Re: Cortex-r52 FP double precision

[Thomas Preudhomme <thomas dot preudhomme at foss dot arm dot com>]

Hi Alexander,

An evaluation needs to be made of the benefit of a new multilib for Armv8-R 
(instead of reusing an existing multilib) compared to the extra maintenance and 
build time.

Best regards,

Thomas

On 31/01/18 14:31, Alexander Fedotov wrote:
> Hi Thomas
>
> Don't you know by the way why ARMv8-R is not listed in t-rmprofile multilib ?
>
> Alex
>
> On Fri, Jan 26, 2018 at 9:25 PM, Thomas Preudhomme
> <thomas.preudhomme@foss.arm.com> wrote:
> > That or just use -mfpu=auto (as in -mcpu=cortex-r52 -mfpu=auto
> > -mfloat-abi=(softfp|hard)).
> >
> > Best regards,
> >
> > Thomas
> >
> >
> > On 26/01/18 16:44, Alexander Fedotov wrote:
> > > Thank you Thomas
> > >
> > > So in order to set dp+Neon for armv8-r I should to use switch
> > > "neon-fp-armv8". Not an fpv5-d16. Right ?
> > >
> > > On Fri, Jan 26, 2018 at 12:52 PM, Thomas Preudhomme
> > > <thomas.preudhomme@foss.arm.com> wrote:
> > > > Hi Alexander,
> > > >
> > > > As mentioned in [1], Arm Cortex-R52 can have either single precision or
> > > > double precision + Neon. This is reflected in GCC 8 by -mcpu=cortex-r52
> > > > defaulting to the latter (double precision + Neon) and
> > > > -mcpu=cortex-r52+nofp.dp giving you the former (single precision).
> > > >
> > > > [1] https://developer.arm.com/products/processors/cortex-r/cortex-r52
> > > >
> > > > Best regards,
> > > >
> > > > Thomas
> > > >
> > > >
> > > > On 25/01/18 21:22, Alexander Fedotov wrote:
> > > > > Hi,
> > > > > As I understand from this
> > > > > https://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html current master
> > > > > branch doesn't have support of double-precision FPv5 floating-point
> > > > > instructions for ARMv8-R (Cortex-r52).
> > > > > If yes, are there any chances to see them in GCC 8 ?
> > > > >
> > > > > Alex