Accelerating ParaView on Crusher with VTK-m
- Science Problem
- DOE has a set of well-established tools to perform visualization for scientific discovery.
- The software for these tools, dating back to the 1990’s, is designed for CPU processing.
- New software is required to utilize the GPU acceleration needed for interactive visualization.
- Technical Solution
- The VTK-m visualization library is redesigning visualization algorithms for efficient execution on GPU devices.
- VTK-m is being integrated into tools such as ParaView.
- Science Impact
- ParaView was demonstrated running on Crusher (Frontier early-access hardware) GPUs.
- VTK-m is integrated to replace existing functionality with no effort on the users’ part.
- Visualization tools continue to support science discovery.
A screenshot of ParaView visualizing a WarpX simulation of magnetic vortex acceleration of a hydrogen beam (data courtesy of Axel Huebel). The visualization of this 13.6 billion element data set was performed using 384 GPUs on 94 nodes of Crusher.
The Z component of the electric field from the aforementioned simulation as visualized by ParaView using VTK-m.