PCMS: Parallel Coupler For Multimodel Simulations

TL;DR

PCMS addresses the challenge of coupling diverse simulation codes at exascale by delivering a GPU-accelerated, generalized coupler with a two-part architecture (server plus lightweight shims) and multiple field-transfer strategies, including local weighted polynomial fitting and mesh-intersection projections. It enables volume coupling across high-dimensional, nonconformal meshes while preserving conservation where possible and supports distributed control via rendezvous-based communication and ADIOS2 interfaces. The authors demonstrate realistic, multi-physics couplings (XGC/DEGAS2; GNET/GTC) and a 5D GNET–GTC coupling, achieving scalable performance on Frontier with strong IO engine improvements and overall weak-scaling efficiency around 85%. The work advances fusion-relevant simulations by enabling flexible workflows, integration with management tools, and plans for future enhancements such as time-step control and axisymmetric lifting.

Abstract

This paper presents the Parallel Coupler for Multimodel Simulations (PCMS), a new GPU accelerated generalized coupling framework for coupling simulation codes on leadership class supercomputers. PCMS includes distributed control and field mapping methods for up to five dimensions. For field mapping PCMS can utilize discretization and field information to accommodate physics constraints. PCMS is demonstrated with a coupling of the gyrokinetic microturbulence code XGC with a Monte Carlo neutral transport code DEGAS2 and with a 5D distribution function coupling of an energetic particle transport code (GNET) to a gyrokinetic microturbulence code (GTC). Weak scaling is also demonstrated on up to 2,080 GPUs of Frontier with a weak scaling efficiency of 85%.

Figures (12)

• Figure 1: Overview of the PCMS components and operations reprinted from Shephard_2024. Each application uses native datastructures and the coupler uses an intermediate representations. Inter application communication is performed with ADIOS2 and EFFIS provides job control.
• Figure 2: Radial basis functions implemented in PCMS (a = 2).
• Figure 3: Example mesh and field of WEST tokamak geometry obtained courtesy of Davide Curreli.
• Figure 4: Error of the linear field recovery when mapping from element centroids to vertices using C4 RBF.
• Figure 5: Error of the quadratic field recovery when mapping field A from element centroids to vertices using C4 RBF.