Between The Lines

I recently met with Forte Design’s CEO Sean Dart and the company’s marketing and sales vice president Brett Cline and we got to talking about a number of topics and eventually the ESL tools market. To me ESL’s been somewhat of an enigma--a segment of the EDA industry that has been hyped to confusion and has yet to really truly deliver and make a notable impact on the EDA market. I commented to the Forte folks, that to me ESL’s become “the boy who cried wolf” segment of the EDA industry, as I typically hear a lot of presentations and claims from vendors and standards and analyst Gary Smith’s been talking it up for years but I rarely hear anything glowing from customers and certainly haven’t seen industry numbers spiking because of it.

Interestingly Cline and Dart agreed and said ESL has been way over hyped in ESL and that the extraordinary claims made by would-be ESL tool providers have raised the expectations with industry watchers and watered down the true value proposition of the market, in turn slowing ESL’s adoption by users.

That’s pretty much what I’ve experienced also. Pretty much every ESL vendor I’ve ever talked to claims to provide a “seamless” link from a system level language (either C, C++, ANSI C or algorithms) to RTL but typically what I hear from the user community is that there is no such thing as “seamless” in ESL and customers have to do or have to hire the EDA vendor to do a lot of manual work to bring ESL into the real hardware design flow.

There are numerous vendors in the ESL space and many types of ESL tools including: SoC ESL architectural level design tools, simulation and modeling tools, formal analysis (I’m not sure if that’s come to fruition yet) and even synthesis. Some ESL tools are targeted at helping architects come up with rough estimates of what functions should go in hardware and what should go into software and what hardware blocks they can reuse vs. what blocks they need to develop, some tools are targeted at getting software designers access to quasi-hardware earlier in the process. Some ESL tools are simply targeted at speeding up software functions by implementing them in hardware—typically in FPGAs. And some tools are really targeted at taking an ES level architectural description of a design and bringing it down to the RT level for true hardware design. And the various tools target different customers: Some are targeted at system architects, some are targeted at SoC architects, some are targeted at software engineers, some are targeted at embedded engineers, some are targeted at DSP programmers, some are targeted at hardware engineers. About the only thing common is they are all targeted at anyone willing to pay hardware tool prices, which of course can be a tall order for systems, software and embedded engineers who are used to paying low to no cost for their tools.

So, indeed, there is a broad swath of tools, methods and potential customers for ESL. Cline said what’s really needed to get to the next level is a bit of order to the ESL space and a standard flow. The latter would be hard to do of course because each user has their own way of doing things. Personally, I’d be happy if there were just some solid sub categories or buckets so it would be easier to sort out who does what and to whom they really serve. That would be a good starting point. Then I’d like to hear from real users of each vendor’s tools (without the vendor babysitting) to find out what they are really using their tools for. Perhaps if the plethora of ESL companies want to take the industry to the next level they should form their own baby EDAC or standards group specifically for ESL. Obviously there can’t be one ESL flow for everyone but perhaps there could be one standardized flow for every sub category? Perhaps if folks in the ESL space, defined the ESL segment a bit better (and a bit more realistically—less hype), they’d get a bit more traction and a broader user adoption and maybe in-turn the Big 4 EDA vendors would start to take the space a bit more seriously.

Vendors like Mentor with Catapult C and Synopsys with its Virtio tools have some cool point offerings in the space but pretty much all the big vendors have yet to push out comprehensive ESL flows. Maybe they are as confused as I am about ESL. Likely not. Maybe it’s just that those software tool price points to an undefined user base still don’t seem worth the R&D effort when it’s a sure thing there’s millions to be made in tried and true tool segments closer to silicon. With Mentor now jumping into the physical design space, seemingly tightening the competition in P&R, it will be interesting to see if the big vendors push more dollars into P&R R&D and sales efforts and if they do, if that will come at the expense of ESL. If that’s the case, ESL vendors are likely going to have to grow on their own for a bit longer. Perhaps it’s a good time to get united and organized. What do you think?