provides efficient memory management, object-oriented features, and high performance.
enables GPU power for accelerating computational tasks.
allows for efficient communication between GPUs, accelerating parallel computing in heterogeneous systems.
empowers high-performance data storage and retrieval across distributed systems.
facilitates post-processing and data analysis.
Amitis supports multi-species ions, limited only by memory, enabling precise simulations of complex plasma environments.
Amitis supports defining planetary exosphere profiles and simulating the production of planetary ions.
Amitis supports the definition and modeling of planetary internal magnetic fields in the form of dipole, quadrupole, and higher magnetic moments for advanced magnetosphere simulations.
Amitis supports modeling crustal magnetic fields for the Moon and Mars, empowering accurate surface interaction research.
Amitis allows custom planetary interior models for studying time-dependent electromagnetic sounding and induction response of interior structures.
Amitis enables virtual spacecraft flight through simulations, collecting time-series data on electromagnetic fields and particles.
We conducted nine simulation runs, doubling the complexity with each subsequent run. We then compared the scaling of Amitis using a single CPU-GPU pair to the execution time of the CPU-based model Flash, developed by the FLASH Center for Computational Science at the University of Chicago. Remarkably, the single CPU-GPU version of Amitis is six times faster than Flash running on 768 CPUs.
For more information, click here to read the Amitis paper!
Weak scaling tests of our 3D model to assess how performance changes as both the problem size and the number of GPUs increase proportionally. A100s and V100s were on different super-computers.
Strong scaling tests of our 3D model to measure how the performance of Amitis improves as the number of GPUs increases, while keeping the problem size fixed.