A Large Scale, Homogenous, Fully Distributed Parallel Machine, II

Abstract

Intended ApplicationsThe present paper complements the hardware proposal for a fully distributed parallel processor presented in "A Large Scale, Homogeneous, Fully Distributed Parallel Machine, I, " by describing a suitable software structure for its operating system management. It is shown that the most basic operating system functions can be performed on a purely local basis, i.e. in such a fashion that identical pieces of the operating systems are concurrently executed by each of the processing elements.

On characterizing bandwidth requirements of parallel applications

“…The hypercube is assumed to have serial (1-bit wide) unidirectional links and uses the -cube routing algorithm [29]. Messages are circuitswitched using a wormhole routing strategy and the switching delay is assumed to be zero.…”

A load-balanced congestion-aware wireless network-on-chip design for multi-core platforms

“…In [3], performance and feasibility comparison between 7-port and 5-port routers was discussed. 7-port routers with adaptive minimal XY routing algorithm have been proven to have a better performance than 5-port routers with Dimension Order Routing (DOR) [33], ROMM [24] and O1TURN [29] routing algorithms. Furthermore, to provide deadlock free solution for 5-port routers, virtual channels that require more buffering have to be deployed to prevent deadlock situations if using adaptive routing schemes.…”

On schemata and $L$ systems for parallel algorithms

“…Aspects of the hardware implementation of the model proposed in [3] are discussed in [4]. Other inspiring research that provides motivation of this work includes Albrecht [2], Schwenkel [14], Sullivan and Bashkow [15] and Sullivan, Bashkow and Klappholz [16].…”