I can not speak of other than Keil, but I, honestly, believe that a lot of the optimization is TOTALLY misapplied just to make the optimizer look more impressive.

I would LOVE an optimizing COMPILER, there would be not debugging implications.

That a change

from
mov r7,Ralph
mov a,r7

to
mov a,Ralph

is done by an optimizing LINKER is uttely ridiculous. Was it done by the compiler, no implications on debugging would exist. Likewise the use of jmp @ for switch statements should never have been a linker function, but be done by the compiler.

I just came to realize that had they done it in the compiler instead of the linker, it would be far less (liklely to be) buggy.

Erik