The Stickiness Factor and Programming Models

• A programming model, according to one definition, is an abstract conceptual view of the structure and operations of a computer system. The words abstract and conceptual are key here. The whole premise of rapid development of complex chips using ESL techniques is highly dependent on the availability of good programming models that can be use to test, verify and debug applications software before the hardware is manufactured and put out in the market. Programming models today satisfy some of the needs that reference boards have met over time. Reference boards in the embedded market are used to boot operating systems, test applications and network interfaces among other things. In short, reference boards serve to propagate use of a particular architecture among potential users. Since the reference boards often come bundled with tools, the existence of the development ecosystem helps companies to deploy products faster. In theory, shipping a product that has a good tool and library ecosystem is a guarantee that the product will see the light of day.
  
  
  Think of the programming model as the new reference board. Certainly it has limitations such as the lack of peripherals and cables hanging off various hosts. But at the end of the day, having a good, verified model of the hardware helps the virtual test and debug of various software applications that the device will run in the real world without the cost of having to manufacture a non working part. However high level models are labor intensive and therefore cost intensive and users are unwilling to pay for them because they are seen as a sunk cost for the semiconductor manufacturer. From a model creation vendors’ perspective, it often ends up being a services business rather than a product business and therefore it is hard to grow the company.
  
  
  

  The Stickiness Factor

  The stickiness factor is similar to brand loyalties in consumer markets. How does a company go beyond the early adopters and reach the mainstream market? As the market evolves to a scenario where hardware designs need more design flexibility to adapt to various markets and support multiple standards, there is a big potential payoff in keeping customers from migrating to other platforms. This is where the stickiness factor comes in. The company may have come out with the best in class chip design possible using the latest tools and technologies and incorporating every leading-edge design method and advanced algorithm possible. But if the platform cannot be programmed, in this instance read lack of good high-level models, the deployment will be limited to a very small number of users. For vendors wanting to enter the mainstream markets, this will become a serious stumbling block. Providing good, verified models that are a faithful description of the hardware constructs will help to create a stickiness factor and keep users from migrating elsewhere.
  
  Thankfully, many vendors have seen the wisdom of making sure that models are available to go hand in hand with the hardware platform and either through internal investment or through acquisition have developed modeling generation capabilities. Texas Instruments was one of the first companies to understand this trend. In the mid 1990s, they acquired an RTOS company and a tools company. These acquisitions became the foundation of their software platform that ships in various configurations with the TI chips. Additionally they developed their partner ecosystem rapidly to support their current and upcoming platforms. The result is that TI has an enviable position in the wireless chip market that is a significant barrier to entry for any newcomer.
  
  
  Over the past few years, IP companies have entered into similar transactions to help develop their tool ecosystem. ARC for example has been acquiring IP, RTOS and tools companies to help support its platform and improve its stickiness factor. The list of acquisitions includes Precise/MQX, VAutomation, Teja Technologies and now Tenison Technology.
  
  
  Figure 1 below shows a few of the transactions that have taken place over the past few years.
  
  

  Vendor	Acquisition	Year
  ARM	AXYS Design Automation	2004
  Synopsys	Virtio	2006
  ARC	Tension Technology	2007

  
  Source: Gary Smith EDA (July 2007)
  
  A good hardware platform needs a good supporting software platform, programming models, RTOS, protocols and tools as part of the ecosystem. In some ways, the move to bring the platforms in house reflects a return to the 1980s when semiconductor companies provided soup to nuts solutions.
  
  
  Daya Nadamuni