[Intel-gfx] [PATCH] drm/i915: Pretend cursor is always on for ILK-style WM calculations

[Matt Roper]

On Mon, Feb 01, 2016 at 04:22:22PM +0200, Ville Syrjälä wrote:
> On Mon, Feb 01, 2016 at 01:26:14AM -0800, Matt Roper wrote:
> > Due to our lack of two-step watermark programming, our driver has
> > historically pretended that the cursor plane is always on for the
> > purpose of watermark calculations; this helps avoid serious flickering
> > when the cursor turns off/on (e.g., when the user moves the mouse
> > pointer to a different screen).  That workaround was accidentally
> > dropped as we started working toward atomic watermark updates.  Since we
> > still aren't quite there yet with two-stage updates, we need to
> > resurrect the workaround and treat the cursor as always active.
> > 
> > Cc: simdev11 at outlook.com
> > Cc: manfred.kitzbichler at gmail.com
> > Reported-by: simdev11 at outlook.com
> > Reported-by: manfred.kitzbichler at gmail.com
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93892
> > Fixes: 43d59eda1 ("drm/i915: Eliminate usage of plane_wm_parameters from ILK-style WM code (v2)")
> > Signed-off-by: Matt Roper <matthew.d.roper at intel.com>
> > ---
> > I'm traveling at the moment and don't have any HW access, so this patch is
> > completely untested, but I think it will solve the issue being reported.
> > 
> >  drivers/gpu/drm/i915/intel_pm.c | 15 ++++++++++-----
> >  1 file changed, 10 insertions(+), 5 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> > index 31bc4ea..a53c018 100644
> > --- a/drivers/gpu/drm/i915/intel_pm.c
> > +++ b/drivers/gpu/drm/i915/intel_pm.c
> > @@ -1799,16 +1799,21 @@ static uint32_t ilk_compute_cur_wm(const struct intel_crtc_state *cstate,
> >  				   const struct intel_plane_state *pstate,
> >  				   uint32_t mem_value)
> >  {
> > -	int cpp = pstate->base.fb ?
> > -		drm_format_plane_cpp(pstate->base.fb->pixel_format, 0) : 0;
> > +	/*
> > +	 * We treat the cursor plane as always-on for the purposes of watermark
> > +	 * calculation.  Until we have two-stage watermark programming merged,
> > +	 * this is necessary to avoid flickering.
> > +	 */
> > +	int cpp = 4;
> > +	int width = drm_rect_width(&pstate->dst) ?: 64;
> 
> I don't think we can rely on that being correct when the plane is not
> visible. Also for the cursor, I'm not sure we should be even using the
> visible width of the plane when it's partially visible.
> 

But would this cause problems?  I thought higher values were always
"safe" (as long as they don't go over the maximums), just non-optimal
since they translate to the level of FIFO data at which we start
fetching more memory.  Acording to the bspec we don't even need to ever
program watermarks from the values setup by the BIOS if we're willing to
live with non-optimal power/memory bandwidth usage.  

The proper fix is definitely to re-merge the two-step watermark
programming, but we still haven't tracked down why that patch upset the
CI system yet (and I won't have time to look at that in depth until I'm
done traveling).  This seemed like a fairly safe/non-invasive temporary
workaround for the user-visible issues.  I'm open to other workarounds
though if you have an idea for something that will work better.

Thanks.


Matt

> >  
> > -	if (!cstate->base.active || !pstate->visible)
> > +	if (!cstate->base.active)
> >  		return 0;
> >  
> > +
> >  	return ilk_wm_method2(ilk_pipe_pixel_rate(cstate),
> >  			      cstate->base.adjusted_mode.crtc_htotal,
> > -			      drm_rect_width(&pstate->dst),
> > -			      cpp, mem_value);
> > +			      width, cpp, mem_value);
> >  }
> >  
> >  /* Only for WM_LP. */
> > -- 
> > 2.1.4
> > 
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> 
> -- 
> Ville Syrjälä
> Intel OTC

-- 
Matt Roper
Graphics Software Engineer
IoTG Platform Enabling & Development
Intel Corporation
(916) 356-2795