Re: patch to constrained-breaking

[Joe Neeman]

On Sat, 11 Mar 2006 06:13 am, Han-Wen Nienhuys wrote:
> Joe Neeman wrote:
> > This contains a bug-fix, an optimisation (been doing some profiling :))
> > and a
> >
> > +   (combine_demerits): use pointers to reduce copying data
>
> Are you sure about this? AFAIK references are internally handled in the
> same way as pointers.
The actual optimisations were in calc_subproblem. But I needed to be able to 
pass an empty value to combine_demerits. Using a null pointer instead of a 
reference allowed me to use pointers instead of values in calc_subproblem.

Column_x_positions::operator= used to be disturbingly high in the call graph 
part of the profile, but now it's pretty much gone.