Advanced Data Visualization Techniques with Distributed Multi-GPU Clusters
This paper explores advanced volume rendering techniques in the context of large data visualization using distributed memory multi-GPU clusters. The authors, including Thomas Fogal and Hank Childs, discuss the scalability challenges and the necessity of utilizing numerous cores to achieve acceptable rendering speeds. They delve into adaptive mesh refinement (AMR) to address problems related to uneven load distribution through sampling techniques adapted for coarse and fine boundaries. The research highlights future work and collaborations, emphasizing the importance of open-source communities and various institutional resources.
Advanced Data Visualization Techniques with Distributed Multi-GPU Clusters
E N D
Presentation Transcript
Large Data Visualization on Distributed Memory Multi-GPU Clusters Thomas Fogal, Hank Childs, Siddharth Shankar, Jens Krüger, R. Daniel Bergeron, Philip Hatcher
Volume Rendering Scalability • Need lots of cores to render at acceptable speeds • Figure: Howison, Bethel, Childs, MPI-hybrid Parallelism for Volume Rendering on Large, Multi-core Systems, EGPGV 2010
Brick Size NVIDIA Quadro FX 5800's
Future Work - AMR • "Adaptive mesh refinement" data • some spatial regions have higher-resolution • coarse/fine boundaries: sampling • uneven load distribution
Thanks • Funding • VACET • CIBC • C-SAFE • MMCI (Saarland University) • Resources • TACC • ORNL • Open source communities: GLEW, Mesa, VisIt • Anonymous reviewers • You!