Skip to main content

Table 1 Impact of grid map reuse and OpenMP multithreading on I/O.

From: Multilevel Parallelization of AutoDock 4.2

  VN (gm = reload) VN (gm = reuse) SMP (gm = reuse)
  FLD MAP LIG LOG FLD MAP LIG LOG FLD MAP LIG LOG
ave 2.5 25.9 90.0 150.5 0.3 5.1 21.2 12.0 0.0 0.6 0.3 0.3
rms 8.0 50.8 335.8 584.1 1.5 13.8 40.8 55.4 0.1 2.4 0.3 0.3
med 0.3 13.4 1.9 0.2 0.0 0.0 1.5 0.3 0.0 0.0 0.2 0.2
max 247.2 542.3 3567.5 3630.0 54.6 187.9 469.6 546.8 1.7 14.0 2.9 3.2
%      14.6 19.7 23.6 8.0 1.1 2.5 0.4 0.3
  1. A 32,000 copy indinavir library was docked with mpAD4 on a 2,048(8,192) node(core) BG/P system in VN node mode (MPI + OpenMP with 1 OpenMP thread/node, 4× virtual nodes) with either grid map reloading or reuse, and SMP mode (MPI + OpenMP with 4 OpenMP threads/node) with grid map reloading or reuse. Statistics for total duration (in seconds) of .fld map load time (FLD), atomic affinity map load time (MAP), ligand load time (LIG), and docking log write time (LOG) are shown for VN (gm = reload), VN (gm = reuse) and SMP (gm = reuse). Pairwise comparisons of average I/O times (seconds) between VN (gm = reload) and VN (gm = reuse) or SMP (gm = reuse) were calculated, and are shown as percent of VN (gm = reload) times. I/O differences between SMP (gm = reload) and (gm = reuse) were negligible, and are not shown.