From mgu at math.berkeley.edu Fri Apr 1 09:43:45 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Fri Apr 1 10:17:22 2011 Subject: [BANANA] LAPACK seminar on April 6, 2011 Message-ID: <201104011643.p31Ghjdg009684@phoenix.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 6, 2011 Speaker: Artem Napov, LBNL Title: An algebraic multigrid method with quaranteed convergence rate Abstract: Applications in science and engineering frequently require the solution of linear systems of equations. As the number of unknowns grows, driven, for instance, by the needs of better quality and higher accuracy of the modelling, the issues of solution time and computational resources become increasingly important. In such context, scalable iterative methods represent often an attractive (if not the only possible) approach. Among them, multigrid methods are known to exhibit optimal convergence rate in many applications; that is, to substantially (e.g., by a factor of 106) reduce the residual norm in few iterations. In this talk, we show how to design an algebraic (or black-box) multigrid method with guaranteed optimal convergence rate. In particular, we consider multigrid methods based on aggregation and prove a bound on convergence rate for systems with symmetric diagonally dominant (M-)matrices. The cornerstone of our approach is an upper bound on the two-grid convergence rate which is local and accurate. By local we mean that the convergence rate can be bounded above by computing separately for each aggregate a parameter which in some sense measures its quality. It is accurate since, assuming the aggregation pattern sufficiently regular, we show that the bound is asymptotically sharp for a large class of elliptic boundary value problems, including problems with variable and discontinuous coefficients. The two-grid estimate is then used to design an automatic aggregation procedure which builds aggregates ensuring that the quality measure is above a chosen threshold. Next, we combine it with a suitable cycling strategy that, given the known bound on the two-grid convergence rate, ensures a level-independent convergence in multilevel setting. Numerical experiments demonstrating the potentialities of such approach are presented and discussed. Date: April 13, 2011 Speaker: Oded Schwartz, UC Berkeley Title: Minimizing communication costs: new upper and lower bounds using graph expansion considerations From mgu at math.berkeley.edu Sun Apr 3 09:57:56 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Sun Apr 3 10:31:34 2011 Subject: [BANANA] Reminder: LAPACK seminar on April 6, 2011 In-Reply-To: <201104011643.p31Ghjdg009684@phoenix.math.berkeley.edu> References: <201104011643.p31Ghjdg009684@phoenix.math.berkeley.edu> Message-ID: Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 6, 2011 Speaker: Artem Napov, LBNL Title: An algebraic multigrid method with quaranteed convergence rate Date: April 13, 2011 Speaker: Oded Schwartz, UC Berkeley Title: Minimizing communication costs: new upper and lower bounds using graph expansion considerations From mgu at math.berkeley.edu Fri Apr 8 10:35:58 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Fri Apr 8 11:09:29 2011 Subject: [BANANA] LAPACK seminar on April 13, 2011 Message-ID: <201104081735.p38HZw6n002256@phoenix.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Please note that we have a change of speaker for next week. Date: April 13, 2011 Speaker: Ichitaro Yamazaki, LBNL Title: Sparse Matrix Techniques in a Parallel Hybrid Solver for Large-scale Linear Systems Abstract: A parallel hybrid (direct/iterative) linear solver based on the Schur complement method has a great potential to utilize thousands of processors for solving large-scale linear systems that are becoming increasingly difficult to solve using standard techniques. In this talk, we outline the algorithm implemented in our parallel hybrid solver, and discuss the sparse matrix techniques used for achieving high-performance. We also present numerical results of solving highly-indefinite linear systems from real applications. Date: April 20, 2011 Speaker: Aaditya (Adi) Rangan, Courant Institute From mgu at math.berkeley.edu Sun Apr 10 22:43:29 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Sun Apr 10 22:47:46 2011 Subject: [BANANA] Reminder: LAPACK seminar on April 13, 2011 In-Reply-To: <201104081735.p38HZw6n002256@phoenix.math.berkeley.edu> References: <201104081735.p38HZw6n002256@phoenix.math.berkeley.edu> Message-ID: Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 13, 2011 Speaker: Ichitaro