|
|
|
|
|
|
 |
||||||||||
|
Building a Total Quality Experience into Silicon IP - Delivering DesignWare Silicon IP into SoC Designs
Kevin Walsh, Director of Marketing
Synopsys, Inc. Introduction Developing integrated circuits (ICs) at 130, 90 and 65 nanometers is a risky and expensive business but rewarding when the design is successfully implemented. Now more than ever, developers of complicated system-on-chip (SoC) design rely on silicon IP (SIP) both internally developed and from third parties to achieve their time-to-market goals. Too often, SIP providers have oversold the capabilities of their products. Claims that SIP is “easy to integrate,” supports “all the features you need for your application,” is “bug-free,” are less persuasive than ever to a savvy SIP buyer. It’s time for a new supplier-buyer relationship. After nearly 10 years of practice, SIP vendors must deliver a “Total Quality Experience” to SIP buyers. But judging the quality of SIP is more than just answering yes to rows upon rows of questions in a matrix. Quality is measured most effectively through direct customer feedback and systematic quality investigation. Taking a page from so many other industries that have suffered from lack of a quality, SIP providers must now step up and deliver a new experience for their SoC customers. Risk of Failure Understanding the customer reaction to an important purchase begins by understanding the risks and costs involved in making that purchase. The economics of any single point of failure when developing an SoC are quite high. They involve both a time-to-market impact and a direct cost. Direct costs are easily measured. A partial mask set or complete mask set can run from hundreds of thousand of dollars at 130-nanometer technology to over one million dollars at 90 nanometers. Time-to-market schedule delays result in lost market share. These costs can be in the hundreds of millions of dollars depending on your product area and the competition. Exacerbating the situation is that the SIP provided is often critical and serves as the main means of communication for the system, as in the case when a system uses connectivity standards like USB or Firewire. Compliance and certification of these protocol standards in a consumer space is essential for public acceptance of the end product. It’s no wonder that the trade-offs for making versus buying SIP revolve around a balance of costs, developing, deploying, supporting, and retargeting SIP, and judgments about the state of the external SIP versus the ability to fund the development internally. Make-Versus-Buy Make versus buy is a difficult decision process in many businesses. In the SIP buying process, it often runs directly into discussion of basic core competencies, or what companies choose as differentiators for their products. In one case, a company’s expertise may lead it to believe that it should develop certain SIP internally and use it in a business competitive way. Each company approaches these choices differently, but there are some compelling trends that are leading, even the most hardened NIH (not invented here) companies, to rethink many of the choices made in the pre-SoC era. Market pressures are steering companies away from developing functions that can be replicated in multiple designs as competitive differentiators. Industry standards, like USB, or open proprietary standards, like the AMBA bus infrastructure, do not offer SoC chip makers much sustainable competitive advantage. So these companies instead look to minimize the cost and maximize productivity of using these standards when designing complicated systems. They opt for using more configurable, pre-verified SIP, the more the better. The market rewards early entrants, and early entry with a complicated system is a real challenge if you have to build and support every part. Here the semiconductor industry is taking a lesson from the auto and aerospace systems companies by practicing better management of their supply chains. They are partnering to get better and more flexible SIP so that they can improve their delivery and quality.
