Routability-Driven Package Router with Machine Learning
| Autor: | Jia-Hao Yang, 楊家豪 |
|---|---|
| Rok vydání: | 2019 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 In advanced technology nodes, it is important to reduce design turnaround time and human resources by estimating design quality before the time-consuming synthesis process. In a system in package (SiP), it is important to be able to evaluate the quality of placement results before the time-consuming routing stage. However, there are no known solutions to evaluate placement results of a SiP. Furthermore, applications and requirements of SiPs are versatile, which makes modeling all the placement patterns in SiPs impractical. To tackle these problems, in this thesis, we develop a machine learning-based method to measure whether a placement is effective and efficient in substrate layout. Appropriate features are extracted from previous placement results, and tags for supervised learning are provided by expertise. In this way, we can use Support Vector Machine (SVM), Logistic Regression (LR), Decision Tree (DT), K-Nearest Neighbor (KNN) to train an evaluation model to evaluate the placement results of a SiP. In this thesis, we adopt machine learning techniques to help designers predict the routability at placement stage. With the predicted results, we can make proper adjustments in advance to avoid unnecessary design iterations and complete routing accordingly. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|