Links for Parallel Computing

[Repositories/Link Collections] * [Conferences] * [Research Groups]
[Distributed Scientific Computing] * [Distributed Arrays / Structured Grids]
[Distributed Unstructured Grids] * [General Distributed Data Structures]
[Parallel Programming Paradigms] * [Hardware, Tools and Communication Libs]

This collection is rather subjective and by no means a comprehensive compilation of research activities (which would be hopeless, anyhow). In particular, the choice is somewhat slanted towards the needs of Scientific Computing, and in particular numerical PDE solution; that is, focus is slightly more on data parallelism than on task parallelism, and in particular, in distributed grid data structures.
Descriptions of packages and projects in blockquotes are generally taken from the pages themselves.
Comments, corrections and suggestions for additions are always welcome.

Parallel Computing Repositories

ParaScope A Listing of Parallel Computing Sites maintained by David A. Bader
The Internet Parallel Computing Archive at the University of Kent at Canterbury is a good entry point.
The Parallel Tools Consortium brings together representatives from the federal, industrial, and academic sectors to address the factors that inhibit tool use and tool usability on parallel computers. Projects are, among others, parallel unix commands, debugging and performance measurement. Contains links to Parallel Tools Projects Around the World
Parallel C++ Projects Archive

Conferences

The Internet Parallel Computing Archive maintains a list of events.
Todd Veldhuizen's Object-Oriented Numerics Page has a section on conferences that may be also of interest for parallel programming.
A list of upcoming conferences of interest to researchers in the areas of compilers, parallel processing, and supercomputing, maintained by Jerry Roth at Rice University
ParCo'99 (PARALLEL COMPUTING '99, in Delft / Netherlands)
EuroPar'99 has a workshop on HPC
PPAM'99 - Third international conference on parallel processing and applied mathematics
Dynamic load-balancing for parallel mesh-based applications: 3rd DRAMA Steering Workshop

Parallel Computing Research Groups

The Berkeley Network of Workstations (NOW) project seeks to harness the power of clustered machines connected via high-speed switched networks.
PC2 Paderborn Center for Parallel Computing
Concurrent Systems Architecture Group (CSAG) at University of Illinois. Research in CSAG focuses on hardware and software architecture issues in parallel and distributed computer systems (e.g. MPP's, scalable servers, and clusters and workgroups of workstations).
Chair for Computer Science II at Aachen University of Technology. Research focus:

Design and implementation of declarative languages
modelling concurrent systems
Higher-order parallel programming

Distributed Scientific Computing Projects

PETSc PETSc, the Portable, Extensible Toolkit for Scientific Computation is a suite of data structures and routines for the scalable (parallel) solution of scientific application problems modeled by partial differential equations.
POOMA
The Parallel Object-Oriented Methods and Applications (POOMA) FrameWork effort at LANL is an application-driven software infrastructure of templated, layered class libraries designed to increase simulation lifetime, portability, and agility across rapidly evolving high-performance computing architectures. The FrameWork grew from particle simulation software efforts in the Numerical Tokamak community.
Overture
The Overture class library can be used to easily write efficient programs in C++ to solve partial differential equations in complicated domains. Overture classes enscapsulate the details involved in discretizing PDEs in curvlinear domains letting the developer work at a higher level. Operator classes, for example, define differential operators using finite-difference and finite-volume discretizations. Overlapping grids are used to represent complicated geometries. Large difficult problems can be solved using adaptive mesh refinement and moving grids. The Overture library is based on the A++/P++ array class library which provides a portable path to creating solvers for parallel machines.

Distributed Arrays and Structured Grids

The KeLP Programming System at Dept. of Computer Science and Engineering, U.C. San Diego. (Successor of LPARX)
KeLP (Kernel Lattice Parallelism) is a framework for implementing portable scientific applications on distributed memory parallel computers. It is intended for applications with special needs, in particular, that adapt to data-dependent or hardware dependent conditions at run time. KeLP is currently used in full-scale applications including subsurface modeling, turbulence studies, and first principles simulation of real materials.
A++/P++ is a parallel C++ array class library. Array operations nearly identical to FORTRAN 90 are provided via the array classes, but with all the features of C++ object-oriented design available to the user. (This seems to be a part of Overture now, I cannot find an address of A++/P++).
Distributed Dynamic Data-Management (Hierarchical Distributed Dynamic Arrays)

Distributed (Unstructured) Grids

GRIDS - a (finished) project on parallelization of grid-based applications
MG - a parallel multilevel platform for unstructured grids
COUPL+ A Parallel Library for Grid-based Applications
The DRAMA Project (DynamicRe-Allocation of Meshes for parallel Finite Element Applications). October 97 - October 99.
Objectives are:
- To develop parallel mesh re-allocation algorithms, and a corresponding subroutine library, for unstructured Finite Element codes (including industrial Structural Mechanics codes) based on the DRAMA cost model.
- To demonstrate and validate the dynamic mesh re-allocation approaches with leading industrial codes (PAM-CRASH/PAM-STAMP, FORGE3).
- To develop a cost model of unstructured finite element applications which includes: dynamically changing computational and communications requirements; re-allocation costs.
- To enable efficient 3-D parallel adaptive solvers
- To enable fully parallel mesh generation as a by-product via exploitation of the parallel re-partitioning of adaptively generated meshes.

Distributed Data Structures

Klaus Birken's DDD - Dynamic Distributed Data (is used for example in the FEM package UG )
DDD is a library for implementing parallel programs based on algorithms which use dynamic data structures on MIMD computers. It is based on a corresponding formal model of distributed data structures, which assists the application programmer in designing the distributed structure as well as in developing a correct data parallelization method.
Multipol library of distributed data structures (Graphs, Hashtables, DistObj and others).
Workshop Distributed Data and Structures, wdas98

Parallel Programming Paradigms

BSP is a programming and cost model taking into account network bandwidth considerations. BSP Worldwide is an association of people interested in the development of the Bulk Synchronous Parallel (BSP) computing model. Anther entry point to BSP is the BSP page at the Oxford Parallel Applications Centre .
Research Group on Data Parallelism (PRS).
PRS stands for Parallélisme Réseaux et Systèmes and is a french research initiative funded by the Ministère de l'Enseignement Supérieur et de la Recherche (a so-called PRC program) and the CNRS (GDR 1169). It is divided in research actions.
One of these actions federates several research teams involved in the Data Parallel programming model. The research and development activities of the teams concern Language definition, Programs design and proof, Compiling techniques and Algorithms implementations in the Data Parallel programming model framework.

Parallel Hardware, Tools and Message-Passing Libraries

Guntram Berti

Last modified: Wed Sep 20 18:23:28 MET DST 2000