Galactos

• DOI: 10.1145/3126908.3126927
• Publication Date: 2017-11-08T21:02:30Z
• Authors: Brian Friesen; Md. Mostofa Ali P...; Brian Austin; Nadathur Satish; Zachary Slepian; Narayanan Sundaram; Deborah Bard; Daniel J. Eisenstein; Jack Deslippe; Pradeep Dubey; Prabhat
• Keywords: Computational Engineering, Finance, and Science (cs.CE); Performance (cs.PF); FOS: Computer and information sciences; Computer Science - Performance; Cosmology and Nongalactic Astrophysics (astro-ph.CO); FOS: Physical sciences; Computer Science - Computational Engineering, Finance, and Science; Astrophysics - Cosmology and Nongalactic Astrophysics

The nature of dark energy and the complete theory gravity are two central questions currently facing cosmology. A vital tool for addressing them is 3-point correlation function (3PCF), which probes deviations from a spatially random distribution galaxies. However, 3PCF's formidable computational expense has prevented its application to astronomical surveys comprising millions billions We present Galactos, high-performance implementation novel, O(N2) algorithm that uses load-balanced k-d tree spherical harmonic expansions compute anisotropic 3PCF. Our optimized Intel Xeon Phi architecture, exploiting SIMD parallelism, instruction thread concurrency, significant L1 L2 cache reuse, reaching 39% peak performance on single node. Galactos scales full Cori system, achieving 9.8 PF (peak) 5.06 (sustained) across 9636 nodes, making 3PCF easily computable all galaxies in observable universe.