In the latest edition of my embedded edge with Nitin podcast, we learn about some of the trends at this year’s Design Automation Conference (DAC 2020). When we spoke to the program track chairs in the podcast, it’s apparent that the key trends this year are the rise of artificial intelligence (AI) and machine learning in EDA tools, and the prominence of RISC-V, with 25% of papers featuring projects using RISC-V.
This year, DAC 2020 is a virtual event. Harry Foster, DAC vice chair & mentor, and who also happens to be the DAC 2021 general chair said, “We decided very early on to adopt a philosophy of rather than trying to recreate a physical experience of a live conference, which obviously, is something you can’t do it virtually, we decided to focus on achieving DAC’s goals within a virtual environment, such as providing a space for leading research from academics, providing even a place where industry could showcase their latest tools and products, various educational forums and even try to have some fun and virtual networking as much as we can do.”
Foster noted this year has been a remarkable year, with a 21% increase in submitted papers – last year, there were 815, while this year there were 984. “We ended up having to grow the committee that reviewed the papers to over 300 researchers and accepted 228 papers. But the key point here is there’s so much research that’s going on in our industry, it’s just a really an exciting time to be around.”
Foster said there’s continual growth and interest in AI and ML hardware architecture papers. “This has been just mind-boggling growth: we had an 86% increase over last year [in terms of papers].” He said, “The other really hot topic, again, is security. Also, over the past couple years, we’ve seen increase in submissions in embedded systems.”
RISC-V is another area which Foster said is showing more prominence, with 25% of papers showing projects incorporating the open source instruction set architecture. Foster talks about this as well as highlighting areas such as cloud to edge, in-memory computing and 5G.
For the designer and IP track, we spoke to its program chair Renu Mehra. She said, “One of the areas we’re seeing is in traditional back-end CAD, we saw an uptick in papers on machine learning, covering both the design flows and also how to use machine learning techniques inside EDA tools. In other areas on the front end is a large uptick on security and safety papers; also in model based systems and software. And meanwhile, our traditional areas of verification for test validation, emulation and simulation also came out strong.
Mehra also talks about a new session at DAC this year. “I’m very excited about the new session we have planned on chip design. This time, we thought it would be a good idea to look at the many new chips coming up especially with new applications and AI. We approached various design companies to talk about the challenges they are facing and the opportunities in this new decade. So we have speakers from Xilinx, Intel and Lattice talking about the latest challenges they’re facing in the era of AI chips.”
For the embedded track, program chair Rob Oshana shared his excitement at what’s coming up. He said, “Some of the common themes in DAC are finding their way into the embedded space as well, like machine learning as it migrates from cloud to the edge that continues to be popular. And even RISC-V technology, both hardware and software down into the embedded spaces with 32-bit. But there’s a couple other areas that I’m excited about that are getting a lot of interest. One is in wireless connectivity with 5g, especially for embedded in the edge space. That’s becoming very exciting and growing quickly. And also audio and voice processing for embedded and also at the edge. Both of these, we’re seeing a lot of interest in these technologies as they continue to get more popular across a wide swath of embedded markets. So those are two that are becoming more popular in DAC this year.”
Oshana said what surprised him was the continued innovation in machine learning in the embedded space, related to resource constrained devices and memory and processing power. “There’s just a lot of innovation in inferencing for these resource-constrained embedded devices, even in the open source community. There’s a lot of open source technologies being developed in open source machine learning frameworks, but there’s a lot of additional work going on to drive that into the MDN node, into the embedded controller, with what could be a small embedded microcontroller. So that innovation is going very quickly, faster than I would have expected.
“I think people are realizing that the edge in the end node are the next areas that need to be conquered with the whole machine learning started in the cloud, but now, rapidly deploying that down. So that that’s going very quickly. The other thing that quite frankly surprised me from just a year ago is just how quickly RISC-V technology is being deployed across the industry into embedded spaces. It’s really very low cost microcontrollers now based on this technology and the innovation around it with the instruction extensions. It surprised me only in the sense of how quickly It’s moving. So those are a couple things that have surprised me a little bit. And then again, 5G has moved actually quicker than I thought. There was a lot of questions as to whether 5G would really be needed or take off this quickly. But what we’re seeing is that it indeed is moving very quickly.”