On Tue, Aug 28, 2012 at 06:09:26PM +0200, Jean Delvare wrote:
> Hi Guenter,
>
> On Fri, 24 Aug 2012 17:53:17 -0700, Guenter Roeck wrote:
> > DIV_ROUND_CLOSEST returns a bad result for negative dividends:
> > DIV_ROUND_CLOSEST(-2, 2) = 0
> >
> > Most of the time this does not matter. However, in the hardware monitoring
> > subsystem, it is often used on integers which can be negative (such as
> > temperatures). Introduce new macro IDIV_ROUND_CLOSEST which also supports
> > negative dividends.
>
> Good catch. It's been broken for years and I never noticed :(
>
> Acked-by: Jean Delvare <khali@xxxxxxxxxxxx>
>
> >
> > Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > ---
> > I can take this patch through my hwmon tree, but would like to get an Ack first.
> > Alternative would be to put it into include/linux/hwmon.h, but I would prefer
> > to avoid that.
>
> I agree it should sit next to DIV_ROUND_CLOSEST in <linux/kernel.h>.
>
> >
> > Also, if someone has an idea for a simpler implementation, I would really like
> > to know about it.
>
> I can't think of anything better.
>
> Note that your implementation only supports negative dividend not
> divisor. While it won't be a problem for hwmon drivers, and most
> probably not a problem for other drivers either, it might be worth
> putting in a comment so as to prevent false expectations from the
> reader.
>
I noticed. I am not concerned about negative divisors, though, but unsigned
divisor variables are a problem too. I have a new version which also supports
divisors which are declared as unsigned.
Thanks,
Guenter
> >
> > include/linux/kernel.h | 9 +++++++++
> > 1 file changed, 9 insertions(+)
> >
> > diff --git a/include/linux/kernel.h b/include/linux/kernel.h
> > index 6043821..a89483c 100644
> > --- a/include/linux/kernel.h
> > +++ b/include/linux/kernel.h
> > @@ -89,6 +89,15 @@
> > } \
> > )
> >
> > +#define IDIV_ROUND_CLOSEST(x, divisor)( \
> > +{ \
> > + typeof(x) __x = x; \
> > + typeof(divisor) __d1 = divisor; \
> > + typeof(divisor) __d2 = (__x) < 0 ? -(__d1) : (__d1);\
> > + (((__x) + ((__d2) / 2)) / (__d1)); \
> > +} \
> > +)
> > +
> > /*
> > * Multiplies an integer by a fraction, while avoiding unnecessary
> > * overflow or loss of precision.
>
>
> --
> Jean Delvare
>
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
