Skip to main content

Table 2 Difference of parallelization among GPU and MIC offload and native implementations (n th is number of MIC threads).

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

Target of parallelization #threads used for one ligand angle Consumption of accelerator memory
GPU Each process in one ligand angle All GPU threads For only one ligand
MICoffload Each process in one ligand angle All MIC threads For only one ligand
MICnative Loop of rotational angles of ligand One MIC thread For n th ligands