0.000000006 feet under
What does the firing of Gary Smith bode for the EDA industry? Not too good it would seem.
I had commented on the EDA industry in my earlier posts. What looked positively heartening to me was that venture funding in the EDA industry looked to be going up.
Although it can be argued, that the move reflects the loss of funding by two of the biggest EDA vendors to a technical magazine, I hardly see this phenomenon being repeated in, say application software. To digest the fact that the industry is given as being representative of two or three companies, is quite difficult.
Joe Costello’s keynote was an interesting (although high caffeinated) call-to-arms for the EDA industry. After all the managerese, what emerged from the discussion is this - the EDA industry has to constantly re-tread its sales and marketing channels. Its not like the database industry, where I can go to a company, tell them to move all their data to my database and have it work seamless. No. Instead, what I do is have to go and sell a whole new and different way of doing the same thing (although maybe faster and better). I have to teach the end-designer a whole new set of TCL commands, deal with newer and newer storage formats and different ways of dumping reports, etc. Creating these sales channels is much, much more difficult for a startup than a mega-company. That is curtailing the flow of innovation which has worked very well for other sectors.
The technical talent in EDA is, in my opinion, greater and more focussed than in most other aspects of software engineering. Because not only do we have to solve problems of huge datasets, we have to solve them fast and on existing hardware. The graph theory that EDA deals with has proven to be highly resistant to distributed computing and parallelization, which brings in the need to optimize/tweak/twist every aspect of the code to make it run better and better, rather than just throwing more machines at it .
The startup cost for EDA - what is it? We get paid the same salaries as a Web 2.0 employee, we dont need humongous server farms or co-located storage (initially atleast). And the people who come to the table have a history of being mathematically and technically attuned to the task at hand. What gives then?
Sales channels - that is the problem. And the solution is partly what Cadence is pushing so hard for. All its open formats - OpenAccess, PFI, ECSM are beginning to cover the entire gamut of the design cycle. If these formats get widespread use, then it will be much easier for a startup to sneak upon the same sales channels (assuming there is also a common TCL interface - instead of the thousand commands you have to relearn each time).
If you were a big company, would you make it easier for young upstarts to sneak up on you? No. But then what is in it for a big company to open source these formats?