Changeset 1506

Timestamp:

Jun 9, 2005, 11:23:16 AM (19 years ago)

Author:

matthew

Message:

The re-initialization of semi_implicit_update[k] to 0.0 is now done in the _update C function of quantity_ext.c, instead of the method update_conserved_quantities of domain.py. For run_profile.py, with N=128, the time spend in update_conserved_quantities is cut from 14.00 secs to 8.31 secs (averaged over three runs).

Location:

inundation/ga/storm_surge/pyvolution

Files:

• domain.py (modified) (1 diff)
• quantity_ext.c (modified) (3 diffs)

inundation/ga/storm_surge/pyvolution/domain.py

@@ -609 +609 @@
             #Clean up
             #Note that Q.explicit_update is reset by compute_fluxes
-            Q.semi_implicit_update[:] = 0.0
+
+            #MH090605 commented out the following since semi_implicit_update is now re-initialized
+            #at the end of the _update function in quantity_ext.c (This is called by the
+            #preceeding Q.update(timestep) statement above).
+            #For run_profile.py with N=128, the time of update_conserved_quantities is cut from 14.00 secs
+            #to 8.35 secs
+            
+            #Q.semi_implicit_update[:] = 0.0 
 
     def update_ghosts(self):

inundation/ga/storm_surge/pyvolution/quantity_ext.c

@@ -195 +195 @@
   for (k=0; k<N; k++) {
     denominator = 1.0 - timestep*semi_implicit_update[k];
-
     if (denominator == 0.0) {
       return -1;
@@ -203 +202 @@
     }
   }
+  //MH080605 set semi_impliit_update[k] to 0.0 here, rather than in update_conserved_quantities.py
+  for (k=0;k<N;k++)
+    semi_implicit_update[k]=0.0;
   return 0;
 }
@@ -227 +229 @@
 
   N = centroid_values -> dimensions[0];
-
 
   err = _update(N, timestep,