On Wed, Jul 26, 2017 at 03:02:15PM -0500, Segher Boessenkool wrote:
> Hi Mike,
>
> On Wed, Jul 26, 2017 at 12:24:17AM -0400, Michael Meissner wrote:
> > * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
> > -mvsx-small-integer option.
> > (ISA_3_0_MASKS_IEEE): Likewise.
> > (POWERPC_MASKS): Likewise.
>
> (OTHER_VSX_VECTOR_MASKS): Likewise.
Thanks.
> > @@ -2099,20 +2099,8 @@ rs6000_hard_regno_mode_ok (int regno, ma
>
> > - if (TARGET_VSX_SMALL_INTEGER)
> > - {
> > - if (mode == SImode)
> > - return 1;
>
> I don't see why deleting this is correct?
>
> > - if (TARGET_P9_VECTOR && (mode == HImode || mode == QImode))
> > - return 1;
>
> And this.
Hmmm, I was assuming that the previous test for TARGET_VSX would catch all of
the cases, but for power7, SFmode does not allow scalars in Altivec registers,
so it will fall through. Let me change it back, and add a TARGET_P8_VECTOR
test.
Thanks for catching this.
I did start a non-boostrap build on p9 to make sure everything is ok.
> The rest looks fine I think.
>
>
> Segher
>
--
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
