Research Article

Performance of a Code Migration for the Simulation of Supersonic Ejector Flow to SMP, MIC, and GPU Using OpenMP, OpenMP+LEO, and OpenACC Directives

Table 4

Computing times and the corresponding speed-up factors obtained by using OpenMP mounted on the Xeon Phi coprocessor. D1: design 1, D2: design 2, C: collapsed, and V: vectorized.

Threads D1-C D2-C D2-V

59 37 h 07 m 47 s (1.00X) 27 h 47 m 51 s (1.34X) 28 h 52 m 20 s (1.29X)
118 18 h 54 m 38 s (1.91X) 18 h 54 m 38 s (1.96X) 24 h 43 m 53 s (1.50X)
177 20 h 09 m 13 s (1.84X) 19 h 24 m 13 s (1.91X) 22 h 24 m 06 s (1.66X)
236 16 h 40 m 19 s (2.23X) 16 h 10 m 10 s (2.30X) 22 h 07 m 53 s (1.68X)