Add ArborX Gordon Bell blogpost

content/blog/2025-12-hacc-gordon-bell.md

@@ -0,0 +1,59 @@
+---
+authors: ["kokkos-team"]
+title: Kokkos powers large-scale analysis in a Gordon Bell Prize finalist
+date: "2026-01-16"
+tags: ["blog"]
+thumbnail: img/blog/2026-arborx/thumbnail_alt.png
+---
+
+At SC25, the Association for Computing Machinery (ACM) announced `Cosmological
+Hydrodynamics at Exascale: A Trillion-Particle Leap in Capability` project as
+one of the six finalists of the annual [ACM Gordon Bell Prize](https://awards.acm.org/bell),
+awarded for outstanding achievement in high-performance computing.
+
+CRK-HACC, a premier cosmological hydrodynamics code built for extreme
+scalability, executed Frontier-E: a four trillion particle full-sky simulation,
+over an order of magnitude larger than previous efforts. The run achieved 513.1
+PFLOPs peak performance, processing 46.6 billion particles per second and
+writing more than 100 PB of data in just over one week of runtime.
+
+The achievement was possible in part due to in-situ analysis capability. A
+central component of this pipeline is the support for clustering-based analysis
+methods, including DBSCAN (Density-based Spatial Clustering of Applications
+with Noise) halo finding and SOD (Spherical Overdensity). Cluster finding is
+computationally intensive, requiring efficient spatial search and neighborhood
+queries across potentially hundreds of millions of particles per rank.
+
+To enable this at scale, the team relied on the publicly available ArborX library,
+which provides GPU-native spatial indexing and traversal routines. ArborX uses
+Kokkos for all of its parallel constructs and data structures. The team ran
+simulations on top supercomputers in the world, including Frontier (AMD GPUs),
+Perlmutter(Nvidia GPUs), and Aurora (Intel GPUs).
+
+The main index in ArborX is a bounding volume hierarchy (BVH). BVH is a tree
+structure created from a set of objects in a multi-dimension space. Fast BVH
+construction algorithms use a space-filling curve (Z-curve) to improve the
+spatial locality of the user data, followed by a single bottom-up construction
+to produce a binary tree structure (hierarchy). ArborX was developed from
+scratch using Kokkos as the on-node programming model. All the data in ArborX
+are stored as Kokkos views, and all the parallel kernels are implemented using
+Kokkos' parallel regions. The development is also guided by using Kokkos native
+profiling tools, as well as connectors to the vendor provided toolkits.
+
+{{< image src="img/blog/2026-arborx/analysis.pdf" style="float: center; height=10">}}
+
+Visualization of the performance impact of ArborX on analysis steps for a production gravity-only cosmology simulation.
+
+{{< image src="img/blog/2026-arborx/scaling.pdf" style="float: center; height=10">}}
+
+Strong (left axis, in red) and weak (right axis, in blue) scaling of the full
+application from 128 to 9,000 nodes on Frontier, with the lower panel showing
+efficiency relative to the ideal case. Weak scaling is presented in terms of
+the number of particles processed per second by the solver.
+
+
+### References
+
+- https://www.hpcwire.com/off-the-wire/ornl-largest-ever-universe-simulation-up-for-supercomputings-highest-prize/
+- "Cosmological Hydrodynamics at Exascale: A Trillion-Particle Leap in Capability", Frontiere et al., SC '25: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, Nov. 2025 (https://doi.org/10.1145/3712285.3771786)
+- "Advances in ArborX to support exascale applications", Prokopenko et al., The International Journal of High Performance Computing Applications. 2024;39(1):167-176. (https://doi.org/10.1177/10943420241298296)