Reusing semiconductor intellectual property (SIP or IP) blocks is no longer a luxury for the designers of complex chips it is a necessity. As mask costs skyrocket and qualified engineering resources become increasingly scarce, chip design teams are incorporating IP reuse strategies into virtually every chip development plan in order to minimize design cycles, reduce chip respins, and meet time-to-revenue goals.
There have been myriad approaches to SIP reuse over the years, ranging from in-house reuse of low-level design elements to complex system-on-chip integration of high level blocks from various SIP vendors. Fundamentally, successful IP reuse requires well-designed IP, a solid strategy for integration of the IP into the user's design, and dependable support from the IP vendor. Analyst firm Gartner Dataquest recognizes 123 companies as participants in the SIP market, addressing 43 product categories. However, few of these 123 companies have really mastered the secrets to successful IP reuse. Here are some signs that your IP provider might be one of the ones that hasn't figured it out yet:
Warning Sign #1: The IP provider views the development of IP as a "one-time deal." Like any good design team, good IP providers have a stake in the technology they are developing, and their business models and offerings anticipate and allow for reuse. They have product development roadmaps that build on their "core" IP, and they are able to add new features quickly to accommodate the application and integration needs of their customers.
Some ASSP companies that are struggling to sell their products have made the decision to market pieces of an existing design as standalone IP. Since reusable IP can require almost twice as much initial development effort according to some sources, this approach often does not provide the customer with well-designed IP that is truly reusable. If your IP vendor views selling IP as a way to quickly generate revenue, it may be a sign that they don't understand the significant challenges associated with developing, selling, and supporting IP, and you might end up stuck with a piece of "orphaned IP" when they go back to their original business model or go out of business.
Warning Sign #2: The IP provider creates standards-based IP in isolation. At first blush, IP that is based on a standard (SPI4.2 or PCI-Express, for example) seems the most straight-forward to design and deliver. After all, the standards organizations responsible for defining the protocols generally produce a written specification, along with application notes, implementation guides, and so on.
However, these standards and the innumerable documents that support them are often not as solid as you might imagine. There may be ambiguities in the specifications that can only be resolved by applying the combined knowledge and expertise of many implementation teams, each with a different set of customers and requirements (see Warning Sign #6). Good IP providers don't depend solely on written specifications to ensure interoperability but communicate extensively with a variety of standards organizations and implementation teams. If your IP vendor is not an active participant in such exchanges, it may be a sign that they don't recognize the importance of interoperability analysis for standards-based IP.
Warning Sign #3: The IP provider does not offer a complete solution. Just because an IP provider is offering a given piece of IP for sale doesn't mean that what they're offering will completely fill your need. Often, an IP function can be broken into several layers (physical, electrical, logical, protocol, transport) and it's a rare IP provider that truly offers a full solution addressing all of the possible options.
Good IP providers will offer an IP solution that addresses most, if not all, of the layers, even in cases where there may be multiple underlying sources of the IP. Ideally, the entire IP solution is designed by a single source. When this is not possible, partnerships among IP vendors, each with different areas of expertise, will fill the voids that exist for any one vendor.
While it is possible for these types of arrangements to be successful, there is greater risk when using IP that wasn't developed by one entity, especially if you don't understand the relationships between the IP development teams. It is also important to have a single point of contact for any issues that may arise, even if there are multiple third-party sources involved. If your IP provider suggests that you can arbitrarily pair up their IP with someone else's to create the "full" solution, it might be a sign that they don't understand the risks associate with such a solution.
Warning Sign #4: The IP provider supports a very limited number of technology ports. With the possible exception of very technology-specific designs, such as custom mixed-signal circuitry, reusable IP will be designed to be readily portable into new technologies. Even if your IP provider supports your current technology needs, it is inevitable that a given piece of IP will some day need to be ported to a different technology for new products or to provide manufacturing cost reductions for existing products. If your IP provider cannot or will not support your future technology porting needs, you'll be faced with: finding an alternate source for the IP that can be ported into the new technology; redesigning the function yourself in the new technology; or simply abandoning the idea of using a different technology altogether.
A related problem is IP providers that only deliver "hard" macros, even if the underlying design is technology independent. While there are sometimes advantages to hard macros for high-speed designs (proven performance of the IP, less layout work for the integration team, smaller footprint of the IP) such solutions limit IP reusability. If your IP provider tries to convince you that technology portability is not an important parameter in the selection of IP, it might be a sign that they don't appreciate the issues associated with planning for future generations of a system.
Warning Sign #5: The IP provider skimps on support. There are two areas of support that must be considered when choosing an IP provider: pre-sales and post-sales support. Prior to making your final decision about which IP solution to go with, the IP provider must give you access to adequate information, and possibly to the IP design team itself, so that you can develop a high level of confidence in the capabilities of the IP designers.
Inadequate support at this stage could be an early indicator that the post-sales support is going to be lacking, too. There is an adage used by experienced third-party IP consumers that "IP is only as good as the applications engineers that support it." While it is not likely that one person will know the answer to every question, beware if your primary contact person is constantly "getting back to you" with answers to your questions. An unknowledgeable applications engineer might be a sign that your IP implementation may not go smoothly.
Warning Sign #6: The IP provider develops IP without customer input. Good IP providers will have a representative sample of end users, each with unique design needs and environments, involved during initial IP architecture development so that real-world applications can be considered. This beta customer approach ensures reusability and interoperability while allowing optional features and user interfaces to be reviewed. If your IP provider doesn't value beta customers during IP development, it might be a sign that their IP is not going to be as useful or reusable as you need it to be.
Warning Sign #7: The IP provider doesn't offer customization services. Even the best-developed IP cannot possibly meet the unique needs of every customer. The best IP providers are those that also offer expert customization services to their customers, so that a given piece of IP can be modified to conform to specific design or application requirements.
Customizations such as modifying user interfaces to integrate better with the user's logic, changing arbitration schemes to more effectively manage a unique data flow requirement, or adapting internal buffer sizes to more closely match the user's storage needs, are good examples of common customizations for off-the-shelf IP. While some IP vendors, particularly those who supply soft IP in the form of RTL, will encourage users to make their own modifications, this is a risky proposition and should only be undertaken as a last resort. Only the original design team can fully understand the impact of changes to the design, and is capable of performing complete regression testing on the modified design. If your IP provider doesn't offer customization services, it may be a sign that you required modifications may be difficult.
Even with the difficult task of finding a qualified IP provider, it is clear that third party IP is here to stay. As Moore's Law continues to govern integration capabilities of the raw silicon, staffing restrictions, software limitations, and time-to-revenue pressures continue to increase the design gap.