On Tue, Aug 24, 2021 at 03:15:28PM +0300, Andy Shevchenko wrote: > On Tue, Aug 24, 2021 at 2:38 PM Sean Nyekjaer <sean@xxxxxxxxxx> wrote: > > ... > > > Do we have some helper functions to do the 12 bit 2-complement numbers? > > Probably not, look around where sign_extend32() is defined. More on this below. > > ... > > > + return regmap_update_bits(data->regmap, FXLS8962AF_INT_EN, > > + mask, > > + value); > > One line? > > ... > > > + /* > > + * Add the same value to the lower-threshold register with a reversed sign > > + * in 2-complement 12 bit format. > > + */ > > + data->lower_thres = (~val & GENMASK(11, 0)) + 1; > > This looks suspicious. > > 0 => 0xfff + 1 => 0x1000. Is it what is wanted? > I thought that -val & mask is what you need. > > Can you explain more in the comment (maybe with examples) on what is > coming and what is expected? It's a bit messy I know :) Some examples: val = 0 => upper = 0x0, lower = 0x0 val = 500 => upper = 0x1F4, lower = 0xe0c val = 1000 => upper = 0x3e8, lower = 0xc18 Guess it could work if we special case val = 0... It doesn't even makes sense to write 0 to this register as noise would trigger events. > > > + data->upper_thres = val & GENMASK(10, 0); > /Sean