Yamazaki, LBNL Title: Sparse Matrix Techniques in a Parallel Hybrid Solver for Large-scale Linear Systems Date: April 20, 2011 Speaker: Aaditya (Adi) Rangan, Courant Institute From saunders at stanford.edu Mon Apr 11 17:44:46 2011 From: saunders at stanford.edu (Michael Saunders) Date: Mon Apr 11 17:46:36 2011 Subject: [BANANA] LA/Opt seminar Thursday April 14 (Laurent El Ghaoui) Message-ID: Linear Algebra and Optimization Seminar (CME 510) ICME, Stanford University http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday April 14, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Professor Laurent El Ghaoui EECS and IEOR Departments Coleman Fung Risk Management Research Center UC Berkeley elghaoui@eecs.berkeley.edu http://www.eecs.berkeley.edu/~elghaoui/ Safe feature elimination for LASSO and related sparse learning problems We describe a fast method to eliminate features (variables) in l1-penalized least-square regression (lasso) problems. The elimination of features leads to a potentially substantial reduction in running time, especially for large values of the penalty parameter. Our method is not heuristic: it only eliminates features that are guaranteed to be absent after solving the lasso problem. The feature elimination step is easy to parallelize and can test each feature for elimination independently. Moreover, the computational effort of our method is negligible compared to that of solving the lasso problem -- it is roughly the same as a single gradient step. Our method extends the scope of existing lasso algorithms to treat larger data sets previously out of their reach. Forthcoming: This year's SIAM Stanford Student Chapter SCREAM seminars http://www.stanford.edu/group/siam/events.html Thu Apr 28 Parvis Moin Turbulence: Analysis of a multiscale complex system Thu May 05 Nicole Taheri Providing grid services with a fleet of plug-in electric vehicles Thu May 12 Andrew Spann How a Pixar rendering technique turned out to applicable to biological membrane flow problems Thu May 19 Aravindakshan Babu Statistical topology: Zigzags and Reeb graphs Thu May 26 Mike Lesnick Developments in the theory of multidimensional persistence From saunders at stanford.edu Thu Apr 14 14:31:31 2011 From: saunders at stanford.edu (Michael Saunders) Date: Thu Apr 14 14:33:19 2011 Subject: [BANANA] LA/Opt seminar TODAY (Laurent El Ghaoui) Message-ID: Reminder: seminar today Linear Algebra and Optimization Seminar (CME 510) ICME, Stanford University http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday April 14, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Professor Laurent El Ghaoui EECS and IEOR Departments Coleman Fung Risk Management Research Center UC Berkeley elghaoui@eecs.berkeley.edu http://www.eecs.berkeley.edu/~elghaoui/ Safe feature elimination for LASSO and related sparse learning problems We describe a fast method to eliminate features (variables) in l1-penalized least-square regression (lasso) problems. The elimination of features leads to a potentially substantial reduction in running time, especially for large values of the penalty parameter. Our method is not heuristic: it only eliminates features that are guaranteed to be absent after solving the lasso problem. The feature elimination step is easy to parallelize and can test each feature for elimination independently. Moreover, the computational effort of our method is negligible compared to that of solving the lasso problem -- it is roughly the same as a single gradient step. Our method extends the scope of existing lasso algorithms to treat larger data sets previously out of their reach. Forthcoming: This year's SIAM Stanford Student Chapter SCREAM seminars http://www.stanford.edu/group/siam/events.html Thu Apr 28 Parvis Moin Turbulence: Analysis of a multiscale complex system Thu May 05 Nicole Taheri Providing grid services with a fleet of plug-in electric vehicles Thu May 12 Andrew Spann How a Pixar rendering technique turned out to applicable to biological membrane flow problems Thu May 19 Aravindakshan Babu Statistical topology: Zigzags and Reeb graphs Thu May 26 Mike Lesnick Developments in the theory of multidimensional persistence From mgu at math.berkeley.edu Thu Apr 14 20:48:19 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Apr 14 20:49:57 2011 Subject: [BANANA] LAPACK seminar on April 20, 2011 Message-ID: <201104150348.p3F3mJx5008444@panda.