Configurable SoC Network Paths for Tile Failure Bypass
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing system-on-chip (SoC) networks have hardcoded physical paths and unique identification tags, limiting flexibility and rendering the entire network inaccessible if a single tile fails, leading to vulnerabilities and downtime.
Innovation Solution
A dynamically configurable SoC network where nodes are self-programmable and can select multiple parent interfaces, allowing reconfiguration of paths and bypassing failed components, enabling flexible and resilient communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a hardcoded physical path is used in the SoC network, then the network structure is simple and easy to manufacture, but the flexibility is limited and the entire network becomes inaccessible if a single tile fails
Solution Approach 1:
The patent implements dynamic parent interface selection where each node can dynamically choose which parent interface to use based on operational needs. This allows the network topology to change at runtime, enabling flexibility and alternative communication paths while maintaining a relatively simple physical structure. The dynamic selection mechanism resolves the contradiction by making the logical network structure adaptable without significantly complicating the physical device architecture.
Solution Approach 2:
The patent gives each node multiple parent interfaces that can serve different functions. Each parent interface can be selected based on the desired communication path or failure recovery needs. This multi-functionality allows the same physical hardware to support multiple logical network configurations, achieving flexibility without requiring separate dedicated structures for each configuration.
2Reliability
If a single parent interface is assigned to each node, then the device complexity is reduced, but the reliability decreases when a tile fails
Solution Approach 1:
The patent implements beforehand cushioning by providing each node with multiple parent interfaces before any failure occurs. This redundant interface configuration ensures that if one parent interface or connected tile fails, the node already has alternative paths available. The redundancy is built in advance, preventing network inaccessibility without requiring complex dynamic reconfiguration during failure events.
Solution Approach 2:
The patent changes the parameter of parent interface count from one to multiple for each node. This parameter change fundamentally improves reliability by providing alternative communication paths. While it does increase interface configuration complexity, the patent manages this through systematic selection mechanisms that allow nodes to autonomously determine which parent interface to use based on simple criteria, thereby limiting the practical complexity increase.
3Reliability
If multiple parent interfaces are provided for each node, then the resilience to failures is improved, but the device complexity increases
Solution Approach 1:
The patent implements self-service by enabling each node to autonomously select its parent interface without requiring complex centralized control or manual configuration. Each node independently monitors its available parent interfaces and selects the appropriate one based on local information and simple selection criteria. This self-service mechanism manages the complexity of multiple interfaces through distributed autonomous decision-making rather than complex centralized management.
Solution Approach 2:
The patent uses dynamic selection where nodes can change their active parent interface based on operational conditions. This dynamic behavior allows the system to adapt to failures and changing conditions automatically. The dynamic selection process manages interface complexity by using simple selection logic that can be implemented through basic control mechanisms rather than complex management systems.
4Adaptability or versatility
If the network path is fixed during manufacturing, then the manufacturing precision is easier to achieve, but the ability to reconfigure post-implementation is lost
Solution Approach 1:
The patent applies preliminary action by pre-configuring multiple parent interfaces at each node during manufacturing, but leaving the selection of which interface to use for later runtime determination. This preliminary setup of multiple interfaces maintains simple manufacturing requirements while enabling post-implementation reconfiguration. The actual path selection is deferred to runtime when the system can adapt to actual operational needs.
Solution Approach 2:
The patent introduces dynamics by allowing the logical network path to change after manufacturing is complete. While the physical interfaces are fixed during manufacturing, the active parent interface selection is dynamic and can be changed at runtime. This separation of fixed physical structure from dynamic logical configuration achieves reconfiguration capability without compromising manufacturing simplicity or precision.
Data Source
AI summary
A device includes a processing device, a set of local resources coupled to the processing device, a plurality of parent interfaces, and a plurality of child interfaces. Each parent interface and each child interface is configured to couple the processing device to a respective node residing outside of the device when selected. The processing device is to select one parent interface, such that a selection of the one parent interface causes the respective node coupled to the one parent interface to operate as a parent of the processing device. The processing device is to further select one child interface, such that a selection of the one child interface causes the processing device to operate as a parent of the respective node coupled to the one child interface.