IP asset sales in the Northwest

chaos watch mike suns kannikar paired referencing importance vicinity prosperity elevation

Standardization certainly has everything to do with the comments made on arcane languages and environments like Linda and Hypercube. You can make all the claims you want about "easing the path to parallelism," but if you require folks to learn a new programming language, the world is unlikely to beat a path to your door. This was very common in the network-processor world...

Standards create markets. If a standard is done properly, and stops defining things before it commoditizes the market, it creates a very large market, consumers are more trusting of standards-based products (as opposed to some slimy proprietary junk), and the market grows. The FPGA market, because of it's disdain for standards, is only 1/100th the size it could be, because there are no fundamental standards in place. THAT is what standards have to do with anything...

What does standardization have to do with anything? If you all haven't learned the lesson yet: Standardization leads to Commoditization.

In the immortal words of WJC, "I feel your pain" (said as he was feeling his pleasure with WH interns). Unless some influential editor begins to call for standards in the FPGA, tools, and core-IP markets, it is destined to follow the path blazed by things like the Hypercube, Linda, and the "Pet Rock". May the Force be with you...

The problem is getting bigger than that. Given VCs' mass migration from anything in the semiconductor, EDA, or even systems electronics industries in favor of cleantech/greentech, there are plenty of startups out there with good ASSP ideas who failed to get their mezzanine or third round of financing, and had little choice but to shut their doors. I'm worried about the number of companies that may be added to the list.

And this is a surprise? FPGA guys ignore standards,,especially on core interconnects. FPGA tools are squirrelly. Add to that (in one case above) that Amdahl's law shows how much embedded code can possibly be parallelized (very little), and look at the host of multicore processors looking for an application out there as further evidence, and you have a recipe straight out of the Donner Party Cookbook. Anyone chasing proprietary FPGA architectures and tools, in combination with parallel processing, is going to be eaten alive....