High-performance cardiac electrophysiology: Performance and scalability of GPU and MPI open-source simulators

High-performance computing techniques are essential to address the growing computational demands of cardiac electrophysiology modeling, particularly in the context of large-scale, data-driven, and clinically relevant simulations. This work evaluates the computational performance of two widely used open-source cardiac electrophysiology simulators, MonoAlg3D and openCARP , representing distinct parallelization paradigms based on GPU acceleration and distributed MPI execution, respectively. Performance is assessed in two representative scenarios: a well-established benchmark problem and a realistic biventricular simulation. While both simulators produce results in numerical agreement, their performance profiles reveal complementary architectural strengths. MonoAlg3D achieves high computational efficiency on single GPU nodes, delivering performance comparable to openCARP executed on a large number of CPU cores, whereas openCARP demonstrates strong scalability on distributed-memory HPC clusters. These results highlight how different high-performance architectures and parallelization strategies can address specific computational challenges in cardiac electrophysiology, providing guidance for the selection of simulation tools in large-scale and clinically oriented computational pipelines.