Enabling GPU accelerated computing in the SUNDIALS time integration library

As part of the Exascale Computing Project (ECP), a recent focus development efforts for SUite Nonlinear and DIfferential/ALgebraic equation Solvers (SUNDIALS) has been to enable GPU-accelerated time integration in scientific applications at extreme scales. This effort resulted several new GPU-enabled implementations core SUNDIALS data structures, support programming paradigms which are aware heterogeneous architectures, introduction utilities provide points flexibility. In this paper, we discuss our considerations, both internal external, when designing these features present themselves. We also performance results on Summit supercomputer early access hardware Frontier supercomputer, demonstrate negligible overhead resulting from additional infrastructure significant speedups using NVIDIA AMD GPUs.