On Tue, Jul 12, 2016 at 01:21:47PM +0200, Bernd Schmidt wrote:
> On 07/12/2016 12:29 PM, Richard Biener wrote:
>
> >Instead of adding more configury can we please enable LRA on trunk by default
> >_now_? Otherwise the amount of testing it recieves won't really increase.
>
> I hope you mean for ppc only, otherwise you're breaking a lot of ports...
Yes I mean for PowerPC only. I can change the switches name to:
--enable-powerpc-lra
--enable-powerpc-float128
if it would be clearer. If you would prefer the switches to be undocumented,
and just private switches, that is fine also.
I view --enable-lra/--enble-powerpc-lra as a temporary switch to allow us to
straddle the issue until the performance blocker (PR 69847) is resolved, and
the PowerPC compiler switches to LRA in GCC 7. Ideally, it would be nice to
eliminate the support for reload in the PowerPC backend (and hence this
switch). But until we fully transition to lra, it would be useful to have a
way to switch what the default is without having to edit the sandbox.
Likewise, --enable-float128/--enable-powerpc-float128 is a transition switch.
It would be nice to have __float128 on by default without using the -mfloat128
switch, but we have a lot of work in the library arena before we can do this.
--
Michael Meissner, IBM
IBM, M/S 2506R, 550 King Street, Littleton, MA 01460-6245, USA
email: meiss...@linux.vnet.ibm.com, phone: +1 (978) 899-4797
