Proteus: Control and management system

The Proteus is a highly parallel MIMD, multiple instruction, multiple-data machine, optimized for large granularity tasks. The system is designed to use 256 to 1024 RISC processors. Computer vision algorithms consist of sub-algorithms, which can be executed in parallel. Advanced software system for partitioning, scheduling, development, and execution of tasks can utilize this fact in a data flow programming paradigm to increase throughput. The authors describe how these modules interact with each other such that permit the efficient control of large grained parallelism without having to handle the general concurrency problem.

[1]  Robert M. Haralick,et al.  Insight: a Dataflow Language for Programming Vision Algorithms in a Reconfigurable Computational Network , 1987, Int. J. Pattern Recognit. Artif. Intell..

[2]  Robert M. Haralick,et al.  Proteus system architecture and organization , 1991, [1991] Proceedings. The Fifth International Parallel Processing Symposium.

[3]  Xinhua Zhuang,et al.  Image Analysis Using Mathematical Morphology , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[4]  Linda G. Shapiro,et al.  Programming Parallel Vision Algorithms: a Dataflow Language Approach , 1988 .

[5]  Arun K. Somani,et al.  The generalized folding-cube network , 1991, Networks.

[6]  Michael L. Campbell Static Allocation for a Data Flow Multiprocessor , 1985, ICPP.