AW: [avr-gcc-list] ATmega256x trampolines? (Was: Why is there 1200bytes of zeroes in my code?)

[Haase Bjoern (PT-BEU/EMT)]

Hi, you could solve your problem by using -Wl,--relax . The linker first starts 
with worst-case assumptions that imply that it needs a trampoline for any 
function. 

The linker thus reserves the space for all possibly needed trampolines and 
starts the application code behind the trampolines. If it turns out that not 
all of the trampolines are actually needed, this space remains reserved by 
default. 

The space is freed afterwards when using the linker relaxation.

Yours, 

Björn.


-----Ursprüngliche Nachricht-----
Von: address@hidden [mailto:address@hidden Im Auftrag von Erik Christiansen
Gesendet: Mittwoch, 21. März 2007 09:42
An: Joerg Wunsch
Cc: address@hidden
Betreff: Re: [avr-gcc-list] ATmega256x trampolines? (Was: Why is there 
1200bytes of zeroes in my code?)

On Wed, Mar 21, 2007 at 07:07:24AM +0100, Joerg Wunsch wrote:

> 
> >  *fill*         0x000020f3        0x1 00=20
> >                 0x000020f4                __trampolines_start =3D .=20
> >  *(.trampolines)=20
> >  .trampolines   0x000020f4        0x0 linker stubs=20
> >  *(.trampolines*)=20
> >                 0x000020f4                __trampolines_end =3D .=20
> >  *(.jumptables)=20
> >  *(.jumptables*)=20
> 
> Obviously, this is somehow related to the linker stubs, but the
> question is how the space allocation is calculated.

Is it possible to invoke avr-gcc with "-Wl,-verbose" in the winavr
environment, so the we can see the linker script being used? (It would
be enlightening to see what it directs for these sections.)

Erik


_______________________________________________
AVR-GCC-list mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/avr-gcc-list