-
Patrick R. Amestoy,
Iain S. Duff,
Jacko Koster,
and Jean-Yves L'Excellent.
MUMPS: a general purpose distributed memory sparse solver.
In A. H. Gebremedhin,
F. Manne,
R. Moe,
and T. Sorevik, editors,
Proceedings of PARA2000, the Fifth International Workshop on Applied Parallel Computing, Bergen, June 18-21,
pages 122--131,
2000.
Springer-Verlag.
Note: Lecture Notes in Computer Science 1947.
@InProceedings{ InProceedingsAmestoy.ADKL_00,
author = {Patrick R. Amestoy and Iain S. Duff and Jacko Koster and Jean-Yves L'Excellent},
title = {{\tt MUMPS}: a general purpose distributed memory sparse solver},
booktitle = {Proceedings of PARA2000, the Fifth International Workshop on Applied Parallel Computing, Bergen, June 18-21},
editor = {A. H. Gebremedhin and F. Manne and R. Moe and T. S{\o}revik},
note = {Lecture Notes in Computer Science 1947},
publisher = {Springer-Verlag},
year = {2000},
pages = {122--131}
}
-
Patrick R. Amestoy,
Iain S. Duff,
Jean-Yves L'Excellent,
and Xiaoye S. Li.
Performance analysis of general sparse solvers for distributed memory computers.
In Seventh SIAM Conference on Applied Linear Algebra,
Raleigh, North Carolina,
October 2000.
@InProceedings{ InProceedingsAmestoy.ADLL_00,
author = {Patrick R. Amestoy and Iain S. Duff and Jean-Yves L'Excellent and Xiaoye S. Li},
title = {Performance analysis of general sparse solvers for distributed memory computers},
booktitle = {Seventh SIAM Conference on Applied Linear Algebra},
address = {Raleigh, North Carolina},
month = oct,
year = {2000}
}
-
Olivier Beaumont,
Vincent Boudet,
Arnaud Legrand,
Fabrice Rastello,
and Yves Robert.
Heterogeneity Considered Harmful to Algorithm Designers.
In Cluster'2000,
pages 403-404,
2000.
IEEE Computer Society Press.
@InProceedings{ InProceedingsBeaumont.BBLRR_00,
author = {Olivier Beaumont and Vincent Boudet and Arnaud Legrand and Fabrice Rastello and Yves Robert},
booktitle = {{Cluster'2000}},
publisher = {IEEE Computer Society Press},
title = {Heterogeneity Considered Harmful to Algorithm Designers},
pages = {403-404},
year = {2000}
}
-
Olivier Beaumont,
Vincent Boudet,
Arnaud Legrand,
Fabrice Rastello,
and Yves Robert.
Dense linear algebra kernels on heterogeneous platforms.
In Parallel Matrix Algorithms and Applications,
2000.
Université de Neuchâtel.
Note: Voir http://www.unine.ch/iiun/matrix/seminars/pmaa2000/sessions.html.
@InProceedings{ InProceedingsBeaumont.BBLRR_00*1,
author = {Olivier Beaumont and Vincent Boudet and Arnaud Legrand and Fabrice Rastello and Yves Robert},
booktitle = {{Parallel Matrix Algorithms and Applications}},
publisher = {Universit\'e de Neuchâtel},
title = {Dense linear algebra kernels on heterogeneous platforms},
year = {2000},
note = {Voir http://www.unine.ch/iiun/matrix/seminars/pmaa2000/sessions.html}
}
-
Olivier Beaumont,
Vincent Boudet,
Fabrice Rastello,
and Yves Robert.
Matrix-matrix multiplication on heterogeneous platforms.
In 2000 International Conference on Parallel Processing (ICPP'2000),
pages 289-298,
2000.
IEEE Computer Society Press.
@InProceedings{ InProceedingsBeaumont.BBRR_00*1,
author = {Olivier Beaumont and Vincent Boudet and Fabrice Rastello and Yves Robert},
booktitle = {2000 International Conference on Parallel Processing {(ICPP'2000)}},
publisher = {IEEE Computer Society Press},
title = {Matrix-matrix multiplication on heterogeneous platforms},
pages = {289-298},
year = {2000}
}
-
Olivier Beaumont,
Vincent Boudet,
Fabrice Rastello,
and Yves Robert.
Load balancing strategies for dense linear algebra kernels on heterogeneous two-dimensional grids.
