Observation for <<Edge Profiling versus Path Profiling : The Showdown>>


     The idea in this paper is great in that we can consider what we can definitely get and potentially get when we only have partial information about the program behaviour. It fits the scientific programs quite well, but not as good for transaction handling programs such as go and gcc. This is natural since the behaviour of these programs are quite complicated so that the partial informations can not help as much as for the scientific programs.

   zhigang