Re: [PATCH] winex11.drv: Fix incorrect frequency for doublescan and interlaced modes.

[Gabriel Ivăncescu]

On 04/02/2020 10:37, Zhiyi Zhang wrote:
> On 2/4/20 4:27 PM, Dmitry Timoshkov wrote:
> > Zhiyi Zhang  wrote:
> > > +static DWORD get_frequency( const XRRModeInfo *mode )
> > > +{
> > > - if (mode->hTotal && mode->vTotal)
> > > - {
> > > -
> > >    double v_total = mode->vTotal;
> > > -
> > > -
> > >    if (mode->modeFlags & RR_DoubleScan)
> > > -
> > >        v_total *= 2;
> > > -
> > >    if (mode->modeFlags & RR_Interlace)
> > > -
> > >        v_total /= 2;
> > > -
> > > -
> > >    /* Adding 0.05 instead of 0.5 to round so that common frequencies like
> > > -
> > >     * 59.94Hz and 23.976Hz become 59Hz and 24Hz. Using 0.5 would make
> > > -
> > >     * 59.94Hz become 60Hz and would make it seem like there are two 60Hz modes */
> > > -
> > >    return mode->dotClock / (mode->hTotal * v_total) + 0.05;
> > > - }
> > > -
> > > - return 0;
> > >
> > > +}
> >
> > It seems that floating point (and especially double) is not needed here
> > to do a proper calculation.
>
> It's needed. See https://gitlab.freedesktop.org/xorg/app/xrandr/commit/00c795e99fe29ecd56e05e...

I think he meant, can't you just multiply both v_total and 
mode->dotClock by 2 for same effect?

int v_total = mode->vTotal * 2;

...

return mode->dotClock * 2 / (mode->hTotal * v_total);

Admittedly it doesn't handle the 0.05 but that seems somewhat arbitrary.