Re: [ARM] ARM NEON support part 2/7: NEON support

[Mark Mitchell <mark at codesourcery dot com>]

Julian Brown wrote:

>> I think it would be slightly better to rename the existing
>> mangle_fundamental_type hook to just plain mangle_type, and then call it
>> for all types.  Then, if we find yet another case where we need back-end
>> help for mangling, we already have the hook we need.
>>
>> OK with that change.
> 
> I've tried this: it's a slight semantic change for the target hook too,
> since it's now called with a type argument which has not been resolved
> to its main variant (typedefs are not looked through before invoking the
> hook). I've changed tm.texi to that effect, and tweaked several backends
> which used the previous hook to use the new one instead.

Thanks, what you did looks good.

> I've tested (with the rest of the Neon support patch) on ARM with no
> regressions, and also natively on x86. Other affected backends are
> alpha, ia64, rs6000, s390 and sparc, though the changes are pretty
> trivial. I don't think I have the capacity to test all those targets in
> a reasonable time frame though -- so if anyone wants to volunteer to
> test their favourite target with this patch, I'd be grateful.
> 
> Alternatively, is it OK to commit as-is without further testing?

I've looked at the other back-end changes and I think they qualify as
simple enough.

So, I think the patch OK -- but please wait 48 hours for objections, and
please respond quickly if someone runs into trouble.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713