Prover Selection via Feature Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for selecting verification engines in formal verification are inefficient, relying on brute force or engineer intuition, and lack a systematic approach to optimize CPU time and convergence probability, leading to suboptimal resource utilization.
Innovation Solution
A computer-implemented method using a pipeline of transformers and estimators that collects and transforms design features through linear regression and classification to select the most suitable prover for verification tasks, configuring step sizes based on design parameters to minimize runtime and maximize convergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If brute force or engineer intuition is used for prover selection, then implementation is simple, but verification efficiency and CPU time optimization are poor
Solution Approach 1:
The system performs preliminary actions by collecting design features and property characteristics before verification, transforming them into a feature vector, and using a trained classifier to predict the best prover in advance. This preliminary classification avoids trial-and-error approaches and directly selects the most suitable prover, thereby improving verification efficiency without requiring complex runtime adjustments.
Solution Approach 2:
The patent introduces an intermediary classification system that acts as a mediator between the verification problem and the provers. The classifier, trained on historical verification data, serves as an intelligent intermediary that translates design features into prover selections, replacing brute force methods while maintaining manageable system complexity through modular architecture.
2Measurement precision
If multiple provers are tested to find the best one, then prover selection accuracy improves, but CPU time and resources increase
Solution Approach 1:
The system performs prover evaluation preliminarily during the training phase using historical verification data. The classifier learns from past performance patterns without requiring runtime testing of multiple provers. This preliminary learning enables accurate prover selection during actual verification while avoiding the time cost of testing multiple provers for each new verification task.
Solution Approach 2:
The patent uses copying by training the classifier on copies of historical verification data and prover performance patterns. Instead of testing actual provers on new designs (which consumes CPU time), the system uses learned patterns from training data copies to predict the best prover, achieving high selection accuracy without runtime prover testing.
3Reliability
If prover configuration is optimized for each verification task, then verification quality improves, but setup time and complexity increase
Solution Approach 1:
The patent implements a universal classifier system that handles multiple verification tasks and prover selections through a single trained model. The classification module serves multiple functions: it evaluates design features, predicts prover performance, and selects appropriate provers across different verification scenarios. This multi-functionality improves verification quality consistently while avoiding the need for separate configuration procedures for each task.
Solution Approach 2:
The system achieves self-service by automatically configuring the verification process through the trained classifier. The classifier autonomously selects the best prover based on design features without requiring manual configuration or complex setup procedures. This automated self-configuration maintains high verification quality while minimizing setup time and operational complexity.
Data Source
AI summary
A computer implemented method of selecting a prover among a plurality of provers for a design to be verified. The method comprises collecting, by a data module, raw data relating to the design, and extracting from the raw data a plurality of input features, transforming, by a transformer module, the plurality of input features, wherein transforming the plurality of features comprises applying a linear regression to the plurality of features, classifying using a classification module, the provers from the plurality of provers, in which the classification module is adapted to predict a best prover being the prover which solves a property faster than the remaining provers of the plurality of provers, selecting one or more provers based on the results of the classification.