Pages: pp. 72-73
Abstract—Among this month's CEDA Currents newsletter items are the TCAD Best Paper Award, an approach to move beyond von Neumann computing, a note about DAC's "Birds of a Feather" session, the IEEE Embedded Systems Letters' most-accessed articles, and upcoming CEDA events.
Keywords—beyond von Neumann computing, DAC, IEEE Embedded Systems Letter, TCAD Best Paper Award, trends in social media
I recently had the pleasure of a ringside view of the selection process for the annual Donald O. Pederson TCAD ( IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems) Best Paper Award. Any paper that has appeared in TCAD over the past two calendar years is eligible for this award. The winner of this year's award was "FLUTE: Fast Lookup Table Based Rectilinear Steiner Minimal Tree Algorithm for VLSI Design," by Chris Chu and Yiu-Chung Wong ( IEEE Transactions on Computer- Aided Design of Integrated Circuits and Systems, vol. 27, no. 1, 2008, pp. 70-83)—an impressive work on fast routing algorithms.
This research addresses a fundamental problem in computer science: building Steiner minimum trees. This topic has applications in diverse fields such as IC design, networking and communications, traffic engineering, brain networking, and phylogenetic trees in bioinformatics. The problem is easy to state, but hard to solve optimally: given a set of points in space, find the best structure that connects these points.
As we map this statement to the routing problem in chip design, an additional restriction is imposed: the wires must run parallel or perpendicular to one another (that is, on a Manhattan grid), and the goal is to minimize the wire length for the route. This problem is known to be NP complete (ergo difficult) and is certainly not new; several talented researchers have banged their heads against it over the decades.
The novelty of this paper is in developing a computationally efficient algorithm with a clever acronym—Flute (fast lookup table estimation). The core idea of the approach is to classify nets in terms of their pin locations, and to group nets into equivalence classes that have similar properties. The key properties of these equivalence classes are precharacterized and stored in a lookup table. For any routing problem, the appropriate equivalence class is identified, and the entries in the lookup table point to a fast evaluation of the optimal Steiner tree.
However, the size of this lookup table grows rapidly (factorially) with the number of pins. So, this method is practical only for trees with a small number of pins. Fortunately, practical designs contain many nets with nine or fewer pins, for which the problem is tractable. Larger nets are split into smaller subnets, and the lookup table approach is then applied to individual subnets—at some loss in optimality, but with a massive gain in runtime and storage overheads.
This work has had excellent practical impact, and academic researchers and industry practitioners have widely adopted Flute-based methods, making this work an excellent exemplar of how theoretical techniques can be brought to bear upon intensely practical problems. In other words, it is a great choice for the Pederson award!
How was the TCAD Best Paper selected? In accordance with past practice, the procedure involved a detailed nomination and voting process. In the first round, about 20 papers were nominated for consideration, spanning all areas of CAD. Next, the TCAD editorial board was polled, and members were asked to provide their top two choices for the award. These choices were used to generate a short list. Then, in the final round of voting, the Best Paper was selected. Several outstanding papers were in contention for the award. However, throughout the voting process, the Flute paper was consistently ahead in the competition, and was a resounding winner in the final round of voting.
So, that's the story of this year's Best Paper Award.
One last note: at TCAD, we welcome your input at all times. If you have ideas to improve our processes, or to make our journal more relevant, or if you want to suggest special-section topics or anything else, please don't hesitate to get in touch with us.
—Sachin S. Sapatnekar (EIC of TCAD)
Steve Teig, president and chief technology officer (CTO) of Tabula described an approach to move beyond von Neumann computing during a luncheon on 15 June 2010 hosted by the IEEE Council on Electronic Design Automation at this year's Design Automation Conference (DAC).
Inventor of Tabula's Spacetime 3-Dimensional Programmable Logic Architecture, Teig described a simultaneous approach to the design of architecture, hardware, and software. He sees these as aspects of a cohesive whole, similar to the way John von Neumann (inventor of the von Neumann architecture) and Alan Turing (developer of the Turing machine) did. Teig believes that a new approach to architectural design will maximize the opportunities to go beyond von Neumann computing.
Prior to cofounding Tabula, a fabless semiconductor company that has developed a new category of 3D programmable logic devices (3PLDs), Teig was co-CTO of Cadence Design Systems. He had joined Cadence through its acquisition of Simplex Solutions, where he had been CTO. He has cofounded two successful biotechnology companies, CombiChem and BioCAD, as well as an EDA company, Tangent Systems. He holds more than 220 patents.
—Shishpal Rawat, email@example.com
CEDA hosted a "Birds of a Feather" session during DAC on social media, moderated by Shishpal Rawat, vice president of technical activities for CEDA and director at Intel of Business Enabling Programs and EDA Investments.
The discussion focused on trends in social media and ways to better use social-media channels to reach a wider EDA audience. This session attempted to determine which mechanisms the EDA community uses and how best to utilize them.
For more details, visit the CEDA: www.c-eda.org
The top-five accessed articles from IEEE Embedded Systems Letters during April 2010 were as follows:
CEDA conferences provide excellent opportunities for those interested in learning about the latest technical trends in electronic design and automation. If you'd like to participate or you have an idea about new topics of interest for our conferences, please contact William Joyner (firstname.lastname@example.org), CEDA vice president of conferences.
8th International Conference on Formal Methods and Models for Codesign (MEMOCODE)
26–28 July 2010
10th Formal Methods in Computer-Aided Design Conference (FMCAD)
20–23 October 2010
IEEE/ACM 2010 International Conference on Computer-Aided Design (ICCAD)
7–11 November 2010
San Jose, California