The paper “Petascale XCT: 3D Image Reconstruction with Hierarchical Communications on Multi-GPU Nodes” has garnered the Best Paper Award at SC20. The research team is led by Argonne researcher Tekin Bicer and PhD candidate Mert Hidayetoglu at the University of Illinois at Urbana-Champaign (UIUC) and comprising Simon Garcia de Gonzalo (Barcelona Supercomputing Center, BSC), Bin Ren (College of William & Mary), Vincent De Andrade (Argonne), Doga Gursoy (Argonne), Rajkumar Kettimuthu (Argonne), Ian Foster (Argonne) and Wen-mei Hwu (UIUC).
Their work involves three novel optimizations: (1) extends optimization of (back)projection operators for 2D memory-centric approach to 3D; (2) inclusion of hierarchical communications by exploiting “fat-node” architecture with many GPUs; (3) utilization of mixed-precision types while preserving convergence rate and quality.
The team extensively evaluated the proposed optimizations and scaling on the Summit supercomputer. Their largest reconstruction is a mouse brain volume with 9K×11K×11K voxels, where the total reconstruction time is under three minutes using 24,576 GPUs, reaching 65 PFLOPS; 34% of Summit's peak performance.
BSC researcher, Simon García de Gonzalo, states that “this lays the groundwork for faster reconstruction of large images, helping to speed up diagnostic results related to medicine and scientific research”.
Further information here.