Re: (fwd) [avr-gcc-list] RFC: ATmega256x and 16-bit function pointers

[An . Kaiser]

On Sat, 19 Mar 2005 22:00:53 +0100, you wrote:

>The problem is that the jump can be to any of the PUSH instructions
>at the beginning of __prologue_saves__ - not necessarily the first one
>(if fewer registers need to be pushed).  With CALL, you would have to
>pop the return address before pushing anything.

Obviously. And if a prologue for some specific number of saved
registers is used only once, the balance is towards the current
solution. 

But not if it is used multiple times, as each stub only has to pop the
return stack, then jump to the proper "push" entry point. If would
even be possible to save a word of the frame size constant loads,
because the word following the CALL could do the job instead (via
LPM).

However maybe it is not worth the effort and it is just the uglyness
of the prologue/epilogue code which offends me.

Gruss, Andreas