Table 3 The experimental performance results for mass-spring simulation

# of nodes

3858

4737

7282

10,223

20,587

38,233

# of springs

72,951

90,045

142,662

201,072

408,870

759,330

Semi-Euler integration (ms)

 CPU only

33.43

52.68

83.49

112.92

254.83

469.93

 Multi-threaded CPU

24.94

29.74

41.28

95.82

184.28

331.97

 GPU + multi-threaded CPU

10.93

15.29

23.22

31.89

68.38

127.45

Midpoint integration (ms)

 CPU only

41.39

58.15

86.82

130.52

245.70

472.29

 Multi-threaded CPU

28.84

33.92

53.82

98.53

152.94

282.93

 GPU + multi-threaded CPU

11.21

16.84

24.53

34.18

71.82

130.84

4th-order Runge-Kutta integration (ms)

 CPU only

49.24

66.25

94.04

145.29

259.39

513.86

 Multi-threaded CPU

30.84

38.13

59.83

103.21

167.32

325.41

 GPU + multi-threaded CPU

13.34

19.42

27.92

35.33

73.41

138.98