On Tue, Jan 13, 2026 at 12:52 AM Richard Biener
<[email protected]> wrote:
>
> On Mon, Jan 12, 2026 at 11:16 PM Andrew Pinski
> <[email protected]> wrote:
> >
> > On Mon, Jan 12, 2026 at 12:33 AM Richard Biener
> > <[email protected]> wrote:
> > >
> > > On Sun, Jan 11, 2026 at 10:05 PM Andrew Pinski
> > > <[email protected]> wrote:
> > > >
> > > > This adds the simpliciation of:
> > > > ```
> > > >   <unnamed-signed:3> _1;
> > > >
> > > >   _2 = (signed char) _1;
> > > >   _3 = _2 ^ -47;
> > > >   _4 = (<unnamed-signed:3>) _3;
> > > > ```
> > > >
> > > > to:
> > > > ```
> > > >   <unnamed-signed:3> _n;
> > > >   _4 = _1 ^ -47;
> > > > ```
> > > >
> > > > This also fixes PR 122843 by optimizing out the xor such that we get:
> > > > ```
> > > >   _1 = b.a;
> > > >   _21 = (<unnamed-signed:3>) t_23(D);
> > > >   // t_23 in the original testcase was 200 so this is reduced to 0
> > > >   _5 = _1 ^ _21;
> > > >   # .MEM_24 = VDEF <.MEM_13>
> > > >   b.a = _5;
> > > > ```
> > > > And then there is no cast catch this pattern:
> > > > `(bit_xor (convert1? (bit_xor:c @0 @1)) (convert2? (bit_xor:c @0 @2)))`
> > > > As we get:
> > > > ```
> > > >   _21 = (<unnamed-signed:3>) t_23(D);
> > > >   _5 = _1 ^ _21;
> > > >   _22 = (<unnamed-signed:3>) t_23(D);
> > > >   _7 = _5 ^ _22;
> > > >   _25 = (<unnamed-signed:3>) t_23(D);
> > > >   _8 = _7 ^ _25;
> > > >   _26 = (<unnamed-signed:3>) t_23(D);
> > > >   _9 = _7 ^ _26;
> > > > ```
> > > > After unrolling and then fre will optimize away all of those xor.
> > > >
> > > > Bootstrapped and tested on x86_64-linux-gnu.
> > > >
> > > >         PR tree-optimization/122845
> > > >         PR tree-optimization/122843
> > > > gcc/ChangeLog:
> > > >
> > > >         * match.pd (`(T1)(a bit_op (T2)b)`): Also
> > > >         simplify if T1 is the same type as b and T2 is wider
> > > >         type than T1.
> > > >
> > > > gcc/testsuite/ChangeLog:
> > > >
> > > >         * gcc.dg/tree-ssa/bitops-12.c: New test.
> > > >         * gcc.dg/tree-ssa/bitops-13.c: New test.
> > > >         * gcc.dg/store_merging_18.c: xfail store merging.
> > > >
> > > > Signed-off-by: Andrew Pinski <[email protected]>
> > > > ---
> > > >  gcc/match.pd                              | 10 +++++++++-
> > > >  gcc/testsuite/gcc.dg/store_merging_18.c   |  3 ++-
> > > >  gcc/testsuite/gcc.dg/tree-ssa/bitops-12.c | 18 ++++++++++++++++++
> > > >  gcc/testsuite/gcc.dg/tree-ssa/bitops-13.c | 18 ++++++++++++++++++
> > > >  4 files changed, 47 insertions(+), 2 deletions(-)
> > > >  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/bitops-12.c
> > > >  create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/bitops-13.c
> > > >
> > > > diff --git a/gcc/match.pd b/gcc/match.pd
> > > > index 492d88514fc..16ce2d1edd2 100644
> > > > --- a/gcc/match.pd
> > > > +++ b/gcc/match.pd
> > > > @@ -2323,7 +2323,15 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> > > >         && !POINTER_TYPE_P (TREE_TYPE (@0))
> > > >         && TREE_CODE (TREE_TYPE (@0)) != OFFSET_TYPE
> > > >         && TYPE_PRECISION (TREE_TYPE (@0)) > TYPE_PRECISION (type))
> > > > -   (bitop:type (convert @0) (convert @1)))))
> > > > +   (bitop:type (convert @0) (convert @1))
> > > > +   /* Similar as above, but the outer and inner most types match
> > > > +       and it was widening cast; replacing 2 casts with only one.  */
> > > > +   (if (GIMPLE
> > > > +        && INTEGRAL_TYPE_P (type)
> > > > +        && INTEGRAL_TYPE_P (TREE_TYPE (@0))
> > > > +        && types_match (type, TREE_TYPE (@0))
> > > > +        && TYPE_PRECISION (TREE_TYPE (@2)) > TYPE_PRECISION (type))
> > >
> > > I think you need the pointer and offset type checks here as well given
> > > bitops are not valid on those in general.
> >
> > The `INTEGRAL_TYPE_P (type)` check will already reject pointer and
> > offset types. And the type on @2 will already be an integer type
> > because it is on the bitop.
> > So there is no need to add the extra checks.
>
> Hmm, indeed.  Can you remove the redundant hecks on the previous case then:
>
>    (if (GIMPLE
>         && INTEGRAL_TYPE_P (type)
>         && INTEGRAL_TYPE_P (TREE_TYPE (@0))
>         && TREE_CODE (@1) != INTEGER_CST
>         && tree_nop_conversion_p (type, TREE_TYPE (@2))
>         && !POINTER_TYPE_P (TREE_TYPE (@0))
>         && TREE_CODE (TREE_TYPE (@0)) != OFFSET_TYPE
> ?

