Skip to main content

Table 5 Computational times of the SW algorithm depending on the number of GPUs used

From: Protein alignment algorithms with an efficient backtracking routine on multiple GPUs

avg.length

# of sequences

1 GPU

2 GPUs

3 GPUs

4 GPUs

  

time

speedup

time

speedup

time

speedup

time

speedup

51

4000

8,06

1,000

4,04

1,997

2,71

2,971

2,07

3,896

 

8000

31,11

1,000

15,57

1,998

10,39

2,995

7,86

3,961

 

12000

69,08

1,000

34,56

1,999

23,07

2,994

17,44

3,961

154

2000

17,93

1,000

9,05

1,982

6,10

2,939

4,61

3,890

 

4000

67,75

1,000

34,21

1,981

22,93

2,955

17,28

3,920

 

6000

149,03

1,000

74,66

1,996

50,14

2,973

37,62

3,961

257

2000

49,23

1,000

24,72

1,991

16,59

2,968

12,54

3,927

 

4000

186,44

1,000

93,80

1,988

62,80

2,969

47,31

3,941

 

6000

410,26

1,000

205,87

1,993

137,26

2,989

103,28

3,972

459

2000

155,63

1,000

78,16

1,991

52,45

2,967

39,48

3,942

 

4000

594,14

1,000

298,50

1,990

198,57

2,992

149,54

3,973

 

6000

1332,83

1,000

667,31

1,997

444,92

2,996

333,59

3,995

608

800

50,22

1,000

25,66

1,957

17,40

2,886

12,84

3,911

 

1200

106,84

1,000

53,91

1,982

35,77

2,987

27,41

3,898

 

1600

191,79

1,000

95,93

1,999

64,01

2,996

48,46

3,958

1103

800

164,78

1,000

83,84

1,965

56,83

2,900

41,95

3,928

 

1200

359,07

1,000

179,85

1,997

119,89

2,995

89,90

3,994

 

1600

628,85

1,000

314,85

1,997

209,78

2,998

158,70

3,963

  1. Mean computational times (in seconds) of the SW algorithm depending on the number of GPUs used. The algorithm was run on Tesla S1070 in U1 rack case containing four graphics cards. The speedup refers always to the one GPU configuration.