CEDA Nerd Fest @ DAC: Expanding the Reach of DA
June 11th, 2015 by Peggy Aycinena
Most mortals at the end of the day on Monday this week in San Francisco polished off their beers on the DAC Exhibit Hall floor at 7 pm and headed out for dinner and some R&R to prepare for the rigors of Tuesday. For those truly passionate about the future of Design Automation, however, Monday’s labors did not end at 7 pm. They had only begun.
Because in Room 300 from 7 pm to 8 pm, a series of visionary talks were offered — not one of them more than 3 minutes long — each proposing a different direction that the highly skilled DA community might take this industry by capitalizing on its vast expertise in optimization, algorithms, and hardware/software co-design. It was a great hour of presentations, which by the way was also a contest, complete with a panel of judges who got to choose the top three proposals and award cash prizes.
Prior to mid-day Monday, I had known nothing about this event/contest. In fact, it was only due to a random invite from someone I ran into during the afternoon on Monday that I realized there was actually something going on in Room 300 that evening and decided to attend. When I walked in at 7 pm and sat down among the 100+ other people in the room, I had no idea what was going on, or why, or that it was a contest. But everything quickly became clear.
It turns out this was first CEDA-sponsored Design Automation Perspective Challenge meeting at DAC, the result of a request for proposal that had gone out to both EDA academics and industry folks many months ago. Everybody [but me] had known well in advance about the event, and many people had responded to the RFP. Those whose proposals were accepted for presentation would address the meeting during the hour.
The judges sat in the front row facing the stage, a time-keeper sat mid-room and popped up at 3-minute intervals so absolutely nobody could exceed their allotted time at the podium, and UC San Diego’s Andrew Kahng [2009 DAC General Chair] sat on the side of the room MC’ing the event. The podium was on stage, with speakers expected to be ready to step up to give their presentations immediately after the previous speaker was done.
It was really fascinating and actually somewhat fraught with tension; speakers from both academia and industry had difficulty getting their message across in a scant 3 minutes. And frequently, speakers’ ideas were not fully articulated. (Helloooo? Get thee to Speech Class.) Nonetheless, all of the topics were interesting.
Following is a list of the talks with my brief summary attached. The names of speakers and the organizations they represent are not included, because I did not fully note that information during the meeting. I have subsequently searched both the CEDA and DAC websites but could not find the list of speakers, so I’ve left them out. [If someone involved wants to forward that info to me, I’ll happily update this blog.]
Names or no names, Room 300 was definitely the place to be on Monday evening. The ideas offered were varied in concept and detail, but were all compelling as they attempted to answer the eternal questions: What are the ideas that can re-invigorate Design Automation? Where will they come from? How will they be selected and implemented?
These are important issues with implications in academia and industry, and from the looks of the gathering Monday night, questions that will continue to motivate CEDA [IEEE Council on Electronic Design Automation] into the future. Clearly this event/contest is going to be on the program again next year at DAC in Austin [this claim yet to be verified, of course], and for many years beyond that. Look to participate if you can.
But if you do, buy a 3-minute egg timer and practice.
Proposal #1 — How to expand EDA: Open distribution of research artifacts
Consider these case studies: LLVN Compiler Infrastructure, a library of modules, reusable compiler, tool-chain components; OpenTuner, MIT’s open-sourced framework for building multi-objective program auto-tuning tools for different domains; OCCAM, Open Curation for Computer Architecture Modeling. The DA community should facilitate and proliferate the open distribution of research artifacts, which are critical for interdisciplinary research.
Proposal #2 — Pick one of the Grand Challenges, act on it
Working beyond the traditional EDA projects will definitely make a difference in the world. Look at recent Grand Challenges, e.g. Longitude, NSF: Help solve dementia. Restore movement to the paralyzed. Ensure clean water. Preserve resistance to antibiotics. Ensure sustainable food supplies. Fly without damaging the environment. Make solar energy affordable. Energy from fusion. Manage the nitrogen cycle. EDA must work to solve one of these challenges specifically, not just work to enhance general computing to assist a solution. The confluence of domain and algorithm expertise in EDA is not rewarded, however, and rushing to the “next big thing” discourages our students and researchers. Therefore, the DA community must pick one challenge and follow through!
