[weston] gl-renderer: add support of WL_SHM_FORMAT_YUV420

Vincent ABRIOU vincent.abriou at st.com
Fri Aug 5 07:43:13 UTC 2016

Hi Daniel,

Coming back to the asymmetrical subsampling...
I am now able too support NV12 SHM pixel format and you are right about 
the subsampling which is not assymmetrical at all :).
I willl send a new series of patches to support both I420 and NV12 
taking into account your comments.

We can abandon this patch since the new series will come very soon.


On 08/04/2016 06:03 PM, Vincent Abriou wrote:
> Hi Daniel,
> On 08/04/2016 04:32 PM, Daniel Stone wrote:
>> Hi Vincent,
>> On 4 August 2016 at 14:35, Vincent Abriou <vincent.abriou at st.com> wrote:
>>> This patch allow weston to accept WL_SHM_FORMAT_YUV420 buffers.
>>> In a gstreamer pipeline, the support of the WL_SHM_FORMAT_YUV420 by
>>> weston avoid pixel conversion between software decoders and waylandsink.
>>> Indeed, software decoders output I420 (YUV420 planar) that willYou
>>> match with WL_SHM_FORMAT_YUV420.
>> Thanks for the patch. It would be nice if wl_shm used the same FourCC
>> codes as wl_drm though, i.e. this would be I420 rather than YUV420.
> I use the WL_SHM_FORMAT_YUV420 defined in Wayland protocol. It is always
> a mental gymnastics to understand the match. But I don't know how to
> change it except by changing the wayland protocol it self.
>> You could save yourself some pain by unconditionally initialising
>> hsub/vsub/offset rather than in every format code though, and also
> I will propose a new patch to simplify this.
>> perhaps just flatten hsub/vsub into a single field, given that I don't
>> expect we'd support asymmetric subsampling through SHM any time soon.
> When I wrote the code, I was thinking about NV12. The 2nd plane UV size
> should be width * (height/2). It is less a component subsampling
> parameters than a plane width and height divider.
> Vincent.
>> You should also trim the 'Change-Id' line from your patch, as we don't
>> use Gerrit upstream.
>> Aside from that, looks good to me.
>> Cheers,
>> Daniel

More information about the wayland-devel mailing list