-------------------------------- ACES Special Isuue ------------------------------ ACES is the APEC (the Asia Pacific Economic Cooperation) Cooperation for Earthquake Simulation see http://quakes.earth.uq.edu.au/ACES and http://www.quakes.uq.edu.au/ACES/WG/WG5/index.html ---------------------------------------------------------------------------------- C501: Parallelization of a Large-Scale Computational Earthquake Simulation Program http://aspen.csit.fsu.edu/CCPEwebresource/C501tiampo/Parallel.pdf Abstract:Here we detail both the methods and preliminary results of first efforts to parallelize two General Earthquake Model (GEM)-related codes: 1) a relatively simple data mining procedure based on a Genetic Algorithm; and 2) the Virtual California simulation of GEM . These preliminary results, using a simple, heterogeneous processor system, existing freeware, and with an extremely low cost of both manpower and hardware dollars, motivate us to more ambitious work with considerably larger-scale computer earthquake simulations of southern California. The GEM computational problem, which is essentially a Monte Carlo simulation, is well suited to optimization on parallel computers, and we outline how we are proceeding in implementing this new software architecture. K.F. Tiampo , J.B. Rundle , S. Gross and S. McGinnis Colorado Center for Chaos & Complexity, CIRES, University of Colorado, Boulder, CO, 80309, USA Email: kristy@caldera.colorado.edu ---------------------------------------------------------------------------------- C502: Parallel Iterative Solvers for Unstructured Grids using Directive/MPI Hybrid Programming Model for GeoFEM Platform on SMP Cluster Architectures http://aspen.csit.fsu.edu/CCPEwebresource/C502nakajima2/KengoNAKAJIMA_SMP.pdf Abstract:In this study, efficient parallel iterative method for unstructured grid has been developed for SMP (shared memory symmetric multiprocessor) cluster architectures on GeoFEM platform. 3-level hybrid parallel programming model has been applied : 1. message passing for inter SMP node communication, 2. loop directive for intra SMP node parallelization and 3. vectorization for each PE (processing element). Simple 3D elastic linear problems with more than 10 8 DOFs have been solved by 3x3 block ICCG(0) with additive Schwartz domain decomposition and PDJDS/CM-RCM ((Parallel Descending-order Jagged Diagonal Storage/ Cyclic Multicolor-Reverse Cuthil Mckee)) re-ordering on 16 SMP nodes of Hitachi SR8000 and 20 GFLOPS performance has been obtained. PDJDS/CM-RCM reordering method provides excellent vector and parallel performance in SMP nodes. This re-ordering is essential for parallelization of forward/backward substitution in IC/ILU factorization with global data dependency. Developed method was also tested on NEC SX-4 and attained 969 MFLOPS (48.5% of peak performance) using single processor. Additive Schwartz domain decomposition method provided robustness to the GeoFEM's parallel iterative solvers with localized preconditioning. Kengo Nakajima and Hiroshi Okuda Department of Computational Earth Sciences, Research Organization for Information Science and Technology (RIST), Tokyo, Japan Department of Quantum Engineering and Systems Science, The University of Tokyo, Tokyo, Japan Email: nakajima@tokyo.rist.or.jp ---------------------------------------------------------------------------------- C503: Finite Element Modeling of Multibody Contact and Its Application to Active Faults http://aspen.csit.fsu.edu/CCPEwebresource/C503xing9/HuilinXing_1.pdf and http://aspen.csit.fsu.edu/CCPEwebresource/C503xing9/HuilinXing_2.html Abstract:Earthquakes have been recognized as resulting from a stick-slip frictional instability along the faults between deformable rocks. An arbitrarily shaped contact element strategy, named as node-to-point contact element strategy is proposed to handle the friction contact between deformable bodies with stick and finite frictional slip and extended to simulate the active faults in the crust with a more general nonlinear friction law. Also introduced is an efficient contact search algorithm for contact problems among multiple small and finite deformation bodies. Moreover, the efficiency of the parallel sparse solver for the nonlinear friction contact problem is investigated. Finally, a model for the plate movement in the northeast zone of Japan under gravitation is taken as an example to be analyzed with different friction behaviors. H.L. Xing and A. Makinouchi Materials Fabrication Lab, The Institute of Physical and Chemical Research (RIKEN) 2-1 Hirosawa, Wako, Saitama, 351-0198, Japan Email: xing@postman.riken.go.jp ---------------------------------------------------------------------------------- C504: Effective Adaptation Technique for Hexahedral Mesh http://aspen.csit.fsu.edu/CCPEwebresource/C504wada4/YWada_Concurrencty_and_Computation.pdf Abstract:This paper describes simple and effective adaptive techniques for the finite element method.The proposed refinement method is based on a modified octree divsion. This modified octree-based method can be applied to an arbitary hexahedral unstructured mesh. Hex-R is the the name of our implentation and it refines the mesh in cooperation with a finite element viewer: GPPView. The refined mesh retains the topology of the original coarse mesh. Therefore one can easily use the proposed method for multigrid generation. Finally the construction of an adaptive method using Hex-R