[Intel-gfx] [PATCH] drm/i915/dmc: Use RUNTIME_INFO->stp for DMC

[Souza, Jose]

On Wed, 2021-06-30 at 17:01 -0700, Lucas De Marchi wrote:
> Typo: RUNTIME_INFO->stp
> 
> On Wed, Jun 30, 2021 at 04:06:24PM -0700, Anusha Srivatsa wrote:
> > On the dmc side,we maintain a lookup table with different display
> > stepping-substepping combinations.
> > 
> > Instead of adding new table for every new platform, lets ues
> > the stepping info from RUNTIME_INFO(dev_priv)->step
> > Adding the helper intel_get_display_step().
> > 
> > Cc: Lucas De Marchi <lucas.demarchi at intel.com>
> > Signed-off-by: Anusha Srivatsa <anusha.srivatsa at intel.com>
> > ---
> > drivers/gpu/drm/i915/display/intel_dmc.c | 49 ++++++++++++++++++++++--
> > 1 file changed, 45 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/display/intel_dmc.c b/drivers/gpu/drm/i915/display/intel_dmc.c
> > index f8789d4543bf..c7ff7ff3f412 100644
> > --- a/drivers/gpu/drm/i915/display/intel_dmc.c
> > +++ b/drivers/gpu/drm/i915/display/intel_dmc.c
> > @@ -266,14 +266,55 @@ static const struct stepping_info icl_stepping_info[] = {
> > };
> > 
> > static const struct stepping_info no_stepping_info = { '*', '*' };
> > +struct stepping_info *display_step;
> > +
> > +static struct stepping_info *
> > +intel_get_display_stepping(struct intel_step_info step)
> > +{
> > +
> > +	switch (step.display_step) {
> > +	case STEP_A0:
> > +		display_step->stepping = 'A';
> > +		display_step->substepping = '0';
> > +		break;
> > +	case STEP_A2:
> > +		display_step->stepping = 'A';
> > +		display_step->substepping = '2';
> > +		break;
> > +	case STEP_B0:
> > +		display_step->stepping = 'B';
> > +		display_step->substepping = '0';
> > +		break;
> > +	case STEP_B1:
> > +		display_step->stepping = 'B';
> > +		display_step->substepping = '1';
> > +		break;
> > +	case STEP_C0:
> > +		display_step->stepping = 'C';
> > +		display_step->substepping = '0';
> > +		break;
> > +	case STEP_D0:
> > +		display_step->stepping = 'D';
> > +		display_step->substepping = '0';
> > +		break;
> > +	default:
> > +		display_step->stepping = '*';
> > +		display_step->substepping = '*';
> > +		break;
> > +	}
> 
> 
> "crazy" idea that would avoid this type of conversion:
> changing the step enum to be:
> 
> 
> #define make_step(letter, num) (int)(((letter) << 8 ) | (num))
> 
> STEP_A0 = make_step('A', '0'),
> STEP_A1 = make_step('A', '1'),

Looks a good idea to me, we could also keep it u8 using 4bits for each if there is memory concerns.

> 
> and adapt the rest of the code to play with u16 instead of u8, and
> handle the STEP_FUTURE/STEP_NONE/STEP_FOREVER.
> Maybe it is crazy, dunno.
> 
> +Jani / +Jose. Thoughts?
> 
> 
> For this version the next comment is probably more important.
> 
> > +	return display_step;
> > +}
> > 
> > static const struct stepping_info *
> > intel_get_stepping_info(struct drm_i915_private *dev_priv)
> > {
> > 	const struct stepping_info *si;
> > +	struct intel_step_info step = RUNTIME_INFO(dev_priv)->step;
> > 	unsigned int size;
> > 
> > -	if (IS_ICELAKE(dev_priv)) {
> > +	if (DISPLAY_VER(dev_priv) >= 12) {
> > +		si = intel_get_display_stepping(step);
> > +	} else if (IS_ICELAKE(dev_priv)) {
> > 		size = ARRAY_SIZE(icl_stepping_info);
> > 		si = icl_stepping_info;
> 
> can we move the other ones too? Just use display_step for all platforms.
> Notice that before the separation we will have display_step ==
> graphics_step, so it should just work.
> 
> 
> Lucas De Marchi
> 
> > 	} else if (IS_SKYLAKE(dev_priv)) {
> > @@ -287,10 +328,10 @@ intel_get_stepping_info(struct drm_i915_private *dev_priv)
> > 		si = NULL;
> > 	}
> > 
> > -	if (INTEL_REVID(dev_priv) < size)
> > -		return si + INTEL_REVID(dev_priv);
> > +	if (DISPLAY_VER(dev_priv) < 12)
> > +		return INTEL_REVID(dev_priv) < size ? si + INTEL_REVID(dev_priv) : &no_stepping_info;
> > 
> > -	return &no_stepping_info;
> > +	return si;
> > }
> > 
> > static void gen9_set_dc_state_debugmask(struct drm_i915_private *dev_priv)
> > -- 
> > 2.32.0
> >