... While in the `first-solution-case' node expansion can be stopped after a first
solution is found,
all processors must finish searching their
current subtree in the `all-solutions-case'.
Due to different work packet sizes,
which vary most in the last iteration, some processors might
get idle while others are still expanding their last tree.
Most of the overheads in 
can be reduced by implementing a
stack-splitting strategy as in [Kumar and Rao, 1990].