A 4-year $2.6 million grant from the National Institute of Biomedical Imaging and Bioengineering (NIBIB), to perform “real-time ” CT imaging dose calculations (2012 – 2016) 1 Participants: • RPI - Xu, Ji, Carothers, and Shephard • Mass General Hospital – Kalra and Liu • GE Global Research – FitzGerald • LANL - Brown
Transcript
Slide 1
A 4-year $2.6 million grant from the National Institute of
Biomedical Imaging and Bioengineering (NIBIB), to perform real-time
CT imaging dose calculations (2012 2016) 1 Participants: RPI - Xu,
Ji, Carothers, and Shephard Mass General Hospital Kalra and Liu GE
Global Research FitzGerald LANL - Brown
Slide 2
Introduction Monte Carlo radiation computing is the gold
standard, but time-consuming Traditional parallel schemes use CPUs
Multiprocessing multithreading Hardware accelerators are emerging
GPU Coprocessor 2
Slide 3
Exa-scale HPC depends on hardware accelerators (Among Top 10
supercomputer as of June 17, 2013) rankName RmaxRpeak Config 1
Tianhe-2 33.9 PF54.9 PF 32,000 Intel Xeon E5-2692 (12-core) 48,000
Intel Xeon Phi coprocessor 31S1P 2 Titan 17.6 PF27.1 PF 18,688 AMD
Opteron 6274 (16-core) 18,688 NVIDIA K20x GPU
Slide 4
GPU offers: - Massive data-parallel computing power - Cost and
energy efficiency - Flexible programming architecture (CUDA) Stream
Processors Single Instruction, Multiple Threads (SIMT)