Politecnico di Torino (logo)

A Graph Neural Network based HLS QoR prediction.

Zhuowei Li

A Graph Neural Network based HLS QoR prediction.

Rel. Luciano Lavagno, Muhammad Usman Jamal. Politecnico di Torino, Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica), 2023

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (4MB) | Preview

Starting from a C/C++ behavioral-level programming language, High-Level Synthesis (HLS) is a solution for rapid prototyping of application-specific hardware, where designers can apply HLS directives to optimize hardware implementations by trading-off between cost and performance. However, current HLS tools do not provide reliable Quality of Results (QoR) estimations, which prevents quick and precise design-space exploration. Under the background of the widespread use of machine learning (ML) to improve the predictability of EDA tools, this work proposes a Graph Neural Network (GNN) based predictive model that can accurately predict post-route QoR. The experimental results demonstrate that the predictive model not only outperforms state-of-the-art HLS tool in terms of QoR estimation accuracy but also speed.

Relators: Luciano Lavagno, Muhammad Usman Jamal
Academic year: 2022/23
Publication type: Electronic
Number of Pages: 64
Corso di laurea: Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica)
Classe di laurea: New organization > Master science > LM-25 - AUTOMATION ENGINEERING
Aziende collaboratrici: Politecnico di Torino
URI: http://webthesis.biblio.polito.it/id/eprint/26675
Modify record (reserved for operators) Modify record (reserved for operators)