SoC regression strategy developement

Abstract

The objective of the verifcation process of hardware is ensuring that the design does not contain any functional errors. Verifying the correct functionality of a large System-on-Chip (SoC) is a co-design process that is performed by running immature software on immature hardware. Among the key objectives is to ensure the completion of the design before proceeding to fabrication.

Verification is performed using a mix of software simulations that imitate the hardware functions and emulations executed on reconfigurable hardware. Both techniques are time-consuming, the software running perhaps at a billionth and the emulation at thousands of times slower than the targeted system. A good verification strategy reduces the time to market without compromising the testing coverage.

This thesis compares regression verification strategies for a large SoC project. These include different techniques of test case selection, test case prioritization that have been researched in software projects.

There is no single strategy that performs well in SoC throughout the whole development cycle. In the early stages of development time based test case prioritization provides the fastest convergence. Later history based test case prioritization and risk based test case selection gave a good balance between coverage, error detection, execution time, and foundations to predict the time to completion.