Skip to main content

Table 6 Docking calculation time results for the protein complex (PDB ID: [PDB:1JK9]) for each process (in seconds).

From: Protein-protein docking on hardware accelerators: comparison of GPU and MIC architectures

  1CPU 8CPUs GPU MICoffload MICnative
P1. Initialization 0.0 0.0   0.8   4.0   0.7  
P2. Receptor voxel 0.3 0.3 (1.1×) 0.3 (1.1×) 0.3 (1.1×) 4.4 (0.1×)
P3. Receptor FFT 0.1 0.1 (1.0×) 0.0 (1.7×) 1.0 (0.1×) 0.3 (0.2×)
P4. Ligand rot & voxel 12.9 3.4 (3.8×) 2.3 (5.5×) 7.4 (1.7×) 1.2 (11.1×)
P5. Ligand FFT 69.8 14.2 (4.9×) 2.2 (31.1×) 15.1 (4.6×) 7.9 (8.9×)
P6. Convolution 27.4 4.6 (5.9×) 1.1 (25.6×) 13.9 (2.0×) 3.6 (7.7×)
P7. Inverse FFT 71.5 14.1 (5.1×) 2.2 (31.8×) 15.2 (4.7×) 8.3 (8.6×)
P8. Identifying the bests 4.3 1.7 (2.5×) 1.7 (2.5×) 9.8 (0.4×) 0.3 (12.5×)
P9. Post processes 0.0 0.0   0.0   0.0   0.0  
Data transfer     0.6   10.1    
Total 186.4 38.5 (4.8×) 10.8 (17.3×) 77.0 (2.4×) 26.8 (7.0×)
  1. The values shown in parentheses are the acceleration rates relative to one CPU core. MIC native used 171 threads.