On 2006-01-24 17:52:57 Uwe Fechner wrote:
>
>Your patch looks trivial.
>
>Is changing one line of code of avr.c really all that needs to be done
>to make avr-gcc support three bytes per call as return adress?
>
It seems to be. I've discussed this here before. The only case this
matters is functions with a large number
of parameters. In this case
parameter are pushed on the stack before
the call. In the function itself,
the distance between the SP and those
parameters depends on the size
of the PC on the stack (2 vs 3 bytes
in this case).
A "cleaner" solution might
be to introduce an "avr6" architecture for
the 3-bytes PC types. I didn't do this
for two reasons
1) I'm lazy
2) I noticed a wish to leave
the avr1-avr5 structure all together because
it's an arbitrary
partioning, not in sync with Atmel's partioning of device
types.
>I will add your patch to my patch-set,
but I am not shure, when I find
>the time to test it (I must order samples of ATmega2561 first).
>
Are you referring to the "official"
patch-set ? If so, where do I find this ?
My patch now overlaps a bit with other
initiatives (and it's based on an
older tool-set).