Re: [PATCH v2 1/2] iio/adc: ingenic: Fix channel offsets in buffer

[Andy Shevchenko]

On Mon, May 22, 2023 at 2:35 PM Paul Cercueil <paul@xxxxxxxxxxxxxxx> wrote:
> Le lundi 22 mai 2023 à 14:05 +0300, Andy Shevchenko a écrit :
> > On Mon, May 22, 2023 at 1:23 PM Paul Cercueil <paul@xxxxxxxxxxxxxxx>
> > wrote:
> > > Le lundi 22 mai 2023 à 13:18 +0300, Andy Shevchenko a écrit :
> > > > On Mon, May 22, 2023 at 1:15 PM Andy Shevchenko
> > > > <andy.shevchenko@xxxxxxxxx> wrote:
> > > > > On Mon, May 22, 2023 at 1:59 AM Artur Rojek
> > > > > <contact@xxxxxxxxxxxxxx> wrote:

...

> > > > > > +       u16 tdat[6];
> > > > > > +       u32 val;
> > > > > > +
> > > > > > +       memset(tdat, 0, ARRAY_SIZE(tdat));
> > > > >
> > > > > Yeah, as LKP tells us this should be sizeof() instead of
> > > > > ARRAY_SIZE().
> > > > >
> > > > > > +       for (i = 0; mask && i < ARRAY_SIZE(tdat); mask >>= 2)
> > > > > > {
> > > > > > +               if (mask & 0x3) {
> > > > >
> > > > > (for the consistency it has to be GENMASK(), but see below)
> > > > >
> > > > > First of all, strictly speaking we should use the full mask
> > > > > without
> > > > > limiting it to the 0 element in the array (I'm talking about
> > > > > active_scan_mask).
> > > > >
> > > > > That said, we may actually use bit operations here in a better
> > > > > way,
> > > > > i.e.
> > > > >
> > > > >   unsigned long mask = active_scan_mask[0] &
> > > > > (active_scan_mask[0] -
> > > > > 1);
> > > > >
> > > > >   j = 0;
> > > > >   for_each_set_bit(i, active_scan_mask, ...) {
> > > > >     val = readl(...);
> > > > >     /* Two channels per sample. Demux active. */
> > > > >     tdat[j++] = val >> (16 * (i % 2));
> > > >
> > > > Alternatively
> > > >
> > > >      /* Two channels per sample. Demux active. */
> > > >      if (i % 2)
> > > >        tdat[j++] = upper_16_bits(val);
> > > >      else
> > > >        tdat[j++] = lower_16_bits(val);
> > > >
> > > > which may be better to read.
> > >
> > > It's not if/else though. You would check (i % 2) for the upper 16
> > > bits,
> > > and (i % 1) for the lower 16 bits. Both can be valid at the same
> > > time.

(i can't be two bits at the same time in my proposal)

> > Are you sure? Have you looked into the proposed code carefully?
>
> Yes. I co-wrote the original code, I know what it's supposed to do.

Yes, but I'm talking about my version to which you commented and I
think it is the correct approach with 'else'. The problematic part in
my proposal is FIFO reading.
So, I have tried to come up with the working solution, but it seems
it's too premature optimization here that's not needed and code,
taking into account readability, will become a bit longer.

That said, let's go with your version for now (implying the GENMASK()
and upper/lower_16_bits() macros in use).

> > What probably can be done differently is the read part, that can be
> > called once. But looking at it I'm not sure how it's supposed to work
> > at all, since the address is always the same. How does the code and
> > hardware are in sync with the channels?
>
> It's a FIFO.

A-ha.

> > > > >   }
> > > > >
> > > > > > +                       val = readl(adc->base +
> > > > > > JZ_ADC_REG_ADTCH);
> > > > > > +                       /* Two channels per sample. Demux
> > > > > > active.
> > > > > > */
> > > > > > +                       if (mask & BIT(0))
> > > > > > +                               tdat[i++] = val & 0xffff;
> > > > > > +                       if (mask & BIT(1))
> > > > > > +                               tdat[i++] = val >> 16;
> > > > > > +               }
> > > > > >         }

--
With Best Regards,
Andy Shevchenko