992-01-01

An efficient three dimensional unstructured Euler solver is parallelized on a Cray Y-MP C90 shared memory computer and on an Intel Touchstone Delta distributed memory computer. This paper relates the experiences gained and describes the software tools and hardware used in this study. Performance comparisons between two differing architectures are made.

Implementation of a Fully-Balanced Periodic Tridiagonal Solver on a Parallel Distributed Memory Architecture

DTIC Science & Technology

1994-05-01

PARALLEL DISTRIBUTED MEMORY ARCHITECTURE LTJh T. M. Eidson 0 - 8 l 9 5 " G. Erlebacher _ _ _. _ DTIe QUALITY INSPECTED a Contract NAS I - 19480 May 1994...DISTRIBUTED MEMORY ARCHITECTURE T.M. Eidson * High Technology Corporation Hampton, VA 23665 G. Erlebachert Institute for Computer Applications in Science and...developed and evaluated. Simple model calculations as well as timing results are pres.nted to evaluate the various strategies. The particular

Trinary Associative Memory Would Recognize Machine Parts

NASA Technical Reports Server (NTRS)

Liu, Hua-Kuang; Awwal, Abdul Ahad S.; Karim, Mohammad A.

1991-01-01

Trinary associative memory combines merits and overcomes major deficiencies of unipolar and bipolar logics by combining them in three-valued logic that reverts to unipolar or bipolar binary selectively, as needed to perform specific tasks. Advantage of associative memory: one obtains access to all parts of it simultaneously on basis of content, rather than address, of data. Consequently, used to exploit fully parallelism and speed of optical computing.

Numerical methods on some structured matrix algebra problems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jessup, E.R.

1996-06-01

This proposal concerned the design, analysis, and implementation of serial and parallel algorithms for certain structured matrix algebra problems. It emphasized large order problems and so focused on methods that can be implemented efficiently on distributed-memory MIMD multiprocessors. Such machines supply the computing power and extensive memory demanded by the large order problems. We proposed to examine three classes of matrix algebra problems: the symmetric and nonsymmetric eigenvalue problems (especially the tridiagonal cases) and the solution of linear systems with specially structured coefficient matrices. As all of these are of practical interest, a major goal of this work was tomore » translate our research in linear algebra into useful tools for use by the computational scientists interested in these and related applications. Thus, in addition to software specific to the linear algebra problems, we proposed to produce a programming paradigm and library to aid in the design and implementation of programs for distributed-memory MIMD computers. We now report on our progress on each of the problems and on the programming tools.« less

Global Load Balancing with Parallel Mesh Adaption on Distributed-Memory Systems

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Oliker, Leonid; Sohn, Andrew

1996-01-01

Dynamic mesh adaptation on unstructured grids is a powerful tool for efficiently computing unsteady problems to resolve solution features of interest. Unfortunately, this causes load inbalances among processors on a parallel machine. This paper described the parallel implementation of a tetrahedral mesh adaption scheme and a new global load balancing method. A heuristic remapping algorithm is presented that assigns partitions to processors such that the redistribution coast is minimized. Results indicate that the parallel performance of the mesh adaption code depends on the nature of the adaption region and show a 35.5X speedup on 64 processors of an SP2 when 35 percent of the mesh is randomly adapted. For large scale scientific computations, our load balancing strategy gives an almost sixfold reduction in solver execution times over non-balanced loads. Furthermore, our heuristic remappier yields processor assignments that are less than 3 percent of the optimal solutions, but requires only 1 percent of the computational time.

The design and implementation of a parallel unstructured Euler solver using software primitives

NASA Technical Reports Server (NTRS)

Das, R.; Mavriplis, D. J.; Saltz, J.; Gupta, S.; Ponnusamy, R.

1992-01-01

This paper is concerned with the implementation of a three-dimensional unstructured grid Euler-solver on massively parallel distributed-memory computer architectures. The goal is to minimize solution time by achieving high computational rates with a numerically efficient algorithm. An unstructured multigrid algorithm with an edge-based data structure has been adopted, and a number of optimizations have been devised and implemented in order to accelerate the parallel communication rates. The implementation is carried out by creating a set of software tools, which provide an interface between the parallelization issues and the sequential code, while providing a basis for future automatic run-time compilation support. Large practical unstructured grid problems are solved on the Intel iPSC/860 hypercube and Intel Touchstone Delta machine. The quantitative effect of the various optimizations are demonstrated, and we show that the combined effect of these optimizations leads to roughly a factor of three performance improvement. The overall solution efficiency is compared with that obtained on the CRAY-YMP vector supercomputer.

Parallelization of an Object-Oriented Unstructured Aeroacoustics Solver

NASA Technical Reports Server (NTRS)

Baggag, Abdelkader; Atkins, Harold; Oezturan, Can; Keyes, David

1999-01-01

A computational aeroacoustics code based on the discontinuous Galerkin method is ported to several parallel platforms using MPI. The discontinuous Galerkin method is a compact high-order method that retains its accuracy and robustness on non-smooth unstructured meshes. In its semi-discrete form, the discontinuous Galerkin method can be combined with explicit time marching methods making it well suited to time accurate computations. The compact nature of the discontinuous Galerkin method also makes it well suited for distributed memory parallel platforms. The original serial code was written using an object-oriented approach and was previously optimized for cache-based machines. The port to parallel platforms was achieved simply by treating partition boundaries as a type of boundary condition. Code modifications were minimal because boundary conditions were abstractions in the original program. Scalability results are presented for the SCI Origin, IBM SP2, and clusters of SGI and Sun workstations. Slightly superlinear speedup is achieved on a fixed-size problem on the Origin, due to cache effects.

Unstructured Adaptive Meshes: Bad for Your Memory?

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Feng, Hui-Yu; VanderWijngaart, Rob

2003-01-01

This viewgraph presentation explores the need for a NASA Advanced Supercomputing (NAS) parallel benchmark for problems with irregular dynamical memory access. This benchmark is important and necessary because: 1) Problems with localized error source benefit from adaptive nonuniform meshes; 2) Certain machines perform poorly on such problems; 3) Parallel implementation may provide further performance improvement but is difficult. Some examples of problems which use irregular dynamical memory access include: 1) Heat transfer problem; 2) Heat source term; 3) Spectral element method; 4) Base functions; 5) Elemental discrete equations; 6) Global discrete equations. Nonconforming Mesh and Mortar Element Method are covered in greater detail in this presentation.

Parallel algorithms for boundary value problems

NASA Technical Reports Server (NTRS)

Lin, Avi

1990-01-01

A general approach to solve boundary value problems numerically in a parallel environment is discussed. The basic algorithm consists of two steps: the local step where all the P available processors work in parallel, and the global step where one processor solves a tridiagonal linear system of the order P. The main advantages of this approach are two fold. First, this suggested approach is very flexible, especially in the local step and thus the algorithm can be used with any number of processors and with any of the SIMD or MIMD machines. Secondly, the communication complexity is very small and thus can be used as easily with shared memory machines. Several examples for using this strategy are discussed.

Work stealing for GPU-accelerated parallel programs in a global address space framework: WORK STEALING ON GPU-ACCELERATED SYSTEMS

DOE Office of Scientific and Technical Information (OSTI.GOV)

Arafat, Humayun; Dinan, James; Krishnamoorthy, Sriram

Task parallelism is an attractive approach to automatically load balance the computation in a parallel system and adapt to dynamism exhibited by parallel systems. Exploiting task parallelism through work stealing has been extensively studied in shared and distributed-memory contexts. In this paper, we study the design of a system that uses work stealing for dynamic load balancing of task-parallel programs executed on hybrid distributed-memory CPU-graphics processing unit (GPU) systems in a global-address space framework. We take into account the unique nature of the accelerator model employed by GPUs, the significant performance difference between GPU and CPU execution as a functionmore » of problem size, and the distinct CPU and GPU memory domains. We consider various alternatives in designing a distributed work stealing algorithm for CPU-GPU systems, while taking into account the impact of task distribution and data movement overheads. These strategies are evaluated using microbenchmarks that capture various execution configurations as well as the state-of-the-art CCSD(T) application module from the computational chemistry domain.« less

Work stealing for GPU-accelerated parallel programs in a global address space framework

DOE Office of Scientific and Technical Information (OSTI.GOV)

Arafat, Humayun; Dinan, James; Krishnamoorthy, Sriram

Task parallelism is an attractive approach to automatically load balance the computation in a parallel system and adapt to dynamism exhibited by parallel systems. Exploiting task parallelism through work stealing has been extensively studied in shared and distributed-memory contexts. In this paper, we study the design of a system that uses work stealing for dynamic load balancing of task-parallel programs executed on hybrid distributed-memory CPU-graphics processing unit (GPU) systems in a global-address space framework. We take into account the unique nature of the accelerator model employed by GPUs, the significant performance difference between GPU and CPU execution as a functionmore » of problem size, and the distinct CPU and GPU memory domains. We consider various alternatives in designing a distributed work stealing algorithm for CPU-GPU systems, while taking into account the impact of task distribution and data movement overheads. These strategies are evaluated using microbenchmarks that capture various execution configurations as well as the state-of-the-art CCSD(T) application module from the computational chemistry domain« less

Implementation of Parallel Dynamic Simulation on Shared-Memory vs. Distributed-Memory Environments

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jin, Shuangshuang; Chen, Yousu; Wu, Di

2015-12-09

Power system dynamic simulation computes the system response to a sequence of large disturbance, such as sudden changes in generation or load, or a network short circuit followed by protective branch switching operation. It consists of a large set of differential and algebraic equations, which is computational intensive and challenging to solve using single-processor based dynamic simulation solution. High-performance computing (HPC) based parallel computing is a very promising technology to speed up the computation and facilitate the simulation process. This paper presents two different parallel implementations of power grid dynamic simulation using Open Multi-processing (OpenMP) on shared-memory platform, and Messagemore » Passing Interface (MPI) on distributed-memory clusters, respectively. The difference of the parallel simulation algorithms and architectures of the two HPC technologies are illustrated, and their performances for running parallel dynamic simulation are compared and demonstrated.« less

Multi-variants synthesis of Petri nets for FPGA devices

NASA Astrophysics Data System (ADS)

Bukowiec, Arkadiusz; Doligalski, Michał

2015-09-01

There is presented new method of synthesis of application specific logic controllers for FPGA devices. The specification of control algorithm is made with use of control interpreted Petri net (PT type). It allows specifying parallel processes in easy way. The Petri net is decomposed into state-machine type subnets. In this case, each subnet represents one parallel process. For this purpose there are applied algorithms of coloring of Petri nets. There are presented two approaches of such decomposition: with doublers of macroplaces or with one global wait place. Next, subnets are implemented into two-level logic circuit of the controller. The levels of logic circuit are obtained as a result of its architectural decomposition. The first level combinational circuit is responsible for generation of next places and second level decoder is responsible for generation output symbols. There are worked out two variants of such circuits: with one shared operational memory or with many flexible distributed memories as a decoder. Variants of Petri net decomposition and structures of logic circuits can be combined together without any restrictions. It leads to existence of four variants of multi-variants synthesis.

Compile-time estimation of communication costs in multicomputers

NASA Technical Reports Server (NTRS)

Gupta, Manish; Banerjee, Prithviraj

1991-01-01

An important problem facing numerous research projects on parallelizing compilers for distributed memory machines is that of automatically determining a suitable data partitioning scheme for a program. Any strategy for automatic data partitioning needs a mechanism for estimating the performance of a program under a given partitioning scheme, the most crucial part of which involves determining the communication costs incurred by the program. A methodology is described for estimating the communication costs at compile-time as functions of the numbers of processors over which various arrays are distributed. A strategy is described along with its theoretical basis, for making program transformations that expose opportunities for combining of messages, leading to considerable savings in the communication costs. For certain loops with regular dependences, the compiler can detect the possibility of pipelining, and thus estimate communication costs more accurately than it could otherwise. These results are of great significance to any parallelization system supporting numeric applications on multicomputers. In particular, they lay down a framework for effective synthesis of communication on multicomputers from sequential program references.

Directions in parallel programming: HPF, shared virtual memory and object parallelism in pC++

NASA Technical Reports Server (NTRS)

Bodin, Francois; Priol, Thierry; Mehrotra, Piyush; Gannon, Dennis

1994-01-01

Fortran and C++ are the dominant programming languages used in scientific computation. Consequently, extensions to these languages are the most popular for programming massively parallel computers. We discuss two such approaches to parallel Fortran and one approach to C++. The High Performance Fortran Forum has designed HPF with the intent of supporting data parallelism on Fortran 90 applications. HPF works by asking the user to help the compiler distribute and align the data structures with the distributed memory modules in the system. Fortran-S takes a different approach in which the data distribution is managed by the operating system and the user provides annotations to indicate parallel control regions. In the case of C++, we look at pC++ which is based on a concurrent aggregate parallel model.

Low latency messages on distributed memory multiprocessors

NASA Technical Reports Server (NTRS)

Rosing, Matthew; Saltz, Joel

1993-01-01

Many of the issues in developing an efficient interface for communication on distributed memory machines are described and a portable interface is proposed. Although the hardware component of message latency is less than one microsecond on many distributed memory machines, the software latency associated with sending and receiving typed messages is on the order of 50 microseconds. The reason for this imbalance is that the software interface does not match the hardware. By changing the interface to match the hardware more closely, applications with fine grained communication can be put on these machines. Based on several tests that were run on the iPSC/860, an interface that will better match current distributed memory machines is proposed. The model used in the proposed interface consists of a computation processor and a communication processor on each node. Communication between these processors and other nodes in the system is done through a buffered network. Information that is transmitted is either data or procedures to be executed on the remote processor. The dual processor system is better suited for efficiently handling asynchronous communications compared to a single processor system. The ability to send data or procedure is very flexible for minimizing message latency, based on the type of communication being performed. The test performed and the proposed interface are described.

Families of Graph Algorithms: SSSP Case Study

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kanewala Appuhamilage, Thejaka Amila Jay; Zalewski, Marcin J.; Lumsdaine, Andrew

2017-08-28

Single-Source Shortest Paths (SSSP) is a well-studied graph problem. Examples of SSSP algorithms include the original Dijkstra’s algorithm and the parallel Δ-stepping and KLA-SSSP algorithms. In this paper, we use a novel Abstract Graph Machine (AGM) model to show that all these algorithms share a common logic and differ from one another by the order in which they perform work. We use the AGM model to thoroughly analyze the family of algorithms that arises from the common logic. We start with the basic algorithm without any ordering (Chaotic), and then we derive the existing and new algorithms by methodically exploringmore » semantic and spatial ordering of work. Our experimental results show that new derived algorithms show better performance than the existing distributed memory parallel algorithms, especially at higher scales.« less

Scalability of Parallel Spatial Direct Numerical Simulations on Intel Hypercube and IBM SP1 and SP2

NASA Technical Reports Server (NTRS)

Joslin, Ronald D.; Hanebutte, Ulf R.; Zubair, Mohammad

1995-01-01

The implementation and performance of a parallel spatial direct numerical simulation (PSDNS) approach on the Intel iPSC/860 hypercube and IBM SP1 and SP2 parallel computers is documented. Spatially evolving disturbances associated with the laminar-to-turbulent transition in boundary-layer flows are computed with the PSDNS code. The feasibility of using the PSDNS to perform transition studies on these computers is examined. The results indicate that PSDNS approach can effectively be parallelized on a distributed-memory parallel machine by remapping the distributed data structure during the course of the calculation. Scalability information is provided to estimate computational costs to match the actual costs relative to changes in the number of grid points. By increasing the number of processors, slower than linear speedups are achieved with optimized (machine-dependent library) routines. This slower than linear speedup results because the computational cost is dominated by FFT routine, which yields less than ideal speedups. By using appropriate compile options and optimized library routines on the SP1, the serial code achieves 52-56 M ops on a single node of the SP1 (45 percent of theoretical peak performance). The actual performance of the PSDNS code on the SP1 is evaluated with a "real world" simulation that consists of 1.7 million grid points. One time step of this simulation is calculated on eight nodes of the SP1 in the same time as required by a Cray Y/MP supercomputer. For the same simulation, 32-nodes of the SP1 and SP2 are required to reach the performance of a Cray C-90. A 32 node SP1 (SP2) configuration is 2.9 (4.6) times faster than a Cray Y/MP for this simulation, while the hypercube is roughly 2 times slower than the Y/MP for this application. KEY WORDS: Spatial direct numerical simulations; incompressible viscous flows; spectral methods; finite differences; parallel computing.

Parallelization of NAS Benchmarks for Shared Memory Multiprocessors

NASA Technical Reports Server (NTRS)

Waheed, Abdul; Yan, Jerry C.; Saini, Subhash (Technical Monitor)

1998-01-01

This paper presents our experiences of parallelizing the sequential implementation of NAS benchmarks using compiler directives on SGI Origin2000 distributed shared memory (DSM) system. Porting existing applications to new high performance parallel and distributed computing platforms is a challenging task. Ideally, a user develops a sequential version of the application, leaving the task of porting to new generations of high performance computing systems to parallelization tools and compilers. Due to the simplicity of programming shared-memory multiprocessors, compiler developers have provided various facilities to allow the users to exploit parallelism. Native compilers on SGI Origin2000 support multiprocessing directives to allow users to exploit loop-level parallelism in their programs. Additionally, supporting tools can accomplish this process automatically and present the results of parallelization to the users. We experimented with these compiler directives and supporting tools by parallelizing sequential implementation of NAS benchmarks. Results reported in this paper indicate that with minimal effort, the performance gain is comparable with the hand-parallelized, carefully optimized, message-passing implementations of the same benchmarks.

Sparse distributed memory overview

NASA Technical Reports Server (NTRS)

Raugh, Mike

1990-01-01

The Sparse Distributed Memory (SDM) project is investigating the theory and applications of massively parallel computing architecture, called sparse distributed memory, that will support the storage and retrieval of sensory and motor patterns characteristic of autonomous systems. The immediate objectives of the project are centered in studies of the memory itself and in the use of the memory to solve problems in speech, vision, and robotics. Investigation of methods for encoding sensory data is an important part of the research. Examples of NASA missions that may benefit from this work are Space Station, planetary rovers, and solar exploration. Sparse distributed memory offers promising technology for systems that must learn through experience and be capable of adapting to new circumstances, and for operating any large complex system requiring automatic monitoring and control. Sparse distributed memory is a massively parallel architecture motivated by efforts to understand how the human brain works. Sparse distributed memory is an associative memory, able to retrieve information from cues that only partially match patterns stored in the memory. It is able to store long temporal sequences derived from the behavior of a complex system, such as progressive records of the system's sensory data and correlated records of the system's motor controls.

Performance prediction: A case study using a multi-ring KSR-1 machine

NASA Technical Reports Server (NTRS)

Sun, Xian-He; Zhu, Jianping

1995-01-01

While computers with tens of thousands of processors have successfully delivered high performance power for solving some of the so-called 'grand-challenge' applications, the notion of scalability is becoming an important metric in the evaluation of parallel machine architectures and algorithms. In this study, the prediction of scalability and its application are carefully investigated. A simple formula is presented to show the relation between scalability, single processor computing power, and degradation of parallelism. A case study is conducted on a multi-ring KSR1 shared virtual memory machine. Experimental and theoretical results show that the influence of topology variation of an architecture is predictable. Therefore, the performance of an algorithm on a sophisticated, heirarchical architecture can be predicted and the best algorithm-machine combination can be selected for a given application.

Towards a Better Distributed Framework for Learning Big Data

DTIC Science & Technology

2017-06-14

UNLIMITED: PB Public Release 13. SUPPLEMENTARY NOTES 14. ABSTRACT This work aimed at solving issues in distributed machine learning. The PI’s team proposed...communication load. Finally, the team proposed the parallel least-squares policy iteration (parallel LSPI) to parallelize a reinforcement policy learning. 15

Dynamic overset grid communication on distributed memory parallel processors

NASA Technical Reports Server (NTRS)

Barszcz, Eric; Weeratunga, Sisira K.; Meakin, Robert L.

1993-01-01

A parallel distributed memory implementation of intergrid communication for dynamic overset grids is presented. Included are discussions of various options considered during development. Results are presented comparing an Intel iPSC/860 to a single processor Cray Y-MP. Results for grids in relative motion show the iPSC/860 implementation to be faster than the Cray implementation.

Proceedings of the second SISAL users` conference

DOE Office of Scientific and Technical Information (OSTI.GOV)

Feo, J T; Frerking, C; Miller, P J

1992-12-01

This report contains papers on the following topics: A sisal code for computing the fourier transform on S{sub N}; five ways to fill your knapsack; simulating material dislocation motion in sisal; candis as an interface for sisal; parallelisation and performance of the burg algorithm on a shared-memory multiprocessor; use of genetic algorithm in sisal to solve the file design problem; implementing FFT`s in sisal; programming and evaluating the performance of signal processing applications in the sisal programming environment; sisal and Von Neumann-based languages: translation and intercommunication; an IF2 code generator for ADAM architecture; program partitioning for NUMA multiprocessor computer systems;more » mapping functional parallelism on distributed memory machines; implicit array copying: prevention is better than cure ; mathematical syntax for sisal; an approach for optimizing recursive functions; implementing arrays in sisal 2.0; Fol: an object oriented extension to the sisal language; twine: a portable, extensible sisal execution kernel; and investigating the memory performance of the optimizing sisal compiler.« less

Hypercluster - Parallel processing for computational mechanics

NASA Technical Reports Server (NTRS)

Blech, Richard A.

1988-01-01

An account is given of the development status, performance capabilities and implications for further development of NASA-Lewis' testbed 'hypercluster' parallel computer network, in which multiple processors communicate through a shared memory. Processors have local as well as shared memory; the hypercluster is expanded in the same manner as the hypercube, with processor clusters replacing the normal single processor node. The NASA-Lewis machine has three nodes with a vector personality and one node with a scalar personality. Each of the vector nodes uses four board-level vector processors, while the scalar node uses four general-purpose microcomputer boards.

Efficiently modeling neural networks on massively parallel computers

NASA Technical Reports Server (NTRS)

Farber, Robert M.

1993-01-01

Neural networks are a very useful tool for analyzing and modeling complex real world systems. Applying neural network simulations to real world problems generally involves large amounts of data and massive amounts of computation. To efficiently handle the computational requirements of large problems, we have implemented at Los Alamos a highly efficient neural network compiler for serial computers, vector computers, vector parallel computers, and fine grain SIMD computers such as the CM-2 connection machine. This paper describes the mapping used by the compiler to implement feed-forward backpropagation neural networks for a SIMD (Single Instruction Multiple Data) architecture parallel computer. Thinking Machines Corporation has benchmarked our code at 1.3 billion interconnects per second (approximately 3 gigaflops) on a 64,000 processor CM-2 connection machine (Singer 1990). This mapping is applicable to other SIMD computers and can be implemented on MIMD computers such as the CM-5 connection machine. Our mapping has virtually no communications overhead with the exception of the communications required for a global summation across the processors (which has a sub-linear runtime growth on the order of O(log(number of processors)). We can efficiently model very large neural networks which have many neurons and interconnects and our mapping can extend to arbitrarily large networks (within memory limitations) by merging the memory space of separate processors with fast adjacent processor interprocessor communications. This paper will consider the simulation of only feed forward neural network although this method is extendable to recurrent networks.

Parallel definition of tear film maps on distributed-memory clusters for the support of dry eye diagnosis.

PubMed

González-Domínguez, Jorge; Remeseiro, Beatriz; Martín, María J

2017-02-01

The analysis of the interference patterns on the tear film lipid layer is a useful clinical test to diagnose dry eye syndrome. This task can be automated with a high degree of accuracy by means of the use of tear film maps. However, the time required by the existing applications to generate them prevents a wider acceptance of this method by medical experts. Multithreading has been previously successfully employed by the authors to accelerate the tear film map definition on multicore single-node machines. In this work, we propose a hybrid message-passing and multithreading parallel approach that further accelerates the generation of tear film maps by exploiting the computational capabilities of distributed-memory systems such as multicore clusters and supercomputers. The algorithm for drawing tear film maps is parallelized using Message Passing Interface (MPI) for inter-node communications and the multithreading support available in the C++11 standard for intra-node parallelization. The original algorithm is modified to reduce the communications and increase the scalability. The hybrid method has been tested on 32 nodes of an Intel cluster (with two 12-core Haswell 2680v3 processors per node) using 50 representative images. Results show that maximum runtime is reduced from almost two minutes using the previous only-multithreaded approach to less than ten seconds using the hybrid method. The hybrid MPI/multithreaded implementation can be used by medical experts to obtain tear film maps in only a few seconds, which will significantly accelerate and facilitate the diagnosis of the dry eye syndrome. Copyright © 2016 Elsevier Ireland Ltd. All rights reserved.

Parallel machine architecture for production rule systems

DOEpatents

Allen, Jr., John D.; Butler, Philip L.

1989-01-01

A parallel processing system for production rule programs utilizes a host processor for storing production rule right hand sides (RHS) and a plurality of rule processors for storing left hand sides (LHS). The rule processors operate in parallel in the recognize phase of the system recognize -Act Cycle to match their respective LHS's against a stored list of working memory elements (WME) in order to find a self consistent set of WME's. The list of WME is dynamically varied during the Act phase of the system in which the host executes or fires rule RHS's for those rules for which a self-consistent set has been found by the rule processors. The host transmits instructions for creating or deleting working memory elements as dictated by the rule firings until the rule processors are unable to find any further self-consistent working memory element sets at which time the production rule system is halted.

Proceedings of the 14th International Conference on the Numerical Simulation of Plasmas

NASA Astrophysics Data System (ADS)

Partial Contents are as follows: Numerical Simulations of the Vlasov-Maxwell Equations by Coupled Particle-Finite Element Methods on Unstructured Meshes; Electromagnetic PIC Simulations Using Finite Elements on Unstructured Grids; Modelling Travelling Wave Output Structures with the Particle-in-Cell Code CONDOR; SST--A Single-Slice Particle Simulation Code; Graphical Display and Animation of Data Produced by Electromagnetic, Particle-in-Cell Codes; A Post-Processor for the PEST Code; Gray Scale Rendering of Beam Profile Data; A 2D Electromagnetic PIC Code for Distributed Memory Parallel Computers; 3-D Electromagnetic PIC Simulation on the NRL Connection Machine; Plasma PIC Simulations on MIMD Computers; Vlasov-Maxwell Algorithm for Electromagnetic Plasma Simulation on Distributed Architectures; MHD Boundary Layer Calculation Using the Vortex Method; and Eulerian Codes for Plasma Simulations.

First Applications of the New Parallel Krylov Solver for MODFLOW on a National and Global Scale

NASA Astrophysics Data System (ADS)

Verkaik, J.; Hughes, J. D.; Sutanudjaja, E.; van Walsum, P.

2016-12-01

Integrated high-resolution hydrologic models are increasingly being used for evaluating water management measures at field scale. Their drawbacks are large memory requirements and long run times. Examples of such models are The Netherlands Hydrological Instrument (NHI) model and the PCRaster Global Water Balance (PCR-GLOBWB) model. Typical simulation periods are 30-100 years with daily timesteps. The NHI model predicts water demands in periods of drought, supporting operational and long-term water-supply decisions. The NHI is a state-of-the-art coupling of several models: a 7-layer MODFLOW groundwater model ( 6.5M 250m cells), a MetaSWAP model for the unsaturated zone (Richards emulator of 0.5M cells), and a surface water model (MOZART-DM). The PCR-GLOBWB model provides a grid-based representation of global terrestrial hydrology and this work uses the version that includes a 2-layer MODFLOW groundwater model ( 4.5M 10km cells). The Parallel Krylov Solver (PKS) speeds up computation by both distributed memory parallelization (Message Passing Interface) and shared memory parallelization (Open Multi-Processing). PKS includes conjugate gradient, bi-conjugate gradient stabilized, and generalized minimal residual linear accelerators that use an overlapping additive Schwarz domain decomposition preconditioner. PKS can be used for both structured and unstructured grids and has been fully integrated in MODFLOW-USG using METIS partitioning and in iMODFLOW using RCB partitioning. iMODFLOW is an accelerated version of MODFLOW-2005 that is implicitly and online coupled to MetaSWAP. Results for benchmarks carried out on the Cartesius Dutch supercomputer (https://userinfo.surfsara.nl/systems/cartesius) for the PCRGLOB-WB model and on a 2x16 core Windows machine for the NHI model show speedups up to 10-20 and 5-10, respectively.

Parallel-vector out-of-core equation solver for computational mechanics

NASA Technical Reports Server (NTRS)

Qin, J.; Agarwal, T. K.; Storaasli, O. O.; Nguyen, D. T.; Baddourah, M. A.

1993-01-01

A parallel/vector out-of-core equation solver is developed for shared-memory computers, such as the Cray Y-MP machine. The input/ output (I/O) time is reduced by using the a synchronous BUFFER IN and BUFFER OUT, which can be executed simultaneously with the CPU instructions. The parallel and vector capability provided by the supercomputers is also exploited to enhance the performance. Numerical applications in large-scale structural analysis are given to demonstrate the efficiency of the present out-of-core solver.

Distributed parallel messaging for multiprocessor systems

DOEpatents

Chen, Dong; Heidelberger, Philip; Salapura, Valentina; Senger, Robert M; Steinmacher-Burrow, Burhard; Sugawara, Yutaka

2013-06-04

A method and apparatus for distributed parallel messaging in a parallel computing system. The apparatus includes, at each node of a multiprocessor network, multiple injection messaging engine units and reception messaging engine units, each implementing a DMA engine and each supporting both multiple packet injection into and multiple reception from a network, in parallel. The reception side of the messaging unit (MU) includes a switch interface enabling writing of data of a packet received from the network to the memory system. The transmission side of the messaging unit, includes switch interface for reading from the memory system when injecting packets into the network.

Accelerating atomistic calculations of quantum energy eigenstates on graphic cards

NASA Astrophysics Data System (ADS)

Rodrigues, Walter; Pecchia, A.; Lopez, M.; Auf der Maur, M.; Di Carlo, A.

2014-10-01

Electronic properties of nanoscale materials require the calculation of eigenvalues and eigenvectors of large matrices. This bottleneck can be overcome by parallel computing techniques or the introduction of faster algorithms. In this paper we report a custom implementation of the Lanczos algorithm with simple restart, optimized for graphical processing units (GPUs). The whole algorithm has been developed using CUDA and runs entirely on the GPU, with a specialized implementation that spares memory and reduces at most machine-to-device data transfers. Furthermore parallel distribution over several GPUs has been attained using the standard message passing interface (MPI). Benchmark calculations performed on a GaN/AlGaN wurtzite quantum dot with up to 600,000 atoms are presented. The empirical tight-binding (ETB) model with an sp3d5s∗+spin-orbit parametrization has been used to build the system Hamiltonian (H).

Gregarious Data Re-structuring in a Many Core Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shrestha, Sunil; Manzano Franco, Joseph B.; Marquez, Andres

this paper, we have developed a new methodology that takes in consideration the access patterns from a single parallel actor (e.g. a thread), as well as, the access patterns of “grouped” parallel actors that share a resource (e.g. a distributed Level 3 cache). We start with a hierarchical tile code for our target machine and apply a series of transformations at the tile level to improve data residence in a given memory hierarchy level. The contribution of this paper includes (a) collaborative data restructuring for group reuse and (b) low overhead transformation technique to improve access pattern and bring closelymore » connected data elements together. Preliminary results in a many core architecture, Tilera TileGX, shows promising improvements over optimized OpenMP code (up to 31% increase in GFLOPS) and over our own previous work on fine grained runtimes (up to 16%) for selected kernels« less

Computer-Aided Parallelizer and Optimizer

NASA Technical Reports Server (NTRS)

Jin, Haoqiang

2011-01-01

The Computer-Aided Parallelizer and Optimizer (CAPO) automates the insertion of compiler directives (see figure) to facilitate parallel processing on Shared Memory Parallel (SMP) machines. While CAPO currently is integrated seamlessly into CAPTools (developed at the University of Greenwich, now marketed as ParaWise), CAPO was independently developed at Ames Research Center as one of the components for the Legacy Code Modernization (LCM) project. The current version takes serial FORTRAN programs, performs interprocedural data dependence analysis, and generates OpenMP directives. Due to the widely supported OpenMP standard, the generated OpenMP codes have the potential to run on a wide range of SMP machines. CAPO relies on accurate interprocedural data dependence information currently provided by CAPTools. Compiler directives are generated through identification of parallel loops in the outermost level, construction of parallel regions around parallel loops and optimization of parallel regions, and insertion of directives with automatic identification of private, reduction, induction, and shared variables. Attempts also have been made to identify potential pipeline parallelism (implemented with point-to-point synchronization). Although directives are generated automatically, user interaction with the tool is still important for producing good parallel codes. A comprehensive graphical user interface is included for users to interact with the parallelization process.

Efficient computation of aerodynamic influence coefficients for aeroelastic analysis on a transputer network

NASA Technical Reports Server (NTRS)

Janetzke, David C.; Murthy, Durbha V.

1991-01-01

Aeroelastic analysis is multi-disciplinary and computationally expensive. Hence, it can greatly benefit from parallel processing. As part of an effort to develop an aeroelastic capability on a distributed memory transputer network, a parallel algorithm for the computation of aerodynamic influence coefficients is implemented on a network of 32 transputers. The aerodynamic influence coefficients are calculated using a 3-D unsteady aerodynamic model and a parallel discretization. Efficiencies up to 85 percent were demonstrated using 32 processors. The effect of subtask ordering, problem size, and network topology are presented. A comparison to results on a shared memory computer indicates that higher speedup is achieved on the distributed memory system.

Memory-Intensive Benchmarks: IRAM vs. Cache-Based Machines

NASA Technical Reports Server (NTRS)

Biswas, Rupak; Gaeke, Brian R.; Husbands, Parry; Li, Xiaoye S.; Oliker, Leonid; Yelick, Katherine A.; Biegel, Bryan (Technical Monitor)

2002-01-01

The increasing gap between processor and memory performance has lead to new architectural models for memory-intensive applications. In this paper, we explore the performance of a set of memory-intensive benchmarks and use them to compare the performance of conventional cache-based microprocessors to a mixed logic and DRAM processor called VIRAM. The benchmarks are based on problem statements, rather than specific implementations, and in each case we explore the fundamental hardware requirements of the problem, as well as alternative algorithms and data structures that can help expose fine-grained parallelism or simplify memory access patterns. The benchmarks are characterized by their memory access patterns, their basic control structures, and the ratio of computation to memory operation.

Parallel Optimization of Polynomials for Large-scale Problems in Stability and Control

NASA Astrophysics Data System (ADS)

Kamyar, Reza

In this thesis, we focus on some of the NP-hard problems in control theory. Thanks to the converse Lyapunov theory, these problems can often be modeled as optimization over polynomials. To avoid the problem of intractability, we establish a trade off between accuracy and complexity. In particular, we develop a sequence of tractable optimization problems --- in the form of Linear Programs (LPs) and/or Semi-Definite Programs (SDPs) --- whose solutions converge to the exact solution of the NP-hard problem. However, the computational and memory complexity of these LPs and SDPs grow exponentially with the progress of the sequence - meaning that improving the accuracy of the solutions requires solving SDPs with tens of thousands of decision variables and constraints. Setting up and solving such problems is a significant challenge. The existing optimization algorithms and software are only designed to use desktop computers or small cluster computers --- machines which do not have sufficient memory for solving such large SDPs. Moreover, the speed-up of these algorithms does not scale beyond dozens of processors. This in fact is the reason we seek parallel algorithms for setting-up and solving large SDPs on large cluster- and/or super-computers. We propose parallel algorithms for stability analysis of two classes of systems: 1) Linear systems with a large number of uncertain parameters; 2) Nonlinear systems defined by polynomial vector fields. First, we develop a distributed parallel algorithm which applies Polya's and/or Handelman's theorems to some variants of parameter-dependent Lyapunov inequalities with parameters defined over the standard simplex. The result is a sequence of SDPs which possess a block-diagonal structure. We then develop a parallel SDP solver which exploits this structure in order to map the computation, memory and communication to a distributed parallel environment. Numerical tests on a supercomputer demonstrate the ability of the algorithm to efficiently utilize hundreds and potentially thousands of processors, and analyze systems with 100+ dimensional state-space. Furthermore, we extend our algorithms to analyze robust stability over more complicated geometries such as hypercubes and arbitrary convex polytopes. Our algorithms can be readily extended to address a wide variety of problems in control such as Hinfinity synthesis for systems with parametric uncertainty and computing control Lyapunov functions.

Support for non-locking parallel reception of packets belonging to a single memory reception FIFO

DOEpatents

Chen, Dong [Yorktown Heights, NY; Heidelberger, Philip [Yorktown Heights, NY; Salapura, Valentina [Yorktown Heights, NY; Senger, Robert M [Yorktown Heights, NY; Steinmacher-Burow, Burkhard [Boeblingen, DE; Sugawara, Yutaka [Yorktown Heights, NY

2011-01-27

A method and apparatus for distributed parallel messaging in a parallel computing system. A plurality of DMA engine units are configured in a multiprocessor system to operate in parallel, one DMA engine unit for transferring a current packet received at a network reception queue to a memory location in a memory FIFO (rmFIFO) region of a memory. A control unit implements logic to determine whether any prior received packet destined for that rmFIFO is still in a process of being stored in the associated memory by another DMA engine unit of the plurality, and prevent the one DMA engine unit from indicating completion of storing the current received packet in the reception memory FIFO (rmFIFO) until all prior received packets destined for that rmFIFO are completely stored by the other DMA engine units. Thus, there is provided non-locking support so that multiple packets destined for a single rmFIFO are transferred and stored in parallel to predetermined locations in a memory.

Advances in Parallelization for Large Scale Oct-Tree Mesh Generation

NASA Technical Reports Server (NTRS)

O'Connell, Matthew; Karman, Steve L.

2015-01-01

Despite great advancements in the parallelization of numerical simulation codes over the last 20 years, it is still common to perform grid generation in serial. Generating large scale grids in serial often requires using special "grid generation" compute machines that can have more than ten times the memory of average machines. While some parallel mesh generation techniques have been proposed, generating very large meshes for LES or aeroacoustic simulations is still a challenging problem. An automated method for the parallel generation of very large scale off-body hierarchical meshes is presented here. This work enables large scale parallel generation of off-body meshes by using a novel combination of parallel grid generation techniques and a hybrid "top down" and "bottom up" oct-tree method. Meshes are generated using hardware commonly found in parallel compute clusters. The capability to generate very large meshes is demonstrated by the generation of off-body meshes surrounding complex aerospace geometries. Results are shown including a one billion cell mesh generated around a Predator Unmanned Aerial Vehicle geometry, which was generated on 64 processors in under 45 minutes.

Automatic Data Traffic Control on DSM Architecture

NASA Technical Reports Server (NTRS)

Frumkin, Michael; Jin, Hao-Qiang; Yan, Jerry; Kwak, Dochan (Technical Monitor)

2000-01-01

We study data traffic on distributed shared memory machines and conclude that data placement and grouping improve performance of scientific codes. We present several methods which user can employ to improve data traffic in his code. We report on implementation of a tool which detects the code fragments causing data congestions and advises user on improvements of data routing in these fragments. The capabilities of the tool include deduction of data alignment and affinity from the source code; detection of the code constructs having abnormally high cache or TLB misses; generation of data placement constructs. We demonstrate the capabilities of the tool on experiments with NAS parallel benchmarks and with a simple computational fluid dynamics application ARC3D.

A distributed version of the NASA Engine Performance Program

NASA Technical Reports Server (NTRS)

Cours, Jeffrey T.; Curlett, Brian P.

1993-01-01

Distributed NEPP, a version of the NASA Engine Performance Program, uses the original NEPP code but executes it in a distributed computer environment. Multiple workstations connected by a network increase the program's speed and, more importantly, the complexity of the cases it can handle in a reasonable time. Distributed NEPP uses the public domain software package, called Parallel Virtual Machine, allowing it to execute on clusters of machines containing many different architectures. It includes the capability to link with other computers, allowing them to process NEPP jobs in parallel. This paper discusses the design issues and granularity considerations that entered into programming Distributed NEPP and presents the results of timing runs.

Parallel Programming Paradigms

DTIC Science & Technology

1987-07-01

Unclassified IS.. DECLASSIFICATIONIOOWNGRADIN G 16. DISTRIBUTION STATEMENT (of this Report) Distribution of this report is unlimited. 17...8416878 and by the Office of Naval Research Contracts No. N00014-86-K-0264 and No. N00014-85- K-0328. 8 ?~~ O . G 1 49 II Parallel Programming Paradigms...processors -. "to fetch from the same memory cell (list head) and thus seems to favor a shared memory - g implementation [37). In this dissertation, we

Multi-objective problem of the modified distributed parallel machine and assembly scheduling problem (MDPMASP) with eligibility constraints

NASA Astrophysics Data System (ADS)

Amallynda, I.; Santosa, B.

2017-11-01

This paper proposes a new generalization of the distributed parallel machine and assembly scheduling problem (DPMASP) with eligibility constraints referred to as the modified distributed parallel machine and assembly scheduling problem (MDPMASP) with eligibility constraints. Within this generalization, we assume that there are a set non-identical factories or production lines, each one with a set unrelated parallel machine with different speeds in processing them disposed to a single assembly machine in series. A set of different products that are manufactured through an assembly program of a set of components (jobs) according to the requested demand. Each product requires several kinds of jobs with different sizes. Beside that we also consider to the multi-objective problem (MOP) of minimizing mean flow time and the number of tardy products simultaneously. This is known to be NP-Hard problem, is important to practice, as the former criterions to reflect the customer's demand and manufacturer's perspective. This is a realistic and complex problem with wide range of possible solutions, we propose four simple heuristics and two metaheuristics to solve it. Various parameters of the proposed metaheuristic algorithms are discussed and calibrated by means of Taguchi technique. All proposed algorithms are tested by Matlab software. Our computational experiments indicate that the proposed problem and fourth proposed algorithms are able to be implemented and can be used to solve moderately-sized instances, and giving efficient solutions, which are close to optimum in most cases.

Tuning collective communication for Partitioned Global Address Space programming models

DOE PAGES

Nishtala, Rajesh; Zheng, Yili; Hargrove, Paul H.; ...

2011-06-12

Partitioned Global Address Space (PGAS) languages offer programmers the convenience of a shared memory programming style combined with locality control necessary to run on large-scale distributed memory systems. Even within a PGAS language programmers often need to perform global communication operations such as broadcasts or reductions, which are best performed as collective operations in which a group of threads work together to perform the operation. In this study we consider the problem of implementing collective communication within PGAS languages and explore some of the design trade-offs in both the interface and implementation. In particular, PGAS collectives have semantic issues thatmore » are different than in send–receive style message passing programs, and different implementation approaches that take advantage of the one-sided communication style in these languages. We present an implementation framework for PGAS collectives as part of the GASNet communication layer, which supports shared memory, distributed memory and hybrids. The framework supports a broad set of algorithms for each collective, over which the implementation may be automatically tuned. In conclusion, we demonstrate the benefit of optimized GASNet collectives using application benchmarks written in UPC, and demonstrate that the GASNet collectives can deliver scalable performance on a variety of state-of-the-art parallel machines including a Cray XT4, an IBM BlueGene/P, and a Sun Constellation system with InfiniBand interconnect.« less

Performance Measurement, Visualization and Modeling of Parallel and Distributed Programs

NASA Technical Reports Server (NTRS)

Yan, Jerry C.; Sarukkai, Sekhar R.; Mehra, Pankaj; Lum, Henry, Jr. (Technical Monitor)

1994-01-01

This paper presents a methodology for debugging the performance of message-passing programs on both tightly coupled and loosely coupled distributed-memory machines. The AIMS (Automated Instrumentation and Monitoring System) toolkit, a suite of software tools for measurement and analysis of performance, is introduced and its application illustrated using several benchmark programs drawn from the field of computational fluid dynamics. AIMS includes (i) Xinstrument, a powerful source-code instrumentor, which supports both Fortran77 and C as well as a number of different message-passing libraries including Intel's NX Thinking Machines' CMMD, and PVM; (ii) Monitor, a library of timestamping and trace -collection routines that run on supercomputers (such as Intel's iPSC/860, Delta, and Paragon and Thinking Machines' CM5) as well as on networks of workstations (including Convex Cluster and SparcStations connected by a LAN); (iii) Visualization Kernel, a trace-animation facility that supports source-code clickback, simultaneous visualization of computation and communication patterns, as well as analysis of data movements; (iv) Statistics Kernel, an advanced profiling facility, that associates a variety of performance data with various syntactic components of a parallel program; (v) Index Kernel, a diagnostic tool that helps pinpoint performance bottlenecks through the use of abstract indices; (vi) Modeling Kernel, a facility for automated modeling of message-passing programs that supports both simulation -based and analytical approaches to performance prediction and scalability analysis; (vii) Intrusion Compensator, a utility for recovering true performance from observed performance by removing the overheads of monitoring and their effects on the communication pattern of the program; and (viii) Compatibility Tools, that convert AIMS-generated traces into formats used by other performance-visualization tools, such as ParaGraph, Pablo, and certain AVS/Explorer modules.

Tensorial Basis Spline Collocation Method for Poisson's Equation

NASA Astrophysics Data System (ADS)

Plagne, Laurent; Berthou, Jean-Yves

2000-01-01

This paper aims to describe the tensorial basis spline collocation method applied to Poisson's equation. In the case of a localized 3D charge distribution in vacuum, this direct method based on a tensorial decomposition of the differential operator is shown to be competitive with both iterative BSCM and FFT-based methods. We emphasize the O(h4) and O(h6) convergence of TBSCM for cubic and quintic splines, respectively. We describe the implementation of this method on a distributed memory parallel machine. Performance measurements on a Cray T3E are reported. Our code exhibits high performance and good scalability: As an example, a 27 Gflops performance is obtained when solving Poisson's equation on a 2563 non-uniform 3D Cartesian mesh by using 128 T3E-750 processors. This represents 215 Mflops per processors.

PIPS-SBB: A Parallel Distributed-Memory Branch-and-Bound Algorithm for Stochastic Mixed-Integer Programs

DOE PAGES

Munguia, Lluis-Miquel; Oxberry, Geoffrey; Rajan, Deepak

2016-05-01

Stochastic mixed-integer programs (SMIPs) deal with optimization under uncertainty at many levels of the decision-making process. When solved as extensive formulation mixed- integer programs, problem instances can exceed available memory on a single workstation. In order to overcome this limitation, we present PIPS-SBB: a distributed-memory parallel stochastic MIP solver that takes advantage of parallelism at multiple levels of the optimization process. We also show promising results on the SIPLIB benchmark by combining methods known for accelerating Branch and Bound (B&B) methods with new ideas that leverage the structure of SMIPs. Finally, we expect the performance of PIPS-SBB to improve furthermore » as more functionality is added in the future.« less

A scalable parallel black oil simulator on distributed memory parallel computers

NASA Astrophysics Data System (ADS)

Wang, Kun; Liu, Hui; Chen, Zhangxin

2015-11-01

This paper presents our work on developing a parallel black oil simulator for distributed memory computers based on our in-house parallel platform. The parallel simulator is designed to overcome the performance issues of common simulators that are implemented for personal computers and workstations. The finite difference method is applied to discretize the black oil model. In addition, some advanced techniques are employed to strengthen the robustness and parallel scalability of the simulator, including an inexact Newton method, matrix decoupling methods, and algebraic multigrid methods. A new multi-stage preconditioner is proposed to accelerate the solution of linear systems from the Newton methods. Numerical experiments show that our simulator is scalable and efficient, and is capable of simulating extremely large-scale black oil problems with tens of millions of grid blocks using thousands of MPI processes on parallel computers.

Parallel computing of a digital hologram and particle searching for microdigital-holographic particle-tracking velocimetry

DOE Office of Scientific and Technical Information (OSTI.GOV)

Satake, Shin-ichi; Kanamori, Hiroyuki; Kunugi, Tomoaki

2007-02-01

We have developed a parallel algorithm for microdigital-holographic particle-tracking velocimetry. The algorithm is used in (1) numerical reconstruction of a particle image computer using a digital hologram, and (2) searching for particles. The numerical reconstruction from the digital hologram makes use of the Fresnel diffraction equation and the FFT (fast Fourier transform),whereas the particle search algorithm looks for local maximum graduation in a reconstruction field represented by a 3D matrix. To achieve high performance computing for both calculations (reconstruction and particle search), two memory partitions are allocated to the 3D matrix. In this matrix, the reconstruction part consists of horizontallymore » placed 2D memory partitions on the x-y plane for the FFT, whereas, the particle search part consists of vertically placed 2D memory partitions set along the z axes.Consequently, the scalability can be obtained for the proportion of processor elements,where the benchmarks are carried out for parallel computation by a SGI Altix machine.« less

Algorithms for Automatic Alignment of Arrays

NASA Technical Reports Server (NTRS)

Chatterjee, Siddhartha; Gilbert, John R.; Oliker, Leonid; Schreiber, Robert; Sheffler, Thomas J.

1996-01-01

Aggregate data objects (such as arrays) are distributed across the processor memories when compiling a data-parallel language for a distributed-memory machine. The mapping determines the amount of communication needed to bring operands of parallel operations into alignment with each other. A common approach is to break the mapping into two stages: an alignment that maps all the objects to an abstract template, followed by a distribution that maps the template to the processors. This paper describes algorithms for solving the various facets of the alignment problem: axis and stride alignment, static and mobile offset alignment, and replication labeling. We show that optimal axis and stride alignment is NP-complete for general program graphs, and give a heuristic method that can explore the space of possible solutions in a number of ways. We show that some of these strategies can give better solutions than a simple greedy approach proposed earlier. We also show how local graph contractions can reduce the size of the problem significantly without changing the best solution. This allows more complex and effective heuristics to be used. We show how to model the static offset alignment problem using linear programming, and we show that loop-dependent mobile offset alignment is sometimes necessary for optimum performance. We describe an algorithm with for determining mobile alignments for objects within do loops. We also identify situations in which replicated alignment is either required by the program itself or can be used to improve performance. We describe an algorithm based on network flow that replicates objects so as to minimize the total amount of broadcast communication in replication.

Impact of the implementation of MPI point-to-point communications on the performance of two general sparse solvers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Amestoy, Patrick R.; Duff, Iain S.; L'Excellent, Jean-Yves

2001-10-10

We examine the mechanics of the send and receive mechanism of MPI and in particular how we can implement message passing in a robust way so that our performance is not significantly affected by changes to the MPI system. This leads us to using the Isend/Irecv protocol which will entail sometimes significant algorithmic changes. We discuss this within the context of two different algorithms for sparse Gaussian elimination that we have parallelized. One is a multifrontal solver called MUMPS, the other is a supernodal solver called SuperLU. Both algorithms are difficult to parallelize on distributed memory machines. Our initial strategiesmore » were based on simple MPI point-to-point communication primitives. With such approaches, the parallel performance of both codes are very sensitive to the MPI implementation, the way MPI internal buffers are used in particular. We then modified our codes to use more sophisticated nonblocking versions of MPI communication. This significantly improved the performance robustness (independent of the MPI buffering mechanism) and scalability, but at the cost of increased code complexity.« less

Dynamically programmable cache

NASA Astrophysics Data System (ADS)

Nakkar, Mouna; Harding, John A.; Schwartz, David A.; Franzon, Paul D.; Conte, Thomas

1998-10-01

Reconfigurable machines have recently been used as co- processors to accelerate the execution of certain algorithms or program subroutines. The problems with the above approach include high reconfiguration time and limited partial reconfiguration. By far the most critical problems are: (1) the small on-chip memory which results in slower execution time, and (2) small FPGA areas that cannot implement large subroutines. Dynamically Programmable Cache (DPC) is a novel architecture for embedded processors which offers solutions to the above problems. To solve memory access problems, DPC processors merge reconfigurable arrays with the data cache at various cache levels to create a multi-level reconfigurable machines. As a result DPC machines have both higher data accessibility and FPGA memory bandwidth. To solve the limited FPGA resource problem, DPC processors implemented multi-context switching (Virtualization) concept. Virtualization allows implementation of large subroutines with fewer FPGA cells. Additionally, DPC processors can parallelize the execution of several operations resulting in faster execution time. In this paper, the speedup improvement for DPC machines are shown to be 5X faster than an Altera FLEX10K FPGA chip and 2X faster than a Sun Ultral SPARC station for two different algorithms (convolution and motion estimation).

