One key to making parallel algorithms efficient is to minimize the amount of communication between cores.


http://web.mit.edu/newsoffice/2011/m...es-4-0228.html