ABSTRACT
Feedback from users of the first generation ISP system plus the development and availability of new VLSI based technologies drove the design of the iPSC/2 node. The new node design was broadly governed by the following goals:
Provide true 32-bit node architecture and performance.
Match the communication performance to the computational performance.
Increase on board memory capacity by using new RAM technology.
Employ a modular functional elements to easily incorporate future technology.
Allow plug compatibility with existing iPSC systems, including interface to co-processors.
Ensure software compatibility for existing iPSC applications.
This paper describes how the iPSC/2 node achieved these goals by leveraging Intel's VLSI expertise and products, surface mount, CMOS and gate array technologies, and small daughter boards to implement modular subsystems.
- 1.Nugent, Steve, TheiPSCI2 Direct-Connect Technology, Proceedings, 3rd Conference on Hypercube Concurrent Computers and Applications, 1988. Google ScholarDigital Library
Index Terms
- The iPSC/2 node architecture
Recommendations
The iPSC/2 direct-connect communications technology
C3P: Proceedings of the third conference on Hypercube concurrent computers and applications: Architecture, software, computer systems, and general issues - Volume 1This paper describes the hardware architecture and protocol of the message routing system used in the iPSC®/2 concurrent computer. The Direct-Connect router was developed by Intel Scientific Computers to replace the store-and-forward message passing ...
Broadcast communication delay metric for the iPSC/2 and iPSC/860 hypercubes
ACM-SE 30: Proceedings of the 30th annual Southeast regional conferenceWhen determining the granularity of a program to be executed in parallel, it is important to have valid information on the cost of communication for the system on which the program is to be run. Previous studies have measured the node-to-node ...
Comments