Event parallelism: Distributed memory parallel computing for high energy physics experiments

NASA Astrophysics Data System (ADS)

Nash, Thomas

1989-12-01

This paper describes the present and expected future development of distributed memory parallel computers for high energy physics experiments. It covers the use of event parallel microprocessor farms, particularly at Fermilab, including both ACP multiprocessors and farms of MicroVAXES. These systems have proven very cost effective in the past. A case is made for moving to the more open environment of UNIX and RISC processors. The 2nd Generation ACP Multiprocessor System, which is based on powerful RISC system, is described. Given the promise of still more extraordinary increases in processor performance, a new emphasis on point to point, rather than bussed, communication will be required. Developments in this direction are described.

Shared virtual memory and generalized speedup

NASA Technical Reports Server (NTRS)

Sun, Xian-He; Zhu, Jianping

1994-01-01

Generalized speedup is defined as parallel speed over sequential speed. The generalized speedup and its relation with other existing performance metrics, such as traditional speedup, efficiency, scalability, etc., are carefully studied. In terms of the introduced asymptotic speed, it was shown that the difference between the generalized speedup and the traditional speedup lies in the definition of the efficiency of uniprocessor processing, which is a very important issue in shared virtual memory machines. A scientific application was implemented on a KSR-1 parallel computer. Experimental and theoretical results show that the generalized speedup is distinct from the traditional speedup and provides a more reasonable measurement. In the study of different speedups, various causes of superlinear speedup are also presented.

Design of multiple sequence alignment algorithms on parallel, distributed memory supercomputers.

PubMed

Church, Philip C; Goscinski, Andrzej; Holt, Kathryn; Inouye, Michael; Ghoting, Amol; Makarychev, Konstantin; Reumann, Matthias

2011-01-01

The challenge of comparing two or more genomes that have undergone recombination and substantial amounts of segmental loss and gain has recently been addressed for small numbers of genomes. However, datasets of hundreds of genomes are now common and their sizes will only increase in the future. Multiple sequence alignment of hundreds of genomes remains an intractable problem due to quadratic increases in compute time and memory footprint. To date, most alignment algorithms are designed for commodity clusters without parallelism. Hence, we propose the design of a multiple sequence alignment algorithm on massively parallel, distributed memory supercomputers to enable research into comparative genomics on large data sets. Following the methodology of the sequential progressiveMauve algorithm, we design data structures including sequences and sorted k-mer lists on the IBM Blue Gene/P supercomputer (BG/P). Preliminary results show that we can reduce the memory footprint so that we can potentially align over 250 bacterial genomes on a single BG/P compute node. We verify our results on a dataset of E.coli, Shigella and S.pneumoniae genomes. Our implementation returns results matching those of the original algorithm but in 1/2 the time and with 1/4 the memory footprint for scaffold building. In this study, we have laid the basis for multiple sequence alignment of large-scale datasets on a massively parallel, distributed memory supercomputer, thus enabling comparison of hundreds instead of a few genome sequences within reasonable time.

3-D parallel program for numerical calculation of gas dynamics problems with heat conductivity on distributed memory computational systems (CS)

DOE Office of Scientific and Technical Information (OSTI.GOV)

Sofronov, I.D.; Voronin, B.L.; Butnev, O.I.

1997-12-31

The aim of the work performed is to develop a 3D parallel program for numerical calculation of gas dynamics problem with heat conductivity on distributed memory computational systems (CS), satisfying the condition of numerical result independence from the number of processors involved. Two basically different approaches to the structure of massive parallel computations have been developed. The first approach uses the 3D data matrix decomposition reconstructed at temporal cycle and is a development of parallelization algorithms for multiprocessor CS with shareable memory. The second approach is based on using a 3D data matrix decomposition not reconstructed during a temporal cycle.more » The program was developed on 8-processor CS MP-3 made in VNIIEF and was adapted to a massive parallel CS Meiko-2 in LLNL by joint efforts of VNIIEF and LLNL staffs. A large number of numerical experiments has been carried out with different number of processors up to 256 and the efficiency of parallelization has been evaluated in dependence on processor number and their parameters.« less

Parallel Computing for Probabilistic Response Analysis of High Temperature Composites

NASA Technical Reports Server (NTRS)

Sues, R. H.; Lua, Y. J.; Smith, M. D.

1994-01-01

The objective of this Phase I research was to establish the required software and hardware strategies to achieve large scale parallelism in solving PCM problems. To meet this objective, several investigations were conducted. First, we identified the multiple levels of parallelism in PCM and the computational strategies to exploit these parallelisms. Next, several software and hardware efficiency investigations were conducted. These involved the use of three different parallel programming paradigms and solution of two example problems on both a shared-memory multiprocessor and a distributed-memory network of workstations.

Fast and Accurate Support Vector Machines on Large Scale Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Vishnu, Abhinav; Narasimhan, Jayenthi; Holder, Larry

Support Vector Machines (SVM) is a supervised Machine Learning and Data Mining (MLDM) algorithm, which has become ubiquitous largely due to its high accuracy and obliviousness to dimensionality. The objective of SVM is to find an optimal boundary --- also known as hyperplane --- which separates the samples (examples in a dataset) of different classes by a maximum margin. Usually, very few samples contribute to the definition of the boundary. However, existing parallel algorithms use the entire dataset for finding the boundary, which is sub-optimal for performance reasons. In this paper, we propose a novel distributed memory algorithm to eliminatemore » the samples which do not contribute to the boundary definition in SVM. We propose several heuristics, which range from early (aggressive) to late (conservative) elimination of the samples, such that the overall time for generating the boundary is reduced considerably. In a few cases, a sample may be eliminated (shrunk) pre-emptively --- potentially resulting in an incorrect boundary. We propose a scalable approach to synchronize the necessary data structures such that the proposed algorithm maintains its accuracy. We consider the necessary trade-offs of single/multiple synchronization using in-depth time-space complexity analysis. We implement the proposed algorithm using MPI and compare it with libsvm--- de facto sequential SVM software --- which we enhance with OpenMP for multi-core/many-core parallelism. Our proposed approach shows excellent efficiency using up to 4096 processes on several large datasets such as UCI HIGGS Boson dataset and Offending URL dataset.« less

Porting Gravitational Wave Signal Extraction to Parallel Virtual Machine (PVM)

NASA Technical Reports Server (NTRS)

Thirumalainambi, Rajkumar; Thompson, David E.; Redmon, Jeffery

2009-01-01

Laser Interferometer Space Antenna (LISA) is a planned NASA-ESA mission to be launched around 2012. The Gravitational Wave detection is fundamentally the determination of frequency, source parameters, and waveform amplitude derived in a specific order from the interferometric time-series of the rotating LISA spacecrafts. The LISA Science Team has developed a Mock LISA Data Challenge intended to promote the testing of complicated nested search algorithms to detect the 100-1 millihertz frequency signals at amplitudes of 10E-21. However, it has become clear that, sequential search of the parameters is very time consuming and ultra-sensitive; hence, a new strategy has been developed. Parallelization of existing sequential search algorithms of Gravitational Wave signal identification consists of decomposing sequential search loops, beginning with outermost loops and working inward. In this process, the main challenge is to detect interdependencies among loops and partitioning the loops so as to preserve concurrency. Existing parallel programs are based upon either shared memory or distributed memory paradigms. In PVM, master and node programs are used to execute parallelization and process spawning. The PVM can handle process management and process addressing schemes using a virtual machine configuration. The task scheduling and the messaging and signaling can be implemented efficiently for the LISA Gravitational Wave search process using a master and 6 nodes. This approach is accomplished using a server that is available at NASA Ames Research Center, and has been dedicated to the LISA Data Challenge Competition. Historically, gravitational wave and source identification parameters have taken around 7 days in this dedicated single thread Linux based server. Using PVM approach, the parameter extraction problem can be reduced to within a day. The low frequency computation and a proxy signal-to-noise ratio are calculated in separate nodes that are controlled by the master using message and vector of data passing. The message passing among nodes follows a pattern of synchronous and asynchronous send-and-receive protocols. The communication model and the message buffers are allocated dynamically to address rapid search of gravitational wave source information in the Mock LISA data sets.

Communication overhead on the Intel Paragon, IBM SP2 and Meiko CS-2

NASA Technical Reports Server (NTRS)

Bokhari, Shahid H.

1995-01-01

Interprocessor communication overhead is a crucial measure of the power of parallel computing systems-its impact can severely limit the performance of parallel programs. This report presents measurements of communication overhead on three contemporary commercial multicomputer systems: the Intel Paragon, the IBM SP2 and the Meiko CS-2. In each case the time to communicate between processors is presented as a function of message length. The time for global synchronization and memory access is discussed. The performance of these machines in emulating hypercubes and executing random pairwise exchanges is also investigated. It is shown that the interprocessor communication time depends heavily on the specific communication pattern required. These observations contradict the commonly held belief that communication overhead on contemporary machines is independent of the placement of tasks on processors. The information presented in this report permits the evaluation of the efficiency of parallel algorithm implementations against standard baselines.

Real-time implementations of image segmentation algorithms on shared memory multicore architecture: a survey (Conference Presentation)

NASA Astrophysics Data System (ADS)

Akil, Mohamed

2017-05-01

The real-time processing is getting more and more important in many image processing applications. Image segmentation is one of the most fundamental tasks image analysis. As a consequence, many different approaches for image segmentation have been proposed. The watershed transform is a well-known image segmentation tool. The watershed transform is a very data intensive task. To achieve acceleration and obtain real-time processing of watershed algorithms, parallel architectures and programming models for multicore computing have been developed. This paper focuses on the survey of the approaches for parallel implementation of sequential watershed algorithms on multicore general purpose CPUs: homogeneous multicore processor with shared memory. To achieve an efficient parallel implementation, it's necessary to explore different strategies (parallelization/distribution/distributed scheduling) combined with different acceleration and optimization techniques to enhance parallelism. In this paper, we give a comparison of various parallelization of sequential watershed algorithms on shared memory multicore architecture. We analyze the performance measurements of each parallel implementation and the impact of the different sources of overhead on the performance of the parallel implementations. In this comparison study, we also discuss the advantages and disadvantages of the parallel programming models. Thus, we compare the OpenMP (an application programming interface for multi-Processing) with Ptheads (POSIX Threads) to illustrate the impact of each parallel programming model on the performance of the parallel implementations.

Dust Dynamics in Protoplanetary Disks: Parallel Computing with PVM

NASA Astrophysics Data System (ADS)

de La Fuente Marcos, Carlos; Barge, Pierre; de La Fuente Marcos, Raúl

2002-03-01

We describe a parallel version of our high-order-accuracy particle-mesh code for the simulation of collisionless protoplanetary disks. We use this code to carry out a massively parallel, two-dimensional, time-dependent, numerical simulation, which includes dust particles, to study the potential role of large-scale, gaseous vortices in protoplanetary disks. This noncollisional problem is easy to parallelize on message-passing multicomputer architectures. We performed the simulations on a cache-coherent nonuniform memory access Origin 2000 machine, using both the parallel virtual machine (PVM) and message-passing interface (MPI) message-passing libraries. Our performance analysis suggests that, for our problem, PVM is about 25% faster than MPI. Using PVM and MPI made it possible to reduce CPU time and increase code performance. This allows for simulations with a large number of particles (N ~ 105-106) in reasonable CPU times. The performances of our implementation of the pa! rallel code on an Origin 2000 supercomputer are presented and discussed. They exhibit very good speedup behavior and low load unbalancing. Our results confirm that giant gaseous vortices can play a dominant role in giant planet formation.

A new parallel-vector finite element analysis software on distributed-memory computers

NASA Technical Reports Server (NTRS)

Qin, Jiangning; Nguyen, Duc T.

1993-01-01

A new parallel-vector finite element analysis software package MPFEA (Massively Parallel-vector Finite Element Analysis) is developed for large-scale structural analysis on massively parallel computers with distributed-memory. MPFEA is designed for parallel generation and assembly of the global finite element stiffness matrices as well as parallel solution of the simultaneous linear equations, since these are often the major time-consuming parts of a finite element analysis. Block-skyline storage scheme along with vector-unrolling techniques are used to enhance the vector performance. Communications among processors are carried out concurrently with arithmetic operations to reduce the total execution time. Numerical results on the Intel iPSC/860 computers (such as the Intel Gamma with 128 processors and the Intel Touchstone Delta with 512 processors) are presented, including an aircraft structure and some very large truss structures, to demonstrate the efficiency and accuracy of MPFEA.

Computation of Coupled Thermal-Fluid Problems in Distributed Memory Environment

NASA Technical Reports Server (NTRS)

Wei, H.; Shang, H. M.; Chen, Y. S.

2001-01-01

The thermal-fluid coupling problems are very important to aerospace and engineering applications. Instead of analyzing heat transfer and fluid flow separately, this study merged two well-accepted engineering solution methods, SINDA for thermal analysis and FDNS for fluid flow simulation, into a unified multi-disciplinary thermal fluid prediction method. A fully conservative patched grid interface algorithm for arbitrary two-dimensional and three-dimensional geometry has been developed. The state-of-the-art parallel computing concept was used to couple SINDA and FDNS for the communication of boundary conditions through PVM (Parallel Virtual Machine) libraries. Therefore, the thermal analysis performed by SINDA and the fluid flow calculated by FDNS are fully coupled to obtain steady state or transient solutions. The natural convection between two thick-walled eccentric tubes was calculated and the predicted results match the experiment data perfectly. A 3-D rocket engine model and a real 3-D SSME geometry were used to test the current model, and the reasonable temperature field was obtained.

AFL-1: A programming Language for Massively Concurrent Computers.

DTIC Science & Technology

1986-11-01

Bibliography Ackley, D.H., Hinton, G.E., Sejnowski, T.J., "A Learning Algorithm for boltzmann Machines", Cognitive Science, 1985, 9, 147-169. Agre...P.E., "Routines", Memo 828, MIT AI Laboratory, Many 1985. Ballard, D.H., Hayes, P.J., "Parallel Logical Inference", Conference of the Cognitive Science...34Experiments on Semantic Memory and Language Com- 125 prehension", in L.W. Greg (Ed.), Cognition in Learning and Memory, New York, Wiley, 1972._ Collins

Kmerind: A Flexible Parallel Library for K-mer Indexing of Biological Sequences on Distributed Memory Systems.

PubMed

Pan, Tony; Flick, Patrick; Jain, Chirag; Liu, Yongchao; Aluru, Srinivas

2017-10-09

Counting and indexing fixed length substrings, or k-mers, in biological sequences is a key step in many bioinformatics tasks including genome alignment and mapping, genome assembly, and error correction. While advances in next generation sequencing technologies have dramatically reduced the cost and improved latency and throughput, few bioinformatics tools can efficiently process the datasets at the current generation rate of 1.8 terabases every 3 days. We present Kmerind, a high performance parallel k-mer indexing library for distributed memory environments. The Kmerind library provides a set of simple and consistent APIs with sequential semantics and parallel implementations that are designed to be flexible and extensible. Kmerind's k-mer counter performs similarly or better than the best existing k-mer counting tools even on shared memory systems. In a distributed memory environment, Kmerind counts k-mers in a 120 GB sequence read dataset in less than 13 seconds on 1024 Xeon CPU cores, and fully indexes their positions in approximately 17 seconds. Querying for 1% of the k-mers in these indices can be completed in 0.23 seconds and 28 seconds, respectively. Kmerind is the first k-mer indexing library for distributed memory environments, and the first extensible library for general k-mer indexing and counting. Kmerind is available at https://github.com/ParBLiSS/kmerind.

Clock Agreement Among Parallel Supercomputer Nodes

DOE Data Explorer

Jones, Terry R.; Koenig, Gregory A.

2014-04-30

This dataset presents measurements that quantify the clock synchronization time-agreement characteristics among several high performance computers including the current world's most powerful machine for open science, the U.S. Department of Energy's Titan machine sited at Oak Ridge National Laboratory. These ultra-fast machines derive much of their computational capability from extreme node counts (over 18000 nodes in the case of the Titan machine). Time-agreement is commonly utilized by parallel programming applications and tools, distributed programming application and tools, and system software. Our time-agreement measurements detail the degree of time variance between nodes and how that variance changes over time. The dataset includes empirical measurements and the accompanying spreadsheets.

Accelerating adaptive inverse distance weighting interpolation algorithm on a graphics processing unit

PubMed Central

Xu, Liangliang; Xu, Nengxiong

2017-01-01

This paper focuses on designing and implementing parallel adaptive inverse distance weighting (AIDW) interpolation algorithms by using the graphics processing unit (GPU). The AIDW is an improved version of the standard IDW, which can adaptively determine the power parameter according to the data points’ spatial distribution pattern and achieve more accurate predictions than those predicted by IDW. In this paper, we first present two versions of the GPU-accelerated AIDW, i.e. the naive version without profiting from the shared memory and the tiled version taking advantage of the shared memory. We also implement the naive version and the tiled version using two data layouts, structure of arrays and array of aligned structures, on both single and double precision. We then evaluate the performance of parallel AIDW by comparing it with its corresponding serial algorithm on three different machines equipped with the GPUs GT730M, M5000 and K40c. The experimental results indicate that: (i) there is no significant difference in the computational efficiency when different data layouts are employed; (ii) the tiled version is always slightly faster than the naive version; and (iii) on single precision the achieved speed-up can be up to 763 (on the GPU M5000), while on double precision the obtained highest speed-up is 197 (on the GPU K40c). To benefit the community, all source code and testing data related to the presented parallel AIDW algorithm are publicly available. PMID:28989754

Accelerating adaptive inverse distance weighting interpolation algorithm on a graphics processing unit.

PubMed

Mei, Gang; Xu, Liangliang; Xu, Nengxiong

2017-09-01

This paper focuses on designing and implementing parallel adaptive inverse distance weighting (AIDW) interpolation algorithms by using the graphics processing unit (GPU). The AIDW is an improved version of the standard IDW, which can adaptively determine the power parameter according to the data points' spatial distribution pattern and achieve more accurate predictions than those predicted by IDW. In this paper, we first present two versions of the GPU-accelerated AIDW, i.e. the naive version without profiting from the shared memory and the tiled version taking advantage of the shared memory. We also implement the naive version and the tiled version using two data layouts, structure of arrays and array of aligned structures, on both single and double precision. We then evaluate the performance of parallel AIDW by comparing it with its corresponding serial algorithm on three different machines equipped with the GPUs GT730M, M5000 and K40c. The experimental results indicate that: (i) there is no significant difference in the computational efficiency when different data layouts are employed; (ii) the tiled version is always slightly faster than the naive version; and (iii) on single precision the achieved speed-up can be up to 763 (on the GPU M5000), while on double precision the obtained highest speed-up is 197 (on the GPU K40c). To benefit the community, all source code and testing data related to the presented parallel AIDW algorithm are publicly available.

A robot arm simulation with a shared memory multiprocessor machine

NASA Technical Reports Server (NTRS)

Kim, Sung-Soo; Chuang, Li-Ping

1989-01-01

A parallel processing scheme for a single chain robot arm is presented for high speed computation on a shared memory multiprocessor. A recursive formulation that is derived from a virtual work form of the d'Alembert equations of motion is utilized for robot arm dynamics. A joint drive system that consists of a motor rotor and gears is included in the arm dynamics model, in order to take into account gyroscopic effects due to the spinning of the rotor. The fine grain parallelism of mechanical and control subsystem models is exploited, based on independent computation associated with bodies, joint drive systems, and controllers. Efficiency and effectiveness of the parallel scheme are demonstrated through simulations of a telerobotic manipulator arm. Two different mechanical subsystem models, i.e., with and without gyroscopic effects, are compared, to show the trade-off between efficiency and accuracy.

Introduction to Parallel Computing

DTIC Science & Technology

1992-05-01

Instruction Stream, Multiple Data Stream Machines .................... 19 2.4 Networks of M achines...independent memory units and connecting them to the processors by an interconnection network . Many different interconnection schemes have been considered, and...connected to the same processor at the same time. Crossbar switching networks are still too expensive to be practical for connecting large numbers of

Parallel Distributed Processing at 25: further explorations in the microstructure of cognition.

PubMed

Rogers, Timothy T; McClelland, James L

2014-08-01

This paper introduces a special issue of Cognitive Science initiated on the 25th anniversary of the publication of Parallel Distributed Processing (PDP), a two-volume work that introduced the use of neural network models as vehicles for understanding cognition. The collection surveys the core commitments of the PDP framework, the key issues the framework has addressed, and the debates the framework has spawned, and presents viewpoints on the current status of these issues. The articles focus on both historical roots and contemporary developments in learning, optimality theory, perception, memory, language, conceptual knowledge, cognitive control, and consciousness. Here we consider the approach more generally, reviewing the original motivations, the resulting framework, and the central tenets of the underlying theory. We then evaluate the impact of PDP both on the field at large and within specific subdomains of cognitive science and consider the current role of PDP models within the broader landscape of contemporary theoretical frameworks in cognitive science. Looking to the future, we consider the implications for cognitive science of the recent success of machine learning systems called "deep networks"-systems that build on key ideas presented in the PDP volumes. Copyright © 2014 Cognitive Science Society, Inc.

A mixed parallel strategy for the solution of coupled multi-scale problems at finite strains

NASA Astrophysics Data System (ADS)

Lopes, I. A. Rodrigues; Pires, F. M. Andrade; Reis, F. J. P.

2018-02-01

A mixed parallel strategy for the solution of homogenization-based multi-scale constitutive problems undergoing finite strains is proposed. The approach aims to reduce the computational time and memory requirements of non-linear coupled simulations that use finite element discretization at both scales (FE^2). In the first level of the algorithm, a non-conforming domain decomposition technique, based on the FETI method combined with a mortar discretization at the interface of macroscopic subdomains, is employed. A master-slave scheme, which distributes tasks by macroscopic element and adopts dynamic scheduling, is then used for each macroscopic subdomain composing the second level of the algorithm. This strategy allows the parallelization of FE^2 simulations in computers with either shared memory or distributed memory architectures. The proposed strategy preserves the quadratic rates of asymptotic convergence that characterize the Newton-Raphson scheme. Several examples are presented to demonstrate the robustness and efficiency of the proposed parallel strategy.

Full Parallel Implementation of an All-Electron Four-Component Dirac-Kohn-Sham Program.

PubMed

Rampino, Sergio; Belpassi, Leonardo; Tarantelli, Francesco; Storchi, Loriano

2014-09-09

A full distributed-memory implementation of the Dirac-Kohn-Sham (DKS) module of the program BERTHA (Belpassi et al., Phys. Chem. Chem. Phys. 2011, 13, 12368-12394) is presented, where the self-consistent field (SCF) procedure is replicated on all the parallel processes, each process working on subsets of the global matrices. The key feature of the implementation is an efficient procedure for switching between two matrix distribution schemes, one (integral-driven) optimal for the parallel computation of the matrix elements and another (block-cyclic) optimal for the parallel linear algebra operations. This approach, making both CPU-time and memory scalable with the number of processors used, virtually overcomes at once both time and memory barriers associated with DKS calculations. Performance, portability, and numerical stability of the code are illustrated on the basis of test calculations on three gold clusters of increasing size, an organometallic compound, and a perovskite model. The calculations are performed on a Beowulf and a BlueGene/Q system.

A manual for PARTI runtime primitives

NASA Technical Reports Server (NTRS)

Berryman, Harry; Saltz, Joel

1990-01-01

Primitives are presented that are designed to help users efficiently program irregular problems (e.g., unstructured mesh sweeps, sparse matrix codes, adaptive mesh partial differential equations solvers) on distributed memory machines. These primitives are also designed for use in compilers for distributed memory multiprocessors. Communications patterns are captured at runtime, and the appropriate send and receive messages are automatically generated.

Notes on implementation of sparsely distributed memory

NASA Technical Reports Server (NTRS)

Keeler, J. D.; Denning, P. J.

1986-01-01

The Sparsely Distributed Memory (SDM) developed by Kanerva is an unconventional memory design with very interesting and desirable properties. The memory works in a manner that is closely related to modern theories of human memory. The SDM model is discussed in terms of its implementation in hardware. Two appendices discuss the unconventional approaches of the SDM: Appendix A treats a resistive circuit for fast, parallel address decoding; and Appendix B treats a systolic array for high throughput read and write operations.

Parallelization of Lower-Upper Symmetric Gauss-Seidel Method for Chemically Reacting Flow

NASA Technical Reports Server (NTRS)

Yoon, Seokkwan; Jost, Gabriele; Chang, Sherry

2005-01-01

Development of technologies for exploration of the solar system has revived an interest in computational simulation of chemically reacting flows since planetary probe vehicles exhibit non-equilibrium phenomena during the atmospheric entry of a planet or a moon as well as the reentry to the Earth. Stability in combustion is essential for new propulsion systems. Numerical solution of real-gas flows often increases computational work by an order-of-magnitude compared to perfect gas flow partly because of the increased complexity of equations to solve. Recently, as part of Project Columbia, NASA has integrated a cluster of interconnected SGI Altix systems to provide a ten-fold increase in current supercomputing capacity that includes an SGI Origin system. Both the new and existing machines are based on cache coherent non-uniform memory access architecture. Lower-Upper Symmetric Gauss-Seidel (LU-SGS) relaxation method has been implemented into both perfect and real gas flow codes including Real-Gas Aerodynamic Simulator (RGAS). However, the vectorized RGAS code runs inefficiently on cache-based shared-memory machines such as SGI system. Parallelization of a Gauss-Seidel method is nontrivial due to its sequential nature. The LU-SGS method has been vectorized on an oblique plane in INS3D-LU code that has been one of the base codes for NAS Parallel benchmarks. The oblique plane has been called a hyperplane by computer scientists. It is straightforward to parallelize a Gauss-Seidel method by partitioning the hyperplanes once they are formed. Another way of parallelization is to schedule processors like a pipeline using software. Both hyperplane and pipeline methods have been implemented using openMP directives. The present paper reports the performance of the parallelized RGAS code on SGI Origin and Altix systems.

Supporting shared data structures on distributed memory architectures

NASA Technical Reports Server (NTRS)

Koelbel, Charles; Mehrotra, Piyush; Vanrosendale, John

1990-01-01

Programming nonshared memory systems is more difficult than programming shared memory systems, since there is no support for shared data structures. Current programming languages for distributed memory architectures force the user to decompose all data structures into separate pieces, with each piece owned by one of the processors in the machine, and with all communication explicitly specified by low-level message-passing primitives. A new programming environment is presented for distributed memory architectures, providing a global name space and allowing direct access to remote parts of data values. The analysis and program transformations required to implement this environment are described, and the efficiency of the resulting code on the NCUBE/7 and IPSC/2 hypercubes are described.

Vienna FORTRAN: A FORTRAN language extension for distributed memory multiprocessors

NASA Technical Reports Server (NTRS)

Chapman, Barbara; Mehrotra, Piyush; Zima, Hans

1991-01-01

Exploiting the performance potential of distributed memory machines requires a careful distribution of data across the processors. Vienna FORTRAN is a language extension of FORTRAN which provides the user with a wide range of facilities for such mapping of data structures. However, programs in Vienna FORTRAN are written using global data references. Thus, the user has the advantage of a shared memory programming paradigm while explicitly controlling the placement of data. The basic features of Vienna FORTRAN are presented along with a set of examples illustrating the use of these features.

Buffered coscheduling for parallel programming and enhanced fault tolerance

DOEpatents

Petrini, Fabrizio [Los Alamos, NM; Feng, Wu-chun [Los Alamos, NM

2006-01-31

A computer implemented method schedules processor jobs on a network of parallel machine processors or distributed system processors. Control information communications generated by each process performed by each processor during a defined time interval is accumulated in buffers, where adjacent time intervals are separated by strobe intervals for a global exchange of control information. A global exchange of the control information communications at the end of each defined time interval is performed during an intervening strobe interval so that each processor is informed by all of the other processors of the number of incoming jobs to be received by each processor in a subsequent time interval. The buffered coscheduling method of this invention also enhances the fault tolerance of a network of parallel machine processors or distributed system processors

Concurrent Collections (CnC): A new approach to parallel programming

DOE Office of Scientific and Technical Information (OSTI.GOV)

Knobe, Kathleen

2010-05-07

A common approach in designing parallel languages is to provide some high level handles to manipulate the use of the parallel platform. This exposes some aspects of the target platform, for example, shared vs. distributed memory. It may expose some but not all types of parallelism, for example, data parallelism but not task parallelism. This approach must find a balance between the desire to provide a simple view for the domain expert and provide sufficient power for tuning. This is hard for any given architecture and harder if the language is to apply to a range of architectures. Either simplicitymore » or power is lost. Instead of viewing the language design problem as one of providing the programmer with high level handles, we view the problem as one of designing an interface. On one side of this interface is the programmer (domain expert) who knows the application but needs no knowledge of any aspects of the platform. On the other side of the interface is the performance expert (programmer or program) who demands maximal flexibility for optimizing the mapping to a wide range of target platforms (parallel / serial, shared / distributed, homogeneous / heterogeneous, etc.) but needs no knowledge of the domain. Concurrent Collections (CnC) is based on this separation of concerns. The talk will present CnC and its benefits. About the speaker. Kathleen Knobe has focused throughout her career on parallelism especially compiler technology, runtime system design and language design. She worked at Compass (aka Massachusetts Computer Associates) from 1980 to 1991 designing compilers for a wide range of parallel platforms for Thinking Machines, MasPar, Alliant, Numerix, and several government projects. In 1991 she decided to finish her education. After graduating from MIT in 1997, she joined Digital Equipment’s Cambridge Research Lab (CRL). She stayed through the DEC/Compaq/HP mergers and when CRL was acquired and absorbed by Intel. She currently works in the Software and Services Group / Technology Pathfinding and Innovation.« less

Concurrent Collections (CnC): A new approach to parallel programming

ScienceCinema

Knobe, Kathleen

2018-04-16

A common approach in designing parallel languages is to provide some high level handles to manipulate the use of the parallel platform. This exposes some aspects of the target platform, for example, shared vs. distributed memory. It may expose some but not all types of parallelism, for example, data parallelism but not task parallelism. This approach must find a balance between the desire to provide a simple view for the domain expert and provide sufficient power for tuning. This is hard for any given architecture and harder if the language is to apply to a range of architectures. Either simplicity or power is lost. Instead of viewing the language design problem as one of providing the programmer with high level handles, we view the problem as one of designing an interface. On one side of this interface is the programmer (domain expert) who knows the application but needs no knowledge of any aspects of the platform. On the other side of the interface is the performance expert (programmer or program) who demands maximal flexibility for optimizing the mapping to a wide range of target platforms (parallel / serial, shared / distributed, homogeneous / heterogeneous, etc.) but needs no knowledge of the domain. Concurrent Collections (CnC) is based on this separation of concerns. The talk will present CnC and its benefits. About the speaker. Kathleen Knobe has focused throughout her career on parallelism especially compiler technology, runtime system design and language design. She worked at Compass (aka Massachusetts Computer Associates) from 1980 to 1991 designing compilers for a wide range of parallel platforms for Thinking Machines, MasPar, Alliant, Numerix, and several government projects. In 1991 she decided to finish her education. After graduating from MIT in 1997, she joined Digital Equipment’s Cambridge Research Lab (CRL). She stayed through the DEC/Compaq/HP mergers and when CRL was acquired and absorbed by Intel. She currently works in the Software and Services Group / Technology Pathfinding and Innovation.

A Parameter Communication Optimization Strategy for Distributed Machine Learning in Sensors.

PubMed

Zhang, Jilin; Tu, Hangdi; Ren, Yongjian; Wan, Jian; Zhou, Li; Li, Mingwei; Wang, Jue; Yu, Lifeng; Zhao, Chang; Zhang, Lei

2017-09-21

In order to utilize the distributed characteristic of sensors, distributed machine learning has become the mainstream approach, but the different computing capability of sensors and network delays greatly influence the accuracy and the convergence rate of the machine learning model. Our paper describes a reasonable parameter communication optimization strategy to balance the training overhead and the communication overhead. We extend the fault tolerance of iterative-convergent machine learning algorithms and propose the Dynamic Finite Fault Tolerance (DFFT). Based on the DFFT, we implement a parameter communication optimization strategy for distributed machine learning, named Dynamic Synchronous Parallel Strategy (DSP), which uses the performance monitoring model to dynamically adjust the parameter synchronization strategy between worker nodes and the Parameter Server (PS). This strategy makes full use of the computing power of each sensor, ensures the accuracy of the machine learning model, and avoids the situation that the model training is disturbed by any tasks unrelated to the sensors.

Limits in decision making arise from limits in memory retrieval.

PubMed

Giguère, Gyslain; Love, Bradley C

2013-05-07

Some decisions, such as predicting the winner of a baseball game, are challenging in part because outcomes are probabilistic. When making such decisions, one view is that humans stochastically and selectively retrieve a small set of relevant memories that provides evidence for competing options. We show that optimal performance at test is impossible when retrieving information in this fashion, no matter how extensive training is, because limited retrieval introduces noise into the decision process that cannot be overcome. One implication is that people should be more accurate in predicting future events when trained on idealized rather than on the actual distributions of items. In other words, we predict the best way to convey information to people is to present it in a distorted, idealized form. Idealization of training distributions is predicted to reduce the harmful noise induced by immutable bottlenecks in people's memory retrieval processes. In contrast, machine learning systems that selectively weight (i.e., retrieve) all training examples at test should not benefit from idealization. These conjectures are strongly supported by several studies and supporting analyses. Unlike machine systems, people's test performance on a target distribution is higher when they are trained on an idealized version of the distribution rather than on the actual target distribution. Optimal machine classifiers modified to selectively and stochastically sample from memory match the pattern of human performance. These results suggest firm limits on human rationality and have broad implications for how to train humans tasked with important classification decisions, such as radiologists, baggage screeners, intelligence analysts, and gamblers.

Limits in decision making arise from limits in memory retrieval

PubMed Central

Giguère, Gyslain; Love, Bradley C.

2013-01-01

Some decisions, such as predicting the winner of a baseball game, are challenging in part because outcomes are probabilistic. When making such decisions, one view is that humans stochastically and selectively retrieve a small set of relevant memories that provides evidence for competing options. We show that optimal performance at test is impossible when retrieving information in this fashion, no matter how extensive training is, because limited retrieval introduces noise into the decision process that cannot be overcome. One implication is that people should be more accurate in predicting future events when trained on idealized rather than on the actual distributions of items. In other words, we predict the best way to convey information to people is to present it in a distorted, idealized form. Idealization of training distributions is predicted to reduce the harmful noise induced by immutable bottlenecks in people’s memory retrieval processes. In contrast, machine learning systems that selectively weight (i.e., retrieve) all training examples at test should not benefit from idealization. These conjectures are strongly supported by several studies and supporting analyses. Unlike machine systems, people’s test performance on a target distribution is higher when they are trained on an idealized version of the distribution rather than on the actual target distribution. Optimal machine classifiers modified to selectively and stochastically sample from memory match the pattern of human performance. These results suggest firm limits on human rationality and have broad implications for how to train humans tasked with important classification decisions, such as radiologists, baggage screeners, intelligence analysts, and gamblers. PMID:23610402

UPC++ Programmer’s Guide (v1.0 2017.9)

DOE Office of Scientific and Technical Information (OSTI.GOV)

Bachan, J.; Baden, S.; Bonachea, D.

UPC++ is a C++11 library that provides Asynchronous Partitioned Global Address Space (APGAS) programming. It is designed for writing parallel programs that run efficiently and scale well on distributed-memory parallel computers. The APGAS model is single program, multiple-data (SPMD), with each separate thread of execution (referred to as a rank, a term borrowed from MPI) having access to local memory as it would in C++. However, APGAS also provides access to a global address space, which is allocated in shared segments that are distributed over the ranks. UPC++ provides numerous methods for accessing and using global memory. In UPC++, allmore » operations that access remote memory are explicit, which encourages programmers to be aware of the cost of communication and data movement. Moreover, all remote-memory access operations are by default asynchronous, to enable programmers to write code that scales well even on hundreds of thousands of cores.« less

UPC++ Programmer’s Guide, v1.0-2018.3.0

DOE Office of Scientific and Technical Information (OSTI.GOV)

Bachan, J.; Baden, S.; Bonachea, Dan

UPC++ is a C++11 library that provides Partitioned Global Address Space (PGAS) programming. It is designed for writing parallel programs that run efficiently and scale well on distributed-memory parallel computers. The PGAS model is single program, multiple-data (SPMD), with each separate thread of execution (referred to as a rank, a term borrowed from MPI) having access to local memory as it would in C++. However, PGAS also provides access to a global address space, which is allocated in shared segments that are distributed over the ranks. UPC++ provides numerous methods for accessing and using global memory. In UPC++, all operationsmore » that access remote memory are explicit, which encourages programmers to be aware of the cost of communication and data movement. Moreover, all remote-memory access operations are by default asynchronous, to enable programmers to write code that scales well even on hundreds of thousands of cores.« less

TESS

DOE Office of Scientific and Technical Information (OSTI.GOV)

Dmitriy Morozov, Tom Peterka

2014-07-29

Computing a Voronoi or Delaunay tessellation from a set of points is a core part of the analysis of many simulated and measured datasets. As the scale of simulations and observations surpasses billions of particles, a distributed-memory scalable parallel algorithm is the only feasible approach. The primary contribution of this software is a distributed-memory parallel Delaunay and Voronoi tessellation algorithm based on existing serial computational geometry libraries that automatically determines which neighbor points need to be exchanged among the subdomains of a spatial decomposition. Other contributions include the addition of periodic and wall boundary conditions.

A self-defining hierarchical data system

NASA Technical Reports Server (NTRS)

Bailey, J.

1992-01-01

The Self-Defining Data System (SDS) is a system which allows the creation of self-defining hierarchical data structures in a form which allows the data to be moved between different machine architectures. Because the structures are self-defining they can be used for communication between independent modules in a distributed system. Unlike disk-based hierarchical data systems such as Starlink's HDS, SDS works entirely in memory and is very fast. Data structures are created and manipulated as internal dynamic structures in memory managed by SDS itself. A structure may then be exported into a caller supplied memory buffer in a defined external format. This structure can be written as a file or sent as a message to another machine. It remains static in structure until it is reimported into SDS. SDS is written in portable C and has been run on a number of different machine architectures. Structures are portable between machines with SDS looking after conversion of byte order, floating point format, and alignment. A Fortran callable version is also available for some machines.

Programming distributed memory architectures using Kali

NASA Technical Reports Server (NTRS)

Mehrotra, Piyush; Vanrosendale, John

1990-01-01

Programming nonshared memory systems is more difficult than programming shared memory systems, in part because of the relatively low level of current programming environments for such machines. A new programming environment is presented, Kali, which provides a global name space and allows direct access to remote data values. In order to retain efficiency, Kali provides a system on annotations, allowing the user to control those aspects of the program critical to performance, such as data distribution and load balancing. The primitives and constructs provided by the language is described, and some of the issues raised in translating a Kali program for execution on distributed memory systems are also discussed.

High-Performance Computation of Distributed-Memory Parallel 3D Voronoi and Delaunay Tessellation

DOE Office of Scientific and Technical Information (OSTI.GOV)

Peterka, Tom; Morozov, Dmitriy; Phillips, Carolyn

2014-11-14

Computing a Voronoi or Delaunay tessellation from a set of points is a core part of the analysis of many simulated and measured datasets: N-body simulations, molecular dynamics codes, and LIDAR point clouds are just a few examples. Such computational geometry methods are common in data analysis and visualization; but as the scale of simulations and observations surpasses billions of particles, the existing serial and shared-memory algorithms no longer suffice. A distributed-memory scalable parallel algorithm is the only feasible approach. The primary contribution of this paper is a new parallel Delaunay and Voronoi tessellation algorithm that automatically determines which neighbormore » points need to be exchanged among the subdomains of a spatial decomposition. Other contributions include periodic and wall boundary conditions, comparison of our method using two popular serial libraries, and application to numerous science datasets.« less

Hybrid Parallelism for Volume Rendering on Large-, Multi-, and Many-Core Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Howison, Mark; Bethel, E. Wes; Childs, Hank

2012-01-01

With the computing industry trending towards multi- and many-core processors, we study how a standard visualization algorithm, ray-casting volume rendering, can benefit from a hybrid parallelism approach. Hybrid parallelism provides the best of both worlds: using distributed-memory parallelism across a large numbers of nodes increases available FLOPs and memory, while exploiting shared-memory parallelism among the cores within each node ensures that each node performs its portion of the larger calculation as efficiently as possible. We demonstrate results from weak and strong scaling studies, at levels of concurrency ranging up to 216,000, and with datasets as large as 12.2 trillion cells.more » The greatest benefit from hybrid parallelism lies in the communication portion of the algorithm, the dominant cost at higher levels of concurrency. We show that reducing the number of participants with a hybrid approach significantly improves performance.« less

Distributed-Memory Fast Maximal Independent Set

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kanewala Appuhamilage, Thejaka Amila J.; Zalewski, Marcin J.; Lumsdaine, Andrew

The Maximal Independent Set (MIS) graph problem arises in many applications such as computer vision, information theory, molecular biology, and process scheduling. The growing scale of MIS problems suggests the use of distributed-memory hardware as a cost-effective approach to providing necessary compute and memory resources. Luby proposed four randomized algorithms to solve the MIS problem. All those algorithms are designed focusing on shared-memory machines and are analyzed using the PRAM model. These algorithms do not have direct efficient distributed-memory implementations. In this paper, we extend two of Luby’s seminal MIS algorithms, “Luby(A)” and “Luby(B),” to distributed-memory execution, and we evaluatemore » their performance. We compare our results with the “Filtered MIS” implementation in the Combinatorial BLAS library for two types of synthetic graph inputs.« less

A Data Parallel Multizone Navier-Stokes Code

NASA Technical Reports Server (NTRS)

Jespersen, Dennis C.; Levit, Creon; Kwak, Dochan (Technical Monitor)

1995-01-01

We have developed a data parallel multizone compressible Navier-Stokes code on the Connection Machine CM-5. The code is set up for implicit time-stepping on single or multiple structured grids. For multiple grids and geometrically complex problems, we follow the "chimera" approach, where flow data on one zone is interpolated onto another in the region of overlap. We will describe our design philosophy and give some timing results for the current code. The design choices can be summarized as: 1. finite differences on structured grids; 2. implicit time-stepping with either distributed solves or data motion and local solves; 3. sequential stepping through multiple zones with interzone data transfer via a distributed data structure. We have implemented these ideas on the CM-5 using CMF (Connection Machine Fortran), a data parallel language which combines elements of Fortran 90 and certain extensions, and which bears a strong similarity to High Performance Fortran (HPF). One interesting feature is the issue of turbulence modeling, where the architecture of a parallel machine makes the use of an algebraic turbulence model awkward, whereas models based on transport equations are more natural. We will present some performance figures for the code on the CM-5, and consider the issues involved in transitioning the code to HPF for portability to other parallel platforms.

Parallelization of Program to Optimize Simulated Trajectories (POST3D)

NASA Technical Reports Server (NTRS)

Hammond, Dana P.; Korte, John J. (Technical Monitor)

2001-01-01

This paper describes the parallelization of the Program to Optimize Simulated Trajectories (POST3D). POST3D uses a gradient-based optimization algorithm that reaches an optimum design point by moving from one design point to the next. The gradient calculations required to complete the optimization process, dominate the computational time and have been parallelized using a Single Program Multiple Data (SPMD) on a distributed memory NUMA (non-uniform memory access) architecture. The Origin2000 was used for the tests presented.

Hypercluster Parallel Processor

NASA Technical Reports Server (NTRS)

Blech, Richard A.; Cole, Gary L.; Milner, Edward J.; Quealy, Angela

1992-01-01

Hypercluster computer system includes multiple digital processors, operation of which coordinated through specialized software. Configurable according to various parallel-computing architectures of shared-memory or distributed-memory class, including scalar computer, vector computer, reduced-instruction-set computer, and complex-instruction-set computer. Designed as flexible, relatively inexpensive system that provides single programming and operating environment within which one can investigate effects of various parallel-computing architectures and combinations on performance in solution of complicated problems like those of three-dimensional flows in turbomachines. Hypercluster software and architectural concepts are in public domain.

Parallel algorithms for quantum chemistry. I. Integral transformations on a hypercube multiprocessor

DOE Office of Scientific and Technical Information (OSTI.GOV)

Whiteside, R.A.; Binkley, J.S.; Colvin, M.E.

1987-02-15

For many years it has been recognized that fundamental physical constraints such as the speed of light will limit the ultimate speed of single processor computers to less than about three billion floating point operations per second (3 GFLOPS). This limitation is becoming increasingly restrictive as commercially available machines are now within an order of magnitude of this asymptotic limit. A natural way to avoid this limit is to harness together many processors to work on a single computational problem. In principle, these parallel processing computers have speeds limited only by the number of processors one chooses to acquire. Themore » usefulness of potentially unlimited processing speed to a computationally intensive field such as quantum chemistry is obvious. If these methods are to be applied to significantly larger chemical systems, parallel schemes will have to be employed. For this reason we have developed distributed-memory algorithms for a number of standard quantum chemical methods. We are currently implementing these on a 32 processor Intel hypercube. In this paper we present our algorithm and benchmark results for one of the bottleneck steps in quantum chemical calculations: the four index integral transformation.« less

A manual for PARTI runtime primitives, revision 1

NASA Technical Reports Server (NTRS)

Das, Raja; Saltz, Joel; Berryman, Harry

1991-01-01

Primitives are presented that are designed to help users efficiently program irregular problems (e.g., unstructured mesh sweeps, sparse matrix codes, adaptive mesh partial differential equations solvers) on distributed memory machines. These primitives are also designed for use in compilers for distributed memory multiprocessors. Communications patterns are captured at runtime, and the appropriate send and receive messages are automatically generated.

Spatial light modulators and applications III; Proceedings of the Meeting, San Diego, CA, Aug. 7, 8, 1989

NASA Astrophysics Data System (ADS)

Efron, Uzi

Recent advances in the technology and applications of spatial light modulators (SLMs) are discussed in review essays by leading experts. Topics addressed include materials for SLMs, SLM devices and device technology, applications to optical data processing, and applications to artificial neural networks. Particular attention is given to nonlinear optical polymers, liquid crystals, magnetooptic SLMs, multiple-quantum-well SLMs, deformable-mirror SLMs, three-dimensional optical memories, applications of photorefractive devices to optical computing, photonic neurocomputers and learning machines, holographic associative memories, SLMs as parallel memories for optoelectronic neural networks, and coherent-optics implementations of neural-network models.

Spatial light modulators and applications III; Proceedings of the Meeting, San Diego, CA, Aug. 7, 8, 1989

NASA Technical Reports Server (NTRS)

Efron, Uzi (Editor)

1990-01-01

Recent advances in the technology and applications of spatial light modulators (SLMs) are discussed in review essays by leading experts. Topics addressed include materials for SLMs, SLM devices and device technology, applications to optical data processing, and applications to artificial neural networks. Particular attention is given to nonlinear optical polymers, liquid crystals, magnetooptic SLMs, multiple-quantum-well SLMs, deformable-mirror SLMs, three-dimensional optical memories, applications of photorefractive devices to optical computing, photonic neurocomputers and learning machines, holographic associative memories, SLMs as parallel memories for optoelectronic neural networks, and coherent-optics implementations of neural-network models.

Multiprogramming performance degradation - Case study on a shared memory multiprocessor

NASA Technical Reports Server (NTRS)

Dimpsey, R. T.; Iyer, R. K.

1989-01-01

The performance degradation due to multiprogramming overhead is quantified for a parallel-processing machine. Measurements of real workloads were taken, and it was found that there is a moderate correlation between the completion time of a program and the amount of system overhead measured during program execution. Experiments in controlled environments were then conducted to calculate a lower bound on the performance degradation of parallel jobs caused by multiprogramming overhead. The results show that the multiprogramming overhead of parallel jobs consumes at least 4 percent of the processor time. When two or more serial jobs are introduced into the system, this amount increases to 5.3 percent

The force on the flex: Global parallelism and portability

NASA Technical Reports Server (NTRS)

Jordan, H. F.

1986-01-01

A parallel programming methodology, called the force, supports the construction of programs to be executed in parallel by an unspecified, but potentially large, number of processes. The methodology was originally developed on a pipelined, shared memory multiprocessor, the Denelcor HEP, and embodies the primitive operations of the force in a set of macros which expand into multiprocessor Fortran code. A small set of primitives is sufficient to write large parallel programs, and the system has been used to produce 10,000 line programs in computational fluid dynamics. The level of complexity of the force primitives is intermediate. It is high enough to mask detailed architectural differences between multiprocessors but low enough to give the user control over performance. The system is being ported to a medium scale multiprocessor, the Flex/32, which is a 20 processor system with a mixture of shared and local memory. Memory organization and the type of processor synchronization supported by the hardware on the two machines lead to some differences in efficient implementations of the force primitives, but the user interface remains the same. An initial implementation was done by retargeting the macros to Flexible Computer Corporation's ConCurrent C language. Subsequently, the macros were caused to directly produce the system calls which form the basis for ConCurrent C. The implementation of the Fortran based system is in step with Flexible Computer Corporations's implementation of a Fortran system in the parallel environment.

Phase space simulation of collisionless stellar systems on the massively parallel processor

NASA Technical Reports Server (NTRS)

White, Richard L.

1987-01-01

A numerical technique for solving the collisionless Boltzmann equation describing the time evolution of a self gravitating fluid in phase space was implemented on the Massively Parallel Processor (MPP). The code performs calculations for a two dimensional phase space grid (with one space and one velocity dimension). Some results from calculations are presented. The execution speed of the code is comparable to the speed of a single processor of a Cray-XMP. Advantages and disadvantages of the MPP architecture for this type of problem are discussed. The nearest neighbor connectivity of the MPP array does not pose a significant obstacle. Future MPP-like machines should have much more local memory and easier access to staging memory and disks in order to be effective for this type of problem.

Using CLIPS in the domain of knowledge-based massively parallel programming

NASA Technical Reports Server (NTRS)

Dvorak, Jiri J.

1994-01-01

The Program Development Environment (PDE) is a tool for massively parallel programming of distributed-memory architectures. Adopting a knowledge-based approach, the PDE eliminates the complexity introduced by parallel hardware with distributed memory and offers complete transparency in respect of parallelism exploitation. The knowledge-based part of the PDE is realized in CLIPS. Its principal task is to find an efficient parallel realization of the application specified by the user in a comfortable, abstract, domain-oriented formalism. A large collection of fine-grain parallel algorithmic skeletons, represented as COOL objects in a tree hierarchy, contains the algorithmic knowledge. A hybrid knowledge base with rule modules and procedural parts, encoding expertise about application domain, parallel programming, software engineering, and parallel hardware, enables a high degree of automation in the software development process. In this paper, important aspects of the implementation of the PDE using CLIPS and COOL are shown, including the embedding of CLIPS with C++-based parts of the PDE. The appropriateness of the chosen approach and of the CLIPS language for knowledge-based software engineering are discussed.

A Parallel Multiclassification Algorithm for Big Data Using an Extreme Learning Machine.

