Table 1.
Test results for three data sets using different numbers of processors p Running time of the parallelized version of DIALIGN on protein sequences (average length: 381 residues) using different numbers p of processors. The values for p = 1 are measured with DIALIGN, the others with ParDIALIGN. All running times in the table are the fastest achieved values of several runs. The speed-up value is defined as the time needed for p = 1 divided by the time for ParDIALIGN. The percentages show the reduction of the time with respect to the sequential version of DIALIGN.
| 20 sequences | 55 sequences | 100 sequences | ||||
| p | time (sec) | speed-up | time (sec) | speed-up | time (sec) | speed-up |
| 1 | 37.69 | - | 334.17 | - | 1283.80 | - |
| 4 | 10.40 | 3.62 (72.41%) | 104.61 | 3.19 (68.70%) | 552.68 | 2.32 (56.95%) |
| 8 | 5.93 | 6.36 (84.27%) | 65.69 | 5.09 (80.34%) | 421.69 | 3.04 (67.15%) |
| 16 | 3.57 | 10.56 (90.53%) | 46.25 | 7.23 (86.16%) | 359.50 | 3.57 (72.00%) |
| 32 | 2.65 | 14.22 (92.97%) | 36.92 | 9.05 (88.95%) | 332.22 | 3.86 (74.12%) |
| 64 | 1.95 | 19.32 (94.82%) | 30.74 | 10.87 (90.80%) | 309.43 | 4.15 (75.90%) |