[meta-ti] omap3-sgx-modules-5.01.01.xx ONLY supports hardfp mode for now

[Ankur Tyagi]

I just checked that inside "ti33x.inc",
DEFAULTTUNE ?= "cortexa8thf-neon"

But it is overwritten by inclusion of arch-armv7a.inc
DEFAULTTUNE ?= "armv7a"
TUNE_FEATURES_tune-armv7a ?= "armv7a vfp"

But TUNER_FEATURES also include "callconvention-hard".

I am bit confused here.


"Knowledge is knowing tomato is a fruit, wisdom is not putting it in a
fruit salad"

On Thu, Aug 6, 2015 at 3:49 PM, Ankur Tyagi <ankur.tyagi85 at gmail.com> wrote:

> thanks for help.
>
> Will try with above suggestion
>
> "Knowledge is knowing tomato is a fruit, wisdom is not putting it in a
> fruit salad"
>
> On Thu, Aug 6, 2015 at 3:46 PM, Khem Raj <raj.khem at gmail.com> wrote:
>
>> On Wed, Aug 5, 2015 at 8:41 PM, Ankur Tyagi <ankur.tyagi85 at gmail.com>
>> wrote:
>> > Thanks for your response.
>> >
>> > So it is worth trying with TUNE_FEATURE "hfp" and measure the
>> performance.
>> > Do I have to set TARGET_FPU as "hfp-neon" as default value is "vfp-neon"
>> >
>>
>> all you might need is to select a proper DEFAULTTUNE in machine config
>> something like
>>
>> DEFAULTTUNE ?= "cortexa8thf-neon"
>>
>> > Does it make sense to push this change upstream as well ?
>> >
>> > "Knowledge is knowing tomato is a fruit, wisdom is not putting it in a
>> fruit
>> > salad"
>> >
>> > On Thu, Aug 6, 2015 at 3:00 PM, Khem Raj <raj.khem at gmail.com> wrote:
>> >>
>> >> On Wed, Aug 5, 2015 at 7:47 PM, Ankur Tyagi <ankur.tyagi85 at gmail.com>
>> >> wrote:
>> >> > Hi,
>> >> >
>> >> > I am using TI AM335x-EVM kit and while building image using
>> >> > "arago-daisy-config.txt", I saw this warning :
>> >> >
>> >> > omap3-sgx-modules-5.01.01.02 ONLY supports hardfp mode for now
>> >> > omap3-sgx-modules-5.01.01.01 ONLY supports hardfp mode for now
>> >> >
>> >> > and TUNE_FEATUREis "vfp"
>> >> >
>> >> > Does that mean GPU is not getting used anymore ?
>> >>
>> >> not really, what its telling is that you dont have
>> >> callconvention-hard in TUNE_FEATURES
>> >> which means that compiler wont use hardfloat ABI when making function
>> >> calls.
>> >> but neon/vfp will still be used e.g. when doing vectorization etc. if
>> >> using hardfp ABI will radically improve performance totally depends
>> >> upon kind of workload that you are measuring. It may vary.
>> >>
>> >> > I am using QT5 with eglfs platform and need GPU support for
>> performance.
>> >> >
>> >> > Please clarify the situation.
>> >> >
>> >> > Thanks
>> >> > Ankur
>> >> >
>> >> > --
>> >> > _______________________________________________
>> >> > meta-ti mailing list
>> >> > meta-ti at yoctoproject.org
>> >> > https://lists.yoctoproject.org/listinfo/meta-ti
>> >> >
>> >
>> >
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.yoctoproject.org/pipermail/meta-ti/attachments/20150806/b41ae1a1/attachment.html>