PubMed

Duan, Mingxing; Li, Kenli; Liao, Xiangke; Li, Keqin

2018-06-01

As data sets become larger and more complicated, an extreme learning machine (ELM) that runs in a traditional serial environment cannot realize its ability to be fast and effective. Although a parallel ELM (PELM) based on MapReduce to process large-scale data shows more efficient learning speed than identical ELM algorithms in a serial environment, some operations, such as intermediate results stored on disks and multiple copies for each task, are indispensable, and these operations create a large amount of extra overhead and degrade the learning speed and efficiency of the PELMs. In this paper, an efficient ELM based on the Spark framework (SELM), which includes three parallel subalgorithms, is proposed for big data classification. By partitioning the corresponding data sets reasonably, the hidden layer output matrix calculation algorithm, matrix decomposition algorithm, and matrix decomposition algorithm perform most of the computations locally. At the same time, they retain the intermediate results in distributed memory and cache the diagonal matrix as broadcast variables instead of several copies for each task to reduce a large amount of the costs, and these actions strengthen the learning ability of the SELM. Finally, we implement our SELM algorithm to classify large data sets. Extensive experiments have been conducted to validate the effectiveness of the proposed algorithms. As shown, our SELM achieves an speedup on a cluster with ten nodes, and reaches a speedup with 15 nodes, an speedup with 20 nodes, a speedup with 25 nodes, a speedup with 30 nodes, and a speedup with 35 nodes.

Sparse distributed memory

NASA Technical Reports Server (NTRS)

Denning, Peter J.

1989-01-01

Sparse distributed memory was proposed be Pentti Kanerva as a realizable architecture that could store large patterns and retrieve them based on partial matches with patterns representing current sensory inputs. This memory exhibits behaviors, both in theory and in experiment, that resemble those previously unapproached by machines - e.g., rapid recognition of faces or odors, discovery of new connections between seemingly unrelated ideas, continuation of a sequence of events when given a cue from the middle, knowing that one doesn't know, or getting stuck with an answer on the tip of one's tongue. These behaviors are now within reach of machines that can be incorporated into the computing systems of robots capable of seeing, talking, and manipulating. Kanerva's theory is a break with the Western rationalistic tradition, allowing a new interpretation of learning and cognition that respects biology and the mysteries of individual human beings.

Distributed computing methodology for training neural networks in an image-guided diagnostic application.

PubMed

Plagianakos, V P; Magoulas, G D; Vrahatis, M N

2006-03-01

Distributed computing is a process through which a set of computers connected by a network is used collectively to solve a single problem. In this paper, we propose a distributed computing methodology for training neural networks for the detection of lesions in colonoscopy. Our approach is based on partitioning the training set across multiple processors using a parallel virtual machine. In this way, interconnected computers of varied architectures can be used for the distributed evaluation of the error function and gradient values, and, thus, training neural networks utilizing various learning methods. The proposed methodology has large granularity and low synchronization, and has been implemented and tested. Our results indicate that the parallel virtual machine implementation of the training algorithms developed leads to considerable speedup, especially when large network architectures and training sets are used.

A performance comparison of the IBM RS/6000 and the Astronautics ZS-1

DOE Office of Scientific and Technical Information (OSTI.GOV)

Smith, W.M.; Abraham, S.G.; Davidson, E.S.

1991-01-01

Concurrent uniprocessor architectures, of which vector and superscalar are two examples, are designed to capitalize on fine-grain parallelism. The authors have developed a performance evaluation method for comparing and improving these architectures, and in this article they present the methodology and a detailed case study of two machines. The runtime of many programs is dominated by time spent in loop constructs - for example, Fortran Do-loops. Loops generally comprise two logical processes: The access process generates addresses for memory operations while the execute process operates on floating-point data. Memory access patterns typically can be generated independently of the data inmore » the execute process. This independence allows the access process to slip ahead, thereby hiding memory latency. The IBM 360/91 was designed in 1967 to achieve slip dynamically, at runtime. One CPU unit executes integer operations while another handles floating-point operations. Other machines, including the VAX 9000 and the IBM RS/6000, use a similar approach.« less

Enabling the High Level Synthesis of Data Analytics Accelerators

DOE Office of Scientific and Technical Information (OSTI.GOV)

Minutoli, Marco; Castellana, Vito G.; Tumeo, Antonino

Conventional High Level Synthesis (HLS) tools mainly tar- get compute intensive kernels typical of digital signal pro- cessing applications. We are developing techniques and ar- chitectural templates to enable HLS of data analytics appli- cations. These applications are memory intensive, present fine-grained, unpredictable data accesses, and irregular, dy- namic task parallelism. We discuss an architectural tem- plate based around a distributed controller to efficiently ex- ploit thread level parallelism. We present a memory in- terface that supports parallel memory subsystems and en- ables implementing atomic memory operations. We intro- duce a dynamic task scheduling approach to efficiently ex- ecute heavilymore » unbalanced workload. The templates are val- idated by synthesizing queries from the Lehigh University Benchmark (LUBM), a well know SPARQL benchmark.« less

Using Coarrays to Parallelize Legacy Fortran Applications: Strategy and Case Study

DOE PAGES

Radhakrishnan, Hari; Rouson, Damian W. I.; Morris, Karla; ...

2015-01-01

This paper summarizes a strategy for parallelizing a legacy Fortran 77 program using the object-oriented (OO) and coarray features that entered Fortran in the 2003 and 2008 standards, respectively. OO programming (OOP) facilitates the construction of an extensible suite of model-verification and performance tests that drive the development. Coarray parallel programming facilitates a rapid evolution from a serial application to a parallel application capable of running on multicore processors and many-core accelerators in shared and distributed memory. We delineate 17 code modernization steps used to refactor and parallelize the program and study the resulting performance. Our initial studies were donemore » using the Intel Fortran compiler on a 32-core shared memory server. Scaling behavior was very poor, and profile analysis using TAU showed that the bottleneck in the performance was due to our implementation of a collective, sequential summation procedure. We were able to improve the scalability and achieve nearly linear speedup by replacing the sequential summation with a parallel, binary tree algorithm. We also tested the Cray compiler, which provides its own collective summation procedure. Intel provides no collective reductions. With Cray, the program shows linear speedup even in distributed-memory execution. We anticipate similar results with other compilers once they support the new collective procedures proposed for Fortran 2015.« less

Parallel discrete event simulation: A shared memory approach

NASA Technical Reports Server (NTRS)

Reed, Daniel A.; Malony, Allen D.; Mccredie, Bradley D.

1987-01-01

With traditional event list techniques, evaluating a detailed discrete event simulation model can often require hours or even days of computation time. Parallel simulation mimics the interacting servers and queues of a real system by assigning each simulated entity to a processor. By eliminating the event list and maintaining only sufficient synchronization to insure causality, parallel simulation can potentially provide speedups that are linear in the number of processors. A set of shared memory experiments is presented using the Chandy-Misra distributed simulation algorithm to simulate networks of queues. Parameters include queueing network topology and routing probabilities, number of processors, and assignment of network nodes to processors. These experiments show that Chandy-Misra distributed simulation is a questionable alternative to sequential simulation of most queueing network models.

A Multi-Level Parallelization Concept for High-Fidelity Multi-Block Solvers

NASA Technical Reports Server (NTRS)

Hatay, Ferhat F.; Jespersen, Dennis C.; Guruswamy, Guru P.; Rizk, Yehia M.; Byun, Chansup; Gee, Ken; VanDalsem, William R. (Technical Monitor)

1997-01-01

The integration of high-fidelity Computational Fluid Dynamics (CFD) analysis tools with the industrial design process benefits greatly from the robust implementations that are transportable across a wide range of computer architectures. In the present work, a hybrid domain-decomposition and parallelization concept was developed and implemented into the widely-used NASA multi-block Computational Fluid Dynamics (CFD) packages implemented in ENSAERO and OVERFLOW. The new parallel solver concept, PENS (Parallel Euler Navier-Stokes Solver), employs both fine and coarse granularity in data partitioning as well as data coalescing to obtain the desired load-balance characteristics on the available computer platforms. This multi-level parallelism implementation itself introduces no changes to the numerical results, hence the original fidelity of the packages are identically preserved. The present implementation uses the Message Passing Interface (MPI) library for interprocessor message passing and memory accessing. By choosing an appropriate combination of the available partitioning and coalescing capabilities only during the execution stage, the PENS solver becomes adaptable to different computer architectures from shared-memory to distributed-memory platforms with varying degrees of parallelism. The PENS implementation on the IBM SP2 distributed memory environment at the NASA Ames Research Center obtains 85 percent scalable parallel performance using fine-grain partitioning of single-block CFD domains using up to 128 wide computational nodes. Multi-block CFD simulations of complete aircraft simulations achieve 75 percent perfect load-balanced executions using data coalescing and the two levels of parallelism. SGI PowerChallenge, SGI Origin 2000, and a cluster of workstations are the other platforms where the robustness of the implementation is tested. The performance behavior on the other computer platforms with a variety of realistic problems will be included as this on-going study progresses.

Performance of the Heavy Flavor Tracker (HFT) detector in star experiment at RHIC

NASA Astrophysics Data System (ADS)

Alruwaili, Manal

With the growing technology, the number of the processors is becoming massive. Current supercomputer processing will be available on desktops in the next decade. For mass scale application software development on massive parallel computing available on desktops, existing popular languages with large libraries have to be augmented with new constructs and paradigms that exploit massive parallel computing and distributed memory models while retaining the user-friendliness. Currently, available object oriented languages for massive parallel computing such as Chapel, X10 and UPC++ exploit distributed computing, data parallel computing and thread-parallelism at the process level in the PGAS (Partitioned Global Address Space) memory model. However, they do not incorporate: 1) any extension at for object distribution to exploit PGAS model; 2) the programs lack the flexibility of migrating or cloning an object between places to exploit load balancing; and 3) lack the programming paradigms that will result from the integration of data and thread-level parallelism and object distribution. In the proposed thesis, I compare different languages in PGAS model; propose new constructs that extend C++ with object distribution and object migration; and integrate PGAS based process constructs with these extensions on distributed objects. Object cloning and object migration. Also a new paradigm MIDD (Multiple Invocation Distributed Data) is presented when different copies of the same class can be invoked, and work on different elements of a distributed data concurrently using remote method invocations. I present new constructs, their grammar and their behavior. The new constructs have been explained using simple programs utilizing these constructs.

Using PVM to host CLIPS in distributed environments

NASA Technical Reports Server (NTRS)

Myers, Leonard; Pohl, Kym

1994-01-01

It is relatively easy to enhance CLIPS (C Language Integrated Production System) to support multiple expert systems running in a distributed environment with heterogeneous machines. The task is minimized by using the PVM (Parallel Virtual Machine) code from Oak Ridge Labs to provide the distributed utility. PVM is a library of C and FORTRAN subprograms that supports distributive computing on many different UNIX platforms. A PVM deamon is easily installed on each CPU that enters the virtual machine environment. Any user with rsh or rexec access to a machine can use the one PVM deamon to obtain a generous set of distributed facilities. The ready availability of both CLIPS and PVM makes the combination of software particularly attractive for budget conscious experimentation of heterogeneous distributive computing with multiple CLIPS executables. This paper presents a design that is sufficient to provide essential message passing functions in CLIPS and enable the full range of PVM facilities.

A Pervasive Parallel Processing Framework for Data Visualization and Analysis at Extreme Scale

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ma, Kwan-Liu

Most of today’s visualization libraries and applications are based off of what is known today as the visualization pipeline. In the visualization pipeline model, algorithms are encapsulated as “filtering” components with inputs and outputs. These components can be combined by connecting the outputs of one filter to the inputs of another filter. The visualization pipeline model is popular because it provides a convenient abstraction that allows users to combine algorithms in powerful ways. Unfortunately, the visualization pipeline cannot run effectively on exascale computers. Experts agree that the exascale machine will comprise processors that contain many cores. Furthermore, physical limitations willmore » prevent data movement in and out of the chip (that is, between main memory and the processing cores) from keeping pace with improvements in overall compute performance. To use these processors to their fullest capability, it is essential to carefully consider memory access. This is where the visualization pipeline fails. Each filtering component in the visualization library is expected to take a data set in its entirety, perform some computation across all of the elements, and output the complete results. The process of iterating over all elements must be repeated in each filter, which is one of the worst possible ways to traverse memory when trying to maximize the number of executions per memory access. This project investigates a new type of visualization framework that exhibits a pervasive parallelism necessary to run on exascale machines. Our framework achieves this by defining algorithms in terms of functors, which are localized, stateless operations. Functors can be composited in much the same way as filters in the visualization pipeline. But, functors’ design allows them to be concurrently running on massive amounts of lightweight threads. Only with such fine-grained parallelism can we hope to fill the billions of threads we expect will be necessary for efficient computation on an exascale computer. This project concludes with a functional prototype containing pervasively parallel algorithms that perform demonstratively well on many-core processors. These algorithms are fundamental for performing data analysis and visualization at extreme scale.« less

Introducing concurrency in the Gaudi data processing framework

NASA Astrophysics Data System (ADS)

Clemencic, Marco; Hegner, Benedikt; Mato, Pere; Piparo, Danilo

2014-06-01

In the past, the increasing demands for HEP processing resources could be fulfilled by the ever increasing clock-frequencies and by distributing the work to more and more physical machines. Limitations in power consumption of both CPUs and entire data centres are bringing an end to this era of easy scalability. To get the most CPU performance per watt, future hardware will be characterised by less and less memory per processor, as well as thinner, more specialized and more numerous cores per die, and rather heterogeneous resources. To fully exploit the potential of the many cores, HEP data processing frameworks need to allow for parallel execution of reconstruction or simulation algorithms on several events simultaneously. We describe our experience in introducing concurrency related capabilities into Gaudi, a generic data processing software framework, which is currently being used by several HEP experiments, including the ATLAS and LHCb experiments at the LHC. After a description of the concurrent framework and the most relevant design choices driving its development, we describe the behaviour of the framework in a more realistic environment, using a subset of the real LHCb reconstruction workflow, and present our strategy and the used tools to validate the physics outcome of the parallel framework against the results of the present, purely sequential LHCb software. We then summarize the measurement of the code performance of the multithreaded application in terms of memory and CPU usage.

Parallel language constructs for tensor product computations on loosely coupled architectures

NASA Technical Reports Server (NTRS)

Mehrotra, Piyush; Vanrosendale, John

1989-01-01

Distributed memory architectures offer high levels of performance and flexibility, but have proven awkard to program. Current languages for nonshared memory architectures provide a relatively low level programming environment, and are poorly suited to modular programming, and to the construction of libraries. A set of language primitives designed to allow the specification of parallel numerical algorithms at a higher level is described. Tensor product array computations are focused on along with a simple but important class of numerical algorithms. The problem of programming 1-D kernal routines is focused on first, such as parallel tridiagonal solvers, and then how such parallel kernels can be combined to form parallel tensor product algorithms is examined.

Precision Parameter Estimation and Machine Learning

NASA Astrophysics Data System (ADS)

Wandelt, Benjamin D.

2008-12-01

I discuss the strategy of ``Acceleration by Parallel Precomputation and Learning'' (AP-PLe) that can vastly accelerate parameter estimation in high-dimensional parameter spaces and costly likelihood functions, using trivially parallel computing to speed up sequential exploration of parameter space. This strategy combines the power of distributed computing with machine learning and Markov-Chain Monte Carlo techniques efficiently to explore a likelihood function, posterior distribution or χ2-surface. This strategy is particularly successful in cases where computing the likelihood is costly and the number of parameters is moderate or large. We apply this technique to two central problems in cosmology: the solution of the cosmological parameter estimation problem with sufficient accuracy for the Planck data using PICo; and the detailed calculation of cosmological helium and hydrogen recombination with RICO. Since the APPLe approach is designed to be able to use massively parallel resources to speed up problems that are inherently serial, we can bring the power of distributed computing to bear on parameter estimation problems. We have demonstrated this with the CosmologyatHome project.

A parallel Monte Carlo code for planar and SPECT imaging: implementation, verification and applications in (131)I SPECT.

PubMed

Dewaraja, Yuni K; Ljungberg, Michael; Majumdar, Amitava; Bose, Abhijit; Koral, Kenneth F

2002-02-01

This paper reports the implementation of the SIMIND Monte Carlo code on an IBM SP2 distributed memory parallel computer. Basic aspects of running Monte Carlo particle transport calculations on parallel architectures are described. Our parallelization is based on equally partitioning photons among the processors and uses the Message Passing Interface (MPI) library for interprocessor communication and the Scalable Parallel Random Number Generator (SPRNG) to generate uncorrelated random number streams. These parallelization techniques are also applicable to other distributed memory architectures. A linear increase in computing speed with the number of processors is demonstrated for up to 32 processors. This speed-up is especially significant in Single Photon Emission Computed Tomography (SPECT) simulations involving higher energy photon emitters, where explicit modeling of the phantom and collimator is required. For (131)I, the accuracy of the parallel code is demonstrated by comparing simulated and experimental SPECT images from a heart/thorax phantom. Clinically realistic SPECT simulations using the voxel-man phantom are carried out to assess scatter and attenuation correction.

Multiprocessing the Sieve of Eratosthenes

NASA Technical Reports Server (NTRS)

Bokhari, S.

1986-01-01

The Sieve of Eratosthenes for finding prime numbers in recent years has seen much use as a benchmark algorithm for serial computers while its intrinsically parallel nature has gone largely unnoticed. The implementation of a parallel version of this algorithm for a real parallel computer, the Flex/32, is described and its performance discussed. It is shown that the algorithm is sensitive to several fundamental performance parameters of parallel machines, such as spawning time, signaling time, memory access, and overhead of process switching. Because of the nature of the algorithm, it is impossible to get any speedup beyond 4 or 5 processors unless some form of dynamic load balancing is employed. We describe the performance of our algorithm with and without load balancing and compare it with theoretical lower bounds and simulated results. It is straightforward to understand this algorithm and to check the final results. However, its efficient implementation on a real parallel machine requires thoughtful design, especially if dynamic load balancing is desired. The fundamental operations required by the algorithm are very simple: this means that the slightest overhead appears prominently in performance data. The Sieve thus serves not only as a very severe test of the capabilities of a parallel processor but is also an interesting challenge for the programmer.

YAPPA: a Compiler-Based Parallelization Framework for Irregular Applications on MPSoCs

DOE Office of Scientific and Technical Information (OSTI.GOV)

Lovergine, Silvia; Tumeo, Antonino; Villa, Oreste

Modern embedded systems include hundreds of cores. Because of the difficulty in providing a fast, coherent memory architecture, these systems usually rely on non-coherent, non-uniform memory architectures with private memories for each core. However, programming these systems poses significant challenges. The developer must extract large amounts of parallelism, while orchestrating communication among cores to optimize application performance. These issues become even more significant with irregular applications, which present data sets difficult to partition, unpredictable memory accesses, unbalanced control flow and fine grained communication. Hand-optimizing every single aspect is hard and time-consuming, and it often does not lead to the expectedmore » performance. There is a growing gap between such complex and highly-parallel architectures and the high level languages used to describe the specification, which were designed for simpler systems and do not consider these new issues. In this paper we introduce YAPPA (Yet Another Parallel Programming Approach), a compilation framework for the automatic parallelization of irregular applications on modern MPSoCs based on LLVM. We start by considering an efficient parallel programming approach for irregular applications on distributed memory systems. We then propose a set of transformations that can reduce the development and optimization effort. The results of our initial prototype confirm the correctness of the proposed approach.« less

Myria: Scalable Analytics as a Service

NASA Astrophysics Data System (ADS)

Howe, B.; Halperin, D.; Whitaker, A.

2014-12-01

At the UW eScience Institute, we're working to empower non-experts, especially in the sciences, to write and use data-parallel algorithms. To this end, we are building Myria, a web-based platform for scalable analytics and data-parallel programming. Myria's internal model of computation is the relational algebra extended with iteration, such that every program is inherently data-parallel, just as every query in a database is inherently data-parallel. But unlike databases, iteration is a first class concept, allowing us to express machine learning tasks, graph traversal tasks, and more. Programs can be expressed in a number of languages and can be executed on a number of execution environments, but we emphasize a particular language called MyriaL that supports both imperative and declarative styles and a particular execution engine called MyriaX that uses an in-memory column-oriented representation and asynchronous iteration. We deliver Myria over the web as a service, providing an editor, performance analysis tools, and catalog browsing features in a single environment. We find that this web-based "delivery vector" is critical in reaching non-experts: they are insulated from irrelevant effort technical work associated with installation, configuration, and resource management. The MyriaX backend, one of several execution runtimes we support, is a main-memory, column-oriented, RDBMS-on-the-worker system that supports cyclic data flows as a first-class citizen and has been shown to outperform competitive systems on 100-machine cluster sizes. I will describe the Myria system, give a demo, and present some new results in large-scale oceanographic microbiology.

A Parameter Communication Optimization Strategy for Distributed Machine Learning in Sensors

PubMed Central

Zhang, Jilin; Tu, Hangdi; Ren, Yongjian; Wan, Jian; Zhou, Li; Li, Mingwei; Wang, Jue; Yu, Lifeng; Zhao, Chang; Zhang, Lei

2017-01-01

In order to utilize the distributed characteristic of sensors, distributed machine learning has become the mainstream approach, but the different computing capability of sensors and network delays greatly influence the accuracy and the convergence rate of the machine learning model. Our paper describes a reasonable parameter communication optimization strategy to balance the training overhead and the communication overhead. We extend the fault tolerance of iterative-convergent machine learning algorithms and propose the Dynamic Finite Fault Tolerance (DFFT). Based on the DFFT, we implement a parameter communication optimization strategy for distributed machine learning, named Dynamic Synchronous Parallel Strategy (DSP), which uses the performance monitoring model to dynamically adjust the parameter synchronization strategy between worker nodes and the Parameter Server (PS). This strategy makes full use of the computing power of each sensor, ensures the accuracy of the machine learning model, and avoids the situation that the model training is disturbed by any tasks unrelated to the sensors. PMID:28934163

Parallel algorithms for modeling flow in permeable media. Annual report, February 15, 1995 - February 14, 1996

DOE Office of Scientific and Technical Information (OSTI.GOV)

G.A. Pope; K. Sephernoori; D.C. McKinney

1996-03-15

This report describes the application of distributed-memory parallel programming techniques to a compositional simulator called UTCHEM. The University of Texas Chemical Flooding reservoir simulator (UTCHEM) is a general-purpose vectorized chemical flooding simulator that models the transport of chemical species in three-dimensional, multiphase flow through permeable media. The parallel version of UTCHEM addresses solving large-scale problems by reducing the amount of time that is required to obtain the solution as well as providing a flexible and portable programming environment. In this work, the original parallel version of UTCHEM was modified and ported to CRAY T3D and CRAY T3E, distributed-memory, multiprocessor computersmore » using CRAY-PVM as the interprocessor communication library. Also, the data communication routines were modified such that the portability of the original code across different computer architectures was mad possible.« less

Efficacy of Code Optimization on Cache-based Processors

NASA Technical Reports Server (NTRS)

VanderWijngaart, Rob F.; Chancellor, Marisa K. (Technical Monitor)

1997-01-01

The current common wisdom in the U.S. is that the powerful, cost-effective supercomputers of tomorrow will be based on commodity (RISC) micro-processors with cache memories. Already, most distributed systems in the world use such hardware as building blocks. This shift away from vector supercomputers and towards cache-based systems has brought about a change in programming paradigm, even when ignoring issues of parallelism. Vector machines require inner-loop independence and regular, non-pathological memory strides (usually this means: non-power-of-two strides) to allow efficient vectorization of array operations. Cache-based systems require spatial and temporal locality of data, so that data once read from main memory and stored in high-speed cache memory is used optimally before being written back to main memory. This means that the most cache-friendly array operations are those that feature zero or unit stride, so that each unit of data read from main memory (a cache line) contains information for the next iteration in the loop. Moreover, loops ought to be 'fat', meaning that as many operations as possible are performed on cache data-provided instruction caches do not overflow and enough registers are available. If unit stride is not possible, for example because of some data dependency, then care must be taken to avoid pathological strides, just ads on vector computers. For cache-based systems the issues are more complex, due to the effects of associativity and of non-unit block (cache line) size. But there is more to the story. Most modern micro-processors are superscalar, which means that they can issue several (arithmetic) instructions per clock cycle, provided that there are enough independent instructions in the loop body. This is another argument for providing fat loop bodies. With these restrictions, it appears fairly straightforward to produce code that will run efficiently on any cache-based system. It can be argued that although some of the important computational algorithms employed at NASA Ames require different programming styles on vector machines and cache-based machines, respectively, neither architecture class appeared to be favored by particular algorithms in principle. Practice tells us that the situation is more complicated. This report presents observations and some analysis of performance tuning for cache-based systems. We point out several counterintuitive results that serve as a cautionary reminder that memory accesses are not the only factors that determine performance, and that within the class of cache-based systems, significant differences exist.

High Performance Distributed Computing in a Supercomputer Environment: Computational Services and Applications Issues

NASA Technical Reports Server (NTRS)

Kramer, Williams T. C.; Simon, Horst D.

1994-01-01

This tutorial proposes to be a practical guide for the uninitiated to the main topics and themes of high-performance computing (HPC), with particular emphasis to distributed computing. The intent is first to provide some guidance and directions in the rapidly increasing field of scientific computing using both massively parallel and traditional supercomputers. Because of their considerable potential computational power, loosely or tightly coupled clusters of workstations are increasingly considered as a third alternative to both the more conventional supercomputers based on a small number of powerful vector processors, as well as high massively parallel processors. Even though many research issues concerning the effective use of workstation clusters and their integration into a large scale production facility are still unresolved, such clusters are already used for production computing. In this tutorial we will utilize the unique experience made at the NAS facility at NASA Ames Research Center. Over the last five years at NAS massively parallel supercomputers such as the Connection Machines CM-2 and CM-5 from Thinking Machines Corporation and the iPSC/860 (Touchstone Gamma Machine) and Paragon Machines from Intel were used in a production supercomputer center alongside with traditional vector supercomputers such as the Cray Y-MP and C90.

Large Scale Analysis of Geospatial Data with Dask and XArray

NASA Astrophysics Data System (ADS)

Zender, C. S.; Hamman, J.; Abernathey, R.; Evans, K. J.; Rocklin, M.; Zender, C. S.; Rocklin, M.

2017-12-01

The analysis of geospatial data with high level languages has acceleratedinnovation and the impact of existing data resources. However, as datasetsgrow beyond single-machine memory, data structures within these high levellanguages can become a bottleneck. New libraries like Dask and XArray resolve some of these scalability issues,providing interactive workflows that are both familiar tohigh-level-language researchers while also scaling out to much largerdatasets. This broadens the access of researchers to larger datasets on highperformance computers and, through interactive development, reducestime-to-insight when compared to traditional parallel programming techniques(MPI). This talk describes Dask, a distributed dynamic task scheduler, Dask.array, amulti-dimensional array that copies the popular NumPy interface, and XArray,a library that wraps NumPy/Dask.array with labeled and indexes axes,implementing the CF conventions. We discuss both the basic design of theselibraries and how they change interactive analysis of geospatial data, and alsorecent benefits and challenges of distributed computing on clusters ofmachines.

Quantum Monte Carlo for large chemical systems: implementing efficient strategies for petascale platforms and beyond.

PubMed

Scemama, Anthony; Caffarel, Michel; Oseret, Emmanuel; Jalby, William

2013-04-30

Various strategies to implement efficiently quantum Monte Carlo (QMC) simulations for large chemical systems are presented. These include: (i) the introduction of an efficient algorithm to calculate the computationally expensive Slater matrices. This novel scheme is based on the use of the highly localized character of atomic Gaussian basis functions (not the molecular orbitals as usually done), (ii) the possibility of keeping the memory footprint minimal, (iii) the important enhancement of single-core performance when efficient optimization tools are used, and (iv) the definition of a universal, dynamic, fault-tolerant, and load-balanced framework adapted to all kinds of computational platforms (massively parallel machines, clusters, or distributed grids). These strategies have been implemented in the QMC=Chem code developed at Toulouse and illustrated with numerical applications on small peptides of increasing sizes (158, 434, 1056, and 1731 electrons). Using 10-80 k computing cores of the Curie machine (GENCI-TGCC-CEA, France), QMC=Chem has been shown to be capable of running at the petascale level, thus demonstrating that for this machine a large part of the peak performance can be achieved. Implementation of large-scale QMC simulations for future exascale platforms with a comparable level of efficiency is expected to be feasible. Copyright © 2013 Wiley Periodicals, Inc.

Parallelising a molecular dynamics algorithm on a multi-processor workstation

NASA Astrophysics Data System (ADS)

Müller-Plathe, Florian

1990-12-01

The Verlet neighbour-list algorithm is parallelised for a multi-processor Hewlett-Packard/Apollo DN10000 workstation. The implementation makes use of memory shared between the processors. It is a genuine master-slave approach by which most of the computational tasks are kept in the master process and the slaves are only called to do part of the nonbonded forces calculation. The implementation features elements of both fine-grain and coarse-grain parallelism. Apart from three calls to library routines, two of which are standard UNIX calls, and two machine-specific language extensions, the whole code is written in standard Fortran 77. Hence, it may be expected that this parallelisation concept can be transfered in parts or as a whole to other multi-processor shared-memory computers. The parallel code is routinely used in production work.

A Screen Space GPGPU Surface LIC Algorithm for Distributed Memory Data Parallel Sort Last Rendering Infrastructures

NASA Astrophysics Data System (ADS)

Loring, B.; Karimabadi, H.; Rortershteyn, V.

2015-10-01

The surface line integral convolution(LIC) visualization technique produces dense visualization of vector fields on arbitrary surfaces. We present a screen space surface LIC algorithm for use in distributed memory data parallel sort last rendering infrastructures. The motivations for our work are to support analysis of datasets that are too large to fit in the main memory of a single computer and compatibility with prevalent parallel scientific visualization tools such as ParaView and VisIt. By working in screen space using OpenGL we can leverage the computational power of GPUs when they are available and run without them when they are not. We address efficiency and performance issues that arise from the transformation of data from physical to screen space by selecting an alternate screen space domain decomposition. We analyze the algorithm's scaling behavior with and without GPUs on two high performance computing systems using data from turbulent plasma simulations.

A Screen Space GPGPU Surface LIC Algorithm for Distributed Memory Data Parallel Sort Last Rendering Infrastructures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Loring, Burlen; Karimabadi, Homa; Rortershteyn, Vadim

2014-07-01

The surface line integral convolution(LIC) visualization technique produces dense visualization of vector fields on arbitrary surfaces. We present a screen space surface LIC algorithm for use in distributed memory data parallel sort last rendering infrastructures. The motivations for our work are to support analysis of datasets that are too large to fit in the main memory of a single computer and compatibility with prevalent parallel scientific visualization tools such as ParaView and VisIt. By working in screen space using OpenGL we can leverage the computational power of GPUs when they are available and run without them when they are not.more » We address efficiency and performance issues that arise from the transformation of data from physical to screen space by selecting an alternate screen space domain decomposition. We analyze the algorithm's scaling behavior with and without GPUs on two high performance computing systems using data from turbulent plasma simulations.« less

Parallel Navier-Stokes computations on shared and distributed memory architectures

NASA Technical Reports Server (NTRS)

Hayder, M. Ehtesham; Jayasimha, D. N.; Pillay, Sasi Kumar

1995-01-01

We study a high order finite difference scheme to solve the time accurate flow field of a jet using the compressible Navier-Stokes equations. As part of our ongoing efforts, we have implemented our numerical model on three parallel computing platforms to study the computational, communication, and scalability characteristics. The platforms chosen for this study are a cluster of workstations connected through fast networks (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), and a distributed memory multiprocessor (the IBM SPI). Our focus in this study is on the LACE testbed. We present some results for the Cray YMP and the IBM SP1 mainly for comparison purposes. On the LACE testbed, we study: (1) the communication characteristics of Ethernet, FDDI, and the ALLNODE networks and (2) the overheads induced by the PVM message passing library used for parallelizing the application. We demonstrate that clustering of workstations is effective and has the potential to be computationally competitive with supercomputers at a fraction of the cost.

SKIRT: Hybrid parallelization of radiative transfer simulations

NASA Astrophysics Data System (ADS)

Verstocken, S.; Van De Putte, D.; Camps, P.; Baes, M.

2017-07-01

We describe the design, implementation and performance of the new hybrid parallelization scheme in our Monte Carlo radiative transfer code SKIRT, which has been used extensively for modelling the continuum radiation of dusty astrophysical systems including late-type galaxies and dusty tori. The hybrid scheme combines distributed memory parallelization, using the standard Message Passing Interface (MPI) to communicate between processes, and shared memory parallelization, providing multiple execution threads within each process to avoid duplication of data structures. The synchronization between multiple threads is accomplished through atomic operations without high-level locking (also called lock-free programming). This improves the scaling behaviour of the code and substantially simplifies the implementation of the hybrid scheme. The result is an extremely flexible solution that adjusts to the number of available nodes, processors and memory, and consequently performs well on a wide variety of computing architectures.

Solving very large, sparse linear systems on mesh-connected parallel computers

NASA Technical Reports Server (NTRS)

Opsahl, Torstein; Reif, John

1987-01-01

The implementation of Pan and Reif's Parallel Nested Dissection (PND) algorithm on mesh connected parallel computers is described. This is the first known algorithm that allows very large, sparse linear systems of equations to be solved efficiently in polylog time using a small number of processors. How the processor bound of PND can be matched to the number of processors available on a given parallel computer by slowing down the algorithm by constant factors is described. Also, for the important class of problems where G(A) is a grid graph, a unique memory mapping that reduces the inter-processor communication requirements of PND to those that can be executed on mesh connected parallel machines is detailed. A description of an implementation on the Goodyear Massively Parallel Processor (MPP), located at Goddard is given. Also, a detailed discussion of data mappings and performance issues is given.

Static analysis of the hull plate using the finite element method

NASA Astrophysics Data System (ADS)

Ion, A.

2015-11-01

This paper aims at presenting the static analysis for two levels of a container ship's construction as follows: the first level is at the girder / hull plate and the second level is conducted at the entire strength hull of the vessel. This article will describe the work for the static analysis of a hull plate. We shall use the software package ANSYS Mechanical 14.5. The program is run on a computer with four Intel Xeon X5260 CPU processors at 3.33 GHz, 32 GB memory installed. In terms of software, the shared memory parallel version of ANSYS refers to running ANSYS across multiple cores on a SMP system. The distributed memory parallel version of ANSYS (Distributed ANSYS) refers to running ANSYS across multiple processors on SMP systems or DMP systems.

Hadoop neural network for parallel and distributed feature selection.

PubMed

Hodge, Victoria J; O'Keefe, Simon; Austin, Jim

2016-06-01

In this paper, we introduce a theoretical basis for a Hadoop-based neural network for parallel and distributed feature selection in Big Data sets. It is underpinned by an associative memory (binary) neural network which is highly amenable to parallel and distributed processing and fits with the Hadoop paradigm. There are many feature selectors described in the literature which all have various strengths and weaknesses. We present the implementation details of five feature selection algorithms constructed using our artificial neural network framework embedded in Hadoop YARN. Hadoop allows parallel and distributed processing. Each feature selector can be divided into subtasks and the subtasks can then be processed in parallel. Multiple feature selectors can also be processed simultaneously (in parallel) allowing multiple feature selectors to be compared. We identify commonalities among the five features selectors. All can be processed in the framework using a single representation and the overall processing can also be greatly reduced by only processing the common aspects of the feature selectors once and propagating these aspects across all five feature selectors as necessary. This allows the best feature selector and the actual features to select to be identified for large and high dimensional data sets through exploiting the efficiency and flexibility of embedding the binary associative-memory neural network in Hadoop. Copyright © 2015 The Authors. Published by Elsevier Ltd.. All rights reserved.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Sewell, Christopher Meyer

This is a set of slides from a guest lecture for a class at the University of Texas, El Paso on visualization and data analysis for high-performance computing. The topics covered are the following: trends in high-performance computing; scientific visualization, such as OpenGL, ray tracing and volume rendering, VTK, and ParaView; data science at scale, such as in-situ visualization, image databases, distributed memory parallelism, shared memory parallelism, VTK-m, "big data", and then an analysis example.

Supercomputing '91; Proceedings of the 4th Annual Conference on High Performance Computing, Albuquerque, NM, Nov. 18-22, 1991

NASA Technical Reports Server (NTRS)

1991-01-01

Various papers on supercomputing are presented. The general topics addressed include: program analysis/data dependence, memory access, distributed memory code generation, numerical algorithms, supercomputer benchmarks, latency tolerance, parallel programming, applications, processor design, networks, performance tools, mapping and scheduling, characterization affecting performance, parallelism packaging, computing climate change, combinatorial algorithms, hardware and software performance issues, system issues. (No individual items are abstracted in this volume)

Orchid: a novel management, annotation and machine learning framework for analyzing cancer mutations.

PubMed

Cario, Clinton L; Witte, John S

2018-03-15

As whole-genome tumor sequence and biological annotation datasets grow in size, number and content, there is an increasing basic science and clinical need for efficient and accurate data management and analysis software. With the emergence of increasingly sophisticated data stores, execution environments and machine learning algorithms, there is also a need for the integration of functionality across frameworks. We present orchid, a python based software package for the management, annotation and machine learning of cancer mutations. Building on technologies of parallel workflow execution, in-memory database storage and machine learning analytics, orchid efficiently handles millions of mutations and hundreds of features in an easy-to-use manner. We describe the implementation of orchid and demonstrate its ability to distinguish tissue of origin in 12 tumor types based on 339 features using a random forest classifier. Orchid and our annotated tumor mutation database are freely available at https://github.com/wittelab/orchid. Software is implemented in python 2.7, and makes use of MySQL or MemSQL databases. Groovy 2.4.5 is optionally required for parallel workflow execution. JWitte@ucsf.edu. Supplementary data are available at Bioinformatics online.

Task Assignment Heuristics for Distributed CFD Applications

NASA Technical Reports Server (NTRS)

Lopez-Benitez, N.; Djomehri, M. J.; Biswas, R.; Biegel, Bryan (Technical Monitor)

2001-01-01

CFD applications require high-performance computational platforms: 1. Complex physics and domain configuration demand strongly coupled solutions; 2. Applications are CPU and memory intensive; and 3. Huge resource requirements can only be satisfied by teraflop-scale machines or distributed computing.

An analysis of options available for developing a common laser ray tracing package for Ares and Kull code frameworks

DOE Office of Scientific and Technical Information (OSTI.GOV)

Weeratunga, S K

Ares and Kull are mature code frameworks that support ALE hydrodynamics for a variety of HEDP applications at LLNL, using two widely different meshing approaches. While Ares is based on a 2-D/3-D block-structured mesh data base, Kull is designed to support unstructured, arbitrary polygonal/polyhedral meshes. In addition, both frameworks are capable of running applications on large, distributed-memory parallel machines. Currently, both these frameworks separately support assorted collections of physics packages related to HEDP, including one for the energy deposition by laser/ion-beam ray tracing. This study analyzes the options available for developing a common laser/ion-beam ray tracing package that can bemore » easily shared between these two code frameworks and concludes with a set of recommendations for its development.« less

BIRD: A general interface for sparse distributed memory simulators

NASA Technical Reports Server (NTRS)

Rogers, David

1990-01-01

Kanerva's sparse distributed memory (SDM) has now been implemented for at least six different computers, including SUN3 workstations, the Apple Macintosh, and the Connection Machine. A common interface for input of commands would both aid testing of programs on a broad range of computer architectures and assist users in transferring results from research environments to applications. A common interface also allows secondary programs to generate command sequences for a sparse distributed memory, which may then be executed on the appropriate hardware. The BIRD program is an attempt to create such an interface. Simplifying access to different simulators should assist developers in finding appropriate uses for SDM.

Parallel, Distributed Scripting with Python

DOE Office of Scientific and Technical Information (OSTI.GOV)

Miller, P J

2002-05-24

Parallel computers used to be, for the most part, one-of-a-kind systems which were extremely difficult to program portably. With SMP architectures, the advent of the POSIX thread API and OpenMP gave developers ways to portably exploit on-the-box shared memory parallelism. Since these architectures didn't scale cost-effectively, distributed memory clusters were developed. The associated MPI message passing libraries gave these systems a portable paradigm too. Having programmers effectively use this paradigm is a somewhat different question. Distributed data has to be explicitly transported via the messaging system in order for it to be useful. In high level languages, the MPI librarymore » gives access to data distribution routines in C, C++, and FORTRAN. But we need more than that. Many reasonable and common tasks are best done in (or as extensions to) scripting languages. Consider sysadm tools such as password crackers, file purgers, etc ... These are simple to write in a scripting language such as Python (an open source, portable, and freely available interpreter). But these tasks beg to be done in parallel. Consider the a password checker that checks an encrypted password against a 25,000 word dictionary. This can take around 10 seconds in Python (6 seconds in C). It is trivial to parallelize if you can distribute the information and co-ordinate the work.« less

The implementation of an aeronautical CFD flow code onto distributed memory parallel systems

NASA Astrophysics Data System (ADS)

Ierotheou, C. S.; Forsey, C. R.; Leatham, M.

2000-04-01

The parallelization of an industrially important in-house computational fluid dynamics (CFD) code for calculating the airflow over complex aircraft configurations using the Euler or Navier-Stokes equations is presented. The code discussed is the flow solver module of the SAUNA CFD suite. This suite uses a novel grid system that may include block-structured hexahedral or pyramidal grids, unstructured tetrahedral grids or a hybrid combination of both. To assist in the rapid convergence to a solution, a number of convergence acceleration techniques are employed including implicit residual smoothing and a multigrid full approximation storage scheme (FAS). Key features of the parallelization approach are the use of domain decomposition and encapsulated message passing to enable the execution in parallel using a single programme multiple data (SPMD) paradigm. In the case where a hybrid grid is used, a unified grid partitioning scheme is employed to define the decomposition of the mesh. The parallel code has been tested using both structured and hybrid grids on a number of different distributed memory parallel systems and is now routinely used to perform industrial scale aeronautical simulations. Copyright

Compacting de Bruijn graphs from sequencing data quickly and in low memory.

PubMed

Chikhi, Rayan; Limasset, Antoine; Medvedev, Paul

2016-06-15

As the quantity of data per sequencing experiment increases, the challenges of fragment assembly are becoming increasingly computational. The de Bruijn graph is a widely used data structure in fragment assembly algorithms, used to represent the information from a set of reads. Compaction is an important data reduction step in most de Bruijn graph based algorithms where long simple paths are compacted into single vertices. Compaction has recently become the bottleneck in assembly pipelines, and improving its running time and memory usage is an important problem. We present an algorithm and a tool bcalm 2 for the compaction of de Bruijn graphs. bcalm 2 is a parallel algorithm that distributes the input based on a minimizer hashing technique, allowing for good balance of memory usage throughout its execution. For human sequencing data, bcalm 2 reduces the computational burden of compacting the de Bruijn graph to roughly an hour and 3 GB of memory. We also applied bcalm 2 to the 22 Gbp loblolly pine and 20 Gbp white spruce sequencing datasets. Compacted graphs were constructed from raw reads in less than 2 days and 40 GB of memory on a single machine. Hence, bcalm 2 is at least an order of magnitude more efficient than other available methods. Source code of bcalm 2 is freely available at: https://github.com/GATB/bcalm rayan.chikhi@univ-lille1.fr. © The Author 2016. Published by Oxford University Press.

[Series: Medical Applications of the PHITS Code (2): Acceleration by Parallel Computing].

PubMed

Furuta, Takuya; Sato, Tatsuhiko

2015-01-01

Time-consuming Monte Carlo dose calculation becomes feasible owing to the development of computer technology. However, the recent development is due to emergence of the multi-core high performance computers. Therefore, parallel computing becomes a key to achieve good performance of software programs. A Monte Carlo simulation code PHITS contains two parallel computing functions, the distributed-memory parallelization using protocols of message passing interface (MPI) and the shared-memory parallelization using open multi-processing (OpenMP) directives. Users can choose the two functions according to their needs. This paper gives the explanation of the two functions with their advantages and disadvantages. Some test applications are also provided to show their performance using a typical multi-core high performance workstation.

Algorithm implementation on the Navier-Stokes computer

DOE Office of Scientific and Technical Information (OSTI.GOV)

Krist, S.E.; Zang, T.A.

1987-03-01

The Navier-Stokes Computer is a multi-purpose parallel-processing supercomputer which is currently under development at Princeton University. It consists of multiple local memory parallel processors, called Nodes, which are interconnected in a hypercube network. Details of the procedures involved in implementing an algorithm on the Navier-Stokes computer are presented. The particular finite difference algorithm considered in this analysis was developed for simulation of laminar-turbulent transition in wall bounded shear flows. Projected timing results for implementing this algorithm indicate that operation rates in excess of 42 GFLOPS are feasible on a 128 Node machine.

Algorithm implementation on the Navier-Stokes computer

NASA Technical Reports Server (NTRS)

Krist, Steven E.; Zang, Thomas A.

1987-01-01

The Navier-Stokes Computer is a multi-purpose parallel-processing supercomputer which is currently under development at Princeton University. It consists of multiple local memory parallel processors, called Nodes, which are interconnected in a hypercube network. Details of the procedures involved in implementing an algorithm on the Navier-Stokes computer are presented. The particular finite difference algorithm considered in this analysis was developed for simulation of laminar-turbulent transition in wall bounded shear flows. Projected timing results for implementing this algorithm indicate that operation rates in excess of 42 GFLOPS are feasible on a 128 Node machine.

Distributed memory parallel Markov random fields using graph partitioning

DOE Office of Scientific and Technical Information (OSTI.GOV)

Heinemann, C.; Perciano, T.; Ushizima, D.

Markov random fields (MRF) based algorithms have attracted a large amount of interest in image analysis due to their ability to exploit contextual information about data. Image data generated by experimental facilities, though, continues to grow larger and more complex, making it more difficult to analyze in a reasonable amount of time. Applying image processing algorithms to large datasets requires alternative approaches to circumvent performance problems. Aiming to provide scientists with a new tool to recover valuable information from such datasets, we developed a general purpose distributed memory parallel MRF-based image analysis framework (MPI-PMRF). MPI-PMRF overcomes performance and memory limitationsmore » by distributing data and computations across processors. The proposed approach was successfully tested with synthetic and experimental datasets. Additionally, the performance of the MPI-PMRF framework is analyzed through a detailed scalability study. We show that a performance increase is obtained while maintaining an accuracy of the segmentation results higher than 98%. The contributions of this paper are: (a) development of a distributed memory MRF framework; (b) measurement of the performance increase of the proposed approach; (c) verification of segmentation accuracy in both synthetic and experimental, real-world datasets« less

Scalable Triadic Analysis of Large-Scale Graphs: Multi-Core vs. Multi-Processor vs. Multi-Threaded Shared Memory Architectures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chin, George; Marquez, Andres; Choudhury, Sutanay

2012-09-01

Triadic analysis encompasses a useful set of graph mining methods that is centered on the concept of a triad, which is a subgraph of three nodes and the configuration of directed edges across the nodes. Such methods are often applied in the social sciences as well as many other diverse fields. Triadic methods commonly operate on a triad census that counts the number of triads of every possible edge configuration in a graph. Like other graph algorithms, triadic census algorithms do not scale well when graphs reach tens of millions to billions of nodes. To enable the triadic analysis ofmore » large-scale graphs, we developed and optimized a triad census algorithm to efficiently execute on shared memory architectures. We will retrace the development and evolution of a parallel triad census algorithm. Over the course of several versions, we continually adapted the code’s data structures and program logic to expose more opportunities to exploit parallelism on shared memory that would translate into improved computational performance. We will recall the critical steps and modifications that occurred during code development and optimization. Furthermore, we will compare the performances of triad census algorithm versions on three specific systems: Cray XMT, HP Superdome, and AMD multi-core NUMA machine. These three systems have shared memory architectures but with markedly different hardware capabilities to manage parallelism.« less

Parallel discrete event simulation using shared memory

NASA Technical Reports Server (NTRS)

Reed, Daniel A.; Malony, Allen D.; Mccredie, Bradley D.

1988-01-01

With traditional event-list techniques, evaluating a detailed discrete-event simulation-model can often require hours or even days of computation time. By eliminating the event list and maintaining only sufficient synchronization to ensure causality, parallel simulation can potentially provide speedups that are linear in the numbers of processors. A set of shared-memory experiments, using the Chandy-Misra distributed-simulation algorithm, to simulate networks of queues is presented. Parameters of the study include queueing network topology and routing probabilities, number of processors, and assignment of network nodes to processors. These experiments show that Chandy-Misra distributed simulation is a questionable alternative to sequential-simulation of most queueing network models.

A matrix-algebraic formulation of distributed-memory maximal cardinality matching algorithms in bipartite graphs

DOE PAGES

Azad, Ariful; Buluç, Aydın

2016-05-16

We describe parallel algorithms for computing maximal cardinality matching in a bipartite graph on distributed-memory systems. Unlike traditional algorithms that match one vertex at a time, our algorithms process many unmatched vertices simultaneously using a matrix-algebraic formulation of maximal matching. This generic matrix-algebraic framework is used to develop three efficient maximal matching algorithms with minimal changes. The newly developed algorithms have two benefits over existing graph-based algorithms. First, unlike existing parallel algorithms, cardinality of matching obtained by the new algorithms stays constant with increasing processor counts, which is important for predictable and reproducible performance. Second, relying on bulk-synchronous matrix operations,more » these algorithms expose a higher degree of parallelism on distributed-memory platforms than existing graph-based algorithms. We report high-performance implementations of three maximal matching algorithms using hybrid OpenMP-MPI and evaluate the performance of these algorithm using more than 35 real and randomly generated graphs. On real instances, our algorithms achieve up to 200 × speedup on 2048 cores of a Cray XC30 supercomputer. Even higher speedups are obtained on larger synthetically generated graphs where our algorithms show good scaling on up to 16,384 cores.« less

A hybrid dynamic harmony search algorithm for identical parallel machines scheduling

NASA Astrophysics Data System (ADS)

Chen, Jing; Pan, Quan-Ke; Wang, Ling; Li, Jun-Qing

2012-02-01

In this article, a dynamic harmony search (DHS) algorithm is proposed for the identical parallel machines scheduling problem with the objective to minimize makespan. First, an encoding scheme based on a list scheduling rule is developed to convert the continuous harmony vectors to discrete job assignments. Second, the whole harmony memory (HM) is divided into multiple small-sized sub-HMs, and each sub-HM performs evolution independently and exchanges information with others periodically by using a regrouping schedule. Third, a novel improvisation process is applied to generate a new harmony by making use of the information of harmony vectors in each sub-HM. Moreover, a local search strategy is presented and incorporated into the DHS algorithm to find promising solutions. Simulation results show that the hybrid DHS (DHS_LS) is very competitive in comparison to its competitors in terms of mean performance and average computational time.

Supporting 64-bit global indices in Epetra and other Trilinos packages :

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jhurani, Chetan; Austin, Travis M.; Heroux, Michael Allen

The Trilinos Project is an effort to facilitate the design, development, integration and ongoing support of mathematical software libraries within an object-oriented framework. It is intended for large-scale, complex multiphysics engineering and scientific applications [2, 4, 3]. Epetra is one of its basic packages. It provides serial and parallel linear algebra capabilities. Before Trilinos version 11.0, released in 2012, Epetra used the C++ int data-type for storing global and local indices for degrees of freedom (DOFs). Since int is typically 32-bit, this limited the largest problem size to be smaller than approximately two billion DOFs. This was true even ifmore » a distributed memory machine could handle larger problems. We have added optional support for C++ long long data-type, which is at least 64-bit wide, for global indices. To save memory, maintain the speed of memory-bound operations, and reduce further changes to the code, the local indices are still 32-bit. We document the changes required to achieve this feature and how the new functionality can be used. We also report on the lessons learned in modifying a mature and popular package from various perspectives design goals, backward compatibility, engineering decisions, C++ language features, effects on existing users and other packages, and build integration.« less

Memory Benchmarks for SMP-Based High Performance Parallel Computers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Yoo, A B; de Supinski, B; Mueller, F

2001-11-20

As the speed gap between CPU and main memory continues to grow, memory accesses increasingly dominates the performance of many applications. The problem is particularly acute for symmetric multiprocessor (SMP) systems, where the shared memory may be accessed concurrently by a group of threads running on separate CPUs. Unfortunately, several key issues governing memory system performance in current systems are not well understood. Complex interactions between the levels of the memory hierarchy, buses or switches, DRAM back-ends, system software, and application access patterns can make it difficult to pinpoint bottlenecks and determine appropriate optimizations, and the situation is even moremore » complex for SMP systems. To partially address this problem, we formulated a set of multi-threaded microbenchmarks for characterizing and measuring the performance of the underlying memory system in SMP-based high-performance computers. We report our use of these microbenchmarks on two important SMP-based machines. This paper has four primary contributions. First, we introduce a microbenchmark suite to systematically assess and compare the performance of different levels in SMP memory hierarchies. Second, we present a new tool based on hardware performance monitors to determine a wide array of memory system characteristics, such as cache sizes, quickly and easily; by using this tool, memory performance studies can be targeted to the full spectrum of performance regimes with many fewer data points than is otherwise required. Third, we present experimental results indicating that the performance of applications with large memory footprints remains largely constrained by memory. Fourth, we demonstrate that thread-level parallelism further degrades memory performance, even for the latest SMPs with hardware prefetching and switch-based memory interconnects.« less

Performance and Application of Parallel OVERFLOW Codes on Distributed and Shared Memory Platforms

NASA Technical Reports Server (NTRS)

Djomehri, M. Jahed; Rizk, Yehia M.

