Area Estimation Framework for Digital Hardware Design using Machine Learning

Conference: MBMV 2020 – Methoden und Beschreibungssprachen zur Modellierung und Verifikation von Schaltungen und Systemen - GMM/ITG/GI-Workshop
03/19/2020 - 03/20/2020 at Stuttgart, Deutschland

Proceedings: GMM-Fb. 96: MBMV 2020

Pages: 10Language: englishTyp: PDF

Froemmer, Jens; Gowayed, Yara (Robert Bosch GmbH, 70442 Stuttgart, Germany & TU Kaiserslautern, 67663 Kaiserslautern, Germany)
Bannow, Nico (Robert Bosch GmbH, 70442 Stuttgart, Germany)
Kunz, Wolfgang; Grimm, Christoph; Schneider, Klaus (TU Kaiserslautern, 67663 Kaiserslautern, Germany)

Digital hardware design usually includes a design space exploration phase in order to determine a viable trade-off between the area and other metrics of the design, such as the performance. As the design space grows exponentially with the number and the range of the design parameters, synthesizing the digital hardware design for a large number of design candidates quickly becomes infeasible. The proposed area estimation framework not only enables a designer to automatically synthesize different design candidates and store the results in a database, but also utilizes machine learning algorithms to approximate the area for future design candidates. Due to its universal approach, the area estimation framework can easily be extended to support arbitrary hardware description languages, synthesis tools, and machine learning algorithms. First experiments using only basic automated hyperparameter tuning already show promising results of above 90% accuracy.