SC19 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

SOLAR Consortium: Accelerated Ray Tracing for Scientific Simulations


Authors: Paul Navratil (Texas Advanced Computing Center (TACC), University of Texas), Estelle Dirand (Total, France), Pascal Grosset (Los Alamos National Laboratory), Peter Messmer (Nvidia Corporation), Jeff Amstutz (Intel Corporation), Dmitry Kozlov (Advanced Micro Devices (AMD) Inc)

Abstract: Many physical simulations incorporate vector mathematics to model phenomena such as radiative transfer and to compute behavior such as particle advection. Hardware-optimized ray tracing engines, tuned by processor manufacturer engineers, can accelerate simulation critical sections that depend on ray-based traversals and intersections. This BoF will serve as a venue for developers and users of simulations, visual analysis codes, and ray tracers to discuss interfaces, capabilities, and performance. This meeting will continue the conversation toward standardization of ray tracer interfaces to facilitate their expanding role throughout scientific workflows, including data evaluation, insight formulation, discovery communication, and presentation-quality artifact generation.

Long Description: This BOF aims to build awareness and develop community for using hardware-optimized ray tracing engines in simulation codes that make extensive use of ray-like vector calculations, such as computing radiative transfer with Monte Carlo methods or advecting particles through a vector field. Using hardware-tuned methods for traversal and intersection in the inner loops of such simulations can provide significant performance enhancement, particularly on next-generation GPU hardware with dedicated ray processing logic. Hardware-tuned ray tracers, however, have not yet been widely adopted due to the graphics-centric nature of their interfaces and terminology. This BOF seeks to build community among simulation users and developers, visualization developers, and ray tracing engine developers in order to standardize interfaces for ray tracer use for simulation and for coprocessing of simulation and analysis.

Goals: This BOF will build community among scientific domains and research teams that utilize ray-like vector math in their simulations and gather requirements from these teams to improve the ability of hardware-tuned ray tracing engines to support simulation science. By bridging community and terminology divides between graphics and simulation users, we hope to improve simulation performance for science teams and guide ray tracing engine developers to improve support for scientific use cases.

Topic: The BOF will seed community discussion with three-minute lightning presentations providing examples of ray tracing’s use in scientific visualization (Paul Navratil, TACC), industrial simulation and analysis (Estelle Dirand, TOTAL), and astrophysical simulations (Hannah Ross, LBL, pending availability); as well as a summary of current capabilities from major ray tracing engine developers (Jeff Amstutz, OSPRay @ Intel; Peter Messmer, OptiX @ NVIDIA; RaedonRays @ AMD pending availability). We will then organize participants into small groups to discuss experiences, needs, and requirements.

Relevance to HPC: Use of hardware-optimized ray tracers in scientific simulation represent an untapped opportunity for performance improvement, particularly for current and near-future machines with accelerated ray tracing hardware. This BOF aims to represent a diverse set of experiences and viewpoints across HPC.

Prior events: The inaugural SOLAR meeting was held May 30-31, 2019 in Santa Fe and attracted 38 participants (20 national lab, 11 industry, 7 academia), which resulted in a vendor-neutral Khronos Group proposal to standardize ray tracing interfaces for scientific rendering, as well as initial discussion toward a standardized ray tracing interface for simulation operations. We hope to leverage the broader audience of SC to increase SOLAR community visibility and reach to a wider audience of scientists to capture broad HPC needs in a future simulation interface standard.

Expected outcome: we expect to (1) produce a white paper summarizing the community insights toward proposing a standardized ray tracing interface for simulation, and (2) develop proof-of-concept collaborations between simulation teams and ray tracing engine developers to quantify the performance improvements made possible by using a hardware-optimized ray tracing engine.


URL: https://sites.google.com/tacc.utexas.edu/solar


Back to Birds of a Feather Archive Listing