An overview of the Advanced CompuTational Software (ACTS) collection

Drummond, L. A.; Marques, Osni

doi:10.1145/1089014.1089016

Cited by 12 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

Section: Description Of the Problemmentioning

confidence: 99%

“…Here we apply direct and iterative strategies that are available in open source software packages to investigate the spectral properties of the problem. Specifically, we use packages available in the ACTS Collection of the US Department of Energy (DOE) [4].Basically, we seek to solveis the first exponential-integral function. An integral equation of this type can be solved by a discretization mechanism, for instance by projecting it into a finite dimensional subspace.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Comparison of parallel eigensolvers for a discretized radiative transfer problem

Vasconcelos

Marques

2007

Proc Appl Math and Mech

Self Cite

View full text Add to dashboard Cite

In this work we consider the numerical computation of eigenpairs of integral operators associated to a radiative transfer problem. The target problem models the emission of photons in stellar atmospheres and is formulated in terms of a weakly singular Fredholm integral equation defined on a Banach space. We examine parallel, direct and iterative strategies for the eigensolution phase. Description of the problemIn this contribution we focus on the numerical computation of eigenpairs of integral operators associated to a model for the emission of photons in stellar atmospheres; see [1][2] for details of the problem. In a previous work we have investigated numerical techniques for the solution of the associated radiative transfer equation [3], considering parallel distributed processing. Here we apply direct and iterative strategies that are available in open source software packages to investigate the spectral properties of the problem. Specifically, we use packages available in the ACTS Collection of the US Department of Energy (DOE) [4].Basically, we seek to solve T ϕ = λϕ where T is an integral operator defined inwhere τ is the optical depth of the stellar atmosphere, τ * is the optical thickness of the atmosphere, ∈ [0, 1] is the albedo, and Solution strategyOur goal is to approximate T m ϕ m = λ m ϕ m by solving an associated eigenvalue problem Ax = λx for large values of m, where A is nonsymmetric, band and sparse. In order to achieve this, we parallelized the generation of A (using MPI) and experimented with two numerical strategies for the eigensolution phase: direct methods implemented in the ScaLAPACK library to compute the Schur decomposition of A, and iterative methods implemented in the SLEPc library to compute a few eigenpairs of A. In ScaLAPACK we first need to reduce A to Hessenberg form (subroutine pdgehrd) and then compute the Schur decomposition of the Hessenberg matrix (subroutine pdlahqr). Concerning SLEPc, we used a recently implemented version of the Krylov-Schur algorithm, which has proven to be faster and more accurate than the Arnoldi algorithm in a number of cases. Our tests were performed on three machines: Jacquard and Bassi, located at the National Energy Research Center (NERSC) of the US DOE, and GridUP, located at the University of Porto. Jacquard is an AMD Opteron cluster with 356 dual-processor nodes, 2.2 GHz processors, 6 GB of memory per node, interconnected with a high-speed InfiniBand network. Bassi is an IBM p575 POWER 5 system with 122 8-processor nodes and 32 GB of memory per node. GridUP is an AMD Opteron Cluster with 24 dual-processor nodes, 2.4 GHz processors, 4 GB of memory per node, interconnected with Gigabit network. On Jacquard and on GridUP we used the basic linear algebra subroutines (BLAS) available in the ACML library, and on Bassi we used the BLAS available in the ESSL library. For all cases showed here we used = 0.75 (tests with larger values of required similar computing times). We used a relative error ε ≤ 10 −12 for the iterative method in order to obtain ...

show abstract

Section: Description Of the Problemmentioning

confidence: 99%

mentioning

confidence: 99%

Comparison of parallel eigensolvers for a discretized radiative transfer problem

Vasconcelos

Marques

2007

Proc Appl Math and Mech

Self Cite

View full text Add to dashboard Cite

show abstract

“…We promote high-quality and general purpose software tools that provide a plethora of computational services to the growing computational sciences community. The Advanced CompuTational Software (ACTS) Collection [1,2,3] is a set of computational tools developed primarily at DOE laboratories and is aimed at simplifying the solution of common and important computational problems. The use of the tools reduces the development time for new codes and the tools provide functionality that might not otherwise be available.…”

mentioning

confidence: 99%

High Performance Computing for Computational Science - VECPAR 2006

Daydé¹,

Palma²,

Coutinho³

et al. 2007

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Software reusability has proven to be an effective practice to speed-up the development of complex high-performance scientific and engineering applications. We promote the reuse of high quality software and general purpose libraries through the Advance CompuTational Software (ACTS) Collection. ACTS tools have continued to provide solutions to many of today's computational problems. In addition, ACTS tools have been successfully ported to a variety of computer platforms; therefore tremendously facilitating the porting of applications that rely on ACTS functionalities. In this contribution we discuss a high-level user interface that provides a faster code prototype and user familiarization with ACTS tools. The high-level user interfaces have been built using Python. Here we focus on Python based interfaces to ScaLAPACK, the PyScaLAPACK component of PyACTS. We briefly introduce their use, functionalities, and benefits. We illustrate a few simple example of their use, as well as exemplar utilization inside large scientific applications. We also comment on existing Python interfaces to other ACTS tools. We present some comparative performance results of PyACTS based versus direct LAPACK and ScaLAPACK code implementations.

show abstract

“…The tools were developed in various laboratories and universities and have allowed a wide spectrum of important computational problems to be solved to content [10]. We refer the reader to [9] for an overview of the project and available numerical tools, and also to the ACTS Information Center [12] for details about all tools available in the Collection. In this paper we are interested in solving equation (1) on a fine mesh.…”

Section: Acts: Tools Of the Tradementioning

confidence: 99%

“…In these lines, we are interested in investigating the trade offs and capabilities implemented in several packages, in particular the ones that are available in the DOE Advanced CompuTational Software (ACTS) Collection [9]. In the following sections we outline the projection and matrix formulation that we use to tackle the integral operator.…”

Section: Introduction and Problem Overviewmentioning

confidence: 99%

Evaluation of Linear Solvers for Astrophysics Transfer Problems

Marques

Vasconcelos

Lecture Notes in Computer Science

View full text Add to dashboard Cite

An overview of the Advanced CompuTational Software (ACTS) collection

Cited by 12 publications

References 22 publications

Comparison of parallel eigensolvers for a discretized radiative transfer problem

Comparison of parallel eigensolvers for a discretized radiative transfer problem

High Performance Computing for Computational Science - VECPAR 2006

Evaluation of Linear Solvers for Astrophysics Transfer Problems

Contact Info

Product

Resources

About