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 3

Computing times and their corresponding speed-up factors obtained with HT enabled for the compact and scatter affinities. HT: hyper threading, V: vectorization, and A: activated.

Threads	HT-A, V-A, compact	HT-A V-A, scatter
1	43 h 22 m 22 s (1.00X)	43 h 19 m 58 s (1.00X)
2	27 h 17 m 46 s (1.59X)	24 h 05 m 16 s (1.80X)
3	21 h 38 m 07 (2.00X)	15 h 41 m 05 (2.76X)
4	14 h 37 m 51 s (2.96X)	12 h 15 m 17 s (3.54X)
5	12 h 12 m 27 s (3.55X)	11 h 18 m 27 s (3.83X)
6	11 h 44 m 25 s (3.69X)	08 h 55 m 26 s (4.86X)
7	11 h 36 m 13 s (3.74X)	07 h 46 m 44 s (5.57X)
8	10 h 15 m 49 s (4.23X)	07 h 21 m 59 s (5.88X)
9	09 h 49 m 17 s (4.42X)	06 h 31 m 26 s (6.64X)
10	9 h 51 m 28 s (4.40X)	06 h 16 m 28 s (6.91X)
11	09 h 26 m 09 s (4.60X)	05 h 33 m 41 s (7.79X)
12	09 h 27 m 33 s (4.59X)	06 h 03 m 58 s (7.14X)
13	09 h 01 m 54 s (4.8X)	06 h 29 m 24 s (6.68X)
14	08 h 37 m 13 s (5.03X)	07 h 01 m 00 s (6.18X)
15	08 h 12 m 43 s (5.28X)	06 h 27 m 47 s (6.70X)
16	07 h 56 m 28 s (5.46X)	06 h 31 m 54 s (6.63X)
17	07 h 35 m 52 s (5.71X)	06 h 26 m 55 s (6.72X)
18	07 h 27 m 38 s (5.81X)	06 h 35 m 04 s (6.58X)
19	07 h 09 m 47 s (6.06X)	06 h 20 m 00 s (6.84X)
20	06 h 26 m 38 s (6.73X)	06 h 19 m 51 s (6.84X)
21	06 h 18 m 32 s (6.87X)	06 h 12 m 58 s (6.97X)
22	06 h 00 m 01 s (7.23X)	06 h 20 m 40 s (6.83X)
23	05 h 51 m 48 s (7.40X)	06 h 24 m 12 s (6.77X)
24	05 h 51 m 49 s (7.40X)	06 h 27 m 04 s (6.72X)