Re: [PATCH v5 05/10] iio: afe: rescale: add INT_PLUS_{MICRO,NANO} support

["Liam Beguin" <liambeguin@gmail.com>]

On Sat Jul 17, 2021 at 12:55 PM EDT, Jonathan Cameron wrote:
> On Fri, 16 Jul 2021 15:18:33 -0400
> "Liam Beguin" <liambeguin@gmail.com> wrote:
>
> > On Thu Jul 15, 2021 at 5:48 AM EDT, Peter Rosin wrote:
> > >
> > > On 2021-07-15 05:12, Liam Beguin wrote:  
> > > > From: Liam Beguin <lvb@xiphos.com>
> > > > 
> > > > Some ADCs use IIO_VAL_INT_PLUS_{NANO,MICRO} scale types.
> > > > Add support for these to allow using the iio-rescaler with them.
> > > > 
> > > > Signed-off-by: Liam Beguin <lvb@xiphos.com>
> > > > ---
> > > >  drivers/iio/afe/iio-rescale.c | 15 +++++++++++++++
> > > >  1 file changed, 15 insertions(+)
> > > > 
> > > > diff --git a/drivers/iio/afe/iio-rescale.c b/drivers/iio/afe/iio-rescale.c
> > > > index 4c3cfd4d5181..a2b220b5ba86 100644
> > > > --- a/drivers/iio/afe/iio-rescale.c
> > > > +++ b/drivers/iio/afe/iio-rescale.c
> > > > @@ -92,7 +92,22 @@ static int rescale_read_raw(struct iio_dev *indio_dev,
> > > >  			do_div(tmp, 1000000000LL);
> > > >  			*val = tmp;
> > > >  			return ret;
> > > > +		case IIO_VAL_INT_PLUS_NANO:
> > > > +			tmp = ((s64)*val * 1000000000LL + *val2) * rescale->numerator;
> > > > +			do_div(tmp, rescale->denominator);
> > > > +
> > > > +			*val = div_s64(tmp, 1000000000LL);
> > > > +			*val2 = tmp - *val * 1000000000LL;
> > > > +			return ret;  
> > >
> > > This is too simplistic and prone to overflow. We need something like
> > > this
> > > (untested)
> > >
> > > tmp = (s64)*val * rescale->numerator;
> > > rem = do_div(tmp, rescale->denominator);
> > > *val = tmp;
> > > tmp = ((s64)rem * 1000000000LL + (s64)*val2) * rescale->numerator;
> > > do_div(tmp, rescale->denominator);
> > > *val2 = tmp;
> > >
> > > Still not very safe with numerator and denominator both "large", but
> > > much
> > > better. And then we need normalizing the fraction part after the above,
> > > of
> > > course.
> > >  
> > 
> > Understood, I'll test that.
> > 
> > > And, of course, I'm not sure what *val == -1 and *val2 == 500000000
> > > really
> > > means. Is that -1.5 or -0.5? The above may very well need adjusting for
> > > negative values...
> > >  
> > 
> > I would've assumed the correct answer is -1 + 500000000e-9 = -0.5
> > but adding a test case to iio-test-format.c seems to return -1.5...
>

Hi Jonathan,

> No. -1.5 is as intended, though the IIO_VAL_PLUS_MICRO is rather
> confusing
> naming :( We should perhaps add more documentation for that. Signs were
> always a bit of a pain with this two integer scheme for fixed point.
>
> The intent is to have moderately readable look up tables with the
> problem that
> we don't have a signed 0 available. Meh, maybe this decision a long time
> back wasn't a the right one, but it may be a pain to change now as too
> many
> drivers to check!
>
> 1, 0000000 == 1
> 0, 5000000 == 0.5
> 0, 0000000 == 0
> 0, -5000000 == -0.5
> -1, 5000000 == -1.5
>

Understood, thanks for clearing that out.

Liam

>
> > 
> > I believe that's a bug but we can work around if for now by moving the
> > integer part of *val2 to *val.
>
> Yup. Fiddly corner cases..
>
> Jonathan
>
> > 
> > Liam
> > 
> > > Cheers,
> > > Peter
> > >  
> > > > +		case IIO_VAL_INT_PLUS_MICRO:
> > > > +			tmp = ((s64)*val * 1000000LL + *val2) * rescale->numerator;
> > > > +			do_div(tmp, rescale->denominator);
> > > > +
> > > > +			*val = div_s64(tmp, 1000000LL);
> > > > +			*val2 = tmp - *val * 1000000LL;
> > > > +			return ret;
> > > >  		default:
> > > > +			dev_err(&indio_dev->dev, "unsupported type %d\n", ret);
> > > >  			return -EOPNOTSUPP;
> > > >  		}
> > > >  	default:
> > > >   
> >