When Open Source goes hardware

THREE MONTHS ago, an announcement by Minneapolis based U.S. semiconductor firm Silicore went almost unnoticed. But it was a ground-breaking move, as it announced the company's adoption of "an open source business model for semiconductor IP".

To put its money where its mouth is, the second move in this direction was the release of its SLC1657 microcontroller core under the GNU Lesser General Public License (LGPL). Both source code and related documentation are available for download from the Silicore website.

For anyone wondering, this is an eight bit RISC processor that according to the company is used on "sensors, medical devices and consumer electronics". Under the LGPL, Silicore "retains the copyright but allows any integrator to copy, modify, reuse and distribute the core without cost". And "If the integrator modifies the core it is obliged to share the resulting (derivative) core under the same licensing conditions" - according to the company's press release.

Wade Peterson, Silicore President and CEO said "Our basic conclusion that soft semiconductor IP is just plain old software. This might be a radical notion to some, but we feel that if it looks like a duck, quacks like a duck and walks like a duck, then it probably is a duck. Software is just another word for machine instructions, and semiconductor IP is just that... instructions for FPGA and ASIC machines. That means we can borrow the licensing and business models from the GNU/Linux community for use in semiconductor IP. We don't need to reinvent any of those wheels."

Now that Microsoft is supposedly getting into CPU design - for its Xbox - I can only wonder if the open source battle is going to move, in the long run, from software into hardware as well, turning this into the ultimate battle for world domination, between closed and open systems. This time, hardware and software.

Peterson said: "We have been studying the semiconductor market for some time now, and have concluded that the best way to create and distribute soft IP is under an open source business model. It's the same model that's used by Red Hat and other GNU/Linux distributors. Until now we've operated under a traditional software business model that's been used by companies like Microsoft and Oracle. Under that approach we viewed IP as a product, but now we see it as a service. We believe that System-on-Chip [SoC] integrators will prefer this model because it lowers costs, simplifies licensing, reduces parts obsolescence, improves security and allows them more control over their system level IP."

Mr. Peterson didn't save words about his current views of the semiconductor market: "As it stands today the semiconductor IP market resembles the computer industry in the early 1980s. That market was fragmented into a small number of sellers markets, with each controlled by a mainframe or minicomputer manufacturer. That meant that you had to sell your equipment or software into a niche market that was controlled by your competitor. We're seeing the same trend in System-on-Chip, only this time around the niches are controlled by the semiconductor manufacturers and tool makers...both of which are heavily investing in IP. We believe this trend is bad for the system integrators...or more accurately those who are paying for the system integration. Right now they're overpaying for their chips and IP because they have to buy everything in these closed markets. They want open, buyer's markets where sellers compete belly-to-belly on price, quality and service. It's the same reasoning that produced the desktop PC and the microcomputer bus markets like VMEbus and PCI. Those were successful because the system integrators preferred them...and that's important because they're the ones ultimately paying for everything."

The Road Ahead?
Almost three years ago I wrote that, in my opinion, one of the main benefits of open source software is that it allows software to "kill time and live forever". In other words, it removes the time variable from the success equation, and its associated uncertainty over ownership and its owner's fortunes.

Open Source projects cannot be "killed" by killing the "owner", since there is none. Yes, there might be a corporate "sponsor" or initiator, but once the code is out there in the wild, it's up for grabs for anyone else to pick and continue enhancing and maintaining it, if there is interest. And if not, the project can be kept "sleeping" until someone, some day, in some dark corner of the planet, discovers it and brings it back to life, just because someone finds the work already done interesting to use as a base for his work.

Just imagine for a second what would happen if chips like IDE-to-USB bridges, WIFI controllers, and more complex semiconductors were "open source". And the holy grail... what about an open source CPU? I'm not advocating this is the solution to all the problems in the semiconductor business, just thinking "what if".

There are more questions than answers, but I applaud Silicore's bold decision to travel this unpaved road heading for the great unknown. Is there a place for "open source hardware" in the marketplace? Let me know. µ

L'INQS
The OpenCores.org project The OpenRISC architecture