Reimplemented MinMeanCycle to be much more efficient.
The new version implements Howard's algorithm instead of Karp's algorithm and
it is at least 10-20 times faster on all the 40-50 random graphs we have tested.

#!/bin/bash

echo '#include<'$1'>' >seedeps_tmp.cc

gcc -MM -I. -I.. -I../.. $2 seedeps_tmp.cc|tr ':' ' '|tr '\' ' '|tr ' ' '\n'|

egrep -v '^$'|egrep -v '.o$'|egrep -v '.cc$'|grep -v '^'$1'$'|sort

rm seedeps_tmp.cc