Re: [PATCH 1/9] drm/i915: Skip modifying PCH DREF if not changing clock sources

[Daniel Vetter <daniel@ffwll.ch>]

On Thu, Mar 28, 2013 at 09:27:31AM +0200, Jani Nikula wrote:
> On Wed, 27 Mar 2013, Jesse Barnes <jbarnes@virtuousgeek.org> wrote:
> > From: Chris Wilson <chris@chris-wilson.co.uk>
> >
> > Modifying the clock sources (via the DREF control on the PCH) is a slow
> > multi-stage process as we need to let the clocks stabilise between each
> > stage. If we are not actually changing the clock sources, then we can
> > return early.
> >
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > ---
> >  drivers/gpu/drm/i915/intel_display.c |   83 +++++++++++++++++++++++++---------
> >  1 file changed, 61 insertions(+), 22 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 8f0db8c..9d05b30 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -4833,7 +4833,7 @@ static void ironlake_init_pch_refclk(struct drm_device *dev)
> >  	struct drm_i915_private *dev_priv = dev->dev_private;
> >  	struct drm_mode_config *mode_config = &dev->mode_config;
> >  	struct intel_encoder *encoder;
> > -	u32 temp;
> > +	u32 val, final;
> >  	bool has_lvds = false;
> >  	bool has_cpu_edp = false;
> >  	bool has_pch_edp = false;
> > @@ -4876,70 +4876,109 @@ static void ironlake_init_pch_refclk(struct drm_device *dev)
> >  	 * PCH B stepping, previous chipset stepping should be
> >  	 * ignoring this setting.
> >  	 */
> > -	temp = I915_READ(PCH_DREF_CONTROL);
> > +	val = I915_READ(PCH_DREF_CONTROL);
> > +
> > +	/* As we must carefully and slowly disable/enable each source in turn,
> > +	 * compute the final state we want first and check if we need to
> > +	 * make any changes at all.
> > +	 */
> > +	final = val;
> > +	final &= ~DREF_NONSPREAD_SOURCE_MASK;
> > +	if (has_ck505)
> > +		final |= DREF_NONSPREAD_CK505_ENABLE;
> > +	else
> > +		final |= DREF_NONSPREAD_SOURCE_ENABLE;
> > +
> > +	final &= ~DREF_SSC_SOURCE_MASK;
> > +	final &= ~DREF_CPU_SOURCE_OUTPUT_MASK;
> > +	final &= ~DREF_SSC1_ENABLE;
> > +
> > +	if (has_panel) {
> > +		final |= DREF_SSC_SOURCE_ENABLE;
> > +
> > +		if (intel_panel_use_ssc(dev_priv) && can_ssc)
> > +			final |= DREF_SSC1_ENABLE;
> > +
> > +		if (has_cpu_edp) {
> > +			if (intel_panel_use_ssc(dev_priv) && can_ssc)
> > +				final |= DREF_CPU_SOURCE_OUTPUT_DOWNSPREAD;
> > +			else
> > +				final |= DREF_CPU_SOURCE_OUTPUT_NONSPREAD;
> > +		} else
> > +			final |= DREF_CPU_SOURCE_OUTPUT_DISABLE;
> 
> I've been assimilated, I dislike not having braces in all branches if
> one branch requires them... but that's bikeshedding. On the stuff that
> matters,

I've punted on this bikeshed (checkpatch didn't yell), but fixed another
one ;-)
> 
> Reviewed-by: Jani Nikula <jani.nikula@intel.com>

I've figured that speeding this up and moving it into
->global_modeset_resources are rather orthogonal, since even with fastboot
we might want to adjust pm state a bit (e.g. with the power wells stuff).
So even when this is move into the modeset code and thought more clever
tricks, we'll still run it in the boot-up modeset path.

Queued for -next, thanks for the patch.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch