[Intel-gfx] [RFC v2] drm/i915: Android native sync support

[Daniel Vetter]

On Mon, Jan 26, 2015 at 11:08:43AM +0000, Tvrtko Ursulin wrote:
> 
> On 01/24/2015 09:47 AM, Daniel Vetter wrote:
> >On Fri, Jan 23, 2015 at 5:49 PM, Tvrtko Ursulin
> ><tvrtko.ursulin at linux.intel.com> wrote:
> >>Could you please translate this into something understandable by newcomers?
> >>:)
> >
> >I don't know which parts are confusing without questions so please ask
> >them ... the questions below about scheduler interactions seem fairly
> >advanced ;-)
> 
> I suppose for starters, when I took over this work I read the last thread
> from December.  There you were saying to Jesse to convert the separate ioctl
> into execbuf interface for in & out fences.
> 
> That's what I did I thought. So is that not the fashion of the day any more
> or I misunderstood something?
> 
> People will scream very soon for something along these lines anyway since it
> is kind of packaged with the scheduler I am told and both will be very much
> desired soon.

That's still the proper fashion imo and what I've thought I've explained
...

About the screaming: The real blocker here is destaging android fences.
I've been telling this to everyone for a while already, but can't hurt to
reinforce ;-)

> Idea is to allow submitting of work and not block in userspace rather let
> the scheduler queue and shuffle depending on fences.

Well that's also possible without explicit fences, but not if your
userspace assumes explicit fences exist ofc.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch