Change when conversion Realf -> Real happens in acceleration.

This is faster, but leads to slightly different numerical results as the accumulation is now done in single and not double precision. Difference within expected range, accuracty of SP.

.gitignore

@@ -1,4 +1,5 @@
 *.o
 version.cpp
 \#*\#
-*~
+*~
+vlasiator

vlasovsolver/cpu_acc_map.hpp

@@ -391,7 +391,8 @@ bool map_1d(SpatialCell* spatial_cell,
 	  for(uint target_i = 0; target_i < 4;target_i ++ ){
 	    const uint tblock=target_block[target_i];
 	    const uint tcell=target_cell[target_i];
-	    const Real tval=target_density[target_i];
+            /*do the conversion from Real to Realf here, faster than doin in accumulation*/
+	    const Realf tval=target_density[target_i];
             /*check that we are within sane limits. If gk is negative,
              * or above blocks_per_dim * blockcells_per_dim then we
              * are outside of the target grid.*/