is mentioned. Yoshitaka Wada Department of Computational Earth Sciences, Research Organization for Information Science and Technology (RIST), Tokyo, Japan Email: wada@tokyo.rist.or.jp ---------------------------------------------------------------------------------- C505: Thermal Convection Analysis in a Rotating Shell by a Parallel FEM - Development of a Thermal-Hydraulic Subsystem of GeoFEM http://aspen.csit.fsu.edu/CCPEwebresource/C505matsui8/H_matsui_ACES_S_ISSUE.pdf Abstract:We carry out a numerical simulation of thermally driven convection in a rotating spherical shell modeled on the Earth ’s outer core using Thermal-Hydraulic subsystem of GeoFEM,which gives a parallel FEM platform.This simulation is for understanding of the origin of the geomagnetic field and dynamics of a fluid in the Earth’s outer core.We solve a three-dimensional and time dependent process of a Boussinesq fluid in a rotating spherical shell under effects of self gravity and the Coriolis force.In this simulation,tri-linear hexahedron element is applied for spatial distribution,and 2nd-order Adams-Bashforth scheme are applied for time integration of the temperature and velocity.To satisfy the mass conservation,a parallel iterative solver of GeoFEM is applied for solving pressure and correction of the velocity fields.To verify our simulation code,results of the simulation are compared with the same analysis by the spectral method.In these results,outline of convection is almost same in these cases;that is,three pairs of convection columns are formed,and these columns propagate to the westward in quasi-steady state.However, magnitude of kinetic energy averaged over the shell is about 93%of that in the case by the spectral method,and drift frequency of the columns in the GeoFEM case is larger than that in the case by the spectral method. Hiroaki Matsui and Hiroshi Okuda Department of Research for Computational Earth Science, Research Organization for Information Science & Technology (RIST), Tokyo Department of Quantum Engineering and System Science, The University of Tokyo, Tokyo, Japan Email: matsui@tokyo.rist.or.jp ---------------------------------------------------------------------------------- C506: Performance Optimization of GeoFEM on Various Computer Architecture http://aspen.csit.fsu.edu/CCPEwebresource/C506minami4/K.minami.pdf Abstract:We have a prospect that Geofem get good parallel performance on various computer architecture by the research which has been made in GeoFEM team. In this research, we focus a target to performance with a single processor, and common data structure / coding manner to make GeoFEM running high performance on various computer architecture was researched. Test coding for Solver Part of the structure and the fluid analysis code Data structure and coding manner was evaluated on scalar/vector/pseudo vector architecture. A new data structure and a new direct access manner was introduced in the fluid analysis solver. As a result, 21% performance for peak performance was obtained on pseudo vector architecture. And We got as good performance as the pseudo-vector execution performance in scalar execution. 25% performance for peak performance was obtained on vector architecture. A new direct access manner was introduced in the structure code solver. As a result, 27% performance for peak performance was obtained on pseudo vector architecture. 34% performance for peak performance was obtained on vector architecture. Test Code for Matrix Assemble Part of the structure analysis code Coding of removing dependency was finished and performance was evaluated on vector/scalar machine. 736.8MFlops was obtained at matrix assembling process on SX- 4. 900.7MFlops was obtained at whole test Code on SX-4, and 2.06GFlops was obtained on VPP5000 (peak : 9.6GFlops). 124MFlops was obtained at matrix assembling process on Alpha system (Alpha 21164 533MHz). Kazuo Minami Department of Computational Earth Sciences, Research Organization for Information Science and Technology (RIST), Tokyo, Japan Email: minami@tokyo.rist.or.jp ---------------------------------------------------------------------------------- C507: Parallel Multilevel Iterative Linear Solvers with Unstructured Adaptive Grids for Simulations in Earth Science http://aspen.csit.fsu.edu/CCPEwebresource/C507nakajima10/KengoNAKAJIMA_MG.pdf Abstract:Preconditioned iterative solver is one of the most powerful choice such as IC (Incomplete Cholesky) or ILU (Incomplete LU) factorization method for large- scale scientific computation. But in these methods, iteration number until convergence increases as the problem size becomes larger. In multigrid solvers, the rate of convergence is independent of problem size and the number of iterations remains fairly constant. Multigrid is also a good preconditioner for Krylov iterative solvers. In this study, multigrid preconditioned conjugate gradient iterative method (MGCG) on parallel computers has been developed and applied to the Poisson equation in the region between dual sphere surfaces on semi-unstructured prismatic grids generated adaptively. Moreover this procedure has been applied to the grids with local refinement. Computational results on Hitachi SR2201 using up to 128 processors show good scalability of the developed method. Kengo Nakajima Department of Computational Earth Sciences, Research Organization for Information Science and Technology (RIST), Tokyo, Japan Email: nakajima@tokyo.rist.or.jp