DocumentCode
688252
Title
Solving the Cardiac Model Using Multi-core CPU and Many Integrated Cores (MIC)
Author
Jing Yang ; Jun Chai ; Mei Wen ; Nan Wu ; Chunyuan Zhang
Author_Institution
Coll. of Comput., Nat. Univ. of Defence Technol., Changsha, China
fYear
2013
fDate
13-15 Nov. 2013
Firstpage
1009
Lastpage
1015
Abstract
Simulation of the cardiac model through mathematical methods is a computationally demanding task, due to its higher spatial and temporary resolution requirements. Adopting the hybrid CPU and MIC programming platform may meet these requirements, since the brand new KNC provides higher computational power. We explored the hybrid CPU and MIC computational model based on a chosen PDE-ODE Solver. First, we adopt the parallel strategies including OpenMP, AVX, memory align to our experimental case on the CPU. Then we complete offload mode on our hybrid platform and acquired a speedup of 24. In the end, we complete the COI mode aiming to overcome the startup overhead of offload mode and acquired the computational performance of 61.11 DP Gflops on our CPU with single MIC platform and the speedup has reached 24.5.
Keywords
cardiology; medical computing; message passing; multiprocessing systems; parallel processing; partial differential equations; AVX; KNC; MIC programming platform; OpenMP; PDE-ODE Solver; cardiac model; computational performance; hybrid CPU; many integrated cores; mathematical methods; memory align; multicore CPU; offload mode; parallel strategies; spatial resolution requirements; temporary resolution requirements; Computational modeling; Coprocessors; Data transfer; Handheld computers; Mathematical model; Message systems; Microwave integrated circuits; AVX; COI; MIC; SCIF; offload;
fLanguage
English
Publisher
ieee
Conference_Titel
High Performance Computing and Communications & 2013 IEEE International Conference on Embedded and Ubiquitous Computing (HPCC_EUC), 2013 IEEE 10th International Conference on
Conference_Location
Zhangjiajie
Type
conf
DOI
10.1109/HPCC.and.EUC.2013.143
Filename
6832025
Link To Document