As the push towards exascale hardware has increased the diversity of system architectures, performance portability has become a critical aspect for scientific software. We describe the Kokkos Performance Portable Programming Model that allows developers to write single source applications for diverse high-performance computing architectures. Kokkos provides key abstractions for both the compute and memory hierarchy of modern hardware. We describe the novel abstractions that have been added to Kokkos version 3 such as hierarchical parallelism, containers, task graphs, and arbitrary-sized atomic operations to prepare for exascale era architectures. We demonstrate the performance of these new features with reproducible benchmarks on CPUs and GPUs.

Diffusion Tensor Imaging (DTI) is a non-invasive magnetic resonance technique that produces in vivo images of biological tissues with local microstructural characteristics such as water diffusion. It can be used, for example, to localize white matter lesions, or in neuro-navigation surgery of brain tumors. Diffusion tensor maps are usually computed on a voxel-by-voxel basis by fitting the signal intensities of diffusion weighted images as a function of their corresponding data acquisition parameters. This processing is highly computation-intensive and can be time-consuming which constraints the clinical use of DTI. This study presents the application of using high performance GPU clusters in diffusion tensor estimation by accelerating the multivariate non-linear regression. The results are tested in simulated DTI brain datasets and show significant performance gain in tensor fitting in addition to favorable scalability characteristics. The proposed GPU implementation framework can further promote the clinical use of DTI, and can be used to accelerate statistical analysis of DTI where Monte Carlo simulations are employed, or readily applied to quantitative assessment of DTI using bootstrap analysis.