The economics of make versus buy, absent the differentiation, are compelling. As long as the industry has providers in these standards spaces, that are providing alternatives that meet the needs of the market, there is no reason to develop this SIP internally. But here is the challenge to the industry. How do you become a reliable supplier to companies who are looking to leverage the SIP industry advantage in their own designs? For them, quality stands front and center. And that is why each company in the SIP industry must address, and embrace, the total quality experience. Elements of a Total Quality Experience Quality is often easy to identify when it’s missing. Dissatisfaction with a product is an emotional response, usually unprompted. Complaints come unsolicited as a natural way to get a problem rectified. So producing quality involves delivering satisfaction across each experience the customer has with a product. In the case of SIP, as with many other products, this means delivering what the customer wants as part of the buying experience, deployment experience and service experience. In SIP, these map to methods used to acquire SIP: the purchase, experiences in using the SIP, placing it into a design, and how the SIP is serviced, if a problem appears. The Buying Experience The buying experience begins during SIP evaluation and ends with a purchase. During this period, a number of interactions occur that often create dissatisfaction in the customer. SIP comes in many flavors; the most sophisticated flavor is source form. Source SIP material is generally delivered in original form and must be protected to retain any value for the seller. Giving original material for evaluation is problematic from a legal and commercial perspective. Providing mechanisms for easy evaluation is essential. Synopsys improves this experience by offering tools that can configure and do unit testing using simulation without exposing the original source material. We also provide extensive technical documentation and presentation material. Finally, if the original material must be viewed or used, we can set up a separately controlled environment or a separate source evaluation procedure. All these options need to be explained to a customer up front so they can judge the time and energy commitment necessary to complete an evaluation. Once the decision is made, customers need the flexibility to purchase SIP in a number of forms depending on their economic situation. These include an individual use purchase, individual and derivative works purchases, and general access agreements. Some SIP requires royalty payments. Royalty limits or royalty buyouts might be required to fix the overall costs of the project SIP. Regardless of the methods of purchase, customers need a simple way to procure SIP without complicated recurring negotiations. All valuable SIP is controlled by an SIP license agreement. This outlines the limits of use and sharing, as well as the liabilities of misuse. Because of the complexities of these agreements, SIP frame licensing is a method of handling the legal conditions governing SIP across a number of different individual products. An SIP frame license grants customers the rights to purchase multiple and separate units by referencing one legal governance agreement. This makes the process of acquisition simpler, since the legal aspects of SIP acquisition are covered generally and then just referenced in any individual purchase. An SIP Access Agreement combines these two concepts. This type of agreement has the elements of easy licensing as well as easy procurement. The access agreement is a volume purchase limited in time or number of uses and typically controlled by a dollar commitment. In this type of agreement, a customer has access to a complete portfolio of SIP, and the legal aspects of acquiring the SIP are covered when the access agreement is signed and govern all the licensing aspects of any core acquired in the future. These agreements may be limited to selection on hand, however, they can be amended in a “re-spin” to get access to new SIP. The result is that customers have an easy way to acquire and use SIP eliminating costly purchasing and legal complications. The buying experience is made pleasant by careful attention to handling evaluations and coordinating purchasing. Careful planning and diligence here improves the quality experience of customers and goes a long way in setting the proper stage for the deployment of the SIP in a chip. The Deployment Experience The deployment experience begins with delivery of the SIP for initiation of a design project and ends when an end product is released. Many situations occur in this process that can lead to raising the dissatisfaction level of a customer. Deployment or a particular project includes the download of the SIP and unpacks it, continues through configuration and integration into the design, and includes setting up both unit test of the target configuration and system-level verification of the SIP. The first step, getting access to the SIP, usually involves downloading that SIP from a network site. During this process, the ownership passes from seller to buyer. Convenience and security are essential during this transfer. Keys are often used to provide security. After the encrypted source material is downloaded, the key serves to un-encrypt the SIP into source form. Once delivered and unpacked, the customer now has the responsibility to make sure the SIP is protected from misuse. Usually SIP is delivered to a particular project group, since they may well be the only group that should have access to the SIP. Increasingly, however, companies are setting up SIP directories, which give many design groups access to SIP. In these cases, customers receiving the SIP must make sure that the groups that have access to the SIP understand the rights and responsibilities of using the SIP. The next step in the process involves configuration. The construction and architecture of SIP must allow for some configuration for customizing to a particular application. How to properly build the