IP Core Regression Testing via Dependency Knowledge Graph
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing complexity of semiconductor IP cores and their constituent blocks leads to lengthy and resource-intensive regression testing, which can result in delayed validation of design changes and increased chances of faulty commits due to the complexity and size of development teams.
Innovation Solution
A regression suite apparatus that utilizes a knowledge graph to identify dependencies between functional blocks, allowing for the selection of a subset of tests using a nearest neighbor algorithm in a multidimensional space, reducing the number of tests required to run by up to 90% while ensuring high qualification and detection of faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a full regression suite is run on updated IP core designs, then comprehensive validation is achieved, but testing time increases significantly (up to 12 hours per run)
Solution Approach 1:
The patent segments the full regression suite into multiple levels (level 1: input functional blocks, level 2: dependent blocks, level 3: transitively dependent blocks). By dividing the test suite hierarchically based on dependency relationships, the system can selectively execute only necessary test levels rather than running the complete suite, thereby reducing testing time while maintaining validation reliability for the updated portions.
Solution Approach 2:
The patent applies partial action by executing only the subset of tests relevant to the updated functional blocks and their dependents. Instead of running 100% of the regression suite, the system identifies and executes only the necessary portion (e.g., level 1 and level 2 tests for a given change), achieving sufficient validation without the excessive time cost of comprehensive testing.
2Reliability
If the regression suite is run overnight to accommodate complex testing, then design validation can be completed, but the debug cycle becomes excessively long and complex
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing the dependency graph of functional blocks before regression testing. This dependency map is built in advance, allowing the system to quickly determine which test levels are necessary based on the specific functional blocks that were updated, rather than running through the entire complex debug cycle every time.
Solution Approach 2:
The patent introduces dynamics by making the regression test execution adaptive and flexible. The system dynamically determines the appropriate test levels to execute based on the specific changes made to functional blocks, rather than following a fixed, static testing schedule. This dynamic approach adjusts the debug cycle complexity to match the actual scope of changes.
3Productivity
If team members work on different blocks independently, then parallel development is achieved, but the chance of faulty commits increases due to lack of full IP understanding
Solution Approach 1:
The patent implements feedback by running targeted regression tests immediately after each commit to the version control system. The system provides rapid feedback on whether a commit introduced faults by executing the relevant subset of tests (based on the dependency graph), allowing team members to verify their changes before they can contaminate the broader codebase. This feedback loop maintains commit quality while preserving parallel development benefits.
Solution Approach 2:
The patent introduces an intermediary mechanism - the automated regression suite apparatus with dependency-based test selection - that mediates between independent team member commits and overall IP quality. This intermediary automatically identifies and executes the necessary tests for each commit, bridging the gap between individual block development and holistic IP validation without requiring team members to understand the entire IP.
4Reliability
If a single bug contaminates the entire regression suite run, then validation of other commits becomes impossible, but running multiple separate tests increases overall testing time
Solution Approach 1:
The patent segments the regression suite execution into independent level-based groups (level 1, level 2, level 3). If a bug is detected in tests at one level, the system can stop execution and report the issue without contaminating the entire suite run. Other levels remain valid and can be executed separately, allowing partial validation results to be preserved and reducing the need to re-run the complete suite.
Solution Approach 2:
The patent applies partial action by executing only the necessary test levels for each commit rather than the full suite. This reduces the window of opportunity for bug contamination and allows the system to achieve validation with a subset of tests, minimizing the impact of potential bugs while reducing total testing time.
Data Source
AI summary
A regression suite apparatus for semiconductor IP having a plurality of functional blocks provides a knowledge graph defining, for each pair of first and second functional blocks, an indication of whether the second functional block depends on the first functional block; in case of a dependency, the knowledge graph comprises a weight representing a degree of dependency of the second functional block on the first functional block; an input for indicating functional blocks changed in an updated design; a selector arranged to select a subset of the tests by selecting numbers of tests of each of a plurality of different levels, wherein a first level selects tests corresponding to an input functional block and a second level selects tests corresponding to the input functional block in combination with second level blocks dependent on the input functional block; and apparatus for running the selected subset of tests on the updated design.