Although many existing journals publish research on parallel and distributed computing, the changes in the computational practice over the last five years warrant introducing this new journal. Parallel computing has become mainstream. The old powerhouses of high performance computing represent a slowly diminishing proportion of the market. PC vendors are the major producers of multiprocessor systems and PC software providers are developing tools to support the new hardware. The declining share of the parallel processing market serviced by traditional supercomputers and specialized parallel architectures, together with the increasing role of clusters of workstations, created conditions for the rapid spread of parallel computing in government and industry. The successful manufacturers of parallel machines leverage technology and performance gains driven by the general market (workstations and personal computers, LANs, etc.) to improve performance of their parallel products. Parallel processing has become ubiquitous at all levels of computing technology. In microprocessor design, for example, super-scalar techniques (executing multiple instructions simultaneously) are now a standard. Not only is parallel computing invading the world of everyday computing through multiprocessor desktop systems, but there is also a rapid increase in distributed computing related to the accelerating development of Intranets, the exponential growth of the Internet and to the development of distributed computing programming tools (like Java). To address the practice of parallel and distributed computing as it continues to develop is one of the aims of this journal.
As parallel computing matures, the research acquires a practical focus which starts to dominate the field. Most new algorithms are implemented on real parallel hardware or distributed across a network, and are used to solve real- life problems of realistic size. The number of professionals directly involved in parallel and/or distributed computing is increasing and their interests focus on applications and efficiency. They have to confront the fact that the complexity of parallel and/or distributed computing is one order higher than that of sequential computing. In addition, the issue of efficiency has several dimensions: algorithm design, parallel programming language constructs, user- friendly efficient parallel programming tools, optimizing and parallelizing compilers. Orienting the readers in this multidimensional space is another major aim of this journal.
The practical and experimental flavor of the journal can be partially assessed from the range of papers in its first issue. Parallel Computing with Generalized Cellular Automata by W. A. Maniatty, B. K. Szymanski and T. Caraco presents experimental results from SIMD and MIMD Systems (both tightly and loosely coupled). D. Bollman, J. Seguel and J. Feo (A Functional Approach to Radix-r FFTs) report on practical results obtained on the Cray C-90. O. Naim and A. J. G. Hey (Visualization of Do-Loop Performance Using AVS) illustrate their results with data collected on an IBM SP-2. The concluding papers (Parallel Algorithms for Multidimensional Integration by J. M. Bull and T. L. Freeman and Comparing Parallel Multidimensional Integration Algorithms by I. Gladwell and M. A. Napierala) together provide a state of the art summary and report numerical experiments on a KSR-1 and an IBM SP-2, respectively.
Aside from standard research papers, the first issue contains an overview paper: Asynchrony in Parallel Computing: From Dataflow to Multithreading by J. Silc, B. Robic and T. Ungerer. It is an extensive study (with 236 references), which provides an excellent source of information. Such overview papers play an important role in summarizing the state of knowledge in this rapidly changing field and they will be a regular feature. In addition to the standard research and overview papers, short research notes will also be published. In general, the journal is committed to fast turnaround and publication time which are necessary to match the speed with which the field develops.
The contents of the first issue represent only a part of the planned long-term coverage. In the context of our journal, parallel and distributed computing practice is viewed broadly. Many parallel algorithms have multilevel (and multidisciplinary) applications and must be viewed holistically. As examples, consider three applications. (A) Multithreaded parallel applications may be designed for non-intrusive visualization and monitoring and as a means of low-overhead steering of real-time processes. (B) Hybrid experimental-computational simulations integrate physical measurements into computer-hosted real-time simulations. At present, experiments are limited by the difficulty of extracting dense measurements in real-time, while computation is limited by inadequate physical models (e.g. of turbulence, radiation or permeability). To improve the quality of experiments and modeling, parallelism may be used for job-latency reduction rather than simply for memory expansion. (C) In many areas of scientific computing (e.g. computational fluid dynamics) it is necessary to assure seamless cooperation between multiple software components: grid generation, adaptive grid refinement, problem partitioning, domain decomposition, automatic discretization, automatic differentiation, linear and nonlinear system solution, visualization, and I/O on massive data sets. At present, these components are written independently using data structures optimized for particular query-response access patterns, making inter-operation a challenge. Research on distributed data structures to support such applications is necessary. We plan to address these and similar holistic perspectives on the practice of parallel and distributed computing. We will also look to the future by studying the limitations of current approaches to parallelism, analyzing engineering trends and their consequences, as well as presenting novel approaches to parallel computing (in hardware, middleware and software).
Providing useful information of a type not available in most journals is among our major tasks. This issue includes a number of calls for papers and/or for participation. Future issues will have a calendar of upcoming events related to parallel and distributed computing. Later in the first volume, book and software review sections will be introduced. Finally, in the upcoming issues, members of the editorial board will make personal commentaries on recent developments in parallel and distributed computing.
We wish to express our gratitude to all those who helped us start this journal. In particular, we wish to thank Katarzyna Paprzycka for the graphical design of the journal and the web-site, and Joseph Kolibal for developing the LaTeX support.