1999-01-01

The presentation discusses recent studies on the performance of the two parallel versions of the aerodynamics CFD code, OVERFLOW_MPI and _MLP. Developed at NASA Ames, the serial version, OVERFLOW, is a multidimensional Navier-Stokes flow solver based on overset (Chimera) grid technology. The code has recently been parallelized in two ways. One is based on the explicit message-passing interface (MPI) across processors and uses the _MPI communication package. This approach is primarily suited for distributed memory systems and workstation clusters. The second, termed the multi-level parallel (MLP) method, is simple and uses shared memory for all communications. The _MLP code is suitable on distributed-shared memory systems. For both methods, the message passing takes place across the processors or processes at the advancement of each time step. This procedure is, in effect, the Chimera boundary conditions update, which is done in an explicit "Jacobi" style. In contrast, the update in the serial code is done in more of the "Gauss-Sidel" fashion. The programming efforts for the _MPI code is more complicated than for the _MLP code; the former requires modification of the outer and some inner shells of the serial code, whereas the latter focuses only on the outer shell of the code. The _MPI version offers a great deal of flexibility in distributing grid zones across a specified number of processors in order to achieve load balancing. The approach is capable of partitioning zones across multiple processors or sending each zone and/or cluster of several zones into a single processor. The message passing across the processors consists of Chimera boundary and/or an overlap of "halo" boundary points for each partitioned zone. The MLP version is a new coarse-grain parallel concept at the zonal and intra-zonal levels. A grouping strategy is used to distribute zones into several groups forming sub-processes which will run in parallel. The total volume of grid points in each group are approximately balanced. A proper number of threads are initially allocated to each group, and in subsequent iterations during the run-time, the number of threads are adjusted to achieve load balancing across the processes. Each process exploits the multitasking directives already established in Overflow.

Gilgamesh: A Multithreaded Processor-In-Memory Architecture for Petaflops Computing

NASA Technical Reports Server (NTRS)

Sterling, T. L.; Zima, H. P.

2002-01-01

Processor-in-Memory (PIM) architectures avoid the von Neumann bottleneck in conventional machines by integrating high-density DRAM and CMOS logic on the same chip. Parallel systems based on this new technology are expected to provide higher scalability, adaptability, robustness, fault tolerance and lower power consumption than current MPPs or commodity clusters. In this paper we describe the design of Gilgamesh, a PIM-based massively parallel architecture, and elements of its execution model. Gilgamesh extends existing PIM capabilities by incorporating advanced mechanisms for virtualizing tasks and data and providing adaptive resource management for load balancing and latency tolerance. The Gilgamesh execution model is based on macroservers, a middleware layer which supports object-based runtime management of data and threads allowing explicit and dynamic control of locality and load balancing. The paper concludes with a discussion of related research activities and an outlook to future work.

Programming in Vienna Fortran

NASA Technical Reports Server (NTRS)

Chapman, Barbara; Mehrotra, Piyush; Zima, Hans

1992-01-01

Exploiting the full performance potential of distributed memory machines requires a careful distribution of data across the processors. Vienna Fortran is a language extension of Fortran which provides the user with a wide range of facilities for such mapping of data structures. In contrast to current programming practice, programs in Vienna Fortran are written using global data references. Thus, the user has the advantages of a shared memory programming paradigm while explicitly controlling the data distribution. In this paper, we present the language features of Vienna Fortran for FORTRAN 77, together with examples illustrating the use of these features.

Center for Parallel Optimization

DTIC Science & Technology

1993-09-30

BOLLING AFB DC 20332-0001 _ii _ 11. SUPPLEMENTARY NOTES 12a. DISTRIBUTION/ AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE APPROVED FOR PUBLIC RELEASE...Machines Corporation, March 16-19, 1993 , A Branch- and-Bound Method for Mixed Integer Programming on the CM-.5 "* Dr. Roberto Musmanno, University of

Application of an onboard processor to the OAO C spacecraft

NASA Technical Reports Server (NTRS)

Stewart, W. N.; Hartenstein, R. G.; Trevathan, C.

1972-01-01

The design of a stored program computer for spacecraft use and its application on the fourth Orbiting Astronomical Observatory (OAO) is reported. The computer is a medium scale, parallel machine with a memory capacity of 16384 words of 18 bits each. It possesses a comprehensive instruction repertoire and operates on 45 W of power (including the dc-to-dc converter). The machine operates at a 500-kHz rate and executes an add instruction in 10 microseconds. Its primary functions on OAO C will be auxiliary command storage, spacecraft monitoring and malfunction reporting, data compression and status summary, and possible performance of emergency corrective action for certain anomalous situations.

Expressing Parallelism with ROOT

NASA Astrophysics Data System (ADS)

Piparo, D.; Tejedor, E.; Guiraud, E.; Ganis, G.; Mato, P.; Moneta, L.; Valls Pla, X.; Canal, P.

2017-10-01

The need for processing the ever-increasing amount of data generated by the LHC experiments in a more efficient way has motivated ROOT to further develop its support for parallelism. Such support is being tackled both for shared-memory and distributed-memory environments. The incarnations of the aforementioned parallelism are multi-threading, multi-processing and cluster-wide executions. In the area of multi-threading, we discuss the new implicit parallelism and related interfaces, as well as the new building blocks to safely operate with ROOT objects in a multi-threaded environment. Regarding multi-processing, we review the new MultiProc framework, comparing it with similar tools (e.g. multiprocessing module in Python). Finally, as an alternative to PROOF for cluster-wide executions, we introduce the efforts on integrating ROOT with state-of-the-art distributed data processing technologies like Spark, both in terms of programming model and runtime design (with EOS as one of the main components). For all the levels of parallelism, we discuss, based on real-life examples and measurements, how our proposals can increase the productivity of scientists.

Expressing Parallelism with ROOT

DOE Office of Scientific and Technical Information (OSTI.GOV)

Piparo, D.; Tejedor, E.; Guiraud, E.

The need for processing the ever-increasing amount of data generated by the LHC experiments in a more efficient way has motivated ROOT to further develop its support for parallelism. Such support is being tackled both for shared-memory and distributed-memory environments. The incarnations of the aforementioned parallelism are multi-threading, multi-processing and cluster-wide executions. In the area of multi-threading, we discuss the new implicit parallelism and related interfaces, as well as the new building blocks to safely operate with ROOT objects in a multi-threaded environment. Regarding multi-processing, we review the new MultiProc framework, comparing it with similar tools (e.g. multiprocessing module inmore » Python). Finally, as an alternative to PROOF for cluster-wide executions, we introduce the efforts on integrating ROOT with state-of-the-art distributed data processing technologies like Spark, both in terms of programming model and runtime design (with EOS as one of the main components). For all the levels of parallelism, we discuss, based on real-life examples and measurements, how our proposals can increase the productivity of scientists.« less

Running ATLAS workloads within massively parallel distributed applications using Athena Multi-Process framework (AthenaMP)

NASA Astrophysics Data System (ADS)

Calafiura, Paolo; Leggett, Charles; Seuster, Rolf; Tsulaia, Vakhtang; Van Gemmeren, Peter

2015-12-01

AthenaMP is a multi-process version of the ATLAS reconstruction, simulation and data analysis framework Athena. By leveraging Linux fork and copy-on-write mechanisms, it allows for sharing of memory pages between event processors running on the same compute node with little to no change in the application code. Originally targeted to optimize the memory footprint of reconstruction jobs, AthenaMP has demonstrated that it can reduce the memory usage of certain configurations of ATLAS production jobs by a factor of 2. AthenaMP has also evolved to become the parallel event-processing core of the recently developed ATLAS infrastructure for fine-grained event processing (Event Service) which allows the running of AthenaMP inside massively parallel distributed applications on hundreds of compute nodes simultaneously. We present the architecture of AthenaMP, various strategies implemented by AthenaMP for scheduling workload to worker processes (for example: Shared Event Queue and Shared Distributor of Event Tokens) and the usage of AthenaMP in the diversity of ATLAS event processing workloads on various computing resources: Grid, opportunistic resources and HPC.

Integrating Cache Performance Modeling and Tuning Support in Parallelization Tools

NASA Technical Reports Server (NTRS)

Waheed, Abdul; Yan, Jerry; Saini, Subhash (Technical Monitor)

1998-01-01

With the resurgence of distributed shared memory (DSM) systems based on cache-coherent Non Uniform Memory Access (ccNUMA) architectures and increasing disparity between memory and processors speeds, data locality overheads are becoming the greatest bottlenecks in the way of realizing potential high performance of these systems. While parallelization tools and compilers facilitate the users in porting their sequential applications to a DSM system, a lot of time and effort is needed to tune the memory performance of these applications to achieve reasonable speedup. In this paper, we show that integrating cache performance modeling and tuning support within a parallelization environment can alleviate this problem. The Cache Performance Modeling and Prediction Tool (CPMP), employs trace-driven simulation techniques without the overhead of generating and managing detailed address traces. CPMP predicts the cache performance impact of source code level "what-if" modifications in a program to assist a user in the tuning process. CPMP is built on top of a customized version of the Computer Aided Parallelization Tools (CAPTools) environment. Finally, we demonstrate how CPMP can be applied to tune a real Computational Fluid Dynamics (CFD) application.

Finite element computation on nearest neighbor connected machines

NASA Technical Reports Server (NTRS)

Mcaulay, A. D.

1984-01-01

Research aimed at faster, more cost effective parallel machines and algorithms for improving designer productivity with finite element computations is discussed. A set of 8 boards, containing 4 nearest neighbor connected arrays of commercially available floating point chips and substantial memory, are inserted into a commercially available machine. One-tenth Mflop (64 bit operation) processors provide an 89% efficiency when solving the equations arising in a finite element problem for a single variable regular grid of size 40 by 40 by 40. This is approximately 15 to 20 times faster than a much more expensive machine such as a VAX 11/780 used in double precision. The efficiency falls off as faster or more processors are envisaged because communication times become dominant. A novel successive overrelaxation algorithm which uses cyclic reduction in order to permit data transfer and computation to overlap in time is proposed.

Machine Learning Toolkit for Extreme Scale

DOE Office of Scientific and Technical Information (OSTI.GOV)

2014-03-31

Support Vector Machines (SVM) is a popular machine learning technique, which has been applied to a wide range of domains such as science, finance, and social networks for supervised learning. MaTEx undertakes the challenge of designing a scalable parallel SVM training algorithm for large scale systems, which includes commodity multi-core machines, tightly connected supercomputers and cloud computing systems. Several techniques are proposed for improved speed and memory space usage including adaptive and aggressive elimination of samples for faster convergence , and sparse format representation of data samples. Several heuristics for earliest possible to lazy elimination of non-contributing samples are consideredmore » in MaTEx. In many cases, where an early sample elimination might result in a false positive, low overhead mechanisms for reconstruction of key data structures are proposed. The proposed algorithm and heuristics are implemented and evaluated on various publicly available datasets« less

Rock sampling. [method for controlling particle size distribution

NASA Technical Reports Server (NTRS)

Blum, P. (Inventor)

1971-01-01

A method for sampling rock and other brittle materials and for controlling resultant particle sizes is described. The method involves cutting grooves in the rock surface to provide a grouping of parallel ridges and subsequently machining the ridges to provide a powder specimen. The machining step may comprise milling, drilling, lathe cutting or the like; but a planing step is advantageous. Control of the particle size distribution is effected primarily by changing the height and width of these ridges. This control exceeds that obtainable by conventional grinding.

Parallel and distributed computation for fault-tolerant object recognition

NASA Technical Reports Server (NTRS)

Wechsler, Harry

1988-01-01

The distributed associative memory (DAM) model is suggested for distributed and fault-tolerant computation as it relates to object recognition tasks. The fault-tolerance is with respect to geometrical distortions (scale and rotation), noisy inputs, occulsion/overlap, and memory faults. An experimental system was developed for fault-tolerant structure recognition which shows the feasibility of such an approach. The approach is futher extended to the problem of multisensory data integration and applied successfully to the recognition of colored polyhedral objects.

Kanerva's sparse distributed memory with multiple hamming thresholds

NASA Technical Reports Server (NTRS)

Pohja, Seppo; Kaski, Kimmo

1992-01-01

If the stored input patterns of Kanerva's Sparse Distributed Memory (SDM) are highly correlated, utilization of the storage capacity is very low compared to the case of uniformly distributed random input patterns. We consider a variation of SDM that has a better storage capacity utilization for correlated input patterns. This approach uses a separate selection threshold for each physical storage address or hard location. The selection of the hard locations for reading or writing can be done in parallel of which SDM implementations can benefit.

An efficient 3-dim FFT for plane wave electronic structure calculations on massively parallel machines composed of multiprocessor nodes

NASA Astrophysics Data System (ADS)

Goedecker, Stefan; Boulet, Mireille; Deutsch, Thierry

2003-08-01

Three-dimensional Fast Fourier Transforms (FFTs) are the main computational task in plane wave electronic structure calculations. Obtaining a high performance on a large numbers of processors is non-trivial on the latest generation of parallel computers that consist of nodes made up of a shared memory multiprocessors. A non-dogmatic method for obtaining high performance for such 3-dim FFTs in a combined MPI/OpenMP programming paradigm will be presented. Exploiting the peculiarities of plane wave electronic structure calculations, speedups of up to 160 and speeds of up to 130 Gflops were obtained on 256 processors.

Scaling Irregular Applications through Data Aggregation and Software Multithreading

DOE Office of Scientific and Technical Information (OSTI.GOV)

Morari, Alessandro; Tumeo, Antonino; Chavarría-Miranda, Daniel

Bioinformatics, data analytics, semantic databases, knowledge discovery are emerging high performance application areas that exploit dynamic, linked data structures such as graphs, unbalanced trees or unstructured grids. These data structures usually are very large, requiring significantly more memory than available on single shared memory systems. Additionally, these data structures are difficult to partition on distributed memory systems. They also present poor spatial and temporal locality, thus generating unpredictable memory and network accesses. The Partitioned Global Address Space (PGAS) programming model seems suitable for these applications, because it allows using a shared memory abstraction across distributed-memory clusters. However, current PGAS languagesmore » and libraries are built to target regular remote data accesses and block transfers. Furthermore, they usually rely on the Single Program Multiple Data (SPMD) parallel control model, which is not well suited to the fine grained, dynamic and unbalanced parallelism of irregular applications. In this paper we present {\\bf GMT} (Global Memory and Threading library), a custom runtime library that enables efficient execution of irregular applications on commodity clusters. GMT integrates a PGAS data substrate with simple fork/join parallelism and provides automatic load balancing on a per node basis. It implements multi-level aggregation and lightweight multithreading to maximize memory and network bandwidth with fine-grained data accesses and tolerate long data access latencies. A key innovation in the GMT runtime is its thread specialization (workers, helpers and communication threads) that realize the overall functionality. We compare our approach with other PGAS models, such as UPC running using GASNet, and hand-optimized MPI code on a set of typical large-scale irregular applications, demonstrating speedups of an order of magnitude.« less

Locality Aware Concurrent Start for Stencil Applications

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shrestha, Sunil; Gao, Guang R.; Manzano Franco, Joseph B.

Stencil computations are at the heart of many physical simulations used in scientific codes. Thus, there exists a plethora of optimization efforts for this family of computations. Among these techniques, tiling techniques that allow concurrent start have proven to be very efficient in providing better performance for these critical kernels. Nevertheless, with many core designs being the norm, these optimization techniques might not be able to fully exploit locality (both spatial and temporal) on multiple levels of the memory hierarchy without compromising parallelism. It is no longer true that the machine can be seen as a homogeneous collection of nodesmore » with caches, main memory and an interconnect network. New architectural designs exhibit complex grouping of nodes, cores, threads, caches and memory connected by an ever evolving network-on-chip design. These new designs may benefit greatly from carefully crafted schedules and groupings that encourage parallel actors (i.e. threads, cores or nodes) to be aware of the computational history of other actors in close proximity. In this paper, we provide an efficient tiling technique that allows hierarchical concurrent start for memory hierarchy aware tile groups. Each execution schedule and tile shape exploit the available parallelism, load balance and locality present in the given applications. We demonstrate our technique on the Intel Xeon Phi architecture with selected and representative stencil kernels. We show improvement ranging from 5.58% to 31.17% over existing state-of-the-art techniques.« less

Architectural requirements for the Red Storm computing system.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Camp, William J.; Tomkins, James Lee

This report is based on the Statement of Work (SOW) describing the various requirements for delivering 3 new supercomputer system to Sandia National Laboratories (Sandia) as part of the Department of Energy's (DOE) Accelerated Strategic Computing Initiative (ASCI) program. This system is named Red Storm and will be a distributed memory, massively parallel processor (MPP) machine built primarily out of commodity parts. The requirements presented here distill extensive architectural and design experience accumulated over a decade and a half of research, development and production operation of similar machines at Sandia. Red Storm will have an unusually high bandwidth, low latencymore » interconnect, specially designed hardware and software reliability features, a light weight kernel compute node operating system and the ability to rapidly switch major sections of the machine between classified and unclassified computing environments. Particular attention has been paid to architectural balance in the design of Red Storm, and it is therefore expected to achieve an atypically high fraction of its peak speed of 41 TeraOPS on real scientific computing applications. In addition, Red Storm is designed to be upgradeable to many times this initial peak capability while still retaining appropriate balance in key design dimensions. Installation of the Red Storm computer system at Sandia's New Mexico site is planned for 2004, and it is expected that the system will be operated for a minimum of five years following installation.« less

Dopaminergic neurons write and update memories with cell-type-specific rules

PubMed Central

Aso, Yoshinori; Rubin, Gerald M

2016-01-01

Associative learning is thought to involve parallel and distributed mechanisms of memory formation and storage. In Drosophila, the mushroom body (MB) is the major site of associative odor memory formation. Previously we described the anatomy of the adult MB and defined 20 types of dopaminergic neurons (DANs) that each innervate distinct MB compartments (Aso et al., 2014a, 2014b). Here we compare the properties of memories formed by optogenetic activation of individual DAN cell types. We found extensive differences in training requirements for memory formation, decay dynamics, storage capacity and flexibility to learn new associations. Even a single DAN cell type can either write or reduce an aversive memory, or write an appetitive memory, depending on when it is activated relative to odor delivery. Our results show that different learning rules are executed in seemingly parallel memory systems, providing multiple distinct circuit-based strategies to predict future events from past experiences. DOI: http://dx.doi.org/10.7554/eLife.16135.001 PMID:27441388

Computer program MCAP-TOSS calculates steady-state fluid dynamics of coolant in parallel channels and temperature distribution in surrounding heat-generating solid

NASA Technical Reports Server (NTRS)

Lee, A. Y.

1967-01-01

Computer program calculates the steady state fluid distribution, temperature rise, and pressure drop of a coolant, the material temperature distribution of a heat generating solid, and the heat flux distributions at the fluid-solid interfaces. It performs the necessary iterations automatically within the computer, in one machine run.

Distributed intelligence for supervisory control

NASA Technical Reports Server (NTRS)

Wolfe, W. J.; Raney, S. D.

1987-01-01

Supervisory control systems must deal with various types of intelligence distributed throughout the layers of control. Typical layers are real-time servo control, off-line planning and reasoning subsystems and finally, the human operator. Design methodologies must account for the fact that the majority of the intelligence will reside with the human operator. Hierarchical decompositions and feedback loops as conceptual building blocks that provide a common ground for man-machine interaction are discussed. Examples of types of parallelism and parallel implementation on several classes of computer architecture are also discussed.

Network Model of Decreased Context Utilization in Autism Spectrum Disorder

ERIC Educational Resources Information Center

Beversdorf, David Q.; Narayanan, Ananth; Hillier, Ashleigh; Hughes, John D.

2007-01-01

Individuals with autism spectrum disorders (ASD) demonstrate impaired utilization of context, which allows for superior performance on the "false memory" task. We report the application of a simplified parallel distributed processing model of context utilization to the false memory task. For individuals without ASD, experiments support a model…

Virtual Machine Language 2.1

NASA Technical Reports Server (NTRS)

Riedel, Joseph E.; Grasso, Christopher A.

2012-01-01

VML (Virtual Machine Language) is an advanced computing environment that allows spacecraft to operate using mechanisms ranging from simple, time-oriented sequencing to advanced, multicomponent reactive systems. VML has developed in four evolutionary stages. VML 0 is a core execution capability providing multi-threaded command execution, integer data types, and rudimentary branching. VML 1 added named parameterized procedures, extensive polymorphism, data typing, branching, looping issuance of commands using run-time parameters, and named global variables. VML 2 added for loops, data verification, telemetry reaction, and an open flight adaptation architecture. VML 2.1 contains major advances in control flow capabilities for executable state machines. On the resource requirements front, VML 2.1 features a reduced memory footprint in order to fit more capability into modestly sized flight processors, and endian-neutral data access for compatibility with Intel little-endian processors. Sequence packaging has been improved with object-oriented programming constructs and the use of implicit (rather than explicit) time tags on statements. Sequence event detection has been significantly enhanced with multi-variable waiting, which allows a sequence to detect and react to conditions defined by complex expressions with multiple global variables. This multi-variable waiting serves as the basis for implementing parallel rule checking, which in turn, makes possible executable state machines. The new state machine feature in VML 2.1 allows the creation of sophisticated autonomous reactive systems without the need to develop expensive flight software. Users specify named states and transitions, along with the truth conditions required, before taking transitions. Transitions with the same signal name allow separate state machines to coordinate actions: the conditions distributed across all state machines necessary to arm a particular signal are evaluated, and once found true, that signal is raised. The selected signal then causes all identically named transitions in all present state machines to be taken simultaneously. VML 2.1 has relevance to all potential space missions, both manned and unmanned. It was under consideration for use on Orion.

Thread-Level Parallelization and Optimization of NWChem for the Intel MIC Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shan, Hongzhang; Williams, Samuel; Jong, Wibe de

In the multicore era it was possible to exploit the increase in on-chip parallelism by simply running multiple MPI processes per chip. Unfortunately, manycore processors' greatly increased thread- and data-level parallelism coupled with a reduced memory capacity demand an altogether different approach. In this paper we explore augmenting two NWChem modules, triples correction of the CCSD(T) and Fock matrix construction, with OpenMP in order that they might run efficiently on future manycore architectures. As the next NERSC machine will be a self-hosted Intel MIC (Xeon Phi) based supercomputer, we leverage an existing MIC testbed at NERSC to evaluate our experiments.more » In order to proxy the fact that future MIC machines will not have a host processor, we run all of our experiments in tt native mode. We found that while straightforward application of OpenMP to the deep loop nests associated with the tensor contractions of CCSD(T) was sufficient in attaining high performance, significant effort was required to safely and efficiently thread the TEXAS integral package when constructing the Fock matrix. Ultimately, our new MPI OpenMP hybrid implementations attain up to 65x better performance for the triples part of the CCSD(T) due in large part to the fact that the limited on-card memory limits the existing MPI implementation to a single process per card. Additionally, we obtain up to 1.6x better performance on Fock matrix constructions when compared with the best MPI implementations running multiple processes per card.« less

Thread-level parallelization and optimization of NWChem for the Intel MIC architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Shan, Hongzhang; Williams, Samuel; de Jong, Wibe

In the multicore era it was possible to exploit the increase in on-chip parallelism by simply running multiple MPI processes per chip. Unfortunately, manycore processors' greatly increased thread- and data-level parallelism coupled with a reduced memory capacity demand an altogether different approach. In this paper we explore augmenting two NWChem modules, triples correction of the CCSD(T) and Fock matrix construction, with OpenMP in order that they might run efficiently on future manycore architectures. As the next NERSC machine will be a self-hosted Intel MIC (Xeon Phi) based supercomputer, we leverage an existing MIC testbed at NERSC to evaluate our experiments.more » In order to proxy the fact that future MIC machines will not have a host processor, we run all of our experiments in native mode. We found that while straightforward application of OpenMP to the deep loop nests associated with the tensor contractions of CCSD(T) was sufficient in attaining high performance, significant e ort was required to safely and efeciently thread the TEXAS integral package when constructing the Fock matrix. Ultimately, our new MPI+OpenMP hybrid implementations attain up to 65× better performance for the triples part of the CCSD(T) due in large part to the fact that the limited on-card memory limits the existing MPI implementation to a single process per card. Additionally, we obtain up to 1.6× better performance on Fock matrix constructions when compared with the best MPI implementations running multiple processes per card.« less

NavP: Structured and Multithreaded Distributed Parallel Programming

NASA Technical Reports Server (NTRS)

Pan, Lei

2007-01-01

We present Navigational Programming (NavP) -- a distributed parallel programming methodology based on the principles of migrating computations and multithreading. The four major steps of NavP are: (1) Distribute the data using the data communication pattern in a given algorithm; (2) Insert navigational commands for the computation to migrate and follow large-sized distributed data; (3) Cut the sequential migrating thread and construct a mobile pipeline; and (4) Loop back for refinement. NavP is significantly different from the current prevailing Message Passing (MP) approach. The advantages of NavP include: (1) NavP is structured distributed programming and it does not change the code structure of an original algorithm. This is in sharp contrast to MP as MP implementations in general do not resemble the original sequential code; (2) NavP implementations are always competitive with the best MPI implementations in terms of performance. Approaches such as DSM or HPF have failed to deliver satisfying performance as of today in contrast, even if they are relatively easy to use compared to MP; (3) NavP provides incremental parallelization, which is beyond the reach of MP; and (4) NavP is a unifying approach that allows us to exploit both fine- (multithreading on shared memory) and coarse- (pipelined tasks on distributed memory) grained parallelism. This is in contrast to the currently popular hybrid use of MP+OpenMP, which is known to be complex to use. We present experimental results that demonstrate the effectiveness of NavP.

VINE-A NUMERICAL CODE FOR SIMULATING ASTROPHYSICAL SYSTEMS USING PARTICLES. II. IMPLEMENTATION AND PERFORMANCE CHARACTERISTICS

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nelson, Andrew F.; Wetzstein, M.; Naab, T.

2009-10-01

We continue our presentation of VINE. In this paper, we begin with a description of relevant architectural properties of the serial and shared memory parallel computers on which VINE is intended to run, and describe their influences on the design of the code itself. We continue with a detailed description of a number of optimizations made to the layout of the particle data in memory and to our implementation of a binary tree used to access that data for use in gravitational force calculations and searches for smoothed particle hydrodynamics (SPH) neighbor particles. We describe the modifications to the codemore » necessary to obtain forces efficiently from special purpose 'GRAPE' hardware, the interfaces required to allow transparent substitution of those forces in the code instead of those obtained from the tree, and the modifications necessary to use both tree and GRAPE together as a fused GRAPE/tree combination. We conclude with an extensive series of performance tests, which demonstrate that the code can be run efficiently and without modification in serial on small workstations or in parallel using the OpenMP compiler directives on large-scale, shared memory parallel machines. We analyze the effects of the code optimizations and estimate that they improve its overall performance by more than an order of magnitude over that obtained by many other tree codes. Scaled parallel performance of the gravity and SPH calculations, together the most costly components of most simulations, is nearly linear up to at least 120 processors on moderate sized test problems using the Origin 3000 architecture, and to the maximum machine sizes available to us on several other architectures. At similar accuracy, performance of VINE, used in GRAPE-tree mode, is approximately a factor 2 slower than that of VINE, used in host-only mode. Further optimizations of the GRAPE/host communications could improve the speed by as much as a factor of 3, but have not yet been implemented in VINE. Finally, we find that although parallel performance on small problems may reach a plateau beyond which more processors bring no additional speedup, performance never decreases, a factor important for running large simulations on many processors with individual time steps, where only a small fraction of the total particles require updates at any given moment.« less

What Multilevel Parallel Programs do when you are not Watching: A Performance Analysis Case Study Comparing MPI/OpenMP, MLP, and Nested OpenMP

NASA Technical Reports Server (NTRS)

Jost, Gabriele; Labarta, Jesus; Gimenez, Judit

2004-01-01

With the current trend in parallel computer architectures towards clusters of shared memory symmetric multi-processors, parallel programming techniques have evolved that support parallelism beyond a single level. When comparing the performance of applications based on different programming paradigms, it is important to differentiate between the influence of the programming model itself and other factors, such as implementation specific behavior of the operating system (OS) or architectural issues. Rewriting-a large scientific application in order to employ a new programming paradigms is usually a time consuming and error prone task. Before embarking on such an endeavor it is important to determine that there is really a gain that would not be possible with the current implementation. A detailed performance analysis is crucial to clarify these issues. The multilevel programming paradigms considered in this study are hybrid MPI/OpenMP, MLP, and nested OpenMP. The hybrid MPI/OpenMP approach is based on using MPI [7] for the coarse grained parallelization and OpenMP [9] for fine grained loop level parallelism. The MPI programming paradigm assumes a private address space for each process. Data is transferred by explicitly exchanging messages via calls to the MPI library. This model was originally designed for distributed memory architectures but is also suitable for shared memory systems. The second paradigm under consideration is MLP which was developed by Taft. The approach is similar to MPi/OpenMP, using a mix of coarse grain process level parallelization and loop level OpenMP parallelization. As it is the case with MPI, a private address space is assumed for each process. The MLP approach was developed for ccNUMA architectures and explicitly takes advantage of the availability of shared memory. A shared memory arena which is accessible by all processes is required. Communication is done by reading from and writing to the shared memory.

Evaluating the performance of the particle finite element method in parallel architectures

NASA Astrophysics Data System (ADS)

Gimenez, Juan M.; Nigro, Norberto M.; Idelsohn, Sergio R.

2014-05-01

This paper presents a high performance implementation for the particle-mesh based method called particle finite element method two (PFEM-2). It consists of a material derivative based formulation of the equations with a hybrid spatial discretization which uses an Eulerian mesh and Lagrangian particles. The main aim of PFEM-2 is to solve transport equations as fast as possible keeping some level of accuracy. The method was found to be competitive with classical Eulerian alternatives for these targets, even in their range of optimal application. To evaluate the goodness of the method with large simulations, it is imperative to use of parallel environments. Parallel strategies for Finite Element Method have been widely studied and many libraries can be used to solve Eulerian stages of PFEM-2. However, Lagrangian stages, such as streamline integration, must be developed considering the parallel strategy selected. The main drawback of PFEM-2 is the large amount of memory needed, which limits its application to large problems with only one computer. Therefore, a distributed-memory implementation is urgently needed. Unlike a shared-memory approach, using domain decomposition the memory is automatically isolated, thus avoiding race conditions; however new issues appear due to data distribution over the processes. Thus, a domain decomposition strategy for both particle and mesh is adopted, which minimizes the communication between processes. Finally, performance analysis running over multicore and multinode architectures are presented. The Courant-Friedrichs-Lewy number used influences the efficiency of the parallelization and, in some cases, a weighted partitioning can be used to improve the speed-up. However the total cputime for cases presented is lower than that obtained when using classical Eulerian strategies.

Efficient iteration in data-parallel programs with irregular and dynamically distributed data structures

DOE Office of Scientific and Technical Information (OSTI.GOV)

Littlefield, R.J.

1990-02-01

To implement an efficient data-parallel program on a non-shared memory MIMD multicomputer, data and computations must be properly partitioned to achieve good load balance and locality of reference. Programs with irregular data reference patterns often require irregular partitions. Although good partitions may be easy to determine, they can be difficult or impossible to implement in programming languages that provide only regular data distributions, such as blocked or cyclic arrays. We are developing Onyx, a programming system that provides a shared memory model of distributed data structures and extends the concept of data distribution to include irregular and dynamic distributions. Thismore » provides a powerful means to specify irregular partitions. Perhaps surprisingly, programs using it can also execute efficiently. In this paper, we describe and evaluate the Onyx implementation of a model problem that repeatedly executes an irregular but fixed data reference pattern. On an NCUBE hypercube, the speed of the Onyx implementation is comparable to that of carefully handwritten message-passing code.« less

A communication-avoiding, hybrid-parallel, rank-revealing orthogonalization method.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hoemmen, Mark

2010-11-01

Orthogonalization consumes much of the run time of many iterative methods for solving sparse linear systems and eigenvalue problems. Commonly used algorithms, such as variants of Gram-Schmidt or Householder QR, have performance dominated by communication. Here, 'communication' includes both data movement between the CPU and memory, and messages between processors in parallel. Our Tall Skinny QR (TSQR) family of algorithms requires asymptotically fewer messages between processors and data movement between CPU and memory than typical orthogonalization methods, yet achieves the same accuracy as Householder QR factorization. Furthermore, in block orthogonalizations, TSQR is faster and more accurate than existing approaches formore » orthogonalizing the vectors within each block ('normalization'). TSQR's rank-revealing capability also makes it useful for detecting deflation in block iterative methods, for which existing approaches sacrifice performance, accuracy, or both. We have implemented a version of TSQR that exploits both distributed-memory and shared-memory parallelism, and supports real and complex arithmetic. Our implementation is optimized for the case of orthogonalizing a small number (5-20) of very long vectors. The shared-memory parallel component uses Intel's Threading Building Blocks, though its modular design supports other shared-memory programming models as well, including computation on the GPU. Our implementation achieves speedups of 2 times or more over competing orthogonalizations. It is available now in the development branch of the Trilinos software package, and will be included in the 10.8 release.« less

Parallel State Space Construction for a Model Checking Based on Maximality Semantics

NASA Astrophysics Data System (ADS)

El Abidine Bouneb, Zine; Saīdouni, Djamel Eddine

2009-03-01

The main limiting factor of the model checker integrated in the concurrency verification environment FOCOVE [1, 2], which use the maximality based labeled transition system (noted MLTS) as a true concurrency model[3, 4], is currently the amount of available physical memory. Many techniques have been developed to reduce the size of a state space. An interesting technique among them is the alpha equivalence reduction. Distributed memory execution environment offers yet another choice. The main contribution of the paper is to show that the parallel state space construction algorithm proposed in [5], which is based on interleaving semantics using LTS as semantic model, may be adapted easily to the distributed implementation of the alpha equivalence reduction for the maximality based labeled transition systems.

PELEC

DOE Office of Scientific and Technical Information (OSTI.GOV)

2017-05-17

PeleC is an adaptive-mesh compressible hydrodynamics code for reacting flows. It solves the compressible Navier-Stokes with multispecies transport in a block structured framework. The resulting algorithm is well suited for flows with localized resolution requirements and robust to discontinuities. User controllable refinement crieteria has the potential to result in extremely small numerical dissipation and dispersion, making this code appropriate for both research and applied usage. The code is built on the AMReX library which facilitates hierarchical parallelism and manages distributed memory parallism. PeleC algorithms are implemented to express shared memory parallelism.

A Massively Parallel Code for Polarization Calculations

NASA Astrophysics Data System (ADS)

Akiyama, Shizuka; Höflich, Peter

2001-03-01

We present an implementation of our Monte-Carlo radiation transport method for rapidly expanding, NLTE atmospheres for massively parallel computers which utilizes both the distributed and shared memory models. This allows us to take full advantage of the fast communication and low latency inherent to nodes with multiple CPUs, and to stretch the limits of scalability with the number of nodes compared to a version which is based on the shared memory model. Test calculations on a local 20-node Beowulf cluster with dual CPUs showed an improved scalability by about 40%.

A multiarchitecture parallel-processing development environment

NASA Technical Reports Server (NTRS)

Townsend, Scott; Blech, Richard; Cole, Gary

1993-01-01

A description is given of the hardware and software of a multiprocessor test bed - the second generation Hypercluster system. The Hypercluster architecture consists of a standard hypercube distributed-memory topology, with multiprocessor shared-memory nodes. By using standard, off-the-shelf hardware, the system can be upgraded to use rapidly improving computer technology. The Hypercluster's multiarchitecture nature makes it suitable for researching parallel algorithms in computational field simulation applications (e.g., computational fluid dynamics). The dedicated test-bed environment of the Hypercluster and its custom-built software allows experiments with various parallel-processing concepts such as message passing algorithms, debugging tools, and computational 'steering'. Such research would be difficult, if not impossible, to achieve on shared, commercial systems.

Thermodynamic Model of Spatial Memory

NASA Astrophysics Data System (ADS)

Kaufman, Miron; Allen, P.

1998-03-01

We develop and test a thermodynamic model of spatial memory. Our model is an application of statistical thermodynamics to cognitive science. It is related to applications of the statistical mechanics framework in parallel distributed processes research. Our macroscopic model allows us to evaluate an entropy associated with spatial memory tasks. We find that older adults exhibit higher levels of entropy than younger adults. Thurstone's Law of Categorical Judgment, according to which the discriminal processes along the psychological continuum produced by presentations of a single stimulus are normally distributed, is explained by using a Hooke spring model of spatial memory. We have also analyzed a nonlinear modification of the ideal spring model of spatial memory. This work is supported by NIH/NIA grant AG09282-06.

Users manual for the Chameleon parallel programming tools

DOE Office of Scientific and Technical Information (OSTI.GOV)

Gropp, W.; Smith, B.

1993-06-01

Message passing is a common method for writing programs for distributed-memory parallel computers. Unfortunately, the lack of a standard for message passing has hampered the construction of portable and efficient parallel programs. In an attempt to remedy this problem, a number of groups have developed their own message-passing systems, each with its own strengths and weaknesses. Chameleon is a second-generation system of this type. Rather than replacing these existing systems, Chameleon is meant to supplement them by providing a uniform way to access many of these systems. Chameleon`s goals are to (a) be very lightweight (low over-head), (b) be highlymore » portable, and (c) help standardize program startup and the use of emerging message-passing operations such as collective operations on subsets of processors. Chameleon also provides a way to port programs written using PICL or Intel NX message passing to other systems, including collections of workstations. Chameleon is tracking the Message-Passing Interface (MPI) draft standard and will provide both an MPI implementation and an MPI transport layer. Chameleon provides support for heterogeneous computing by using p4 and PVM. Chameleon`s support for homogeneous computing includes the portable libraries p4, PICL, and PVM and vendor-specific implementation for Intel NX, IBM EUI (SP-1), and Thinking Machines CMMD (CM-5). Support for Ncube and PVM 3.x is also under development.« less

Monitoring Data-Structure Evolution in Distributed Message-Passing Programs

NASA Technical Reports Server (NTRS)

Sarukkai, Sekhar R.; Beers, Andrew; Woodrow, Thomas S. (Technical Monitor)

1996-01-01

Monitoring the evolution of data structures in parallel and distributed programs, is critical for debugging its semantics and performance. However, the current state-of-art in tracking and presenting data-structure information on parallel and distributed environments is cumbersome and does not scale. In this paper we present a methodology that automatically tracks memory bindings (not the actual contents) of static and dynamic data-structures of message-passing C programs, using PVM. With the help of a number of examples we show that in addition to determining the impact of memory allocation overheads on program performance, graphical views can help in debugging the semantics of program execution. Scalable animations of virtual address bindings of source-level data-structures are used for debugging the semantics of parallel programs across all processors. In conjunction with light-weight core-files, this technique can be used to complement traditional debuggers on single processors. Detailed information (such as data-structure contents), on specific nodes, can be determined using traditional debuggers after the data structure evolution leading to the semantic error is observed graphically.

Hedging bets: Applying New Zealand's gambling machine regime to cannabis legalization.

PubMed

Caulkins, Jonathan P

2018-03-01

Cannabis legalization is often falsely depicted as a binary choice between status quo prohibition and legalizing production and distribution by (regulated) for-profit industry. There are, however, many more prudent architectures for legalization, such as restricting production and distribution licenses to not-for-profit entities. Wilkins describes how New Zealand applied that concept to gambling machines and proposes a parallel for cannabis legalization. Greater investment in proposing good designs along these lines, including attending to governance structures, would be valuable. Copyright © 2017 Elsevier B.V. All rights reserved.

An Evaluation of Architectural Platforms for Parallel Navier-Stokes Computations

NASA Technical Reports Server (NTRS)

Jayasimha, D. N.; Hayder, M. E.; Pillay, S. K.

1996-01-01

We study the computational, communication, and scalability characteristics of a computational fluid dynamics application, which solves the time accurate flow field of a jet using the compressible Navier-Stokes equations, on a variety of parallel architecture platforms. The platforms chosen for this study are a cluster of workstations (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), and distributed memory multiprocessors with different topologies - the IBM SP and the Cray T3D. We investigate the impact of various networks connecting the cluster of workstations on the performance of the application and the overheads induced by popular message passing libraries used for parallelization. The work also highlights the importance of matching the memory bandwidth to the processor speed for good single processor performance. By studying the performance of an application on a variety of architectures, we are able to point out the strengths and weaknesses of each of the example computing platforms.

Parallelizing Navier-Stokes Computations on a Variety of Architectural Platforms

NASA Technical Reports Server (NTRS)

Jayasimha, D. N.; Hayder, M. E.; Pillay, S. K.

1997-01-01

We study the computational, communication, and scalability characteristics of a Computational Fluid Dynamics application, which solves the time accurate flow field of a jet using the compressible Navier-Stokes equations, on a variety of parallel architectural platforms. The platforms chosen for this study are a cluster of workstations (the LACE experimental testbed at NASA Lewis), a shared memory multiprocessor (the Cray YMP), distributed memory multiprocessors with different topologies-the IBM SP and the Cray T3D. We investigate the impact of various networks, connecting the cluster of workstations, on the performance of the application and the overheads induced by popular message passing libraries used for parallelization. The work also highlights the importance of matching the memory bandwidth to the processor speed for good single processor performance. By studying the performance of an application on a variety of architectures, we are able to point out the strengths and weaknesses of each of the example computing platforms.

Distributed communications and control network for robotic mining

NASA Technical Reports Server (NTRS)

Schiffbauer, William H.

1989-01-01

The application of robotics to coal mining machines is one approach pursued to increase productivity while providing enhanced safety for the coal miner. Toward that end, a network composed of microcontrollers, computers, expert systems, real time operating systems, and a variety of program languages are being integrated that will act as the backbone for intelligent machine operation. Actual mining machines, including a few customized ones, have been given telerobotic semiautonomous capabilities by applying the described network. Control devices, intelligent sensors and computers onboard these machines are showing promise of achieving improved mining productivity and safety benefits. Current research using these machines involves navigation, multiple machine interaction, machine diagnostics, mineral detection, and graphical machine representation. Guidance sensors and systems employed include: sonar, laser rangers, gyroscopes, magnetometers, clinometers, and accelerometers. Information on the network of hardware/software and its implementation on mining machines are presented. Anticipated coal production operations using the network are discussed. A parallelism is also drawn between the direction of present day underground coal mining research to how the lunar soil (regolith) may be mined. A conceptual lunar mining operation that employs a distributed communication and control network is detailed.

A Framework for Parallel Unstructured Grid Generation for Complex Aerodynamic Simulations

NASA Technical Reports Server (NTRS)

Zagaris, George; Pirzadeh, Shahyar Z.; Chrisochoides, Nikos

2009-01-01

A framework for parallel unstructured grid generation targeting both shared memory multi-processors and distributed memory architectures is presented. The two fundamental building-blocks of the framework consist of: (1) the Advancing-Partition (AP) method used for domain decomposition and (2) the Advancing Front (AF) method used for mesh generation. Starting from the surface mesh of the computational domain, the AP method is applied recursively to generate a set of sub-domains. Next, the sub-domains are meshed in parallel using the AF method. The recursive nature of domain decomposition naturally maps to a divide-and-conquer algorithm which exhibits inherent parallelism. For the parallel implementation, the Master/Worker pattern is employed to dynamically balance the varying workloads of each task on the set of available CPUs. Performance results by this approach are presented and discussed in detail as well as future work and improvements.

Massively Multithreaded Maxflow for Image Segmentation on the Cray XMT-2

PubMed Central

Bokhari, Shahid H.; Çatalyürek, Ümit V.; Gurcan, Metin N.

2014-01-01

SUMMARY Image segmentation is a very important step in the computerized analysis of digital images. The maxflow mincut approach has been successfully used to obtain minimum energy segmentations of images in many fields. Classical algorithms for maxflow in networks do not directly lend themselves to efficient parallel implementations on contemporary parallel processors. We present the results of an implementation of Goldberg-Tarjan preflow-push algorithm on the Cray XMT-2 massively multithreaded supercomputer. This machine has hardware support for 128 threads in each physical processor, a uniformly accessible shared memory of up to 4 TB and hardware synchronization for each 64 bit word. It is thus well-suited to the parallelization of graph theoretic algorithms, such as preflow-push. We describe the implementation of the preflow-push code on the XMT-2 and present the results of timing experiments on a series of synthetically generated as well as real images. Our results indicate very good performance on large images and pave the way for practical applications of this machine architecture for image analysis in a production setting. The largest images we have run are 320002 pixels in size, which are well beyond the largest previously reported in the literature. PMID:25598745

Computational mechanics analysis tools for parallel-vector supercomputers

NASA Technical Reports Server (NTRS)

Storaasli, Olaf O.; Nguyen, Duc T.; Baddourah, Majdi; Qin, Jiangning

1993-01-01

Computational algorithms for structural analysis on parallel-vector supercomputers are reviewed. These parallel algorithms, developed by the authors, are for the assembly of structural equations, 'out-of-core' strategies for linear equation solution, massively distributed-memory equation solution, unsymmetric equation solution, general eigensolution, geometrically nonlinear finite element analysis, design sensitivity analysis for structural dynamics, optimization search analysis and domain decomposition. The source code for many of these algorithms is available.

Aerodynamic Shape Optimization of Supersonic Aircraft Configurations via an Adjoint Formulation on Parallel Computers

NASA Technical Reports Server (NTRS)

Reuther, James; Alonso, Juan Jose; Rimlinger, Mark J.; Jameson, Antony

1996-01-01

This work describes the application of a control theory-based aerodynamic shape optimization method to the problem of supersonic aircraft design. The design process is greatly accelerated through the use of both control theory and a parallel implementation on distributed memory computers. Control theory is employed to derive the adjoint differential equations whose solution allows for the evaluation of design gradient information at a fraction of the computational cost required by previous design methods. The resulting problem is then implemented on parallel distributed memory architectures using a domain decomposition approach, an optimized communication schedule, and the MPI (Message Passing Interface) Standard for portability and efficiency. The final result achieves very rapid aerodynamic design based on higher order computational fluid dynamics methods (CFD). In our earlier studies, the serial implementation of this design method was shown to be effective for the optimization of airfoils, wings, wing-bodies, and complex aircraft configurations using both the potential equation and the Euler equations. In our most recent paper, the Euler method was extended to treat complete aircraft configurations via a new multiblock implementation. Furthermore, during the same conference, we also presented preliminary results demonstrating that this basic methodology could be ported to distributed memory parallel computing architectures. In this paper, our concern will be to demonstrate that the combined power of these new technologies can be used routinely in an industrial design environment by applying it to the case study of the design of typical supersonic transport configurations. A particular difficulty of this test case is posed by the propulsion/airframe integration.

Parallel processing for scientific computations

NASA Technical Reports Server (NTRS)

Alkhatib, Hasan S.

1995-01-01

The scope of this project dealt with the investigation of the requirements to support distributed computing of scientific computations over a cluster of cooperative workstations. Various experiments on computations for the solution of simultaneous linear equations were performed in the early phase of the project to gain experience in the general nature and requirements of scientific applications. A specification of a distributed integrated computing environment, DICE, based on a distributed shared memory communication paradigm has been developed and evaluated. The distributed shared memory model facilitates porting existing parallel algorithms that have been designed for shared memory multiprocessor systems to the new environment. The potential of this new environment is to provide supercomputing capability through the utilization of the aggregate power of workstations cooperating in a cluster interconnected via a local area network. Workstations, generally, do not have the computing power to tackle complex scientific applications, making them primarily useful for visualization, data reduction, and filtering as far as complex scientific applications are concerned. There is a tremendous amount of computing power that is left unused in a network of workstations. Very often a workstation is simply sitting idle on a desk. A set of tools can be developed to take advantage of this potential computing power to create a platform suitable for large scientific computations. The integration of several workstations into a logical cluster of distributed, cooperative, computing stations presents an alternative to shared memory multiprocessor systems. In this project we designed and evaluated such a system.

Exploring Machine Learning Techniques For Dynamic Modeling on Future Exascale Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Song, Shuaiwen; Tallent, Nathan R.; Vishnu, Abhinav

2013-09-23

Future exascale systems must be optimized for both power and performance at scale in order to achieve DOE’s goal of a sustained petaflop within 20 Megawatts by 2022 [1]. Massive parallelism of the future systems combined with complex memory hierarchies will form a barrier to efficient application and architecture design. These challenges are exacerbated with emerging complex architectures such as GPGPUs and Intel Xeon Phi as parallelism increases orders of magnitude and system power consumption can easily triple or quadruple. Therefore, we need techniques that can reduce the search space for optimization, isolate power-performance bottlenecks, identify root causes for software/hardwaremore » inefficiency, and effectively direct runtime scheduling.« less

Parallel volume ray-casting for unstructured-grid data on distributed-memory architectures

NASA Technical Reports Server (NTRS)

Ma, Kwan-Liu

1995-01-01

As computing technology continues to advance, computational modeling of scientific and engineering problems produces data of increasing complexity: large in size and unstructured in shape. Volume visualization of such data is a challenging problem. This paper proposes a distributed parallel solution that makes ray-casting volume rendering of unstructured-grid data practical. Both the data and the rendering process are distributed among processors. At each processor, ray-casting of local data is performed independent of the other processors. The global image composing processes, which require inter-processor communication, are overlapped with the local ray-casting processes to achieve maximum parallel efficiency. This algorithm differs from previous ones in four ways: it is completely distributed, less view-dependent, reasonably scalable, and flexible. Without using dynamic load balancing, test results on the Intel Paragon using from two to 128 processors show, on average, about 60% parallel efficiency.

Implementation and performance of FDPS: a framework for developing parallel particle simulation codes

NASA Astrophysics Data System (ADS)

Iwasawa, Masaki; Tanikawa, Ataru; Hosono, Natsuki; Nitadori, Keigo; Muranushi, Takayuki; Makino, Junichiro

2016-08-01

We present the basic idea, implementation, measured performance, and performance model of FDPS (Framework for Developing Particle Simulators). FDPS is an application-development framework which helps researchers to develop simulation programs using particle methods for large-scale distributed-memory parallel supercomputers. A particle-based simulation program for distributed-memory parallel computers needs to perform domain decomposition, exchange of particles which are not in the domain of each computing node, and gathering of the particle information in other nodes which are necessary for interaction calculation. Also, even if distributed-memory parallel computers are not used, in order to reduce the amount of computation, algorithms such as the Barnes-Hut tree algorithm or the Fast Multipole Method should be used in the case of long-range interactions. For short-range interactions, some methods to limit the calculation to neighbor particles are required. FDPS provides all of these functions which are necessary for efficient parallel execution of particle-based simulations as "templates," which are independent of the actual data structure of particles and the functional form of the particle-particle interaction. By using FDPS, researchers can write their programs with the amount of work necessary to write a simple, sequential and unoptimized program of O(N2) calculation cost, and yet the program, once compiled with FDPS, will run efficiently on large-scale parallel supercomputers. A simple gravitational N-body program can be written in around 120 lines. We report the actual performance of these programs and the performance model. The weak scaling performance is very good, and almost linear speed-up was obtained for up to the full system of the K computer. The minimum calculation time per timestep is in the range of 30 ms (N = 107) to 300 ms (N = 109). These are currently limited by the time for the calculation of the domain decomposition and communication necessary for the interaction calculation. We discuss how we can overcome these bottlenecks.

