The Scalability of Mesh Improvement Algorithms PDF Download

Author: Lori Freitag
Publisher:
ISBN:
Category : Differential equations, Partial
Languages : en
Pages : 23

View

Book Description
Abstract: "In this paper we develop a common framework to explore the scalability of three improvement strategies for unstructured meshes: adaptive refinement, vertex smoothing, and edge flipping. We give a general parallel algorithm for these strategies based on defining, for each algorithm, an elemental operation and a task graph. By choosing the correct task graph, we can ensure the correct parallel execution of the algorithms independent of implementation. Finally, we present experimental results obtained on an IBM SP system and use these results to investigate, in practice, the scaling and relative costs of these algorithms."

Author: Michael T. Heath
Publisher: Springer Science & Business Media
ISBN: 1461215161
Category : Mathematics
Languages : en
Pages : 373

View

Book Description
This IMA Volume in Mathematics and its Applications ALGORITHMS FOR PARALLEL PROCESSING is based on the proceedings of a workshop that was an integral part of the 1996-97 IMA program on "MATHEMATICS IN HIGH-PERFORMANCE COMPUTING. " The workshop brought together algorithm developers from theory, combinatorics, and scientific computing. The topics ranged over models, linear algebra, sorting, randomization, and graph algorithms and their analysis. We thank Michael T. Heath of University of lllinois at Urbana (Com puter Science), Abhiram Ranade of the Indian Institute of Technology (Computer Science and Engineering), and Robert S. Schreiber of Hewlett Packard Laboratories for their excellent work in organizing the workshop and editing the proceedings. We also take this opportunity to thank the National Science Founda tion (NSF) and the Army Research Office (ARO), whose financial support made the workshop possible. A vner Friedman Robert Gulliver v PREFACE The Workshop on Algorithms for Parallel Processing was held at the IMA September 16 - 20, 1996; it was the first workshop of the IMA year dedicated to the mathematics of high performance computing. The work shop organizers were Abhiram Ranade of The Indian Institute of Tech nology, Bombay, Michael Heath of the University of Illinois, and Robert Schreiber of Hewlett Packard Laboratories. Our idea was to bring together researchers who do innovative, exciting, parallel algorithms research on a wide range of topics, and by sharing insights, problems, tools, and methods to learn something of value from one another.

Author: Rainer Keller
Publisher: Springer
ISBN: 3642358934
Category : Computers
Languages : en
Pages : 156

View

Book Description
This state-of-the-art survey features topics related to the impact of multicore, manycore, and coprocessor technologies in science and large-scale applications in an interdisciplinary environment. The papers included in this survey cover research in mathematical modeling, design of parallel algorithms, aspects of microprocessor architecture, parallel programming languages, hardware-aware computing, heterogeneous platforms, manycore technologies, performance tuning, and requirements for large-scale applications. The contributions presented in this volume are an outcome of an inspiring conference conceived and organized by the editors at the University of Applied Sciences (HfT) in Stuttgart, Germany, in September 2012. The 10 revised full papers selected from 21 submissions are presented together with the twelve poster abstracts and focus on combination of new aspects of microprocessor technologies, parallel applications, numerical simulation, and software development; thus they clearly show the potential of emerging technologies in the area of multicore and manycore processors that are paving the way towards personal supercomputing and very likely towards exascale computing.

Author: Simona Perotto
Publisher: Springer
ISBN: 3319060538
Category : Mathematics
Languages : en
Pages : 326

View

Book Description
This volume collects selected contributions from the “Fourth Tetrahedron Workshop on Grid Generation for Numerical Computations”, which was held in Verbania, Italy in July 2013. The previous editions of this Workshop were hosted by the Weierstrass Institute in Berlin (2005), by INRIA Rocquencourt in Paris (2007), and by Swansea University (2010). This book covers different, though related, aspects of the field: the generation of quality grids for complex three-dimensional geometries; parallel mesh generation algorithms; mesh adaptation, including both theoretical and implementation aspects; grid generation and adaptation on surfaces – all with an interesting mix of numerical analysis, computer science and strongly application-oriented problems.

