Table 2.
Execution time in seconds for various numbers of microRNA sequences against a reference dataset.
| # of queries | 1 GPU | 2 GPUs | 4 GPUs | miRanda on CPU | 4GPUs vs. CPU | |||
|---|---|---|---|---|---|---|---|---|
| Time | GCUPs | Time | GCUPs | Time | GCUPs | Time | Speedup | |
| 1 | 2.48 | 0.06 | 1.44 | 0.1 | 1.25 | 0.12 | 7 | 5x |
| 32 | 2.59 | 1.8 | 1.92 | 2.44 | 1.39 | 3.38 | 177 | 127x |
| 64 | 5.08 | 1.79 | 2.86 | 3.17 | 1.97 | 4.62 | 288 | 146x |
| 128 | 9.89 | 1.79 | 5.58 | 3.17 | 3.63 | 4.89 | 580 | 159x |
| 256 | 19.25 | 1.8 | 10.49 | 3.3 | 6.47 | 5.36 | 1076 | 166x |
The reference dataset contained 256 sequences with average length of 20K. The orignial miRanda 3.3 on CPU and our proposed CUDA-miRanda on GPUs were compared.