In 14th International Parallel and Distributed Processing Symposium (IPDPS'2000),
pages 783-792,
2000.
IEEE Computer Society Press.
@InProceedings{ InProceedingsBeaumont.BBRR_00,
author = {Olivier Beaumont and Vincent Boudet and Fabrice Rastello and Yves Robert},
booktitle = {14th International Parallel and Distributed Processing Symposium {(IPDPS'2000)}},
publisher = {IEEE Computer Society Press},
pages = {783-792},
title = {Load balancing strategies for dense linear algebra kernels on heterogeneous two-dimensional grids},
year = {2000}
}
-
Eddy Caron,
Dominique Lazure,
and Frédéric Suter.
Manipulation de données de grande taille dans Scilab//.
In Douzièmes Rencontres Francophones du Parallélisme,
Besançon,
June 2000.
Keywords:
Scilab//.
@InProceedings{ InProceedingsCaron.CLS_00,
address = {Besan{\c c}on},
author = {Caron, Eddy and Lazure, Dominique and Suter, Fr{\'e}d{\'e}ric},
booktitle = {Douzi{\`e}mes Rencontres Francophones du Parall{\'e}lisme},
keywords = {Scilab//},
month = jun,
title = {Manipulation de donn{\'e}es de grande taille dans Scilab//},
year = 2000
}
-
Eddy Caron,
Dominique Lazure,
and Gil Utard.
Performance Modeling and Analysis of Parallel Out-of-Core Matrix Factorization.
In HiPC'2000. 7th International Conference on High Performance Computing. Bangalore, India,
December 2000.
Keywords:
Out-of-core.
@InProceedings{ InProceedingsCaron.CLU_00*2,
author = {Caron, Eddy and Lazure, Dominique and Utard, Gil},
booktitle = {{HiPC'2000}. 7th International Conference on High Performance Computing. Bangalore, India},
keywords = {Out-of-core},
month = dec,
postscript = {http://www.ens-lyon.fr/~ecaron/biblio/papers/hipc00.ps.gz},
title = {Performance Modeling and Analysis of Parallel Out-of-Core Matrix Factorization},
year = {2000}
}
-
Eddy Caron,
Dominique Lazure,
and Gil Utard.
Inversion of Huge Matrix on Cluster.
In Cluster 2000. IEEE International Conference on Cluster Computing. Technische Universität Chemnitz, Saxony, Germany. (Poster),
November 2000.
Keywords:
Out-of-core.
@InProceedings{ InProceedingsCaron.CLU_00*1,
author = {Caron, Eddy and Lazure, Dominique and Utard, Gil},
booktitle = {{Cluster 2000}. IEEE International Conference on Cluster Computing. Technische Universit{\"a}t Chemnitz, Saxony, Germany. (Poster)},
keywords = {Out-of-core},
month = nov,
title = {Inversion of Huge Matrix on Cluster},
year = 2000
}
-
Eddy Caron,
Dominique Lazure,
and Gil Utard.
Modélisation et analyse des performances de la factorisation parallel LU out-of-core.
In RenPar 2000. 12èmes Rencontres francophones du parallélisme. Besançon. France Rencontres francophones du parallélisme.,
Besançon. France,
June 2000.
Keywords:
Out-of-core.
@InProceedings{ InProceedingsCaron.CLU_00,
address = {Besan{\c c}on. France},
author = {Caron, Eddy and Lazure, Dominique and Utard, Gil},
booktitle = {RenPar 2000. 12{\`e}mes Rencontres francophones du parall{\'e}lisme. Besan{\c c}on. France Rencontres francophones du parall{\'e}lisme.},
keywords = {Out-of-core},
month = jun,
title = {Mod{\'e}lisation et analyse des performances de la factorisation parallel LU out-of-core},
year = {2000}
}
-
Henri Casanova,
Arnaud Legrand,
Dmitrii Zagorodnov,
and Francine Berman.
Heuristics for Scheduling Parameter Sweep Applications in Grid Environments.
