Programming Model for Configurable Hardware

John McCaskill, Uwe Tangen, and Helmut Weberpals
Research period: 1994-1996

In biology as well as in computer science there is a dualism between data and programs, between storing and processing of information. Many attempts to understand this dualism in the context of evolution are based on models which Manfred Eigen developed in 1971. These models imply that the space in which the species interact plays an important role in creating and stabilizing information.

Simulating species distributed in space requires computing power that exceeds the capabilities of a monoprocessor and calls for parallel processing. To overcome this difficulty, John McCaskill et al. designed and built a parallel computer using configurable hardware in 1992. Following the guidelines of molecular biology we designed the following programming model. Information is coded into sequences consisting of atomic units with each atom representing a datum as well as a program. While these sequences are traversing the network like worms, their route is determined by the processors. Whenever a pair of atoms meet in a processor, there is an interaction by which the processor executes the data stored in the atoms as a program. Since this programming model is cast into configurable hardware, it achieves high parallel performance.