Re: [PATCH 2/2] drm/msm/mdp5: Add hardware cursor support

[Daniel Vetter]

On Sat, Jan 17, 2015 at 08:46:05AM -0500, Rob Clark wrote:
> On Fri, Jan 16, 2015 at 11:06 PM, Daniel Vetter <daniel@xxxxxxxx> wrote:
> > On Thu, Jan 15, 2015 at 08:46:46AM -0500, Rob Clark wrote:
> >> On Wed, Jan 14, 2015 at 7:55 PM, Daniel Vetter <daniel@xxxxxxxx> wrote:
> >> > On Tue, Jan 13, 2015 at 05:18:04PM -0500, Stephane Viau wrote:
> >> >> From: Beeresh Gopal <gbeeresh@xxxxxxxxxxxxxx>
> >> >>
> >> >> This patch implements the hardware accelarated cursor
> >> >> support for MDP5 platforms.
> >> >>
> >> >> Signed-off-by: Beeresh Gopal <gbeeresh@xxxxxxxxxxxxxx>
> >> >> Signed-off-by: Wentao Xu <wentaox@xxxxxxxxxxxxxx>
> >> >> Signed-off-by: Stephane Viau <sviau@xxxxxxxxxxxxxx>
> >> >
> >> > Imo implementing legacy cursor support instead of with universal planes
> >> > makes no sense. Especially since msm is converted to atomic already, and
> >> > you can't move the cursor with atomic when it's legacy only. See the
> >> > cursor argument for the drm_crtc_init_with_planes function and how it's
> >> > used in e.g. i915.
> >> >
> >>
> >> well, I'm still not 100% convinced about going through the whole
> >> atomic mechanism for cursors..  in particular stuff that tries to
> >> enable/disable the cursor at 1000fps, goes *really* badly when things
> >> start waiting for vsync.
> >>
> >> I'll probably try some experiments with it at some point, but at this
> >> point something that works with x11 is a lot more interesting for me
> >> (since every time I switch from mdp4 device to mdp5 device I forget to
> >> disable hw cursor the first time I start x)
> >
> > Well for one this uses the legacy cursor callbacks directly, at least a
> > cursor plane is imo in order.
> >
> > Otoh we just need to fix up the cursor atomic implementation to allow
> > drivers to do fully async updates which get merged down to one update.
> > Which Ville's original atomic stuff already had. So all recoverable by
> > adding a flag somewhere and setting that in the plane_update-on-atomic
> > function.
> 
> something that could merge multiple intra-vsync updates would, I
> think, make cursor planes usable
> 
> > Merging legacy code just because the new stuff isn't 100% perfect yet imo
> > just doesn't make that much sense.
> 
> but merging legacy because new stuff isn't usable yet does ;-)

I've thought up a quick&tiny hack which should give us perfect (as in
"matches old i915 semantics") legacy cursor ioctls on top of universal
planes + atomic helpers. I need it for i915 too ;-)
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/