Here is the final version of this patch, implementing Richard's latest suggestion,
as committed to mainline.  This revision has been tested on x86_64-pc-linux-gnu
with a "make bootstrap" and "make -k check" with no new regressions.

Many thanks for the helpful feedback about match.pd's BUILT_IN_ iterators;
I'll post a follow-up patch to clean-up/make use of this idiom elsewhere in
match.pd, as it is much nicer.

Cheers,
Roger
--

-----Original Message-----
From: Richard Biener <richard.guenther@gmail.com> 
Sent: 23 July 2020 10:02
To: GCC Patches <gcc-patches@gcc.gnu.org>
Cc: Roger Sayle <roger@nextmovesoftware.com>
Subject: Re: [PATCH] middle-end: Fold popcount(x&4) to (x>>2)&1 and friends.

On Thu, Jul 23, 2020 at 10:15 AM Marc Glisse <marc.glisse@inria.fr> wrote:
>
> On Wed, 22 Jul 2020, Roger Sayle wrote:
>
> > Many thanks for the peer review and feedback.  I completely agree 
> > that POPCOUNT and PARITY iterators simplifies things and handle the IFN_ variants.
>
> Is there a reason why the iterators cannot be used for this one?
>
> +(for popcount (BUILT_IN_POPCOUNT BUILT_IN_POPCOUNTL BUILT_IN_POPCOUNTLL
> +              BUILT_IN_POPCOUNTIMAX)
> +     parity (BUILT_IN_PARITY BUILT_IN_PARITYL BUILT_IN_PARITYLL
> +            BUILT_IN_PARITYIMAX)
> +  (simplify
> +    (bit_and (popcount @0) integer_onep)
> +    (parity @0)))

I should indeed work fine to even do

 (simplify
   (bit_and (POPCOUNT @0) integer_onep)
   (PARITY @0))

Likewise for the existing

+/* popcount(X) == 0 is X == 0, and related (in)equalities.  */ (for 
+popcount (POPCOUNT)
   (for cmp (le eq ne gt)
        rep (eq eq ne ne)
     (simplify
       (cmp (popcount @0) integer_zerop)
       (rep @0 { build_zero_cst (TREE_TYPE (@0)); }))))

you can elide the (for ...)

OK with those changes.

Richard.

>
> --
> Marc Glisse