In part 1, we tackled the concept of IP as a partnership between the IP supplier and the user. In the second installment, we looked at how sophisticated the industry has become and if it is now too complex for start-ups.
Taking part in this discussion are: Mike Gianfagna, vice president of Corporate Marketing, Atrenta; Warren Savage, president and CEO of IPextreme; John Koeter, vice president of Marketing, Solutions Group, Synopsys; and Chris Rowen, Cadence Fellow and CTO, Tensilica.
EE Times: IP used to be fairly simple blocks, but today there appears to be a trend for larger IP sub-systems becoming the norm. While some IP suppliers are shipping the complete sub-system, are we perhaps going to see a tiered approach in the industry where a company may integrate several smaller pieces of IP together and then resell that as a sub-system?
Mike Gianfagna: On the engineering side of it, it's a matter of who is smart enough to do it. What applications do you want to focus on? I think that is absolutely a commercial opportunity, and it is something we are going to see more of.
Warren Savage: I think it's more challenging than it seems. What I am seeing is verticalisation. Verticalisation enables specialists in areas, such as automotive. It often requires special certification and special expertise in those areas. You may see a consolidator for a sub-system in particular markets, whether it be medical, automotive. It requires that there are enough specialties in those areas that it's not horizontal.
Chris Rowen: I see two distinct flavours of opportunity. One, there are more vertical applications that have reached a critical mass in terms of popularity and ubiquity—there is room for IP sub-systems to serve those markets. As systems grows, there will be more of these niches where an individual core is too small, and somebody's going to want a complete 802.11ac base band, or a complete ZigBee base band, or a complete vision augmented reality sub-system.
The other flavour is the set of IP, tools, and methodologies needed to integrate sub-systems together. The number of sub-systems will soon outstrip the number of individual IP cores we used to integrate. The integration of sub-systems is different qualitatively from the integration of blocks. The amount of software and application content in a sub-system is much greater, and it takes on a whole new character.
Gianfagna: I think there's a subtext here around standard interfaces and standard architectures. To the extent that those kinds of things flourish, you will have more of this. Without standard interface and standard architecture, you'll have a hard time with these sub-systems.
EE Times: Are we talking about hardware or software standards?
Rowen: Today, it's kind of a mess. Certainly, there is a fair amount of knowledge about things such as Android. The software interfaces, the knowledge of the software interfaces, the configurability of the software interfaces within Android will become central concerns. But the way others do it is different. I think there are very significant challenges ahead as we pop up to that next level of interfaces. With hardware interfaces we have known for a long time that they were important, and there's been a lot of work there.
Gianfagna: Standards make us all say, that's going to take forever, but actually, there's been progress.
Rowen: Some of them are de facto standards, such as AXI and good verification IP to make it easy to stitch it all together. IO interface standardisation is key. So the hardware piece is in okay shape. But moving up from there, it gets tougher.
Gianfagna: We all know that it's the hardware and software bundled together that constitutes delivery that people want. So there's an interesting impediment to progress.
John Koeter: The biggest thing with software is to start it earlier. That's where virtualisation technologies and prototyping technologies are so important. The waterfall method that some people did: chip then software—it's just not viable.
Savage: Everybody's going Agile, on the software side and even on the hardware side.
Does the integration of larger sub-systems make your job easier or more complex? How are people getting on with software integration?