Applications Performance on NAS Intel Paragon XP/S - 15#

NASA Technical Reports Server (NTRS)

Saini, Subhash; Simon, Horst D.; Copper, D. M. (Technical Monitor)

1994-01-01

The Numerical Aerodynamic Simulation (NAS) Systems Division received an Intel Touchstone Sigma prototype model Paragon XP/S- 15 in February, 1993. The i860 XP microprocessor with an integrated floating point unit and operating in dual -instruction mode gives peak performance of 75 million floating point operations (NIFLOPS) per second for 64 bit floating point arithmetic. It is used in the Paragon XP/S-15 which has been installed at NAS, NASA Ames Research Center. The NAS Paragon has 208 nodes and its peak performance is 15.6 GFLOPS. Here, we will report on early experience using the Paragon XP/S- 15. We have tested its performance using both kernels and applications of interest to NAS. We have measured the performance of BLAS 1, 2 and 3 both assembly-coded and Fortran coded on NAS Paragon XP/S- 15. Furthermore, we have investigated the performance of a single node one-dimensional FFT, a distributed two-dimensional FFT and a distributed three-dimensional FFT Finally, we measured the performance of NAS Parallel Benchmarks (NPB) on the Paragon and compare it with the performance obtained on other highly parallel machines, such as CM-5, CRAY T3D, IBM SP I, etc. In particular, we investigated the following issues, which can strongly affect the performance of the Paragon: a. Impact of the operating system: Intel currently uses as a default an operating system OSF/1 AD from the Open Software Foundation. The paging of Open Software Foundation (OSF) server at 22 MB to make more memory available for the application degrades the performance. We found that when the limit of 26 NIB per node out of 32 MB available is reached, the application is paged out of main memory using virtual memory. When the application starts paging, the performance is considerably reduced. We found that dynamic memory allocation can help applications performance under certain circumstances. b. Impact of data cache on the i860/XP: We measured the performance of the BLAS both assembly coded and Fortran coded. We found that the measured performance of assembly-coded BLAS is much less than what memory bandwidth limitation would predict. The influence of data cache on different sizes of vectors is also investigated using one-dimensional FFTs. c. Impact of processor layout: There are several different ways processors can be laid out within the two-dimensional grid of processors on the Paragon. We have used the FFT example to investigate performance differences based on processors layout.

Dynamic Load Balancing for Grid Partitioning on a SP-2 Multiprocessor: A Framework

NASA Technical Reports Server (NTRS)

Sohn, Andrew; Simon, Horst; Lasinski, T. A. (Technical Monitor)

1994-01-01

Computational requirements of full scale computational fluid dynamics change as computation progresses on a parallel machine. The change in computational intensity causes workload imbalance of processors, which in turn requires a large amount of data movement at runtime. If parallel CFD is to be successful on a parallel or massively parallel machine, balancing of the runtime load is indispensable. Here a framework is presented for dynamic load balancing for CFD applications, called Jove. One processor is designated as a decision maker Jove while others are assigned to computational fluid dynamics. Processors running CFD send flags to Jove in a predetermined number of iterations to initiate load balancing. Jove starts working on load balancing while other processors continue working with the current data and load distribution. Jove goes through several steps to decide if the new data should be taken, including preliminary evaluate, partition, processor reassignment, cost evaluation, and decision. Jove running on a single EBM SP2 node has been completely implemented. Preliminary experimental results show that the Jove approach to dynamic load balancing can be effective for full scale grid partitioning on the target machine IBM SP2.

Dynamic Load Balancing For Grid Partitioning on a SP-2 Multiprocessor: A Framework

NASA Technical Reports Server (NTRS)

Sohn, Andrew; Simon, Horst; Lasinski, T. A. (Technical Monitor)

1994-01-01

Computational requirements of full scale computational fluid dynamics change as computation progresses on a parallel machine. The change in computational intensity causes workload imbalance of processors, which in turn requires a large amount of data movement at runtime. If parallel CFD is to be successful on a parallel or massively parallel machine, balancing of the runtime load is indispensable. Here a framework is presented for dynamic load balancing for CFD applications, called Jove. One processor is designated as a decision maker Jove while others are assigned to computational fluid dynamics. Processors running CFD send flags to Jove in a predetermined number of iterations to initiate load balancing. Jove starts working on load balancing while other processors continue working with the current data and load distribution. Jove goes through several steps to decide if the new data should be taken, including preliminary evaluate, partition, processor reassignment, cost evaluation, and decision. Jove running on a single IBM SP2 node has been completely implemented. Preliminary experimental results show that the Jove approach to dynamic load balancing can be effective for full scale grid partitioning on the target machine IBM SP2.

Parallelizing serial code for a distributed processing environment with an application to high frequency electromagnetic scattering

NASA Astrophysics Data System (ADS)

Work, Paul R.

1991-12-01

This thesis investigates the parallelization of existing serial programs in computational electromagnetics for use in a parallel environment. Existing algorithms for calculating the radar cross section of an object are covered, and a ray-tracing code is chosen for implementation on a parallel machine. Current parallel architectures are introduced and a suitable parallel machine is selected for the implementation of the chosen ray-tracing algorithm. The standard techniques for the parallelization of serial codes are discussed, including load balancing and decomposition considerations, and appropriate methods for the parallelization effort are selected. A load balancing algorithm is modified to increase the efficiency of the application, and a high level design of the structure of the serial program is presented. A detailed design of the modifications for the parallel implementation is also included, with both the high level and the detailed design specified in a high level design language called UNITY. The correctness of the design is proven using UNITY and standard logic operations. The theoretical and empirical results show that it is possible to achieve an efficient parallel application for a serial computational electromagnetic program where the characteristics of the algorithm and the target architecture critically influence the development of such an implementation.

Exascale Hardware Architectures Working Group

DOE Office of Scientific and Technical Information (OSTI.GOV)

Hemmert, S; Ang, J; Chiang, P

2011-03-15

The ASC Exascale Hardware Architecture working group is challenged to provide input on the following areas impacting the future use and usability of potential exascale computer systems: processor, memory, and interconnect architectures, as well as the power and resilience of these systems. Going forward, there are many challenging issues that will need to be addressed. First, power constraints in processor technologies will lead to steady increases in parallelism within a socket. Additionally, all cores may not be fully independent nor fully general purpose. Second, there is a clear trend toward less balanced machines, in terms of compute capability compared tomore » memory and interconnect performance. In order to mitigate the memory issues, memory technologies will introduce 3D stacking, eventually moving on-socket and likely on-die, providing greatly increased bandwidth but unfortunately also likely providing smaller memory capacity per core. Off-socket memory, possibly in the form of non-volatile memory, will create a complex memory hierarchy. Third, communication energy will dominate the energy required to compute, such that interconnect power and bandwidth will have a significant impact. All of the above changes are driven by the need for greatly increased energy efficiency, as current technology will prove unsuitable for exascale, due to unsustainable power requirements of such a system. These changes will have the most significant impact on programming models and algorithms, but they will be felt across all layers of the machine. There is clear need to engage all ASC working groups in planning for how to deal with technological changes of this magnitude. The primary function of the Hardware Architecture Working Group is to facilitate codesign with hardware vendors to ensure future exascale platforms are capable of efficiently supporting the ASC applications, which in turn need to meet the mission needs of the NNSA Stockpile Stewardship Program. This issue is relatively immediate, as there is only a small window of opportunity to influence hardware design for 2018 machines. Given the short timeline a firm co-design methodology with vendors is of prime importance.« less

Hardware/software codesign for embedded RISC core

NASA Astrophysics Data System (ADS)

Liu, Peng

2001-12-01

This paper describes hardware/software codesign method of the extendible embedded RISC core VIRGO, which based on MIPS-I instruction set architecture. VIRGO is described by Verilog hardware description language that has five-stage pipeline with shared 32-bit cache/memory interface, and it is controlled by distributed control scheme. Every pipeline stage has one small controller, which controls the pipeline stage status and cooperation among the pipeline phase. Since description use high level language and structure is distributed, VIRGO core has highly extension that can meet the requirements of application. We take look at the high-definition television MPEG2 MPHL decoder chip, constructed the hardware/software codesign virtual prototyping machine that can research on VIRGO core instruction set architecture, and system on chip memory size requirements, and system on chip software, etc. We also can evaluate the system on chip design and RISC instruction set based on the virtual prototyping machine platform.

Efficient implementation of multidimensional fast fourier transform on a distributed-memory parallel multi-node computer

DOEpatents

Bhanot, Gyan V [Princeton, NJ; Chen, Dong [Croton-On-Hudson, NY; Gara, Alan G [Mount Kisco, NY; Giampapa, Mark E [Irvington, NY; Heidelberger, Philip [Cortlandt Manor, NY; Steinmacher-Burow, Burkhard D [Mount Kisco, NY; Vranas, Pavlos M [Bedford Hills, NY

2012-01-10

The present in invention is directed to a method, system and program storage device for efficiently implementing a multidimensional Fast Fourier Transform (FFT) of a multidimensional array comprising a plurality of elements initially distributed in a multi-node computer system comprising a plurality of nodes in communication over a network, comprising: distributing the plurality of elements of the array in a first dimension across the plurality of nodes of the computer system over the network to facilitate a first one-dimensional FFT; performing the first one-dimensional FFT on the elements of the array distributed at each node in the first dimension; re-distributing the one-dimensional FFT-transformed elements at each node in a second dimension via "all-to-all" distribution in random order across other nodes of the computer system over the network; and performing a second one-dimensional FFT on elements of the array re-distributed at each node in the second dimension, wherein the random order facilitates efficient utilization of the network thereby efficiently implementing the multidimensional FFT. The "all-to-all" re-distribution of array elements is further efficiently implemented in applications other than the multidimensional FFT on the distributed-memory parallel supercomputer.

Efficient implementation of a multidimensional fast fourier transform on a distributed-memory parallel multi-node computer

DOEpatents

Bhanot, Gyan V [Princeton, NJ; Chen, Dong [Croton-On-Hudson, NY; Gara, Alan G [Mount Kisco, NY; Giampapa, Mark E [Irvington, NY; Heidelberger, Philip [Cortlandt Manor, NY; Steinmacher-Burow, Burkhard D [Mount Kisco, NY; Vranas, Pavlos M [Bedford Hills, NY

2008-01-01

The present in invention is directed to a method, system and program storage device for efficiently implementing a multidimensional Fast Fourier Transform (FFT) of a multidimensional array comprising a plurality of elements initially distributed in a multi-node computer system comprising a plurality of nodes in communication over a network, comprising: distributing the plurality of elements of the array in a first dimension across the plurality of nodes of the computer system over the network to facilitate a first one-dimensional FFT; performing the first one-dimensional FFT on the elements of the array distributed at each node in the first dimension; re-distributing the one-dimensional FFT-transformed elements at each node in a second dimension via "all-to-all" distribution in random order across other nodes of the computer system over the network; and performing a second one-dimensional FFT on elements of the array re-distributed at each node in the second dimension, wherein the random order facilitates efficient utilization of the network thereby efficiently implementing the multidimensional FFT. The "all-to-all" re-distribution of array elements is further efficiently implemented in applications other than the multidimensional FFT on the distributed-memory parallel supercomputer.

A Parallel Rendering Algorithm for MIMD Architectures

NASA Technical Reports Server (NTRS)

Crockett, Thomas W.; Orloff, Tobias

1991-01-01

Applications such as animation and scientific visualization demand high performance rendering of complex three dimensional scenes. To deliver the necessary rendering rates, highly parallel hardware architectures are required. The challenge is then to design algorithms and software which effectively use the hardware parallelism. A rendering algorithm targeted to distributed memory MIMD architectures is described. For maximum performance, the algorithm exploits both object-level and pixel-level parallelism. The behavior of the algorithm is examined both analytically and experimentally. Its performance for large numbers of processors is found to be limited primarily by communication overheads. An experimental implementation for the Intel iPSC/860 shows increasing performance from 1 to 128 processors across a wide range of scene complexities. It is shown that minimal modifications to the algorithm will adapt it for use on shared memory architectures as well.

A parallel implementation of a multisensor feature-based range-estimation method

NASA Technical Reports Server (NTRS)

Suorsa, Raymond E.; Sridhar, Banavar

1993-01-01

There are many proposed vision based methods to perform obstacle detection and avoidance for autonomous or semi-autonomous vehicles. All methods, however, will require very high processing rates to achieve real time performance. A system capable of supporting autonomous helicopter navigation will need to extract obstacle information from imagery at rates varying from ten frames per second to thirty or more frames per second depending on the vehicle speed. Such a system will need to sustain billions of operations per second. To reach such high processing rates using current technology, a parallel implementation of the obstacle detection/ranging method is required. This paper describes an efficient and flexible parallel implementation of a multisensor feature-based range-estimation algorithm, targeted for helicopter flight, realized on both a distributed-memory and shared-memory parallel computer.

Array distribution in data-parallel programs

NASA Technical Reports Server (NTRS)

Chatterjee, Siddhartha; Gilbert, John R.; Schreiber, Robert; Sheffler, Thomas J.

1994-01-01

We consider distribution at compile time of the array data in a distributed-memory implementation of a data-parallel program written in a language like Fortran 90. We allow dynamic redistribution of data and define a heuristic algorithmic framework that chooses distribution parameters to minimize an estimate of program completion time. We represent the program as an alignment-distribution graph. We propose a divide-and-conquer algorithm for distribution that initially assigns a common distribution to each node of the graph and successively refines this assignment, taking computation, realignment, and redistribution costs into account. We explain how to estimate the effect of distribution on computation cost and how to choose a candidate set of distributions. We present the results of an implementation of our algorithms on several test problems.

What does fault tolerant Deep Learning need from MPI?

DOE Office of Scientific and Technical Information (OSTI.GOV)

Amatya, Vinay C.; Vishnu, Abhinav; Siegel, Charles M.

Deep Learning (DL) algorithms have become the {\\em de facto} Machine Learning (ML) algorithm for large scale data analysis. DL algorithms are computationally expensive -- even distributed DL implementations which use MPI require days of training (model learning) time on commonly studied datasets. Long running DL applications become susceptible to faults -- requiring development of a fault tolerant system infrastructure, in addition to fault tolerant DL algorithms. This raises an important question: {\\em What is needed from MPI for designing fault tolerant DL implementations?} In this paper, we address this problem for permanent faults. We motivate the need for amore » fault tolerant MPI specification by an in-depth consideration of recent innovations in DL algorithms and their properties, which drive the need for specific fault tolerance features. We present an in-depth discussion on the suitability of different parallelism types (model, data and hybrid); a need (or lack thereof) for check-pointing of any critical data structures; and most importantly, consideration for several fault tolerance proposals (user-level fault mitigation (ULFM), Reinit) in MPI and their applicability to fault tolerant DL implementations. We leverage a distributed memory implementation of Caffe, currently available under the Machine Learning Toolkit for Extreme Scale (MaTEx). We implement our approaches by extending MaTEx-Caffe for using ULFM-based implementation. Our evaluation using the ImageNet dataset and AlexNet neural network topology demonstrates the effectiveness of the proposed fault tolerant DL implementation using OpenMPI based ULFM.« less

Distributed and parallel approach for handle and perform huge datasets

NASA Astrophysics Data System (ADS)

Konopko, Joanna

2015-12-01

Big Data refers to the dynamic, large and disparate volumes of data comes from many different sources (tools, machines, sensors, mobile devices) uncorrelated with each others. It requires new, innovative and scalable technology to collect, host and analytically process the vast amount of data. Proper architecture of the system that perform huge data sets is needed. In this paper, the comparison of distributed and parallel system architecture is presented on the example of MapReduce (MR) Hadoop platform and parallel database platform (DBMS). This paper also analyzes the problem of performing and handling valuable information from petabytes of data. The both paradigms: MapReduce and parallel DBMS are described and compared. The hybrid architecture approach is also proposed and could be used to solve the analyzed problem of storing and processing Big Data.

Computational mechanics analysis tools for parallel-vector supercomputers

NASA Technical Reports Server (NTRS)

Storaasli, O. O.; Nguyen, D. T.; Baddourah, M. A.; Qin, J.

1993-01-01

Computational algorithms for structural analysis on parallel-vector supercomputers are reviewed. These parallel algorithms, developed by the authors, are for the assembly of structural equations, 'out-of-core' strategies for linear equation solution, massively distributed-memory equation solution, unsymmetric equation solution, general eigen-solution, geometrically nonlinear finite element analysis, design sensitivity analysis for structural dynamics, optimization algorithm and domain decomposition. The source code for many of these algorithms is available from NASA Langley.

Incremental Parallelization of Non-Data-Parallel Programs Using the Charon Message-Passing Library

NASA Technical Reports Server (NTRS)

VanderWijngaart, Rob F.

2000-01-01

Message passing is among the most popular techniques for parallelizing scientific programs on distributed-memory architectures. The reasons for its success are wide availability (MPI), efficiency, and full tuning control provided to the programmer. A major drawback, however, is that incremental parallelization, as offered by compiler directives, is not generally possible, because all data structures have to be changed throughout the program simultaneously. Charon remedies this situation through mappings between distributed and non-distributed data. It allows breaking up the parallelization into small steps, guaranteeing correctness at every stage. Several tools are available to help convert legacy codes into high-performance message-passing programs. They usually target data-parallel applications, whose loops carrying most of the work can be distributed among all processors without much dependency analysis. Others do a full dependency analysis and then convert the code virtually automatically. Even more toolkits are available that aid construction from scratch of message passing programs. None, however, allows piecemeal translation of codes with complex data dependencies (i.e. non-data-parallel programs) into message passing codes. The Charon library (available in both C and Fortran) provides incremental parallelization capabilities by linking legacy code arrays with distributed arrays. During the conversion process, non-distributed and distributed arrays exist side by side, and simple mapping functions allow the programmer to switch between the two in any location in the program. Charon also provides wrapper functions that leave the structure of the legacy code intact, but that allow execution on truly distributed data. Finally, the library provides a rich set of communication functions that support virtually all patterns of remote data demands in realistic structured grid scientific programs, including transposition, nearest-neighbor communication, pipelining, gather/scatter, and redistribution. At the end of the conversion process most intermediate Charon function calls will have been removed, the non-distributed arrays will have been deleted, and virtually the only remaining Charon functions calls are the high-level, highly optimized communications. Distribution of the data is under complete control of the programmer, although a wide range of useful distributions is easily available through predefined functions. A crucial aspect of the library is that it does not allocate space for distributed arrays, but accepts programmer-specified memory. This has two major consequences. First, codes parallelized using Charon do not suffer from encapsulation; user data is always directly accessible. This provides high efficiency, and also retains the possibility of using message passing directly for highly irregular communications. Second, non-distributed arrays can be interpreted as (trivial) distributions in the Charon sense, which allows them to be mapped to truly distributed arrays, and vice versa. This is the mechanism that enables incremental parallelization. In this paper we provide a brief introduction of the library and then focus on the actual steps in the parallelization process, using some representative examples from, among others, the NAS Parallel Benchmarks. We show how a complicated two-dimensional pipeline-the prototypical non-data-parallel algorithm- can be constructed with ease. To demonstrate the flexibility of the library, we give examples of the stepwise, efficient parallel implementation of nonlocal boundary conditions common in aircraft simulations, as well as the construction of the sequence of grids required for multigrid.

From biological neural networks to thinking machines: Transitioning biological organizational principles to computer technology

NASA Technical Reports Server (NTRS)

Ross, Muriel D.

1991-01-01

The three-dimensional organization of the vestibular macula is under study by computer assisted reconstruction and simulation methods as a model for more complex neural systems. One goal of this research is to transition knowledge of biological neural network architecture and functioning to computer technology, to contribute to the development of thinking computers. Maculas are organized as weighted neural networks for parallel distributed processing of information. The network is characterized by non-linearity of its terminal/receptive fields. Wiring appears to develop through constrained randomness. A further property is the presence of two main circuits, highly channeled and distributed modifying, that are connected through feedforward-feedback collaterals and biasing subcircuit. Computer simulations demonstrate that differences in geometry of the feedback (afferent) collaterals affects the timing and the magnitude of voltage changes delivered to the spike initiation zone. Feedforward (efferent) collaterals act as voltage followers and likely inhibit neurons of the distributed modifying circuit. These results illustrate the importance of feedforward-feedback loops, of timing, and of inhibition in refining neural network output. They also suggest that it is the distributed modifying network that is most involved in adaptation, memory, and learning. Tests of macular adaptation, through hyper- and microgravitational studies, support this hypothesis since synapses in the distributed modifying circuit, but not the channeled circuit, are altered. Transitioning knowledge of biological systems to computer technology, however, remains problematical.

Confessions of a robot lobotomist

NASA Technical Reports Server (NTRS)

Gottshall, R. Marc

1994-01-01

Since its inception, numerically controlled (NC) machining methods have been used throughout the aerospace industry to mill, drill, and turn complex shapes by sequentially stepping through motion programs. However, the recent demand for more precision, faster feeds, exotic sensors, and branching execution have existing computer numerical control (CNC) and distributed numerical control (DNC) systems running at maximum controller capacity. Typical disadvantages of current CNC's include fixed memory capacities, limited communication ports, and the use of multiple control languages. The need to tailor CNC's to meet specific applications, whether it be expanded memory, additional communications, or integrated vision, often requires replacing the original controller supplied with the commercial machine tool with a more powerful and capable system. This paper briefly describes the process and equipment requirements for new controllers and their evolutionary implementation in an aerospace environment. The process of controller retrofit with currently available machines is examined, along with several case studies and their computational and architectural implications.

GSRP/David Marshall: Fully Automated Cartesian Grid CFD Application for MDO in High Speed Flows

NASA Technical Reports Server (NTRS)

2003-01-01

With the renewed interest in Cartesian gridding methodologies for the ease and speed of gridding complex geometries in addition to the simplicity of the control volumes used in the computations, it has become important to investigate ways of extending the existing Cartesian grid solver functionalities. This includes developing methods of modeling the viscous effects in order to utilize Cartesian grids solvers for accurate drag predictions and addressing the issues related to the distributed memory parallelization of Cartesian solvers. This research presents advances in two areas of interest in Cartesian grid solvers, viscous effects modeling and MPI parallelization. The development of viscous effects modeling using solely Cartesian grids has been hampered by the widely varying control volume sizes associated with the mesh refinement and the cut cells associated with the solid surface. This problem is being addressed by using physically based modeling techniques to update the state vectors of the cut cells and removing them from the finite volume integration scheme. This work is performed on a new Cartesian grid solver, NASCART-GT, with modifications to its cut cell functionality. The development of MPI parallelization addresses issues associated with utilizing Cartesian solvers on distributed memory parallel environments. This work is performed on an existing Cartesian grid solver, CART3D, with modifications to its parallelization methodology.

Application Characterization at Scale: Lessons learned from developing a distributed Open Community Runtime system for High Performance Computing

DOE Office of Scientific and Technical Information (OSTI.GOV)

Landwehr, Joshua B.; Suetterlein, Joshua D.; Marquez, Andres

2016-05-16

Since 2012, the U.S. Department of Energy’s X-Stack program has been developing solutions including runtime systems, programming models, languages, compilers, and tools for the Exascale system software to address crucial performance and power requirements. Fine grain programming models and runtime systems show a great potential to efficiently utilize the underlying hardware. Thus, they are essential to many X-Stack efforts. An abundant amount of small tasks can better utilize the vast parallelism available on current and future machines. Moreover, finer tasks can recover faster and adapt better, due to a decrease in state and control. Nevertheless, current applications have been writtenmore » to exploit old paradigms (such as Communicating Sequential Processor and Bulk Synchronous Parallel processing). To fully utilize the advantages of these new systems, applications need to be adapted to these new paradigms. As part of the applications’ porting process, in-depth characterization studies, focused on both application characteristics and runtime features, need to take place to fully understand the application performance bottlenecks and how to resolve them. This paper presents a characterization study for a novel high performance runtime system, called the Open Community Runtime, using key HPC kernels as its vehicle. This study has the following contributions: one of the first high performance, fine grain, distributed memory runtime system implementing the OCR standard (version 0.99a); and a characterization study of key HPC kernels in terms of runtime primitives running on both intra and inter node environments. Running on a general purpose cluster, we have found up to 1635x relative speed-up for a parallel tiled Cholesky Kernels on 128 nodes with 16 cores each and a 1864x relative speed-up for a parallel tiled Smith-Waterman kernel on 128 nodes with 30 cores.« less

Optical Symbolic Computing

NASA Astrophysics Data System (ADS)

Neff, John A.

1989-12-01

Experiments originating from Gestalt psychology have shown that representing information in a symbolic form provides a more effective means to understanding. Computer scientists have been struggling for the last two decades to determine how best to create, manipulate, and store collections of symbolic structures. In the past, much of this struggling led to software innovations because that was the path of least resistance. For example, the development of heuristics for organizing the searching through knowledge bases was much less expensive than building massively parallel machines that could search in parallel. That is now beginning to change with the emergence of parallel architectures which are showing the potential for handling symbolic structures. This paper will review the relationships between symbolic computing and parallel computing architectures, and will identify opportunities for optics to significantly impact the performance of such computing machines. Although neural networks are an exciting subset of massively parallel computing structures, this paper will not touch on this area since it is receiving a great deal of attention in the literature. That is, the concepts presented herein do not consider the distributed representation of knowledge.

Aerodynamic Shape Optimization of Supersonic Aircraft Configurations via an Adjoint Formulation on Parallel Computers

NASA Technical Reports Server (NTRS)

Reuther, James; Alonso, Juan Jose; Rimlinger, Mark J.; Jameson, Antony

1996-01-01

This work describes the application of a control theory-based aerodynamic shape optimization method to the problem of supersonic aircraft design. The design process is greatly accelerated through the use of both control theory and a parallel implementation on distributed memory computers. Control theory is employed to derive the adjoint differential equations whose solution allows for the evaluation of design gradient information at a fraction of the computational cost required by previous design methods (13, 12, 44, 38). The resulting problem is then implemented on parallel distributed memory architectures using a domain decomposition approach, an optimized communication schedule, and the MPI (Message Passing Interface) Standard for portability and efficiency. The final result achieves very rapid aerodynamic design based on higher order computational fluid dynamics methods (CFD). In our earlier studies, the serial implementation of this design method (19, 20, 21, 23, 39, 25, 40, 41, 42, 43, 9) was shown to be effective for the optimization of airfoils, wings, wing-bodies, and complex aircraft configurations using both the potential equation and the Euler equations (39, 25). In our most recent paper, the Euler method was extended to treat complete aircraft configurations via a new multiblock implementation. Furthermore, during the same conference, we also presented preliminary results demonstrating that the basic methodology could be ported to distributed memory parallel computing architectures [241. In this paper, our concem will be to demonstrate that the combined power of these new technologies can be used routinely in an industrial design environment by applying it to the case study of the design of typical supersonic transport configurations. A particular difficulty of this test case is posed by the propulsion/airframe integration.

A sample implementation for parallelizing Divide-and-Conquer algorithms on the GPU.

PubMed

Mei, Gang; Zhang, Jiayin; Xu, Nengxiong; Zhao, Kunyang

2018-01-01

The strategy of Divide-and-Conquer (D&C) is one of the frequently used programming patterns to design efficient algorithms in computer science, which has been parallelized on shared memory systems and distributed memory systems. Tzeng and Owens specifically developed a generic paradigm for parallelizing D&C algorithms on modern Graphics Processing Units (GPUs). In this paper, by following the generic paradigm proposed by Tzeng and Owens, we provide a new and publicly available GPU implementation of the famous D&C algorithm, QuickHull, to give a sample and guide for parallelizing D&C algorithms on the GPU. The experimental results demonstrate the practicality of our sample GPU implementation. Our research objective in this paper is to present a sample GPU implementation of a classical D&C algorithm to help interested readers to develop their own efficient GPU implementations with fewer efforts.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jones, J.P.; Bangs, A.L.; Butler, P.L.

Hetero Helix is a programming environment which simulates shared memory on a heterogeneous network of distributed-memory computers. The machines in the network may vary with respect to their native operating systems and internal representation of numbers. Hetero Helix presents a simple programming model to developers, and also considers the needs of designers, system integrators, and maintainers. The key software technology underlying Hetero Helix is the use of a compiler'' which analyzes the data structures in shared memory and automatically generates code which translates data representations from the format native to each machine into a common format, and vice versa. Themore » design of Hetero Helix was motivated in particular by the requirements of robotics applications. Hetero Helix has been used successfully in an integration effort involving 27 CPUs in a heterogeneous network and a body of software totaling roughly 100,00 lines of code. 25 refs., 6 figs.« less

Integrated Network Decompositions and Dynamic Programming for Graph Optimization (INDDGO)

DOE Office of Scientific and Technical Information (OSTI.GOV)

The INDDGO software package offers a set of tools for finding exact solutions to graph optimization problems via tree decompositions and dynamic programming algorithms. Currently the framework offers serial and parallel (distributed memory) algorithms for finding tree decompositions and solving the maximum weighted independent set problem. The parallel dynamic programming algorithm is implemented on top of the MADNESS task-based runtime.

Equation solvers for distributed-memory computers

NASA Technical Reports Server (NTRS)

Storaasli, Olaf O.

1994-01-01

A large number of scientific and engineering problems require the rapid solution of large systems of simultaneous equations. The performance of parallel computers in this area now dwarfs traditional vector computers by nearly an order of magnitude. This talk describes the major issues involved in parallel equation solvers with particular emphasis on the Intel Paragon, IBM SP-1 and SP-2 processors.

Design and Implementation of a Parallel Multivariate Ensemble Kalman Filter for the Poseidon Ocean General Circulation Model

NASA Technical Reports Server (NTRS)

Keppenne, Christian L.; Rienecker, Michele M.; Koblinsky, Chester (Technical Monitor)

2001-01-01

A multivariate ensemble Kalman filter (MvEnKF) implemented on a massively parallel computer architecture has been implemented for the Poseidon ocean circulation model and tested with a Pacific Basin model configuration. There are about two million prognostic state-vector variables. Parallelism for the data assimilation step is achieved by regionalization of the background-error covariances that are calculated from the phase-space distribution of the ensemble. Each processing element (PE) collects elements of a matrix measurement functional from nearby PEs. To avoid the introduction of spurious long-range covariances associated with finite ensemble sizes, the background-error covariances are given compact support by means of a Hadamard (element by element) product with a three-dimensional canonical correlation function. The methodology and the MvEnKF configuration are discussed. It is shown that the regionalization of the background covariances; has a negligible impact on the quality of the analyses. The parallel algorithm is very efficient for large numbers of observations but does not scale well beyond 100 PEs at the current model resolution. On a platform with distributed memory, memory rather than speed is the limiting factor.

Dynamic programming on a shared-memory multiprocessor

NASA Technical Reports Server (NTRS)

Edmonds, Phil; Chu, Eleanor; George, Alan

1993-01-01

Three new algorithms for solving dynamic programming problems on a shared-memory parallel computer are described. All three algorithms attempt to balance work load, while keeping synchronization cost low. In particular, for a multiprocessor having p processors, an analysis of the best algorithm shows that the arithmetic cost is O(n-cubed/6p) and that the synchronization cost is O(absolute value of log sub C n) if p much less than n, where C = (2p-1)/(2p + 1) and n is the size of the problem. The low synchronization cost is important for machines where synchronization is expensive. Analysis and experiments show that the best algorithm is effective in balancing the work load and producing high efficiency.

Efficient parallelization for AMR MHD multiphysics calculations; implementation in AstroBEAR

NASA Astrophysics Data System (ADS)

Carroll-Nellenback, Jonathan J.; Shroyer, Brandon; Frank, Adam; Ding, Chen

2013-03-01

Current adaptive mesh refinement (AMR) simulations require algorithms that are highly parallelized and manage memory efficiently. As compute engines grow larger, AMR simulations will require algorithms that achieve new levels of efficient parallelization and memory management. We have attempted to employ new techniques to achieve both of these goals. Patch or grid based AMR often employs ghost cells to decouple the hyperbolic advances of each grid on a given refinement level. This decoupling allows each grid to be advanced independently. In AstroBEAR we utilize this independence by threading the grid advances on each level with preference going to the finer level grids. This allows for global load balancing instead of level by level load balancing and allows for greater parallelization across both physical space and AMR level. Threading of level advances can also improve performance by interleaving communication with computation, especially in deep simulations with many levels of refinement. While we see improvements of up to 30% on deep simulations run on a few cores, the speedup is typically more modest (5-20%) for larger scale simulations. To improve memory management we have employed a distributed tree algorithm that requires processors to only store and communicate local sections of the AMR tree structure with neighboring processors. Using this distributed approach we are able to get reasonable scaling efficiency (>80%) out to 12288 cores and up to 8 levels of AMR - independent of the use of threading.

Sparse distributed memory prototype: Principles of operation

NASA Technical Reports Server (NTRS)

Flynn, Michael J.; Kanerva, Pentti; Ahanin, Bahram; Bhadkamkar, Neal; Flaherty, Paul; Hickey, Philip

1988-01-01

Sparse distributed memory is a generalized random access memory (RAM) for long binary words. Such words can be written into and read from the memory, and they can be used to address the memory. The main attribute of the memory is sensitivity to similarity, meaning that a word can be read back not only by giving the original right address but also by giving one close to it as measured by the Hamming distance between addresses. Large memories of this kind are expected to have wide use in speech and scene analysis, in signal detection and verification, and in adaptive control of automated equipment. The memory can be realized as a simple, massively parallel computer. Digital technology has reached a point where building large memories is becoming practical. The research is aimed at resolving major design issues that have to be faced in building the memories. The design of a prototype memory with 256-bit addresses and from 8K to 128K locations for 256-bit words is described. A key aspect of the design is extensive use of dynamic RAM and other standard components.

A message passing kernel for the hypercluster parallel processing test bed

NASA Technical Reports Server (NTRS)

Blech, Richard A.; Quealy, Angela; Cole, Gary L.

1989-01-01

A Message-Passing Kernel (MPK) for the Hypercluster parallel-processing test bed is described. The Hypercluster is being developed at the NASA Lewis Research Center to support investigations of parallel algorithms and architectures for computational fluid and structural mechanics applications. The Hypercluster resembles the hypercube architecture except that each node consists of multiple processors communicating through shared memory. The MPK efficiently routes information through the Hypercluster, using a message-passing protocol when necessary and faster shared-memory communication whenever possible. The MPK also interfaces all of the processors with the Hypercluster operating system (HYCLOPS), which runs on a Front-End Processor (FEP). This approach distributes many of the I/O tasks to the Hypercluster processors and eliminates the need for a separate I/O support program on the FEP.

ELAPSE - NASA AMES LISP AND ADA BENCHMARK SUITE: EFFICIENCY OF LISP AND ADA PROCESSING - A SYSTEM EVALUATION

NASA Technical Reports Server (NTRS)

Davis, G. J.

1994-01-01

One area of research of the Information Sciences Division at NASA Ames Research Center is devoted to the analysis and enhancement of processors and advanced computer architectures, specifically in support of automation and robotic systems. To compare systems' abilities to efficiently process Lisp and Ada, scientists at Ames Research Center have developed a suite of non-parallel benchmarks called ELAPSE. The benchmark suite was designed to test a single computer's efficiency as well as alternate machine comparisons on Lisp, and/or Ada languages. ELAPSE tests the efficiency with which a machine can execute the various routines in each environment. The sample routines are based on numeric and symbolic manipulations and include two-dimensional fast Fourier transformations, Cholesky decomposition and substitution, Gaussian elimination, high-level data processing, and symbol-list references. Also included is a routine based on a Bayesian classification program sorting data into optimized groups. The ELAPSE benchmarks are available for any computer with a validated Ada compiler and/or Common Lisp system. Of the 18 routines that comprise ELAPSE, provided within this package are 14 developed or translated at Ames. The others are readily available through literature. The benchmark that requires the most memory is CHOLESKY.ADA. Under VAX/VMS, CHOLESKY.ADA requires 760K of main memory. ELAPSE is available on either two 5.25 inch 360K MS-DOS format diskettes (standard distribution) or a 9-track 1600 BPI ASCII CARD IMAGE format magnetic tape. The contents of the diskettes are compressed using the PKWARE archiving tools. The utility to unarchive the files, PKUNZIP.EXE, is included. The ELAPSE benchmarks were written in 1990. VAX and VMS are trademarks of Digital Equipment Corporation. MS-DOS is a registered trademark of Microsoft Corporation.

HARP: A Dynamic Inertial Spectral Partitioner

NASA Technical Reports Server (NTRS)

Simon, Horst D.; Sohn, Andrew; Biswas, Rupak

1997-01-01

Partitioning unstructured graphs is central to the parallel solution of computational science and engineering problems. Spectral partitioners, such recursive spectral bisection (RSB), have proven effecfive in generating high-quality partitions of realistically-sized meshes. The major problem which hindered their wide-spread use was their long execution times. This paper presents a new inertial spectral partitioner, called HARP. The main objective of the proposed approach is to quickly partition the meshes at runtime in a manner that works efficiently for real applications in the context of distributed-memory machines. The underlying principle of HARP is to find the eigenvectors of the unpartitioned vertices and then project them onto the eigerivectors of the original mesh. Results for various meshes ranging in size from 1000 to 100,000 vertices indicate that HARP can indeed partition meshes rapidly at runtime. Experimental results show that our largest mesh can be partitioned sequentially in only a few seconds on an SP2 which is several times faster than other spectral partitioners while maintaining the solution quality of the proven RSB method. A parallel WI version of HARP has also been implemented on IBM SP2 and Cray T3E. Parallel HARP, running on 64 processors SP2 and T3E, can partition a mesh containing more than 100,000 vertices into 64 subgrids in about half a second. These results indicate that graph partitioning can now be truly embedded in dynamically-changing real-world applications.

Comparison of Origin 2000 and Origin 3000 Using NAS Parallel Benchmarks

NASA Technical Reports Server (NTRS)

Turney, Raymond D.

2001-01-01

This report describes results of benchmark tests on the Origin 3000 system currently being installed at the NASA Ames National Advanced Supercomputing facility. This machine will ultimately contain 1024 R14K processors. The first part of the system, installed in November, 2000 and named mendel, is an Origin 3000 with 128 R12K processors. For comparison purposes, the tests were also run on lomax, an Origin 2000 with R12K processors. The BT, LU, and SP application benchmarks in the NAS Parallel Benchmark Suite and the kernel benchmark FT were chosen to determine system performance and measure the impact of changes on the machine as it evolves. Having been written to measure performance on Computational Fluid Dynamics applications, these benchmarks are assumed appropriate to represent the NAS workload. Since the NAS runs both message passing (MPI) and shared-memory, compiler directive type codes, both MPI and OpenMP versions of the benchmarks were used. The MPI versions used were the latest official release of the NAS Parallel Benchmarks, version 2.3. The OpenMP versiqns used were PBN3b2, a beta version that is in the process of being released. NPB 2.3 and PBN 3b2 are technically different benchmarks, and NPB results are not directly comparable to PBN results.

Enhancing Application Performance Using Mini-Apps: Comparison of Hybrid Parallel Programming Paradigms

NASA Technical Reports Server (NTRS)

Lawson, Gary; Poteat, Michael; Sosonkina, Masha; Baurle, Robert; Hammond, Dana

2016-01-01

In this work, several mini-apps have been created to enhance a real-world application performance, namely the VULCAN code for complex flow analysis developed at the NASA Langley Research Center. These mini-apps explore hybrid parallel programming paradigms with Message Passing Interface (MPI) for distributed memory access and either Shared MPI (SMPI) or OpenMP for shared memory accesses. Performance testing shows that MPI+SMPI yields the best execution performance, while requiring the largest number of code changes. A maximum speedup of 23X was measured for MPI+SMPI, but only 10X was measured for MPI+OpenMP.

HPC-NMF: A High-Performance Parallel Algorithm for Nonnegative Matrix Factorization

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kannan, Ramakrishnan; Sukumar, Sreenivas R.; Ballard, Grey M.

NMF is a useful tool for many applications in different domains such as topic modeling in text mining, background separation in video analysis, and community detection in social networks. Despite its popularity in the data mining community, there is a lack of efficient distributed algorithms to solve the problem for big data sets. We propose a high-performance distributed-memory parallel algorithm that computes the factorization by iteratively solving alternating non-negative least squares (NLS) subproblems formore » $$\\WW$$ and $$\\HH$$. It maintains the data and factor matrices in memory (distributed across processors), uses MPI for interprocessor communication, and, in the dense case, provably minimizes communication costs (under mild assumptions). As opposed to previous implementation, our algorithm is also flexible: It performs well for both dense and sparse matrices, and allows the user to choose any one of the multiple algorithms for solving the updates to low rank factors $$\\WW$$ and $$\\HH$$ within the alternating iterations.« less

Method and structure for skewed block-cyclic distribution of lower-dimensional data arrays in higher-dimensional processor grids

DOEpatents

Chatterjee, Siddhartha [Yorktown Heights, NY; Gunnels, John A [Brewster, NY

2011-11-08

A method and structure of distributing elements of an array of data in a computer memory to a specific processor of a multi-dimensional mesh of parallel processors includes designating a distribution of elements of at least a portion of the array to be executed by specific processors in the multi-dimensional mesh of parallel processors. The pattern of the designating includes a cyclical repetitive pattern of the parallel processor mesh, as modified to have a skew in at least one dimension so that both a row of data in the array and a column of data in the array map to respective contiguous groupings of the processors such that a dimension of the contiguous groupings is greater than one.

An OpenACC-Based Unified Programming Model for Multi-accelerator Systems

DOE Office of Scientific and Technical Information (OSTI.GOV)

Kim, Jungwon; Lee, Seyong; Vetter, Jeffrey S

2015-01-01

This paper proposes a novel SPMD programming model of OpenACC. Our model integrates the different granularities of parallelism from vector-level parallelism to node-level parallelism into a single, unified model based on OpenACC. It allows programmers to write programs for multiple accelerators using a uniform programming model whether they are in shared or distributed memory systems. We implement a prototype of our model and evaluate its performance with a GPU-based supercomputer using three benchmark applications.

Newton-like methods for Navier-Stokes solution

NASA Astrophysics Data System (ADS)

Qin, N.; Xu, X.; Richards, B. E.

1992-12-01

The paper reports on Newton-like methods called SFDN-alpha-GMRES and SQN-alpha-GMRES methods that have been devised and proven as powerful schemes for large nonlinear problems typical of viscous compressible Navier-Stokes solutions. They can be applied using a partially converged solution from a conventional explicit or approximate implicit method. Developments have included the efficient parallelization of the schemes on a distributed memory parallel computer. The methods are illustrated using a RISC workstation and a transputer parallel system respectively to solve a hypersonic vortical flow.

Avalanches and generalized memory associativity in a network model for conscious and unconscious mental functioning

NASA Astrophysics Data System (ADS)

Siddiqui, Maheen; Wedemann, Roseli S.; Jensen, Henrik Jeldtoft

2018-01-01

We explore statistical characteristics of avalanches associated with the dynamics of a complex-network model, where two modules corresponding to sensorial and symbolic memories interact, representing unconscious and conscious mental processes. The model illustrates Freud's ideas regarding the neuroses and that consciousness is related with symbolic and linguistic memory activity in the brain. It incorporates the Stariolo-Tsallis generalization of the Boltzmann Machine in order to model memory retrieval and associativity. In the present work, we define and measure avalanche size distributions during memory retrieval, in order to gain insight regarding basic aspects of the functioning of these complex networks. The avalanche sizes defined for our model should be related to the time consumed and also to the size of the neuronal region which is activated, during memory retrieval. This allows the qualitative comparison of the behaviour of the distribution of cluster sizes, obtained during fMRI measurements of the propagation of signals in the brain, with the distribution of avalanche sizes obtained in our simulation experiments. This comparison corroborates the indication that the Nonextensive Statistical Mechanics formalism may indeed be more well suited to model the complex networks which constitute brain and mental structure.

SciSpark: Highly Interactive and Scalable Model Evaluation and Climate Metrics

NASA Astrophysics Data System (ADS)

Wilson, B. D.; Palamuttam, R. S.; Mogrovejo, R. M.; Whitehall, K. D.; Mattmann, C. A.; Verma, R.; Waliser, D. E.; Lee, H.

2015-12-01

Remote sensing data and climate model output are multi-dimensional arrays of massive sizes locked away in heterogeneous file formats (HDF5/4, NetCDF 3/4) and metadata models (HDF-EOS, CF) making it difficult to perform multi-stage, iterative science processing since each stage requires writing and reading data to and from disk. We are developing a lightning fast Big Data technology called SciSpark based on ApacheTM Spark under a NASA AIST grant (PI Mattmann). Spark implements the map-reduce paradigm for parallel computing on a cluster, but emphasizes in-memory computation, "spilling" to disk only as needed, and so outperforms the disk-based ApacheTM Hadoop by 100x in memory and by 10x on disk. SciSpark will enable scalable model evaluation by executing large-scale comparisons of A-Train satellite observations to model grids on a cluster of 10 to 1000 compute nodes. This 2nd generation capability for NASA's Regional Climate Model Evaluation System (RCMES) will compute simple climate metrics at interactive speeds, and extend to quite sophisticated iterative algorithms such as machine-learning based clustering of temperature PDFs, and even graph-based algorithms for searching for Mesocale Convective Complexes. We have implemented a parallel data ingest capability in which the user specifies desired variables (arrays) as several time-sorted lists of URL's (i.e. using OPeNDAP model.nc?varname, or local files). The specified variables are partitioned by time/space and then each Spark node pulls its bundle of arrays into memory to begin a computation pipeline. We also investigated the performance of several N-dim. array libraries (scala breeze, java jblas & netlib-java, and ND4J). We are currently developing science codes using ND4J and studying memory behavior on the JVM. On the pyspark side, many of our science codes already use the numpy and SciPy ecosystems. The talk will cover: the architecture of SciSpark, the design of the scientific RDD (sRDD) data structure, our efforts to integrate climate science algorithms in Python and Scala, parallel ingest and partitioning of A-Train satellite observations from HDF files and model grids from netCDF files, first parallel runs to compute comparison statistics and PDF's, and first metrics quantifying parallel speedups and memory & disk usage.

A Parallel Distributed-Memory Particle Method Enables Acquisition-Rate Segmentation of Large Fluorescence Microscopy Images

PubMed Central

Afshar, Yaser; Sbalzarini, Ivo F.

2016-01-01

Modern fluorescence microscopy modalities, such as light-sheet microscopy, are capable of acquiring large three-dimensional images at high data rate. This creates a bottleneck in computational processing and analysis of the acquired images, as the rate of acquisition outpaces the speed of processing. Moreover, images can be so large that they do not fit the main memory of a single computer. We address both issues by developing a distributed parallel algorithm for segmentation of large fluorescence microscopy images. The method is based on the versatile Discrete Region Competition algorithm, which has previously proven useful in microscopy image segmentation. The present distributed implementation decomposes the input image into smaller sub-images that are distributed across multiple computers. Using network communication, the computers orchestrate the collectively solving of the global segmentation problem. This not only enables segmentation of large images (we test images of up to 1010 pixels), but also accelerates segmentation to match the time scale of image acquisition. Such acquisition-rate image segmentation is a prerequisite for the smart microscopes of the future and enables online data compression and interactive experiments. PMID:27046144

A Parallel Distributed-Memory Particle Method Enables Acquisition-Rate Segmentation of Large Fluorescence Microscopy Images.

PubMed

Afshar, Yaser; Sbalzarini, Ivo F

2016-01-01

Modern fluorescence microscopy modalities, such as light-sheet microscopy, are capable of acquiring large three-dimensional images at high data rate. This creates a bottleneck in computational processing and analysis of the acquired images, as the rate of acquisition outpaces the speed of processing. Moreover, images can be so large that they do not fit the main memory of a single computer. We address both issues by developing a distributed parallel algorithm for segmentation of large fluorescence microscopy images. The method is based on the versatile Discrete Region Competition algorithm, which has previously proven useful in microscopy image segmentation. The present distributed implementation decomposes the input image into smaller sub-images that are distributed across multiple computers. Using network communication, the computers orchestrate the collectively solving of the global segmentation problem. This not only enables segmentation of large images (we test images of up to 10(10) pixels), but also accelerates segmentation to match the time scale of image acquisition. Such acquisition-rate image segmentation is a prerequisite for the smart microscopes of the future and enables online data compression and interactive experiments.

Implementation of the force decomposition machine for molecular dynamics simulations.

PubMed

Borštnik, Urban; Miller, Benjamin T; Brooks, Bernard R; Janežič, Dušanka

2012-09-01

We present the design and implementation of the force decomposition machine (FDM), a cluster of personal computers (PCs) that is tailored to running molecular dynamics (MD) simulations using the distributed diagonal force decomposition (DDFD) parallelization method. The cluster interconnect architecture is optimized for the communication pattern of the DDFD method. Our implementation of the FDM relies on standard commodity components even for networking. Although the cluster is meant for DDFD MD simulations, it remains general enough for other parallel computations. An analysis of several MD simulation runs on both the FDM and a standard PC cluster demonstrates that the FDM's interconnect architecture provides a greater performance compared to a more general cluster interconnect. Copyright © 2012 Elsevier Inc. All rights reserved.

Improved locality of the phase-field lattice-Boltzmann model for immiscible fluids at high density ratios

NASA Astrophysics Data System (ADS)

Fakhari, Abbas; Mitchell, Travis; Leonardi, Christopher; Bolster, Diogo

2017-11-01

Based on phase-field theory, we introduce a robust lattice-Boltzmann equation for modeling immiscible multiphase flows at large density and viscosity contrasts. Our approach is built by modifying the method proposed by Zu and He [Phys. Rev. E 87, 043301 (2013), 10.1103/PhysRevE.87.043301] in such a way as to improve efficiency and numerical stability. In particular, we employ a different interface-tracking equation based on the so-called conservative phase-field model, a simplified equilibrium distribution that decouples pressure and velocity calculations, and a local scheme based on the hydrodynamic distribution functions for calculation of the stress tensor. In addition to two distribution functions for interface tracking and recovery of hydrodynamic properties, the only nonlocal variable in the proposed model is the phase field. Moreover, within our framework there is no need to use biased or mixed difference stencils for numerical stability and accuracy at high density ratios. This not only simplifies the implementation and efficiency of the model, but also leads to a model that is better suited to parallel implementation on distributed-memory machines. Several benchmark cases are considered to assess the efficacy of the proposed model, including the layered Poiseuille flow in a rectangular channel, Rayleigh-Taylor instability, and the rise of a Taylor bubble in a duct. The numerical results are in good agreement with available numerical and experimental data.

Distributed Memory Parallel Computing with SEAWAT

NASA Astrophysics Data System (ADS)

Verkaik, J.; Huizer, S.; van Engelen, J.; Oude Essink, G.; Ram, R.; Vuik, K.