Done with the attached patch which I pushed too.

Thanks,
Andrew

>
> > >
> > > OK with adding those.
> >
> > I pushed it without the extra checks added.
> >
> > Thanks,
> > Andrew Pinski
> >
> >
> >
> > > Richard.
> > >
> > > > +    (bitop:type @0 (convert @1))))))
> > > >
> > > >  (for bitop (bit_and bit_ior)
> > > >       rbitop (bit_ior bit_and)
> > > > diff --git a/gcc/testsuite/gcc.dg/store_merging_18.c 
> > > > b/gcc/testsuite/gcc.dg/store_merging_18.c
> > > > index fdff6b4d812..546c17dc7c4 100644
> > > > --- a/gcc/testsuite/gcc.dg/store_merging_18.c
> > > > +++ b/gcc/testsuite/gcc.dg/store_merging_18.c
> > > > @@ -1,7 +1,8 @@
> > > >  /* PR tree-optimization/83843 */
> > > >  /* { dg-do run } */
> > > >  /* { dg-options "-O2 -fno-tree-vectorize -fdump-tree-store-merging" } 
> > > > */
> > > > -/* { dg-final { scan-tree-dump-times "Merging successful" 3 
> > > > "store-merging" { target { store_merge && { ! arm*-*-* } } } } } */
> > > > +/* xfailed after PR 122845, see PR 123541 ( */
> > > > +/* { dg-final { scan-tree-dump-times "Merging successful" 3 
> > > > "store-merging" { target { store_merge && { ! arm*-*-* } } xfail *-*-* 
> > > > } } } */
> > > >
> > > >  __attribute__((noipa)) void
> > > >  foo (unsigned char *buf, unsigned char *tab)
> > > > diff --git a/gcc/testsuite/gcc.dg/tree-ssa/bitops-12.c 
> > > > b/gcc/testsuite/gcc.dg/tree-ssa/bitops-12.c
> > > > new file mode 100644
> > > > index 00000000000..50d8dc2180f
> > > > --- /dev/null
> > > > +++ b/gcc/testsuite/gcc.dg/tree-ssa/bitops-12.c
> > > > @@ -0,0 +1,18 @@
> > > > +/* { dg-do compile } */
> > > > +/* { dg-options "-O1 -fdump-tree-optimized-raw" } */
> > > > +/* PR tree-optimization/122845 */
> > > > +
> > > > +struct s1
> > > > +{
> > > > +  signed f1:3;
> > > > +};
> > > > +struct s1 p;
> > > > +void f(void)
> > > > +{
> > > > + p.f1 ^= -47;
> > > > + p.f1 ^= -47;
> > > > +}
> > > > +
> > > > +/* There should be no gimple stmts left except for return.
> > > > +   That is the ^ has been optimized away.   */
> > > > +/* { dg-final { scan-tree-dump-not "gimple_assign <" "optimized"  } } 
> > > > */
> > > > diff --git a/gcc/testsuite/gcc.dg/tree-ssa/bitops-13.c 
> > > > b/gcc/testsuite/gcc.dg/tree-ssa/bitops-13.c
> > > > new file mode 100644
> > > > index 00000000000..ecf69738267
> > > > --- /dev/null
> > > > +++ b/gcc/testsuite/gcc.dg/tree-ssa/bitops-13.c
> > > > @@ -0,0 +1,18 @@
> > > > +/* { dg-do compile } */
> > > > +/* { dg-options "-O1 -fdump-tree-optimized-raw" } */
> > > > +/* PR tree-optimization/122845 */
> > > > +
> > > > +int f(signed char a, int  b)
> > > > +{
> > > > +  int aa = a;
> > > > +  int bb = b;
> > > > +  aa &= bb;
> > > > +  signed char t = aa;
> > > > +  aa = t;
> > > > +  aa &= bb;
> > > > +  t = aa;
> > > > +  return t;
> > > > +}
> > > > +
> > > > +/* There should be only 1 bit_and_expr left as both ands are the same. 
> > > > */
> > > > +/* { dg-final { scan-tree-dump-times "bit_and_expr, " 1 "optimized"  } 
> > > > } */
> > > > --
> > > > 2.43.0
> > > >

Attachment: 0001-match-Remove-redundant-type-checks-from-T1-a-bit_op-.patch
Description: Binary data

Reply via email to