Content

Version/Status

  • Content created at: 2025-04-17
  • Status: Active (at the time of publication 2025-10-01)

Description

Given a graph, KaMinPar aims to divide its vertices into k disjoint blocks of approximately equal weight, minimizing the number of edges crossing between blocks. KaMinPar offers high efficiency and low memory overheads while achieving partitions of similar quality as the widely used Metis algorithm. For example, it can partition the massive hyperlink-2012 graph (approx. 3.5 billion vertices and 112 billion edges) into 30,000 blocks in under 6 minutes on 96 cores, using around 300 GiB RAM. Notably, KaMinPar is also optimized for large k, where it often achieves an order-of-magnitude speedup over competing partitioners. For unweighted input graphs, it guarantees strict adherence to the balance constraint.

Metrics

Questions / Remarks / Feedback
X

Feedback for Dagstuhl Publishing


Thanks for your feedback!

Feedback submitted

Could not send message

Please try again later or send an E-mail