SIP will be covered later. The issue here is to provide a convenient way for users to construct the configured SIP and do unit testing of that SIP to verify that the configuration is correct. Synopsys DesignWare IP is packaged with a tool called coreConsultant. This tool makes it easy to configure and test the newly purchased SIP. The coreConsultant tool controls the configuration options and constructs the unit test environment for validation. The tool leads the design team carefully through the process guiding the configuration so that problems are avoided. During deployment, it’s likely that questions will come up on how to configure, integrate and test a core. Gaining access to protocol expertise may also be required during this time. Setting up a clean support and service capability is essential to cost effectively deploying the SIP. Delays in schedule often occur because of misunderstandings on how the SIP is configured and tested. Further, questions often arise about how particular aspects of a standard protocol are handled. Synopsys’ DesignWare core support is right there, offering easy access to configuration and protocol knowledgeable engineers. Individually configured SIP is then integrated into complete systems. Here it’s important to understand the interface options. DesignWare IP is designed to link either directly to an AMBA bus or as a Peripheral Virtual Component Interface (PVCI). Adapting these interfaces for each system needs to be carefully managed. Good documentation and assembly tools provide the best foundation to make these interfaces bug-free in the final design. The Service Experience The service experience begins when any problem or issue arises and ends when it has been resolved. Even with effective unit test, sophisticated systems verification environments for complicated systems will uncover problems in SIP. Most interesting problems are found during system test, which is really controlled by the SIP buyer. Service organizations that focus on helping a customer design-in the SIP are really best positioned to service the ultimate goal of the SIP buyer, making a fully functioning product. SIP vendors have to know how the SIP is supposed to work in a system, they need to be able to isolate and determine workarounds if problems occur, and they need to make sure the SIP is getting through the SoC tool flow as effectively as possible. Response in the service experience is an important measure in satisfying the SIP buyer. Having an experienced support staff and enough of them to service the complete customer base will provide the smoothest path through any problem. Elements of Quality in DesignWare IP The DesignWare IP engineers at Synopsys are dedicated to building quality in all aspects of our SIP. In order to produce a quality SIP product, Synopsys leverages the extensive experience it has gained in developing SIP over nearly ten years. In fact, Synopsys engineers have coauthored three books on the subject of reuse (The Reuse Methodology Manuals). These materials are now standard methodology texts in the SIP area. No SIP company invests more in producing quality. Our experience in developing SIP can form the basis of a vendor audit that you can use to choose SIP. Producing a high quality SIP product is a journey, and it never stops. But to make that journey fruitful you need to be on the right path. The compass we use to measure our progress has eights points. Function, Use, Support, and Competence form the major compass points, supported by Methodology, Test, Interface, and Interoperability. Understanding how your SIP develops with these directions in mind will prove invaluable in making a satisfying choice in SIP.
Of these eight directions, function is where most of the aggravation on SIP quality comes into play. Bottom line – the SIP has to work in the systems it’s targeted for. If it doesn’t work, there is absolutely no advantage to using it, in fact, poor functioning SIP can negatively impact your entire chip project. Improving functional quality is evolving and in this area there is a lot of interplay between methods and tools to design and validate SIP. There is clear customer benefit from having the best methods and tools in place at the SIP provider. Savvy buyers probe in this area in much the same way as each of us manage our own person health. Here, as in general fitness, good habits or practices, are the best measure of long-term benefit. Prevention is the operative word. SIP suppliers must have access to tools licenses to perform an appropriate level of verification for the SIP they produce. For both hard PHY-level SIP and for synthesizable data link-level SIP, access to sufficient, if not copious copies of validation tools, like simulators are important parts of proving SIP quality. Access to these tools also allows you to run the state-of-the-art methods, like randomized testing. Best practices in this area include architectural review, code reviews, and two types of coding inspections, buddy review and extreme programming. In the buddy review, code is shared between two developers. The lead does the programming and passes it to a buddy. The buddy reviews each element of the code for functional correctness. Extreme programming is a method used with new developers. Here, an experienced developer mentors a new coder in an “over the shoulder” model. Of course the most common best practices, architectural and code reviews, build on the premise that building quality into the design is better, and more effective, than running extensive regressions and debugging. Buyers, beware of SIP constructed with shoddy or incomplete methods and tools. Again, stateof- the-art buying demands inspecting the methods of your suppliers. Build quality in and make sure your suppliers do this as well. Auditing your SIP developer’s process is the best way to determine whether these best practices are part of the design process. Functional Audit The following elements form a basic set of audit checks for functional quality. Exploring these items with your SIP provider should be part of your SIP buying process.