In Heterogeneous Computing Workshop,
pages 349--363,
2000.
| Abstract: |
The Computational Grid provides a promising platform for the efficient execution of parameter sweep applications over very large parameter spaces. Scheduling such applications is challenging because target resources are heterogeneous, because their load and availability varies dynamically, and because independent tasks may share common data files. In this paper, we propose an adaptive scheduling algorithm for parameter sweep applications on the Grid. We modify standard heuristics for task/host assignment in perfectly predictable environments (Max-min, Min-min, Sufferage), and we propose an extension of Sufferage called XSufferage. Using simulation, we demonstrate that XSufferage can take advantage of file sharing to achieve better performance than the other heuristics. We also study the impact of inaccurate performance prediction on scheduling. Our study shows that: (i) different heuristics behave differently when predictions are inaccurate; (ii) increased adaptivity leads to better performance. |
@InProceedings{ InProceedingsCasanova.CLZB_00,
author = {Henri Casanova and Arnaud Legrand and Dmitrii Zagorodnov and Francine Berman},
title = {Heuristics for Scheduling Parameter Sweep Applications in Grid Environments},
booktitle = {Heterogeneous Computing Workshop},
pages = {349--363},
year = {2000},
postscript = {http://apples.ucsd.edu/pubs/hcw00_pst.ps},
abstract = {The Computational Grid provides a promising platform for the efficient execution of parameter sweep applications over very large parameter spaces. Scheduling such applications is challenging because target resources are heterogeneous, because their load and availability varies dynamically, and because independent tasks may share common data files. In this paper, we propose an adaptive scheduling algorithm for parameter sweep applications on the Grid. We modify standard heuristics for task/host assignment in perfectly predictable environments (Max-min, Min-min, Sufferage), and we propose an extension of Sufferage called XSufferage. Using simulation, we demonstrate that XSufferage can take advantage of file sharing to achieve better performance than the other heuristics. We also study the impact of inaccurate performance prediction on scheduling. Our study shows that: (i) different heuristics behave differently when predictions are inaccurate; (ii) increased adaptivity leads to better performance.}
}
-
Sylvain Contassot-Vivier,
Frédéric Lombard,
Jean-Marc Nicod,
and Laurent Philippe.
Specification of a Scilab Meta-Computing extension.
In International Workshop on Metacomputing Systems and Applications (MSA'00),
Toronto, Canada,
pages 207--214,
August 2000.
@InProceedings{ InProceedingsContassot-Vivier.CLNP_00*1,
author = {Contassot-Vivier, Sylvain and Lombard, Fr\'ed\'eric and Nicod, Jean-Marc and Philippe, Laurent},
title = {Specification of a Scilab Meta-Computing extension},
booktitle = {International Workshop on Metacomputing Systems and Applications (MSA'00)},
address = {Toronto, Canada},
pages = {207--214},
month = aug,
year = 2000
}
-
Sylvain Contassot-Vivier,
Frédéric Lombard,
Jean-Marc Nicod,
and Laurent Philippe.
Spécification de services sous CORBA pour une extension méta-computing de Scilab.
In Douzièmes rencontres Francophones du Parallélisme (RENPAR'12),
Besançon, France,
pages 33--38,
June 2000.
@InProceedings{ InProceedingsContassot-Vivier.CLNP_00,
author = {Contassot-Vivier, Sylvain and Lombard, Fr\'ed\'eric and Nicod, Jean-Marc and Philippe, Laurent},
title = {Sp\'ecification de services sous {CORBA} pour une extension m\'eta-computing de {S}cilab},
booktitle = {Douzi\`emes rencontres Francophones du Parall\'elisme (RENPAR'12)},
pages = {33--38},
address = {Besan\c{c}on, France},
month = jun,
year = 2000
}
-
Holly Dail,
Graziano Obertelli,
Fran Berman,
Rich Wolski,
and A. Grimshaw.
Application-aware scheduling of a magnetohydrodynamics application in the Legion metasystem.
In Proceedings of the 9th Heterogeneous Computing Workshop,
Cancun, Mexico,
2000.
@InProceedings{ InProceedingsDail.DOBWG_00,
author = {Holly Dail and Graziano Obertelli and Fran Berman and Rich Wolski and A. Grimshaw},
title = {Application-aware scheduling of a magnetohydrodynamics application in the Legion metasystem},
booktitle = {Proceedings of the 9th Heterogeneous Computing Workshop},
address = {Cancun, Mexico},
year = {2000}
}
-
Alain Darte,
Claude Diderich,
Marc Gengler,
and Frédéric Vivien.
Scheduling the Computations of a Loop Nest with Respect to a Given Mapping.
In Proceedings of Euro-Par 2000,
volume 1900 of LNCS,
Munich, Germany,
pages 405--414,
September 2000.
