On Fri, Dec 13, 2019 at 07:28:28PM -0600, Joshua Watt wrote: > On Fri, Dec 13, 2019, 7:22 PM Adrian Bunk <[email protected]> wrote: > > > On Fri, Dec 13, 2019 at 06:47:33PM -0600, Joshua Watt wrote: > > > On Fri, Dec 13, 2019 at 5:10 PM Adrian Bunk <[email protected]> wrote: > > > > > > > > On Fri, Dec 13, 2019 at 04:03:09PM -0600, Joshua Watt wrote: > > > > > Adds a tune for Cortex-A72 Cortex-A53 big.LITTLE SoCs, e.g. Rockchip > > > > > RK3399 > > > > >... > > > > > +TUNE_FEATURES_tune-cortexa72-cortexa53 = > > "${TUNE_FEATURES_tune-aarch64} cortexa72-cortexa53" > > > > >... > > > > > > > > Looking at the gcc sources cortexa72-cortexa53 defaults to crc enabled, > > > > so TUNE_FEATURES_tune-cortexa72-cortexa53 must contain crc. > > > > > > I think that's armv8.1-a, the cortex A72 is only armv8-a (at least > > > AFAICT), so I think crc would still be an optional feature. > > >... > > > > It is optional for armv8-a, but not for Cortex A53 or A72. > > > > What matters in practice is that gcc automatically enables it for > > cortex-a72.cortex-a53, so the resulting code might not run on > > armv8-a hardware without support for crc - things will break > > if the tune features are not 100% aligned with whatever gcc > > is doing. > > I agree, but I can't find anywhere that says crc is enabled by default when > the tune is cortexa72.cortexa53. Maybe I'm looking in the wrong location?
https://github.com/gcc-mirror/gcc/blob/master/gcc/config/arm/arm-cpus.in#L1278 > Thanks cu Adrian -- _______________________________________________ Openembedded-core mailing list [email protected] http://lists.openembedded.org/mailman/listinfo/openembedded-core
