AMROC Main     Blockstructured Adaptive Mesh Refinement in object-oriented C++

Explosion in a Box - 3 Grid Levels (density, isosurface at 1.8)

Problem description       Results:   3 Levels
previous     next

 

t=0
t=0.125
t=0.25









t=0.375
t=0.5



Benchmark

Adaptive computation with 3 grid levels
Task
P=1
P=2
P=4
s
%
s
%
s
%
Integration
 26392
  94.7
 12486
  84.3
  6166
  79.3
Flux correction
   428
   1.5
   670
   4.5
   423
   5.4
Boundary setting
   233
   0.8
  1066
   7.2
   809
  10.4
Recomposition
   402
   1.4
   323
   2.2
   207
   2.7
Clustering
   165
   0.6
    93
   0.6
    52
   0.7
Misc.
   250
   0.9
   171
   1.2
   109
   1.5
Total / Parallel Efficiency
 27870
 100.0
 14810
  94.1
  7766
  89.7

Uniform refinement
Task
P=1
P=2
P=4
s
%
s
%
s
%
Integration
 74957
  99.1
 16884
  80.8
  8418
  97.7
Flux correction
     0
   0.0
     0
   0.0
     0
   0.0
Boundary setting
    15
   0.0
   105
   0.5
    41
   0.5
Recomposition
     0
   0.0
     0
   0.0
     0
   0.0
Clustering
     0
   0.0
     0
   0.0
     0
   0.0
Misc.
   634
   0.9
  3904
  18.7
   145
   1.8
Total / Parallel Efficiency
 75606
 100.0
 20894
 180.9
  8604
 219.7


In this example large regions of the computational domain require adaption. The mere computational time of an adaptive and a uniformly refined calculation are similar. But, the adaptive computation reduces the memory demands. Therefore, the adaptive computation is much faster on a single node (P=1), but nearly equal times are achieved on four nodes (P=4).

Comparison between adaptive and uniformly refined computation

top

Quickstart       Users Guide     Programmers Reference     Installation      Examples     Download


AMROC Main      Home      Contact
last update: 6/1/04