Author:
Publisher:
ISBN:
Category :
Languages : en
Pages : 2

View

Book Description

Author: Dimitris P. Vartziotis
Publisher: CRC Press
ISBN: 0429680090
Category : Computers
Languages : en
Pages : 412

View

Book Description
High quality meshes play a key role in many applications based on digital modeling and simulation. The finite element method is a paragon for such an approach and it is well known that quality meshes can significantly improve computational efficiency and solution accuracy of this method. Therefore, a lot of effort has been put in methods for improving mesh quality. These range from simple geometric approaches, like Laplacian smoothing, with a high computational efficiency but possible low resulting mesh quality, to global optimization-based methods, resulting in an excellent mesh quality at the cost of an increased computational and implementational complexity. The geometric element transformation method (GETMe) aims to fill the gap between these two approaches. It is based on geometric mesh element transformations, which iteratively transform polygonal and polyhedral elements into their regular counterparts or into elements with a prescribed shape. GETMe combines a Laplacian smoothing-like computational efficiency with a global optimization-like effectiveness. The method is straightforward to implement and its variants can also be used to improve tangled and anisotropic meshes. This book describes the mathematical theory of geometric element transformations as foundation for mesh smoothing. It gives a thorough introduction to GETMe-based mesh smoothing and its algorithms providing a framework to focus on effectively improving key mesh quality aspects. It addresses the improvement of planar, surface, volumetric, mixed, isotropic, and anisotropic meshes and addresses aspects of combining mesh smoothing with topological mesh modification. The advantages of GETMe-based mesh smoothing are demonstrated by the example of various numerical tests. These include smoothing of real world meshes from engineering applications as well as smoothing of synthetic meshes for demonstrating key aspects of GETMe-based mesh improvement. Results are compared with those of other smoothing methods in terms of runtime behavior, mesh quality, and resulting finite element solution efficiency and accuracy. Features: • Helps to improve finite element mesh quality by applying geometry-driven mesh smoothing approaches. • Supports the reader in understanding and implementing GETMe-based mesh smoothing. • Discusses aspects and properties of GETMe smoothing variants and thus provides guidance for choosing the appropriate mesh improvement algorithm. • Addresses smoothing of various mesh types: planar, surface, volumetric, isotropic, anisotropic, non-mixed, and mixed. • Provides and analyzes geometric element transformations for polygonal and polyhedral elements with regular and non-regular limits. • Includes a broad range of numerical examples and compares results with those of other smoothing methods.

Author: Michael E. Henderson
Publisher: SIAM
ISBN: 9780898714456
Category : Technology & Engineering
Languages : en
Pages : 340

View

Book Description
Contains papers presented at the October 1998 SIAM Workshop on Object Oriented Methods for Interoperable Scientific and Engineering Computing that covered a variety of topics and issues related to designing and implementing computational tools for science and engineering.

Author: Thap Panitanarak
Publisher:
ISBN:
Category :
Languages : en
Pages :

