lkml.org 
[lkml]   [2021]   [Jul]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v3 06/10] iio: afe: rescale: add offset support
On Sun Jul 4, 2021 at 12:53 PM EDT, Jonathan Cameron wrote:
> On Wed, 30 Jun 2021 21:00:30 -0400
> Liam Beguin <liambeguin@gmail.com> wrote:
>
> > From: Liam Beguin <lvb@xiphos.com>
> >
> > This is a preparatory change required for the addition of temperature
> > sensing front ends.
> >
> > Signed-off-by: Liam Beguin <lvb@xiphos.com>
> Hi Liam.
>
> A few remaining things inline.
>
> Looking good in general.
>
> Jonathan
>
> > ---
> > drivers/iio/afe/iio-rescale.c | 63 +++++++++++++++++++++++++++++++++++
> > 1 file changed, 63 insertions(+)
> >
> > diff --git a/drivers/iio/afe/iio-rescale.c b/drivers/iio/afe/iio-rescale.c
> > index 8f79c582519c..c8750286c308 100644
> > --- a/drivers/iio/afe/iio-rescale.c
> > +++ b/drivers/iio/afe/iio-rescale.c
> > @@ -32,6 +32,7 @@ struct rescale {
> > bool chan_processed;
> > s32 numerator;
> > s32 denominator;
> > + s32 offset;
> > };
> >
> > static int rescale_read_raw(struct iio_dev *indio_dev,
> > @@ -39,6 +40,8 @@ static int rescale_read_raw(struct iio_dev *indio_dev,
> > int *val, int *val2, long mask)
> > {
> > struct rescale *rescale = iio_priv(indio_dev);
> > + int scale, scale2;
> > + int schan_off = 0;
> > s64 tmp, tmp2;
> > u32 factor;
> > int ret;
> > @@ -99,6 +102,62 @@ static int rescale_read_raw(struct iio_dev *indio_dev,
> > dev_err(&indio_dev->dev, "unsupported type %d\n", ret);
> > return -EOPNOTSUPP;
> > }
> > + case IIO_CHAN_INFO_OFFSET:
> > + /*
> > + * Processed channels are scaled 1-to-1 and source offset is
> > + * already taken into account.
> > + *
> > + * In other cases, the final offset value is defined by:
> > + * offset = schan_offset + rescaler_offset / schan_scale
>
> Maths is right, but perhaps useful to express how this is derived as I
> had
> to scribble it down to check.
>
> Want to express real world measurement as:
> scale * (raw + offset)
> Given we applying scale and offset recursively we have.
> rescaler_scale * (schan_scale * (raw + schan_offset) + rescaler_offset)
> which can be rearrange into the correct form at.
> rescaler_scale *schan_scale* (raw + (schan_offset +
> rescaler_offset/schan_scale))
> Thus the scale and offset we expose to userspace should be.
> scale = rescaler_scale * schan_scale
> offset = schan_offset + rescaler_offset/schan_scale;
>

Understood, I'll update the comment with more details.

>
> > + */
> > + if (rescale->chan_processed) {
> > + *val = rescale->offset;
> > + return IIO_VAL_INT;
> > + }
> > +
> > + if (iio_channel_has_info(rescale->source->channel,
> > + IIO_CHAN_INFO_OFFSET)) {
> > + ret = iio_read_channel_offset(rescale->source,
> > + &schan_off, NULL);
> > + if (ret < 0)
> > + return ret;
> if you've returned in the first branch of the if, no need to worry about
> the
> else for checking the second condition.
>
> if (ret != IIO_VAL_INT)
> return...

Right, I'll combine both statements.

>
>
> > + else if (ret != IIO_VAL_INT)
> > + return -EOPNOTSUPP;
> > + }
> > +
> > + ret = iio_read_channel_scale(rescale->source, &scale, &scale2);
> > + switch (ret) {
> > + case IIO_VAL_FRACTIONAL:
> > + tmp = (s64)rescale->offset * scale2;
> > + *val = div_s64(tmp, scale) + schan_off;
> > + return IIO_VAL_INT;
> > + case IIO_VAL_INT:
> > + *val = div_s64(rescale->offset, scale) + schan_off;
> > + return IIO_VAL_INT;
> > + case IIO_VAL_FRACTIONAL_LOG2:
> > + tmp = (s64)rescale->offset * (1 << scale2);
> > + *val = div_s64(tmp, scale) + schan_off;
> > + return IIO_VAL_INT;
> > + case IIO_VAL_INT_PLUS_NANO:
> > + tmp = (s64)rescale->offset * 1000000000UL;
> > + tmp2 = ((s64)scale * 1000000000UL) + scale2;
> > + factor = gcd(tmp, tmp2);
> > + tmp /= factor;
> > + tmp2 /= factor;
>
> What is the benefit to doing gcd division before div_s64?
>

You're right the GCD division is unnecessary here, will drop.

> > + *val = div_s64(tmp, tmp2) + schan_off;
> > + return IIO_VAL_INT;
> > + case IIO_VAL_INT_PLUS_MICRO:
> > + tmp = (s64)rescale->offset * 1000000UL;
> > + tmp2 = ((s64)scale * 1000000UL) + scale2;
> > + factor = gcd(tmp, tmp2);
> > + tmp /= factor;
> > + tmp2 /= factor;
> > + *val = div_s64(tmp, tmp2) + schan_off;
> > + return IIO_VAL_INT;
> > + default:
> > + dev_err(&indio_dev->dev, "unsupported type %d\n", ret);
> > + return -EOPNOTSUPP;
> > + }
> > default:
> > return -EINVAL;
> > }
> > @@ -175,6 +234,9 @@ static int rescale_configure_channel(struct device *dev,
> > chan->info_mask_separate = BIT(IIO_CHAN_INFO_RAW) |
> > BIT(IIO_CHAN_INFO_SCALE);
> >
> > + if (rescale->offset)
> > + chan->info_mask_separate |= BIT(IIO_CHAN_INFO_OFFSET);
> > +
> > /*
> > * Using .read_avail() is fringe to begin with and makes no sense
> > * whatsoever for processed channels, so we make sure that this cannot
> > @@ -339,6 +401,7 @@ static int rescale_probe(struct platform_device *pdev)
> > rescale->cfg = of_device_get_match_data(dev);
> > rescale->numerator = 1;
> > rescale->denominator = 1;
> > + rescale->offset = 0;
> >
> > ret = rescale->cfg->props(dev, rescale);
> > if (ret)

\
 
 \ /
  Last update: 2021-07-05 06:27    [W:0.272 / U:1.996 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site