Re: [PATCH v8 3/3] mmc: Add driver for LiteX's LiteSDCard interface

From: Gabriel L. Somlo
Date: Sat Jan 08 2022 - 18:55:36 EST


On Sat, Jan 08, 2022 at 07:47:32PM +0200, Andy Shevchenko wrote:
> On Sat, Jan 8, 2022 at 3:20 PM Gabriel L. Somlo <gsomlo@xxxxxxxxx> wrote:
> > On Sat, Jan 08, 2022 at 01:26:08PM +0200, Andy Shevchenko wrote:
> > > On Sat, Jan 8, 2022 at 3:57 AM Gabriel L. Somlo <gsomlo@xxxxxxxxx> wrote:
> > > > On Sat, Jan 08, 2022 at 01:54:07AM +0200, Andy Shevchenko wrote:
> > > > > On Saturday, January 8, 2022, Gabriel Somlo <gsomlo@xxxxxxxxx> wrote:
>
> ...
>
> > > > > + u32 div;
> > > > > +
> > > > > + div = freq ? host->ref_clk / freq : 256U;
> > > > >
> > > > > + div = roundup_pow_of_two(div);
> > > > > + div = clamp(div, 2U, 256U);
> > > > >
> > > > > Logically seems to me that you may join these two together, because clamped
> > > > > range is power-of-2 one.
> > > >
> > > > `div` needs to be a power-of-2 when written to the LITEX_PHY_CLOCKERDIV
> > > > register (below). And clamp() will just enforce a min/max range, so if
> > > > (div = ref_clk / freq) ends up e.g., 5, I need both roundup_pow_of_two()
> > > > to bump it to 8, and clamp() to enforce that it's between 2 and 256.
> > > >
> > > > Unless you mean I should simply write it like:
> > > >
> > > > div = clamp(roundup_pow_of_two(div), 2U, 256U);
> > > >
> > > > ... as a single line?
> > >
> > > Yes, that's what I meant.
> >
> > Turns out, clamp really hates being passed roundup_pow_of_two()
> > directly (see below). I think it's probably better if we leave
> > them as-is, to avoid going the explicit cast route which Geert
> > recommended against.
>
> I see, then ignore my comment on this matter in v9.
> Perhaps add a comment in the code explaining that roundup_pow_of_two()
> may not be unified with clamp()?

I worry that commenting on why things are not done some other way at
that location would detract from the legibility of the code itself.

Perhaps we could use a cast after all, and write it out like this:

div = clamp((u32)roundup_pow_of_two(div), 2U, 256U);

which compiles fine without any warnings, accomplishes your "do it in
a single line" desired behavior, and doesn't require me commenting on
which linux library functions do or don't work well with others... :)

Geert, what do you think?

Thanks,
--Gabriel