Assuring proper functionality begins with proper architecture. Here, understanding the systems where the SIP will be used is critical. Since reuse is an important benefit of SIP, architecture for reuse is a necessity to get maximum value out of the SIP you purchase. A systems perspective applied across multiple applications provides the best formula for constructing a flexible architecture. Find out how the SIP is architected. Inspect clock domains and check on clock boundary assumptions. Looking at the gate-level output of synthesis, logic depth, will give you a view to whether the SIP will achieve its timing constraints. A best practice here is an architectural specification which you can view. Look at a sample of an architectural specification from your SIP supplier. If they don’t have one, you will be hard-pressed to tell what design assumptions were overlooked. This will likely produce downstream problems when the SIP is used in a configuration not anticipated. The result is a real catastrophic problem, like a system hanging. And these problems can really upset your end customer. Methodology Audit Methodology was highlighted already as key to producing quality SIP. Just like in overall chip design, a well thought out methodology for building SIP gives you the best chance at producing superior measurable quality. Look at how your SIP supplier develops SIP. Do they do code reviews? Do they practice buddy or extreme programming techniques? How do they measure their progress against the product specification? Do they have an engineering specification for the products they are building? Without an engineering specification, products are developed ad-hoc. You have no formal way of reflecting that your architectural choices are being implemented without a document to check and measure against. Inspect the development processes of your potential SIP providers. Look for best practices that compare to your own methods. Don’t settle for SIP that is developed in weaker methods than your own. Verification Audit Verification is the obvious step to producing measurable quality metrics. Here, the key to superior quality lies in the test plan. Ask your suppliers to show you the test plan for the SIP you want to buy. Look at the techniques they use for functional validation. Best practices here involve function coverage metrics, supported by both directed and random testing. Assertions are also used both internally for validating operation, and externally for validating actual protocol correctness. SIP vendors should be tracking their progress against verification metrics. Ask your vendor for their function coverage metrics, their line and toggle coverage metrics. Ask them where they are in their verification plan. Not having a comprehensive verification plan and tracking against it will likely lead to major holes in SIP testing, and this will lead to bugs in your systems downstream. Has your SIP provider prototyped SIP in a design? This is the next audit question. Putting SIP in an actual design is an essential proof point for customers who want to see it working in an actual design as a confidence builder. But SIP developers get value from this prototype as well. Not only does is this an acid test of the SIP, where you are able to test power up sequences and resets that are difficult to capture in simulation, but prototypes give you an application platform for certification. Make sure your SIP provider is prototyping their SIP. Certification is an important step in SIP validation. The operational characteristics of most connectivity SIP is controlled by some standards body. For example, the USB-IF administers the testing and logo certification of USB product. The logo is then displayed prominently on the product packaging so that consumers can be confident that the product will interoperate with other products at the other end of a USB cable. Ask to see the certification reports. These will show that the SIP at least meets minimum performance. Use Audit Even SIP that is of high functional quality may be lacking in important aspects that make the SIP usable in a design. Inspect and determine whether your SIP vendor provides the documentation and support that will make your use experience satisfying. Complete documentation can save an incredible amount of engineering time. You should review the documentation on any piece of SIP you intend to buy. This should be more than just a tableof- contents review. Ask yourself some basic questions about the SIP you’re considering, and then test whether the documentation provides answers. Is a topic covered, completely, clearly, in language that will not be misunderstood by different members of your design team? SIP support is measured by the success customers have in designing-in the SIP into their chips. Delivering this support involves theory of operation or technical assistance and involves:
Suspected defect support or hot fix support is the next level and includes:
Finally, flow support includes:
Integration Audit In order to successfully deploy SIP on your SoC, your SIP vendor must provide all essential elements for integration. Evaluating SIP up front for breadth and depth of support for integration will prevent costly slips to your product release timetable. You should investigate in detail the following important elements of integration:
The application side of SIP should interface to a standard bus fabric like AMBA™. This type of common interface can be augmented by an interface like PVCI, which allows you to interface to any proprietary bus. On the other side of digital SIP, you should see a PHY interface like ULPI or PIPE. On the PHY pin side of hard SIP, you should see support for pads structures capable of supporting the ESD requirements of your product. Both the digital SIP and mixed-signal SIP have to be configurable (or available in a number of configurations for hard SIP). This is one of the reuse benefits of properly architected SIP. These options allow you to target the SIP for your end application. This targeting is a form of optimization of gate count versus function and performance. Choices here will save you silicon area and power, and match the performance requirements desired. Hard mixed-signal SIP should also be available in different forms. Configurations of x1, x2, x4, x8 and above lanes for PCI Express® are essential to achieve the bandwidth and speed required for different applications. Look carefully at the configuration choices of digital and mixed-signal SIP, especially if you intend to reuse it in future designs. End product requirements often change, and you will need to have other configurations available to meet those changing market requirements. Finally your integration audit should include looking at SIP assembly and verification. No piece of SIP sits on its own. SIP is part of a more complicated system. Productive assembly requires SIP that has built-in design intent. This intent takes the form of connectivity and verification rules. Your SIP vendor should be able to provide both the tools to assemble multiple pieces of SIP and SIP which has the rules necessary to make sure that you connect and validate the SIP correctly. In Synopsys’ case, coreAssembler is the tool that takes care of SIP subsystem assembly. In Synopsys' case, the coreAssembler tool (Figure 3) reads the connectivity intent, and after configuration choices are made, makes sure that all the components are correctly connected. It then can drive synthesis and verification. This type of capability eliminates many simple, yet costly problems of integrating SIP together. These problems have measurable impact on design cycles. Companies using tools like coreAssembler have seen 60+% improvements in their design processes when using SIP. And the Synopsys set of coreTools allows you to develop internal SIP and integrate other third-party SIP.
