Comments on Supercomputing as an Interdisciplinary Activity

Abstract

I would like to discuss supercomputing as an interdisciplinary activity. It has always been true that computational science is a field that exploits the results of many others. Our computers are built by electrical engineers, usually produced by commercial companies. We use languages, compilers, and other software produced by computer scientists. So how is the situation changing? The answer lies primarily in the greatly increased pace of change and the greatly broadened scope of the disciplines and tools involved. It is no longer simply a matter of writing a program in some well-established language, debugging it, producing some results, and then writing papers about the results. First of all, we use hardware that is evolving very rapidly. Again two of the talks before mine have emphasized how true that is: we now exploit highly parallel computing, much more powerful vector computing, and combinations of the two. As the hardware progresses, we have to use still more powerful software, software that evolves while we use it.