Parallel implementation of the finite element method using compressed data structures

Ribeiro, Fernando L. B.; Ferreira, Iuri Alves

doi:10.1007/s00466-007-0166-x

Cited by 16 publications

(19 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The code allows for the simulation of a layered construction, i.e., the geometry of the structure may change to simulate the sequential pouring of the several concrete lifts. The parallel implementation is designed for a distributed memory architecture and the data structure known as compressed storage row/columns (CSRC) is used to store the coefficient matrix and obtain iterative solutions, in a subdomain-by-subdomain approach (SBS) using the method of preconditioned conjugate gradients (PCG) [16,21].…”

Section: Principles Of the Thermo-chemo-mechanical Numerical Modelmentioning

confidence: 99%

Optimization of Mass Concrete Construction Using a Twofold Parallel Genetic Algorithm

et al. 2018

View full text Add to dashboard Cite

This paper presents a solution strategy, based on a parallel Genetic Algorithm (GA), to optimize the construction of massive concrete structures. The optimization process aims at minimizing the construction cost, considering the following design variables: the concrete mixes, the placing temperature, the height of the lifts, and the time intervals between placing the lifts. The cracking tendency is taken into account by a penalty scheme imposed to the fitness function of the GA. A thermo-chemo-mechanical model is used to calculate the transient fields of hydration, temperature, stress, strain, and cracking tendency. This model is implemented in a finite element code that is, in turn, parallelized. To demonstrate the efficiency of the proposed methodology, the simulation of the construction of a structure similar to the real thick foundation of an industrial building is presented. It shows that the optimization procedure here presented is feasible and is ready to be used in real engineering applications.

show abstract

Section: Principles Of the Thermo-chemo-mechanical Numerical Modelmentioning

confidence: 99%

Optimization of Mass Concrete Construction Using a Twofold Parallel Genetic Algorithm

et al. 2018

View full text Add to dashboard Cite

show abstract

“…The algorithms hereby presented are now part of a distributed-memory implementation of the finite element method [26]. Currently, we conduct experiments on the effect of coupling both coarse-and fine-grained parallelisms.…”

Section: Resultsmentioning

confidence: 99%

“…The compressed sparse row-column (CSRC) format is a specialization of CSR for structurally symmetric matrices arising in finite element modelling [26], which is the target domain application of this work. Given an arbitrary n × n global matrix A = (a ij ), with nnz non-zeros, the CSRC decomposes A into the sum A D + A L + A U , where A D , A L , and A U correspond to the diagonal, lower and upper parts of A, respectively.…”

Section: The Csrc Storage Formatmentioning

confidence: 99%

“…In the overlapping strategy implemented in any distributed-memory finite element code using a subdomain-by-subdomain approach [2,26], it is normal the occurrence of rectangular matrices with a remarkable property. An n × m matrix A, with m > n, can always be written as the sum A S + A R , where A S and A R are of order n × n and n × k, respectively, with k = m − n. In addition, the A S matrix has symmetric non-zero pattern, and it is occasionally numerically symmetric.…”

Section: Extension To Rectangular Matricesmentioning

confidence: 99%

“…In this case, it would be sufficient to store, roughly, half of the matrix connectivity. The compressed sparse row-column (CSRC) format was designed to take benefit from this fact [26]. Basically, it stores the column indices for only half of the offdiagonal entries.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Parallel Structurally-Symmetric Sparse Matrix-Vector Products on Multi-Core Processors

Batista¹,

Ainsworth²,

Ribeiro³

Proceedings of the Third International Conference on Parallel, Distributed, Grid and Cloud Computing for Engineering

View full text Add to dashboard Cite

We consider the problem of developing an efficient multi-threaded implementation of the matrix-vector multiplication algorithm for sparse matrices with structural symmetry. Matrices are stored using the compressed sparse row-column format (CSRC), designed for profiting from the symmetric non-zero pattern observed in global finite element matrices. Unlike classical compressed storage formats, performing the sparse matrix-vector product using the CSRC requires thread-safe access to the destination vector. To avoid race conditions, we have implemented two partitioning strategies. In the first one, each thread allocates an array for storing its contributions, which are later combined in an accumulation step. We analyze how to perform this accumulation in four different ways. The second strategy employs a coloring algorithm for grouping rows that can be concurrently processed by threads. Our results indicate that, although incurring an increase in the working set size, the former approach leads to the best performance improvements for most matrices.

show abstract

A constitutive model for mechanical and chemo‐mechanical compaction in sedimentary basins and finite element analysis

Brüch

Maghous

Ribeiro

et al. 2016

Num Anal Meth Geomechanics

View full text Add to dashboard Cite

Summary Compaction and associated fluid flow are fundamental processes in sedimentary basin deformation. Purely mechanical compaction originates mainly from pore fluid expulsion and rearrangement of solid particles during burial, while chemo‐mechanical compaction results from Intergranular Pressure‐Solution (IPS) and represents a major mechanism of deformation in sedimentary basins during diagenesis. The aim of the present contribution is to provide a comprehensive 3D framework for constitutive and numerical modeling of purely mechanical and chemo‐mechanical compaction in sedimentary basins. Extending the concepts that have been previously proposed for the modeling of purely mechanical compaction in finite poroplasticity, deformation by IPS is addressed herein by means of additional viscoplastic terms in the state equations of the porous material. The finite element model integrates the poroplastic and poroviscoplastic components of deformation at large strains. The corresponding implementation allows for numerical simulation of sediments accretion/erosion periods by progressive activation/deactivation of the gravity forces within a fictitious closed material system. Validation of the numerical approach is assessed by means of comparison with closed‐form solutions derived in the context of a simplified compaction model. The last part of the paper presents the results of numerical basin simulation performed in one dimensional setting, demonstrating the ability of the modeling to capture the main features in elastoplastic and viscoplastic compaction. Copyright © 2016 John Wiley & Sons, Ltd.

show abstract

Parallel implementation of the finite element method using compressed data structures

Cited by 16 publications

References 27 publications

Optimization of Mass Concrete Construction Using a Twofold Parallel Genetic Algorithm

Optimization of Mass Concrete Construction Using a Twofold Parallel Genetic Algorithm

Parallel Structurally-Symmetric Sparse Matrix-Vector Products on Multi-Core Processors

A constitutive model for mechanical and chemo‐mechanical compaction in sedimentary basins and finite element analysis

Contact Info

Product

Resources

About