Most often when people "optimise" code they are just guessing where the problems are, grasping at straws and doing things they once heard someone say would make things faster (and usually they don't of course).

a mocking framework won't tell you what needs optimising.
A profiler may, but not a mock object (if anything it will likely hide the reason for any performance problems).

nice, but how are you going to trace the time needed for the functionality you replaced by those mocks (which usually is the expensive stuff like database and network operations)?

You don't mock any objects ! Just define pointcuts for entry/exit of all classes/functions you wanna measure and add before()/after() to trace.
Will post teh code if possible.