008. The development of WinALT simulation system

Motivation and Aim: In recent years, simulation models of structures and algorithms with fine-grain parallelism have gained wide acceptance in describing physical, chemical, biological and social processes. Also as the technology of VLSI construction evolves, they are becoming more relevant to the design of computing devices with massive parallelism. Models with large amounts of data that take into consideration the essential features of simulated objects are interesting from both scientific and practical points of view. The participation of various specialists and the use of supercomputers are required in order to construct models of this kind. WinALT simulation system is constructed with these factors kept in mind. It is developed as a system with free access to modification of functions by a user and with an active use of the Internet both by an individual user and by a group of developers. The description of WinALT architecture, its simulation language, technology of model construction, a set of libraries that can be extended by user in order to tune WinALT to her or his problem domain are described in the report. The samples of models are presented as well.

Methods and Algorithms: An algorithmic system [1], which worked well in describing fine-grain algorithms and structures, was put into the basis of WinALT language. An open modular architecture makes it possible for a user to actively participate in the extension of its functionality by adding new modules. It is demonstrated how external modules can be used with the help the technology of component assembly, for example for the construction of such an application component as the WinALT site. A component for simulation on the basis of clusters of Windows hosts was implemented in order to improve the efficiency of simulation. Also, the requirements to its new version are formulated and its architecture and current state of its implementation is outlined. The new version provides significantly wider functionality. It supports parallelization of a bigger set of models with fine-grain parallelism (the older version supported only models of cellular automata including those with Margolus neighborhood). It can use all main kinds of contemporary parallel architectures (MPP, SMP, NOW, GPU) and some of their combinations for simulations.
Results: The stable release of WinALT simulation system is formed. The distributive package of the system is published at the WinALT site (http://winalt.sscc.ru/). The system has been tested on numerous samples, some of which are documented and published at the WinALT site.
Conclusion: The report discusses ways to improve system’s functionality in order to turn it into such an active and versatile simulation environment (from the point of view of capabilities of simulation) with Web access that includes WinALT itself and resources of the Siberian Supercomputer Center, SB RAS
Availability: Free distributive, documentation and samples as well as overviews of various issues concerning fine-grain parallelism is available at the WinALT site.
References: 1. S.M. Achasova, O.L. Bandman, V.P. Markova, S.V. Piskunov: Parallel substitution algorithm. Theory and Application. World Scientific, Singapore (1994)