Ring Path Control Bus for Low Latency On-Chip Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional buses in processing devices do not scale well for larger systems, leading to high latency in on-chip communication between components due to inefficient routing structures.
Innovation Solution
A control bus with a ring path configuration that propagates read and write requests until they are serviced by receiving circuits, allowing for efficient circulation and reducing latency by supporting a large number of requesting and receiving circuits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a tree structure bus is used for routing data packets between on-chip components, then routing capability is provided, but latency becomes unacceptably high in larger chips with more components
Solution Approach 1:
The bus system is segmented into multiple independent ring paths instead of using a single hierarchical tree structure. Each ring path operates autonomously to service requests, dividing the communication workload and reducing contention-induced latency as the number of components increases
Solution Approach 2:
The patent transitions from a two-dimensional hierarchical routing approach to a multi-dimensional ring-based topology. Requests can circulate through multiple rings simultaneously, adding a temporal dimension to routing where requests that miss their target can continue circulating rather than requiring complex backtracking through hierarchical levels
2Adaptability or versatility
If conventional bus schemes are used to support a large number of requesting and receiving circuits, then routing capacity is limited, but latency increases significantly
Solution Approach 1:
Requests that are not serviced on their first pass continue circulating through the ring path without being discarded or requiring complex re-routing. This continuous circulation allows requests to be serviced on subsequent passes, maintaining system productivity even as the number of circuits increases and request contention rises
Solution Approach 2:
The system changes the fundamental parameter of request handling from 'service once or discard' to 'service when encountered in circulation'. This parameter change allows the system to scale to more circuits because requests persist through multiple cycles, increasing the probability of successful service without requiring additional routing infrastructure
Data Source
AI summary
A device comprising: a control bus; a plurality of requesting circuits each accessible on the control bus, wherein each of the plurality of requesting circuits is operable to dispatch read or write requests to the control bus for delivery to at least one of a plurality of receiving circuits, and the plurality of receiving circuits each accessible on the control bus, and each of which is operable to receive requests from the at least one control bus and service the requests by providing at least one of read or write access to storage associated with the respective receiving circuit, wherein the control bus provides a ring path configured to support, the requests in circulation in the ring path, wherein the control bus is configured to propagate each of at least some of the requests at least until those requests have been serviced by at least one of the receiving circuits.