Skip to content

WIP: New Benchmark: GraphChi

John Zhang requested to merge new-bench-graphchi into master

Description

Adding GraphChi as a benchmark.

Using Netflix and LiveJournal as input datasets as suggested by Shoaib Akram.

TODO

  • Benchmark calibration: decided whether netflix and livejournal are good sizes.
  • Validation: since the stdout output is different each time (threads), what should be the way to validate a successful run of the benchmark? Another option is to validate the files produced (a lot).
  • External data: the datasets downloaded are quite large. When !14 is merged (Issue #11), this data can potentially be downloaded and kept externally from the resulting dacapo.jar.

Merge request reports