[Intel-gfx] [PATCH v6 2/7] drm/i915: update the QGV point frequency calculations

[Govindapillai, Vinod]

On Tue, 2023-05-23 at 16:14 +0300, Jani Nikula wrote:
> On Tue, 23 May 2023, "Govindapillai, Vinod" <vinod.govindapillai at intel.com> wrote:
> > On Tue, 2023-05-23 at 12:01 +0300, Jani Nikula wrote:
> > > On Tue, 23 May 2023, Vinod Govindapillai <vinod.govindapillai at intel.com> wrote:
> > > > From MTL onwwards, pcode locks the QGV point based on peak BW of
> > > > the intended QGV point passed by the driver. So the peak BW
> > > > calculation must match the value expected by the pcode. Update
> > > > the calculations as per the Bspec.
> > > > 
> > > > v2: use DIV_ROUND_* macro for the calculations (Ville)
> > > > 
> > > > Bspec: 64636
> > > > 
> > > > Signed-off-by: Vinod Govindapillai <vinod.govindapillai at intel.com>
> > > > Reviewed-by: Stanislav Lisovskiy <stanislav.lisovskiy at intel.com>
> > > > ---
> > > >  drivers/gpu/drm/i915/display/intel_bw.c | 2 +-
> > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/i915/display/intel_bw.c
> > > > b/drivers/gpu/drm/i915/display/intel_bw.c
> > > > index ab405c48ca3a..c8075a37c3ab 100644
> > > > --- a/drivers/gpu/drm/i915/display/intel_bw.c
> > > > +++ b/drivers/gpu/drm/i915/display/intel_bw.c
> > > > @@ -182,7 +182,7 @@ static int mtl_read_qgv_point_info(struct drm_i915_private *dev_priv,
> > > >         val2 = intel_uncore_read(&dev_priv->uncore,
> > > >                                  MTL_MEM_SS_INFO_QGV_POINT_HIGH(point));
> > > >         dclk = REG_FIELD_GET(MTL_DCLK_MASK, val);
> > > > -       sp->dclk = DIV_ROUND_UP((16667 * dclk), 1000);
> > > > +       sp->dclk = DIV_ROUND_CLOSEST(16667 * dclk + 500, 1000);
> > > 
> > > What's with the +500 there?
> > 
> > This is what pcode expects. Somehow pcode use this formula and we have to exactly match this.
> > Got
> > this confirmed from Art.
> 
> I'm guessing all it really means is to round to closest, and they
> specified it like that instead of saying "round to closest".
> 
> Essentially (x + (div/2)) / div is what DIV_ROUND_CLOSEST() does.
> 
> It's odd to do it *twice*, and surely not what they expect.
> 
> BR,
> Jani.

Ah.. ok! Thanks! will update this!

BR
Vinod

> 
> > 
> > BR
> > Vinod
> > 
> > > 
> > > BR,
> > > Jani.
> > > 
> > > 
> > > >         sp->t_rp = REG_FIELD_GET(MTL_TRP_MASK, val);
> > > >         sp->t_rcd = REG_FIELD_GET(MTL_TRCD_MASK, val);
> > > 
> > 
>