On computing power

Abstract

We analyze in details some implementations of a challenging, yet simple application: CERN’s calorimeter. We try both general purpose computer architectures (single and multi processors, Simd and Mimd), and special purpose electronics (full-custom, gate-array, FPGA) on the problem.All measures are expressed in a single common unit for computing power: the Gbops. It applies to all forms of digital processors, and across technologies. What's more, Noyce's thesis provides a reliable way to extrapolate Gbops benchmarks through future time, say up to year 2001.The quantitative result of our analysis shows that special purpose processing is an order of magnitude more efficient than general purpose processing, on our specific problem. We show how to map the calorimeter on a programmable active memory PAM, at performance and cost comparable to those of fully dedicated implementations: orders of magnitude better than any general purpose implementation, in 1992. We argue that this current computational power advantage for PAM technology will increase with time.Finally, we discuss how to program such novel virtual PAM computers in the 2Z language, for very large synchronous designs.KeywordsComputing PowerDigital Equipment CorporationMulti ProcessorVirtual Power2adic NumberThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.