Flop Vulnerability Categorization Using Circuit Data Dependencies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing safety or reliability analysis techniques for data processing circuits fail to accurately assess the vulnerability of flops to transient errors, as they do not adequately consider data dependencies, leading to an incomplete view of circuit reliability.
Innovation Solution
An apparatus and method that categorize flops into vulnerable, conditional, and isolated categories based on data dependencies, using receiving circuitry to determine these dependencies and outputting the categorization, which indicates the vulnerability of flops to transient errors, allowing for a more accurate assessment of circuit reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional activity-based analysis is used to assess flop vulnerability, then the analysis is simple to perform, but the reliability assessment accuracy deteriorates
Solution Approach 1:
The patent segments flops into three distinct categories (vulnerable, conditional, isolated) based on their data dependency characteristics. This segmentation allows for differentiated reliability assessment, where vulnerable flops receive stricter analysis while isolated flops can be quickly identified as low-risk, thereby improving overall assessment accuracy without uniformly increasing complexity across all flops.
Solution Approach 2:
The patent performs preliminary classification of flops into vulnerability categories before conducting detailed reliability analysis. By pre-categorizing flops based on their data dependency patterns, the system prepares the analysis groundwork in advance, enabling more accurate subsequent assessment while reducing the computational burden during the actual reliability evaluation phase.
2Reliability
If comprehensive data dependency analysis is performed to improve vulnerability categorization, then the reliability assessment improves, but the analysis time increases
Solution Approach 1:
The patent applies local quality by tailoring the depth of data dependency analysis to each flop's specific characteristics. Vulnerable flops undergo comprehensive data dependency tracing to ensure accurate reliability assessment, while isolated flops receive minimal analysis since their impact on overall circuit reliability is negligible. This localized approach maintains high reliability assessment quality for critical components while reducing total analysis time.
3Reliability
If fault injection testing is used to assess circuit reliability, then the results are comprehensive, but the testing process is slow and has limited applicability
Solution Approach 1:
The patent replaces the mechanical fault injection testing process with a data dependency-based analytical model. Instead of physically injecting faults and observing circuit behavior, the system uses logical analysis of data dependencies to predict vulnerability. This substitution dramatically increases testing speed and applicability while maintaining reliable vulnerability assessment, as the analytical approach can be applied universally without physical testing constraints.
Data Source
AI summary
An apparatus is provided comprising receiving circuitry to receive a representation of a circuit comprising a plurality of flops. Categorisation circuitry determines data dependencies between the flops from the representation and generates a categorisation of the flops into one of at least: a vulnerable category, a conditional category, and an isolated category, in dependence on the data dependencies. The categorisation indicates the vulnerability of the flops to transient errors. Output circuitry outputs the categorisation of the flops. The conditional category comprises those of the flops whose change in value is indicated by a change in a value in a corresponding flop in the flops or corresponding signal. The vulnerable category comprises those of the flops that are absent from the conditional category and whose change in value is affected by one of the flops or a signal and the isolated category comprises the flops that are absent from the conditional category and that are absent from the vulnerable category.