Semiconductor IP News and Trends Blog
Semiconductor IP faces API-ification and Wicked Problems
Service oriented APIs and user experience-based designs are affecting traditional software design approaches. Will IP developers face the same challenges?
Here are a few follow-on thoughts to my fanciful story about the evolving relationship between hardware, software and innovation; Does Innovation Lie Beyond Software?
Two recent articles add a new perspective to my observation that software is moving toward a new trajectory – and not simply due to the commoditization of chip and board hardware. The first article deals with the API-ification of software development while the second one ponders the need for classical mathematics in software engineering:
What I find interesting about the API-ification of software development is that the rapid growth and dependence on APIs seems to represent a transition from objects and class libraries to service oriented (cloud-based) architectures. Most of us can remember the shift from fully loaded programs to a reliance on libraries and DLLs in software design from the last decade, e.g., Hummingbird libs for TCP/IP and PowerBuilder for user interfaces. API-ification extends that idea to service oriented applets in the cloud (or at least on a networked server).
What does this have to do with semiconductor IP design? Simply put, IP designers eventually take their development cues from the world of software application development. Hence a trend in the later sooner or later becomes a trend in the former.
Which brings me to my second story which suggests that classical math is not really needed for today’s application developers. The argument runs like this:
“It just doesn’t take as much math to write an operating system as it does to design a printed circuit board. Programming is rigidly structured and, at the same time, an evolving art form—neither of which is especially amenable to mathematical analysis.”
The question, then, is whether classical mathematics remains relevant for semiconductor IP developers. One would imagine so, since IP designers are closer to the hardware than software application engineers. But is this really the case? Today’s electronic engineers (the “hardware” guys) use EDA-based register-level and even higher level ESL tools to create chips. These tools are far removed for actual transistor-level design.
What does this trend portent? Will the engineering discipline of hardware design follow the commoditization trend of the end-product? If so, then chip design will become increasingly automated and further abstracted from the transistor building blocks. In that case, only a few architectural gurus and device physicists will be needed. Everyone else will essentially be a software designer.
Finally, the mathematical article touches on a new class of “wicked” problems that add social, political, economics and under-fined issues to existing technical challenges. All of these former factors are not easily modeled by mathematical equations. Some engineers are already experiencing the wicked class of problems as they begin to deal with user-designed product development:
The trend toward changes in hardware and software design continues. Will these changes be spiral in nature (cycling outward) or will they represent fundamental changes in developing semiconductor-based systems? We’ll find out together.