Proposal #3 — DA & Test Challenges for Flexible Hybrid Electronics
FHEs are thin-film, light-weight, low-cost, bendable, durable electronics, often with large form factors, that offer the opportunity to create unique industrial designs for everything from biomedical apps to consumer electronics and IoT. The challenge is integrating the heterogeneous system parts, which require additive manufacturing, reliability under bending, robust design/test techniques, and tools for physical design and verification. The DOD Institute will soon begin to build an ecosystem for FHE manufacturing innovation, and need computationally efficient models that can capture process-dependent variations and time-dependent degradation, as well as printing processes that are compatible with CMOS EDA tools. The ability to perform CMOS-FHE co-simulation is also needed.
Proposal #4 — Integrate Simulation & Design for Optical Components
We’re using electronics for computing, photons for communication, and ions for memory including memristers and NVMs. We want to develop PDKs for Pcells for various optical components. The proposed approach: GDS-driven simulation of devices, supplemented by experimental data to deliver a validated Verilog model of the device. Proposed workflow: Design on-channel packing, develop a Pcell, choose a foundry, simulate optical behavior, look at DC/phase efficiency, and put it all together for an ideal, compact model.
Proposal #5 — Integrate Co-design for Design Automation of Things, Apply to EVs
The strength of the EDA community lies in computer architecture, circuit design, and system theory. All of this can be combined for the design automation of things, which is in particular need of a systematic approach for problem formulations, solution methods, validation methods. Electric vehicles are a great target for these skills, particularly EV power-consumption modeling, minimum energy traction/motor capacity design space exploration, and run-time optimization to match the appropriate driving methods with each set of driving conditions.
Proposal #6 — Integrate System-Package-Chip co-design
Today design practices continue to exist in silos, and for good reason; each silo comes with lots of complexity. But we need an integrated system-package-chip co-design environment to stop the need to design with excess margins, the overall system impact of design choices being not well understood by the ‘silo’ tools. We need a paradigm shift across disciplines: electrical, thermal, mechanical. And across practices: cost, power, performance, reliability, and manufacturability. This is an important challenge and opportunity for the DA community. Do it now.
Proposal #7 — Transitioning to Autonomous Driving Systems
DA is one of few area that has been able to build strong ties with niche areas such as embedded systems, synthetic biology, security, heterogeneous compute systems, and semiconductor manufacturing. DA also shows great promise with respect to autonomous driving vehicles, which should each be able to go from source to destination within predefined spacial accuracy limits — minimizing danger, the number of accidents, and the energy required for each trip, as well as competition for limited energy resources and parking spots. Current research should focus on high-frequency radar systems and algorithmic solutions to facilitate the transition from today’s human-operated vehicles to a future of autonomous vehicles behaving properly and consistently within a sea of vehicles. Models are needed for the design of vehicles that address issues of reliability, security, and privacy issues.
Proposal # 8 — EDA Challenges in Neuromorphic Computing
Defined as an interdisciplinary technology inspired by biology, physics, math, computer science, and electrical engineering, and targeted at designing artificial neural systems, neuromorphic computing has the potential to compensate for the weaknesses of von Neuman architectures in proving cognitive applications. Relevant research is being well funded by various agencies in the US and China, etc. Questions of current interest include: 1) Although a distinguished list of researchers have attempted to understand the human brain, they’ve only partially succeeded; do we actually need to understand the brain to emulate it? 2) Which is the correct platform for neuromorphic computing, a single platform or a hybrid of several different options? 3) Are conventional CMOS and EDA technologies capable of supporting long-term research and development in nueromorphic computing? 4) Is this a Life Science Project or an Engineering Project?
Proposal #9 — Design Automation in Energy Systems
Calgary is the energy capital of Canada, so graduate courses in optimization are of great interest and focus heavily on problems related to energy systems: Crude oil extraction; Best ways to move oil; Optimizing the size and location of oil storage facilities. Design automation opportunities exist in multiple areas including computer models for energy systems, reducing graph theory complexity, algorithms for processing Big Data, and systems planning for minimizing carbon emissions. Long-term benefits from applying DA techniques to energy systems include reduced pollution, sustained economic growth, and political stability. To achieve this, an open-source framework should be created that includes a design platform for sharing tools and a social platform for sustained collaboration across organizations.
