A high performance interconnection concept for dataflow- or other closely coupled multiprocessors
The interconnection network is a crucial part in many existing or proposed multiprocessor systems. Generalized cube- or crossbar-type networks featuring packet-switching [Sieg85] are often proposed if the bandwith demand exceeds the limited capacity of common-bus solutions.
The paper outlines an upgraded version of the intercommunication memory Intercom being successfully implemented in the ETH- Multiprocessor Empress [Bueh82]. This n-way multiport memory type concept is well suited to realize a high bandwith packet-switched network at reasonable cost despite the fact that its complexity is of the order n * n (n equals the number of connected processors) rather than n * log n as in a generalized cube network. Key properties are particularily congestion- (i.e. collision) elimination and minimal network latency. The Intercom technique avoids, therefore, the potential network bottleneck in many multiprocessor concepts as long as the number of connected processors is not exceeding an upper limit.
KeywordsData Packet Interconnection Network Multiprocessor System Memory Block Target Processor
Unable to display preview. Download preview PDF.
- [Ager82]Agerwala T. and Arvind, "Data Flow Systems. Guest Editors' Introduction", IEEE Computer, Vol. 15, No. 2, Feb. 1982, pp. 10–13. [Arvi83] Arvind and members of the Functional Languages and Architectures Group, "The Tagged Token Dataflow Architecture", Memo, Computation Structures Group, Laboratory for Computer Science, MIT, Cambridge, Mass., March 1983.Google Scholar
- [Bueh82]Buehrer R.E., Brundiers H.J., Benz H., Bron B., Friess H.M., Haelg W., Halin H.J., Isacson A., and Tadjan M., "The ETH-Multiprocessor Empress: A Dynamically Configurable MIMD System", IEEE Transactions on Computers, Vol. 31, No. 11, Nov. 1982, pp. 1035–1044.Google Scholar
- [Denn80]Dennis J.B., "Data Flow Supercomputers", IEEE Computer, Vol. 13, No. 4, Nov. 1980, pp. 48–56.Google Scholar
- [Gais82]Gajski D.D, Padua D.A., Kuck D.J., and Kuhn R.H., "A Second Opinion on Data Flow Machines and Languages", IEEE Computer, February 1982, pp. 58–69.Google Scholar
- [Hali80]Halin H.J., Buehrer R., Haelg W., Benz H., Bron B., Brundiers H.J., Isacson A., and Tadjan M., "The ETH-Multiprocessor Project: Parallel Simulation of Continuous Systems", Simulation, Oct. 1980, pp. 109–123.Google Scholar
- [Hwan84]Hwang K. and Briggs F.A., "Computer Architecture and Parallel Processing" McGraw-Hill, Inc., U.S.A., 1984Google Scholar
- [Levi81]Levi P., "Betriebssystem für Realzeitanwendungen", CCG Texte 3, Datakontext-Verlag, Koeln, Germany, 1981.Google Scholar
- [Pope83]Pope K.W., "Asynchronous dual-port RAM simplifies multiprocessor systems", EDN, September 1983, pp. 147–154.Google Scholar
- [Pfis85]Pfister G.F., Norton V.A., "'Hot Spot’ Contention and Combining in Multistage Interconnection Networks", Proceedings of the 1985 International Conference on Parallel Processing, August 1985, pp.790–797.Google Scholar
- [Sieg85]Siegel H.J., "Interconnection Networks for Large-Scale Parallel Processing", Lexington Books, Lexington, U.S.A., 1985.Google Scholar
- [Texa84]Texas Instruments, "TMS 9650, Multiprocessor Interface (MPIF) Data Manual", Texas Instruments Inc., May 1984.Google Scholar