Re: [PATCH v2 1/5] bitops: Introduce find_next_andnot_bit()

[Yury Norov <yury.norov@gmail.com>]

On Fri, Aug 19, 2022 at 11:34:01AM +0100, Valentin Schneider wrote:
> On 18/08/22 22:04, Andy Shevchenko wrote:
> > On Thu, Aug 18, 2022 at 8:18 PM Steven Rostedt <rostedt@goodmis.org> wrote:
> >> On Thu, 18 Aug 2022 17:26:43 +0100
> >> Valentin Schneider <vschneid@redhat.com> wrote:
> >>
> >> > How about:
> >>
> >> >
> >> >   find the next set bit in (*addr1 & ~*addr2)
> >>
> >> I understand the above better. But to convert that into English, we could
> >> say:
> >>
> >>
> >>   Find the next bit in *addr1 excluding all the bits in *addr2.
> >>
> >> or
> >>
> >>   Find the next bit in *addr1 that is not set in *addr2.
> >
> > With this explanation I'm wondering how different this is to
> > bitmap_bitremap(), with adjusting to using an inverted mask. If they
> > have something in common, perhaps make them in the same namespace with
> > similar naming convention?
> >
> 
> I'm trying to wrap my head around the whole remap thing, IIUC we could have
> something like remap *addr1 to ~*addr2 and stop rather than continue with a
> wraparound, but that really feels like shoehorning.

Old and new maps create a simple forward-looking mapping, like this:
    #0   #4
old: 0111 0 ...
     | \\\|
New: 00 111 ...

So if you pass #0, it's wired to 0; but #1 will skip 1 bit and would be
wired to 2; and so on. There is some puzzling when wraparound comes in
play, but the general idea is like that.

I think there's nothing common with bitmap_and{,_not}.

Thanks,
Yury