Never miss a story from Cadence Academic Network. Subscribe for in-depth analysis and articles.
Last week, I visited the Cadathlon@ICCAD event at the 2019 International Conference on Computer Aided Design . It was my first CADathlon and I was quite intrigued, since the organizers webpage announced it boldly as the “Olympic Games of EDA,” which required the participants to qualify for the event.
CADathlon has nothing to do with a hackathon, though the names are quite similar. The 2-person groups receive six problems in advance, which are described in scientific papers. The problems are provided either by organizing professors or from EDA companies one week before the contest; William Chow, from Cadence, has provided one of the scientific papers this year. As previous participants admitted, they have never read scientific papers more carefully than the provided ones!
There is no internet access on the day of the event, so the solutions have to be written directly. The contestants receive test data, but the full set of test data is not available, so the solution must be robust, in order to pass the test with unknown inputs. At the end, the solutions are verified by a checker program and the team with most points is the winner.
The teams this year were international, but traditionally, PhD-students from Taiwanese universities are the majority. Out of nine teams, five were from Taiwan, two from US, one from Korea and one from Egypt. The CADathlon started at 9 in the morning and was supposed to end at 5, but the organizers decided to extend by an hour and a half to allow the teams to develop more clear solutions to the problems. This means that the contestants were programming for nine and a half hours without taking a break! I really admire the endurance of these young people, working in deep concentration on difficult mathematical problems for so long. Needless to say, the winners have a very good chance to get a job at an EDA company; However, the recruiting during the contest was not a good idea, since the last thing I wanted to do was to distract them while working on the tasks.
Now some words about the problems. Most of them were about classical EDA problems, like synthesis, routing, and logic simulation, but there were also some new kinds of problems, like optimization of spintronic devices and neural networks. An additional difficulty is the maximum allowed runtime of the program, if it is above a certain limit, the testcase is considered a failure. All the problems were highly technical and complex.
Siang-Yun Lee and Po-Chun Chien (Team ALCOM's Midnight Program) from EPFL/Switzerland and NTU/Taiwan
Jiayuan He and Yi-Shan Lu (Team UTCS) from UT Austin/US
Pei-Wei Chen and He-Teng Zhang (Team Smug Response) from NTU/Taiwan
I would like to thank the organizers of Cadathlon for their efforts: Tsung-Wei Huang from University of Utah, Yu-Guang Chen from National Central University, Taiwan and Pei-Yu Lee from Maxeda Technology, Taiwan.