Re: [PATCH 1/5] drm: Add drm_fixed_16_16 helper

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

On Wed, Sep 01, 2021 at 09:35:40PM -0400, Alyssa Rosenzweig wrote:
> > Missing documentation :-)
> 
> Ack.
> 
> > > +static inline int drm_fixed_16_16(s32 mult, s32 div)
> > 
> > You should return a s32.
> 
> Ack.
> 
> > The function name isn't very explicit, and departs from the naming
> > scheme of other functions in the same file. As fixed-point numbers are
> > stored in a s64 for the drm_fixp_* helpers, we shouldn't rese the
> > drm_fixp_ prefix, maybe drm_fixp_s16_16_ would be a good prefix. The
> > function should probably be named drm_fixp_s16_16 from_fraction() then,
> > but then the same logic should possibly be replicated to ensure optimal
> > precision. I wonder if it wouldn't be best to simply use
> > drm_fixp_from_fraction() and shift the result right by 16 bits.
> 
> Sure, I'm not attached to the naming ... will wait to hear what colours
> everyone else wants the bikehed painted.
> 
> As for the implementation, I just went with what was used across
> multiple drivers already (no chance of regressions that way) but could
> reuse other helpers if it's better..? If the behaviour changes this goes
> from a trivial cleanup to a much more invasive changeset. I don't own
> half of the hardware here.

But except for msm which may need testing, all other drivers use the
existing FRAC_16_16() macro with constant arguments, so it shouldn't be
difficult to ensure that the new function gives the same results.

-- 
Regards,

Laurent Pinchart