Closed campreilly closed 1 year ago
Adding a note here only, as Sean and I had talked about this can be accomplished with functors. I have a few concepts that could address this concern, but will not begin development at this time.
I was thinking that I'd steered you in the wrong direction slighting. The change to this line:
initialize<DATA_TYPE>::zero(dy0, dy1, dy2, dy3) ;
should be
initialize<DATA_TYPE>::zero(dy0, dy1, dy2, dy3,*_data) ;
which will properly address the known issue with VS2010 and provide a sizing element required by the functor.
Had some valgrind errors.
Fixed valgrid errors.
There a cases where it can be useful to interpolate more than one ordinate using the same abscissa. However, There are several instances were we assume that the abscissa is the same date type as the ordinate. It may be much faster to interpolate these items in parallel than it is to create a separate data_grid for each ordinate.
This is pretty easy to do for the linear version of interpolation
Before the change:
After the change:
But, is much harder for pchip, basically because of the cases where we check the sing of the slope. I'm also unclear on how it should effect the ability to write to a netCDF file.
So we'll put this off for another day.