Abstract. The paper is dedicated to an open T-system (OpenTS) -a programming system that supports automatic parallelization of computations for high-performance and distributed applications. In this paper, we describe the system architecture and input programming language as well as system's distinctive features. The paper focuses on the achievements of the last two years of development, including support of distributed, meta-cluster computations. Open T-System OutlineOpen T-System (Open TS) is a recent dynamic program parallelization technology for high-performance and distributed applications. It originates from functional and metaprogramming technologies [1, 2] and tries to achieve maximum performance of single/multi-processors, supercomputers, clusters and meta-clusters. Another goal was the development of easy-to-use tools for parallel programming, with high learning curve and easy legacy code support. With initial implementations of T-system dated back to nineties and end of eighties of the last century, Open TS is a third generation of the T-system [3]. The Open TS approach allows addressing in a uniform way parallel computing problem for mutli-core processors, SMP systems, computational clusters and distributed systems. As well, Open TS facilitates parallel applications with non-uniform parallelism grains or parallelism grains defined at runtime. Related WorkThe Open TS design utilizes many concepts of parallel computing. First of all, it devises high-level parallelizing approach, while many of them currently exist [4]. Secondly it utilizes an extension of C++ language to express parallelism, while many other extensions of C and C++ for parallel computing were developed [5]. Thirdly, the concept of T-system is based upon functional programming approach [1], that make it very similar to parallel implementations of functional languages [6]. At last, Open TS runtime implementation utilizes Distributed Shared Memory (DSM) [7], mutli-tier architecture [8] and C++ template-based design [9]. Here we note separately only small fraction of all works in this field, not comprehensive but representative, as we hope:

No abstract

No abstract