2017-12-01

Fresh groundwater reserves in coastal aquifers are threatened by sea-level rise, extreme weather conditions, increasing urbanization and associated groundwater extraction rates. To counteract these threats, accurate high-resolution numerical models are required to optimize the management of these precious reserves. The major model drawbacks are long run times and large memory requirements, limiting the predictive power of these models. Distributed memory parallel computing is an efficient technique for reducing run times and memory requirements, where the problem is divided over multiple processor cores. A new Parallel Krylov Solver (PKS) for SEAWAT is presented. PKS has recently been applied to MODFLOW and includes Conjugate Gradient (CG) and Biconjugate Gradient Stabilized (BiCGSTAB) linear accelerators. Both accelerators are preconditioned by an overlapping additive Schwarz preconditioner in a way that: a) subdomains are partitioned using Recursive Coordinate Bisection (RCB) load balancing, b) each subdomain uses local memory only and communicates with other subdomains by Message Passing Interface (MPI) within the linear accelerator, c) it is fully integrated in SEAWAT. Within SEAWAT, the PKS-CG solver replaces the Preconditioned Conjugate Gradient (PCG) solver for solving the variable-density groundwater flow equation and the PKS-BiCGSTAB solver replaces the Generalized Conjugate Gradient (GCG) solver for solving the advection-diffusion equation. PKS supports the third-order Total Variation Diminishing (TVD) scheme for computing advection. Benchmarks were performed on the Dutch national supercomputer (https://userinfo.surfsara.nl/systems/cartesius) using up to 128 cores, for a synthetic 3D Henry model (100 million cells) and the real-life Sand Engine model ( 10 million cells). The Sand Engine model was used to investigate the potential effect of the long-term morphological evolution of a large sand replenishment and climate change on fresh groundwater resources. Speed-ups up to 40 were obtained with the new PKS solver.

A divide and conquer approach to the nonsymmetric eigenvalue problem

DOE Office of Scientific and Technical Information (OSTI.GOV)

Jessup, E.R.

1991-01-01

Serial computation combined with high communication costs on distributed-memory multiprocessors make parallel implementations of the QR method for the nonsymmetric eigenvalue problem inefficient. This paper introduces an alternative algorithm for the nonsymmetric tridiagonal eigenvalue problem based on rank two tearing and updating of the matrix. The parallelism of this divide and conquer approach stems from independent solution of the updating problems. 11 refs.

Design and Implementation of Distributed Crawler System Based on Scrapy

NASA Astrophysics Data System (ADS)

Fan, Yuhao

2018-01-01

At present, some large-scale search engines at home and abroad only provide users with non-custom search services, and a single-machine web crawler cannot sovle the difficult task. In this paper, Through the study and research of the original Scrapy framework, the original Scrapy framework is improved by combining Scrapy and Redis, a distributed crawler system based on Web information Scrapy framework is designed and implemented, and Bloom Filter algorithm is applied to dupefilter modul to reduce memory consumption. The movie information captured from douban is stored in MongoDB, so that the data can be processed and analyzed. The results show that distributed crawler system based on Scrapy framework is more efficient and stable than the single-machine web crawler system.

A parallel solver for huge dense linear systems

NASA Astrophysics Data System (ADS)

Badia, J. M.; Movilla, J. L.; Climente, J. I.; Castillo, M.; Marqués, M.; Mayo, R.; Quintana-Ortí, E. S.; Planelles, J.

2011-11-01

HDSS (Huge Dense Linear System Solver) is a Fortran Application Programming Interface (API) to facilitate the parallel solution of very large dense systems to scientists and engineers. The API makes use of parallelism to yield an efficient solution of the systems on a wide range of parallel platforms, from clusters of processors to massively parallel multiprocessors. It exploits out-of-core strategies to leverage the secondary memory in order to solve huge linear systems O(100.000). The API is based on the parallel linear algebra library PLAPACK, and on its Out-Of-Core (OOC) extension POOCLAPACK. Both PLAPACK and POOCLAPACK use the Message Passing Interface (MPI) as the communication layer and BLAS to perform the local matrix operations. The API provides a friendly interface to the users, hiding almost all the technical aspects related to the parallel execution of the code and the use of the secondary memory to solve the systems. In particular, the API can automatically select the best way to store and solve the systems, depending of the dimension of the system, the number of processes and the main memory of the platform. Experimental results on several parallel platforms report high performance, reaching more than 1 TFLOP with 64 cores to solve a system with more than 200 000 equations and more than 10 000 right-hand side vectors. New version program summaryProgram title: Huge Dense System Solver (HDSS) Catalogue identifier: AEHU_v1_1 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEHU_v1_1.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 87 062 No. of bytes in distributed program, including test data, etc.: 1 069 110 Distribution format: tar.gz Programming language: Fortran90, C Computer: Parallel architectures: multiprocessors, computer clusters Operating system: Linux/Unix Has the code been vectorized or parallelized?: Yes, includes MPI primitives. RAM: Tested for up to 190 GB Classification: 6.5 External routines: MPI ( http://www.mpi-forum.org/), BLAS ( http://www.netlib.org/blas/), PLAPACK ( http://www.cs.utexas.edu/~plapack/), POOCLAPACK ( ftp://ftp.cs.utexas.edu/pub/rvdg/PLAPACK/pooclapack.ps) (code for PLAPACK and POOCLAPACK is included in the distribution). Catalogue identifier of previous version: AEHU_v1_0 Journal reference of previous version: Comput. Phys. Comm. 182 (2011) 533 Does the new version supersede the previous version?: Yes Nature of problem: Huge scale dense systems of linear equations, Ax=B, beyond standard LAPACK capabilities. Solution method: The linear systems are solved by means of parallelized routines based on the LU factorization, using efficient secondary storage algorithms when the available main memory is insufficient. Reasons for new version: In many applications we need to guarantee a high accuracy in the solution of very large linear systems and we can do it by using double-precision arithmetic. Summary of revisions: Version 1.1 Can be used to solve linear systems using double-precision arithmetic. New version of the initialization routine. The user can choose the kind of arithmetic and the values of several parameters of the environment. Running time: About 5 hours to solve a system with more than 200 000 equations and more than 10 000 right-hand side vectors using double-precision arithmetic on an eight-node commodity cluster with a total of 64 Intel cores.

A parallel algorithm for the two-dimensional time fractional diffusion equation with implicit difference method.

PubMed

Gong, Chunye; Bao, Weimin; Tang, Guojian; Jiang, Yuewen; Liu, Jie

2014-01-01

It is very time consuming to solve fractional differential equations. The computational complexity of two-dimensional fractional differential equation (2D-TFDE) with iterative implicit finite difference method is O(M(x)M(y)N(2)). In this paper, we present a parallel algorithm for 2D-TFDE and give an in-depth discussion about this algorithm. A task distribution model and data layout with virtual boundary are designed for this parallel algorithm. The experimental results show that the parallel algorithm compares well with the exact solution. The parallel algorithm on single Intel Xeon X5540 CPU runs 3.16-4.17 times faster than the serial algorithm on single CPU core. The parallel efficiency of 81 processes is up to 88.24% compared with 9 processes on a distributed memory cluster system. We do think that the parallel computing technology will become a very basic method for the computational intensive fractional applications in the near future.

Some fast elliptic solvers on parallel architectures and their complexities

NASA Technical Reports Server (NTRS)

Gallopoulos, E.; Saad, Y.

1989-01-01

The discretization of separable elliptic partial differential equations leads to linear systems with special block tridiagonal matrices. Several methods are known to solve these systems, the most general of which is the Block Cyclic Reduction (BCR) algorithm which handles equations with nonconstant coefficients. A method was recently proposed to parallelize and vectorize BCR. In this paper, the mapping of BCR on distributed memory architectures is discussed, and its complexity is compared with that of other approaches including the Alternating-Direction method. A fast parallel solver is also described, based on an explicit formula for the solution, which has parallel computational compelxity lower than that of parallel BCR.

Some fast elliptic solvers on parallel architectures and their complexities

NASA Technical Reports Server (NTRS)

Gallopoulos, E.; Saad, Youcef

1989-01-01

The discretization of separable elliptic partial differential equations leads to linear systems with special block triangular matrices. Several methods are known to solve these systems, the most general of which is the Block Cyclic Reduction (BCR) algorithm which handles equations with nonconsistant coefficients. A method was recently proposed to parallelize and vectorize BCR. Here, the mapping of BCR on distributed memory architectures is discussed, and its complexity is compared with that of other approaches, including the Alternating-Direction method. A fast parallel solver is also described, based on an explicit formula for the solution, which has parallel computational complexity lower than that of parallel BCR.

First Annual Workshop on Space Operations Automation and Robotics (SOAR 87)

NASA Technical Reports Server (NTRS)

Griffin, Sandy (Editor)

1987-01-01

Several topics relative to automation and robotics technology are discussed. Automation of checkout, ground support, and logistics; automated software development; man-machine interfaces; neural networks; systems engineering and distributed/parallel processing architectures; and artificial intelligence/expert systems are among the topics covered.

Scalable Parallel Density-based Clustering and Applications

NASA Astrophysics Data System (ADS)

Patwary, Mostofa Ali

2014-04-01

Recently, density-based clustering algorithms (DBSCAN and OPTICS) have gotten significant attention of the scientific community due to their unique capability of discovering arbitrary shaped clusters and eliminating noise data. These algorithms have several applications, which require high performance computing, including finding halos and subhalos (clusters) from massive cosmology data in astrophysics, analyzing satellite images, X-ray crystallography, and anomaly detection. However, parallelization of these algorithms are extremely challenging as they exhibit inherent sequential data access order, unbalanced workload resulting in low parallel efficiency. To break the data access sequentiality and to achieve high parallelism, we develop new parallel algorithms, both for DBSCAN and OPTICS, designed using graph algorithmic techniques. For example, our parallel DBSCAN algorithm exploits the similarities between DBSCAN and computing connected components. Using datasets containing up to a billion floating point numbers, we show that our parallel density-based clustering algorithms significantly outperform the existing algorithms, achieving speedups up to 27.5 on 40 cores on shared memory architecture and speedups up to 5,765 using 8,192 cores on distributed memory architecture. In our experiments, we found that while achieving the scalability, our algorithms produce clustering results with comparable quality to the classical algorithms.

Parallel computing method for simulating hydrological processesof large rivers under climate change

NASA Astrophysics Data System (ADS)

Wang, H.; Chen, Y.

2016-12-01

Climate change is one of the proverbial global environmental problems in the world.Climate change has altered the watershed hydrological processes in time and space distribution, especially in worldlarge rivers.Watershed hydrological process simulation based on physically based distributed hydrological model can could have better results compared with the lumped models.However, watershed hydrological process simulation includes large amount of calculations, especially in large rivers, thus needing huge computing resources that may not be steadily available for the researchers or at high expense, this seriously restricted the research and application. To solve this problem, the current parallel method are mostly parallel computing in space and time dimensions.They calculate the natural features orderly thatbased on distributed hydrological model by grid (unit, a basin) from upstream to downstream.This articleproposes ahigh-performancecomputing method of hydrological process simulation with high speedratio and parallel efficiency.It combinedthe runoff characteristics of time and space of distributed hydrological model withthe methods adopting distributed data storage, memory database, distributed computing, parallel computing based on computing power unit.The method has strong adaptability and extensibility,which means it canmake full use of the computing and storage resources under the condition of limited computing resources, and the computing efficiency can be improved linearly with the increase of computing resources .This method can satisfy the parallel computing requirements ofhydrological process simulation in small, medium and large rivers.

Line-by-line spectroscopic simulations on graphics processing units

NASA Astrophysics Data System (ADS)

Collange, Sylvain; Daumas, Marc; Defour, David

2008-01-01

We report here on software that performs line-by-line spectroscopic simulations on gases. Elaborate models (such as narrow band and correlated-K) are accurate and efficient for bands where various components are not simultaneously and significantly active. Line-by-line is probably the most accurate model in the infrared for blends of gases that contain high proportions of H 2O and CO 2 as this was the case for our prototype simulation. Our implementation on graphics processing units sustains a speedup close to 330 on computation-intensive tasks and 12 on memory intensive tasks compared to implementations on one core of high-end processors. This speedup is due to data parallelism, efficient memory access for specific patterns and some dedicated hardware operators only available in graphics processing units. It is obtained leaving most of processor resources available and it would scale linearly with the number of graphics processing units in parallel machines. Line-by-line simulation coupled with simulation of fluid dynamics was long believed to be economically intractable but our work shows that it could be done with some affordable additional resources compared to what is necessary to perform simulations on fluid dynamics alone. Program summaryProgram title: GPU4RE Catalogue identifier: ADZY_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/ADZY_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 62 776 No. of bytes in distributed program, including test data, etc.: 1 513 247 Distribution format: tar.gz Programming language: C++ Computer: x86 PC Operating system: Linux, Microsoft Windows. Compilation requires either gcc/g++ under Linux or Visual C++ 2003/2005 and Cygwin under Windows. It has been tested using gcc 4.1.2 under Ubuntu Linux 7.04 and using Visual C++ 2005 with Cygwin 1.5.24 under Windows XP. RAM: 1 gigabyte Classification: 21.2 External routines: OpenGL ( http://www.opengl.org) Nature of problem: Simulating radiative transfer on high-temperature high-pressure gases. Solution method: Line-by-line Monte-Carlo ray-tracing. Unusual features: Parallel computations are moved to the GPU. Additional comments: nVidia GeForce 7000 or ATI Radeon X1000 series graphics processing unit is required. Running time: A few minutes.

HTM Spatial Pooler With Memristor Crossbar Circuits for Sparse Biometric Recognition.

PubMed

James, Alex Pappachen; Fedorova, Irina; Ibrayev, Timur; Kudithipudi, Dhireesha

2017-06-01

Hierarchical Temporal Memory (HTM) is an online machine learning algorithm that emulates the neo-cortex. The development of a scalable on-chip HTM architecture is an open research area. The two core substructures of HTM are spatial pooler and temporal memory. In this work, we propose a new Spatial Pooler circuit design with parallel memristive crossbar arrays for the 2D columns. The proposed design was validated on two different benchmark datasets, face recognition, and speech recognition. The circuits are simulated and analyzed using a practical memristor device model and 0.18 μm IBM CMOS technology model. The databases AR, YALE, ORL, and UFI, are used to test the performance of the design in face recognition. TIMIT dataset is used for the speech recognition.

Balancing Contention and Synchronization on the Intel Paragon

NASA Technical Reports Server (NTRS)

Bokhari, Shahid H.; Nicol, David M.

1996-01-01

The Intel Paragon is a mesh-connected distributed memory parallel computer. It uses an oblivious and deterministic message routing algorithm: this permits us to develop highly optimized schedules for frequently needed communication patterns. The complete exchange is one such pattern. Several approaches are available for carrying it out on the mesh. We study an algorithm developed by Scott. This algorithm assumes that a communication link can carry one message at a time and that a node can only transmit one message at a time. It requires global synchronization to enforce a schedule of transmissions. Unfortunately global synchronization has substantial overhead on the Paragon. At the same time the powerful interconnection mechanism of this machine permits 2 or 3 messages to share a communication link with minor overhead. It can also overlap multiple message transmission from the same node to some extent. We develop a generalization of Scott's algorithm that executes complete exchange with a prescribed contention. Schedules that incur greater contention require fewer synchronization steps. This permits us to tradeoff contention against synchronization overhead. We describe the performance of this algorithm and compare it with Scott's original algorithm as well as with a naive algorithm that does not take interconnection structure into account. The Bounded contention algorithm is always better than Scott's algorithm and outperforms the naive algorithm for all but the smallest message sizes. The naive algorithm fails to work on meshes larger than 12 x 12. These results show that due consideration of processor interconnect and machine performance parameters is necessary to obtain peak performance from the Paragon and its successor mesh machines.

A Distributed-Memory Package for Dense Hierarchically Semi-Separable Matrix Computations Using Randomization

DOE Office of Scientific and Technical Information (OSTI.GOV)

Rouet, François-Henry; Li, Xiaoye S.; Ghysels, Pieter

In this paper, we present a distributed-memory library for computations with dense structured matrices. A matrix is considered structured if its off-diagonal blocks can be approximated by a rank-deficient matrix with low numerical rank. Here, we use Hierarchically Semi-Separable (HSS) representations. Such matrices appear in many applications, for example, finite-element methods, boundary element methods, and so on. Exploiting this structure allows for fast solution of linear systems and/or fast computation of matrix-vector products, which are the two main building blocks of matrix computations. The compression algorithm that we use, that computes the HSS form of an input dense matrix, reliesmore » on randomized sampling with a novel adaptive sampling mechanism. We discuss the parallelization of this algorithm and also present the parallelization of structured matrix-vector product, structured factorization, and solution routines. The efficiency of the approach is demonstrated on large problems from different academic and industrial applications, on up to 8,000 cores. Finally, this work is part of a more global effort, the STRUctured Matrices PACKage (STRUMPACK) software package for computations with sparse and dense structured matrices. Hence, although useful on their own right, the routines also represent a step in the direction of a distributed-memory sparse solver.« less

A Distributed-Memory Package for Dense Hierarchically Semi-Separable Matrix Computations Using Randomization

DOE PAGES

Rouet, François-Henry; Li, Xiaoye S.; Ghysels, Pieter; ...

2016-06-30

In this paper, we present a distributed-memory library for computations with dense structured matrices. A matrix is considered structured if its off-diagonal blocks can be approximated by a rank-deficient matrix with low numerical rank. Here, we use Hierarchically Semi-Separable (HSS) representations. Such matrices appear in many applications, for example, finite-element methods, boundary element methods, and so on. Exploiting this structure allows for fast solution of linear systems and/or fast computation of matrix-vector products, which are the two main building blocks of matrix computations. The compression algorithm that we use, that computes the HSS form of an input dense matrix, reliesmore » on randomized sampling with a novel adaptive sampling mechanism. We discuss the parallelization of this algorithm and also present the parallelization of structured matrix-vector product, structured factorization, and solution routines. The efficiency of the approach is demonstrated on large problems from different academic and industrial applications, on up to 8,000 cores. Finally, this work is part of a more global effort, the STRUctured Matrices PACKage (STRUMPACK) software package for computations with sparse and dense structured matrices. Hence, although useful on their own right, the routines also represent a step in the direction of a distributed-memory sparse solver.« less

Sparse distributed memory: Principles and operation

NASA Technical Reports Server (NTRS)

Flynn, M. J.; Kanerva, P.; Bhadkamkar, N.

1989-01-01

Sparse distributed memory is a generalized random access memory (RAM) for long (1000 bit) binary words. Such words can be written into and read from the memory, and they can also be used to address the memory. The main attribute of the memory is sensitivity to similarity, meaning that a word can be read back not only by giving the original write address but also by giving one close to it as measured by the Hamming distance between addresses. Large memories of this kind are expected to have wide use in speech recognition and scene analysis, in signal detection and verification, and in adaptive control of automated equipment, in general, in dealing with real world information in real time. The memory can be realized as a simple, massively parallel computer. Digital technology has reached a point where building large memories is becoming practical. Major design issues were resolved which were faced in building the memories. The design is described of a prototype memory with 256 bit addresses and from 8 to 128 K locations for 256 bit words. A key aspect of the design is extensive use of dynamic RAM and other standard components.

Parallel Computation of the Regional Ocean Modeling System (ROMS)

DOE Office of Scientific and Technical Information (OSTI.GOV)

Wang, P; Song, Y T; Chao, Y

2005-04-05

The Regional Ocean Modeling System (ROMS) is a regional ocean general circulation modeling system solving the free surface, hydrostatic, primitive equations over varying topography. It is free software distributed world-wide for studying both complex coastal ocean problems and the basin-to-global scale ocean circulation. The original ROMS code could only be run on shared-memory systems. With the increasing need to simulate larger model domains with finer resolutions and on a variety of computer platforms, there is a need in the ocean-modeling community to have a ROMS code that can be run on any parallel computer ranging from 10 to hundreds ofmore » processors. Recently, we have explored parallelization for ROMS using the MPI programming model. In this paper, an efficient parallelization strategy for such a large-scale scientific software package, based on an existing shared-memory computing model, is presented. In addition, scientific applications and data-performance issues on a couple of SGI systems, including Columbia, the world's third-fastest supercomputer, are discussed.« less

PISCES: An environment for parallel scientific computation

NASA Technical Reports Server (NTRS)

Pratt, T. W.

1985-01-01

The parallel implementation of scientific computing environment (PISCES) is a project to provide high-level programming environments for parallel MIMD computers. Pisces 1, the first of these environments, is a FORTRAN 77 based environment which runs under the UNIX operating system. The Pisces 1 user programs in Pisces FORTRAN, an extension of FORTRAN 77 for parallel processing. The major emphasis in the Pisces 1 design is in providing a carefully specified virtual machine that defines the run-time environment within which Pisces FORTRAN programs are executed. Each implementation then provides the same virtual machine, regardless of differences in the underlying architecture. The design is intended to be portable to a variety of architectures. Currently Pisces 1 is implemented on a network of Apollo workstations and on a DEC VAX uniprocessor via simulation of the task level parallelism. An implementation for the Flexible Computing Corp. FLEX/32 is under construction. An introduction to the Pisces 1 virtual computer and the FORTRAN 77 extensions is presented. An example of an algorithm for the iterative solution of a system of equations is given. The most notable features of the design are the provision for several granularities of parallelism in programs and the provision of a window mechanism for distributed access to large arrays of data.

Xyce parallel electronic simulator users guide, version 6.1

DOE Office of Scientific and Technical Information (OSTI.GOV)

Keiter, Eric R; Mei, Ting; Russo, Thomas V.

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been designed as a SPICE-compatible, high-performance analog circuit simulator, and has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability over the current state-of-the-art in the following areas; Capability to solve extremely large circuit problems by supporting large-scale parallel computing platforms (up to thousands of processors). This includes support for most popular parallel and serial computers; A differential-algebraic-equation (DAE) formulation, which better isolates the device model package from solver algorithms. This allows one to developmore » new types of analysis without requiring the implementation of analysis-specific device models; Device models that are specifically tailored to meet Sandia's needs, including some radiationaware devices (for Sandia users only); and Object-oriented code design and implementation using modern coding practices. Xyce is a parallel code in the most general sense of the phrase-a message passing parallel implementation-which allows it to run efficiently a wide range of computing platforms. These include serial, shared-memory and distributed-memory parallel platforms. Attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows.« less

Xyce parallel electronic simulator users' guide, Version 6.0.1.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Keiter, Eric R; Mei, Ting; Russo, Thomas V.

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been designed as a SPICE-compatible, high-performance analog circuit simulator, and has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability over the current state-of-the-art in the following areas: Capability to solve extremely large circuit problems by supporting large-scale parallel computing platforms (up to thousands of processors). This includes support for most popular parallel and serial computers. A differential-algebraic-equation (DAE) formulation, which better isolates the device model package from solver algorithms. This allows one to developmore » new types of analysis without requiring the implementation of analysis-specific device models. Device models that are specifically tailored to meet Sandias needs, including some radiationaware devices (for Sandia users only). Object-oriented code design and implementation using modern coding practices. Xyce is a parallel code in the most general sense of the phrase a message passing parallel implementation which allows it to run efficiently a wide range of computing platforms. These include serial, shared-memory and distributed-memory parallel platforms. Attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows.« less

Xyce parallel electronic simulator users guide, version 6.0.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Keiter, Eric R; Mei, Ting; Russo, Thomas V.

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been designed as a SPICE-compatible, high-performance analog circuit simulator, and has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability over the current state-of-the-art in the following areas: Capability to solve extremely large circuit problems by supporting large-scale parallel computing platforms (up to thousands of processors). This includes support for most popular parallel and serial computers. A differential-algebraic-equation (DAE) formulation, which better isolates the device model package from solver algorithms. This allows one to developmore » new types of analysis without requiring the implementation of analysis-specific device models. Device models that are specifically tailored to meet Sandias needs, including some radiationaware devices (for Sandia users only). Object-oriented code design and implementation using modern coding practices. Xyce is a parallel code in the most general sense of the phrase a message passing parallel implementation which allows it to run efficiently a wide range of computing platforms. These include serial, shared-memory and distributed-memory parallel platforms. Attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows.« less

JuxtaView - A tool for interactive visualization of large imagery on scalable tiled displays

USGS Publications Warehouse

Krishnaprasad, N.K.; Vishwanath, V.; Venkataraman, S.; Rao, A.G.; Renambot, L.; Leigh, J.; Johnson, A.E.; Davis, B.

2004-01-01

JuxtaView is a cluster-based application for viewing ultra-high-resolution images on scalable tiled displays. We present in JuxtaView, a new parallel computing and distributed memory approach for out-of-core montage visualization, using LambdaRAM, a software-based network-level cache system. The ultimate goal of JuxtaView is to enable a user to interactively roam through potentially terabytes of distributed, spatially referenced image data such as those from electron microscopes, satellites and aerial photographs. In working towards this goal, we describe our first prototype implemented over a local area network, where the image is distributed using LambdaRAM, on the memory of all nodes of a PC cluster driving a tiled display wall. Aggressive pre-fetching schemes employed by LambdaRAM help to reduce latency involved in remote memory access. We compare LambdaRAM with a more traditional memory-mapped file approach for out-of-core visualization. ?? 2004 IEEE.

Ropes: Support for collective opertions among distributed threads

NASA Technical Reports Server (NTRS)

Haines, Matthew; Mehrotra, Piyush; Cronk, David

1995-01-01

Lightweight threads are becoming increasingly useful in supporting parallelism and asynchronous control structures in applications and language implementations. Recently, systems have been designed and implemented to support interprocessor communication between lightweight threads so that threads can be exploited in a distributed memory system. Their use, in this setting, has been largely restricted to supporting latency hiding techniques and functional parallelism within a single application. However, to execute data parallel codes independent of other threads in the system, collective operations and relative indexing among threads are required. This paper describes the design of ropes: a scoping mechanism for collective operations and relative indexing among threads. We present the design of ropes in the context of the Chant system, and provide performance results evaluating our initial design decisions.

Exploiting Thread Parallelism for Ocean Modeling on Cray XC Supercomputers

DOE Office of Scientific and Technical Information (OSTI.GOV)

Sarje, Abhinav; Jacobsen, Douglas W.; Williams, Samuel W.

The incorporation of increasing core counts in modern processors used to build state-of-the-art supercomputers is driving application development towards exploitation of thread parallelism, in addition to distributed memory parallelism, with the goal of delivering efficient high-performance codes. In this work we describe the exploitation of threading and our experiences with it with respect to a real-world ocean modeling application code, MPAS-Ocean. We present detailed performance analysis and comparisons of various approaches and configurations for threading on the Cray XC series supercomputers.

Design consideration in constructing high performance embedded Knowledge-Based Systems (KBS)

NASA Technical Reports Server (NTRS)

Dalton, Shelly D.; Daley, Philip C.

1988-01-01

As the hardware trends for artificial intelligence (AI) involve more and more complexity, the process of optimizing the computer system design for a particular problem will also increase in complexity. Space applications of knowledge based systems (KBS) will often require an ability to perform both numerically intensive vector computations and real time symbolic computations. Although parallel machines can theoretically achieve the speeds necessary for most of these problems, if the application itself is not highly parallel, the machine's power cannot be utilized. A scheme is presented which will provide the computer systems engineer with a tool for analyzing machines with various configurations of array, symbolic, scaler, and multiprocessors. High speed networks and interconnections make customized, distributed, intelligent systems feasible for the application of AI in space. The method presented can be used to optimize such AI system configurations and to make comparisons between existing computer systems. It is an open question whether or not, for a given mission requirement, a suitable computer system design can be constructed for any amount of money.

Parallel machine architecture and compiler design facilities

NASA Technical Reports Server (NTRS)

Kuck, David J.; Yew, Pen-Chung; Padua, David; Sameh, Ahmed; Veidenbaum, Alex

1990-01-01

The objective is to provide an integrated simulation environment for studying and evaluating various issues in designing parallel systems, including machine architectures, parallelizing compiler techniques, and parallel algorithms. The status of Delta project (which objective is to provide a facility to allow rapid prototyping of parallelized compilers that can target toward different machine architectures) is summarized. Included are the surveys of the program manipulation tools developed, the environmental software supporting Delta, and the compiler research projects in which Delta has played a role.

DCL System Using Deep Learning Approaches for Land-based or Ship-based Real-Time Recognition and Localization of Marine Mammals

DTIC Science & Technology

2012-09-30

platform (HPC) was developed, called the HPC-Acoustic Data Accelerator, or HPC-ADA for short. The HPC-ADA was designed based on fielded systems [1-4...software (Detection cLassificaiton for MAchine learning - High Peformance Computing). The software package was designed to utilize parallel and...Sedna [7] and is designed using a parallel architecture2, allowing existing algorithms to distribute to the various processing nodes with minimal changes

The Harmonic Sieve: A Novel Application of Fourier Analysis to Machine Learning Theory and Practice.

DTIC Science & Technology

1995-08-23

1987. [Ros62] F. Rosenblatt. Principles of Neurodynamics : Perceptrons and the Theory of Brain Mechanisms. Spartan Books, 1962. [RHW86] D. E...editors, Parallel Distributed Processing: Explorations in the Micro structures of Cognition , volume 1, chapter 8, pages 318-362. MIT Press, 1986

Turbomachinery CFD on parallel computers

NASA Technical Reports Server (NTRS)

Blech, Richard A.; Milner, Edward J.; Quealy, Angela; Townsend, Scott E.

1992-01-01

The role of multistage turbomachinery simulation in the development of propulsion system models is discussed. Particularly, the need for simulations with higher fidelity and faster turnaround time is highlighted. It is shown how such fast simulations can be used in engineering-oriented environments. The use of parallel processing to achieve the required turnaround times is discussed. Current work by several researchers in this area is summarized. Parallel turbomachinery CFD research at the NASA Lewis Research Center is then highlighted. These efforts are focused on implementing the average-passage turbomachinery model on MIMD, distributed memory parallel computers. Performance results are given for inviscid, single blade row and viscous, multistage applications on several parallel computers, including networked workstations.

PANDA: A distributed multiprocessor operating system

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chubb, P.

1989-01-01

PANDA is a design for a distributed multiprocessor and an operating system. PANDA is designed to allow easy expansion of both hardware and software. As such, the PANDA kernel provides only message passing and memory and process management. The other features needed for the system (device drivers, secondary storage management, etc.) are provided as replaceable user tasks. The thesis presents PANDA's design and implementation, both hardware and software. PANDA uses multiple 68010 processors sharing memory on a VME bus, each such node potentially connected to others via a high speed network. The machine is completely homogeneous: there are no differencesmore » between processors that are detectable by programs running on the machine. A single two-processor node has been constructed. Each processor contains memory management circuits designed to allow processors to share page tables safely. PANDA presents a programmers' model similar to the hardware model: a job is divided into multiple tasks, each having its own address space. Within each task, multiple processes share code and data. Tasks can send messages to each other, and set up virtual circuits between themselves. Peripheral devices such as disc drives are represented within PANDA by tasks. PANDA divides secondary storage into volumes, each volume being accessed by a volume access task, or VAT. All knowledge about the way that data is stored on a disc is kept in its volume's VAT. The design is such that PANDA should provide a useful testbed for file systems and device drivers, as these can be installed without recompiling PANDA itself, and without rebooting the machine.« less

Studying an Eulerian Computer Model on Different High-performance Computer Platforms and Some Applications

NASA Astrophysics Data System (ADS)

Georgiev, K.; Zlatev, Z.

2010-11-01

The Danish Eulerian Model (DEM) is an Eulerian model for studying the transport of air pollutants on large scale. Originally, the model was developed at the National Environmental Research Institute of Denmark. The model computational domain covers Europe and some neighbour parts belong to the Atlantic Ocean, Asia and Africa. If DEM model is to be applied by using fine grids, then its discretization leads to a huge computational problem. This implies that such a model as DEM must be run only on high-performance computer architectures. The implementation and tuning of such a complex large-scale model on each different computer is a non-trivial task. Here, some comparison results of running of this model on different kind of vector (CRAY C92A, Fujitsu, etc.), parallel computers with distributed memory (IBM SP, CRAY T3E, Beowulf clusters, Macintosh G4 clusters, etc.), parallel computers with shared memory (SGI Origin, SUN, etc.) and parallel computers with two levels of parallelism (IBM SMP, IBM BlueGene/P, clusters of multiprocessor nodes, etc.) will be presented. The main idea in the parallel version of DEM is domain partitioning approach. Discussions according to the effective use of the cache and hierarchical memories of the modern computers as well as the performance, speed-ups and efficiency achieved will be done. The parallel code of DEM, created by using MPI standard library, appears to be highly portable and shows good efficiency and scalability on different kind of vector and parallel computers. Some important applications of the computer model output are presented in short.

Memory consolidation reconfigures neural pathways involved in the suppression of emotional memories

PubMed Central

Liu, Yunzhe; Lin, Wanjun; Liu, Chao; Luo, Yuejia; Wu, Jianhui; Bayley, Peter J.; Qin, Shaozheng

2016-01-01

The ability to suppress unwanted emotional memories is crucial for human mental health. Through consolidation over time, emotional memories often become resistant to change. However, how consolidation impacts the effectiveness of emotional memory suppression is still unknown. Using event-related fMRI while concurrently recording skin conductance, we investigated the neurobiological processes underlying the suppression of aversive memories before and after overnight consolidation. Here we report that consolidated aversive memories retain their emotional reactivity and become more resistant to suppression. Suppression of consolidated memories involves higher prefrontal engagement, and less concomitant hippocampal and amygdala disengagement. In parallel, we show a shift away from hippocampal-dependent representational patterns to distributed neocortical representational patterns in the suppression of aversive memories after consolidation. These findings demonstrate rapid changes in emotional memory organization with overnight consolidation, and suggest possible neurobiological bases underlying the resistance to suppression of emotional memories in affective disorders. PMID:27898050

Large-Scale Distributed Computational Fluid Dynamics on the Information Power Grid Using Globus

NASA Technical Reports Server (NTRS)

Barnard, Stephen; Biswas, Rupak; Saini, Subhash; VanderWijngaart, Robertus; Yarrow, Maurice; Zechtzer, Lou; Foster, Ian; Larsson, Olle

1999-01-01

This paper describes an experiment in which a large-scale scientific application development for tightly-coupled parallel machines is adapted to the distributed execution environment of the Information Power Grid (IPG). A brief overview of the IPG and a description of the computational fluid dynamics (CFD) algorithm are given. The Globus metacomputing toolkit is used as the enabling device for the geographically-distributed computation. Modifications related to latency hiding and Load balancing were required for an efficient implementation of the CFD application in the IPG environment. Performance results on a pair of SGI Origin 2000 machines indicate that real scientific applications can be effectively implemented on the IPG; however, a significant amount of continued effort is required to make such an environment useful and accessible to scientists and engineers.

Dome: Distributed Object Migration Environment

DTIC Science & Technology

1994-05-01

Best Available Copy AD-A281 134 Computer Science Dome: Distributed object migration environment Adam Beguelin Erik Seligman Michael Starkey May 1994...Beguelin Erik Seligman Michael Starkey May 1994 CMU-CS-94-153 School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 Abstract Dome... Linda [4], Isis [2], and Express [6] allow a pro- grammer to treat a heterogeneous network of computers as a parallel machine. These tools allow the

A Distributed Parallel Genetic Algorithm of Placement Strategy for Virtual Machines Deployment on Cloud Platform

PubMed Central

Dong, Yu-Shuang; Xu, Gao-Chao; Fu, Xiao-Dong

2014-01-01

The cloud platform provides various services to users. More and more cloud centers provide infrastructure as the main way of operating. To improve the utilization rate of the cloud center and to decrease the operating cost, the cloud center provides services according to requirements of users by sharding the resources with virtualization. Considering both QoS for users and cost saving for cloud computing providers, we try to maximize performance and minimize energy cost as well. In this paper, we propose a distributed parallel genetic algorithm (DPGA) of placement strategy for virtual machines deployment on cloud platform. It executes the genetic algorithm parallelly and distributedly on several selected physical hosts in the first stage. Then it continues to execute the genetic algorithm of the second stage with solutions obtained from the first stage as the initial population. The solution calculated by the genetic algorithm of the second stage is the optimal one of the proposed approach. The experimental results show that the proposed placement strategy of VM deployment can ensure QoS for users and it is more effective and more energy efficient than other placement strategies on the cloud platform. PMID:25097872

A distributed parallel genetic algorithm of placement strategy for virtual machines deployment on cloud platform.

PubMed

Dong, Yu-Shuang; Xu, Gao-Chao; Fu, Xiao-Dong

2014-01-01

The cloud platform provides various services to users. More and more cloud centers provide infrastructure as the main way of operating. To improve the utilization rate of the cloud center and to decrease the operating cost, the cloud center provides services according to requirements of users by sharding the resources with virtualization. Considering both QoS for users and cost saving for cloud computing providers, we try to maximize performance and minimize energy cost as well. In this paper, we propose a distributed parallel genetic algorithm (DPGA) of placement strategy for virtual machines deployment on cloud platform. It executes the genetic algorithm parallelly and distributedly on several selected physical hosts in the first stage. Then it continues to execute the genetic algorithm of the second stage with solutions obtained from the first stage as the initial population. The solution calculated by the genetic algorithm of the second stage is the optimal one of the proposed approach. The experimental results show that the proposed placement strategy of VM deployment can ensure QoS for users and it is more effective and more energy efficient than other placement strategies on the cloud platform.

The Distributed Diagonal Force Decomposition Method for Parallelizing Molecular Dynamics Simulations

PubMed Central

Boršnik, Urban; Miller, Benjamin T.; Brooks, Bernard R.; Janežič, Dušanka

2011-01-01

Parallelization is an effective way to reduce the computational time needed for molecular dynamics simulations. We describe a new parallelization method, the distributed-diagonal force decomposition method, with which we extend and improve the existing force decomposition methods. Our new method requires less data communication during molecular dynamics simulations than replicated data and current force decomposition methods, increasing the parallel efficiency. It also dynamically load-balances the processors' computational load throughout the simulation. The method is readily implemented in existing molecular dynamics codes and it has been incorporated into the CHARMM program, allowing its immediate use in conjunction with the many molecular dynamics simulation techniques that are already present in the program. We also present the design of the Force Decomposition Machine, a cluster of personal computers and networks that is tailored to running molecular dynamics simulations using the distributed diagonal force decomposition method. The design is expandable and provides various degrees of fault resilience. This approach is easily adaptable to computers with Graphics Processing Units because it is independent of the processor type being used. PMID:21793007

Design considerations for parallel graphics libraries

NASA Technical Reports Server (NTRS)

Crockett, Thomas W.

1994-01-01

Applications which run on parallel supercomputers are often characterized by massive datasets. Converting these vast collections of numbers to visual form has proven to be a powerful aid to comprehension. For a variety of reasons, it may be desirable to provide this visual feedback at runtime. One way to accomplish this is to exploit the available parallelism to perform graphics operations in place. In order to do this, we need appropriate parallel rendering algorithms and library interfaces. This paper provides a tutorial introduction to some of the issues which arise in designing parallel graphics libraries and their underlying rendering algorithms. The focus is on polygon rendering for distributed memory message-passing systems. We illustrate our discussion with examples from PGL, a parallel graphics library which has been developed on the Intel family of parallel systems.

Effects of Ordering Strategies and Programming Paradigms on Sparse Matrix Computations

NASA Technical Reports Server (NTRS)

Oliker, Leonid; Li, Xiaoye; Husbands, Parry; Biswas, Rupak; Biegel, Bryan (Technical Monitor)

2002-01-01

The Conjugate Gradient (CG) algorithm is perhaps the best-known iterative technique to solve sparse linear systems that are symmetric and positive definite. For systems that are ill-conditioned, it is often necessary to use a preconditioning technique. In this paper, we investigate the effects of various ordering and partitioning strategies on the performance of parallel CG and ILU(O) preconditioned CG (PCG) using different programming paradigms and architectures. Results show that for this class of applications: ordering significantly improves overall performance on both distributed and distributed shared-memory systems, that cache reuse may be more important than reducing communication, that it is possible to achieve message-passing performance using shared-memory constructs through careful data ordering and distribution, and that a hybrid MPI+OpenMP paradigm increases programming complexity with little performance gains. A implementation of CG on the Cray MTA does not require special ordering or partitioning to obtain high efficiency and scalability, giving it a distinct advantage for adaptive applications; however, it shows limited scalability for PCG due to a lack of thread level parallelism.

Experiences with hypercube operating system instrumentation

NASA Technical Reports Server (NTRS)

Reed, Daniel A.; Rudolph, David C.

1989-01-01

The difficulties in conceptualizing the interactions among a large number of processors make it difficult both to identify the sources of inefficiencies and to determine how a parallel program could be made more efficient. This paper describes an instrumentation system that can trace the execution of distributed memory parallel programs by recording the occurrence of parallel program events. The resulting event traces can be used to compile summary statistics that provide a global view of program performance. In addition, visualization tools permit the graphic display of event traces. Visual presentation of performance data is particularly useful, indeed, necessary for large-scale parallel computers; the enormous volume of performance data mandates visual display.

Distributed Sensor Systems and Electromechanical Analog Facility

DTIC Science & Technology

1980-01-01

interfaces (parallel I/O, modems , etc.) real time operating systems (perhaps a short survey of what is available in the industry today), data...consists of a LSI-11 microprocessor, 56K bytes of memory, and serial and parallel I/O boards. 2.1.7 Disk controller The standard disk controller...with MTS via the modems connected to the LSI-lls. This pseudodevice cannot be reassigned. OSWIT I/O AND INTERRUPT STRUCTURE 137 OSWIT

Experiments and Analyses of Data Transfers Over Wide-Area Dedicated Connections

DOE Office of Scientific and Technical Information (OSTI.GOV)

Rao, Nageswara S.; Liu, Qiang; Sen, Satyabrata

Dedicated wide-area network connections are increasingly employed in high-performance computing and big data scenarios. One might expect the performance and dynamics of data transfers over such connections to be easy to analyze due to the lack of competing traffic. However, non-linear transport dynamics and end-system complexities (e.g., multi-core hosts and distributed filesystems) can in fact make analysis surprisingly challenging. We present extensive measurements of memory-to-memory and disk-to-disk file transfers over 10 Gbps physical and emulated connections with 0–366 ms round trip times (RTTs). For memory-to-memory transfers, profiles of both TCP and UDT throughput as a function of RTT show concavemore » and convex regions; large buffer sizes and more parallel flows lead to wider concave regions, which are highly desirable. TCP and UDT both also display complex throughput dynamics, as indicated by their Poincare maps and Lyapunov exponents. For disk-to-disk transfers, we determine that high throughput can be achieved via a combination of parallel I/O threads, parallel network threads, and direct I/O mode. Our measurements also show that Lustre filesystems can be mounted over long-haul connections using LNet routers, although challenges remain in jointly optimizing file I/O and transport method parameters to achieve peak throughput.« less

Dynamic data distributions in Vienna Fortran

NASA Technical Reports Server (NTRS)

Chapman, Barbara; Mehrotra, Piyush; Moritsch, Hans; Zima, Hans

1993-01-01

Vienna Fortran is a machine-independent language extension of Fortran, which is based upon the Single-Program-Multiple-Data (SPMD) paradigm and allows the user to write programs for distributed-memory systems using global addresses. The language features focus mainly on the issue of distributing data across virtual processor structures. Those features of Vienna Fortran that allow the data distributions of arrays to change dynamically, depending on runtime conditions are discussed. The relevant language features are discussed, their implementation is outlined, and how they may be used in applications is described.

User-Defined Data Distributions in High-Level Programming Languages

NASA Technical Reports Server (NTRS)

Diaconescu, Roxana E.; Zima, Hans P.

2006-01-01

One of the characteristic features of today s high performance computing systems is a physically distributed memory. Efficient management of locality is essential for meeting key performance requirements for these architectures. The standard technique for dealing with this issue has involved the extension of traditional sequential programming languages with explicit message passing, in the context of a processor-centric view of parallel computation. This has resulted in complex and error-prone assembly-style codes in which algorithms and communication are inextricably interwoven. This paper presents a high-level approach to the design and implementation of data distributions. Our work is motivated by the need to improve the current parallel programming methodology by introducing a paradigm supporting the development of efficient and reusable parallel code. This approach is currently being implemented in the context of a new programming language called Chapel, which is designed in the HPCS project Cascade.

Parallelization and checkpointing of GPU applications through program transformation

DOE Office of Scientific and Technical Information (OSTI.GOV)

Solano-Quinde, Lizandro Damian

2012-01-01

GPUs have emerged as a powerful tool for accelerating general-purpose applications. The availability of programming languages that makes writing general-purpose applications for running on GPUs tractable have consolidated GPUs as an alternative for accelerating general purpose applications. Among the areas that have benefited from GPU acceleration are: signal and image processing, computational fluid dynamics, quantum chemistry, and, in general, the High Performance Computing (HPC) Industry. In order to continue to exploit higher levels of parallelism with GPUs, multi-GPU systems are gaining popularity. In this context, single-GPU applications are parallelized for running in multi-GPU systems. Furthermore, multi-GPU systems help to solvemore » the GPU memory limitation for applications with large application memory footprint. Parallelizing single-GPU applications has been approached by libraries that distribute the workload at runtime, however, they impose execution overhead and are not portable. On the other hand, on traditional CPU systems, parallelization has been approached through application transformation at pre-compile time, which enhances the application to distribute the workload at application level and does not have the issues of library-based approaches. Hence, a parallelization scheme for GPU systems based on application transformation is needed. Like any computing engine of today, reliability is also a concern in GPUs. GPUs are vulnerable to transient and permanent failures. Current checkpoint/restart techniques are not suitable for systems with GPUs. Checkpointing for GPU systems present new and interesting challenges, primarily due to the natural differences imposed by the hardware design, the memory subsystem architecture, the massive number of threads, and the limited amount of synchronization among threads. Therefore, a checkpoint/restart technique suitable for GPU systems is needed. The goal of this work is to exploit higher levels of parallelism and to develop support for application-level fault tolerance in applications using multiple GPUs. Our techniques reduce the burden of enhancing single-GPU applications to support these features. To achieve our goal, this work designs and implements a framework for enhancing a single-GPU OpenCL application through application transformation.« less

Feasibility of using the Massively Parallel Processor for large eddy simulations and other Computational Fluid Dynamics applications

NASA Technical Reports Server (NTRS)

Bruno, John

1984-01-01

The results of an investigation into the feasibility of using the MPP for direct and large eddy simulations of the Navier-Stokes equations is presented. A major part of this study was devoted to the implementation of two of the standard numerical algorithms for CFD. These implementations were not run on the Massively Parallel Processor (MPP) since the machine delivered to NASA Goddard does not have sufficient capacity. Instead, a detailed implementation plan was designed and from these were derived estimates of the time and space requirements of the algorithms on a suitably configured MPP. In addition, other issues related to the practical implementation of these algorithms on an MPP-like architecture were considered; namely, adaptive grid generation, zonal boundary conditions, the table lookup problem, and the software interface. Performance estimates show that the architectural components of the MPP, the Staging Memory and the Array Unit, appear to be well suited to the numerical algorithms of CFD. This combined with the prospect of building a faster and larger MMP-like machine holds the promise of achieving sustained gigaflop rates that are required for the numerical simulations in CFD.

MSAProbs-MPI: parallel multiple sequence aligner for distributed-memory systems.

PubMed

González-Domínguez, Jorge; Liu, Yongchao; Touriño, Juan; Schmidt, Bertil

2016-12-15

MSAProbs is a state-of-the-art protein multiple sequence alignment tool based on hidden Markov models. It can achieve high alignment accuracy at the expense of relatively long runtimes for large-scale input datasets. In this work we present MSAProbs-MPI, a distributed-memory parallel version of the multithreaded MSAProbs tool that is able to reduce runtimes by exploiting the compute capabilities of common multicore CPU clusters. Our performance evaluation on a cluster with 32 nodes (each containing two Intel Haswell processors) shows reductions in execution time of over one order of magnitude for typical input datasets. Furthermore, MSAProbs-MPI using eight nodes is faster than the GPU-accelerated QuickProbs running on a Tesla K20. Another strong point is that MSAProbs-MPI can deal with large datasets for which MSAProbs and QuickProbs might fail due to time and memory constraints, respectively. Source code in C ++ and MPI running on Linux systems as well as a reference manual are available at http://msaprobs.sourceforge.net CONTACT: jgonzalezd@udc.esSupplementary information: Supplementary data are available at Bioinformatics online. © The Author 2016. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.

Investigating a method of producing "red and dead" galaxies

NASA Astrophysics Data System (ADS)

Skory, Stephen

2010-08-01

In optical wavelengths, galaxies are observed to be either red or blue. The overall color of a galaxy is due to the distribution of the ages of its stellar population. Galaxies with currently active star formation appear blue, while those with no recent star formation at all (greater than about a Gyr) have only old, red stars. This strong bimodality has lead to the idea of star formation quenching, and various proposed physical mechanisms. In this dissertation, I attempt to reproduce with Enzo the results of Naab et al. (2007), in which red and dead galaxies are formed using gravitational quenching, rather than with one of the more typical methods of quenching. My initial attempts are unsuccessful, and I explore the reasons why I think they failed. Then using simpler methods better suited to Enzo + AMR, I am successful in producing a galaxy that appears to be similar in color and formation history to those in Naab et al. However, quenching is achieved using unphysically high star formation efficiencies, which is a different mechanism than Naab et al. suggests. Preliminary results of a much higher resolution, follow-on simulation of the above show some possible contradiction with the results of Naab et al. Cold gas is streaming into the galaxy to fuel starbursts, while at a similar epoch the galaxies in Naab et al. have largely already ceased forming stars in the galaxy. On the other hand, the results of the high resolution simulation are qualitatively similar to other works in the literature that show a somewhat different gravitational quenching mechanism than Naab et al. I also discuss my work using halo finders to analyze simulated cosmological data, and my work improving the Enzo/AMR analysis tool "yt". This includes two parallelizations of the halo finder HOP (Eisenstein and Hut, 1998) which allows analysis of very large cosmological datasets on parallel machines. The first version is "yt-HOP," which works well for datasets between about 2563 and 5123 particles, but has memory bottlenecks as the datasets get larger. These bottlenecks inspired the second version, "Parallel HOP," which is a fully parallelized method and implementation of HOP that has worked on datasets with more than 20483 particles on hundreds of processing cores. Both methods are described in detail, as are the various effects of performance-related runtime options. Additionally, both halo finders are subjected to a full suite of performance benchmarks varying both dataset sizes and computational resources used. I conclude with descriptions of four new tools I added to yt. A Parallel Structure Function Generator allows analysis of two-point functions, such as correlation functions, using memory- and workload-parallelism. A Parallel Merger Tree Generator leverages the parallel halo finders in yt, such as Parallel HOP, to build the merger tree of halos in a cosmological simulation, and outputs the result to a SQLite database for simple and powerful data extraction. A Star Particle Analysis toolkit takes a group of star particles and can output the rate of formation as a function of time, and/or a synthetic Spectral Energy Distribution (S.E.D.) using the Bruzual and Charlot (2003) data tables. Finally, a Halo Mass Function toolkit takes as input a list of halo masses and can output the halo mass function for the halos, as well as an analytical fit for those halos using several previously published fits.

A distributed-memory approximation algorithm for maximum weight perfect bipartite matching

DOE Office of Scientific and Technical Information (OSTI.GOV)

Azad, Ariful; Buluc, Aydin; Li, Xiaoye S.

We design and implement an efficient parallel approximation algorithm for the problem of maximum weight perfect matching in bipartite graphs, i.e. the problem of finding a set of non-adjacent edges that covers all vertices and has maximum weight. This problem differs from the maximum weight matching problem, for which scalable approximation algorithms are known. It is primarily motivated by finding good pivots in scalable sparse direct solvers before factorization where sequential implementations of maximum weight perfect matching algorithms, such as those available in MC64, are widely used due to the lack of scalable alternatives. To overcome this limitation, we proposemore » a fully parallel distributed memory algorithm that first generates a perfect matching and then searches for weightaugmenting cycles of length four in parallel and iteratively augments the matching with a vertex disjoint set of such cycles. For most practical problems the weights of the perfect matchings generated by our algorithm are very close to the optimum. An efficient implementation of the algorithm scales up to 256 nodes (17,408 cores) on a Cray XC40 supercomputer and can solve instances that are too large to be handled by a single node using the sequential algorithm.« less

Implementation of a fully-balanced periodic tridiagonal solver on a parallel distributed memory architecture

NASA Technical Reports Server (NTRS)

Eidson, T. M.; Erlebacher, G.