@InProceedings{ InProceedingsDarte.DDGV_00,
author = {Alain Darte and Claude Diderich and Marc Gengler and Fr\'ed\'eric Vivien},
title = {Scheduling the {C}omputations of a {L}oop {N}est with {R}espect to a {G}iven {M}apping},
booktitle = {Proceedings of Euro-Par 2000},
pages = {405--414},
year = 2000,
volume = 1900,
series = {LNCS},
address = {Munich, Germany},
month = sep
}
-
Alain Darte,
Robert Schreiber,
B. Ramakrishna Rau,
and Frédéric Vivien.
A Constructive Solution to the Juggling Problem in Processor Array Synthesis.
In Proceedings of the International Parallel and Distributed Processing Symposium (IPDPS 2000),
Cancun, Mexico,
May 2000.
@InProceedings{ InProceedingsDarte.DSRV_00,
author = {Alain Darte and Robert Schreiber and B. Ramakrishna Rau and Fr\'ed\'eric Vivien},
title = {A {C}onstructive {S}olution to the {J}uggling {P}roblem in {P}rocessor {A}rray {S}ynthesis},
booktitle = {Proceedings of the International Parallel and Distributed Processing Symposium (IPDPS 2000)},
year = 2000,
month = may,
address = {Cancun, Mexico}
}
-
F. Desprez,
E. Fleury,
E. Jeannot,
J.-M. Nicod,
and F. Suter.
Computational Servers in a Metacomputing Environment.
In Parallel Matrix Algorithms and Applications,
Neuchâtel, Switzerland,
August 2000.
SIAM.
@InProceedings{ InProceedingsDesprez.DFJNS_00,
author = "Desprez, F. and Fleury, E. and Jeannot, E. and Nicod, J.-M. and Suter, F.",
title = "{Computational Servers in a Metacomputing Environment}",
booktitle = "Parallel Matrix Algorithms and Applications",
year = 2000,
organization = "SIAM",
address = "Neuch\^atel, Switzerland",
month = aug
}
-
Benjamin Gaidioz,
Rich Wolski,
and Bernard Tourancheau.
Synchronizing Network Probes to avoid Measurement Intrusiveness with the Network Weather Service.
In Proceedings of HPDC : High Performance Distributed Computing,
Pittsburgh, USA,
pages 147--154,
August 2000.
IEEE.
@InProceedings{ InProceedingsGaidioz.GWT_00,
address = {Pittsburgh, {USA}},
author = {Gaidioz, Benjamin and Wolski, Rich and Tourancheau, Bernard},
booktitle = {Proceedings of HPDC : High Performance Distributed Computing},
month = aug,
pages = {147--154},
publisher = {IEEE},
title = {Synchronizing Network Probes to avoid Measurement Intrusiveness with the Network Weather Service},
url = {http://www.computer.org/proceedings/hpdc/0783/07830147abs.htm},
year = {2000}
}
-
Patrick Geoffray and Bernard Tourancheau.
Optimiser la latence du broadcast MPI pour les reseaux locaux haute performance.
In INRIA Press, editor,
Algotel 2000,
La Rochelle, France,
May 2000.
@InProceedings{ InProceedingsGeoffray.GT_00,
address = {La Rochelle, France},
author = {Patrick Geoffray and Bernard Tourancheau},
booktitle = {Algotel 2000},
editor = {INRIA Press},
month = may,
title = {Optimiser la latence du broadcast {MPI} pour les reseaux locaux haute performance},
year = {2000}
}
-
A. Petitet,
H. Casanova,
J. Dongarra,
Y. Robert,
and R.C. Whaley.
Parallel and Distributed Scientific Computing: A Numerical Linear Algebra Problem Solving Environment Designer's Perspective.
In J. Blazewicz,
K. Ecker,
B. Plateau,
and D.Trystram, editors,
Handbook on Parallel and Distributed Processing,
International Handbook on Information Systems, 3,
pages 464-504,
2000.
Springer Verlag.
Note: Available as LAPACK Working Note 139.
@InProceedings{ InProceedingsPetitet.PCDRW_00,
author = {A. Petitet and H. Casanova and J. Dongarra and Y. Robert and R.C. Whaley},
title = {Parallel and Distributed Scientific Computing: A Numerical Linear Algebra Problem Solving Environment Designer's Perspective},
booktitle = {Handbook on Parallel and Distributed Processing},
series = {International Handbook on Information Systems, 3},
year = 2000,
pages = {464-504},
editor = {J. Blazewicz and K. Ecker and B. Plateau and D.Trystram},
publisher = {Springer Verlag},
note = {Available as LAPACK Working Note 139.}
}
-
Bernard Tourancheau and Roland Westrelin.
