A new toolkit for analyzing galaxy simulations
The Octavius toolkit provides faster and more memory efficient analysis of large scale cosmological simulations.
Researchers created Octavius, a high-performance toolkit for analyzing galaxy simulations. This tool handles the big data computational challenge of simulation outputs. It uses a hybrid-parallel approach to process information.
The toolkit ingests a simulation snapshot and a halo catalogue. It uses a 6D friends-of-friends algorithm to locate galaxies. It then produces an HDF5 catalogue of properties for haloes, subhaloes, and galaxies.
A halo catalogue is a list of associated data for haloes. A snapshot is a single moment in a simulation. The toolkit is written in Python and uses a plugin-based architecture.
Octavius is faster than the predecessor toolkit, Caesar. It delivers order-of-magnitude improvements in speed. It also provides a factor of $>5$ reduction in peak memory.
The pipeline is presently I/O-limited. Compute times scale with the superlinear computational load of larger haloes. The code supports SIMBA, SWIFT-KIARA, SWIFT-EAGLE, SWIFT-COLIBRE, and TNG snapshots.
Key facts
- Octavius delivers order-of-magnitude improvements in speed compared to Caesar.
- The toolkit provides a factor of $>5$ reduction in peak memory.
- The pipeline is currently limited by I/O.