1994-01-01

While parallel computers offer significant computational performance, it is generally necessary to evaluate several programming strategies. Two programming strategies for a fairly common problem - a periodic tridiagonal solver - are developed and evaluated. Simple model calculations as well as timing results are presented to evaluate the various strategies. The particular tridiagonal solver evaluated is used in many computational fluid dynamic simulation codes. The feature that makes this algorithm unique is that these simulation codes usually require simultaneous solutions for multiple right-hand-sides (RHS) of the system of equations. Each RHS solutions is independent and thus can be computed in parallel. Thus a Gaussian elimination type algorithm can be used in a parallel computation and the more complicated approaches such as cyclic reduction are not required. The two strategies are a transpose strategy and a distributed solver strategy. For the transpose strategy, the data is moved so that a subset of all the RHS problems is solved on each of the several processors. This usually requires significant data movement between processor memories across a network. The second strategy attempts to have the algorithm allow the data across processor boundaries in a chained manner. This usually requires significantly less data movement. An approach to accomplish this second strategy in a near-perfect load-balanced manner is developed. In addition, an algorithm will be shown to directly transform a sequential Gaussian elimination type algorithm into the parallel chained, load-balanced algorithm.

Analytical modeling and feasibility study of a multi-GPU cloud-based server (MGCS) framework for non-voxel-based dose calculations.

PubMed

Neylon, J; Min, Y; Kupelian, P; Low, D A; Santhanam, A

2017-04-01

In this paper, a multi-GPU cloud-based server (MGCS) framework is presented for dose calculations, exploring the feasibility of remote computing power for parallelization and acceleration of computationally and time intensive radiotherapy tasks in moving toward online adaptive therapies. An analytical model was developed to estimate theoretical MGCS performance acceleration and intelligently determine workload distribution. Numerical studies were performed with a computing setup of 14 GPUs distributed over 4 servers interconnected by a 1 Gigabits per second (Gbps) network. Inter-process communication methods were optimized to facilitate resource distribution and minimize data transfers over the server interconnect. The analytically predicted computation time predicted matched experimentally observations within 1-5 %. MGCS performance approached a theoretical limit of acceleration proportional to the number of GPUs utilized when computational tasks far outweighed memory operations. The MGCS implementation reproduced ground-truth dose computations with negligible differences, by distributing the work among several processes and implemented optimization strategies. The results showed that a cloud-based computation engine was a feasible solution for enabling clinics to make use of fast dose calculations for advanced treatment planning and adaptive radiotherapy. The cloud-based system was able to exceed the performance of a local machine even for optimized calculations, and provided significant acceleration for computationally intensive tasks. Such a framework can provide access to advanced technology and computational methods to many clinics, providing an avenue for standardization across institutions without the requirements of purchasing, maintaining, and continually updating hardware.

A Cerebellar-model Associative Memory as a Generalized Random-access Memory

NASA Technical Reports Server (NTRS)

Kanerva, Pentti

1989-01-01

A versatile neural-net model is explained in terms familiar to computer scientists and engineers. It is called the sparse distributed memory, and it is a random-access memory for very long words (for patterns with thousands of bits). Its potential utility is the result of several factors: (1) a large pattern representing an object or a scene or a moment can encode a large amount of information about what it represents; (2) this information can serve as an address to the memory, and it can also serve as data; (3) the memory is noise tolerant--the information need not be exact; (4) the memory can be made arbitrarily large and hence an arbitrary amount of information can be stored in it; and (5) the architecture is inherently parallel, allowing large memories to be fast. Such memories can become important components of future computers.

Experiences using OpenMP based on Computer Directed Software DSM on a PC Cluster

NASA Technical Reports Server (NTRS)

Hess, Matthias; Jost, Gabriele; Mueller, Matthias; Ruehle, Roland

2003-01-01

In this work we report on our experiences running OpenMP programs on a commodity cluster of PCs running a software distributed shared memory (DSM) system. We describe our test environment and report on the performance of a subset of the NAS Parallel Benchmarks that have been automaticaly parallelized for OpenMP. We compare the performance of the OpenMP implementations with that of their message passing counterparts and discuss performance differences.

Flexible language constructs for large parallel programs

NASA Technical Reports Server (NTRS)

Rosing, Matthew; Schnabel, Robert

1993-01-01

The goal of the research described is to develop flexible language constructs for writing large data parallel numerical programs for distributed memory (MIMD) multiprocessors. Previously, several models have been developed to support synchronization and communication. Models for global synchronization include SIMD (Single Instruction Multiple Data), SPMD (Single Program Multiple Data), and sequential programs annotated with data distribution statements. The two primary models for communication include implicit communication based on shared memory and explicit communication based on messages. None of these models by themselves seem sufficient to permit the natural and efficient expression of the variety of algorithms that occur in large scientific computations. An overview of a new language that combines many of these programming models in a clean manner is given. This is done in a modular fashion such that different models can be combined to support large programs. Within a module, the selection of a model depends on the algorithm and its efficiency requirements. An overview of the language and discussion of some of the critical implementation details is given.

Optimizing Irregular Applications for Energy and Performance on the Tilera Many-core Architecture

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chavarría-Miranda, Daniel; Panyala, Ajay R.; Halappanavar, Mahantesh

Optimizing applications simultaneously for energy and performance is a complex problem. High performance, parallel, irregular applications are notoriously hard to optimize due to their data-dependent memory accesses, lack of structured locality and complex data structures and code patterns. Irregular kernels are growing in importance in applications such as machine learning, graph analytics and combinatorial scientific computing. Performance- and energy-efficient implementation of these kernels on modern, energy efficient, multicore and many-core platforms is therefore an important and challenging problem. We present results from optimizing two irregular applications { the Louvain method for community detection (Grappolo), and high-performance conjugate gradient (HPCCG) {more » on the Tilera many-core system. We have significantly extended MIT's OpenTuner auto-tuning framework to conduct a detailed study of platform-independent and platform-specific optimizations to improve performance as well as reduce total energy consumption. We explore the optimization design space along three dimensions: memory layout schemes, compiler-based code transformations, and optimization of parallel loop schedules. Using auto-tuning, we demonstrate whole node energy savings of up to 41% relative to a baseline instantiation, and up to 31% relative to manually optimized variants.« less

Livermore Big Artificial Neural Network Toolkit

DOE Office of Scientific and Technical Information (OSTI.GOV)

Essen, Brian Van; Jacobs, Sam; Kim, Hyojin

2016-07-01

LBANN is a toolkit that is designed to train artificial neural networks efficiently on high performance computing architectures. It is optimized to take advantages of key High Performance Computing features to accelerate neural network training. Specifically it is optimized for low-latency, high bandwidth interconnects, node-local NVRAM, node-local GPU accelerators, and high bandwidth parallel file systems. It is built on top of the open source Elemental distributed-memory dense and spars-direct linear algebra and optimization library that is released under the BSD license. The algorithms contained within LBANN are drawn from the academic literature and implemented to work within a distributed-memory framework.

New NAS Parallel Benchmarks Results

NASA Technical Reports Server (NTRS)

Yarrow, Maurice; Saphir, William; VanderWijngaart, Rob; Woo, Alex; Kutler, Paul (Technical Monitor)

1997-01-01

NPB2 (NAS (NASA Advanced Supercomputing) Parallel Benchmarks 2) is an implementation, based on Fortran and the MPI (message passing interface) message passing standard, of the original NAS Parallel Benchmark specifications. NPB2 programs are run with little or no tuning, in contrast to NPB vendor implementations, which are highly optimized for specific architectures. NPB2 results complement, rather than replace, NPB results. Because they have not been optimized by vendors, NPB2 implementations approximate the performance a typical user can expect for a portable parallel program on distributed memory parallel computers. Together these results provide an insightful comparison of the real-world performance of high-performance computers. New NPB2 features: New implementation (CG), new workstation class problem sizes, new serial sample versions, more performance statistics.

Xyce Parallel Electronic Simulator Users' Guide Version 6.8

DOE Office of Scientific and Technical Information (OSTI.GOV)

Keiter, Eric R.; Aadithya, Karthik Venkatraman; Mei, Ting

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been de- signed as a SPICE-compatible, high-performance analog circuit simulator, and has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability over the current state-of-the-art in the following areas: Capability to solve extremely large circuit problems by supporting large-scale parallel com- puting platforms (up to thousands of processors). This includes support for most popular parallel and serial computers. A differential-algebraic-equation (DAE) formulation, which better isolates the device model package from solver algorithms. This allows onemore » to develop new types of analysis without requiring the implementation of analysis-specific device models. Device models that are specifically tailored to meet Sandia's needs, including some radiation- aware devices (for Sandia users only). Object-oriented code design and implementation using modern coding practices. Xyce is a parallel code in the most general sense of the phrase$-$ a message passing parallel implementation $-$ which allows it to run efficiently a wide range of computing platforms. These include serial, shared-memory and distributed-memory parallel platforms. Attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows.« less

Programming model for distributed intelligent systems

NASA Technical Reports Server (NTRS)

Sztipanovits, J.; Biegl, C.; Karsai, G.; Bogunovic, N.; Purves, B.; Williams, R.; Christiansen, T.

1988-01-01

A programming model and architecture which was developed for the design and implementation of complex, heterogeneous measurement and control systems is described. The Multigraph Architecture integrates artificial intelligence techniques with conventional software technologies, offers a unified framework for distributed and shared memory based parallel computational models and supports multiple programming paradigms. The system can be implemented on different hardware architectures and can be adapted to strongly different applications.

Running accuracy analysis of a 3-RRR parallel kinematic machine considering the deformations of the links

NASA Astrophysics Data System (ADS)

Wang, Liping; Jiang, Yao; Li, Tiemin

2014-09-01

Parallel kinematic machines have drawn considerable attention and have been widely used in some special fields. However, high precision is still one of the challenges when they are used for advanced machine tools. One of the main reasons is that the kinematic chains of parallel kinematic machines are composed of elongated links that can easily suffer deformations, especially at high speeds and under heavy loads. A 3-RRR parallel kinematic machine is taken as a study object for investigating its accuracy with the consideration of the deformations of its links during the motion process. Based on the dynamic model constructed by the Newton-Euler method, all the inertia loads and constraint forces of the links are computed and their deformations are derived. Then the kinematic errors of the machine are derived with the consideration of the deformations of the links. Through further derivation, the accuracy of the machine is given in a simple explicit expression, which will be helpful to increase the calculating speed. The accuracy of this machine when following a selected circle path is simulated. The influences of magnitude of the maximum acceleration and external loads on the running accuracy of the machine are investigated. The results show that the external loads will deteriorate the accuracy of the machine tremendously when their direction coincides with the direction of the worst stiffness of the machine. The proposed method provides a solution for predicting the running accuracy of the parallel kinematic machines and can also be used in their design optimization as well as selection of suitable running parameters.

A parallel implementation of an off-lattice individual-based model of multicellular populations

NASA Astrophysics Data System (ADS)

Harvey, Daniel G.; Fletcher, Alexander G.; Osborne, James M.; Pitt-Francis, Joe

2015-07-01

As computational models of multicellular populations include ever more detailed descriptions of biophysical and biochemical processes, the computational cost of simulating such models limits their ability to generate novel scientific hypotheses and testable predictions. While developments in microchip technology continue to increase the power of individual processors, parallel computing offers an immediate increase in available processing power. To make full use of parallel computing technology, it is necessary to develop specialised algorithms. To this end, we present a parallel algorithm for a class of off-lattice individual-based models of multicellular populations. The algorithm divides the spatial domain between computing processes and comprises communication routines that ensure the model is correctly simulated on multiple processors. The parallel algorithm is shown to accurately reproduce the results of a deterministic simulation performed using a pre-existing serial implementation. We test the scaling of computation time, memory use and load balancing as more processes are used to simulate a cell population of fixed size. We find approximate linear scaling of both speed-up and memory consumption on up to 32 processor cores. Dynamic load balancing is shown to provide speed-up for non-regular spatial distributions of cells in the case of a growing population.

Parallel computation for biological sequence comparison: comparing a portable model to the native model for the Intel Hypercube.

PubMed

Nadkarni, P M; Miller, P L

1991-01-01

A parallel program for inter-database sequence comparison was developed on the Intel Hypercube using two models of parallel programming. One version was built using machine-specific Hypercube parallel programming commands. The other version was built using Linda, a machine-independent parallel programming language. The two versions of the program provide a case study comparing these two approaches to parallelization in an important biological application area. Benchmark tests with both programs gave comparable results with a small number of processors. As the number of processors was increased, the Linda version was somewhat less efficient. The Linda version was also run without change on Network Linda, a virtual parallel machine running on a network of desktop workstations.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Chrisochoides, N.; Sukup, F.

In this paper we present a parallel implementation of the Bowyer-Watson (BW) algorithm using the task-parallel programming model. The BW algorithm constitutes an ideal mesh refinement strategy for implementing a large class of unstructured mesh generation techniques on both sequential and parallel computers, by preventing the need for global mesh refinement. Its implementation on distributed memory multicomputes using the traditional data-parallel model has been proven very inefficient due to excessive synchronization needed among processors. In this paper we demonstrate that with the task-parallel model we can tolerate synchronization costs inherent to data-parallel methods by exploring concurrency in the processor level.more » Our preliminary performance data indicate that the task- parallel approach: (i) is almost four times faster than the existing data-parallel methods, (ii) scales linearly, and (iii) introduces minimum overheads compared to the {open_quotes}best{close_quotes} sequential implementation of the BW algorithm.« less

Concurrent Image Processing Executive (CIPE)

NASA Technical Reports Server (NTRS)

Lee, Meemong; Cooper, Gregory T.; Groom, Steven L.; Mazer, Alan S.; Williams, Winifred I.

1988-01-01

The design and implementation of a Concurrent Image Processing Executive (CIPE), which is intended to become the support system software for a prototype high performance science analysis workstation are discussed. The target machine for this software is a JPL/Caltech Mark IIIfp Hypercube hosted by either a MASSCOMP 5600 or a Sun-3, Sun-4 workstation; however, the design will accommodate other concurrent machines of similar architecture, i.e., local memory, multiple-instruction-multiple-data (MIMD) machines. The CIPE system provides both a multimode user interface and an applications programmer interface, and has been designed around four loosely coupled modules; (1) user interface, (2) host-resident executive, (3) hypercube-resident executive, and (4) application functions. The loose coupling between modules allows modification of a particular module without significantly affecting the other modules in the system. In order to enhance hypercube memory utilization and to allow expansion of image processing capabilities, a specialized program management method, incremental loading, was devised. To minimize data transfer between host and hypercube a data management method which distributes, redistributes, and tracks data set information was implemented.

Equalizer: a scalable parallel rendering framework.

PubMed

Eilemann, Stefan; Makhinya, Maxim; Pajarola, Renato

2009-01-01

Continuing improvements in CPU and GPU performances as well as increasing multi-core processor and cluster-based parallelism demand for flexible and scalable parallel rendering solutions that can exploit multipipe hardware accelerated graphics. In fact, to achieve interactive visualization, scalable rendering systems are essential to cope with the rapid growth of data sets. However, parallel rendering systems are non-trivial to develop and often only application specific implementations have been proposed. The task of developing a scalable parallel rendering framework is even more difficult if it should be generic to support various types of data and visualization applications, and at the same time work efficiently on a cluster with distributed graphics cards. In this paper we introduce a novel system called Equalizer, a toolkit for scalable parallel rendering based on OpenGL which provides an application programming interface (API) to develop scalable graphics applications for a wide range of systems ranging from large distributed visualization clusters and multi-processor multipipe graphics systems to single-processor single-pipe desktop machines. We describe the system architecture, the basic API, discuss its advantages over previous approaches, present example configurations and usage scenarios as well as scalability results.

Parallel implementation of an adaptive and parameter-free N-body integrator

NASA Astrophysics Data System (ADS)

Pruett, C. David; Ingham, William H.; Herman, Ralph D.

2011-05-01

Previously, Pruett et al. (2003) [3] described an N-body integrator of arbitrarily high order M with an asymptotic operation count of O(MN). The algorithm's structure lends itself readily to data parallelization, which we document and demonstrate here in the integration of point-mass systems subject to Newtonian gravitation. High order is shown to benefit parallel efficiency. The resulting N-body integrator is robust, parameter-free, highly accurate, and adaptive in both time-step and order. Moreover, it exhibits linear speedup on distributed parallel processors, provided that each processor is assigned at least a handful of bodies. Program summaryProgram title: PNB.f90 Catalogue identifier: AEIK_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEIK_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC license, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 3052 No. of bytes in distributed program, including test data, etc.: 68 600 Distribution format: tar.gz Programming language: Fortran 90 and OpenMPI Computer: All shared or distributed memory parallel processors Operating system: Unix/Linux Has the code been vectorized or parallelized?: The code has been parallelized but has not been explicitly vectorized. RAM: Dependent upon N Classification: 4.3, 4.12, 6.5 Nature of problem: High accuracy numerical evaluation of trajectories of N point masses each subject to Newtonian gravitation. Solution method: Parallel and adaptive extrapolation in time via power series of arbitrary degree. Running time: 5.1 s for the demo program supplied with the package.

Visual Computing Environment

NASA Technical Reports Server (NTRS)

Lawrence, Charles; Putt, Charles W.

1997-01-01

The Visual Computing Environment (VCE) is a NASA Lewis Research Center project to develop a framework for intercomponent and multidisciplinary computational simulations. Many current engineering analysis codes simulate various aspects of aircraft engine operation. For example, existing computational fluid dynamics (CFD) codes can model the airflow through individual engine components such as the inlet, compressor, combustor, turbine, or nozzle. Currently, these codes are run in isolation, making intercomponent and complete system simulations very difficult to perform. In addition, management and utilization of these engineering codes for coupled component simulations is a complex, laborious task, requiring substantial experience and effort. To facilitate multicomponent aircraft engine analysis, the CFD Research Corporation (CFDRC) is developing the VCE system. This system, which is part of NASA's Numerical Propulsion Simulation System (NPSS) program, can couple various engineering disciplines, such as CFD, structural analysis, and thermal analysis. The objectives of VCE are to (1) develop a visual computing environment for controlling the execution of individual simulation codes that are running in parallel and are distributed on heterogeneous host machines in a networked environment, (2) develop numerical coupling algorithms for interchanging boundary conditions between codes with arbitrary grid matching and different levels of dimensionality, (3) provide a graphical interface for simulation setup and control, and (4) provide tools for online visualization and plotting. VCE was designed to provide a distributed, object-oriented environment. Mechanisms are provided for creating and manipulating objects, such as grids, boundary conditions, and solution data. This environment includes parallel virtual machine (PVM) for distributed processing. Users can interactively select and couple any set of codes that have been modified to run in a parallel distributed fashion on a cluster of heterogeneous workstations. A scripting facility allows users to dictate the sequence of events that make up the particular simulation.

A data distributed parallel algorithm for ray-traced volume rendering

NASA Technical Reports Server (NTRS)

Ma, Kwan-Liu; Painter, James S.; Hansen, Charles D.; Krogh, Michael F.

1993-01-01

This paper presents a divide-and-conquer ray-traced volume rendering algorithm and a parallel image compositing method, along with their implementation and performance on the Connection Machine CM-5, and networked workstations. This algorithm distributes both the data and the computations to individual processing units to achieve fast, high-quality rendering of high-resolution data. The volume data, once distributed, is left intact. The processing nodes perform local ray tracing of their subvolume concurrently. No communication between processing units is needed during this locally ray-tracing process. A subimage is generated by each processing unit and the final image is obtained by compositing subimages in the proper order, which can be determined a priori. Test results on both the CM-5 and a group of networked workstations demonstrate the practicality of our rendering algorithm and compositing method.

Research in Computational Aeroscience Applications Implemented on Advanced Parallel Computing Systems

NASA Technical Reports Server (NTRS)

Wigton, Larry

1996-01-01

Improving the numerical linear algebra routines for use in new Navier-Stokes codes, specifically Tim Barth's unstructured grid code, with spin-offs to TRANAIR is reported. A fast distance calculation routine for Navier-Stokes codes using the new one-equation turbulence models is written. The primary focus of this work was devoted to improving matrix-iterative methods. New algorithms have been developed which activate the full potential of classical Cray-class computers as well as distributed-memory parallel computers.

Experiences Using OpenMP Based on Compiler Directed Software DSM on a PC Cluster

NASA Technical Reports Server (NTRS)

Hess, Matthias; Jost, Gabriele; Mueller, Matthias; Ruehle, Roland; Biegel, Bryan (Technical Monitor)

2002-01-01

In this work we report on our experiences running OpenMP (message passing) programs on a commodity cluster of PCs (personal computers) running a software distributed shared memory (DSM) system. We describe our test environment and report on the performance of a subset of the NAS (NASA Advanced Supercomputing) Parallel Benchmarks that have been automatically parallelized for OpenMP. We compare the performance of the OpenMP implementations with that of their message passing counterparts and discuss performance differences.

Automatic Generation of Directive-Based Parallel Programs for Shared Memory Parallel Systems

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Yan, Jerry; Frumkin, Michael

2000-01-01

The shared-memory programming model is a very effective way to achieve parallelism on shared memory parallel computers. As great progress was made in hardware and software technologies, performance of parallel programs with compiler directives has demonstrated large improvement. The introduction of OpenMP directives, the industrial standard for shared-memory programming, has minimized the issue of portability. Due to its ease of programming and its good performance, the technique has become very popular. In this study, we have extended CAPTools, a computer-aided parallelization toolkit, to automatically generate directive-based, OpenMP, parallel programs. We outline techniques used in the implementation of the tool and present test results on the NAS parallel benchmarks and ARC3D, a CFD application. This work demonstrates the great potential of using computer-aided tools to quickly port parallel programs and also achieve good performance.

Reverse time migration: A seismic processing application on the connection machine

NASA Technical Reports Server (NTRS)

Fiebrich, Rolf-Dieter

1987-01-01

The implementation of a reverse time migration algorithm on the Connection Machine, a massively parallel computer is described. Essential architectural features of this machine as well as programming concepts are presented. The data structures and parallel operations for the implementation of the reverse time migration algorithm are described. The algorithm matches the Connection Machine architecture closely and executes almost at the peak performance of this machine.

Parallel Clustering Algorithm for Large-Scale Biological Data Sets

PubMed Central

Wang, Minchao; Zhang, Wu; Ding, Wang; Dai, Dongbo; Zhang, Huiran; Xie, Hao; Chen, Luonan; Guo, Yike; Xie, Jiang

2014-01-01

Backgrounds Recent explosion of biological data brings a great challenge for the traditional clustering algorithms. With increasing scale of data sets, much larger memory and longer runtime are required for the cluster identification problems. The affinity propagation algorithm outperforms many other classical clustering algorithms and is widely applied into the biological researches. However, the time and space complexity become a great bottleneck when handling the large-scale data sets. Moreover, the similarity matrix, whose constructing procedure takes long runtime, is required before running the affinity propagation algorithm, since the algorithm clusters data sets based on the similarities between data pairs. Methods Two types of parallel architectures are proposed in this paper to accelerate the similarity matrix constructing procedure and the affinity propagation algorithm. The memory-shared architecture is used to construct the similarity matrix, and the distributed system is taken for the affinity propagation algorithm, because of its large memory size and great computing capacity. An appropriate way of data partition and reduction is designed in our method, in order to minimize the global communication cost among processes. Result A speedup of 100 is gained with 128 cores. The runtime is reduced from serval hours to a few seconds, which indicates that parallel algorithm is capable of handling large-scale data sets effectively. The parallel affinity propagation also achieves a good performance when clustering large-scale gene data (microarray) and detecting families in large protein superfamilies. PMID:24705246

Parallelization of a Monte Carlo particle transport simulation code

NASA Astrophysics Data System (ADS)

Hadjidoukas, P.; Bousis, C.; Emfietzoglou, D.

2010-05-01

We have developed a high performance version of the Monte Carlo particle transport simulation code MC4. The original application code, developed in Visual Basic for Applications (VBA) for Microsoft Excel, was first rewritten in the C programming language for improving code portability. Several pseudo-random number generators have been also integrated and studied. The new MC4 version was then parallelized for shared and distributed-memory multiprocessor systems using the Message Passing Interface. Two parallel pseudo-random number generator libraries (SPRNG and DCMT) have been seamlessly integrated. The performance speedup of parallel MC4 has been studied on a variety of parallel computing architectures including an Intel Xeon server with 4 dual-core processors, a Sun cluster consisting of 16 nodes of 2 dual-core AMD Opteron processors and a 200 dual-processor HP cluster. For large problem size, which is limited only by the physical memory of the multiprocessor server, the speedup results are almost linear on all systems. We have validated the parallel implementation against the serial VBA and C implementations using the same random number generator. Our experimental results on the transport and energy loss of electrons in a water medium show that the serial and parallel codes are equivalent in accuracy. The present improvements allow for studying of higher particle energies with the use of more accurate physical models, and improve statistics as more particles tracks can be simulated in low response time.

Massive parallelization of a 3D finite difference electromagnetic forward solution using domain decomposition methods on multiple CUDA enabled GPUs

NASA Astrophysics Data System (ADS)

Schultz, A.

2010-12-01

3D forward solvers lie at the core of inverse formulations used to image the variation of electrical conductivity within the Earth's interior. This property is associated with variations in temperature, composition, phase, presence of volatiles, and in specific settings, the presence of groundwater, geothermal resources, oil/gas or minerals. The high cost of 3D solutions has been a stumbling block to wider adoption of 3D methods. Parallel algorithms for modeling frequency domain 3D EM problems have not achieved wide scale adoption, with emphasis on fairly coarse grained parallelism using MPI and similar approaches. The communications bandwidth as well as the latency required to send and receive network communication packets is a limiting factor in implementing fine grained parallel strategies, inhibiting wide adoption of these algorithms. Leading Graphics Processor Unit (GPU) companies now produce GPUs with hundreds of GPU processor cores per die. The footprint, in silicon, of the GPU's restricted instruction set is much smaller than the general purpose instruction set required of a CPU. Consequently, the density of processor cores on a GPU can be much greater than on a CPU. GPUs also have local memory, registers and high speed communication with host CPUs, usually through PCIe type interconnects. The extremely low cost and high computational power of GPUs provides the EM geophysics community with an opportunity to achieve fine grained (i.e. massive) parallelization of codes on low cost hardware. The current generation of GPUs (e.g. NVidia Fermi) provides 3 billion transistors per chip die, with nearly 500 processor cores and up to 6 GB of fast (DDR5) GPU memory. This latest generation of GPU supports fast hardware double precision (64 bit) floating point operations of the type required for frequency domain EM forward solutions. Each Fermi GPU board can sustain nearly 1 TFLOP in double precision, and multiple boards can be installed in the host computer system. We describe our ongoing efforts to achieve massive parallelization on a novel hybrid GPU testbed machine currently configured with 12 Intel Westmere Xeon CPU cores (or 24 parallel computational threads) with 96 GB DDR3 system memory, 4 GPU subsystems which in aggregate contain 960 NVidia Tesla GPU cores with 16 GB dedicated DDR3 GPU memory, and a second interleved bank of 4 GPU subsystems containing in aggregate 1792 NVidia Fermi GPU cores with 12 GB dedicated DDR5 GPU memory. We are applying domain decomposition methods to a modified version of Weiss' (2001) 3D frequency domain full physics EM finite difference code, an open source GPL licensed f90 code available for download from www.OpenEM.org. This will be the core of a new hybrid 3D inversion that parallelizes frequencies across CPUs and individual forward solutions across GPUs. We describe progress made in modifying the code to use direct solvers in GPU cores dedicated to each small subdomain, iteratively improving the solution by matching adjacent subdomain boundary solutions, rather than iterative Krylov space sparse solvers as currently applied to the whole domain.

Parallel computation for biological sequence comparison: comparing a portable model to the native model for the Intel Hypercube.

PubMed Central

Nadkarni, P. M.; Miller, P. L.

1991-01-01

A parallel program for inter-database sequence comparison was developed on the Intel Hypercube using two models of parallel programming. One version was built using machine-specific Hypercube parallel programming commands. The other version was built using Linda, a machine-independent parallel programming language. The two versions of the program provide a case study comparing these two approaches to parallelization in an important biological application area. Benchmark tests with both programs gave comparable results with a small number of processors. As the number of processors was increased, the Linda version was somewhat less efficient. The Linda version was also run without change on Network Linda, a virtual parallel machine running on a network of desktop workstations. PMID:1807632

Structure and Deterioration of Semantic Memory: A Neuropsychological and Computational Investigation

ERIC Educational Resources Information Center

Rogers, Timothy T.; Lambon Ralph, Matthew A.; Garrard, Peter; Bozeat, Sasha; McClelland, James L.; Hodges, John R.; Patterson, Karalyn

2004-01-01

Wernicke (1900, as cited in G. H. Eggert, 1977) suggested that semantic knowledge arises from the interaction of perceptual representations of objects and words. The authors present a parallel distributed processing implementation of this theory, in which semantic representations emerge from mechanisms that acquire the mappings between visual…

Slices: A Scalable Partitioner for Finite Element Meshes

NASA Technical Reports Server (NTRS)

Ding, H. Q.; Ferraro, R. D.

1995-01-01

A parallel partitioner for partitioning unstructured finite element meshes on distributed memory architectures is developed. The element based partitioner can handle mixtures of different element types. All algorithms adopted in the partitioner are scalable, including a communication template for unpredictable incoming messages, as shown in actual timing measurements.

Multistable wireless micro-actuator based on antagonistic pre-shaped double beams

NASA Astrophysics Data System (ADS)

Liu, X.; Lamarque, F.; Doré, E.; Pouille, P.

2015-07-01

This paper presents a monolithic multistable micro-actuator based on antagonistic pre-shaped double beams. The designed micro-actuator is formed by two rows of bistable micro-actuators providing four stable positions. The bistable mechanism for each row is a pair of antagonistic pre-shaped beams. This bistable mechanism has an easier pre-load operation compared to the pre-compressed bistable beams method. Furthermore, it solves the asymmetrical force output problem of parallel pre-shaped bistable double beams. At the same time, the geometrical limit is lower than parallel pre-shaped bistable double beams, which ensures a smaller stroke of the micro-actuator with the same dimensions. The designed micro-actuator is fabricated using laser cutting machine on medium density fiberboard (MDF). The bistability and merits of antagonistic pre-shaped double beams are experimentally validated. Finally, a contactless actuation test is performed using 660 nm wavelength laser heating shape memory alloy (SMA) active elements.

Enhancing Application Performance Using Mini-Apps: Comparison of Hybrid Parallel Programming Paradigms

NASA Technical Reports Server (NTRS)

Lawson, Gary; Sosonkina, Masha; Baurle, Robert; Hammond, Dana

2017-01-01

In many fields, real-world applications for High Performance Computing have already been developed. For these applications to stay up-to-date, new parallel strategies must be explored to yield the best performance; however, restructuring or modifying a real-world application may be daunting depending on the size of the code. In this case, a mini-app may be employed to quickly explore such options without modifying the entire code. In this work, several mini-apps have been created to enhance a real-world application performance, namely the VULCAN code for complex flow analysis developed at the NASA Langley Research Center. These mini-apps explore hybrid parallel programming paradigms with Message Passing Interface (MPI) for distributed memory access and either Shared MPI (SMPI) or OpenMP for shared memory accesses. Performance testing shows that MPI+SMPI yields the best execution performance, while requiring the largest number of code changes. A maximum speedup of 23 was measured for MPI+SMPI, but only 11 was measured for MPI+OpenMP.

Parallel and fault-tolerant algorithms for hypercube multiprocessors

DOE Office of Scientific and Technical Information (OSTI.GOV)

Aykanat, C.

1988-01-01

Several techniques for increasing the performance of parallel algorithms on distributed-memory message-passing multi-processor systems are investigated. These techniques are effectively implemented for the parallelization of the Scaled Conjugate Gradient (SCG) algorithm on a hypercube connected message-passing multi-processor. Significant performance improvement is achieved by using these techniques. The SCG algorithm is used for the solution phase of an FE modeling system. Almost linear speed-up is achieved, and it is shown that hypercube topology is scalable for an FE class of problem. The SCG algorithm is also shown to be suitable for vectorization, and near supercomputer performance is achieved on a vectormore » hypercube multiprocessor by exploiting both parallelization and vectorization. Fault-tolerance issues for the parallel SCG algorithm and for the hypercube topology are also addressed.« less

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ibrahim, Khaled Z.; Epifanovsky, Evgeny; Williams, Samuel W.

Coupled-cluster methods provide highly accurate models of molecular structure by explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix-matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts tomore » extend the Libtensor framework to work in the distributed memory environment in a scalable and energy efficient manner. We achieve up to 240 speedup compared with the best optimized shared memory implementation. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures, (Cray XC30&XC40, BlueGene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance. Nevertheless, we preserve a uni ed interface to both programming models to maintain the productivity of computational quantum chemists.« less

HipMCL: a high-performance parallel implementation of the Markov clustering algorithm for large-scale networks

PubMed Central

Azad, Ariful; Ouzounis, Christos A; Kyrpides, Nikos C; Buluç, Aydin

2018-01-01

Abstract Biological networks capture structural or functional properties of relevant entities such as molecules, proteins or genes. Characteristic examples are gene expression networks or protein–protein interaction networks, which hold information about functional affinities or structural similarities. Such networks have been expanding in size due to increasing scale and abundance of biological data. While various clustering algorithms have been proposed to find highly connected regions, Markov Clustering (MCL) has been one of the most successful approaches to cluster sequence similarity or expression networks. Despite its popularity, MCL’s scalability to cluster large datasets still remains a bottleneck due to high running times and memory demands. Here, we present High-performance MCL (HipMCL), a parallel implementation of the original MCL algorithm that can run on distributed-memory computers. We show that HipMCL can efficiently utilize 2000 compute nodes and cluster a network of ∼70 million nodes with ∼68 billion edges in ∼2.4 h. By exploiting distributed-memory environments, HipMCL clusters large-scale networks several orders of magnitude faster than MCL and enables clustering of even bigger networks. HipMCL is based on MPI and OpenMP and is freely available under a modified BSD license. PMID:29315405

HipMCL: a high-performance parallel implementation of the Markov clustering algorithm for large-scale networks

DOE PAGES

Azad, Ariful; Pavlopoulos, Georgios A.; Ouzounis, Christos A.; ...

2018-01-05

Biological networks capture structural or functional properties of relevant entities such as molecules, proteins or genes. Characteristic examples are gene expression networks or protein–protein interaction networks, which hold information about functional affinities or structural similarities. Such networks have been expanding in size due to increasing scale and abundance of biological data. While various clustering algorithms have been proposed to find highly connected regions, Markov Clustering (MCL) has been one of the most successful approaches to cluster sequence similarity or expression networks. Despite its popularity, MCL’s scalability to cluster large datasets still remains a bottleneck due to high running times andmore » memory demands. In this paper, we present High-performance MCL (HipMCL), a parallel implementation of the original MCL algorithm that can run on distributed-memory computers. We show that HipMCL can efficiently utilize 2000 compute nodes and cluster a network of ~70 million nodes with ~68 billion edges in ~2.4 h. By exploiting distributed-memory environments, HipMCL clusters large-scale networks several orders of magnitude faster than MCL and enables clustering of even bigger networks. Finally, HipMCL is based on MPI and OpenMP and is freely available under a modified BSD license.« less

HipMCL: a high-performance parallel implementation of the Markov clustering algorithm for large-scale networks

DOE Office of Scientific and Technical Information (OSTI.GOV)

Azad, Ariful; Pavlopoulos, Georgios A.; Ouzounis, Christos A.

Biological networks capture structural or functional properties of relevant entities such as molecules, proteins or genes. Characteristic examples are gene expression networks or protein–protein interaction networks, which hold information about functional affinities or structural similarities. Such networks have been expanding in size due to increasing scale and abundance of biological data. While various clustering algorithms have been proposed to find highly connected regions, Markov Clustering (MCL) has been one of the most successful approaches to cluster sequence similarity or expression networks. Despite its popularity, MCL’s scalability to cluster large datasets still remains a bottleneck due to high running times andmore » memory demands. In this paper, we present High-performance MCL (HipMCL), a parallel implementation of the original MCL algorithm that can run on distributed-memory computers. We show that HipMCL can efficiently utilize 2000 compute nodes and cluster a network of ~70 million nodes with ~68 billion edges in ~2.4 h. By exploiting distributed-memory environments, HipMCL clusters large-scale networks several orders of magnitude faster than MCL and enables clustering of even bigger networks. Finally, HipMCL is based on MPI and OpenMP and is freely available under a modified BSD license.« less

An Efficient Multiblock Method for Aerodynamic Analysis and Design on Distributed Memory Systems

NASA Technical Reports Server (NTRS)

Reuther, James; Alonso, Juan Jose; Vassberg, John C.; Jameson, Antony; Martinelli, Luigi

1997-01-01

The work presented in this paper describes the application of a multiblock gridding strategy to the solution of aerodynamic design optimization problems involving complex configurations. The design process is parallelized using the MPI (Message Passing Interface) Standard such that it can be efficiently run on a variety of distributed memory systems ranging from traditional parallel computers to networks of workstations. Substantial improvements to the parallel performance of the baseline method are presented, with particular attention to their impact on the scalability of the program as a function of the mesh size. Drag minimization calculations at a fixed coefficient of lift are presented for a business jet configuration that includes the wing, body, pylon, aft-mounted nacelle, and vertical and horizontal tails. An aerodynamic design optimization is performed with both the Euler and Reynolds Averaged Navier-Stokes (RANS) equations governing the flow solution and the results are compared. These sample calculations establish the feasibility of efficient aerodynamic optimization of complete aircraft configurations using the RANS equations as the flow model. There still exists, however, the need for detailed studies of the importance of a true viscous adjoint method which holds the promise of tackling the minimization of not only the wave and induced components of drag, but also the viscous drag.

Design of object-oriented distributed simulation classes

NASA Technical Reports Server (NTRS)

Schoeffler, James D. (Principal Investigator)

1995-01-01

Distributed simulation of aircraft engines as part of a computer aided design package is being developed by NASA Lewis Research Center for the aircraft industry. The project is called NPSS, an acronym for 'Numerical Propulsion Simulation System'. NPSS is a flexible object-oriented simulation of aircraft engines requiring high computing speed. It is desirable to run the simulation on a distributed computer system with multiple processors executing portions of the simulation in parallel. The purpose of this research was to investigate object-oriented structures such that individual objects could be distributed. The set of classes used in the simulation must be designed to facilitate parallel computation. Since the portions of the simulation carried out in parallel are not independent of one another, there is the need for communication among the parallel executing processors which in turn implies need for their synchronization. Communication and synchronization can lead to decreased throughput as parallel processors wait for data or synchronization signals from other processors. As a result of this research, the following have been accomplished. The design and implementation of a set of simulation classes which result in a distributed simulation control program have been completed. The design is based upon MIT 'Actor' model of a concurrent object and uses 'connectors' to structure dynamic connections between simulation components. Connectors may be dynamically created according to the distribution of objects among machines at execution time without any programming changes. Measurements of the basic performance have been carried out with the result that communication overhead of the distributed design is swamped by the computation time of modules unless modules have very short execution times per iteration or time step. An analytical performance model based upon queuing network theory has been designed and implemented. Its application to realistic configurations has not been carried out.

Design of Object-Oriented Distributed Simulation Classes

NASA Technical Reports Server (NTRS)

Schoeffler, James D.

1995-01-01

Distributed simulation of aircraft engines as part of a computer aided design package being developed by NASA Lewis Research Center for the aircraft industry. The project is called NPSS, an acronym for "Numerical Propulsion Simulation System". NPSS is a flexible object-oriented simulation of aircraft engines requiring high computing speed. It is desirable to run the simulation on a distributed computer system with multiple processors executing portions of the simulation in parallel. The purpose of this research was to investigate object-oriented structures such that individual objects could be distributed. The set of classes used in the simulation must be designed to facilitate parallel computation. Since the portions of the simulation carried out in parallel are not independent of one another, there is the need for communication among the parallel executing processors which in turn implies need for their synchronization. Communication and synchronization can lead to decreased throughput as parallel processors wait for data or synchronization signals from other processors. As a result of this research, the following have been accomplished. The design and implementation of a set of simulation classes which result in a distributed simulation control program have been completed. The design is based upon MIT "Actor" model of a concurrent object and uses "connectors" to structure dynamic connections between simulation components. Connectors may be dynamically created according to the distribution of objects among machines at execution time without any programming changes. Measurements of the basic performance have been carried out with the result that communication overhead of the distributed design is swamped by the computation time of modules unless modules have very short execution times per iteration or time step. An analytical performance model based upon queuing network theory has been designed and implemented. Its application to realistic configurations has not been carried out.

Line-drawing algorithms for parallel machines

NASA Technical Reports Server (NTRS)

Pang, Alex T.

1990-01-01

The fact that conventional line-drawing algorithms, when applied directly on parallel machines, can lead to very inefficient codes is addressed. It is suggested that instead of modifying an existing algorithm for a parallel machine, a more efficient implementation can be produced by going back to the invariants in the definition. Popular line-drawing algorithms are compared with two alternatives; distance to a line (a point is on the line if sufficiently close to it) and intersection with a line (a point on the line if an intersection point). For massively parallel single-instruction-multiple-data (SIMD) machines (with thousands of processors and up), the alternatives provide viable line-drawing algorithms. Because of the pixel-per-processor mapping, their performance is independent of the line length and orientation.

Reducing Interprocessor Dependence in Recoverable Distributed Shared Memory

NASA Technical Reports Server (NTRS)

Janssens, Bob; Fuchs, W. Kent

1994-01-01

Checkpointing techniques in parallel systems use dependency tracking and/or message logging to ensure that a system rolls back to a consistent state. Traditional dependency tracking in distributed shared memory (DSM) systems is expensive because of high communication frequency. In this paper we show that, if designed correctly, a DSM system only needs to consider dependencies due to the transfer of blocks of data, resulting in reduced dependency tracking overhead and reduced potential for rollback propagation. We develop an ownership timestamp scheme to tolerate the loss of block state information and develop a passive server model of execution where interactions between processors are considered atomic. With our scheme, dependencies are significantly reduced compared to the traditional message-passing model.

Sequential behavior and its inherent tolerance to memory faults.

NASA Technical Reports Server (NTRS)

Meyer, J. F.

1972-01-01

Representation of a memory fault of a sequential machine M by a function mu on the states of M and the result of the fault by an appropriately determined machine M(mu). Given some sequential behavior B, its inherent tolerance to memory faults can then be measured in terms of the minimum memory redundancy required to realize B with a state-assigned machine having fault tolerance type tau and fault tolerance level t. A behavior having maximum inherent tolerance is exhibited, and it is shown that behaviors of the same size can have different inherent tolerance.

Hierarchical analytical and simulation modelling of human-machine systems with interference

NASA Astrophysics Data System (ADS)

Braginsky, M. Ya; Tarakanov, D. V.; Tsapko, S. G.; Tsapko, I. V.; Baglaeva, E. A.

2017-01-01

The article considers the principles of building the analytical and simulation model of the human operator and the industrial control system hardware and software. E-networks as the extension of Petri nets are used as the mathematical apparatus. This approach allows simulating complex parallel distributed processes in human-machine systems. The structural and hierarchical approach is used as the building method for the mathematical model of the human operator. The upper level of the human operator is represented by the logical dynamic model of decision making based on E-networks. The lower level reflects psychophysiological characteristics of the human-operator.

A high performance linear equation solver on the VPP500 parallel supercomputer

DOE Office of Scientific and Technical Information (OSTI.GOV)

Nakanishi, Makoto; Ina, Hiroshi; Miura, Kenichi

1994-12-31

This paper describes the implementation of two high performance linear equation solvers developed for the Fujitsu VPP500, a distributed memory parallel supercomputer system. The solvers take advantage of the key architectural features of VPP500--(1) scalability for an arbitrary number of processors up to 222 processors, (2) flexible data transfer among processors provided by a crossbar interconnection network, (3) vector processing capability on each processor, and (4) overlapped computation and transfer. The general linear equation solver based on the blocked LU decomposition method achieves 120.0 GFLOPS performance with 100 processors in the LIN-PACK Highly Parallel Computing benchmark.

Disk-based k-mer counting on a PC

PubMed Central

2013-01-01

Background The k-mer counting problem, which is to build the histogram of occurrences of every k-symbol long substring in a given text, is important for many bioinformatics applications. They include developing de Bruijn graph genome assemblers, fast multiple sequence alignment and repeat detection. Results We propose a simple, yet efficient, parallel disk-based algorithm for counting k-mers. Experiments show that it usually offers the fastest solution to the considered problem, while demanding a relatively small amount of memory. In particular, it is capable of counting the statistics for short-read human genome data, in input gzipped FASTQ file, in less than 40 minutes on a PC with 16 GB of RAM and 6 CPU cores, and for long-read human genome data in less than 70 minutes. On a more powerful machine, using 32 GB of RAM and 32 CPU cores, the tasks are accomplished in less than half the time. No other algorithm for most tested settings of this problem and mammalian-size data can accomplish this task in comparable time. Our solution also belongs to memory-frugal ones; most competitive algorithms cannot efficiently work on a PC with 16 GB of memory for such massive data. Conclusions By making use of cheap disk space and exploiting CPU and I/O parallelism we propose a very competitive k-mer counting procedure, called KMC. Our results suggest that judicious resource management may allow to solve at least some bioinformatics problems with massive data on a commodity personal computer. PMID:23679007

Comparison of two paradigms for distributed shared memory

DOE Office of Scientific and Technical Information (OSTI.GOV)

Levelt, W.G.; Kaashoek, M.F.; Bal, H.E.

1990-08-01

The paper compares two paradigms for Distributed Shared Memory on loosely coupled computing systems: the shared data-object model as used in Orca, a programming language specially designed for loosely coupled computing systems and the Shared Virtual Memory model. For both paradigms the authors have implemented two systems, one using only point-to-point messages, the other using broadcasting as well. They briefly describe these two paradigms and their implementations. Then they compare their performance on four applications: the traveling salesman problem, alpha-beta search, matrix multiplication and the all pairs shortest paths problem. The measurements show that both paradigms can be used efficientlymore » for programming large-grain parallel applications. Significant speedups were obtained on all applications. The unstructured Shared Virtual Memory paradigm achieves the best absolute performance, although this is largely due to the preliminary nature of the Orca compiler used. The structured shared data-object model achieves the highest speedups and is much easier to program and to debug.« less

Xyce™ Parallel Electronic Simulator Users' Guide, Version 6.5.

DOE Office of Scientific and Technical Information (OSTI.GOV)

Keiter, Eric R.; Aadithya, Karthik V.; Mei, Ting

This manual describes the use of the Xyce Parallel Electronic Simulator. Xyce has been designed as a SPICE-compatible, high-performance analog circuit simulator, and has been written to support the simulation needs of the Sandia National Laboratories electrical designers. This development has focused on improving capability over the current state-of-the-art in the following areas: Capability to solve extremely large circuit problems by supporting large-scale parallel computing platforms (up to thousands of processors). This includes support for most popular parallel and serial computers. A differential-algebraic-equation (DAE) formulation, which better isolates the device model package from solver algorithms. This allows one to developmore » new types of analysis without requiring the implementation of analysis-specific device models. Device models that are specifically tailored to meet Sandia's needs, including some radiation- aware devices (for Sandia users only). Object-oriented code design and implementation using modern coding practices. Xyce is a parallel code in the most general sense of the phrase -- a message passing parallel implementation -- which allows it to run efficiently a wide range of computing platforms. These include serial, shared-memory and distributed-memory parallel platforms. Attention has been paid to the specific nature of circuit-simulation problems to ensure that optimal parallel efficiency is achieved as the number of processors grows. The information herein is subject to change without notice. Copyright © 2002-2016 Sandia Corporation. All rights reserved.« less

Two-dimensional shape recognition using sparse distributed memory

NASA Technical Reports Server (NTRS)

Kanerva, Pentti; Olshausen, Bruno

1990-01-01

Researchers propose a method for recognizing two-dimensional shapes (hand-drawn characters, for example) with an associative memory. The method consists of two stages: first, the image is preprocessed to extract tangents to the contour of the shape; second, the set of tangents is converted to a long bit string for recognition with sparse distributed memory (SDM). SDM provides a simple, massively parallel architecture for an associative memory. Long bit vectors (256 to 1000 bits, for example) serve as both data and addresses to the memory, and patterns are grouped or classified according to similarity in Hamming distance. At the moment, tangents are extracted in a simple manner by progressively blurring the image and then using a Canny-type edge detector (Canny, 1986) to find edges at each stage of blurring. This results in a grid of tangents. While the technique used for obtaining the tangents is at present rather ad hoc, researchers plan to adopt an existing framework for extracting edge orientation information over a variety of resolutions, such as suggested by Watson (1987, 1983), Marr and Hildreth (1980), or Canny (1986).

Shared direct memory access on the Explorer 2-LX

NASA Technical Reports Server (NTRS)

Musgrave, Jeffrey L.

1990-01-01

Advances in Expert System technology and Artificial Intelligence have provided a framework for applying automated Intelligence to the solution of problems which were generally perceived as intractable using more classical approaches. As a result, hybrid architectures and parallel processing capability have become more common in computing environments. The Texas Instruments Explorer II-LX is an example of a machine which combines a symbolic processing environment, and a computationally oriented environment in a single chassis for integrated problem solutions. This user's manual is an attempt to make these capabilities more accessible to a wider range of engineers and programmers with problems well suited to solution in such an environment.

National Combustion Code Parallel Performance Enhancements

NASA Technical Reports Server (NTRS)

Quealy, Angela; Benyo, Theresa (Technical Monitor)

2002-01-01

The National Combustion Code (NCC) is being developed by an industry-government team for the design and analysis of combustion systems. The unstructured grid, reacting flow code uses a distributed memory, message passing model for its parallel implementation. The focus of the present effort has been to improve the performance of the NCC code to meet combustor designer requirements for model accuracy and analysis turnaround time. Improving the performance of this code contributes significantly to the overall reduction in time and cost of the combustor design cycle. This report describes recent parallel processing modifications to NCC that have improved the parallel scalability of the code, enabling a two hour turnaround for a 1.3 million element fully reacting combustion simulation on an SGI Origin 2000.

The Effects of Different Electrode Types for Obtaining Surface Machining Shape on Shape Memory Alloy Using Electrochemical Machining

NASA Astrophysics Data System (ADS)

Choi, S. G.; Kim, S. H.; Choi, W. K.; Moon, G. C.; Lee, E. S.

2017-06-01

Shape memory alloy (SMA) is important material used for the medicine and aerospace industry due to its characteristics called the shape memory effect, which involves the recovery of deformed alloy to its original state through the application of temperature or stress. Consumers in modern society demand stability in parts. Electrochemical machining is one of the methods for obtained these stabilities in parts requirements. These parts of shape memory alloy require fine patterns in some applications. In order to machine a fine pattern, the electrochemical machining method is suitable. For precision electrochemical machining using different shape electrodes, the current density should be controlled precisely. And electrode shape is required for precise electrochemical machining. It is possible to obtain precise square holes on the SMA if the insulation layer controlled the unnecessary current between electrode and workpiece. If it is adjusting the unnecessary current to obtain the desired shape, it will be a great contribution to the medical industry and the aerospace industry. It is possible to process a desired shape to the shape memory alloy by micro controlling the unnecessary current. In case of the square electrode without insulation layer, it derives inexact square holes due to the unnecessary current. The results using the insulated electrode in only side show precise square holes. The removal rate improved in case of insulated electrode than others because insulation layer concentrate the applied current to the machining zone.

A portable MPI-based parallel vector template library

NASA Technical Reports Server (NTRS)

Sheffler, Thomas J.

1995-01-01

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C++ by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of C or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

A Portable MPI-Based Parallel Vector Template Library

NASA Technical Reports Server (NTRS)

Sheffler, Thomas J.

