Lightweight Graph Reordering (IISWC18)

Repo for the IISWC18 paper - "When is Graph Reordering an Optimization? Studying the Effect of Lightweight Graph Reordering Across Applications and Input Graphs" pre-print

Contents

The repo contains source code:

• Packing Factor computation
• Lightweight Reordering techniques (incorporated in the Ligra applications):
  • Hub Sorting (based on [1])
  • Hub Clustering

Details about finding the packing factor of graphs (must be in ligra-format) and applying lightweight reordering can be found in the Ligra and GAP directories

Tools

For running algorithms beyond those in the GAP and Ligra benchmark suites, there is a tool to generate a reordered graph (as an edgelist or in CSR format) which can be used as an input for the new algorithm.

In the GAP/* directories, graph-reorderer will output reordered graphs. The -l input flag allows choosing between out-degree vs in-degree based reordering. The -e flag outputs the graph in the edgelist format (note that the edgelist is sorted by source ids). The -b flag outputs the graph in the serialized graph format used in GAP

Requirements

Ligra and GAP require g++ version >= 5.3.0

Both benchmark suites are parallelized with OpenMP

Tested on Debian Stretch with g++ version 6.3.0

Issues

For issues/information, please feel free to send an email at [email protected]

References

[1]: Zhang, Yunming, et al. "Making caches work for graph analytics." Big Data (Big Data), 2017 IEEE International Conference on. IEEE, 2017