Fast Approximate AIG-Based Synthesis
Heil A, Keszöcze O (2022)
Publication Language: English
Publication Type: Conference contribution, Original article
Publication year: 2022
Event location: Bremen
Abstract
Approximate Computing is a design paradigm that trades off computational accuracy for gains in non-functional aspects such as smaller area, shorter delay, or less power consumption. Synthesizing approximate circuits is a difficult task due to the inherent complexity of evaluating error metrics used to indicate the quality of the approximated circuit. We present a novel and fast greedy, bucket-based Approximate Logic Synthesis technique working on AND-Inverter Graphs that iteratively performs XOR and constant replacements. It minimizes both, the area of the approximated circuit and, the error introduced by the approximations. We exploit properties of the replacement operations to speed up the computation of the weighted Hamming distance error metric, allowing the algorithm to have a very short execution time. The effectiveness of the presented method is evaluated in a set of experiments, including the state-of-the-art benchmark library EvoApproxLibLITE
Authors with CRIS profile
Related research project(s)
How to cite
APA:
Heil, A., & Keszöcze, O. (2022). Fast Approximate AIG-Based Synthesis. In Proceedings of the International Workshop on Boolean Problems. Bremen, DE.
MLA:
Heil, Annika, and Oliver Keszöcze. "Fast Approximate AIG-Based Synthesis." Proceedings of the International Workshop on Boolean Problems, Bremen 2022.
BibTeX: Download