Ref:
http://sources.redhat.com/ml/gdb/2002-04/msg00245.html

The attached is a revised frame->unwind and frame cache patch.  I think 
this one is looking much better - it now drops into place for existing 
targets.

It introduces two changes:


frame->register_unwind()

Per other e-mail, there is a per-frame unwind function.  Two 
implementations are provided:
	- a generic dummy frame
	- the traditional ->saved_regs frame
The method for each frame is currently selected according to the PC 
(override mechanism?).  Those functions can make use of a cacheing 
mechanism (although in the patch it is #if 0'd out).  Running 
restore.exp native suggests a 2% reduction in the system time (which 
makes sense :-).  Testing continues.

I suspect the interface to this method will evolve over time.


``info frame''

I've updated this to use the register-unwind function.  This eliminates 
the problem of using ->saved_regs to determine the location of all the 
registers.  In the case of the SP, it will now also indicate a value on 
the stack or in another register.


Doesn't show any regressions for the rs6000.


I'll have to try it on a few other targets.

Andrew