Multiprocessor Data Exchange via Segmented Routing and Buffers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data exchange mechanisms in multiprocessor systems are limited by serialization, resource-intensive connections, and scalability issues, which hinder parallel execution and efficiency in embedded applications.
Innovation Solution
A system and method utilizing data exchange buffers and routing modules that allow flexible data exchange between processing modules, enabling simultaneous data transfer and routing through predetermined paths, with modes like ping-pong and FIFO, and a rotating priority scheme for efficient channel allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a common bus with arbitrator is used for data exchange, then any processor can communicate with any other processor, but communication gets serialized and processing is delayed
Solution Approach 1:
The patent segments the communication system into multiple independent point-to-point links between processors, replacing the single common bus. Each link operates independently, allowing simultaneous data transfers between different processor pairs, thus eliminating serialization while maintaining communication flexibility.
Solution Approach 2:
The patent introduces dedicated link controllers as intermediary devices that manage individual point-to-point connections. These controllers handle protocol management, flow control, and arbitration locally, enabling parallel communication without requiring a centralized bus arbitrator that serializes all traffic.
2Productivity
If dedicated links connect every processor directly to every other processor, then parallel execution is achieved, but resource consumption increases
Solution Approach 1:
The patent implements local quality by providing direct dedicated links only for communication paths that are most frequently used or critical for parallel execution. Not all processor pairs have direct links; instead, the system provides high-speed direct connections for essential communication patterns while accepting indirect paths for less critical traffic, thus reducing overall resource consumption while maintaining parallel execution capability.
3Productivity
If shared memory is used for data exchange, then source processor can proceed to next job, but synchronization is required and parallel communication is limited by memory ports
Solution Approach 1:
The patent introduces message buffers as intermediary storage devices located at each communication endpoint. These buffers decouple the producer and consumer processors, allowing the source processor to write data and proceed to the next job without waiting for the destination processor. The buffers automatically manage data availability and eliminate the need for complex synchronization protocols, while enabling parallel communication beyond memory port limitations.
Data Source
AI summary
A system for data exchange in multiprocessor computer system is disclosed. The system includes at least one processing module adapted to communicate with at least one of a plurality of routing modules or a plurality of buffer modules or both and at least one communicating interfaces to facilitate data exchange in multiprocessor computer system.