Maison de la Simulation

Gif-sur-Yvette, France

Maison de la Simulation

Gif-sur-Yvette, France
SEARCH FILTERS
Time filter
Source Type

Fender A.,Nvidia | Fender A.,University of Versailles | Emad N.,Maison de la Simulation | Emad N.,University of Versailles | And 3 more authors.
Procedia Computer Science | Year: 2017

In this paper we develop a parallel approach for computing the modularity clustering often used to identify and analyse communities in social networks. We show that modularity can be approximated by looking at the largest eigenpairs of the weighted graph adjacency matrix that has been perturbed by a rank one update. Also, we generalize this formulation to identify multiple clusters at once. We develop a fast parallel implementation for it that takes advantage of the Lanczos eigenvalue solver and k-means algorithm on the GPU. Finally, we highlight the performance and quality of our approach versus existing state-of-the-art techniques. © 2017 The Authors. Published by Elsevier B.V.


Besse C.,French National Center for Scientific Research | Xing F.,Maison de la Simulation | Xing F.,University of Lille Nord de France | Xing F.,French Institute for Research in Computer Science and Automation
Journal of Scientific Computing | Year: 2017

This paper deals with two domain decomposition methods for two dimensional linear Schrödinger equation, the Schwarz waveform relaxation method and the domain decomposition in space method. After presenting the classical algorithms, we propose a new algorithm for the Schrödinger equation with constant potential and a preconditioned algorithm for the general Schrödinger equation. These algorithms are then studied numerically. The numerical experiments show that the new algorithms can improve the convergence rate and reduce the computation time. Besides of the traditional Robin transmission condition, we also propose to use a newly constructed absorbing condition as the transmission condition. © 2017 Springer Science+Business Media New York


Liu Z.,Maison de la Simulation | Liu Z.,University of Versailles | Emad N.,Maison de la Simulation | Emad N.,University of Versailles | And 2 more authors.
International Journal of Parallel Programming | Year: 2014

A parallel implementation based on implicitly restarted Arnoldi method (MIRAM) is proposed for calculating dominant eigenpair of stochastic matrices derived from very large real networks. Their high damping factor makes many existing algorithms less efficient, while MIRAM could be promising. Also, we apply this method in an epidemic application. We describe in this paper a stochastic model based on PageRank to simulate the epidemic spread, where a PageRank-like infection vector is calculated by MIRAM to help establish efficient vaccination strategy. MIRAM is implemented within the framework of Trilinos, targeting big data and sparse matrices representing scale-free networks, also known as power law networks. Hypergraph partitioning approach is employed to minimize the communication overhead. The algorithm is tested on a nation wide cluster of clusters Grid5000. Experiments on very large networks such as twitter and yahoo with over 1 billion nodes are conducted. With our parallel implementation, a speedup of (Formula presented.) is met compared to the sequential solver. © 2014 Springer Science+Business Media New York


Chen L.,Maison de la Simulation | Petition S.,Lille University of Science and Technology
Proceedings - IEEE International Conference on Cluster Computing, ICCC | Year: 2015

Krylov subspace methods (KSMs) are widely used insolving large-scale sparse linear problems. The orthogonalizationprocess in methods like GMRES would consume a majorityof the time. Since modern manycore architecture based acceleratorshave provided great horsepowers for computations,communication overheads remain a bottleneck, especially inclusters with a great number of nodes. The HA-PACS/TCA ofTsukuba University is a CPU-GPU hybrid cluster equipped withdifferent interconnects for communications among GPUs. We testa group of Krylov basis computation methods with differentsparse matrices and interconnects on HA-PACS/TCA. Resultsshow that an auto-tuning scheme is required to deal with varioustypes of matrices. © 2015 IEEE.


Fender A.,Nvidia | Fender A.,University of Versailles | Emad N.,Maison de la Simulation | Emad N.,University of Versailles | And 3 more authors.
Procedia Computer Science | Year: 2016

Data sets such as graphs are growing so rapidly that performing meaningful data analytics in reasonable time is beyond the ability of common software and hardware for many applications. In this context, performance and efficiency are primary concerns. The spectral analysis of real networks reflects such problematic. In this paper we present a solution based on Krylov methods which combines accelerators to increase the throughput of graphs traversals and latency oriented architectures to solve small problems. We focus on an hybrid acceleration of the implicitly restarted Arnoldi method which targets large non-symmetric problems with irregular sparsity pattern. The result of this cooperation is an efficient solver to compute eigenpairs of real networks. Moreover, this approach can be applied to other methods based on coarsening. © The Authors. Published by Elsevier B.V.


Ye F.,CEA Saclay Nuclear Research Center | Calvin C.,CEA Saclay Nuclear Research Center | Petiton S.G.,Maison de la Simulation | Petiton S.G.,Lille University of Science and Technology
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

The Sparse Matrix-Vector Multiplication (SpMV) is fundamental to a broad spectrum of scientific and engineering applications, such as many iterative numerical methods. The widely used Compressed Sparse Row (CSR) sparse matrix storage format was chosen to carry on this study for sustainability and reusability reasons. We parallelized for Intel Many Integrated Core (MIC) architecture a vectorized SpMV kernel using MPI and OpenMP, both pure and hybrid versions of them. In comparison to pure models and vendor-supplied BLAS libraries across different mainstream architectures (CPU, GPU), the hybrid model exhibits a substantial improvement. To further assess the behavior of hybrid model, we attribute the inadequacy of performances to vectorization rate, irregularity of non-zeros, and load balancing issue. A mathematical relationship between the first two factors and the performance is then proposed based on the experimental data. © Springer International Publishing Switzerland 2015.


Chen L.,Maison de la Simulation | Petiton S.G.,Maison de la Simulation | Petiton S.G.,Lille University of Science and Technology | Drummond L.A.,Lawrence Berkeley National Laboratory | Hugues M.,French Institute for Research in Computer Science and Automation
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2015

Krylov Subspace Methods (KSMs) are widely used for solving large-scale linear systems and eigenproblems. However, the computation of Krylov subspace bases suffers from the overhead of performing global reduction operations when computing the inner vector products in the orthogonalization steps. In this paper, a hypergraph based communication optimization scheme is applied to Arnoldi and incomplete Arnoldi methods of forming Krylov subspace basis from sparse matrix, and features of these methods are compared in a analytical way. Finally, experiments on a CPU-GPU heterogeneous cluster show that our optimization improves the Arnoldi methods implementations for a generic matrix, and a benefit of up to 10x speedup for some special diagonal structured matrix. The performance advantage also varies for different subspace sizes and matrix formats, which requires a further integration of auto-tuning strategy. © Springer International Publishing Switzerland 2015.

Loading Maison de la Simulation collaborators
Loading Maison de la Simulation collaborators