[Mesa-dev] Mesa (master): st/mesa: implement PBO downloads for ReadPixels

Nicolai Hähnle nhaehnle at gmail.com
Thu Jun 2 15:24:51 UTC 2016


Hi Michel,

thanks for letting me know. I can reproduce this on my Tonga, will look 
into it soon.

Cheers,
Nicolai

On 02.06.2016 15:12, Michel Dänzer wrote:
> On 02.06.2016 05:44, Nicolai =?UNKNOWN?Q?H=C3=A4hnle?= wrote:
>> Module: Mesa
>> Branch: master
>> Commit: ed0e9862c56d13381f0414b5a835fd31960edfe6
>> URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=ed0e9862c56d13381f0414b5a835fd31960edfe6
>>
>> Author: Nicolai Hähnle <nicolai.haehnle at amd.com>
>> Date:   Tue Apr 26 13:19:28 2016 -0500
>>
>> st/mesa: implement PBO downloads for ReadPixels
>>
>> v2: require PIPE_CAP_SAMPLER_VIEW_TARGET; technically only needed for some of
>>      the texture targets, but all hardware that has shader images should also
>>      have this cap.
>>
>> Reviewed-by: Brian Paul <brianp at vmware.com>
>> Reviewed-by: Marek Olšák <marek.olsak at amd.com>
>
> This commit seems to have broken the X root weave with glamor on my
> Kaveri and Tonga. I say "seems" because it's only broken about 10% of
> the time, so I'm not 100% sure yet that my bisection was correct. I used
>
>   for i in $(seq 10); do Xephyr :99 -retro -glamor; done
>
> for testing, just checking that all of them come up with the correct
> root weave and then closing the window. With the bad commits, at least
> one out of the 10 Xephyr instances came up bad. With the commit before
> this one, 20 Xephyr instances came up good.
>
>
> Since the problem doesn't always occur and looks basically random, maybe
> some synchronization isn't quite correct yet.
>
>


More information about the mesa-dev mailing list