View

Book Description
Big datasets are now becoming a standard quantity in large-scale data analysis; they involve social and information network, and scientific mesh computations. These datasets are commonly stored and processed across multiple machines due to limited capabilities (such as memory and CPU) of single machines. However, many available analysis tools are still lacking in terms of an ability to fully utilize existing distributed-memory architectures. As these datasets are usually processed and analyzed in the form of graphs or meshes, we propose scalable and efficient approaches for graph and mesh computations for distributed-memory systems in this dissertation. Although graph and mesh computations are closely related regarding their parallelization approaches, some of their unique characteristics still need to be addressed separately. Thus, we organize the dissertation into two parts. The first part is for distributed graph computations, and the second part is for distributed mesh computations.In the first part of the dissertation, we focus on graph computations. First, we study a problem of Single-Source Shortest Path (SSSP) by analyzing and evaluating three well-known SSSP algorithms, i.e, Dijkstra's, Bellman-Ford, and $\Delta$-stepping algorithms. We implement these algorithms to run on distributed-memory systems based on a bulk synchronous parallel model. Their performances are evaluated and compared. Next, we propose our SSSP algorithm by combining advantages of these SSSP algorithms and utilizing a two-dimensional (2D) graph layout for our graph data structures. Then, we extend our study of the 2D graph data structures and optimization approaches to other well-known graph algorithms including breadth-first search, approximate diameter, connected components, and PageRank on various real-world graphs. Our objective is to implement an efficient graph framework for distributed-memory systems that works efficiently for many graph algorithms on various graph types. Finally, we propose graph coloring algorithms that are scalable and can be efficiently used for both graph and mesh applications.In the second part of the dissertation, we focus on parallel mesh computations on distributed-memory systems. First, we propose a domain decomposition method for 2D parallel mesh generation based on the MeTis partitioner with angle improvements. Our method is fast and gives good subdomain quality in terms of subdomain angles and mesh quality. Next, we propose a general-purpose parallel mesh warping method based on a parallel formulation of a sequential, log barrier-based mesh warping algorithm called LBWARP. Our parallel algorithm utilizes a modified distributed graph data structure with a vertex ghosting technique resulting in an efficient mesh warping algorithm which employs minimal communication. Since the algorithm needs to solve a sparse linear system with three right-hand sides (for 3D meshes), i.e., are each for the final $x$-, $y$- and $z$-coordinates in the deformed meshes, we also provide three parallel sparse linear solvers that support multiple right-hand sides for users to choose from based on the size of the problem and the number of available cores. These solvers further improve the overall performance of the algorithm, especially when a sequence of multiple deformations is required.

Author: Sudip Misra
Publisher: Springer Science & Business Media
ISBN: 1848822189
Category : Computers
Languages : en
Pages : 725

View

Book Description
Overview and Goals Wireless communication technologies are undergoing rapid advancements. The last few years have experienced a steep growth in research in the area of wireless sensor networks (WSNs). In WSNs, communication takes place with the help of spatially distributedautonomoussensornodesequippedtosensespeci?cinformation. WSNs, especially the ones that have gained much popularity in the recent years, are, ty- cally, ad hoc in nature and they inherit many characteristics/features of wireless ad hoc networks such as the ability for infrastructure-less setup, minimal or no reliance on network planning, and the ability of the nodes to self-organize and self-con?gure without the involvement of a centralized network manager, router, access point, or a switch. These features help to set up WSNs fast in situations where there is no existing network setup or in times when setting up a ?xed infrastructure network is considered infeasible, for example, in times of emergency or during relief - erations. WSNs ?nd a variety of applications in both the military and the civilian population worldwide such as in cases of enemy intrusion in the battle?eld, object tracking, habitat monitoring, patient monitoring, ?re detection, and so on. Even though sensor networks have emerged to be attractive and they hold great promises for our future, there are several challenges that need to be addressed. Some of the well-known challenges are attributed to issues relating to coverage and deployment, scalability, quality-of-service, size, computational power, energy ef?ciency, and security.

Author: Emmanuel Jeannot
Publisher: Springer Science & Business Media
ISBN: 3642233961
Category : Computers
Languages : en
Pages : 486

View

Book Description
The two-volume set LNCS 6852/6853 constitutes the refereed proceedings of the 17th International Euro-Par Conference held in Bordeaux, France, in August/September 2011. The 81 revised full papers presented were carefully reviewed and selected from 271 submissions. The papers are organized in topical sections on support tools and environments; performance prediction and evaluation; scheduling and load-balancing; high-performance architectures and compilers; parallel and distributed data management; grid, cluster and cloud computing; peer to peer computing; distributed systems and algorithms; parallel and distributed programming; parallel numerical algorithms; multicore and manycore programming; theory and algorithms for parallel computation; high performance networks and mobile ubiquitous computing.