-
M. Faverge,
G. Pichon,
and P. Ramet.
Exploiting Parameterized Task-graph in Sparse Direct Solvers.
In SIAM Conference on Computational Science and Engineering (CSE19),
Spokane, United States,
February 2019.
Keyword(s): Low-rank compression.
Abstract:
Task-based programming models have been widely studied in the context of dense linear algebra, but remains less studied for the more complex sparse solvers. In this talk, we will present the use of two different programming models: Sequential Task Flow from StarPU, and Parameterized Task Graph from PaRSEC to parallelize the factorization step of the PaStiX sparse direct solver. We will present how those programming models have been used to integrate more complex and finer parallelism to take into account new architectures with many computational units. Efficiency of such solutions on homogeneous and heterogeneous architectures with a spectrum of matrices from different applications will be shown. We will also present how such solutions enable, without extra cost to the programmer, better performance on irregular computations such as in the block low-rank implementation of the solver. |
@inproceedings{faverge:hal-01956963,
TITLE = {{Exploiting Parameterized Task-graph in Sparse Direct Solvers}},
AUTHOR = {Faverge, M. and Pichon, G. and Ramet, P.},
URL = {https://hal.inria.fr/hal-01956963},
BOOKTITLE = {{SIAM Conference on Computational Science and Engineering (CSE19)}},
ADDRESS = {Spokane, United States},
YEAR = {2019},
MONTH = Feb,
HAL_ID = {hal-01956963},
HAL_VERSION = {v1},
KEYWORDS = {Low-rank compression},
ABSTRACT = { Task-based programming models have been widely studied in the context of dense linear algebra, but remains less studied for the more complex sparse solvers. In this talk, we will present the use of two different programming models: Sequential Task Flow from StarPU, and Parameterized Task Graph from PaRSEC to parallelize the factorization step of the PaStiX sparse direct solver. We will present how those programming models have been used to integrate more complex and finer parallelism to take into account new architectures with many computational units. Efficiency of such solutions on homogeneous and heterogeneous architectures with a spectrum of matrices from different applications will be shown. We will also present how such solutions enable, without extra cost to the programmer, better performance on irregular computations such as in the block low-rank implementation of the solver. }
}
-
Esragul Korkmaz,
Mathieu Faverge,
Grégoire Pichon,
and Pierre Ramet.
Rank Revealing QR Methods for Sparse Block Low Rank Solvers.
In Sparse Days 2019,
Toulouse, France,
July 2019.
Keyword(s): Randomized.
@inproceedings{korkmaz:hal-02326070,
TITLE = {{Rank Revealing QR Methods for Sparse Block Low Rank Solvers}},
AUTHOR = {Korkmaz, Esragul and Faverge, Mathieu and Pichon, Gr{\'e}goire and Ramet, Pierre},
URL = {https://hal.inria.fr/hal-02326070},
BOOKTITLE = {{Sparse Days 2019}},
ADDRESS = {Toulouse, France},
YEAR = {2019},
MONTH = Jul,
KEYWORDS = {Randomized},
PDF = {https://hal.inria.fr/hal-02326070/file/demo.pdf},
HAL_ID = {hal-02326070},
HAL_VERSION = {v1},
}
-
Esragul Korkmaz,
Mathieu Faverge,
Grégoire Pichon,
and Pierre Ramet.
Rank Revealing QR Methods for Sparse Block Low Rank Solvers.
In COMPAS 2019 - Conférence d'informatique en Parallélisme, Architecture et Système,
Anglet, France,
June 2019.
Keyword(s): Randomized.
@inproceedings{korkmaz:hal-02326084,
TITLE = {{Rank Revealing QR Methods for Sparse Block Low Rank Solvers}},
AUTHOR = {Korkmaz, Esragul and Faverge, Mathieu and Pichon, Gr{\'e}goire and Ramet, Pierre},
URL = {https://hal.inria.fr/hal-02326084},
BOOKTITLE = {{COMPAS 2019 - Conf{\'e}rence d'informatique en Parall{\'e}lisme, Architecture et Syst{\`e}me}},
ADDRESS = {Anglet, France},
YEAR = {2019},
MONTH = Jun,
KEYWORDS = {Randomized},
PDF = {https://hal.inria.fr/hal-02326084/file/demo.pdf},
HAL_ID = {hal-02326084},
HAL_VERSION = {v1},
}
-
G. Pichon,
E. Darve,
M. Faverge,
P. Ramet,
and J. Roman.
