A Multi-GPU-Accelerated Sharp-Interface Immersed-Boundary Solver for Extreme-Scale Flow Simulation with Moving Boundary

Speaker: Sushrut Kumar, Johns Hopkins University
July 16, 2026

Presentation

Abstract

Direct numerical simulation of flows with moving and deforming boundaries remains one of the most computationally demanding problems in CFD, and the irregular data structures, dynamic sparsity, and time-varying operators inherent to immersed boundary methods have historically made them resistant to efficient GPU acceleration. In this talk, I will describe the development of a fully GPU-resident, multi-GPU version of the sharp-interface immersed boundary solver ViCar3D and the algorithmic innovations that make extreme-scale fluid–structure interaction (FSI) simulations practical. The solver couples a hybrid OpenACC/CUDA programming model with GPU-optimized geometric kernels, a matrix decomposition to isolate the dynamically evolving immersed-boundary operators, an asynchronous dual-stream sparse linear solver, and a sparse communication layer built on GPU-aware MPI, NCCL, and NVSHMEM. I will present verification and performance results demonstrating speedups exceeding 20× relative to CPU implementations, along with weak scaling up to 12 billion grid points on the NVIDIA GB200 NVL72 platform. Finally, I will show how this capability enables high-Reynolds-number simulations of moving-boundary and FSI problems at previously impractical resolutions, with applications ranging from biological and bioinspired locomotion to incompressible aerodynamics.

Biography

Sushrut Kumar received his Ph.D. in Mechanical Engineering from Johns Hopkins University, where he worked in Prof. Rajat Mittal's Flow Physics and Computation Laboratory. His doctoral research focused on the multi-GPU implementation of the sharp-interface immersed boundary solver ViCar3D, enabling extreme-scale simulations of flows with moving boundaries and fluid–structure interaction. His work also includes studies of biological and bioinspired locomotion, such as the aero-structural dynamics of bat-inspired membrane wings and flexible fish fins, published in venues including the Journal of Fluid Mechanics, Bioinspiration & Biomimetics, and the AIAA SciTech and Aviation forums. He previously interned at NVIDIA, where he developed neural operators for thermo-fluid simulations in digital twins.

Seminar Participation

Attending In Person

For select seminars, you may wish to attend in-person if you are in the vicinity of NASA Ames Research Center. These AMS seminars are held in Building N258 in Room 127, just off the front lobby. Sessions available to attend in-person will be indicated in the announcement information. In-person participants without NASA badges will need a valid visitor badge (US citizenship or permanent residency is required) to enter the facility.

Join Us Online

To attend an AMS seminar, click on the Microsoft Teams link in the seminar description. You can connect through the desktop Teams App if you have it, or you can select "Watch on the Web Instead" to enter the meeting as a guest. Please ensure that you've muted your microphone in Teams before you select "Join Now" to enter the meeting lobby.