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 20, 2011 Speaker: Aaditya Rangan, NYU Title: Detecting low-rank submatrices: some methods and applications Abstract: Many techniques for matrix-compression, such as matrix-skeletonization, take advantage of the reduction in dimensionality which can be attained when the matrices in question have structurally evident low-rank sub-blocks. A natural question when dealing with a matrix of unknown structure is: how can one detect the submatrices of low rank? In this talk I will describe some techniques for detecting low-rank submatrices. I will also present a few examples where these techniques can be used for matrix-compression, and as tools for data-analysis. Date: April 27, 2011 Speaker: Frederic Gibou, UC Santa Barbara From mgu at math.berkeley.edu Tue Apr 19 09:42:09 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Tue Apr 19 09:43:28 2011 Subject: [BANANA] Reminder: LAPACK seminar on April 20, 2011 In-Reply-To: <201104150348.p3F3mJx5008444@panda.math.berkeley.edu> References: <201104150348.p3F3mJx5008444@panda.math.berkeley.edu> Message-ID: Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 20, 2011 Speaker: Aaditya Rangan, NYU Title: Detecting low-rank submatrices: some methods and applications Date: April 27, 2011 Speaker: Frederic Gibou, UC Santa Barbara Title: Numerical Solvers for Nonlinear Partial Differential Equations on Octree Adaptive Grids From mgu at math.berkeley.edu Thu Apr 21 16:40:54 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Apr 21 16:54:49 2011 Subject: [BANANA] A mathematics weekend at Stanford: BASCD and ICME Open Day (fwd) Message-ID: ---------- Forwarded message ---------- Date: Mon, 18 Apr 2011 22:21:07 -0700 (PDT) From: "margot.gerritsen@stanford.edu" To: Ming Gu Subject: A mathematics weekend at Stanford: BASCD and ICME Open Day We are looking forward to May 8. The complete schedule can now be found at http://icme.stanford.edu/DeptEvents/BASC.html Please use this website to register also. We're also very pleased to invite you to ICME Open Day on May 7, 2011. I've attached the agenda. This is the day before the Bay Area Scientific Computing Day on May 8. Information about Open Day can be found also at http://icme-openday.stanford.edu and the attached flyer. We hope to welcome you to campus that weekend. Please feel free to distribute to colleagues who might be interested. Best wishes, ? Margot Gerritsen Director, Institute for Computational & Mathematical Engineering, Associate Professor, Department of Energy Resources Engineering, Stanford University Mailing address: Huang Engineering Center 475 Via Ortega Room M10 Stanford, CA 94305 Work phone: 650-725-2727 or 650-725-3542 margot.gerritsen@stanford.edu http://margot.stanford.edu, http://smartenergyshow.wordpress.com twitter: @smartenergyshow -------------- next part -------------- A non-text attachment was scrubbed... Name: ICME_OpenDay_lowres.pdf Type: application/pdf Size: 1011652 bytes Desc: Url : http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110421/3c48f171/ICME_OpenDay_lowres-0001.pdf From mgu at math.berkeley.edu Thu Apr 21 22:03:05 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Apr 21 22:04:07 2011 Subject: [BANANA] LAPACK seminar on April 27, 2011 Message-ID: <201104220503.p3M535Nr005356@phoenix.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 27, 2011 Speaker: Frederic Gibou, UC Santa Barbara Title: Numerical Solvers for Nonlinear Partial Differential Equations on Octree Adaptive Grids Abstract: Several phenomena in the physical and the life sciences can be modeled as a time dependent interface problem and nonlinear partial differential equations. Examples include the study of electro-osmotic flows, molecular beam Epitaxy, free surface flows and multiphase flows in porous media. One of the main difficulties in solving numerically these equations stems from the fact that the geometry of the problems is often arbitrary and special care is needed to correctly apply boundary conditions. Another difficulty is associated with the fact that such problems involve dissimilar length scales, with smaller scales influencing larger ones so that nontrivial pattern formation dynamics can be expected to occur at all intermediate scales. Uniform grids are limited in their ability to resolve small scales and are in such situations extremely inefficient in terms of memory storage and CPU requirements. In this talk, I will present recent advances in the numerical treatment of interface problem and describe new numerical solvers for nonlinear partial differential equations in the context of adaptive mesh refinement based on Octree grids. If time permits, I will present a second-order accurate symmetric positive definite monolithic