Study of the medium message performance of BIP/Myrinet.
In IEEE International Conference on Cluster Computing (CLUSTER 2000),
Chemnitz, Germany,
November 2000.
IEEE Society Press.
@InProceedings{ InProceedingsTourancheau.TW_00,
address = {Chemnitz, Germany},
author = {Bernard Tourancheau and Roland Westrelin},
booktitle = {IEEE International Conference on Cluster Computing (CLUSTER 2000)},
month = nov,
publisher = {IEEE Society Press},
title = {Study of the medium message performance of {BIP/Myrinet}},
url = {http://resam.univ-lyon1.fr/~rwestrel/publi/cluster.ps.gz},
year = {2000}
}
-
Patrick R. Amestoy,
Iain S. Duff,
Jean-Yves L'Excellent,
and Xiaoye S. Li.
Analysis, Tuning and Comparison of Two General Sparse Solvers for Distributed Memory Computers.
Research report LBNL-45992,
NERSC, Lawrence Berkeley National Laboratory,
June 2000.
Note: Also France-Berkeley Project final report and ENSEEIHT-IRIT report RT/APO/00/2 and CERFACS report TR/PA/00/72, shortened version appeared in it ACM Trans. Math. Softw.
@TechReport{ TechReportRR-NERSC-LBNL-45992,
author = {Patrick R. Amestoy and Iain S. Duff and Jean-Yves L'Excellent and Xiaoye S. Li},
title = {Analysis, Tuning and Comparison of Two General Sparse Solvers for Distributed Memory Computers},
institution = {NERSC, Lawrence Berkeley National Laboratory},
number = {LBNL-45992},
note = {Also France-Berkeley Project final report and ENSEEIHT-IRIT report RT/APO/00/2 and CERFACS report TR/PA/00/72, shortened version appeared in {\it ACM Trans. Math. Softw.}},
month = jun,
type = {Research report},
year = {2000}
}
-
Olivier Beaumont,
Vincent Boudet,
Arnaud Legrand,
Fabrice Rastello,
and Yves Robert.
Heterogeneity Considered Harmful to Algorithm Designers.
Technical report 2000-24,
LIP,
June 2000.
| Abstract: |
In this paper, we deal with algorithmic issues on heterogeneous platforms. We show that static scheduling and load-balancing strategies are absolutely needed to achieve good performances, in contrast to situation for homogeneous parallel machines where dynamic schemes often turn out to be very satisfactory. However, we also show that static strategies targeted to heterogeneous platforms are difficult to design and implement: intuitively, data distribution must obey a much more refined model than standard block-cyclic distributions to equally balance the load between processors of different speeds. Technically, we state several NP-completeness results that demonstrate the intrinsic difficulty of static load-balancing on heterogeneous platforms. |
@TechReport{ TechReportRR-LIP-2000-24,
author = {Olivier Beaumont and Vincent Boudet and Arnaud Legrand and Fabrice Rastello and Yves Robert},
title = {Heterogeneity Considered Harmful to Algorithm Designers},
institution = {LIP},
year = {2000},
number = {2000-24},
month = jun,
postscript = {ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2000/RR2000-24.ps.Z},
abstract = {In this paper, we deal with algorithmic issues on heterogeneous platforms. We show that static scheduling and load-balancing strategies are absolutely needed to achieve good performances, in contrast to situation for homogeneous parallel machines where dynamic schemes often turn out to be very satisfactory. However, we also show that static strategies targeted to heterogeneous platforms are difficult to design and implement: intuitively, data distribution must obey a much more refined model than standard block-cyclic distributions to equally balance the load between processors of different speeds. Technically, we state several NP-completeness results that demonstrate the intrinsic difficulty of static load-balancing on heterogeneous platforms.}
}
-
Olivier Beaumont,
Vincent Boudet,
Fabrice Rastello,
and Yves Robert.
Heterogeneous Matrix-Matrix Multiplication or Partitioning a Square into Rectangles: NP-Completeness and Approximation Algorithms.
