Open side-bar Menu
 IP Showcase
Peggy Aycinena
Peggy Aycinena
Peggy Aycinena is a contributing editor for EDACafe.Com

Lanza Panel at DAC: Democratization is not Open Source

 
July 14th, 2016 by Peggy Aycinena


Long-time EDA investor Lucio Lanza lead a fascinating, albeit mystifying, discussion
in the DAC Pavilion on Monday, July 6th, in Austin. His panelists included IPextreme’s Warren Savage, Scientific Ventures’ Mark Templeton, and eFabless’ Michael Wishart, with the topic under discussion being open source.

The session was titled “Daring to Move to Open Source” and was described thusly: “The emerging Internet of Things market is destined to upend that time-tested ‘advanced-node’ model, as developers opt for older, less costly process technologies, using commodity design tools and selecting proven IP blocks to quickly and efficiently assemble chips. As demand for IoT devices grows exponentially, might open source EDA tools and IP become viable, or even the winning combination that enables the low-cost design of an IoT SoC?”

Here are some soundbites from the panelists.

* Warren Savage: It’s about the cover, not the content. There is motivation in the design community [for open source], but there is no road map for a ‘GNU’ set of IC design flows. Principally because patents are a problem.

[conclusion: No to Open Source]

* Mark Templeton: When we started Artisan 20 years ago, TSMC would send – without an NDA – 20 pages of design rules. Today if you call TSMC, it takes a year to get the design rules. The slowing of Moore’s Law means [existing] nodes will stay stable and people will want to optimize these nodes for the IoT edge. [In that environment] standard flows will not allow for open source. Furthermore, as markets fragment, change will happen, but we will not cannibalize from EDA.

[conclusion: No to Open Source]

* Lucio Lanza: The traditional market will not go away at all. Now it will be virtual IDMs, and [a market] for system tools that will help create 5-to-10 prototypes.

[conclusion: Unclear]

* Michael Wishart: Our manufacturing partner is XFAB who has worked through things with us, and understands that we control the underlying technology on our site. Our plan is not to serve those already being served. We want to serve new communities without connections to the established electronics community. Competing with the EDA industry is not our goal. Creating new markets is our goal.

[conclusion: No to Open Source]

******************

Lanza DAC 2016Interleaved within these statements, the Pavilion Panel audience heard a lot of ringing endorsements of the Maker Community – praise for their pluck, innovation and courage.

Absolutely admirable sentiments, of course, but not really relevant to Open Source. Except that the Maker Community can’t afford expensive tools. But cheap tools are not the same as Open Source tools. Neither is democratization the same as Open Source.

Open Source is about open source code that’s made freely available, that can be freely modified, stabilized, and evolved to meet a particular need. Stuff that can be commercialized upon, with a requisite nod to the developer community and perhaps even some ink for a GNU license or two.

Clearly, I should have known better than to believe anything within spitting distance of DAC – let alone at the epicenter of the Exhibit Hall – was actually going to “dare to move to open source”.

[conclusion: Disappointing, somewhat aggravating]

******************

Tags: , , , , , , , , , ,

One Response to “Lanza Panel at DAC: Democratization is not Open Source”

  1. Avatar Kevin Cameron says:

    The original GNU project started off with just the the C compiler (gcc), an open-source EDA movement would need something similar, and I would suggest Xyce is that –

    https://xyce.sandia.gov/

    It allows people to agree on reference models in a portable way, and probably out-performs most commercial SPICE simulators, and is a usable base for mixed-signal simulation (with Icarus, QUCS etc.) for applications like IoT.

    The rest of the EDA stack is somewhat clunky and RTL is the wrong level to design at. A move to software-defined-hardware seems like what should be coming next, and I have an open-source plan for that –

    http://parallel.cc

    – which is going to fit with programming the new crop of CNN/AI hardware.

Logged in as . Log out »

Verific: SystemVerilog & VHDL Parsers



© 2024 Internet Business Systems, Inc.
670 Aberdeen Way, Milpitas, CA 95035
+1 (408) 882-6554 — Contact Us, or visit our other sites:
TechJobsCafe - Technical Jobs and Resumes EDACafe - Electronic Design Automation GISCafe - Geographical Information Services  MCADCafe - Mechanical Design and Engineering ShareCG - Share Computer Graphic (CG) Animation, 3D Art and 3D Models
  Privacy PolicyAdvertise