Close

Presentation

SMART Adaptive Regression Using Nearest Neighbours Algorithm
TimeMonday, July 11th2:15pm - 2:30pm PDT
Location2010, Level 2
Event Type
Engineering Tracks
Front-End Design
Topics
AI
Cloud
Front-End Design
DescriptionRunning a full test suite for each design or verification change by every individual during the project development phase is not practical. This costs machine time and long wait for the results to qualify every change.

So, running a compact qualification run is the right way to ensure the basic quality of each update consisting of single or multiple file changes.

Quality of such qualification regression dictates the overall quality of the updates released during the day. High quality qualification regression helps to detect the design and verification bugs immediately during the development phase by stressing the right functional areas and respective interconnected blocks. This tremendously reduces the debug loop by avoiding buggy releases going through the overnight regressions.

So, selecting the right set of tests based on local changes is more efficient and effective than running a fixed set of tests. This paper presents an idea for smartly selecting the right set of tests using a mathematical approach based on nearest neighbors algorithm. All the tests are graded based on different criteria e.g. coverage or certain scenario. User feeds the information about the changes to be qualified and the tool selects the appropriate tests based on their grading.