On Dec 11, 2007 4:28 PM, Nuno Lucas <[EMAIL PROTECTED]> wrote:
>
>
> Allocated memory should always come aligned. That's not the problem.
> I would guess the problem is the array base address in the stack (on
> IA32 the stack is always aligned by 32 bits, but that can not be true
> on other platforms).
>
> You have 3 options:
>
> 1) Use an union - it can not be adequate to the code in question;
> 2) Use memcpy() - assures success even on un-aligned memory;
> 3) Some other algorithm that access the pointed data in an adequate
> way (like the memcpy byte access).
>
> Regards,
> ~Nuno Lucas
>
> P.S.- Only now noticed I wasn't replying to the mailing list. One of
> those gmail things.
>
I've added the mailing list to CC, since your reply might be helpful to someone.
In the meantime I thought; something tells me that it should be
possible to make this kind of access "safe" using some GCC option? It
should mean adding a few assembler instructions before such an access,
for example:
(more code)
(more code)
AND r1, 3
JPZ skip_fixstuff
CALL fixstuff
skip_fixstuff:
LDR r1, r2
(more code)
(more code)
(alert: quasi assembler written by a n00b)
With the option I'm taking about turned on, those three instructions
and the label are inserted, otherwise they do not exist. R1
theoretically contains the address of data to access.
Does such option or its equivalent exist?
--
Ivan Vučica
-- Croatia --
-------------------------------------------------------------------------
SF.Net email is sponsored by:
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Cegcc-devel mailing list
Cegcc-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/cegcc-devel
