Selective Reuse of Integrated Coverage Models for Verification Signoff
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic design automation (EDA) methods are inefficient in verification signoff due to the creation of duplicate coverage models and excessive loading of duplicate information during the merging process, leading to high computation times.
Innovation Solution
The method involves populating a database repository with coverage models, computing signatures for code and functional coverage models, and selectively reusing unique models by creating new ones or copying existing ones based on signature matches, thereby reducing redundant model creation and loading during the merging process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If coverage models are created for every simulation test run, then verification completeness is ensured, but computation time increases due to redundant model creation and merging
Solution Approach 1:
The system performs preliminary actions by creating coverage models only when necessary (when design or testbench changes occur) rather than for every simulation run. The signature-based detection mechanism identifies when coverage models need to be created versus when they can be reused, allowing the system to prepare only the essential models in advance.
Solution Approach 2:
The system creates copies of coverage models from the repository when signatures match, rather than recreating them. This copying mechanism allows rapid duplication of proven coverage models for subsequent simulation runs where the design and testbench remain unchanged, significantly reducing computation time while preserving verification completeness.
2Measurement precision
If coverage models are loaded and merged for every simulation test run, then accurate coverage measurement is achieved, but processing speed decreases due to excessive loading of duplicate information
Solution Approach 1:
The system performs preliminary signature computation and model identification before the loading phase. By pre-determining which coverage models need to be loaded based on signature matching, the system avoids the unnecessary loading of duplicate information, thereby improving processing speed without compromising coverage measurement accuracy.
Solution Approach 2:
Instead of loading and merging duplicate coverage models for every simulation run, the system copies references to previously loaded models when signatures match. This copying approach maintains accurate coverage measurements while dramatically reducing the processing overhead associated with loading and merging redundant data.
3Productivity
If signature-based detection is implemented for coverage model re-use, then redundant processing is reduced, but system complexity increases due to signature computation and storage
Solution Approach 1:
The system uses simple signature strings (hashes) as copies of the essential characteristics of coverage models. These signatures serve as lightweight identifiers that capture the essence of design and testbench configurations without requiring complex data structures. The signature-based matching mechanism is computationally efficient and adds minimal complexity compared to full model comparison.
Solution Approach 2:
The system transforms the complex problem of coverage model equivalence detection into a simpler parameter-based comparison by computing signatures from design and testbench configurations. This parameter transformation approach converts a potentially complex structural comparison into a simple string matching operation, reducing system complexity while maintaining accuracy.
Data Source
AI summary
In the context of electronic design automation and in particular coverage-driven verification, each of a number of integrated coverage models to be merged for coverage analysis are divided between a code coverage model and a functional coverage model. During a coverage model generation phase, new code coverage models or functional coverage models are created only if they are not already in a coverage model database repository; otherwise, they are copied. During a merging phase, code coverage models or functional coverage models are loaded only if they have not already been loaded. Signatures that can be based on checksums can be used to determine whether the models are unique or duplicates of those already created or loaded. Selectivity in the creation and re-use of the respective coverage models provides computational time savings in each of the creation and loading phases.