The TOP-C parallel model and symbolic algebra

Abstract

The TOP-C parallel model was specifically designed to be optimal for many symbolic algebra algorithms. Symbolic algebra problems are irregular and differ from typical problems of numerical analysis in two respects.1. They have large amounts of intermediate swell.2. They are usually not able to make good use of data parallelism.The TOP-C model was developed on the basis of experience over ten years. The model was first implemented for distributed memory for GCL (GNU Common LISP), and is currently available as ParGCL [1]. That work was extended to ParGAP [3], a binding to the interactive language GAP (Groups, Algorithms and Programming). Finally, a C/C++ version, Task Oriented Parallel C/C++, lent its acronym to the TOP-C model [2].A sign of the maturity of TOP-C is the parallelization of Geant4 (Geometry and Tracking) [4]. Geant4 is a million-line program for the simulation of particle-matter interaction used in the design of collider experiments at CERN. Current issues facing parallelization include the slowness of RAM [5] and the need for semi-automated marshalling (serialization of objects) [6].