Figure 3: Synopsys coreAssembler Tool Interoperation Audit Interoperability is essential in a connectivity standard. The metric for this is certification or plugfest participation. Before you purchase SIP, you should collect and review all the certification material on the SIP. Look at the test results. Investigate the platform or product used to obtain certification. See if the SIP vendor has a lab where they do internal testing. Ask about the policy of re-certifying if problems are found or new versions created. Competence Audit Participation in standards, contribution to the standard, helping to define proper SIP interface for the promotion of industry acceptance of the standard, is another important measure of whether you are getting what you need form your SIP provider. You should see active participation from your vendors in the areas they are providing product. Having a vendor stay current on these standards is a major advantage to an SIP buyer. As standards change and evolve, it is an additional expense to have to track these changes. SIP vendors that are participating can warn of changes coming if they have built up the practice of keeping their customers informed. This will free you from having to closely monitor many different types of standards. Further, look at whether your SIP vendors are promoting SIP use within the users groups. The best example of this is defining interfaces which promote the re-use of Open Systems International (OSI) model elements of the protocol, like separate data link and PHY definitions. Interface specifications like UTMI and PSIPE have helped develop industry ecosystems that encourage reuse and cooperation among vendors supplying standards-based products. Have your vendor articulate their involvement in the standards they support. Look for active participation. Look for real contributions. Total Quality So you’ve found the perfect vendor for SIP. But now you have to make sure your vendor is a trusted supplier. Why? Even vendors who follow the best practices in providing quality face market pressures to get new SIP to their customers on time. Market adoption of new standards SIP is often difficult to predict. Standards advance, like moving from a parallel to serial bus architecture, as in the case of PCI to PCI Express. As they do, new standards require early adopters and systems integrators require proof of viability. An ecosystem is typically an essential element in market adoption of a successful standard. This means those systems producers, chip producers, and SIP providers all need to cooperate, define, build, and validate. Two addition elements are required to make the deployment of SIP successful in new markets: partnership and transparency. An SIP partnership is an essential ingredient to get what you need when you need it. It can be thought of as a type of co-development. Partnerships can take the form of close co-operations, where engineering plans are shared, or loose alignments, such as when you want to validate operation between blocks as in interoperability testing. In the former case, suppliers want to become part of the building and testing of new SIP. In a way, the SIP provider becomes part of the buyers own development team. All the essential relationship elements of any successful partnership play in these engagements, and expectations need to be managed. But well planned alignments are becoming common as component SIP builds in complexity and delivery to hit a market window overrides other concerns, like NIH. An example of a loose alignment is the case where partners agree on a common platform (board) often with a daughter card, for testing and certification. In this loose cooperation, vendors demonstrate to system OEMs that they can work together to provide a complete solution, and that a multi-vendor approach will not prevent end customer products from working together. The second element, transparency, applies both to new SIP deployment, and older SIP debugging. Again, in the former case, because markets wait for no one, it may be necessary to send designs to tape-out early. In this case, complete transparency is required so that all the risks are understood. In the latter case, if a serious problem does occur. The SIP vendor must share all the details of how problems will be corrected. Summary Best practices in producing high-quality SIP are only valuable if you evaluate your suppliers against the clear quality metrics you value. Use the audits presented here to determine the level of quality you will get from your suppliers. In the end, you want more than a buyer-beware relationship with your SIP supplier. The monetary risks are too high to leave to chance. Systematic evaluation of your SIP provider will go a long way in making sure that you get what you want, and the industry overall will benefit. Choosing quality suppliers over those that are not investing in quality will eventually move the entire SIP industry forward, providing a new level of quality experience for the entire industry. For more information on Synopsys DesignWare IP, visit www.synopsys.com/designware.
|
|
||||||||
|
|
|
|
|
Home | Feedback | Register | Site Map |
|
|
|
All material on this site Copyright © 2017 Design And Reuse S.A. All rights reserved. |