1995-01-01

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C + + by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of c or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Performance Analysis of Multilevel Parallel Applications on Shared Memory Architectures

NASA Technical Reports Server (NTRS)

Biegel, Bryan A. (Technical Monitor); Jost, G.; Jin, H.; Labarta J.; Gimenez, J.; Caubet, J.

2003-01-01

Parallel programming paradigms include process level parallelism, thread level parallelization, and multilevel parallelism. This viewgraph presentation describes a detailed performance analysis of these paradigms for Shared Memory Architecture (SMA). This analysis uses the Paraver Performance Analysis System. The presentation includes diagrams of a flow of useful computations.

Testing New Programming Paradigms with NAS Parallel Benchmarks

NASA Technical Reports Server (NTRS)

Jin, H.; Frumkin, M.; Schultz, M.; Yan, J.

2000-01-01

Over the past decade, high performance computing has evolved rapidly, not only in hardware architectures but also with increasing complexity of real applications. Technologies have been developing to aim at scaling up to thousands of processors on both distributed and shared memory systems. Development of parallel programs on these computers is always a challenging task. Today, writing parallel programs with message passing (e.g. MPI) is the most popular way of achieving scalability and high performance. However, writing message passing programs is difficult and error prone. Recent years new effort has been made in defining new parallel programming paradigms. The best examples are: HPF (based on data parallelism) and OpenMP (based on shared memory parallelism). Both provide simple and clear extensions to sequential programs, thus greatly simplify the tedious tasks encountered in writing message passing programs. HPF is independent of memory hierarchy, however, due to the immaturity of compiler technology its performance is still questionable. Although use of parallel compiler directives is not new, OpenMP offers a portable solution in the shared-memory domain. Another important development involves the tremendous progress in the internet and its associated technology. Although still in its infancy, Java promisses portability in a heterogeneous environment and offers possibility to "compile once and run anywhere." In light of testing these new technologies, we implemented new parallel versions of the NAS Parallel Benchmarks (NPBs) with HPF and OpenMP directives, and extended the work with Java and Java-threads. The purpose of this study is to examine the effectiveness of alternative programming paradigms. NPBs consist of five kernels and three simulated applications that mimic the computation and data movement of large scale computational fluid dynamics (CFD) applications. We started with the serial version included in NPB2.3. Optimization of memory and cache usage was applied to several benchmarks, noticeably BT and SP, resulting in better sequential performance. In order to overcome the lack of an HPF performance model and guide the development of the HPF codes, we employed an empirical performance model for several primitives found in the benchmarks. We encountered a few limitations of HPF, such as lack of supporting the "REDISTRIBUTION" directive and no easy way to handle irregular computation. The parallelization with OpenMP directives was done at the outer-most loop level to achieve the largest granularity. The performance of six HPF and OpenMP benchmarks is compared with their MPI counterparts for the Class-A problem size in the figure in next page. These results were obtained on an SGI Origin2000 (195MHz) with MIPSpro-f77 compiler 7.2.1 for OpenMP and MPI codes and PGI pghpf-2.4.3 compiler with MPI interface for HPF programs.

A Component-Based FPGA Design Framework for Neuronal Ion Channel Dynamics Simulations

PubMed Central

Mak, Terrence S. T.; Rachmuth, Guy; Lam, Kai-Pui; Poon, Chi-Sang

2008-01-01

Neuron-machine interfaces such as dynamic clamp and brain-implantable neuroprosthetic devices require real-time simulations of neuronal ion channel dynamics. Field Programmable Gate Array (FPGA) has emerged as a high-speed digital platform ideal for such application-specific computations. We propose an efficient and flexible component-based FPGA design framework for neuronal ion channel dynamics simulations, which overcomes certain limitations of the recently proposed memory-based approach. A parallel processing strategy is used to minimize computational delay, and a hardware-efficient factoring approach for calculating exponential and division functions in neuronal ion channel models is used to conserve resource consumption. Performances of the various FPGA design approaches are compared theoretically and experimentally in corresponding implementations of the AMPA and NMDA synaptic ion channel models. Our results suggest that the component-based design framework provides a more memory economic solution as well as more efficient logic utilization for large word lengths, whereas the memory-based approach may be suitable for time-critical applications where a higher throughput rate is desired. PMID:17190033

Implementing Access to Data Distributed on Many Processors

NASA Technical Reports Server (NTRS)

James, Mark

2006-01-01

A reference architecture is defined for an object-oriented implementation of domains, arrays, and distributions written in the programming language Chapel. This technology primarily addresses domains that contain arrays that have regular index sets with the low-level implementation details being beyond the scope of this discussion. What is defined is a complete set of object-oriented operators that allows one to perform data distributions for domain arrays involving regular arithmetic index sets. What is unique is that these operators allow for the arbitrary regions of the arrays to be fragmented and distributed across multiple processors with a single point of access giving the programmer the illusion that all the elements are collocated on a single processor. Today's massively parallel High Productivity Computing Systems (HPCS) are characterized by a modular structure, with a large number of processing and memory units connected by a high-speed network. Locality of access as well as load balancing are primary concerns in these systems that are typically used for high-performance scientific computation. Data distributions address these issues by providing a range of methods for spreading large data sets across the components of a system. Over the past two decades, many languages, systems, tools, and libraries have been developed for the support of distributions. Since the performance of data parallel applications is directly influenced by the distribution strategy, users often resort to low-level programming models that allow fine-tuning of the distribution aspects affecting performance, but, at the same time, are tedious and error-prone. This technology presents a reusable design of a data-distribution framework for data parallel high-performance applications. Distributions are a means to express locality in systems composed of large numbers of processor and memory components connected by a network. Since distributions have a great effect on the performance of applications, it is important that the distribution strategy is flexible, so its behavior can change depending on the needs of the application. At the same time, high productivity concerns require that the user be shielded from error-prone, tedious details such as communication and synchronization.

Some Problems and Solutions in Transferring Ecosystem Simulation Codes to Supercomputers

NASA Technical Reports Server (NTRS)

Skiles, J. W.; Schulbach, C. H.

1994-01-01

Many computer codes for the simulation of ecological systems have been developed in the last twenty-five years. This development took place initially on main-frame computers, then mini-computers, and more recently, on micro-computers and workstations. Recent recognition of ecosystem science as a High Performance Computing and Communications Program Grand Challenge area emphasizes supercomputers (both parallel and distributed systems) as the next set of tools for ecological simulation. Transferring ecosystem simulation codes to such systems is not a matter of simply compiling and executing existing code on the supercomputer since there are significant differences in the system architectures of sequential, scalar computers and parallel and/or vector supercomputers. To more appropriately match the application to the architecture (necessary to achieve reasonable performance), the parallelism (if it exists) of the original application must be exploited. We discuss our work in transferring a general grassland simulation model (developed on a VAX in the FORTRAN computer programming language) to a Cray Y-MP. We show the Cray shared-memory vector-architecture, and discuss our rationale for selecting the Cray. We describe porting the model to the Cray and executing and verifying a baseline version, and we discuss the changes we made to exploit the parallelism in the application and to improve code execution. As a result, the Cray executed the model 30 times faster than the VAX 11/785 and 10 times faster than a Sun 4 workstation. We achieved an additional speed-up of approximately 30 percent over the original Cray run by using the compiler's vectorizing capabilities and the machine's ability to put subroutines and functions "in-line" in the code. With the modifications, the code still runs at only about 5% of the Cray's peak speed because it makes ineffective use of the vector processing capabilities of the Cray. We conclude with a discussion and future plans.

A Parallel Algorithm for Contact in a Finite Element Hydrocode

DOE Office of Scientific and Technical Information (OSTI.GOV)

Pierce, Timothy G.

A parallel algorithm is developed for contact/impact of multiple three dimensional bodies undergoing large deformation. As time progresses the relative positions of contact between the multiple bodies changes as collision and sliding occurs. The parallel algorithm is capable of tracking these changes and enforcing an impenetrability constraint and momentum transfer across the surfaces in contact. Portions of the various surfaces of the bodies are assigned to the processors of a distributed-memory parallel machine in an arbitrary fashion, known as the primary decomposition. A secondary, dynamic decomposition is utilized to bring opposing sections of the contacting surfaces together on the samemore » processors, so that opposing forces may be balanced and the resultant deformation of the bodies calculated. The secondary decomposition is accomplished and updated using only local communication with a limited subset of neighbor processors. Each processor represents both a domain of the primary decomposition and a domain of the secondary, or contact, decomposition. Thus each processor has four sets of neighbor processors: (a) those processors which represent regions adjacent to it in the primary decomposition, (b) those processors which represent regions adjacent to it in the contact decomposition, (c) those processors which send it the data from which it constructs its contact domain, and (d) those processors to which it sends its primary domain data, from which they construct their contact domains. The latter three of these neighbor sets change dynamically as the simulation progresses. By constraining all communication to these sets of neighbors, all global communication, with its attendant nonscalable performance, is avoided. A set of tests are provided to measure the degree of scalability achieved by this algorithm on up to 1024 processors. Issues related to the operating system of the test platform which lead to some degradation of the results are analyzed. This algorithm has been implemented as the contact capability of the ALE3D multiphysics code, and is currently in production use.« less

Facilitating arrhythmia simulation: the method of quantitative cellular automata modeling and parallel running

PubMed Central

Zhu, Hao; Sun, Yan; Rajagopal, Gunaretnam; Mondry, Adrian; Dhar, Pawan

2004-01-01

Background Many arrhythmias are triggered by abnormal electrical activity at the ionic channel and cell level, and then evolve spatio-temporally within the heart. To understand arrhythmias better and to diagnose them more precisely by their ECG waveforms, a whole-heart model is required to explore the association between the massively parallel activities at the channel/cell level and the integrative electrophysiological phenomena at organ level. Methods We have developed a method to build large-scale electrophysiological models by using extended cellular automata, and to run such models on a cluster of shared memory machines. We describe here the method, including the extension of a language-based cellular automaton to implement quantitative computing, the building of a whole-heart model with Visible Human Project data, the parallelization of the model on a cluster of shared memory computers with OpenMP and MPI hybrid programming, and a simulation algorithm that links cellular activity with the ECG. Results We demonstrate that electrical activities at channel, cell, and organ levels can be traced and captured conveniently in our extended cellular automaton system. Examples of some ECG waveforms simulated with a 2-D slice are given to support the ECG simulation algorithm. A performance evaluation of the 3-D model on a four-node cluster is also given. Conclusions Quantitative multicellular modeling with extended cellular automata is a highly efficient and widely applicable method to weave experimental data at different levels into computational models. This process can be used to investigate complex and collective biological activities that can be described neither by their governing differentiation equations nor by discrete parallel computation. Transparent cluster computing is a convenient and effective method to make time-consuming simulation feasible. Arrhythmias, as a typical case, can be effectively simulated with the methods described. PMID:15339335

Cognitive neuroepigenetics: the next evolution in our understanding of the molecular mechanisms underlying learning and memory?

NASA Astrophysics Data System (ADS)

Marshall, Paul; Bredy, Timothy W.

2016-07-01

A complete understanding of the fundamental mechanisms of learning and memory continues to elude neuroscientists. Although many important discoveries have been made, the question of how memories are encoded and maintained at the molecular level remains. So far, this issue has been framed within the context of one of the most dominant concepts in molecular biology, the central dogma, and the result has been a protein-centric view of memory. Here, we discuss the evidence supporting a role for neuroepigenetic mechanisms, which constitute dynamic and reversible, state-dependent modifications at all levels of control over cellular function, and their role in learning and memory. This neuroepigenetic view suggests that DNA, RNA and protein each influence one another to produce a holistic cellular state that contributes to the formation and maintenance of memory, and predicts a parallel and distributed system for the consolidation, storage and retrieval of the engram.

Optical interconnection network for parallel access to multi-rank memory in future computing systems.

PubMed

Wang, Kang; Gu, Huaxi; Yang, Yintang; Wang, Kun

2015-08-10

With the number of cores increasing, there is an emerging need for a high-bandwidth low-latency interconnection network, serving core-to-memory communication. In this paper, aiming at the goal of simultaneous access to multi-rank memory, we propose an optical interconnection network for core-to-memory communication. In the proposed network, the wavelength usage is delicately arranged so that cores can communicate with different ranks at the same time and broadcast for flow control can be achieved. A distributed memory controller architecture that works in a pipeline mode is also designed for efficient optical communication and transaction address processes. The scaling method and wavelength assignment for the proposed network are investigated. Compared with traditional electronic bus-based core-to-memory communication, the simulation results based on the PARSEC benchmark show that the bandwidth enhancement and latency reduction are apparent.

Cognitive neuroepigenetics: the next evolution in our understanding of the molecular mechanisms underlying learning and memory?

PubMed Central

Marshall, Paul; Bredy, Timothy W.

2016-01-01

A complete understanding of the fundamental mechanisms of learning and memory continues to elude neuroscientists. Although many important discoveries have been made, the question of how memories are encoded and maintained at the molecular level remains. To date, this issue has been framed within the context of one of the most dominant concepts in molecular biology, the central dogma, and the result has been a protein-centric view of memory. Here we discuss the evidence supporting a role for neuroepigenetic mechanisms, which constitute dynamic and reversible, state-dependent modifications at all levels of control over cellular function, and their role in learning and memory. This neuroepigenetic view suggests that DNA, RNA and protein each influence one another to produce a holistic cellular state that contributes to the formation and maintenance of memory, and predicts a parallel and distributed system for the consolidation, storage and retrieval of the engram. PMID:27512601

A parallel algorithm for 2D visco-acoustic frequency-domain full-waveform inversion: application to a dense OBS data set

NASA Astrophysics Data System (ADS)

Sourbier, F.; Operto, S.; Virieux, J.

2006-12-01

We present a distributed-memory parallel algorithm for 2D visco-acoustic full-waveform inversion of wide-angle seismic data. Our code is written in fortran90 and use MPI for parallelism. The algorithm was applied to real wide-angle data set recorded by 100 OBSs with a 1-km spacing in the eastern-Nankai trough (Japan) to image the deep structure of the subduction zone. Full-waveform inversion is applied sequentially to discrete frequencies by proceeding from the low to the high frequencies. The inverse problem is solved with a classic gradient method. Full-waveform modeling is performed with a frequency-domain finite-difference method. In the frequency-domain, solving the wave equation requires resolution of a large unsymmetric system of linear equations. We use the massively parallel direct solver MUMPS (http://www.enseeiht.fr/irit/apo/MUMPS) for distributed-memory computer to solve this system. The MUMPS solver is based on a multifrontal method for the parallel factorization. The MUMPS algorithm is subdivided in 3 main steps: a symbolic analysis step that performs re-ordering of the matrix coefficients to minimize the fill-in of the matrix during the subsequent factorization and an estimation of the assembly tree of the matrix. Second, the factorization is performed with dynamic scheduling to accomodate numerical pivoting and provides the LU factors distributed over all the processors. Third, the resolution is performed for multiple sources. To compute the gradient of the cost function, 2 simulations per shot are required (one to compute the forward wavefield and one to back-propagate residuals). The multi-source resolutions can be performed in parallel with MUMPS. In the end, each processor stores in core a sub-domain of all the solutions. These distributed solutions can be exploited to compute in parallel the gradient of the cost function. Since the gradient of the cost function is a weighted stack of the shot and residual solutions of MUMPS, each processor computes the corresponding sub-domain of the gradient. In the end, the gradient is centralized on the master processor using a collective communation. The gradient is scaled by the diagonal elements of the Hessian matrix. This scaling is computed only once per frequency before the first iteration of the inversion. Estimation of the diagonal terms of the Hessian requires performing one simulation per non redondant shot and receiver position. The same strategy that the one used for the gradient is used to compute the diagonal Hessian in parallel. This algorithm was applied to a dense wide-angle data set recorded by 100 OBSs in the eastern Nankai trough, offshore Japan. Thirteen frequencies ranging from 3 and 15 Hz were inverted. Tweny iterations per frequency were computed leading to 260 tomographic velocity models of increasing resolution. The velocity model dimensions are 105 km x 25 km corresponding to a finite-difference grid of 4201 x 1001 grid with a 25-m grid interval. The number of shot was 1005 and the number of inverted OBS gathers was 93. The inversion requires 20 days on 6 32-bits bi-processor nodes with 4 Gbytes of RAM memory per node when only the LU factorization is performed in parallel. Preliminary estimations of the time required to perform the inversion with the fully-parallelized code is 6 and 4 days using 20 and 50 processors respectively.

Using data tagging to improve the performance of Kanerva's sparse distributed memory

NASA Technical Reports Server (NTRS)

Rogers, David

1988-01-01

The standard formulation of Kanerva's sparse distributed memory (SDM) involves the selection of a large number of data storage locations, followed by averaging the data contained in those locations to reconstruct the stored data. A variant of this model is discussed, in which the predominant pattern is the focus of reconstruction. First, one architecture is proposed which returns the predominant pattern rather than the average pattern. However, this model will require too much storage for most uses. Next, a hybrid model is proposed, called tagged SDM, which approximates the results of the predominant pattern machine, but is nearly as efficient as Kanerva's original formulation. Finally, some experimental results are shown which confirm that significant improvements in the recall capability of SDM can be achieved using the tagged architecture.

Compiling global name-space programs for distributed execution

NASA Technical Reports Server (NTRS)

Koelbel, Charles; Mehrotra, Piyush

1990-01-01

Distributed memory machines do not provide hardware support for a global address space. Thus programmers are forced to partition the data across the memories of the architecture and use explicit message passing to communicate data between processors. The compiler support required to allow programmers to express their algorithms using a global name-space is examined. A general method is presented for analysis of a high level source program and along with its translation to a set of independently executing tasks communicating via messages. If the compiler has enough information, this translation can be carried out at compile-time. Otherwise run-time code is generated to implement the required data movement. The analysis required in both situations is described and the performance of the generated code on the Intel iPSC/2 is presented.

Parallel Computational Fluid Dynamics: Current Status and Future Requirements

NASA Technical Reports Server (NTRS)

Simon, Horst D.; VanDalsem, William R.; Dagum, Leonardo; Kutler, Paul (Technical Monitor)

1994-01-01

One or the key objectives of the Applied Research Branch in the Numerical Aerodynamic Simulation (NAS) Systems Division at NASA Allies Research Center is the accelerated introduction of highly parallel machines into a full operational environment. In this report we discuss the performance results obtained from the implementation of some computational fluid dynamics (CFD) applications on the Connection Machine CM-2 and the Intel iPSC/860. We summarize some of the experiences made so far with the parallel testbed machines at the NAS Applied Research Branch. Then we discuss the long term computational requirements for accomplishing some of the grand challenge problems in computational aerosciences. We argue that only massively parallel machines will be able to meet these grand challenge requirements, and we outline the computer science and algorithm research challenges ahead.

Schemas in Problem Solving: An Integrated Model of Learning, Memory, and Instruction

DTIC Science & Technology

1992-01-01

reflected in the title of a recent article: "lybid Coupation, in Cognitive Science: Neural Networks ad Symbl (3. A Andesson, 1990). And, Marvin Mtuky...Rumneihart, D. E (1989). Explorations in parallel distributed processing: A handbook of models, programs, and exercises. Cambridge, MA: The MrT Press. Minsky

Flexible Language Constructs for Large Parallel Programs

DOE PAGES

Rosing, Matt; Schnabel, Robert

1994-01-01

The goal of the research described in this article is to develop flexible language constructs for writing large data parallel numerical programs for distributed memory (multiple instruction multiple data [MIMD]) multiprocessors. Previously, several models have been developed to support synchronization and communication. Models for global synchronization include single instruction multiple data (SIMD), single program multiple data (SPMD), and sequential programs annotated with data distribution statements. The two primary models for communication include implicit communication based on shared memory and explicit communication based on messages. None of these models by themselves seem sufficient to permit the natural and efficient expression ofmore » the variety of algorithms that occur in large scientific computations. In this article, we give an overview of a new language that combines many of these programming models in a clean manner. This is done in a modular fashion such that different models can be combined to support large programs. Within a module, the selection of a model depends on the algorithm and its efficiency requirements. In this article, we give an overview of the language and discuss some of the critical implementation details.« less

Distributed shared memory for roaming large volumes.

PubMed

Castanié, Laurent; Mion, Christophe; Cavin, Xavier; Lévy, Bruno

2006-01-01

We present a cluster-based volume rendering system for roaming very large volumes. This system allows to move a gigabyte-sized probe inside a total volume of several tens or hundreds of gigabytes in real-time. While the size of the probe is limited by the total amount of texture memory on the cluster, the size of the total data set has no theoretical limit. The cluster is used as a distributed graphics processing unit that both aggregates graphics power and graphics memory. A hardware-accelerated volume renderer runs in parallel on the cluster nodes and the final image compositing is implemented using a pipelined sort-last rendering algorithm. Meanwhile, volume bricking and volume paging allow efficient data caching. On each rendering node, a distributed hierarchical cache system implements a global software-based distributed shared memory on the cluster. In case of a cache miss, this system first checks page residency on the other cluster nodes instead of directly accessing local disks. Using two Gigabit Ethernet network interfaces per node, we accelerate data fetching by a factor of 4 compared to directly accessing local disks. The system also implements asynchronous disk access and texture loading, which makes it possible to overlap data loading, volume slicing and rendering for optimal volume roaming.

Automation of Data Traffic Control on DSM Architecture

NASA Technical Reports Server (NTRS)

Frumkin, Michael; Jin, Hao-Qiang; Yan, Jerry

2001-01-01

The design of distributed shared memory (DSM) computers liberates users from the duty to distribute data across processors and allows for the incremental development of parallel programs using, for example, OpenMP or Java threads. DSM architecture greatly simplifies the development of parallel programs having good performance on a few processors. However, to achieve a good program scalability on DSM computers requires that the user understand data flow in the application and use various techniques to avoid data traffic congestions. In this paper we discuss a number of such techniques, including data blocking, data placement, data transposition and page size control and evaluate their efficiency on the NAS (NASA Advanced Supercomputing) Parallel Benchmarks. We also present a tool which automates the detection of constructs causing data congestions in Fortran array oriented codes and advises the user on code transformations for improving data traffic in the application.

High order parallel numerical schemes for solving incompressible flows

NASA Technical Reports Server (NTRS)

Lin, Avi; Milner, Edward J.; Liou, May-Fun; Belch, Richard A.

1992-01-01

The use of parallel computers for numerically solving flow fields has gained much importance in recent years. This paper introduces a new high order numerical scheme for computational fluid dynamics (CFD) specifically designed for parallel computational environments. A distributed MIMD system gives the flexibility of treating different elements of the governing equations with totally different numerical schemes in different regions of the flow field. The parallel decomposition of the governing operator to be solved is the primary parallel split. The primary parallel split was studied using a hypercube like architecture having clusters of shared memory processors at each node. The approach is demonstrated using examples of simple steady state incompressible flows. Future studies should investigate the secondary split because, depending on the numerical scheme that each of the processors applies and the nature of the flow in the specific subdomain, it may be possible for a processor to seek better, or higher order, schemes for its particular subcase.

Partitioning problems in parallel, pipelined and distributed computing

NASA Technical Reports Server (NTRS)

Bokhari, S.

1985-01-01

The problem of optimally assigning the modules of a parallel program over the processors of a multiple computer system is addressed. A Sum-Bottleneck path algorithm is developed that permits the efficient solution of many variants of this problem under some constraints on the structure of the partitions. In particular, the following problems are solved optimally for a single-host, multiple satellite system: partitioning multiple chain structured parallel programs, multiple arbitrarily structured serial programs and single tree structured parallel programs. In addition, the problems of partitioning chain structured parallel programs across chain connected systems and across shared memory (or shared bus) systems are also solved under certain constraints. All solutions for parallel programs are equally applicable to pipelined programs. These results extend prior research in this area by explicitly taking concurrency into account and permit the efficient utilization of multiple computer architectures for a wide range of problems of practical interest.

National Combustion Code: Parallel Implementation and Performance

NASA Technical Reports Server (NTRS)

Quealy, A.; Ryder, R.; Norris, A.; Liu, N.-S.

2000-01-01

The National Combustion Code (NCC) is being developed by an industry-government team for the design and analysis of combustion systems. CORSAIR-CCD is the current baseline reacting flow solver for NCC. This is a parallel, unstructured grid code which uses a distributed memory, message passing model for its parallel implementation. The focus of the present effort has been to improve the performance of the NCC flow solver to meet combustor designer requirements for model accuracy and analysis turnaround time. Improving the performance of this code contributes significantly to the overall reduction in time and cost of the combustor design cycle. This paper describes the parallel implementation of the NCC flow solver and summarizes its current parallel performance on an SGI Origin 2000. Earlier parallel performance results on an IBM SP-2 are also included. The performance improvements which have enabled a turnaround of less than 15 hours for a 1.3 million element fully reacting combustion simulation are described.

Quantum information, cognition, and music.

PubMed

Dalla Chiara, Maria L; Giuntini, Roberto; Leporini, Roberto; Negri, Eleonora; Sergioli, Giuseppe

2015-01-01

Parallelism represents an essential aspect of human mind/brain activities. One can recognize some common features between psychological parallelism and the characteristic parallel structures that arise in quantum theory and in quantum computation. The article is devoted to a discussion of the following questions: a comparison between classical probabilistic Turing machines and quantum Turing machines.possible applications of the quantum computational semantics to cognitive problems.parallelism in music.

Quantum information, cognition, and music

PubMed Central

Dalla Chiara, Maria L.; Giuntini, Roberto; Leporini, Roberto; Negri, Eleonora; Sergioli, Giuseppe

2015-01-01

Parallelism represents an essential aspect of human mind/brain activities. One can recognize some common features between psychological parallelism and the characteristic parallel structures that arise in quantum theory and in quantum computation. The article is devoted to a discussion of the following questions: a comparison between classical probabilistic Turing machines and quantum Turing machines.possible applications of the quantum computational semantics to cognitive problems.parallelism in music. PMID:26539139

Distributed Fading Memory for Stimulus Properties in the Primary Visual Cortex

PubMed Central

Singer, Wolf; Maass, Wolfgang

2009-01-01

It is currently not known how distributed neuronal responses in early visual areas carry stimulus-related information. We made multielectrode recordings from cat primary visual cortex and applied methods from machine learning in order to analyze the temporal evolution of stimulus-related information in the spiking activity of large ensembles of around 100 neurons. We used sequences of up to three different visual stimuli (letters of the alphabet) presented for 100 ms and with intervals of 100 ms or larger. Most of the information about visual stimuli extractable by sophisticated methods of machine learning, i.e., support vector machines with nonlinear kernel functions, was also extractable by simple linear classification such as can be achieved by individual neurons. New stimuli did not erase information about previous stimuli. The responses to the most recent stimulus contained about equal amounts of information about both this and the preceding stimulus. This information was encoded both in the discharge rates (response amplitudes) of the ensemble of neurons and, when using short time constants for integration (e.g., 20 ms), in the precise timing of individual spikes (≤∼20 ms), and persisted for several 100 ms beyond the offset of stimuli. The results indicate that the network from which we recorded is endowed with fading memory and is capable of performing online computations utilizing information about temporally sequential stimuli. This result challenges models assuming frame-by-frame analyses of sequential inputs. PMID:20027205

Machine Learning in Medicine

PubMed Central

Deo, Rahul C.

2015-01-01

Spurred by advances in processing power, memory, storage, and an unprecedented wealth of data, computers are being asked to tackle increasingly complex learning tasks, often with astonishing success. Computers have now mastered a popular variant of poker, learned the laws of physics from experimental data, and become experts in video games – tasks which would have been deemed impossible not too long ago. In parallel, the number of companies centered on applying complex data analysis to varying industries has exploded, and it is thus unsurprising that some analytic companies are turning attention to problems in healthcare. The purpose of this review is to explore what problems in medicine might benefit from such learning approaches and use examples from the literature to introduce basic concepts in machine learning. It is important to note that seemingly large enough medical data sets and adequate learning algorithms have been available for many decades – and yet, although there are thousands of papers applying machine learning algorithms to medical data, very few have contributed meaningfully to clinical care. This lack of impact stands in stark contrast to the enormous relevance of machine learning to many other industries. Thus part of my effort will be to identify what obstacles there may be to changing the practice of medicine through statistical learning approaches, and discuss how these might be overcome. PMID:26572668

Parallelization of the TRIGRS model for rainfall-induced landslides using the message passing interface

USGS Publications Warehouse

Alvioli, M.; Baum, R.L.

2016-01-01

We describe a parallel implementation of TRIGRS, the Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability Model for the timing and distribution of rainfall-induced shallow landslides. We have parallelized the four time-demanding execution modes of TRIGRS, namely both the saturated and unsaturated model with finite and infinite soil depth options, within the Message Passing Interface framework. In addition to new features of the code, we outline details of the parallel implementation and show the performance gain with respect to the serial code. Results are obtained both on commercial hardware and on a high-performance multi-node machine, showing the different limits of applicability of the new code. We also discuss the implications for the application of the model on large-scale areas and as a tool for real-time landslide hazard monitoring.

Implementation of a Parallel Kalman Filter for Stratospheric Chemical Tracer Assimilation

NASA Technical Reports Server (NTRS)

Chang, Lang-Ping; Lyster, Peter M.; Menard, R.; Cohn, S. E.

1998-01-01

A Kalman filter for the assimilation of long-lived atmospheric chemical constituents has been developed for two-dimensional transport models on isentropic surfaces over the globe. An important attribute of the Kalman filter is that it calculates error covariances of the constituent fields using the tracer dynamics. Consequently, the current Kalman-filter assimilation is a five-dimensional problem (coordinates of two points and time), and it can only be handled on computers with large memory and high floating point speed. In this paper, an implementation of the Kalman filter for distributed-memory, message-passing parallel computers is discussed. Two approaches were studied: an operator decomposition and a covariance decomposition. The latter was found to be more scalable than the former, and it possesses the property that the dynamical model does not need to be parallelized, which is of considerable practical advantage. This code is currently used to assimilate constituent data retrieved by limb sounders on the Upper Atmosphere Research Satellite. Tests of the code examined the variance transport and observability properties. Aspects of the parallel implementation, some timing results, and a brief discussion of the physical results will be presented.

A Dual Super-Element Domain Decomposition Approach for Parallel Nonlinear Finite Element Analysis

NASA Astrophysics Data System (ADS)

Jokhio, G. A.; Izzuddin, B. A.

2015-05-01

This article presents a new domain decomposition method for nonlinear finite element analysis introducing the concept of dual partition super-elements. The method extends ideas from the displacement frame method and is ideally suited for parallel nonlinear static/dynamic analysis of structural systems. In the new method, domain decomposition is realized by replacing one or more subdomains in a "parent system," each with a placeholder super-element, where the subdomains are processed separately as "child partitions," each wrapped by a dual super-element along the partition boundary. The analysis of the overall system, including the satisfaction of equilibrium and compatibility at all partition boundaries, is realized through direct communication between all pairs of placeholder and dual super-elements. The proposed method has particular advantages for matrix solution methods based on the frontal scheme, and can be readily implemented for existing finite element analysis programs to achieve parallelization on distributed memory systems with minimal intervention, thus overcoming memory bottlenecks typically faced in the analysis of large-scale problems. Several examples are presented in this article which demonstrate the computational benefits of the proposed parallel domain decomposition approach and its applicability to the nonlinear structural analysis of realistic structural systems.

Simplified Parallel Domain Traversal

DOE Office of Scientific and Technical Information (OSTI.GOV)

Erickson III, David J

2011-01-01

Many data-intensive scientific analysis techniques require global domain traversal, which over the years has been a bottleneck for efficient parallelization across distributed-memory architectures. Inspired by MapReduce and other simplified parallel programming approaches, we have designed DStep, a flexible system that greatly simplifies efficient parallelization of domain traversal techniques at scale. In order to deliver both simplicity to users as well as scalability on HPC platforms, we introduce a novel two-tiered communication architecture for managing and exploiting asynchronous communication loads. We also integrate our design with advanced parallel I/O techniques that operate directly on native simulation output. We demonstrate DStep bymore » performing teleconnection analysis across ensemble runs of terascale atmospheric CO{sub 2} and climate data, and we show scalability results on up to 65,536 IBM BlueGene/P cores.« less

The scheme machine: A case study in progress in design derivation at system levels

NASA Technical Reports Server (NTRS)

Johnson, Steven D.

1995-01-01

The Scheme Machine is one of several design projects of the Digital Design Derivation group at Indiana University. It differs from the other projects in its focus on issues of system design and its connection to surrounding research in programming language semantics, compiler construction, and programming methodology underway at Indiana and elsewhere. The genesis of the project dates to the early 1980's, when digital design derivation research branched from the surrounding research effort in programming languages. Both branches have continued to develop in parallel, with this particular project serving as a bridge. However, by 1990 there remained little real interaction between the branches and recently we have undertaken to reintegrate them. On the software side, researchers have refined a mathematically rigorous (but not mechanized) treatment starting with the fully abstract semantic definition of Scheme and resulting in an efficient implementation consisting of a compiler and virtual machine model, the latter typically realized with a general purpose microprocessor. The derivation includes a number of sophisticated factorizations and representations and is also deep example of the underlying engineering methodology. The hardware research has created a mechanized algebra supporting the tedious and massive transformations often seen at lower levels of design. This work has progressed to the point that large scale devices, such as processors, can be derived from first-order finite state machine specifications. This is roughly where the language oriented research stops; thus, together, the two efforts establish a thread from the highest levels of abstract specification to detailed digital implementation. The Scheme Machine project challenges hardware derivation research in several ways, although the individual components of the system are of a similar scale to those we have worked with before. The machine has a custom dual-ported memory to support garbage collection. It consists of four tightly coupled processes--processor, collector, allocator, memory--with a very non-trivial synchronization relationship. Finally, there are deep issues of representation for the run-time objects of a symbolic processing language. The research centers on verification through integrated formal reasoning systems, but is also involved with modeling and prototyping environments. Since the derivation algebra is basd on an executable modeling language, there is opportunity to incorporate design animation in the design process. We are looking for ways to move smoothly and incrementally from executable specifications into hardware realization. For example, we can run the garbage collector specification, a Scheme program, directly against the physical memory prototype, and similarly, the instruction processor model against the heap implementation.

Automated Instrumentation, Monitoring and Visualization of PVM Programs Using AIMS

NASA Technical Reports Server (NTRS)

Mehra, Pankaj; VanVoorst, Brian; Yan, Jerry; Lum, Henry, Jr. (Technical Monitor)

1994-01-01

We present views and analysis of the execution of several PVM (Parallel Virtual Machine) codes for Computational Fluid Dynamics on a networks of Sparcstations, including: (1) NAS Parallel Benchmarks CG and MG; (2) a multi-partitioning algorithm for NAS Parallel Benchmark SP; and (3) an overset grid flowsolver. These views and analysis were obtained using our Automated Instrumentation and Monitoring System (AIMS) version 3.0, a toolkit for debugging the performance of PVM programs. We will describe the architecture, operation and application of AIMS. The AIMS toolkit contains: (1) Xinstrument, which can automatically instrument various computational and communication constructs in message-passing parallel programs; (2) Monitor, a library of runtime trace-collection routines; (3) VK (Visual Kernel), an execution-animation tool with source-code clickback; and (4) Tally, a tool for statistical analysis of execution profiles. Currently, Xinstrument can handle C and Fortran 77 programs using PVM 3.2.x; Monitor has been implemented and tested on Sun 4 systems running SunOS 4.1.2; and VK uses XIIR5 and Motif 1.2. Data and views obtained using AIMS clearly illustrate several characteristic features of executing parallel programs on networked workstations: (1) the impact of long message latencies; (2) the impact of multiprogramming overheads and associated load imbalance; (3) cache and virtual-memory effects; and (4) significant skews between workstation clocks. Interestingly, AIMS can compensate for constant skew (zero drift) by calibrating the skew between a parent and its spawned children. In addition, AIMS' skew-compensation algorithm can adjust timestamps in a way that eliminates physically impossible communications (e.g., messages going backwards in time). Our current efforts are directed toward creating new views to explain the observed performance of PVM programs. Some of the features planned for the near future include: (1) ConfigView, showing the physical topology of the virtual machine, inferred using specially formatted IP (Internet Protocol) packets: and (2) LoadView, synchronous animation of PVM-program execution and resource-utilization patterns.

A Parallel Cartesian Approach for External Aerodynamics of Vehicles with Complex Geometry

NASA Technical Reports Server (NTRS)

Aftosmis, M. J.; Berger, M. J.; Adomavicius, G.

2001-01-01

This workshop paper presents the current status in the development of a new approach for the solution of the Euler equations on Cartesian meshes with embedded boundaries in three dimensions on distributed and shared memory architectures. The approach uses adaptively refined Cartesian hexahedra to fill the computational domain. Where these cells intersect the geometry, they are cut by the boundary into arbitrarily shaped polyhedra which receive special treatment by the solver. The presentation documents a newly developed multilevel upwind solver based on a flexible domain-decomposition strategy. One novel aspect of the work is its use of space-filling curves (SFC) for memory efficient on-the-fly parallelization, dynamic re-partitioning and automatic coarse mesh generation. Within each subdomain the approach employs a variety reordering techniques so that relevant data are on the same page in memory permitting high-performance on cache-based processors. Details of the on-the-fly SFC based partitioning are presented as are construction rules for the automatic coarse mesh generation. After describing the approach, the paper uses model problems and 3- D configurations to both verify and validate the solver. The model problems demonstrate that second-order accuracy is maintained despite the presence of the irregular cut-cells in the mesh. In addition, it examines both parallel efficiency and convergence behavior. These investigations demonstrate a parallel speed-up in excess of 28 on 32 processors of an SGI Origin 2000 system and confirm that mesh partitioning has no effect on convergence behavior.

Dynamic Load-Balancing for Distributed Heterogeneous Computing of Parallel CFD Problems

NASA Technical Reports Server (NTRS)

Ecer, A.; Chien, Y. P.; Boenisch, T.; Akay, H. U.

2000-01-01

The developed methodology is aimed at improving the efficiency of executing block-structured algorithms on parallel, distributed, heterogeneous computers. The basic approach of these algorithms is to divide the flow domain into many sub- domains called blocks, and solve the governing equations over these blocks. Dynamic load balancing problem is defined as the efficient distribution of the blocks among the available processors over a period of several hours of computations. In environments with computers of different architecture, operating systems, CPU speed, memory size, load, and network speed, balancing the loads and managing the communication between processors becomes crucial. Load balancing software tools for mutually dependent parallel processes have been created to efficiently utilize an advanced computation environment and algorithms. These tools are dynamic in nature because of the chances in the computer environment during execution time. More recently, these tools were extended to a second operating system: NT. In this paper, the problems associated with this application will be discussed. Also, the developed algorithms were combined with the load sharing capability of LSF to efficiently utilize workstation clusters for parallel computing. Finally, results will be presented on running a NASA based code ADPAC to demonstrate the developed tools for dynamic load balancing.

Neural networks and applications tutorial

NASA Astrophysics Data System (ADS)

Guyon, I.

1991-09-01

The importance of neural networks has grown dramatically during this decade. While only a few years ago they were primarily of academic interest, now dozens of companies and many universities are investigating the potential use of these systems and products are beginning to appear. The idea of building a machine whose architecture is inspired by that of the brain has roots which go far back in history. Nowadays, technological advances of computers and the availability of custom integrated circuits, permit simulations of hundreds or even thousands of neurons. In conjunction, the growing interest in learning machines, non-linear dynamics and parallel computation spurred renewed attention in artificial neural networks. Many tentative applications have been proposed, including decision systems (associative memories, classifiers, data compressors and optimizers), or parametric models for signal processing purposes (system identification, automatic control, noise canceling, etc.). While they do not always outperform standard methods, neural network approaches are already used in some real world applications for pattern recognition and signal processing tasks. The tutorial is divided into six lectures, that where presented at the Third Graduate Summer Course on Computational Physics (September 3-7, 1990) on Parallel Architectures and Applications, organized by the European Physical Society: (1) Introduction: machine learning and biological computation. (2) Adaptive artificial neurons (perceptron, ADALINE, sigmoid units, etc.): learning rules and implementations. (3) Neural network systems: architectures, learning algorithms. (4) Applications: pattern recognition, signal processing, etc. (5) Elements of learning theory: how to build networks which generalize. (6) A case study: a neural network for on-line recognition of handwritten alphanumeric characters.

Standardization and validation of a parallel form of the verbal and non-verbal recognition memory test in an Italian population sample.

PubMed

Smirni, Daniela; Smirni, Pietro; Di Martino, Giovanni; Cipolotti, Lisa; Oliveri, Massimiliano; Turriziani, Patrizia

2018-05-04

In the neuropsychological assessment of several neurological conditions, recognition memory evaluation is requested. Recognition seems to be more appropriate than recall to study verbal and non-verbal memory, because interferences of psychological and emotional disorders are less relevant in the recognition than they are in recall memory paradigms. In many neurological disorders, longitudinal repeated assessments are needed to monitor the effectiveness of rehabilitation programs or pharmacological treatments on the recovery of memory. In order to contain the practice effect in repeated neuropsychological evaluations, it is necessary the use of parallel forms of the tests. Having two parallel forms of the same test, that kept administration procedures and scoring constant, is a great advantage in both clinical practice, for the monitoring of memory disorder, and in experimental practice, to allow the repeated evaluation of memory on healthy and neurological subjects. First aim of the present study was to provide normative values in an Italian sample (n = 160) for a parallel form of a verbal and non-verbal recognition memory battery. Multiple regression analysis revealed significant effects of age and education on recognition memory performance, whereas sex did not reach a significant probability level. Inferential cutoffs have been determined and equivalent scores computed. Secondly, the study aimed to validate the equivalence of the two parallel forms of the Recognition Memory Test. The correlations analyses between the total scores of the two versions of the test and correlation between the three subtasks revealed that the two forms are parallel and the subtasks are equivalent for difficulty.

Using Multiple Big Datasets and Machine Learning to Produce a New Global Particulate Dataset: A Technology Challenge Case Study

NASA Astrophysics Data System (ADS)

Lary, D. J.

2013-12-01

A BigData case study is described where multiple datasets from several satellites, high-resolution global meteorological data, social media and in-situ observations are combined using machine learning on a distributed cluster using an automated workflow. The global particulate dataset is relevant to global public health studies and would not be possible to produce without the use of the multiple big datasets, in-situ data and machine learning.To greatly reduce the development time and enhance the functionality a high level language capable of parallel processing has been used (Matlab). A key consideration for the system is high speed access due to the large data volume, persistence of the large data volumes and a precise process time scheduling capability.

Support of Multidimensional Parallelism in the OpenMP Programming Model

NASA Technical Reports Server (NTRS)

Jin, Hao-Qiang; Jost, Gabriele

2003-01-01

OpenMP is the current standard for shared-memory programming. While providing ease of parallel programming, the OpenMP programming model also has limitations which often effect the scalability of applications. Examples for these limitations are work distribution and point-to-point synchronization among threads. We propose extensions to the OpenMP programming model which allow the user to easily distribute the work in multiple dimensions and synchronize the workflow among the threads. The proposed extensions include four new constructs and the associated runtime library. They do not require changes to the source code and can be implemented based on the existing OpenMP standard. We illustrate the concept in a prototype translator and test with benchmark codes and a cloud modeling code.

The paradigm compiler: Mapping a functional language for the connection machine

NASA Technical Reports Server (NTRS)

Dennis, Jack B.

1989-01-01

The Paradigm Compiler implements a new approach to compiling programs written in high level languages for execution on highly parallel computers. The general approach is to identify the principal data structures constructed by the program and to map these structures onto the processing elements of the target machine. The mapping is chosen to maximize performance as determined through compile time global analysis of the source program. The source language is Sisal, a functional language designed for scientific computations, and the target language is Paris, the published low level interface to the Connection Machine. The data structures considered are multidimensional arrays whose dimensions are known at compile time. Computations that build such arrays usually offer opportunities for highly parallel execution; they are data parallel. The Connection Machine is an attractive target for these computations, and the parallel for construct of the Sisal language is a convenient high level notation for data parallel algorithms. The principles and organization of the Paradigm Compiler are discussed.

PARVMEC: An Efficient, Scalable Implementation of the Variational Moments Equilibrium Code

DOE Office of Scientific and Technical Information (OSTI.GOV)

Seal, Sudip K; Hirshman, Steven Paul; Wingen, Andreas

The ability to sustain magnetically confined plasma in a state of stable equilibrium is crucial for optimal and cost-effective operations of fusion devices like tokamaks and stellarators. The Variational Moments Equilibrium Code (VMEC) is the de-facto serial application used by fusion scientists to compute magnetohydrodynamics (MHD) equilibria and study the physics of three dimensional plasmas in confined configurations. Modern fusion energy experiments have larger system scales with more interactive experimental workflows, both demanding faster analysis turnaround times on computational workloads that are stressing the capabilities of sequential VMEC. In this paper, we present PARVMEC, an efficient, parallel version of itsmore » sequential counterpart, capable of scaling to thousands of processors on distributed memory machines. PARVMEC is a non-linear code, with multiple numerical physics modules, each with its own computational complexity. A detailed speedup analysis supported by scaling results on 1,024 cores of a Cray XC30 supercomputer is presented. Depending on the mode of PARVMEC execution, speedup improvements of one to two orders of magnitude are reported. PARVMEC equips fusion scientists for the first time with a state-of-theart capability for rapid, high fidelity analyses of magnetically confined plasmas at unprecedented scales.« less

Protection of Mission-Critical Applications from Untrusted Execution Environment: Resource Efficient Replication and Migration of Virtual Machines

DTIC Science & Technology

2015-09-28

the performance of log-and- replay can degrade significantly for VMs configured with multiple virtual CPUs, since the shared memory communication...whether based on checkpoint replication or log-and- replay , existing HA ap- proaches use in- memory backups. The backup VM sits in the memory of a...efficiently. 15. SUBJECT TERMS High-availability virtual machines, live migration, memory and traffic overheads, application suspension, Java

An M-step preconditioned conjugate gradient method for parallel computation

NASA Technical Reports Server (NTRS)

Adams, L.

1983-01-01

This paper describes a preconditioned conjugate gradient method that can be effectively implemented on both vector machines and parallel arrays to solve sparse symmetric and positive definite systems of linear equations. The implementation on the CYBER 203/205 and on the Finite Element Machine is discussed and results obtained using the method on these machines are given.

Parallel design patterns for a low-power, software-defined compressed video encoder

NASA Astrophysics Data System (ADS)

Bruns, Michael W.; Hunt, Martin A.; Prasad, Durga; Gunupudi, Nageswara R.; Sonachalam, Sekar

2011-06-01

Video compression algorithms such as H.264 offer much potential for parallel processing that is not always exploited by the technology of a particular implementation. Consumer mobile encoding devices often achieve real-time performance and low power consumption through parallel processing in Application Specific Integrated Circuit (ASIC) technology, but many other applications require a software-defined encoder. High quality compression features needed for some applications such as 10-bit sample depth or 4:2:2 chroma format often go beyond the capability of a typical consumer electronics device. An application may also need to efficiently combine compression with other functions such as noise reduction, image stabilization, real time clocks, GPS data, mission/ESD/user data or software-defined radio in a low power, field upgradable implementation. Low power, software-defined encoders may be implemented using a massively parallel memory-network processor array with 100 or more cores and distributed memory. The large number of processor elements allow the silicon device to operate more efficiently than conventional DSP or CPU technology. A dataflow programming methodology may be used to express all of the encoding processes including motion compensation, transform and quantization, and entropy coding. This is a declarative programming model in which the parallelism of the compression algorithm is expressed as a hierarchical graph of tasks with message communication. Data parallel and task parallel design patterns are supported without the need for explicit global synchronization control. An example is described of an H.264 encoder developed for a commercially available, massively parallel memorynetwork processor device.

A numerical differentiation library exploiting parallel architectures

NASA Astrophysics Data System (ADS)

Voglis, C.; Hadjidoukas, P. E.; Lagaris, I. E.; Papageorgiou, D. G.

2009-08-01

We present a software library for numerically estimating first and second order partial derivatives of a function by finite differencing. Various truncation schemes are offered resulting in corresponding formulas that are accurate to order O(h), O(h), and O(h), h being the differencing step. The derivatives are calculated via forward, backward and central differences. Care has been taken that only feasible points are used in the case where bound constraints are imposed on the variables. The Hessian may be approximated either from function or from gradient values. There are three versions of the software: a sequential version, an OpenMP version for shared memory architectures and an MPI version for distributed systems (clusters). The parallel versions exploit the multiprocessing capability offered by computer clusters, as well as modern multi-core systems and due to the independent character of the derivative computation, the speedup scales almost linearly with the number of available processors/cores. Program summaryProgram title: NDL (Numerical Differentiation Library) Catalogue identifier: AEDG_v1_0 Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEDG_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 73 030 No. of bytes in distributed program, including test data, etc.: 630 876 Distribution format: tar.gz Programming language: ANSI FORTRAN-77, ANSI C, MPI, OPENMP Computer: Distributed systems (clusters), shared memory systems Operating system: Linux, Solaris Has the code been vectorised or parallelized?: Yes RAM: The library uses O(N) internal storage, N being the dimension of the problem Classification: 4.9, 4.14, 6.5 Nature of problem: The numerical estimation of derivatives at several accuracy levels is a common requirement in many computational tasks, such as optimization, solution of nonlinear systems, etc. The parallel implementation that exploits systems with multiple CPUs is very important for large scale and computationally expensive problems. Solution method: Finite differencing is used with carefully chosen step that minimizes the sum of the truncation and round-off errors. The parallel versions employ both OpenMP and MPI libraries. Restrictions: The library uses only double precision arithmetic. Unusual features: The software takes into account bound constraints, in the sense that only feasible points are used to evaluate the derivatives, and given the level of the desired accuracy, the proper formula is automatically employed. Running time: Running time depends on the function's complexity. The test run took 15 ms for the serial distribution, 0.6 s for the OpenMP and 4.2 s for the MPI parallel distribution on 2 processors.

[CMACPAR an modified parallel neuro-controller for control processes].

PubMed

Ramos, E; Surós, R

1999-01-01

CMACPAR is a Parallel Neurocontroller oriented to real time systems as for example Control Processes. Its characteristics are mainly a fast learning algorithm, a reduced number of calculations, great generalization capacity, local learning and intrinsic parallelism. This type of neurocontroller is used in real time applications required by refineries, hydroelectric centers, factories, etc. In this work we present the analysis and the parallel implementation of a modified scheme of the Cerebellar Model CMAC for the n-dimensional space projection using a mean granularity parallel neurocontroller. The proposed memory management allows for a significant memory reduction in training time and required memory size.

Cross-scale efficient tensor contractions for coupled cluster computations through multiple programming model backends

DOE Office of Scientific and Technical Information (OSTI.GOV)

Ibrahim, Khaled Z.; Epifanovsky, Evgeny; Williams, Samuel

Coupled-cluster methods provide highly accurate models of molecular structure through explicit numerical calculation of tensors representing the correlation between electrons. These calculations are dominated by a sequence of tensor contractions, motivating the development of numerical libraries for such operations. While based on matrix–matrix multiplication, these libraries are specialized to exploit symmetries in the molecular structure and in electronic interactions, and thus reduce the size of the tensor representation and the complexity of contractions. The resulting algorithms are irregular and their parallelization has been previously achieved via the use of dynamic scheduling or specialized data decompositions. We introduce our efforts tomore » extend the Libtensor framework to work in the distributed memory environment in a scalable and energy-efficient manner. We achieve up to 240× speedup compared with the optimized shared memory implementation of Libtensor. We attain scalability to hundreds of thousands of compute cores on three distributed-memory architectures (Cray XC30 and XC40, and IBM Blue Gene/Q), and on a heterogeneous GPU-CPU system (Cray XK7). As the bottlenecks shift from being compute-bound DGEMM's to communication-bound collectives as the size of the molecular system scales, we adopt two radically different parallelization approaches for handling load-imbalance, tasking and bulk synchronous models. Nevertheless, we preserve a unified interface to both programming models to maintain the productivity of computational quantum chemists.« less