top-execution-time | ← SUNDIALS and hypre: Exascale-Capable Libraries for Adaptive Time-Stepping and Scalable Solvers

Figure 4. Top: Execution time of 2D diffusion benchmark problem on one node of Summit (i.e., CPU problems use all 40 CPU cores, and GPU problems use all 6 V100 GPUs). Bottom: Execution time of 2D diffusion benchmark problem on one node of Spock (60 CPU cores and 4 MI100 GPUs). The CVODE backward differentiation formula integrator is used with a preconditioned conjugate gradient (PCG) linear solver.