On Mon, Nov 21, 2022 at 02:21:30PM +0100, Hans de Goede wrote: > On 11/18/22 19:43, Andy Shevchenko wrote: ... > Thanks, patch looks good to me: > > Reviewed-by: Hans de Goede <hdegoede@xxxxxxxxxx> > > I've added this to my media-atomisp branch now and will > include this in the pull-req which I will send to Mauro after I've > run some quick tests (tonight or tomorrow). Thank you! ... > > res->sc_bayer_origin_x_bqs_on_shading_table = > > ((left_padding_adjusted_bqs + bad_bqs_on_left_before_bs) > > - * bs_hor_ratio_out + bs_hor_ratio_in / 2) / bs_hor_ratio_in > > + * bds.denominator + bds.numerator / 2) / bds.numerator > > + bad_bqs_on_left_after_bs; > > - /* "+ bs_hor_ratio_in/2": rounding for division by bs_hor_ratio_in */ > > + /* "+ bds.numerator / 2": rounding for division by bds.numerator */ > > res->sc_bayer_origin_y_bqs_on_shading_table = > > - (bad_bqs_on_top_before_bs * bs_ver_ratio_out + bs_ver_ratio_in / 2) / bs_ver_ratio_in > > + (bad_bqs_on_top_before_bs * bds.denominator + bds.numerator / 2) / bds.numerator > > + bad_bqs_on_top_after_bs; > > - /* "+ bs_ver_ratio_in/2": rounding for division by bs_ver_ratio_in */ > > + /* "+ bds.numerator / 2": rounding for division by bds.numerator */ While reading this code again I just realized this is open coded DIV_ROUND_CLOSEST(). But, thinking more, I believe we may rather introduce the FRACT_DIV_ROUND_CLOSEST() macro in the math.h and take struct *_fract as a parameter. In any case, it's for the future improvements. -- With Best Regards, Andy Shevchenko
