RE: Re[2]: [avr-gcc-list] Re: GCC-AVR Update (20082003)

[James Dabbs]

> IAR can optimize RCALL vs. CALL in situations where it's clear that an
RCALL
> suffices, gcc currently can only use either of them.  (You can force it to
> use RCALL even on > 8 KB devices, but that's impractical.) This could
perhaps
> also save a lot, where the relative saving is most notable on 16 KB
devices.

That's a nice feature.  Out of curiosity, what would it take to get this
into GCC?  Does the architecture allow this type of optimization (at the
linker level)?

My code uses about 12K, but I see why it works out of blind luck.  The ISR
code is pretty much compartmentalized in terms of the call tree, as it the
foreground loop.  They are also linked pretty much one after another.
Because of my linker order and the fact that all of the shared fg/bg
routines are inline, it works.  How's that for luck!?