A Comprehensive Automation Approach with Jenkins for ARM-based CPU Verification

Abstract

Automation in industries is a growing field that helps increase productivity and shorten the time-to-market period. A typical verification flow in Very Large Scale Integration (VLSI) involves compiling the design builds, compiling the tests, simulating the test, and reporting the results. Only after this process can a developer actually start debugging the tests. Automation in these initial steps can help reduce human error, put all developers on the same page, expedite the qualification of changes, and provide better reporting. In the process of verifying a CPU, this form of automation could be immensely helpful. Usually, simulations at CPU verification scale consist of thousands of tests that are configured to run for billions of clock cycles. This can lead to a lack of compute resources for running the simulations. One of the trending solutions for this is to use cloud computing services such as Google Cloud, Amazon Web Services, etc. And with an automation platform, very regulated resource management can be carried out. Jenkins is an open-source automation server with tons of plugins and services that can bring substantial changes to the verification flow. With Jenkins, multiple developers can test their changes simultaneously, have automatic report generation as per team requirements, and, in turn, reduce development time. It is also easy to integrate it with cloud services. Packaging of simulation libraries can be done in Jenkins, which is then sent to the cloud for generating the results. These results can be fetched back to Jenkins and reported as per team requirements. Jenkins supports multiple scripting languages. Groovy scripts were used in this project. Jenkins servers were first configured to pick changes from users via. Their Gerrit branch. Users can configure their runs, such as enabling coverage on simulation, limiting the number of tests, running specific suites, etc. Once the runs are complete, an email with a link to the report is sent to the respective recipients. The user can also configure the run to be on the cloud or on-premises.