[Intel-gfx] [v2] drm/i915: Enable hw workaround to bypass alpha

[Lankhorst, Maarten]

tor 2018-06-21 klockan 14:09 -0700 skrev Radhakrishna Sripada:
> On Thu, Jun 21, 2018 at 08:43:56PM +0530, Vandita Kulkarni wrote:
> > Alpha blending with alpha 0 and 0xff passes through
> > alpha math and rounding logic causing differences
> > compared to fully transparent or opaque plane,resulting
> > in CRC mismatch.
> > This WA on icl and above enables hardware to bypass alpha
> > math and rounding for per pixel alpha values of 00 and 0xff
> > 
> > v2: Fix patchwork checkpatch warnings.
> > 
> > Signed-off-by: Vandita Kulkarni <vandita.kulkarni at intel.com>
> > ---
> >  drivers/gpu/drm/i915/i915_reg.h      |  8 ++++++++
> >  drivers/gpu/drm/i915/intel_display.c | 12 ++++++++++++
> >  2 files changed, 20 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_reg.h
> > b/drivers/gpu/drm/i915/i915_reg.h
> > index 4bfd7a9..b66ec9b 100644
> > --- a/drivers/gpu/drm/i915/i915_reg.h
> > +++ b/drivers/gpu/drm/i915/i915_reg.h
> > @@ -7366,6 +7366,14 @@ enum {
> >  #define BDW_SCRATCH1					_MMIO(
> > 0xb11c)
> >  #define  GEN9_LBS_SLA_RETRY_TIMER_DECREMENT_ENABLE	(1 << 2)
> >  
> > +/*GEN11 chicken */
> > +#define _PIPEA_CHICKEN			0x70038
> > +#define _PIPEB_CHICKEN			0x71038
> > +#define _PIPEC_CHICKEN			0x72038
> > +#define  PER_PIXEL_ALPHA_BYPASS_EN	(1 << 7)
> > +#define PIPE_CHICKEN(pipe)		_MMIO_PIPE(pipe,
> > _PIPEA_CHICKEN,\
> > +						   _PIPEB_CHICKEN)
> > +
> >  /* PCH */
> >  
> >  /* south display engine interrupt: IBX */
> > diff --git a/drivers/gpu/drm/i915/intel_display.c
> > b/drivers/gpu/drm/i915/intel_display.c
> > index 2c8fef3..3d849ec 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -5632,6 +5632,7 @@ static void haswell_crtc_enable(struct
> > intel_crtc_state *pipe_config,
> >  	struct intel_atomic_state *old_intel_state =
> >  		to_intel_atomic_state(old_state);
> >  	bool psl_clkgate_wa;
> > +	u32 pipe_chicken;
> >  
> >  	if (WARN_ON(intel_crtc->active))
> >  		return;
> > @@ -5691,6 +5692,17 @@ static void haswell_crtc_enable(struct
> > intel_crtc_state *pipe_config,
> >  	 */
> >  	intel_color_load_luts(&pipe_config->base);
> >  
> > +	/*
> > +	 * Display WA #1153: enable hardware to bypass the alpha
> > math
> > +	 * and rounding for per-pixel values 00 and 0xff
> > +	 */
> > +	if (INTEL_GEN(dev_priv) >= 11) {
> > +		pipe_chicken = I915_READ(PIPE_CHICKEN(pipe));
> > +		if (!(pipe_chicken & PER_PIXEL_ALPHA_BYPASS_EN))
> > +			I915_WRITE_FW(PIPE_CHICKEN(pipe),
> > +				      pipe_chicken |
> > PER_PIXEL_ALPHA_BYPASS_EN);
> > +	}
> 
> This would enable the wa by default for gen > 11. Would this impact
> for non 00, 0xff alpha cases?
> In other words, should we enable the wa only when alpha is 00/0xff
> and not enable for other values?
> 
This is about per pixel values, unlike the plane alpha. This looks at
the pixel contents, so if a pixel has (0xff, R, G, B), we pass through
RGB unmodified. If it's (0x00, R, G, B), we do the same for the
background pixel without introducing rounding errors.

When we program plane alpha, we handle 0x00 by disabling the plane, and
0xff by disabling plane alpha and making the plane opaque.

Thanks for the patch, pushed. :)

~Maarten
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3282 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20180622/afba08c4/attachment-0001.bin>