Table 3 Speedup for alignment stages of multithreaded muBLASTP over multithreaded NCBI BLAST on Haswell platform (with 24 threads) and Nehalem platform (with 12 threads) with query batches of different query length
From: muBLASTP: database-indexed protein sequence search on multicore CPUs
Query length | Haswell platform | Â | Nehalem platform | Â | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Â | uniprot_sprot | env_nr | nr | uniprot_sprot | env_nr | nr | ||||||
| Â | Aver | SD | Aver | SD | Aver | SD | Aver | SD | Aver | SD | Aver | SD |
100 | 5.7 | 0.013 | 2.67 | 0.013 | 1.94 | 0.021 | 8.2 | 0.013 | 4.52 | 0.017 | 8.59 | 0.015 |
500 | 3.22 | 0.012 | 1.44 | 0.012 | 1.39 | 0.014 | 2.25 | 0.018 | 1.69 | 0.018 | 2.69 | 0.016 |
1000 | 2.85 | 0.016 | 1.35 | 0.013 | 1.20 | 0.015 | 1.76 | 0.018 | 1.41 | 0.016 | 1.72 | 0.015 |
Mixed | 4.16 | 0.013 | 1.41 | 0.012 | 1.49 | 0.012 | 2.82 | 0.018 | 2.34 | 0.013 | 4.1 | 0.015 |