On Sun, Mar 16, 2003 at 10:35:24PM +1100, matthew green wrote:
> 
>    On Sat, Mar 15, 2003 at 09:25:08PM -0800, Jason Thorpe wrote:
>    > 
>    > On Saturday, March 15, 2003, at 08:09  AM, Juergen Hannken-Illjes wrote:
>    > 
>    > >
>    > >Module Name:	src
>    > >Committed By:	hannken
>    > >Date:		Sat Mar 15 16:09:35 UTC 2003
>    > >
>    > >Modified Files:
>    > >	src/gnu/dist/toolchain/gcc/config/rs6000: netbsd.h
>    > >
>    > >Log Message:
>    > >Add -mstrict-align to the CC1 default options.
>    > >At least the 403 ports cannot handle unaligned access.
>    > >
>    > >Discussed with Jason Thorpe and Matt Thomas.
>    > >
>    > 
>    > Hm.  Maybe what we should do is default to -mstrict-align, but disable 
>    > -mstrict-align for -mcpu=foo where foo doesn't require strict alignment 
>    > (e.g. 603, 604, 750, etc.).  All of those CPUs can do unaligned access 
>    > on integer loads/stores, right?  Is it only the 403 that actually has 
>    > the problem?  (I honestly can't remember if the 405 can do unaligned 
>    > access or not...)
>    
>    From the sources (I have no walnut board) the 405 also traps.
>    
>    At least 1.6 has a shared/powerpc with X, games and text.
>    Therefore the powerpc should use a feature set useable on all cpus.
> 
> that's OK;  if someone uses -m603 they are doing that themselves
> and it doesn't matter ... i think jason's idea is still useful.

I understand. What about

	#define CC1_SPEC "-mno-multiple %{!mcpu*: -mstrict-align}"

to set this flag only when there is no -mXXX?
-- 
Juergen Hannken-Illjes - hannken@eis.cs.tu-bs.de - TU Braunschweig (Germany)