[meta-xilinx] gcc with hard float for Zynq

[Richard Cagley]

On Thu, Oct 16, 2014 at 1:12 AM, Philip Balister <philip at balister.org> wrote:
> On 10/16/2014 08:39 AM, Mike Looijmans wrote:
>> On 10/16/2014 01:23 AM, Richard Cagley wrote:
>>> when I bitbake core-image-minimal-dev I get gcc for native compilation
>>> on the target. But, it doesn't appear to provide hard float support.
>>> Is there a way to configure so I can use -mfloat-abi=hard?
>>
>> The default is to use hard float support, but not between libraries. So
>> all your code will use hard float (mixed NEON and VFP instructions),
>> it's only when passing floats to other libraries it uses the standard
>> registers instead of the floating point stack to exchange the data. This
>> usually has no impact on performance, unless in some very particular
>> circumstances.
>>
>
> Urg, I just realized my earlier answer is for a different problem.
> Answering email late at night ....
>
> Yes the native sdk defaults do not match the tune settings for the
> build. Thanks for reminding me of this, I need to talk to some of the OE
> guys today and see what we can do about this.
>
> Philip

Sorry, now i'm very confused. I'm building an application and linking
against OpenCV/boost pre-built libraries. Is it because of the "OE
guys" setup that I can't use -mfloat-abi=hard for my gcc call? Or,
will your first reply Philip still work?