Skip to main content

Advertisement

Table 8 The overhead of the backtracking procedure

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

avg. length # of sequences SW global NW semiglobal NW
   GCUPS BT share GCUPS BT share GCUPS BT share
51 4000 7,59 1,48 8,35 1,47 8,08 1,47
  8000 7,60 1,46 8,36 1,45 8,31 1,49
  12000 7,61 1,45 8,43 1,45 8,15 1,45
154 2000 8,05 1,50 8,77 1,49 8,89 1,53
  4000 8,24 1,48 9,11 1,48 9,09 1,51
  6000 8,31 1,47 9,17 1,46 9,39 1,52
257 2000 7,76 1,52 8,69 1,53 8,69 1,55
  4000 8,02 1,53 8,65 1,49 8,82 1,53
  6000 7,85 1,51 8,67 1,50 8,65 1,51
459 2000 7,03 1,65 7,84 1,62 7,79 1,63
  4000 7,21 1,58 8,17 1,59 8,01 1,59
  6000 7,19 1,55 8,07 1,56 7,91 1,54
  1. Mean performance (in GCUPS) for different versions of the score-only algorithm - without the backtracking procedure. The BT share columns refer to the overhead of corresponding algorithms with backtracking routine. Tests were run on a single GeForce GTX480.