Program Speedups in Theory and Practice

The aim of this discussion paper is to stimulate (or perhaps to provoke) stronger interactions among theoreticians and practitioners interested in eecient problem solutions. We emphasize computing abilities in relation to programming and speciication languages, as these provide the user's interface with the computer. Due to the breadth of the topic we mainly discuss eeciency as measured by computation time, ignoring space and other measures; and this on sequential, deterministic, individual, and xed computers, not treating parallelism, stochastic algorithms, distributed computing , or dynamically expanding architectures. 1. Theory versus practice? In much current Computer Science research there seems to be a quite unnecessary gap between theory and practice. Work lying on the interface between theory and practice is often subject to skepticism both from practitioner, who may doubt that theory will ever be useful for solving their complex, real-world problems; and from theoreticians, who may be displeased to see their elegant frameworks and concisely stated and powerful results made more complex and less transparent when modiied as inevitably is necessary to approach realistic problems. Some theoretical results seem likely to be forever without practical impact, but others may well eventually have very positive eeects. An analogy can be drawn with with context-free grammars, which were long studied as a purely academic discipline, but which eventually led to todays's mature and practically useful technology for parser generation 1].