[FFmpeg-devel] [PATCH] pixblockdsp: disable altivec optimizations on ppc64be

[Andreas Cadhalpun]

On 01.11.2016 02:04, Michael Niedermayer wrote:
> On Tue, Nov 01, 2016 at 12:27:00AM +0100, Andreas Cadhalpun wrote:
>> The checkasm test fails, see trac ticket 5508.
>>
>> Also, the following tests fail due to this:
>> fate-vsynth1-dnxhd-2k-hr-hq fate-vsynth1-dnxhd-edge1-hr
>> fate-vsynth1-dnxhd-edge2-hr fate-vsynth1-dnxhd-edge3-hr
>> fate-vsynth1-dnxhd-hr-sq-mov fate-vsynth1-dnxhd-hr-hq-mov
>> fate-vsynth2-dnxhd-2k-hr-hq fate-vsynth2-dnxhd-edge1-hr
>> fate-vsynth2-dnxhd-edge2-hr fate-vsynth2-dnxhd-edge3-hr
>> fate-vsynth2-dnxhd-hr-sq-mov fate-vsynth2-dnxhd-hr-hq-mov
>> fate-vsynth3-dnxhd-2k-hr-hq fate-vsynth3-dnxhd-edge1-hr
>> fate-vsynth3-dnxhd-edge2-hr fate-vsynth3-dnxhd-edge3-hr
>> fate-vsynth3-dnxhd-hr-sq-mov fate-vsynth3-dnxhd-hr-hq-mov
>>
>> Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun at googlemail.com>
>> ---
>>
>> Just disabling the checkasm_check_pixblockdsp test for ppc64be,
>> as was done in commit e5d434 for release/3.1, does not make much sense,
>> as the altivec functions actually don't work...
> 
> has this been tested with actual hw or an emulator ?

The tests failed on the Debian buildd [1] and I could reproduce the exact
same failures with qemu. The patch is only tested with qemu.

> is this a regression ? if so since when ?

Well, the commit e5d434 was never on master, but in release/3.1.
And the failing dnxhd tests are new:
eb5f4b1 tests/fate/vcodec: add dnxhr mov tests
44ac2b9 tests/fate/vcodec: add dnxhr edge tests
6108cb2 tests/fate: add dnxhr encoding tests

So this altivec code probably never worked on ppc64be, it was just not
tested.

Best regards,
Andreas


1: https://buildd.debian.org/status/fetch.php?pkg=ffmpeg&arch=ppc64&ver=7%3A3.2-1&stamp=1477927610