A fine-grained parallelization of the immersed boundary method

TL;DR

This paper introduces new parallel algorithms for the immersed boundary method that leverage key-value sort and segmented reduce primitives, achieving near-ideal scaling on GPUs and multicore CPUs.

Contribution

The paper presents novel parallel algorithms for Eulerian-Lagrangian interactions in the immersed boundary method using well-studied primitives, enabling efficient execution on GPUs and shared memory systems.

Findings

Near-ideal scaling on GPUs

Efficient algorithms for scattered points and elastic structures

Applicable to both GPUs and multicore CPUs

Abstract

We present new algorithms for the parallelization of Eulerian-Lagrangian interaction operations in the immersed boundary method. Our algorithms rely on two well-studied parallel primitives: key-value sort and segmented reduce. The use of these parallel primitives allows us to implement our algorithms on both graphics processing units (GPUs) and on other shared memory architectures. We present strong and weak scaling tests on problems involving scattered points and elastic structures. Our tests show that our algorithms exhibit near-ideal scaling on both multicore CPUs and GPUs.