Hacker News Reader

Comment by nickpsecurity | Hacker News Reader

This isn't good news. One of my ideas for open, secure hardware involved a partial or whole acquisition of Mentor by a company that was pro-FOSS. Then, the product lines would be tweaked for easy verification. Plus deals for hardware startups that got them the tools cheap to free early on with guaranteed revenue for Mentor if successful.

Now, one of the Big 3 of EDA is in the hands of a large corporation far from our interests. This kind of thing happens a lot. It's to be expected. Oh well.

View on HN

Used to work at Siemens. Their open-source policies were onerous from a consumption point-of-view. Every revision of every approved open-source product had to be reviewed by their open-source legal hawks to ensure that nothing had changed. I don't remember their mandatory training on open source but I especially don't recall any encouragement to publish open source projects and to contribute to them. Great company but it certainly treated open-source with a sentiment that I can only describe as fear.

That's definitely bad. One of my ideas for Mentor didn't even need open-source, though. The idea was that they enabled export of some intermediate steps in RTL synthesis or optimization so customers could equivalence check them. I don't use their stuff so forgive me if they already do this & I just didn't know. Anyway, the initial, intermediate, and final RTL's could be checked both with testing and formal tools that were open-source. All the magic behind making those steps happen can remain proprietary. Likewise, do some OSS stuff for gate-level testing and material checks to run after Mentor's. Analog stuff done similarly as there's already diverse proprietary and FOSS tools for SPICE, etc.

So, they can differentiate on verifiability while still keeping the trade secrets that make them the most money. There is risk that the intermediate steps could let people reverse engineer some of the magic. However, that's low given people can already analyze their binaries when running those steps and academics [like Mentor funds] come up with better ones all the time. The next step in my Mentor-oriented plan was same buyer acquiring eASIC so the verified, hardware designs could be practically thrown into initial production at little cost. Key IP that almost everyone needs they get for free so long as they use eASIC for the HW design for X years or iterations with a fair, royalty agreement.

Interested in what people in hardware market think of these schemes?

I worked at another one of the similar field and the policies were quite similar.

I agree. It's disappointing to see a news like this. Mentor Graphics did a really great job on their verification tutorials that had done a great benefit to my career. I'm not sure Synopsys is winning the competition though. I just hope some great startup would emerge that's willing to support FOSS or rather some group of people willing to put their effort in creating the next generation and yet competitive and free EDA tools that I'll be interested to join.

Interested in your opinion on this:

https://news.ycombinator.com/item?id=12951128

Why would this happen given that semiconductor industry itself is shrinking? I used to work in EDA until recently, but since money in semiconductor is dwindling, people like me have no interest in pursuing ideas in EDA

The other problem that I noticed is buyers are extremely risk-adverse about synthesis tools. They dont want to pay that mask fee to get broken chips in return. No more than necessary anyway. They were picky even about offerings from Big 3 when I last looked where behavioral was almost always Design Compiler, RTL almost always same product, and most using Mentor for material check. Id see different mixes on occasion but most stuck with what seemed to work for quite a few people.

I think an EDA startup doing these things might have to prove itself with industrial-grade, academic projects first just to get some proof it works. Then, it will get a small slice of a shrinking pie that requires solving a pile of NP-hard problems in a patent minefield. I'm surprised at how many get funded to begin with.

Agree with your sentiment, but wouldn't Cadence or Altium be better candidates than Mentor for that goal?

There's pretty good HLS tools and other things to work at a high level. The real difficulty is on verification, RTL synthesis, material checks, fast simulation, etc. Altium lacks a bit of these whereas Cadence is huge. That Mentor was so tiny in market cap despite having something good for every part of the stack, even software side, made them the best buy.

I have occasionally recommended Altium to people designing new hardware just because they have a nice cost/benefit ratio. So, did Tanner on PCB's and analog. Mentor bought Tanner. Now we're back to Mentor. :)

That makes sense. Rightly or wrongly, I perceive Mentor as having lots of users who cater to companies in industries like telecom, medical, or A&D who aren't very pro-FOSS (on a corporate level, at least).

It would be cool to see them open up access to their tools somehow though; Library Manager, Valor DFM, maybe even something like HyperLynx. Not sure how open people would be to some sort of royalty agreement in exchange for access to CAD tools.

That last part will need some market research. eASIC already seems to do it for silicon. Quite a few vendors in embedded do for their side. There's at least potential HW upstart would want a pile of tooling at zero cost in exchange for a percentage of operating profits they get from it.

Definitely true, seems like it would depend pretty heavily on what's being made & at what scale (as if that isn't true for all mfg haha). Something that takes a few PCBs and needs bare boards, assembly, system integration, packaging, shipping, etc etc, could make a percentage of profits challenging. Makes sense to me for embedded because they're after platform adoption as well.

Reply on news.ycombinator.com