solver for fluid/solid interactions. Date: May 4, 2011 Speaker: Takeshi Iwashita, Kyoto University, Japan Title: Hybrid parallel ordering method for a parallelized multiplicative Schwarz smoother in a multigrid solver for time-harmonic electromagnetic field problems From mgu at math.berkeley.edu Tue Apr 26 01:12:05 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Tue Apr 26 01:13:38 2011 Subject: [BANANA] Reminder: LAPACK seminar on April 27, 2011 In-Reply-To: <201104220503.p3M535Nr005356@phoenix.math.berkeley.edu> References: <201104220503.p3M535Nr005356@phoenix.math.berkeley.edu> Message-ID: Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: April 27, 2011 Speaker: Frederic Gibou, UC Santa Barbara Title: Numerical Solvers for Nonlinear Partial Differential Equations on Octree Adaptive Grids Date: May 4, 2011 Speaker: Takeshi Iwashita, Kyoto University, Japan Title: Hybrid parallel ordering method for a parallelized multiplicative Schwarz smoother in a multigrid solver for time-harmonic electromagnetic field problems From saunders at stanford.edu Tue Apr 26 18:34:03 2011 From: saunders at stanford.edu (Michael Saunders) Date: Tue Apr 26 18:35:59 2011 Subject: [BANANA] LA/Opt SCREAM seminar Thursday April 28 (Parvis Moin) Message-ID: SIAM Stanford Student Chapter SCREAM seminar and Linear Algebra and Optimization Seminar (CME 510) ICME, Stanford University http://www.stanford.edu/group/siam/events.html http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday April 28, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Refreshments served! Professor Parviz Moin Dept of Mechanical Engineering & Center for Turbulence Research Stanford University moin@stanford.edu Turbulence: Analysis of a Multi-Scale Complex System Turbulent flows are ubiquities in environment and engineering applications. Turbulence is an example of a nonlinear multi-scale phenomenon. Turbulent flows are described by the deterministic Navier-Stokes equations, but the solutions consist of both random and deterministic elements. For most applications only ensemble or time-averaged values of the flow variables are needed. However, the nonlinearity of the governing equations for the averaged quantities leads to the so-called closure problem that has plagued the research community for more than a century. In this presentation, I will discuss the nature of turbulence from a physical point of view, the closure problem, research for elucidating the structure and mechanics of turbulent flows, and modern efforts in computing multi-physics turbulent flows. Forthcoming SCREAM seminars: Thu May 05 Nicole Taheri Providing grid services with a fleet of plug-in electric vehicles Thu May 12 Andrew Spann How a Pixar rendering technique turned out to applicable to biological membrane flow problems Thu May 19 Aravindakshan Babu Statistical topology: Zigzags and Reeb graphs Thu May 26 Mike Lesnick Developments in the theory of multidimensional persistence From saunders at stanford.edu Thu Apr 28 13:52:55 2011 From: saunders at stanford.edu (Michael Saunders) Date: Thu Apr 28 13:58:53 2011 Subject: [BANANA] LA/Opt SCREAM seminar TODAY (Parvis Moin) Message-ID: Reminder: SIAM Stanford Student Chapter SCREAM seminar and Linear Algebra and Optimization Seminar (CME 510) ICME, Stanford University http://www.stanford.edu/group/siam/events.html http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday April 28, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Refreshments served! Professor Parviz Moin Dept of Mechanical Engineering & Center for Turbulence Research Stanford University moin@stanford.edu Turbulence: Analysis of a Multi-Scale Complex System Turbulent flows are ubiquities in environment and engineering applications. Turbulence is an example of a nonlinear multi-scale phenomenon. Turbulent flows are described by the deterministic Navier-Stokes equations, but the solutions consist of both random and deterministic elements. For most applications only ensemble or time-averaged values of the flow variables are needed. However, the nonlinearity of the governing equations for the averaged quantities leads to the so-called closure problem that has plagued the research community for more than a century. In this presentation, I will discuss the nature of turbulence from a physical point of view, the closure problem, research for elucidating the structure and mechanics of turbulent flows, and modern efforts in computing multi-physics turbulent flows. Forthcoming SCREAM seminars: Thu May 05 Nicole Taheri Providing grid services with a fleet of plug-in electric vehicles Thu May 12 Andrew Spann How a Pixar rendering technique turned out to be applicable to biological membrane flow problems Thu May 19 Aravindakshan Babu Statistical topology: Zigzags and Reeb graphs Thu May 26 Mike Lesnick Developments in the theory of multidimensional persistence From mgu at math.berkeley.edu Thu Apr 28 15:45:19 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Apr 28 15:47:03 2011 Subject: [BANANA] LAPACK seminar on May 2, 2011 Message-ID: <201104282245.p3SMjJqZ008577@panda.