Workshops Archive

A Distributed Particle-Based Volume Rendering Framework for Fugaku and Beyond


Workshop: ISAV26: In Situ AI, Analysis and Visualization

Authors: Tsukasa Osaki and Jorji Nonaka (RIKEN Center for Computational Science (R-CCS)) and Takuma Kawamura (Center for Computational Science & e-Systems, Japan Atomic Energy Agency (JAEA))

Abstract: Modern flagship HPC systems increasingly rely on GPU-accelerated architectures. While GPUs accelerate simulations, CPU resources may remain available, providing an opportunity for in-situ processing and visualization. This community talk presents an ongoing effort to develop a distributed Particle-Based Volume Rendering (PBVR) framework that exploits this CPU/GPU division. Volumetric fields are converted in situ into compact visualization particles using CPUs on the HPC system, while interactive rendering is performed on a GPU-equipped remote client. We evaluate the approach on Fugaku using OpenFOAM and demonstrate particle generation across up to 2,304 compute nodes (110,592 CPU cores). For a separate 432 x 432 x 432 grid measurement at 576 MPI ranks, PBVR particle files were 98.1% smaller than the uncompressed logical Ux payload over 11 steps. We discuss the implementation, initial performance results, and potential applicability to future heterogeneous HPC systems such as FugakuNEXT.


Back to ISAV26: In Situ AI, Analysis and Visualization Archive Listing

Back to Full Workshop Archive Listing