Thoughts on IC development, EDA, and hardware design languages.

Industry Insights: EDA Development Needs a 10X Productivity Improvement

Posted by: Brad Quinton | Posted on: May 13th, 2015 | 4 Comments

Share on LinkedIn57Share on Facebook0Tweet about this on Twitter0Share on Google+0Email this to someone

How long should it take to develop an EDA application? 2 years? 3 years? 4 years? You will rarely hear the answer in weeks or days, but that is exactly where it should be. Why? Because that is the current economic reality of software development. In EDA today, we seem to be slow to accept the new reality. As a result the software development world is passing us by. Hot-shot developers whip-off $100M SaaS businesses in a few months and we toil for years optimizing our pointer allocations in dark rooms….

We are always hearing “doom and gloom” about the state of the startup EDA industry (Problems Ahead for EDA ): “There is no VC money” , “there is no angel money”, “there are no new start-ups…” ( The Road to Success is Long for EDA Startups). But the lack of investment is, for the most part, completely justified. Given the size of the market and current methodologies, it just doesn’t pay to produce EDA tools. To make matters worse, it takes way too long to develop products to that point where you can get them in front of potential customers to find out if they are actually interested in buying. Investors are smart to put their dollars in places where companies move fast and fail fast, allowing them to start again and find out what their target customers really want. This is no longer pie-in-the-sky startup guru stuff. This is the accepted startup methodology that was clearly articulated 10+ years ago (by people such as Steve Blank, for instance).

Want ROI? Change The Denominator!

We have to start thinking differently in the EDA world. If we want investment, we have to build a model where there is a clear and attainable ROI. So what do we do? Well, we can sit around and hope that the semiconductor companies will pay more for their EDA tools. History and hard-won experience shows us that they won’t. So, if we can’t change the “R” we need to reduce the “I”. And that means adopting rapid, efficient development methodologies to get stuff done. A 10X reduction in “I” will lead to a 10X increase in ROI. Then investors will take notice! But wait! With an ultra-low “I” we may not even need investors!

Lets look around to some of the most vibrant and exciting software development areas: mobile apps, software-as-a-service (SaaS), and gaming. How many of these developers still think it makes sense to write their own low-level parsers, hand craft their own C++ data structures, or re-write algorithms that have been written a 100 times before? Very few (at least of the successful ones). No, they use frameworks, platforms and engines such as: in mobile apps, Unity for gaming, or AWS for SaaS.

Invionics built Invio to be a Platform in exactly the same vein as these other platforms and engines. We want to help semiconductor companies rapidly build and deploy their own internal EDA applications, and allow EDA startups to do exactly the same thing. And we want it all to happen very, very quickly so the cost stays low, and the ROI stays high. Let us handle the HDL parsing, let us build the efficent data structures, let us create efficient search, let us create the GUI widgets, let us parse SDC, IP-XACT, UPF, and whatever else you need. Then you can focus on your value add!

Having announced our Invio Platform last year, we now have some experience under our belts and we are able to share some of these.

Our Invio Experience So Far:

1) We recently engaged with a stealth mode startup (who will remain nameless). They have used Invio to adopt this new 10X paradigm and it is fantastic to watch. With a bit of help from us, and using the Invio Platform, they were in front of a major semiconductor target customer with a live working demonstration of their EDA products in less than 6 weeks from starting from SCRATCH!.

2) We started working with our friends at CVC, and as we explained in our Python API blog, they ported their flag ship application to our new platform in 1hr! 1hr, how’s that for low investment!

3) Our semiconductor customers have been doing well too. One used Invio to build a custom application and had it deployed in production flows in less than 16 weeks.

4) And, just the other day, we had a potential customer e-mail asking for help to write an application to track down critical paths in third party SystemVerilog RTL. Invionics was able to turn around an example of a RTL level critical path explorer, complete with GUI, in just over 24 hours.

So, our first experiences are starting to make us believe it really is possible. EDA could really be 10X more productive.

Skeptical? Well…

Watch Us “Put our money were our mouth is” At DAC:

To show we mean business, Invionics will hold a 48 hour EDA Development Challenge at DAC. Here is how it will work: Come to our booth (#2507) on Monday morning and tell us what front-end EDA application you would like to see us try to build (anything from clock tree explorer, RAM BIST inserter, clock crossing verification, etc.) At 2pm we will pick the app with the most votes, and I will call two of our developers here in Vancouver. They will have until Wednesday at 2pm to complete the app, which we will demonstrate live in our booth Wednesday afternoon. If nothing else it will be fun to see the developers sweat! And for all those who do come back to see the results, we will also be giving away an iPad. Watch for more details to come on this one…

As always, we really want to hear what you think (And I expect I will hear some different opinions on this one!) Leave us a comment, and we will be sure to reply.

See you at DAC (booth #2507)!


Share on LinkedIn57Share on Facebook0Tweet about this on Twitter0Share on Google+0Email this to someone

Comments (4)

  1. Tudor Timi - Reply
    May 13, 2015

    I remember you mentioning in a post that you use Verific parsers for Invio. I like that you also eat your own dog food.

    You make a very good point with the small startups not doing everything from scratch and using frameworks and libraries and such. The software world also has something the hardware world doesn’t, though: more open source projects that can be leveraged to build bigger stuff. Even if we’d get much more open source code available tomorrow, I’m pretty sure a lot of companies still wouldn’t use this (I’ve seen significant resistance against anything non-commercial in the industry). Also, because of the very specialized tools required, it’s pretty much impossible to get any kind of hobbyist effort started and companies aren’t interested in making anything public.

    Nevertheless, hardware companies aren’t EDA companies, so maybe in that space it’s different.

    • Brad Quinton - Reply
      May 14, 2015

      Thanks Tudor! We have often thought about why the open source model doesn’t port to EDA, but experience shows that it hasn’t. I expect that it is because the problems are big and the relative number of qualified developers is small. Maybe there is not critical mass. There is also some industry resistance that doesn’t help.

      In any case, a lot of similar scaling effects can be achieved with commercial platforms in the sense that the cost of developing a platform can be amortized across a large group, keeping the initial costs low.

  2. virendra - Reply
    May 15, 2015

    Great article and novel approach , it address not not only the VC problem which will remain forever for EDA but taking off load engineering efforts where people struggle to retain the talent in a startups.

    • Brad Quinton - Reply
      May 15, 2015

      Thanks Virendra. That’s a good point that I had not thought about. Another benefit of rapid, lightweight development is that you do not need to as many resources, which, as you say, can be difficult to attract these days…


Leave a Comment