Hybrid Verification Environment With AI Models for Digital IC Coverage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing functional verification methods for Digital ICs face challenges in efficiently addressing the increasing design size and complexity, requiring significant computational resources, long simulation times, and manual intervention, with limited flexibility and scalability, especially when integrating AI/ML techniques directly into verification environments.
Innovation Solution
A method that integrates a machine learning cluster with a hybrid verification environment, utilizing AI models for stimuli generation, failure analysis, and output prediction, allowing on-the-fly data requests and responses to enhance verification efficiency and reduce manual effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If simulation-based verification methods are used to verify Digital IC designs, then verification can be performed with existing EDA tools, but the verification process becomes time-consuming and requires significant computational resources as design size increases
Solution Approach 1:
The patent introduces an AI-based intermediary system that acts as a mediator between the verification environment and the design under test. The AI model analyzes simulation data and generates verification stimuli, serving as an intelligent intermediary that accelerates the verification process without requiring direct manual intervention or extensive computational simulation time.
Solution Approach 2:
The patent replaces traditional mechanical simulation processes with AI-based verification mechanisms. Instead of relying solely on exhaustive simulation cycles, the system uses machine learning models to predict verification outcomes and generate targeted test cases, substituting the mechanical simulation approach with an intelligent system that achieves similar verification goals faster.
2Reliability
If manual verification processes are used to ensure high verification coverage, then verification thoroughness improves, but verification time and effort increase significantly
Solution Approach 1:
The patent implements a self-service verification system where the AI model autonomously analyzes verification data, identifies coverage gaps, and generates appropriate stimuli without requiring manual verification engineer intervention. The system serves itself by automatically maintaining verification coverage targets and adapting to design changes, thereby achieving high verification coverage with improved productivity.
Solution Approach 2:
The patent incorporates feedback mechanisms where the AI model continuously monitors verification coverage metrics and adjusts verification strategies accordingly. Simulation data is fed back into the AI system, which uses this information to refine future verification approaches, ensuring high coverage while reducing redundant manual verification efforts through adaptive feedback loops.
3Adaptability or versatility
If verification environments are made more flexible to handle design changes, then adaptability improves, but system complexity increases
Solution Approach 1:
The patent applies parameter changes by using AI models that can adapt to different verification scenarios and design changes through learned patterns rather than requiring structural modifications to the verification environment. The AI system adjusts its behavior based on input data characteristics, providing flexibility without increasing the complexity of the underlying verification infrastructure architecture.
4Reliability
If more simulations are conducted to reduce uncertainty in verification results, then verification reliability improves, but computational resources and verification costs increase
Solution Approach 1:
The patent applies partial action by using AI models to identify and execute only the necessary verification simulations required to achieve adequate coverage, rather than conducting exhaustive simulations. The AI system determines the minimum sufficient verification effort needed, avoiding unnecessary computational resources while maintaining verification reliability through intelligent selection of critical test cases.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method to prepare a verification project in a verification system, the method comprising the following steps: Processing verification artefacts by an integrating system, Configuring a machine learning cluster by the integrating; Configuring a function call management unit of the hybrid verification environment by the integrating system - Instantiating a system under test of the verification project in the hybrid verification environment by the integrating system; Configuring a decision management unit for the verification project in the hybrid verification environment by the integrating system, the decision management unit comprising a definition of processes to be accomplished in the verification project; and Configuring the prompt design module to translate function call management and the selected ai models in the machine learning cluster for the verification project.