Re: [PATCH v2 2/6] units: Add 32- and 64-bit signed values of π

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Sun, Nov 09, 2025 at 05:00:32PM +0000, Jonathan Cameron wrote:
> On Sat, 8 Nov 2025 21:02:54 +0200
> Andy Shevchenko <andy.shevchenko@gmail.com> wrote:
> > пʼятниця, 7 листопада 2025 р. David Lechner <dlechner@baylibre.com> пише:
> > > On 11/7/25 2:03 PM, Andy Shevchenko wrote:  
> > > > There are a few drivers that want to have these values, and
> > > > one more known to come soon. Let's define the values for them.

...

> > > > +/* Value of π * 10⁸ (fits s32 or signed int) */
> > > > +#define PI   314159265  
> > >
> > > It isn't clear to me at all in the other patches that e.g.
> > > DIV_ROUND_UP(PI, 1000) would be π ✕ 10⁴ (rounded to nearest
> > > integer, of course).
> > >
> > > Calling these PI_E8 and PI_E18 or PI_x10_8 and PI_x10_18
> > > would help to clear that up.  
> > 
> > This will be an awful and ugly name(s). I fully disagree on such a
> > proposal. The power is chosen to fit the type with maximum precision, no
> > need to explain this in the name.
> With no indicate of the multiplier we have to check it every time wwe
> use it which isn't ideal either.
> 
> So I agree with David that we have to indicate this somewhere. Maybe hold the
> multiplier in a separate define
> #define PI_MUL
> #define PI_MUL_LL
> 
> Then compute the divisions by (PI_MUL) / KILO or something along those lines?

I see, we have a disagreement here, because I don't like to uglify the pure PI
definition. So, please abandon this series then.

> > > > +/* Value of π * 10¹⁸ (fits s64 or signed long long) */
> > > > +#define PI_LL        3141592653589793238LL

Note, this constant is used a lot in the tests (3+ modules), so if you go with
your version, take this into account.

> > > >  /* Hz based multipliers */
> > > >  #define NANOHZ_PER_HZ                1000000000UL
> > > >  #define MICROHZ_PER_HZ               1000000UL  

-- 
With Best Regards,
Andy Shevchenko