While dodging the employment zombies wandering the exhibits at DAC 2012 ("Jobs! Jobs!"), I ran across an old friend, Gunnar Scholl, at the booth of a brand new company, ProFPGA.
Gunnar was the marketing director at ProDesign a few years ago. He then joined Synopsys when it acquired the ChipIt automated FPGA-based prototyping system from ProDesign.
Gunnar returned to ProDesign late last year to help it spin off ProFPGA and its modular approach to design FPGA prototypes… which could probably fit in well with the Synopsys-Synplicity ChipIt flow! How about that?
I stopped to have a chat with Gunnar and get the skinny on the new company, which -- as of this writing -- doesn't even have a fleshed out Website, so this video may be the most information you'll find about this new company for a while. Take a look at the video and then I'll pose a couple of questions:
OK, so here's what I want to know: Gunnar says the ProFPGA has the capability of mixing and matching FPGAs from Xilinx and Altera, which I thought might be a great way of performing some real benchmarking. But then I wondered, other than benchmarking, why might one wish to mix FPGAs from these two companies?
I can imagine a number of reasons a system architect or design engineer might choose one FPGA vendor over another, including cost, capacity, performance, and application. But would there be some advantages to having a truly agnostic system foundation? Conversely, what might make something like this turn into a nightmare?
Interesting comment speaking as a designer there is little difference (well there is in resources, toolchains etc) but the key thing for an engineer is the understanding of principles which do not change depending upon the target device. Most engineers should be able to learn the tool chains pretty quickly and implement effective designs.
Such could reduce supply chain issues (depending on a single supplier increases risk), increase FPGA vendor innovation (potentially attempting to maintain lock-in by providing exclusive features that are not handled well by such tools but also freeing the vendors to try things that might break less flexible tools), and increase ease of migration from FPGA to gate-ASIC or 'true' ASIC.
Reducing transaction costs (and cost of change might be considered a transaction cost) tends to increase economic activity.
It is possible that such might encourage greater similarity among FPGA products in certain basic aspects (by removing the lock-in advantage in those aspects--though there may be patent barriers [and some differences may have practical implications and not merely be arbitrary choices reinforced by historical weight]) and move distinguishing features into more practical aspects.
The nightmare scenario (IMHO) is when supply chain managers get wind of "agnostic hardware design tools" they will use it as yet another way to drive engineers crazy with cost reduction plans that don't include upfront engineering costs.