What Would Joe Do?
Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
CAD Tools: No longer agnostic w.r.t. end-product systems?
March 30th, 2017 by Peggy Aycinena
This year’s keynotes and skywalks cover a range of topics: From how the IoT will make smart buildings smarter, to why hardware/software co-design is being relabeled as digital twin[ning], why III-V compound semiconductors are the wave of the future, how wearable devices will soon be able to snoop around and find out if you’re having a bad day, and – did we mention the IoT?
New this year, even politics will get its 25 minutes of fame with a keynote outlining what’s up with ICs in China.
In other words, at DAC 2017 the topics are all over the map. You no longer just hear about the nuts and nuances of design; now you hear as much about what’s going on downstream in the system that runs on the IC as you hear about designing the IC itself.
There is a deeper meaning in all of this, of course.
For several years, a number of CAD tool vendors have begun to tailor their offerings to the specific needs of one end-product or another: Cars. Phones. Wearable devices. Data centers. Communication hubs. Systems awash in AI and ML. Not to mention the current cause célèbre, low-power edge devices which, in aggregate, comprise the Internet of Things.
Today, vendors say their tools are ostensibly kitted out with optimizations suited to each of these systems and more. But it is true?
Can you actually craft tools that are better for designing low-power edge devices than they are for designing high-reliability controllers for cars, tools that are better for designing high-speed data center CPUs than they are for designing high-throughput gaming GPUs.
Is this just marketing hype, or can a synthesis or place-and-route tool know that a million transistors placed in this location for that purpose will make for a better end-product than a million transistors placed in that location for this purpose?
Which begs a larger question: If these CAD tools are specific to end-product systems, are they actually system-aware? Do they automatically know that low power is needed here, but performance is needed there?
Do the tools actually know that, or do the designers using the tools know the end-goal and wield the tools accordingly? How far along are we in automating the process of design automation?
This is a pretty interesting question, and one I hope to study further over the next several months leading up to DAC.
I’ll be exploring the offerings of several of the major and minor EDA tool vendors to see what their tools purport to offer by way of end-product specificity, and try to find out what tweaks have been made in the tools to make them so smart.
Hence, by the time I get to the Design Automation Conference in June in Austin I hope to be a more informed consumer of the number one consumable at DAC: Marketing in all its glory.