[Bug 779146] dmabuf: be able to negotiate tiled surfaces

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=779146

--- Comment #21 from Scott D Phillips <scott.d.phillips at intel.com> ---
(In reply to Nicolas Dufresne (stormer) from comment #20)
> In real life, a pipeline like:
> 
> ... ! vaapidecode ! textoverlay ! vaapiencode ! ..
> 
> Will use linear memory backend. So still not copying, but will pay the
> overhead. While if you do:

Well, with gstreamer-vaapi + intel-vaapi-driver as it exists today, the only
situation where linear surfaces are actually created is with dmabuf. In all the
non-dmabuf cases gstreamer-vaapi doesn't specify tiled or linear in allocations
and the intel-vaapi-driver chooses tiled. The reason is that the intel encode
and decode hardware units are able to operate on tiled surfaces only. When a
linear surface is given somehow then it will either by tiled by the cpu
(vaDeriveImage + memcpy) or the gpu (vaPutImage), but someone needs to pay the
tiling cost before encode can happen.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.