[meta-freescale] [PATCH V2] imx-base.inc: Enable thumb2 and vfp capability

[Andreas Müller]

On Thu, May 28, 2015 at 8:16 PM, Max Krummenacher <max.oss.09 at gmail.com> wrote:
> Hi
>
> Khem Raj <raj.khem at ...> writes:
>
>>
>> On Wed, May 27, 2015 at 3:28 PM, Andreas Müller
>> <schnitzeltony <at> googlemail.com> wrote:
>> > On Wed, May 27, 2015 at 11:49 PM, Khem Raj <raj.khem <at> gmail.com> wrote:
>> >>
>> >> Yes just see if you can add vfp to it
>> >> I think oe-core needs fixing probably  for t2 case meanwhile we can live
>> >> with it.
>> > ??? - sorry you lost me here. Just a note: Working with angstrom
>> > (enabling thumb(2))) leads to many packages with vfp/thumb2. What
>> > change do you have in mind?
>> >
>>
>> because in angstrom we change to use t2 when you have 't' in incoming
>> tunes and the processor supports
>> t2 thats why you see what you see. So when you just use 't' it all
>> works out. However OE-Core should not
>> freak out when we use 't2' since thats a valid combination. Hope that
> helps you
>>
>> > Andreas
>
> I believe that DEFAULTTUNE must be set to one of the AVAILTUNES from
> oe-core.
> http://cgit.openembedded.org/openembedded-core/tree/meta/conf/machine/include/arm/arch-armv7a.inc#n27
>
> For the i.MX 6 that would be armv7at-neon and armv7athf-neon.
> (note the missing 2 and vfp)
>
> arch-armv7a.inc will then add thumb vfp and neon to the TUNE_FEATURES, so
> the patch should read:
> -# DEFAULTTUNE_mx6 ?= "cortexa9-neon"
> +# DEFAULTTUNE_mx6 ?= "cortexa9t-neon"
>  # handled by hardware
> -DEFAULTTUNE_mx6 ?= "cortexa9hf-neon"
> +DEFAULTTUNE_mx6 ?= "cortexa9thf-neon"
>
>
> thumb vs. thumb2 is chosen by gcc form the -mcpu, -march arguments.
>
> That oe-core uses t for thumb in AVAILTUNES for arm archs which supports
> thumb or thumb2, but does use t and t2 in the package feed names is a
> bit confusing but IMHO not warranting a change in oe-core breaking a lot
> of BSP layers in the process.
>
>
Hi Max,

how about an Ack for the patch [1] doing exactly that ?

[1] https://lists.yoctoproject.org/pipermail/meta-freescale/2015-May/014042.html

Andreas