Are You Really Confident That You Are Getting the Very Best From Your Verification Resources?

Your last project had your compute grid running constantly at 100 percent capacity and your simulation licenses were maxed out most of the time. The next chip is twice as large so you must have to double or treble your resources.

Or maybe there is another way based on smart combination of the grid and regression management systems to ensure that every verification cycle is a valid cycle. This paper will show how adding control and visibility to these systems, and then better integrating them, will help your organization get the very best from every verification dollar. The paper will also explain how a regression system can be developed with the infrastructure to control and monitor exactly what is happening. This addition of intelligent automation allows dynamic reaction to the current status and create the maximum throughput and capacity. The diagram in Figure 1 shows the major parts of the complete system that will be explained in this paper.

Not getting the most from technical infrastructure used in a product development cycle can be expensive. The cost, which include software licenses, fixed assets, and wasted time, often can exceed that of the original investment in the infrastructure. This is why it's essential to monitor usage on a job-by-job basis, which requires integration between the license and grid scheduling software to provide for complete visibility. Accounting for metrics such as code churn, bug status as a sign of completion, and performance criteria for bus fabrics or bandwidth within communications applications while integrating the grid, regression and simulation software can paint an even more complete picture of what is happening in the verification environment.

This paper will also describe how to combine these metrics to better track both verification resources and overall verification progress, and how to do so within the tight timescales required to produce right-first-time silicon.

^{Figure 1: Block Diagram of Complete System}

The Regression System

Twin demands increased quality and shortened design cycles put pressure on IP and SoC houses to leverage automation throughout their design and verification processes. In the verification space, these pressures require verification engineers to get more efficient contending with tasks such as coverage closure, bug hunting, smoke and soak testing, all of which are done through running lots of regressions. Regression systems can be heavily scripted, and are often developed and understood by just a small handful of people within a larger organization. Historical baggage is carried over from project to project, and sometimes the majority of users don't even know why they have to run a particular script. It just works until it doesn't.