I highly recommend reading the tour of objc_msgSend [2]. It's very eye opening: for example, there's a cache scanning loop even in the fast case.
[1]: http://www.mulle-kybernetik.com/artikel/Optimization/opti-3-... [2]: http://www.friday.com/bbum/2009/12/18/objc_msgsend-part-1-th...
Hmm, check the results in the same page for the 64bit Cocoa runtime. This is where most optimizations are made, because Apple used the 32->64 bit transition as an opportunity to do most binary compatibility breaking changes to the language.
There, in 64bit, the IMP cached msg send is actually FASTER than C++ virtual method dispatch.
And if your app's performance suffers because of method calls / msg sends, you are doing it wrong. They should be an insignificant amount of time compared to the actual processing stuff done by your app.
If you're alluding to cost of message send, then it's pretty low, and can easily be made faster in case it becomes noticeable:
http://www.mikeash.com/pyblog/performance-comparisons-of-com...