Department of Computer Systems

Designing and exploiting libraries of approximate circuits

Czech title:Navrhování a využívání knihoven aproximativních obvodů
Research leader:Sekanina Lukáš
Team members:Bidlo Michal, Strnadel Josef, Vašíček Zdeněk
Agency:Czech Science Foundation
Code:GA19-10137S
Start:2019-01-01
End:2021-12-31
Keywords:approximate circuit, computer aided design, genetic programming, evolvable hardware, digital circuit library
Annotation:
Approximate circuits are building blocks of complex systems on a chip that can exchange the quality of processing for power consumption reduction. The design of an approximate circuit showing good tradeoffs between the error and other parameters is still a challenging task. We propose to reduce the design time and increase the complexity of circuits that can routinely be approximated by (i) developing a comprehensive library of elementary approximate circuits and (ii) introducing suitable compositional schemes exploiting the elementary circuits available in the library. Assembling a complex approximate circuit from approximate sub-circuits is a challenging problem as it is, in general, unclear how to perform the (de)composition and how the errors of these sub-circuits manifest at the global level. The project seeks the following contributions: (1) efficient deterministic and evolutionary compositional algorithms for approximate circuit design, (2) a library containing over 1 million approximate circuits, and (3) new knowledge about approximate circuits by analyzing the library.

Publications

2019HUSA Jakub. Comparison of Genetic Programming Methods on Design of Cryptographic Boolean Functions. In: Genetic Programming 22st European Conference, EuroGP 2019, Proceedings. Cham: Springer International Publishing, 2019, pp. 228-244. ISBN 978-3-030-14811-9.
 KOCNOVÁ Jitka and VAŠÍČEK Zdeněk. EA-based refactoring of mapped logic circuits. In: 2019 IEEE International Symposium on Circuits and Systems (ISCAS). Red Hook, NY: IEEE Computer Society Press, 2019, pp. 1-5. ISBN 978-1-72810-397-6.
 MRÁZEK Vojtěch, HANIF Muhammad A., VAŠÍČEK Zdeněk, SEKANINA Lukáš and SHAFIQUE Muhammad. autoAx: An Automatic Design Space Exploration and Circuit Building Methodology utilizing Libraries of Approximate Components. In: The 56th Annual Design Automation Conference 2019 (DAC '19). Las Vegas: Association for Computing Machinery, 2019, pp. 1-6. ISBN 978-1-4503-6725-7.
 MRÁZEK Vojtěch, VAŠÍČEK Zdeněk, SEKANINA Lukáš, HANIF Muhammad A. and SHAFIQUE Muhammad. ALWANN: Automatic Layer-Wise Approximation of Deep Neural Network Accelerators without Retraining. In: Proceedings of the IEEE/ACM International Conference on Computer-Aided Design. Denver, 2019, pp. 1-8.
 REK Petr and SEKANINA Lukáš. TypeCNN: CNN Development Framework With Flexible Data Types. In: Design, Automation and Test in Europe Conference. Florence: European Design and Automation Association, 2019, pp. 292-295. ISBN 978-3-9819263-2-3.
 VAŠÍČEK Zdeněk, MRÁZEK Vojtěch and SEKANINA Lukáš. Automated Circuit Approximation Method Driven by Data Distribution. In: Design, Automation and Test in Europe Conference. Florence: European Design and Automation Association, 2019, pp. 96-101. ISBN 978-3-9819263-2-3.

Your IPv4 address: 3.226.251.81
Switch to https