Integrating optimized load balancing and asynchronous parallel I/O enabled human cardiac electrophysiology simulations on up to 16,384 compute cores, reducing the realtime lag factor to 240.
High-performance computing enables near-realtime simulation of human cardiac electrophysiology, making in silico experimentation feasible for clinical workflows.
Absolute Event Rate: 240% vs 5600%
In this study, the feasibility of conducting in silico experiments in near-realtime with anatomically realistic, biophysically detailed models of human cardiac electrophysiology is demonstrated using a current national high-performance computing facility. The required performance is achieved by integrating and optimizing load balancing and parallel I/O, which lead to strongly scalable simulations up to 16,384 compute cores. This degree of parallelization enables computer simulations of human cardiac electrophysiology at 240 times slower than real time and activation times can be simulated in approximately 1 min. This unprecedented speed suffices requirements for introducing in silico experimentation into a clinical workflow.
Niederer et al. (Sat,) conducted a other in Left bundle branch block (computational model) (n=1). Optimized load balancing and asynchronous parallel I/O vs. Conventional parallelization techniques (inline I/O and nodal-based partitioning) was evaluated on Realtime lag factor (jr). Integrating optimized load balancing and asynchronous parallel I/O enabled human cardiac electrophysiology simulations on up to 16,384 compute cores, reducing the realtime lag factor to 240.