Technical report 2000-10,
LIP,
March 2000.
| Abstract: |
In this paper, we deal with two geometric problems arising from heterogeneous parallel computing: how to partition the unit square into p rectangles of given area s_1, s_2,..., s_p (such that the sum of the s_i is equal to 1), so as to minimize (i) either the sum of the p perimeters of the rectangles (ii) or the largest perimeter of the p rectangles. For both problems, we prove NP-completeness and we introduce approximation algorithms. |
@TechReport{ TechReportRR-LIP-2000-10,
author = {Olivier Beaumont and Vincent Boudet and Fabrice Rastello and Yves Robert},
title = {Heterogeneous Matrix-Matrix Multiplication or Partitioning a Square into Rectangles: {NP}-Completeness and Approximation Algorithms},
institution = {LIP},
year = {2000},
number = {2000-10},
month = mar,
postscript = {ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2000/RR2000-10.ps.Z},
abstract = {In this paper, we deal with two geometric problems arising from heterogeneous parallel computing: how to partition the unit square into p rectangles of given area s_1, s_2,..., s_p (such that the sum of the s_i is equal to 1), so as to minimize (i) either the sum of the p perimeters of the rectangles (ii) or the largest perimeter of the p rectangles. For both problems, we prove NP-completeness and we introduce approximation algorithms.}
}
-
Olivier Beaumont,
Arnaud Legrand,
Fabrice Rastello,
and Yves Robert.
Dense Linear Algebra Kernels on Heterogeneous Platforms: Redistribution Issues.
Technical report 2000-45,
LIP,
December 2000.
| Abstract: |
In this paper, we deal with redistribution issues for dense linear algebra kernels on heterogeneous platforms. In this context, processors speeds may well vary during the execution of a large kernel, which requires efficient strategies for redistributing the data along the computations. The strategy that we propose is to redistribute data after some well identified static phases and therefore, it is neither fully static nor fully dynamic. We present an optimal algorithm (under some assumptions) for redistributing data when performing matrix matrix multiplication. |
@TechReport{ TechReportRR-LIP-2000-45,
author = {Olivier Beaumont and Arnaud Legrand and Fabrice Rastello and Yves Robert},
title = {Dense Linear Algebra Kernels on Heterogeneous Platforms: Redistribution Issues},
institution = {LIP},
year = {2000},
number = {2000-45},
month = dec,
postscript = {ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2000/RR2000-45.ps.Z},
abstract = {In this paper, we deal with redistribution issues for dense linear algebra kernels on heterogeneous platforms. In this context, processors speeds may well vary during the execution of a large kernel, which requires efficient strategies for redistributing the data along the computations. The strategy that we propose is to redistribute data after some well identified static phases and therefore, it is neither fully static nor fully dynamic. We present an optimal algorithm (under some assumptions) for redistributing data when performing matrix matrix multiplication.}
}
-
Olivier Beaumont,
Arnaud Legrand,
Fabrice Rastello,
and Yves Robert.
Static LU Decomposition on Heterogeneous Platforms.
Technical report 2000-44,
LIP,
December 2000.
| Abstract: |
In this paper, we deal with algorithmic issues on heterogeneous platforms. We concentrate on dense linear algebra kernels, such as matrix multiplication or LU decomposition. Block cyclic distribution techniques used in ScaLAPACK are no longer sufficient to balance the load among processors running at different speeds. The main result of this paper is to provide a static data distribution scheme that leads to an asymptotically perfect load balancing for LU decomposition, thereby providing solid foundations toward the design of a cluster-oriented version of ScaLAPACK. |
@TechReport{ TechReportRR-LIP-2000-44,
author = {Olivier Beaumont and Arnaud Legrand and Fabrice Rastello and Yves Robert},
title = {Static {LU} Decomposition on Heterogeneous Platforms},
institution = {LIP},
year = {2000},
number = {2000-44},
month = dec,
postscript = {ftp://ftp.ens-lyon.fr/pub/LIP/Rapports/RR/RR2000/RR2000-44.ps.Z},
abstract = {In this paper, we deal with algorithmic issues on heterogeneous platforms. We concentrate on dense linear algebra kernels, such as matrix multiplication or LU decomposition. Block cyclic distribution techniques used in ScaLAPACK are no longer sufficient to balance the load among processors running at different speeds. The main result of this paper is to provide a static data distribution scheme that leads to an asymptotically perfect load balancing for LU decomposition, thereby providing solid foundations toward the design of a cluster-oriented version of ScaLAPACK.}
}