[cig-commits] commit: Remove an unused variable and add some comments.

Mercurial hg at geodynamics.org
Fri Apr 6 10:29:02 PDT 2012


changeset:   149:32f74f34089c
user:        Walter Landry <wlandry at caltech.edu>
date:        Fri Apr 06 10:20:15 2012 -0700
files:       compute_coefficients/Interface.hxx
description:
Remove an unused variable and add some comments.


diff -r dd823a902bdc -r 32f74f34089c compute_coefficients/Interface.hxx
--- a/compute_coefficients/Interface.hxx	Mon Apr 02 20:06:55 2012 -0700
+++ b/compute_coefficients/Interface.hxx	Fri Apr 06 10:20:15 2012 -0700
@@ -11,7 +11,6 @@ public:
 public:
   FTensor::Tensor1<double,2> grid_pos, pos[2], corner_pos[2][2],
     anticorner_pos[2];
-  int corner_dir;
   bool intersect_dd[2], intersect_sides[2], intersect_corner[2][2],
     intersect_anticorner[2][2];
   bool intersect_dp;
@@ -31,6 +30,9 @@ public:
                                     const double &center_dist,
                                     const double dist[][ny])
   {
+    /* This has to be consistent with the checks for the second
+       derivatives.  Otherwise, the stencil weights may get too large
+       for off-center points */
     intersect_sides[0]=(sign(dist[i][j])!=sign(dist[i+1][j])
                         && sign(dist[i][j+1])!=sign(dist[i+1][j+1])
                         && sign(dist[i][j])==sign(dist[i][j+1]));
@@ -50,6 +52,12 @@ public:
               tangent(0)=-norm(1);
               tangent(1)=norm(0);
 
+              /* Use the intersection with the element boundary, not
+                 the straight line from the center of the element to
+                 the corner element.  This may cause problems when the
+                 interface just barely skims the element.  We did it
+                 this way because then the weighting of the stencil
+                 always sums to zero.  May want to change this.  */
               /* TODO: this assumes straight boundaries */
               if(std::fabs(tangent(0))>std::fabs(tangent(1)))
                 {



More information about the CIG-COMMITS mailing list