Proposal #10 — Design Automation for Genomics
The Genome is the blueprint of life, which can be read using modern sequencing techniques. The cost of sequencing [in some cases] has dropped below $1000, from initial costs in excess of $100 million. A change in scale comparable to achievements out of EDA with respect to Moore’s Law. We need big computers that can sequence huge bunches of genomes to capitalize on opportunities to develop personalized medical care, e.g. for cancer and COPD. DA provides algorithmic expertise, point tools, Big Data management, systems knowledge, skills with heterogeneous computing and data compression. This domain knowledge is crucial to understanding biology and enhancing genome sequencing technology.
Proposal #11 — Go Meta to pick the Right Design Flow
We should be producing meta knowledge from all of the work in synthesis, modeling, optimization, verification, debugging, simulation, emulation. By looking at the meta theory behind all of this, there is potential for rapid, high-quality design automation when new, faster strategies can be developed for picking the correct solution to a particular problem, strategies that would prove useful way beyond DA for ICs. If we don’t pursue these opportunities and quickly, our best students will continue to go to Google and Facebook.
Proposal #12 — Leveraging EDA Research for Design of Things
Modern processors contain billions of interconnected devices, and are designed using a sophisticated set of CAD tools that guarantee performance and correctness. Significant investment in CAD tools over the decades, including algorithms and techniques, can be applied to virtually all design processes that include: model, simulate, optimize, analyze, re-iterate. For example, energy distribution networks could be optimized, or smart cities. Wherever DA research is going on, create a bigger, more global impact by applying those same principles to problems in the energy area.
Proposal #13 — Trustworthy Trusted Hardware
Trusted commercial hardware is an oxymoron. Intel or ARM does not tell you what’s inside when you buy their products; you don’t own the RTL. Trusted hardware is open-source, formally verified hardware that benefits from an open security community and analyzing security properties. Let’s move to this model.
After the presentations were complete, the judges tabulated their results and Intel’s Shishpal Rawat, CEDA Chair, presented the top three winners with an honorarium: $500 for First Place, $200 for Second, $100 for Third. First however, the audience was allowed to log on and record their own personal favorite.
1st Place: Proposal #3 — DA & Test Challenges for Flexible Hybrid Electronics
2nd Place: Proposal #5 — Co-design for Design Automation of EVs
3rd Place: Proposal #9 — DA for Energy Systems
Honorable Mention: Proposal #13 — Trustworthy Trusted Hardware
Honorable Mention: Proposal #7 — Transitioning to Autonomous Driving Systems
Honorable Mention: Proposal #6 — Integrate System-Package-Chip co-design
Audience Favorite: Proposal #9 — DA for Energy Systems
Audience questions …
To dissuade anyone from thinking that Room 300 was awash with goodness and light from 7 pm to 8 pm, here’s a sample of the types of questions thrown at the presenters by a tough audience.
Q: Proposal #8 — Transistors are stupid, but neurons are wonderful. Why would you use transistors to create brains instead of just laying down neurons?
Q: Proposal #12 — What’s so novel about working on Smart Cities?
From DAC Program …
Electronic Design Automation (EDA) has had a profound impact on the development of modern computing and information technology which in turn has transformed our lives and society. Over the past 5 decades, the vision and efforts of the EDA community have been primarily focused on supporting the successful scaling of the Moore’s law. Despite its dominant focus on electronics, EDA is one of the first fields in engineering that has taken a truly interdisciplinary route: several abstractions, computational models, algorithms, methodologies, and tools have been jointly developed by the chemists, device physicists, electrical engineers, computer scientists, applied math, and optimization experts. Not only the EDA tools are able to take the high level functional description and automatically synthesize and optimize it to a physical entity, but also they can perform the complex tasks of simulation and verification. The growth in traditional EDA industry and development is however, limited due to the physical limits of manufacturing technology and the field’s maturity.
The Design Automation (DA) Perspective Challenge aims to seek visionary proposals describing key long-term research problems in DA of emerging domains that can benefit and further evolve the practices and methodologies developed by the EDA researchers and industry over the past few decades. The perspective proposals shall indicate the challenges in addressing the suggested long-term problems and how addressing the problems could potentially lead to a big scientific or industrial impact. The committee will carefully review the proposals; The proposals with long-term prospects will be included in the DA Challenge Archive on IEEE CEDA web-portal. A limited set of proposals will be selected for presentation at the competition event, to take place in SF Moscone Center on June 8 at 7pm.
Tags: Andrew Kahng, CEDA, DA Perspective Challenge, DAC, Design Automation Conference, IEEE Council on Electronic Design Automation, Shishpal Rawat