An Efficient, Scalable IO Framework for Sparse Data: larcv3

TL;DR

The paper introduces larcv3, an efficient and scalable I/O framework designed for handling sparse, irregular data in particle physics, optimized for high-performance computing and compatible with Python data analysis tools.

Contribution

Develops larcv3, a new C++/Python framework that significantly improves I/O performance for sparse data in neutrino physics experiments, enabling scalable analysis on modern HPC systems.

Findings

Demonstrates improved I/O performance in benchmark tests.

Enables fast processing of ragged and irregular data.

Ensures compatibility with popular Python analysis tools.

Abstract

Neutrino physics is one of the fundamental areas of research into the origins and properties of the Universe. Many experimental neutrino projects use sophisticated detectors to observe properties of these particles, and have turned to deep learning and artificial intelligence techniques to analyze their data. From this, we have developed \texttt{larcv}, a \texttt{C++} and \texttt{Python} based framework for efficient IO of sparse data with particle physics applications in mind. We describe in this paper the \texttt{larcv} framework and some benchmark IO performance tests. \texttt{larcv} is designed to enable fast and efficient IO of ragged and irregular data, at scale on modern HPC systems, and is compatible with the most popular open source data analysis tools in the Python ecosystem.