Kavli Affiliate: Lile Wang
| First 5 Authors: Lile Wang, , , ,
| Summary:
The field of astrophysics has long sought computational tools capable of
harnessing the power of modern GPUs to simulate the complex dynamics of
astrophysical phenomena. The Kratos Framework, a novel GPU-based simulation
system designed to leverage heterogeneous computing architectures, is
introduced to address these challenges. Kratos offers a flexible and efficient
platform for a wide range of astrophysical simulations, by including its device
abstraction layer, multiprocessing communication model, and mesh management
system that serves as the foundation for the physical module container.
Focusing on the hydrodynamics module as an example and foundation for more
complex simulations, optimizations and adaptations have been implemented for
heterogeneous devices that allows for accurate and fast computations,
especially the mixed precision method that maximize its efficiency on
consumer-level GPUs while holding the conservation laws to machine accuracy.
The performance and accuracy of Kratos are verified through a series of
standard hydrodynamic benchmarks, demonstrating its potential as a powerful
tool for astrophysical research.
| Search Query: ArXiv Query: search_query=au:”Lile Wang”&id_list=&start=0&max_results=3