sandbox/Antoonvh/refineing3d.c

A Test of prolongation attributes in 3 dimensions

We test the accuracy of some prolongation attributes for a function c(x,y,z) whos terms are linearly dependent on the x,y and z coordinate,

c(x,y,z)=1+x+y+z+xy+yz+xz+xyz.

#include "grid/octree.h"
#include "utils.h"
#include "additional-attributes.h"

#define func (1.+x+y+z+x*y+x*z+y*z+x*y*z)

scalar c[],err[];
c[top]   = dirichlet(func); // On coarse grid we limit the effect of the boundaries 
c[bottom]= dirichlet(func);
c[left]  = dirichlet(func);
c[right] = dirichlet(func);
c[front] = dirichlet(func);
c[back]  = dirichlet(func);
FILE * fp1;
int cells = 0;
stats s;
int maxlevel = 8;
int main(){
  fp1=fopen("data.dat","w");
  L0=10.;
  X0=Y0=-L0/2;
  init_grid(1<<maxlevel);
  foreach()
    c[]=func;

  for (int n=1;n<5;n++){
    cells=0;
    foreach(reduction(+:cells))
      cells++;
    fprintf(ferr,"%d\t %d\n",n,cells);
    fprintf(fp1,"%d\t%g\t",n,L0/(pow(2.,(double)(maxlevel-n+1.))));
    for (int j=1;j<6;j++){
      if (j==1) c.prolongation=refine_injection;
      if (j==2) c.prolongation=refine_bilinear;
      if (j==3) c.prolongation=refine_linear;
      if (j==4) c.prolongation=refine_quadratic; //Third order?
      if (j==5) c.prolongation=refine_3x3x3_linear;
      wavelet(c,err);
      foreach(){// We are not interested in cells at the boundary. 
	if (abs(x)<3. && abs(y)<3. &&abs(z)<3.) 
	  err[]=fabs(err[]);
	else
	  err[]=0.;
      }
      s=statsf(err);
      fprintf(fp1,"%g\t",s.sum+0.00000001);
    }
    fprintf(fp1,"\n");
    unrefine(level>=(maxlevel-n)); 
  }
}
Results from various interpolation/upsample techniques

Results from various interpolation/upsample techniques

As was found in 2D, the injection method is only first order accurate. As expected the second-order accurate bilinear technique is exact for first-order functions. However, the linear technique does not seem to display this property. This is because it neglecs variations that are not grid alligned (i.e. introduced by the last four terms of the equation above). The quadratic furmulation seems to perform better, but is not exact. The 3x3x3 linear that was designed to be exact for these functions is indeed exact. So apperently the weights are typed over correctly!