Block Low-rank Algebraic Clustering for Sparse Direct Solvers.
In SIAM Conference on Computational Science and Engineering (CSE19),
Spokane, United States,
February 2019.
Keyword(s): Low-rank compression.
Abstract:
In this talk, we adress the Block Low-Rank (BLR) clustering problem, to cluster unknowns within separators appearing during the factorization of sparse matrices. We show that methods considering only intra-separators connectivity (i.e., k-way or recursive bissection) as well as methods managing only interaction between separators have some limitations. The new strategy we propose consider interactions between a separator and its children to pre-select some interactions while reducing the number of off-diagonal blocks. We demonstrate how this method enhance the BLR strategies in the sparse direct supernodal solver PaStiX, and discuss how it can be extended to low-rank formats with more than one level of hierarchy. |
@inproceedings{pichon:hal-01956962,
TITLE = {{Block Low-rank Algebraic Clustering for Sparse Direct Solvers}},
AUTHOR = {Pichon, G. and Darve, E. and Faverge, M. and Ramet, P. and Roman, J.},
URL = {https://hal.inria.fr/hal-01956962},
BOOKTITLE = {{SIAM Conference on Computational Science and Engineering (CSE19)}},
ADDRESS = {Spokane, United States},
YEAR = {2019},
MONTH = Feb,
HAL_ID = {hal-01956962},
HAL_VERSION = {v1},
KEYWORDS = {Low-rank compression},
ABSTRACT = { In this talk, we adress the Block Low-Rank (BLR) clustering problem, to cluster unknowns within separators appearing during the factorization of sparse matrices. We show that methods considering only intra-separators connectivity (i.e., k-way or recursive bissection) as well as methods managing only interaction between separators have some limitations. The new strategy we propose consider interactions between a separator and its children to pre-select some interactions while reducing the number of off-diagonal blocks. We demonstrate how this method enhance the BLR strategies in the sparse direct supernodal solver PaStiX, and discuss how it can be extended to low-rank formats with more than one level of hierarchy. }
}
-
Grégoire Pichon,
E Darve,
Mathieu Faverge,
Esragul Korkmaz,
Pierre Ramet,
and Jean Roman.
Sparse supernodal solver using block low-rank compression: Design, performance and analysis.
JOREK development meeting,
November 2019.
Keyword(s): Low-rank compression.
@misc{pichon:hal-02326407,
TITLE = {{Sparse supernodal solver using block low-rank compression: Design, performance and analysis}},
AUTHOR = {Pichon, Gr{\'e}goire and Darve, E and Faverge, Mathieu and Korkmaz, Esragul and Ramet, Pierre and Roman, Jean},
URL = {https://hal.inria.fr/hal-02326407},
HOWPUBLISHED = {{JOREK development meeting}},
YEAR = {2019},
MONTH = Nov,
KEYWORDS = {Low-rank compression},
PDF = {https://hal.inria.fr/hal-02326407/file/jorek.pdf},
HAL_ID = {hal-02326407},
HAL_VERSION = {v1},
}
-
Pierre Ramet.
Utilisation de techniques de compression low-rank pour un solveur parallèle direct creux.
Journées Ondes du Sud-Ouest,
March 2019.
@misc{ramet:hal-02081108,
TITLE = {{Utilisation de techniques de compression low-rank pour un solveur parall{\`e}le direct creux}},
AUTHOR = {Ramet, Pierre},
URL = {https://hal.archives-ouvertes.fr/hal-02081108},
HOWPUBLISHED = {{Journ{\'e}es Ondes du Sud-Ouest}},
YEAR = {2019},
MONTH = Mar,
PDF = {https://hal.archives-ouvertes.fr/hal-02081108/file/joso.pdf},
HAL_ID = {hal-02081108},
HAL_VERSION = {v1},
}