Novel architectures for declarative languages

Technology has made it possible to create huge numbers of single-chip computers at low cost. Because declarative languages allow parallel evaluation in a natural manner, one attractive possibility for novel architects is to `buy speed? from recent technology by organising large numbers of chips to work in concert on the evaluation of a single declarative program. Following a brief introduction to the field of `declarative architectures?, some of the central issues are developed, and various novel architectures are discussed using a new classification based on the way work is distributed. The conclusion is that, even if the more extreme claims for `buying speed? from VLSI for declarative languages fail to materialise, `super von Neumann? implementations will make the new languages practicable very soon