[PATCH, committed] PPC405 atomic support (PR target/21760)

[Daniel Jacobowitz]

On Fri, Jun 24, 2005 at 10:16:13AM -0400, David Edelsohn wrote:
> >>>>> Daniel Jacobowitz writes:
> 
> Dan> Why does this need to be a preprocessor check - why not do this based
> Dan> on -mcpu=405 instead of --with-cpu=405?
> 
> 	The atomic instruction fix needs to be included in all GCC runtime
> libraries. No PPC405 multilib will be added.  Therefore, GCC needs to
> be configured and built for PPC405 to address this erratum.  If one does
> not configure for PPC405, there is no reason to include the fix.
> 
> 	This also allows users to configure --with-cpu=405 for PPC405CR
> and use -mcpu=405 for all other PPC405 processors which are not affected
> by the erratum. 

There are no other cases in which --with-cpu is different from -mcpu;
it was a conscious design decision for the revised --with-ARG mechanism
to exactly match the behavior of hardcoding options to the compiler. 
I'm pretty unhappy with a divergence.  Why not create -mcpu=405cr, tie
the erratum fix to that option, and document --with-cpu=405cr?

-- 
Daniel Jacobowitz
CodeSourcery, LLC