Even a relatively straightforward protocol like AMBA AHB presents a huge complexity challenge to the verification team. A simple bus read has hundreds of permutations that must be captured both in the human readable verification plan document and in the actual verification process itself.

To comprehensively test such a protocol manually would require a massive, labor-intensive verification plan and process. The limitations of human ability to capture all the verification scenarios make the process even riskier. This is one example of how End-to-End VIP, or a universal verification IP or component, provides a superior result to internally built solutions.

These high-end universal VIPs abstract the protocol details to a level that humans can effectively manage. Instead of being forced to track millions of protocol permutations (such as, "did the bus read from location FFFCA while the grant was enabled during a buffer overflow condition") they provide a higher abstraction level. With this type, VIP engineers will instead ask "have all reads been exercised?"

Some readers may be thinking most of the permutations don't matter since they're essentially the same as others already verified. While it's true that performing a read from address A is not far different than reading from address B, it is still important in the big picture. Human testers will always miss combinations and edge values, thereby weakening verification completeness.

The only approach that enables complete functional coverage employs context sensitive automatic stimulus generation. This ensures that all the important configurations and permutations have been stimulated, even infrequent device conditions such as error conditions. For example, you can automatically generate stimuli for the full range of potential behaviors including error injection.

Furthermore, while simulation cycles are becoming less expensive, they aren't free (and never will be). Therefore it's important not to waste them. End-to-End VIP directs the verification engines to avoid protocol functionality not in the specific design as well as already verified functionality. This is a major advantage of using a functional coverage approach. It addresses the entire process from plan to verification closure.

Use an automatic plan-to-closure process
 

Imagine you're blindfolded and dropped into an open field miles from any city. How would you find your way home? You don't know where you are and you don't know where you're going. That is the conundrum many verification teams find themselves in. They're charging ahead yet they don't have a clear map of the territory. Adding to their woes, such teams are unable to concisely and/or accurately convey to their management where they are in the verification process or when they'll finish.

To address these issues End-to-End VIP must deliver three key elements.

1. End-to-End VIP must provide a clear definition of closure including metrics that everyone on the team understands. This requires defining a DUT functionality matrix spelling out what is and what is not to be tested. The matrix is then codified into an executable verification plan.

2. It must provide a way to observe and measure the verification results relative to the metrics. This enables everyone to understand where the verification stands relative to closure.

3. A reporting mechanism is needed so that the team and their management know precisely where the verification process stands. They can also accurately predict how much more time and resource reaching closure will require.
 

While a plan to closure process is necessary, it is still not sufficient. The verification process must be automated to successfully verify multi-million gate SOCs. Even if you had the thousands of man-months needed, the complexity of managing thousands of individual test cases is beyond the capacity of manual testing. Automation capabilities planned into VIP address three essential elements. First, they provide an executable verification plan. It is both a human readable and machine readable document that spells out the functionality matrix to be verified (see Figure 2 below).

Second, they add in automatic stimulus generation to ensure that each component of the functionality matrix is exercised. And third, they provide the coverage points and coverage metrics to enable you to assess and report on verification completeness.

Verification teams also commonly struggle with the competing tasks of finding bugs and reaching coverage closure. End-to-End VIP alleviates this struggle by automating both. For example, they provide failure triage to find bugs. And, when not finding bugs, they work to deliver maximum coverage. Therefore End-to-End VIP enables each additional simulation seat to find more bugs and/or to increase coverage.

Employ a reuse methodology
 

Verification IP is all about reuse. VIP must be usable (and reusable) by expending only minor effort. This is true of first usage or when moving from block to chip to system level verification. To achieve this goal requires a significant investment in a reusable VIP architecture and a reuse methodology. This can be supplied either by the design team or by the VIP supplier, but it must be in place to reap the full benefits of VIP reuse.

This is not just theory. Users employing a market proven reuse methodology together with End-to-End VIP have reported 50-100X productivity gains when creating and reusing their verification environments.

Page 1: Verification IP takes a broader role
Page 3: Verification IP takes a broader role