A new era in scientific computing: Domain decomposition methods in hybrid CPU–GPU architectures

Papadrakakis, Manolis; Stavroulakis, George; Karatarakis, Alexander

doi:10.1016/j.cma.2011.01.013

Cited by 101 publications

(50 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A solution to this problem can be found in high-performance and cloud computing. Although high-performance computing provides new and interesting opportunities to solve large-scale structural engineering problems (Papadrakakis et al, 2011;Hori et al, 2018), the development of new computational models and algorithms that exploit the unique architecture of these machines still remains a challenge (Adeli and Soegiarso, 1998). Cloud Computing is a model for enabling convenient, ondemand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort (Mell and Grance, 2011).…”

Section: High-performance and Cloud Computingmentioning

confidence: 99%

Computational Structural Engineering: Past Achievements and Future Challenges

Plevris

Tsiatas

2018

Front. Built Environ.

View full text Add to dashboard Cite

PAST ACHIEVEMENTS AND CURRENT TRENDSComputational methods are computer-based methods used to numerically solve mathematical models that describe physical phenomena. The purpose of computational modeling is to study the behavior of complex systems by means of computer simulations and it can be used to make predictions of the system's behavior under different conditions, often for cases in which intuitive analytical solutions are not available (Nature, 2018). Computational methods have emerged in engineering during the 1960s. Since then, structural engineers have been leaders in technological solutions to engineering analysis and design problems. The evolution of electronic computers together with the tremendous increase of computational power has triggered the continuous development of computational methods. Rapid advances in computer hardware have had a profound effect on various engineering disciplines. The applications are numerous and cover a broad field of engineering branches including civil, mechanical, naval, electrical, aerospace, material, biomolecular, among others. Computational structural engineering has evolved as an insightful blend combining both structural analysis and computer science.Among all computational methods, the Finite Element Method (FEM) and the Boundary Element Method (BEM) are the most prevalent ones. Both methods exhibit unique characteristics as well as advantages and disadvantages. The FEM, as a simulation tool, enables sophisticated methods of computational mechanics, computer technology and applied mathematics. It has been broadly adopted in scientific research and engineering applications and it can be considered as the most popular method used for structural analysis, tackling linear and nonlinear problems of systems with various geometries, material properties and loads (Reddy, 2005). In FEM, the solution domain is subdivided into a finite number of elements. Each element approximately reproduces the behavior of a small region of the body it represents, but continuity between the elements is only enforced in an overall minimum energy sense. The FEM is especially suitable for problems with complex geometries, but the whole-body discretization scheme that is utilized inevitably leads to a large number of finite elements and, thus, increased computational cost. Although Professor R. Clough coined the term "Finite Element Method" in his pioneer work dated back in 1960 (Clough, 1960), the answer to the question "who invented FEM in everyday use?" is possibly M. Jonathan (Jon) Turner at Boeing who generalized and perfected the Direct Stiffness Method, and convinced his company to commit resources to it, over the period 1952 -1964(Felippa, 2017. The FEM obtained its real impetus in the 1960s and 1970s by the developments of pioneers J. H. Argyris, R. W. Clough, H. C. Martin, O. C. Zienkiewicz and their co-workers who evolved the method and applied it to a wide range of structural problems and applications.In the years since its first use, FEM has grown and developed into a standard...

show abstract

Section: High-performance and Cloud Computingmentioning

confidence: 99%

Computational Structural Engineering: Past Achievements and Future Challenges

Plevris

Tsiatas

2018

Front. Built Environ.

View full text Add to dashboard Cite

show abstract

“…Linear algebra applications have also been a topic of scientific interest for GPU implementations [17]- [20]. A hybrid CPU-GPU implementation of domain decomposition methods is presented in [21] where speedups of the order of 40x have been achieved with just one GPU.…”

Section: Introductionmentioning

confidence: 99%

Computation of the Isogeometric Analysis Stiffness Matrix on Gpu

Karatarakis¹,

Karakitsios²,

Papadrakakis³

2013

Proceedings of the 3rd South-East European Conference on Computational Mechanics – SEECCM III

View full text Add to dashboard Cite

show abstract

“…There are also excellent results in molecular dynamics simulations [5], in contact detection for DEM (Discret Element Method) simulations [6] and in boundary element applications [7]. As examples of recent contributions in the FEM context, one can cite the work in [8], concerning the implementation of a domain-decomposition method. With the increasing power of GPUs for general purpose programming, multinode, multi-core clusters with one (or more) GPU per node have seen a rapid diffusion.…”

Section: Introductionmentioning

confidence: 99%

“…[2,9,14,10]). It is important to recall, as suggested in [8], that, concerning NVIDIA GPUs, in all the implementations before CUDA compute capability 1.3, only single-precision floating point operations are supported directly by the hardware. Furthermore, double-precision variables consume twice as much of the limited amount of GPU registers and block shared memory, and in view of the possible decay of performances implied by the use of device global memory, this has often been considered as a strong indication towards the use of single-precision arithmetic.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An explicit dynamics GPU structural solver for thin shell finite elements

Bartezzaghi

Cremonesi

Parolini

et al. 2015

Computers & Structures

View full text Add to dashboard Cite

With the availability of user oriented software tools, dedicated architectures, such as the parallel computing platform and programming model CUDA (Compute Unified Device Architecture) released by NVIDIA, one of the main producers of graphics cards, and of improved, highly performing GPU (Graphics Processing Unit) boards, GPGPU (General Purpose programming on GPU) is attracting increasing interest in the engineering community, for the development of analysis tools suitable to be used in validation/verification and virtual reality applications. For their inherent explicit and decoupled structure, explicit dynamics finite element formulations appear to be particularly attractive for implementations on hybrid CPU/GPU or pure GPU architectures. The issue of an optimized, double-precision finite element GPU implementation of an explicit dynamics finite element solver for elastic shell problems in small strains and large displacements and rotations, using unstructured meshes, is here addressed. The conceptual difference between a GPU implementation directly adapted from a standard CPU approach and a new optimized formulation, specifically conceived for GPUs, is discussed and comparatively assessed. It is shown that a speedup factor of about 5 can be achieved by an optimized algorithm reformulation and careful memory management. A speedup of more than 40 is achieved with respect of state-of-the art commercial codes running on CPU, obtaining real-time simulations in some cases, on commodity hardware. When a last generation GPU board is used, it is shown that a problem with more than 16 millions degrees of freedom can be solved in just few hours of computing time, opening the way to virtualization approaches for real large scale engineering problems.

show abstract

A new era in scientific computing: Domain decomposition methods in hybrid CPU–GPU architectures

Cited by 101 publications

References 35 publications

Computational Structural Engineering: Past Achievements and Future Challenges

Computational Structural Engineering: Past Achievements and Future Challenges

Computation of the Isogeometric Analysis Stiffness Matrix on Gpu

An explicit dynamics GPU structural solver for thin shell finite elements

Contact Info

Product

Resources

About