Streami: GPU-Accelerated MPI Library for Fluid Field Lines
GPU-accelerated field line computation for HPC fluid dynamics—now open source.
Streami is a new, extensible GPU-accelerated library designed for high-performance computation of field lines in fluid flows. Developed by Stefan Zellmann, Milan Jaros, Andrea Paris, Ingo Wald, and Tatiana von Landesberger, Streami leverages MPI data-parallelism to scale across distributed GPU clusters. It functions as a thin interoperability layer, supporting both post-hoc analysis (after simulation) and in-situ analysis (during simulation), and seamlessly interfaces with existing MPI-based applications. Key design decisions prioritize high performance and extensibility, allowing users to adapt the library to various fluid flow field representations.
The library also includes a sample application for rapid prototyping and interactive seed point placement, making it practical for researchers and engineers working on computational fluid dynamics. Streami is released under a permissive open-source license, encouraging broad adoption and community contributions. By combining GPU acceleration with MPI parallelism, Streami addresses the growing need for efficient field line tracing in large-scale simulations, with potential applications in aerospace, climate modeling, and energy systems.
- Streami is an MPI data-parallel library that computes field lines on GPUs, targeting HPC fluid dynamics.
- Supports both post-hoc and in-situ analysis, integrating with existing MPI applications.
- Released under a permissive open-source license, includes a sample app for rapid prototyping.
Why It Matters
Enables faster, scalable field line analysis on GPU clusters, accelerating CFD research and industrial simulations.