The ECP ALPINE project: In situ and post hoc visualization infrastructure and analysis capabilities for exascale

Overview

This paper describes the Exascale Computing Project (ECP) ALPINE project, which focused on developing exascale-ready visualization solutions, including in situ processing. The project aimed to address the challenge of data generation outpacing data saving capabilities on exascale computers. ALPINE developed both post hoc visualization tools, ParaView and VisIt, and in situ algorithms and infrastructures, notably Ascent. KVL regularly uses and contributes to these projects.

Work Summary

The ALPINE project addressed the need for efficient visualization and analysis of the massive datasets produced by exascale simulations. The project's contributions included making post hoc tools like ParaView and VisIt exascale-ready by implementing GPU-accelerated filters and enhancing their scalability. It developed several exascale visualization and analysis algorithms, such as topological analysis, adaptive sampling, Lagrangian flow analysis, and optimal viewpoint selection. The project also created Ascent, a lightweight in situ visualization and analysis library designed for multi-physics simulations. Furthermore, ALPINE integrated other ECP software technologies and incorporated its algorithms and infrastructure into ECP applications like Combustion-Pele, WarpX, and MFIX-Exa. The paper details the development and integration of these tools and algorithms, showcasing their application in real-world simulations.

You can see more information on this publication here: https://repository.kaust.edu.sa/items/41826f92-d919-449c-aa76-0d32cfc65b18

Impact

The ALPINE project significantly advanced the field of scientific visualization for exascale computing. By providing robust in situ and post hoc visualization tools and algorithms, it enables researchers to analyze and understand complex simulation data more effectively. The development of Ascent offers a lightweight and efficient solution for in situ processing, reducing I/O bottlenecks. The project's integration with various ECP applications demonstrates its practical impact and its potential to accelerate scientific discovery across multiple domains. The work helps address the critical challenge of managing and visualizing the massive amounts of data generated by exascale simulations, ensuring that valuable insights can be extracted from these computations. KVL's use and continued contributions to these projects ensure that KAUST and Shaheen III can take advantage of performant and cutting edge visualization.