The paper presents a description of par3Dhp- a 3D, parallel, fully automatic hp-adaptive finite element code for elliptic and Maxwell problems. The parallel implementation is an extension of the sequential code 3Dhp90, which generates, in a fully automatic mode, optimal hp meshes for various boundary value problems. The system constitutes an infrastructure for a class of parallel hp adaptive computations. Its modular structure allows for an independent parallelization of each component of the system. The presented work addresses parallelization of these components, including distributed data structures, load balancing and domain redistribution, parallel (multi-frontal) solver, optimal hp mesh refinements, and a main control module. All components communicate through a distributed data structure, and the control module synchronizes work of all components. The concept of ghost elements has been used to simplify the communication algorithms for parallel mesh refinements. The system has been implemented in Fortran 90 and MPI, and the load balancing is done through an interface with the ZOLTAN library. Numerical results are presented for the model Fichera problem.

Abstract. This paper reports a case study in automated composition of application families from components. The case study composes multiple instances of an h-p adaptive finite element code. An application family is represented as a structure of components. Each component is encapsulated with an interface giving a semantic specification of the properties and behavior of the component. Instances of the application family can be automatically assembled from a library of components by a compiler and the application instance can be optimized by component replacement during runtime through runtime component selection and binding. The case study demonstrates the benefits of the component composition approach to application family development and shows that execution efficiency is maintained or improved by the componentized development process.

This paper describes a method for evolutionary component-based development of families of parallel programs to attain performance goals on multiple execution environments for multiple family instances and an implementation of the method. It is based upon combining component-oriented development with integration of parallel/distributed execution and parallel/distributed simulation. Each component may have multiple representations at multiple levels of realization from analytical timing models to production code. Each component is encapsulated with an associative interface specifying its properties and behaviors which enables distinguishing among different implementations (or abstractions) of the same functional behavior which may have different performance behavior. performance model to a complete program and may continue evolution during runtime. Performance can be estimated at any stage of realization. The implementation is a compiler which composes parallel/distributed programs from components encapsulated with associative interfaces and a runtime system which supports integrated execution/simulation of parallel programs composed from components at different levels of abstraction and program evolution at runtime by component replacement. Case studies in the application of the evolutionary development method including performance results are given.

We describe a new parallel direct solver for hp refined meshes, embedded into a 3D selfadaptive hp finite-element method. The solver utilizes a sub-structuring method with multifrontal processing of sub-domain internal nodes over each sub-domain. This method of solution includes a new approach to solve the interface problem. Specifically, the solver utilizes multifrontal processing of the top of the tree of separators associated with the sub-domains on which we apply the Schur complement strategy. The relative efficiency of the solver is both analyzed theoretically and measured on a sequence of meshes generated for a 3D borehole resistivity logging problem in a deviated well. Execution time and memory usage of the solver are compared against the parallel MUMPS solver executed over the entire problem, and the MUMPS-based direct sub-structuring method with the sequential or parallel MUMPS solvers utilized to approach the interface problem. We show that the relative efficiency of the new solver tends to infinity as the polynomial orders of approximation utilized on the hp meshes approaches infinity. This result is achieved under the assumption that the domain decomposition is performed on the level of finite-element faces. Such a behavior comes as a consequence of the fact that the computational cost in the interior of high-order elements is several orders of magnitude higher then the computational cost over high-order elements faces. Based on the performed experiments, it follows that our new solver is up to five times faster than the MUMPS parallel solver when implemented on on 16 processors and if executed on computational meshes with high polynomial orders of approximation. Key words: Parallel direct solvers, Finite Element Method, hp adaptivity, 3D resistivity logging simulations.

This technical report is a preprint of a paper intended for publication in a journal or proceedings. Since changes may be made before publication, this preprint is made available with the understanding that anyone wanting to cite or reproduce it ascertains that no published version in journal or proceedings exists. This work was supported by NSF grant #DMS-0625917 Permission to copy this report is granted for electronic viewing and single-copy printing. Permissible uses are research and browsing. Specifically prohibited are sales of any copy, whether electronic or hardcopy, for any purpose. Also prohibited is copying, excerpting or extensive quoting of any report in another work without the written permission of one of the report’s authors. The University of Texas at Austin and the Texas Advanced Computing Center make no warranty, express or implied, nor assume any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed.

Abstract not found