Research and ACM SRC Posters Archive

Cache-Driven Super-Linear Scaling in a Python Fluid Solver


Poster Type: Research Posters

Author: Harshit Tiwari (New York University (NYU); IIT Kanpur, India), Samar Aseeri (King Abdullah University of Science and Technology (KAUST)), Katepalli R. Sreenivasan (New York University (NYU)), David E. Keyes (King Abdullah University of Science and Technology (KAUST)), Mahendra K. Verma (Indian Institute of Technology Kanpur)

Abstract: Turbulent convection transports heat in stars, planets, and the Earth’s atmosphere, and simulating it demands extreme-scale computing. We present DHARA, a modular fluid solver written in Python (NumPy, CuPy, mpi4py) with compiled stencil kernels for CPUs and GPUs. On the CPU-based Shaheen III (KAUST), DHARA scales to 786,432 cores for compressible convection, and its strong scaling is super-linear: adding cores speeds up each step by more than the ratio of the core counts. Profiling explains why. As the domain is split across more cores, each subdomain’s stencil working set drops into the fast L1/L2 caches, so cache misses collapse, leading to a better than doubling of the per-core throughput. The same code carries over to the Frontier (OLCF) and Polaris (ALCF) GPU systems with near-ideal strong scaling.

Best Poster Finalist (BP): no
Poster: PDF
Poster Summary: PDF


Back to Poster Archive Listing