hhd0@gte.com (Horace Dediu) (10/24/88)
Greetings! I am interested in developing a flexible simulator for multiprocessor architectures. That is, developing a formal description language for multi-functional-unit machines and using the description to simulate the architecture to determine performance. The description would include the interconnection network (a graph) and the processor description (in ISPS notation [Barbacci '80]). I know simulators exist for particular machines, but I want to build a flexible simulator, in which the architecture is a free variable. Note that this means simulation *of* multiprocessors on uniprocessors, not simulation *on* multiprocessors. I would like to know what others have done in this area. Preliminary searches have not yielded much. If anyone knows about research in this area, please mail or post. I promise a summary. I thank you all, Horace Dediu hhd0@gte.com
zenith@uunet.UU.NET (Steven Zenith) (10/27/88)
Greetings! Horace Dediu writes: > I am interested in developing a flexible simulator for multiprocessor > architectures. That is, developing a formal description language for > multi-functional-unit machines and using the description to simulate the > architecture to determine performance. The description would include the > interconnection network (a graph) and the processor description (in ISPS > notation [Barbacci '80]). [...] Maybe I don't understand this question properly (being new to this newsgroup). CSP and Occam are both languages able to provide a formal description of "multi-functional-unit" machines. They do both assume that the system is founded on a Communicating Process *Architecture*. However, a "simulation" of any *configuration* of such units may be written in Occam and performance evaluated on a single transputer. > [...] I want to build a flexible simulator, in which the architecture is a > free variable. Again, I may be misunderstanding, do you mean 'architecture' or 'configuration'. Do you mean you are looking for a generalised notation of some kind? * Steven Ericsson Zenith | zenith@inmos.uucp (INMOS Bristol, England) ** The tao that can be told is not the eternal tao **