A Hypergraph Model and Associated Optimization Strategies for Path Length-Driven Netlist Partitioning - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

A Hypergraph Model and Associated Optimization Strategies for Path Length-Driven Netlist Partitioning

Résumé

Prototyping large circuits on multi-FPGA platforms requires to partition the circuits into sub-circuits, each to be mapped in a given signle FPGA. While most existing partitioning algorithms focus on minimizing cut size, the main issue is not to map long paths across multiple FPGAs, as it may cause an increase in critical path length. To address this problem, we propose a new hypergraph model, for which we design algorithms for initial partitioning and partition refinement. We integrate these algorithm in a multilevel framework, combined with existing min-cut solvers, to tackle simultaneously both path length and cut size objectives. We observe a significant reduction in critical path degradation, by 12%-40%, at the cost of a moderate increase in cut size, compared to path-agnostic min-cut methods.
Fichier sous embargo
Fichier sous embargo
0 2 16
Année Mois Jours
Avant la publication
samedi 20 juillet 2024
Fichier sous embargo
samedi 20 juillet 2024
Connectez-vous pour demander l'accès au fichier

Dates et versions

hal-04379716 , version 1 (15-01-2024)

Licence

Paternité

Identifiants

Citer

Julien Rodriguez, François Galea, François Pellegrini, Lilia Zaourar. A Hypergraph Model and Associated Optimization Strategies for Path Length-Driven Netlist Partitioning. ICCS 2023 - 23rd International Conference on Computational Science, Jul 2023, Prague, Czech Republic. pp.652-660, ⟨10.1007/978-3-031-36024-4_50⟩. ⟨hal-04379716⟩
60 Consultations
8 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More