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: May 2, 2011 Speaker: Takeshi Iwashita, Academic Center for Computing and Media Studies, Kyoto University, Japan Title: Hybrid parallel ordering method for a parallelized multiplicative Schwarz smoother in a multigrid solver for time-harmonic electromagnetic field problems Abstract: This research investigates large-scale parallel time-harmonic electromagnetic field analysis based on the finite element method. The parallel geometric multigrid preconditioned iterative solver for the resulting linear system was developed on a cluster of shared memory parallel computers. We propose a hybrid parallel ordering method for the parallelization of a multiplicative Schwarz smoother, which is a key component of the multigrid solver for electromagnetic field analysis. The method, using domain decomposition ordering for multi-process parallelism and introducing block multi-color ordering for multi-thread parallel processing, attains a high convergence rate with a small number of message passing interface communications and thread synchronizations. The numerical test confirms that the proposed method attains a solver performance more than twice as good as the method based on multi-color ordering. Furthermore, an approximately 800 million degrees of freedom problem is successfully solved on 256 quad-core processors. From mgu at math.berkeley.edu Thu Apr 28 15:50:49 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Apr 28 15:52:30 2011 Subject: [BANANA] Correction: LAPACK seminar on May 2, 2011 In-Reply-To: <201104282245.p3SMjJqZ008577@panda.math.berkeley.edu> References: <201104282245.p3SMjJqZ008577@panda.math.berkeley.edu> Message-ID: Sorry, the seminar is on Wednesday, May 4, 2011. It will be the last talk for this semester. All the best, Ming On Thu, 28 Apr 2011, Ming Gu wrote: > > Math 290, Section 29, CS 298, Section 6, > Spring 2011 > (Matrix Computations and Scientific Computing) > > We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, > Berkeley campus. The coordinators are Profs. J. Demmel > (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). > The program will be a mixture of research talks and > tutorials. The tutorials will provide a partial sequel to > Math 221. > > For the schedule and other details about the seminar, please see > math.berkeley.edu/~mgu/LAPACKSeminar.htm > > Date: May 2, 2011 > Speaker: Takeshi Iwashita, Academic Center for Computing and > Media Studies, Kyoto University, Japan > Title: Hybrid parallel ordering method for a parallelized multiplicative > Schwarz smoother in a multigrid solver for time-harmonic electromagnetic > field problems > Abstract: This research investigates large-scale parallel time-harmonic > electromagnetic field analysis based on the finite element method. The > parallel geometric multigrid preconditioned iterative solver for the > resulting linear system was developed on a cluster of shared memory > parallel computers. We propose a hybrid parallel ordering method for the > parallelization of a multiplicative Schwarz smoother, which is a key > component of the multigrid solver for electromagnetic field analysis. > The method, using domain decomposition ordering for multi-process > parallelism and introducing block multi-color ordering for multi-thread > parallel processing, attains a high convergence rate with a small number > of message passing interface communications and thread synchronizations. > The numerical test confirms that the proposed method attains a solver > performance more than twice as good as the method based on multi-color > ordering. Furthermore, an approximately 800 million degrees of freedom > problem is successfully solved on 256 quad-core processors. > > -- ---------------------------------------------------------------------- Ming Gu tels: (510) 632-3145, 659-8816 Dept. of Mathematics email: mgu@math.berkeley.edu University of California URL: http://www.math.berkeley.edu/~mgu Berkeley, CA 94720-3840 ----------------------------------------------------------------------