This allows to disable the 1-elem-block optimization to be able to compare the running times with and with out it.