Skip to topic | Skip to bottom
Home

Start of topic | Skip to actions

High resolution performance results

24 processors, SHC

300x240 base grid

Method whole time integration time time steps cost/timestep (sec)
Roe- van Albada: (300x240 base grid 2,2,2,2 refinefinement, cfl max 0.9) 6861 1388 1033 1.3
Roe- van Albada: 2x Refined, (300x240 base grid 2,2,2,4 refinement, cfl max 0.9) 21030 8441 1089 7.8
Weno-TCD-bandwidth 4th order: pressure only dcflag, pcurv=1e7 24922 9748 1112 8.8
Weno-TCD-bandwidth 2nd order: pressure only dcflag, pcurv=3e7 20356 7430 1068 7.0
Weno-TCD-bandwidth 4th order: pressure only dcflag, pcurv=3e7 26128 9800 1114 8.8

  • WENO-TCD - 300x240 base grid 2,2,2,2 refinefinement, cfl max 0.85
  • "cost/timestep" = integration time / coarse level timesteps

150x120 base grid

Method whole time integration time time steps cost/timestep (sec) 1/2*(cost/timestep) (sec)
Roe- Min-Mod: inviscid, (150x120 base grid 2,2,4,2 refinement, cfl max 0.9) 3221 494 290 1.7 0.85
Roe- Min-Mod, (150x120 base grid 2,2,4,2 refinement, cfl max 0.9) 5604 967 495 2.0 1.0
Roe- van Albada, (150x120 base grid 2,2,4,2 refinvimement, cfl max 0.9) 6946 1627 518 3.1 1.55
Roe- van Albada: 2x Refined, (150x120 base grid 2,2,2,2,4 refinement, cfl max 0.9) 24607 8388 545 15.4 7.7

  • Note: To compare the 150x120 base grid results with the 300x240 base grid results, the 150x120 "cost/timestep" can be divided by 2 to include the extra 2x refinement that makes the finest grid scales equivalent.

DoubleMachReflectionStudy

You are here: Amroc > DoubleMachReflectionStudy > ConstantPropGFMPerformanceDMR

to top

Copyright © 1997-2024 California Institute of Technology.