Real-time Visualization of Large Scale and Time Varying Nuclear Reactor Simulations on Supercomputers
-100x speedup of visualization using the visualization software "In-Situ PBVR"-

• High performance supercomputers enabled extreme scale simulations. Visualizing the results of such simulations is now a critical issue as conventional visualization approaches post process petabytes (10¹⁵ bytes) of calculation data on storage, which can take several days or longer.
• JAEA developed the visualization software "In-Situ PBVR" that computes visualization data at simulation runtime on the supercomputer. It outputs compressed visualization data (~megabytes, 10⁶ bytes) instead of all calculation data.
• "In-Situ PBVR" made visualization ~100x faster than the conventional method, and compressed the amount of data transfer to one billionth. This enabled real-time visualization of large scale, time varying nuclear reactor simulations for temperature and flow fields on the supercomputer.
• This software was released as an open source software. It is expected to become indispensable for exascale simulations performed on state-of-the-art supercomputers.

Figure 1 Overview of "In-Situ PBVR". On the supercomputer, the particle calculation program attached to the simulation converts the calculation data (~petabytes) into particle data for visualization (~megabytes) and outputs this to the storage. The control program aggregates the particle data and transfers it to the user PC over the network. The rendering program on the user PC renders the particle data on the screen and displays the visualization results using the viewer. The user observes the visualization result and adjusts visualization parameters such as color and opacity. The adjusted visualization parameters are stored on the storage via the control program. The particle calculation program reads the visualization parameters from the storage and performs new visualization computation for them in the following time steps.

【Published article】

Journal：Journal of Visualization

Title:Improvement of Interactive Remote In-Situ Visualization using SIMD-aware Function Parser and Asynchronous Data I/O

Authors：Takuma Kawamura¹, Yasuhiro Idomura¹

Affiliation：¹Japan Atomic Energy Agency

DOI:10.1007/s12650-020-00652-z