Test Set Minimization via Partitioned Coverage Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In microprocessor verification, existing methods require significant computation time and resources to rank and prioritize tests for optimal coverage, often missing tests with unique coverage that provide essential design insights due to focusing solely on high overall coverage tests.
Innovation Solution
A method is developed to minimize test sets by generating an initial empty and minimal test set, then partitioning and merging additional test sets into control and experiment sets, identifying and merging tests with unique coverage to maximize overall coverage without lengthy ranking processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If individual tests are ranked using rank tools to show which tests have superior coverage, then test prioritization is improved, but computation time and resources increase significantly
Solution Approach 1:
The patent segments the test set into multiple partitions and processes them in parallel using multiple threads. Each thread independently analyzes a partition of tests, computing coverage metrics without requiring sequential ranking of all tests. This divides the computationally intensive ranking task into smaller, concurrent units, reducing overall computation time while maintaining effective test prioritization.
Solution Approach 2:
The patent computes coverage metrics for only a predetermined number of tests (a subset) rather than analyzing the entire test set. By selecting and analyzing a representative sample of tests in parallel, the system achieves sufficient prioritization accuracy without the computational cost of exhaustive analysis, effectively applying partial action to resolve the contradiction.
2Measurement precision
If a subset of tests with the best coverage is selected, then coverage quality is improved, but overlap in coverage amongst tests increases
Solution Approach 1:
The patent introduces a coverage database as an intermediary structure that tracks and merges coverage information from multiple test partitions. This database enables the system to identify and eliminate redundant coverage across tests by comparing against previously analyzed partitions, allowing selection of tests with best coverage while minimizing overlap through structured comparison and merging operations.
3Productivity
If tests with low overall coverage are excluded to focus on high coverage tests, then computation efficiency is improved, but unique coverage from low coverage tests is lost
Solution Approach 1:
The patent implements feedback mechanisms where coverage information from each analyzed test is immediately recorded in the coverage database. This feedback allows the system to identify tests that provide unique coverage contributions even if their overall coverage metrics are lower. The feedback loop ensures that tests with unique value are detected and retained, preventing loss of important coverage information while maintaining computational efficiency through parallel processing.
Data Source
AI summary
A method for minimizing a test set for optimal coverage is disclosed. The method includes generating a first test set which is both an empty and minimal test set. Then, generating a second test set with a predetermined number of tests. Further, partitioning the second test set into a control test set and an experiment test set. Subsequently, providing a list of tests for coverage by merging the control test set with the first test set to form a merged list of sets.